diff --git a/gcc/common/config/riscv/riscv-common.c b/gcc/common/config/riscv/riscv-common.c new file mode 100644 index 0000000..5bb2218 --- /dev/null +++ b/gcc/common/config/riscv/riscv-common.c @@ -0,0 +1,140 @@ +/* Common hooks for RISC-V. + Copyright (C) 1989-2014 Free Software Foundation, Inc. + +This file is part of GCC. + +GCC is free software; you can redistribute it and/or modify +it under the terms of the GNU General Public License as published by +the Free Software Foundation; either version 3, or (at your option) +any later version. + +GCC is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +You should have received a copy of the GNU General Public License +along with GCC; see the file COPYING3. If not see +. */ + +#include "config.h" +#include "system.h" +#include "coretypes.h" +#include "tm.h" +#include "common/common-target.h" +#include "common/common-target-def.h" +#include "opts.h" +#include "flags.h" +#include "errors.h" + +/* Parse a RISC-V ISA string into an option mask. */ + +static void +riscv_parse_arch_string (const char *isa, int *flags) +{ + const char *p = isa; + + if (strncmp (p, "RV32", 4) == 0) + *flags |= MASK_32BIT, p += 4; + else if (strncmp (p, "RV64", 4) == 0) + *flags &= ~MASK_32BIT, p += 4; + + if (*p++ != 'I') + { + error ("-march=%s: ISA strings must begin with I, RV32I, or RV64I", isa); + return; + } + + *flags &= ~MASK_MULDIV; + if (*p == 'M') + *flags |= MASK_MULDIV, p++; + + *flags &= ~MASK_ATOMIC; + if (*p == 'A') + *flags |= MASK_ATOMIC, p++; + + *flags |= MASK_SOFT_FLOAT_ABI; + if (*p == 'F') + *flags &= ~MASK_SOFT_FLOAT_ABI, p++; + + if (*p == 'D') + { + p++; + if (!TARGET_HARD_FLOAT) + { + error ("-march=%s: the D extension requires the F extension", isa); + return; + } + } + else if (TARGET_HARD_FLOAT) + { + error ("-march=%s: single-precision-only is not yet supported", isa); + return; + } + + *flags &= ~MASK_RVC; + if (*p == 'C') + *flags |= MASK_RVC, p++; + + /* FIXME: For now we just stop parsing when faced with a + non-standard RISC-V ISA extension, partially becauses of a + problem with the naming scheme. */ + if (*p == 'X') + return; + + if (*p) + { + error ("-march=%s: unsupported ISA substring %s", isa, p); + return; + } +} + +static int +riscv_flags_from_arch_string (const char *isa) +{ + int flags = 0; + riscv_parse_arch_string (isa, &flags); + return flags; +} + +/* Implement TARGET_HANDLE_OPTION. */ + +static bool +riscv_handle_option (struct gcc_options *opts, + struct gcc_options *opts_set ATTRIBUTE_UNUSED, + const struct cl_decoded_option *decoded, + location_t loc ATTRIBUTE_UNUSED) +{ + switch (decoded->opt_index) + { + case OPT_march_: + riscv_parse_arch_string (decoded->arg, &opts->x_target_flags); + return true; + + default: + return true; + } +} + +/* Implement TARGET_OPTION_OPTIMIZATION_TABLE. */ +static const struct default_options riscv_option_optimization_table[] = + { + { OPT_LEVELS_1_PLUS, OPT_fsection_anchors, NULL, 1 }, + { OPT_LEVELS_1_PLUS, OPT_fomit_frame_pointer, NULL, 1 }, + { OPT_LEVELS_SIZE, OPT_msave_restore, NULL, 1 }, + { OPT_LEVELS_NONE, 0, NULL, 0 } + }; + +#undef TARGET_OPTION_OPTIMIZATION_TABLE +#define TARGET_OPTION_OPTIMIZATION_TABLE riscv_option_optimization_table + +#undef TARGET_DEFAULT_TARGET_FLAGS +#define TARGET_DEFAULT_TARGET_FLAGS \ + (TARGET_DEFAULT \ + | riscv_flags_from_arch_string (RISCV_ARCH_STRING_DEFAULT) \ + | (TARGET_64BIT_DEFAULT ? 0 : MASK_32BIT)) + +#undef TARGET_HANDLE_OPTION +#define TARGET_HANDLE_OPTION riscv_handle_option + +struct gcc_targetm_common targetm_common = TARGETM_COMMON_INITIALIZER; diff --git a/gcc/config/riscv/constraints.md b/gcc/config/riscv/constraints.md new file mode 100644 index 0000000..19dbbd7 --- /dev/null +++ b/gcc/config/riscv/constraints.md @@ -0,0 +1,93 @@ +;; Constraint definitions for RISC-V target. +;; Copyright (C) 2011-2014 Free Software Foundation, Inc. +;; Contributed by Andrew Waterman (waterman@cs.berkeley.edu) at UC Berkeley. +;; Based on MIPS target for GNU compiler. +;; +;; This file is part of GCC. +;; +;; GCC is free software; you can redistribute it and/or modify +;; it under the terms of the GNU General Public License as published by +;; the Free Software Foundation; either version 3, or (at your option) +;; any later version. +;; +;; GCC is distributed in the hope that it will be useful, +;; but WITHOUT ANY WARRANTY; without even the implied warranty of +;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +;; GNU General Public License for more details. +;; +;; You should have received a copy of the GNU General Public License +;; along with GCC; see the file COPYING3. If not see +;; . + +;; Register constraints + +(define_register_constraint "f" "TARGET_HARD_FLOAT ? FP_REGS : NO_REGS" + "A floating-point register (if available).") + +(define_register_constraint "b" "ALL_REGS" + "@internal") + +(define_register_constraint "j" "T_REGS" + "@internal") + +(define_register_constraint "l" "JALR_REGS" + "@internal") + +;; Integer constraints + +(define_constraint "Z" + "@internal" + (and (match_code "const_int") + (match_test "1"))) + +(define_constraint "I" + "An I-type 12-bit signed immediate." + (and (match_code "const_int") + (match_test "SMALL_OPERAND (ival)"))) + +(define_constraint "J" + "Integer zero." + (and (match_code "const_int") + (match_test "ival == 0"))) + +;; Floating-point constraints + +(define_constraint "G" + "Floating-point zero." + (and (match_code "const_double") + (match_test "op == CONST0_RTX (mode)"))) + +;; General constraints + +(define_constraint "Q" + "@internal" + (match_operand 0 "const_arith_operand")) + +(define_memory_constraint "A" + "An address that is held in a general-purpose register." + (and (match_code "mem") + (match_test "GET_CODE(XEXP(op,0)) == REG"))) + +(define_constraint "S" + "@internal + A constant call address." + (and (match_operand 0 "call_insn_operand") + (match_test "CONSTANT_P (op)"))) + +(define_constraint "T" + "@internal + A constant @code{move_operand}." + (and (match_operand 0 "move_operand") + (match_test "CONSTANT_P (op)"))) + +(define_memory_constraint "W" + "@internal + A memory address based on a member of @code{BASE_REG_CLASS}." + (and (match_code "mem") + (match_operand 0 "memory_operand"))) + +(define_constraint "YG" + "@internal + A vector zero." + (and (match_code "const_vector") + (match_test "op == CONST0_RTX (mode)"))) diff --git a/gcc/config/riscv/default-32.h b/gcc/config/riscv/default-32.h new file mode 100644 index 0000000..4fac35f --- /dev/null +++ b/gcc/config/riscv/default-32.h @@ -0,0 +1,22 @@ +/* Definitions of target machine for GCC, for RISC-V, + defaulting to 32-bit code generation. + + Copyright (C) 1999-2014 Free Software Foundation, Inc. + +This file is part of GCC. + +GCC is free software; you can redistribute it and/or modify +it under the terms of the GNU General Public License as published by +the Free Software Foundation; either version 3, or (at your option) +any later version. + +GCC is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +You should have received a copy of the GNU General Public License +along with GCC; see the file COPYING3. If not see +. */ + +#define TARGET_64BIT_DEFAULT 0 diff --git a/gcc/config/riscv/elf.h b/gcc/config/riscv/elf.h new file mode 100644 index 0000000..491ec8b --- /dev/null +++ b/gcc/config/riscv/elf.h @@ -0,0 +1,31 @@ +/* Target macros for riscv*-elf targets. + Copyright (C) 1994, 1997, 1999, 2000, 2002, 2003, 2004, 2007, 2010 + Free Software Foundation, Inc. + +This file is part of GCC. + +GCC is free software; you can redistribute it and/or modify +it under the terms of the GNU General Public License as published by +the Free Software Foundation; either version 3, or (at your option) +any later version. + +GCC is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +You should have received a copy of the GNU General Public License +along with GCC; see the file COPYING3. If not see +. */ + +/* Leave the linker script to choose the appropriate libraries. */ +#undef LIB_SPEC +#define LIB_SPEC "" + +#undef STARTFILE_SPEC +#define STARTFILE_SPEC "crt0%O%s crtbegin%O%s" + +#undef ENDFILE_SPEC +#define ENDFILE_SPEC "crtend%O%s" + +#define NO_IMPLICIT_EXTERN_C 1 diff --git a/gcc/config/riscv/generic.md b/gcc/config/riscv/generic.md new file mode 100644 index 0000000..a6dc8fa7a --- /dev/null +++ b/gcc/config/riscv/generic.md @@ -0,0 +1,98 @@ +;; Generic DFA-based pipeline description for RISC-V targets. +;; Copyright (C) 2011-2014 Free Software Foundation, Inc. +;; Contributed by Andrew Waterman (waterman@cs.berkeley.edu) at UC Berkeley. +;; Based on MIPS target for GNU compiler. + +;; This file is part of GCC. + +;; GCC is free software; you can redistribute it and/or modify it +;; under the terms of the GNU General Public License as published +;; by the Free Software Foundation; either version 3, or (at your +;; option) any later version. + +;; GCC is distributed in the hope that it will be useful, but WITHOUT +;; ANY WARRANTY; without even the implied warranty of MERCHANTABILITY +;; or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public +;; License for more details. + +;; You should have received a copy of the GNU General Public License +;; along with GCC; see the file COPYING3. If not see +;; . + + +;; This file is derived from the old define_function_unit description. +;; Each reservation can be overridden on a processor-by-processor basis. + +(define_insn_reservation "generic_alu" 1 + (eq_attr "type" "unknown,const,arith,shift,slt,multi,nop,logical,move") + "alu") + +(define_insn_reservation "generic_load" 3 + (eq_attr "type" "load,fpload,fpidxload") + "alu") + +(define_insn_reservation "generic_store" 1 + (eq_attr "type" "store,fpstore,fpidxstore") + "alu") + +(define_insn_reservation "generic_xfer" 2 + (eq_attr "type" "mfc,mtc") + "alu") + +(define_insn_reservation "generic_branch" 1 + (eq_attr "type" "branch,jump,call") + "alu") + +(define_insn_reservation "generic_imul" 17 + (eq_attr "type" "imul") + "imuldiv*17") + +(define_insn_reservation "generic_idiv" 38 + (eq_attr "type" "idiv") + "imuldiv*38") + +(define_insn_reservation "generic_fcvt" 1 + (eq_attr "type" "fcvt") + "alu") + +(define_insn_reservation "generic_fmove" 2 + (eq_attr "type" "fmove") + "alu") + +(define_insn_reservation "generic_fcmp" 3 + (eq_attr "type" "fcmp") + "alu") + +(define_insn_reservation "generic_fadd" 4 + (eq_attr "type" "fadd") + "alu") + +(define_insn_reservation "generic_fmul_single" 7 + (and (eq_attr "type" "fmul,fmadd") + (eq_attr "mode" "SF")) + "alu") + +(define_insn_reservation "generic_fmul_double" 8 + (and (eq_attr "type" "fmul,fmadd") + (eq_attr "mode" "DF")) + "alu") + +(define_insn_reservation "generic_fdiv_single" 23 + (and (eq_attr "type" "fdiv") + (eq_attr "mode" "SF")) + "alu") + +(define_insn_reservation "generic_fdiv_double" 36 + (and (eq_attr "type" "fdiv") + (eq_attr "mode" "DF")) + "alu") + +(define_insn_reservation "generic_fsqrt_single" 54 + (and (eq_attr "type" "fsqrt") + (eq_attr "mode" "SF")) + "alu") + +(define_insn_reservation "generic_fsqrt_double" 112 + (and (eq_attr "type" "fsqrt") + (eq_attr "mode" "DF")) + "alu") diff --git a/gcc/config/riscv/linux.h b/gcc/config/riscv/linux.h new file mode 100644 index 0000000..3d5f44b --- /dev/null +++ b/gcc/config/riscv/linux.h @@ -0,0 +1,60 @@ +/* Definitions for RISC-V GNU/Linux systems with ELF format. + Copyright (C) 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, + 2007, 2008, 2010, 2011 Free Software Foundation, Inc. + +This file is part of GCC. + +GCC is free software; you can redistribute it and/or modify +it under the terms of the GNU General Public License as published by +the Free Software Foundation; either version 3, or (at your option) +any later version. + +GCC is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +You should have received a copy of the GNU General Public License +along with GCC; see the file COPYING3. If not see +. */ + +#undef WCHAR_TYPE +#define WCHAR_TYPE "int" + +#undef WCHAR_TYPE_SIZE +#define WCHAR_TYPE_SIZE 32 + +#define TARGET_OS_CPP_BUILTINS() \ + do { \ + GNU_USER_TARGET_OS_CPP_BUILTINS(); \ + /* The GNU C++ standard library requires this. */ \ + if (c_dialect_cxx ()) \ + builtin_define ("_GNU_SOURCE"); \ + } while (0) + +#undef SUBTARGET_CPP_SPEC +#define SUBTARGET_CPP_SPEC "%{posix:-D_POSIX_SOURCE} %{pthread:-D_REENTRANT}" + +#define GLIBC_DYNAMIC_LINKER "/lib/ld.so.1" + +/* Borrowed from sparc/linux.h */ +#undef LINK_SPEC +#define LINK_SPEC \ + "%{shared:-shared} \ + %{!shared: \ + %{!static: \ + %{rdynamic:-export-dynamic} \ + -dynamic-linker " GNU_USER_DYNAMIC_LINKER "} \ + %{static:-static}}" + +#undef LIB_SPEC +#define LIB_SPEC "\ +%{pthread:-lpthread} \ +%{shared:-lc} \ +%{!shared: \ + %{profile:-lc_p} %{!profile:-lc}}" + +/* Similar to standard Linux, but adding -ffast-math support. */ +#undef ENDFILE_SPEC +#define ENDFILE_SPEC \ + "%{shared|pie:crtendS.o%s;:crtend.o%s} crtn.o%s" diff --git a/gcc/config/riscv/linux64.h b/gcc/config/riscv/linux64.h new file mode 100644 index 0000000..56e0900 --- /dev/null +++ b/gcc/config/riscv/linux64.h @@ -0,0 +1,43 @@ +/* Definitions for 64-bit RISC-V GNU/Linux systems with ELF format. + Copyright 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2010, 2011 + Free Software Foundation, Inc. + +This file is part of GCC. + +GCC is free software; you can redistribute it and/or modify +it under the terms of the GNU General Public License as published by +the Free Software Foundation; either version 3, or (at your option) +any later version. + +GCC is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +You should have received a copy of the GNU General Public License +along with GCC; see the file COPYING3. If not see +. */ + +/* Force the default ABI flags onto the command line + in order to make the other specs easier to write. */ +#undef LIB_SPEC +#define LIB_SPEC "\ +%{pthread:-lpthread} \ +%{shared:-lc} \ +%{!shared: \ + %{profile:-lc_p} %{!profile:-lc}}" + +#define GLIBC_DYNAMIC_LINKER32 "/lib32/ld.so.1" +#define GLIBC_DYNAMIC_LINKER64 "/lib/ld.so.1" + +#undef LINK_SPEC +#define LINK_SPEC "\ +%{shared} \ + %{!shared: \ + %{!static: \ + %{rdynamic:-export-dynamic} \ + %{" OPT_ARCH64 ": -dynamic-linker " GNU_USER_DYNAMIC_LINKER64 "} \ + %{" OPT_ARCH32 ": -dynamic-linker " GNU_USER_DYNAMIC_LINKER32 "}} \ + %{static:-static}} \ +%{" OPT_ARCH64 ":-melf64lriscv} \ +%{" OPT_ARCH32 ":-melf32lriscv}" diff --git a/gcc/config/riscv/opcode-riscv.h b/gcc/config/riscv/opcode-riscv.h new file mode 100644 index 0000000..a5435ac --- /dev/null +++ b/gcc/config/riscv/opcode-riscv.h @@ -0,0 +1,149 @@ +/* RISC-V ISA encoding. + Copyright (C) 2011-2014 Free Software Foundation, Inc. + Contributed by Andrew Waterman (waterman@cs.berkeley.edu) at UC Berkeley. + Based on MIPS target for GNU compiler. + +This file is part of GDB, GAS, and the GNU binutils. + +GDB, GAS, and the GNU binutils are free software; you can redistribute +them and/or modify them under the terms of the GNU General Public +License as published by the Free Software Foundation; either version +1, or (at your option) any later version. + +GDB, GAS, and the GNU binutils are distributed in the hope that they +will be useful, but WITHOUT ANY WARRANTY; without even the implied +warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See +the GNU General Public License for more details. + +You should have received a copy of the GNU General Public License +along with this file; see the file COPYING. If not, write to the Free +Software Foundation, 51 Franklin Street - Fifth Floor, Boston, MA 02110-1301, USA. */ + +#ifndef _RISCV_H_ +#define _RISCV_H_ + +#define RV_X(x, s, n) (((x) >> (s)) & ((1<<(n))-1)) +#define RV_IMM_SIGN(x) (-(((x) >> 31) & 1)) + +#define EXTRACT_ITYPE_IMM(x) \ + (RV_X(x, 20, 12) | (RV_IMM_SIGN(x) << 12)) +#define EXTRACT_STYPE_IMM(x) \ + (RV_X(x, 7, 5) | (RV_X(x, 25, 7) << 5) | (RV_IMM_SIGN(x) << 12)) +#define EXTRACT_SBTYPE_IMM(x) \ + ((RV_X(x, 8, 4) << 1) | (RV_X(x, 25, 6) << 5) | (RV_X(x, 7, 1) << 11) | (RV_IMM_SIGN(x) << 12)) +#define EXTRACT_UTYPE_IMM(x) \ + ((RV_X(x, 12, 20) << 20) | (RV_IMM_SIGN(x) << 32)) +#define EXTRACT_UJTYPE_IMM(x) \ + ((RV_X(x, 21, 10) << 1) | (RV_X(x, 20, 1) << 11) | (RV_X(x, 12, 8) << 12) | (RV_IMM_SIGN(x) << 20)) + +#define ENCODE_ITYPE_IMM(x) \ + (RV_X(x, 0, 12) << 20) +#define ENCODE_STYPE_IMM(x) \ + ((RV_X(x, 0, 5) << 7) | (RV_X(x, 5, 7) << 25)) +#define ENCODE_SBTYPE_IMM(x) \ + ((RV_X(x, 1, 4) << 8) | (RV_X(x, 5, 6) << 25) | (RV_X(x, 11, 1) << 7) | (RV_X(x, 12, 1) << 31)) +#define ENCODE_UTYPE_IMM(x) \ + (RV_X(x, 12, 20) << 12) +#define ENCODE_UJTYPE_IMM(x) \ + ((RV_X(x, 1, 10) << 21) | (RV_X(x, 11, 1) << 20) | (RV_X(x, 12, 8) << 12) | (RV_X(x, 20, 1) << 31)) + +#define VALID_ITYPE_IMM(x) (EXTRACT_ITYPE_IMM(ENCODE_ITYPE_IMM(x)) == (x)) +#define VALID_STYPE_IMM(x) (EXTRACT_STYPE_IMM(ENCODE_STYPE_IMM(x)) == (x)) +#define VALID_SBTYPE_IMM(x) (EXTRACT_SBTYPE_IMM(ENCODE_SBTYPE_IMM(x)) == (x)) +#define VALID_UTYPE_IMM(x) (EXTRACT_UTYPE_IMM(ENCODE_UTYPE_IMM(x)) == (x)) +#define VALID_UJTYPE_IMM(x) (EXTRACT_UJTYPE_IMM(ENCODE_UJTYPE_IMM(x)) == (x)) + +#define RISCV_RTYPE(insn, rd, rs1, rs2) \ + ((MATCH_ ## insn) | ((rd) << OP_SH_RD) | ((rs1) << OP_SH_RS1) | ((rs2) << OP_SH_RS2)) +#define RISCV_ITYPE(insn, rd, rs1, imm) \ + ((MATCH_ ## insn) | ((rd) << OP_SH_RD) | ((rs1) << OP_SH_RS1) | ENCODE_ITYPE_IMM(imm)) +#define RISCV_STYPE(insn, rs1, rs2, imm) \ + ((MATCH_ ## insn) | ((rs1) << OP_SH_RS1) | ((rs2) << OP_SH_RS2) | ENCODE_STYPE_IMM(imm)) +#define RISCV_SBTYPE(insn, rs1, rs2, target) \ + ((MATCH_ ## insn) | ((rs1) << OP_SH_RS1) | ((rs2) << OP_SH_RS2) | ENCODE_SBTYPE_IMM(target)) +#define RISCV_UTYPE(insn, rd, bigimm) \ + ((MATCH_ ## insn) | ((rd) << OP_SH_RD) | ENCODE_UTYPE_IMM(bigimm)) +#define RISCV_UJTYPE(insn, rd, target) \ + ((MATCH_ ## insn) | ((rd) << OP_SH_RD) | ENCODE_UJTYPE_IMM(target)) + +#define RISCV_NOP RISCV_ITYPE(ADDI, 0, 0, 0) + +#define RISCV_CONST_HIGH_PART(VALUE) \ + (((VALUE) + (RISCV_IMM_REACH/2)) & ~(RISCV_IMM_REACH-1)) +#define RISCV_CONST_LOW_PART(VALUE) ((VALUE) - RISCV_CONST_HIGH_PART (VALUE)) + +/* RV fields */ + +#define OP_MASK_OP 0x7f +#define OP_SH_OP 0 +#define OP_MASK_RS2 0x1f +#define OP_SH_RS2 20 +#define OP_MASK_RS1 0x1f +#define OP_SH_RS1 15 +#define OP_MASK_RS3 0x1f +#define OP_SH_RS3 27 +#define OP_MASK_RD 0x1f +#define OP_SH_RD 7 +#define OP_MASK_SHAMT 0x3f +#define OP_SH_SHAMT 20 +#define OP_MASK_SHAMTW 0x1f +#define OP_SH_SHAMTW 20 +#define OP_MASK_RM 0x7 +#define OP_SH_RM 12 +#define OP_MASK_PRED 0xf +#define OP_SH_PRED 24 +#define OP_MASK_SUCC 0xf +#define OP_SH_SUCC 20 +#define OP_MASK_AQ 0x1 +#define OP_SH_AQ 26 +#define OP_MASK_RL 0x1 +#define OP_SH_RL 25 + +#define OP_MASK_VRD 0x1f +#define OP_SH_VRD 7 +#define OP_MASK_VRS 0x1f +#define OP_SH_VRS 15 +#define OP_MASK_VRT 0x1f +#define OP_SH_VRT 20 +#define OP_MASK_VRR 0x1f +#define OP_SH_VRR 25 + +#define OP_MASK_VFD 0x1f +#define OP_SH_VFD 7 +#define OP_MASK_VFS 0x1f +#define OP_SH_VFS 15 +#define OP_MASK_VFT 0x1f +#define OP_SH_VFT 20 +#define OP_MASK_VFR 0x1f +#define OP_SH_VFR 25 + +#define OP_MASK_IMMNGPR 0x3f +#define OP_SH_IMMNGPR 20 +#define OP_MASK_IMMNFPR 0x3f +#define OP_SH_IMMNFPR 26 +#define OP_MASK_IMMSEGNELM 0x1f +#define OP_SH_IMMSEGNELM 17 +#define OP_MASK_IMMSEGSTNELM 0x1f +#define OP_SH_IMMSEGSTNELM 12 +#define OP_MASK_CUSTOM_IMM 0x7f +#define OP_SH_CUSTOM_IMM 25 + +#define LINK_REG 1 + +#define RISCV_JUMP_BITS RISCV_BIGIMM_BITS +#define RISCV_JUMP_ALIGN_BITS 1 +#define RISCV_JUMP_ALIGN (1 << RISCV_JUMP_ALIGN_BITS) +#define RISCV_JUMP_REACH ((1ULL< SI optimizations +;;........................ + +;; Simplify (int)(a + 1), etc. +(define_peephole2 + [(set (match_operand:DI 0 "register_operand") + (match_operator:DI 4 "modular_operator" + [(match_operand:DI 1 "register_operand") + (match_operand:DI 2 "arith_operand")])) + (set (match_operand:SI 3 "register_operand") + (truncate:SI (match_dup 0)))] + "TARGET_64BIT && (REGNO (operands[0]) == REGNO (operands[3]) || peep2_reg_dead_p (2, operands[0])) + && (GET_CODE (operands[4]) != ASHIFT || (CONST_INT_P (operands[2]) && INTVAL (operands[2]) < 32))" + [(set (match_dup 3) + (truncate:SI + (match_op_dup:DI 4 + [(match_operand:DI 1 "register_operand") + (match_operand:DI 2 "arith_operand")])))]) + +;; Simplify (int)a + 1, etc. +(define_peephole2 + [(set (match_operand:SI 0 "register_operand") + (truncate:SI (match_operand:DI 1 "register_operand"))) + (set (match_operand:SI 3 "register_operand") + (match_operator:SI 4 "modular_operator" + [(match_dup 0) + (match_operand:SI 2 "arith_operand")]))] + "TARGET_64BIT && (REGNO (operands[0]) == REGNO (operands[3]) || peep2_reg_dead_p (2, operands[0]))" + [(set (match_dup 3) + (match_op_dup:SI 4 [(match_dup 1) (match_dup 2)]))]) + +;; Simplify -(int)a, etc. +(define_peephole2 + [(set (match_operand:SI 0 "register_operand") + (truncate:SI (match_operand:DI 2 "register_operand"))) + (set (match_operand:SI 3 "register_operand") + (match_operator:SI 4 "modular_operator" + [(match_operand:SI 1 "reg_or_0_operand") + (match_dup 0)]))] + "TARGET_64BIT && (REGNO (operands[0]) == REGNO (operands[3]) || peep2_reg_dead_p (2, operands[0]))" + [(set (match_dup 3) + (match_op_dup:SI 4 [(match_dup 1) (match_dup 2)]))]) + +;; Simplify PIC loads to static variables. +;; These will go away once we figure out how to emit auipc discretely. +(define_insn "*local_pic_load" + [(set (match_operand:ANYI 0 "register_operand" "=r") + (mem:ANYI (match_operand 1 "absolute_symbolic_operand" "")))] + "flag_pic && SYMBOL_REF_LOCAL_P (operands[1])" + "\t%0,%1" + [(set (attr "length") (const_int 8))]) +(define_insn "*local_pic_load" + [(set (match_operand:ANYF 0 "register_operand" "=f") + (mem:ANYF (match_operand 1 "absolute_symbolic_operand" ""))) + (clobber (match_scratch:DI 2 "=&r"))] + "TARGET_HARD_FLOAT && TARGET_64BIT && flag_pic && SYMBOL_REF_LOCAL_P (operands[1])" + "\t%0,%1,%2" + [(set (attr "length") (const_int 8))]) +(define_insn "*local_pic_load" + [(set (match_operand:ANYF 0 "register_operand" "=f") + (mem:ANYF (match_operand 1 "absolute_symbolic_operand" ""))) + (clobber (match_scratch:SI 2 "=&r"))] + "TARGET_HARD_FLOAT && !TARGET_64BIT && flag_pic && SYMBOL_REF_LOCAL_P (operands[1])" + "\t%0,%1,%2" + [(set (attr "length") (const_int 8))]) +(define_insn "*local_pic_loadu" + [(set (match_operand:SUPERQI 0 "register_operand" "=r") + (zero_extend:SUPERQI (mem:SUBDI (match_operand 1 "absolute_symbolic_operand" ""))))] + "flag_pic && SYMBOL_REF_LOCAL_P (operands[1])" + "u\t%0,%1" + [(set (attr "length") (const_int 8))]) +(define_insn "*local_pic_storedi" + [(set (mem:ANYI (match_operand 0 "absolute_symbolic_operand" "")) + (match_operand:ANYI 1 "reg_or_0_operand" "rJ")) + (clobber (match_scratch:DI 2 "=&r"))] + "TARGET_64BIT && (flag_pic && SYMBOL_REF_LOCAL_P (operands[0]))" + "\t%z1,%0,%2" + [(set (attr "length") (const_int 8))]) +(define_insn "*local_pic_storesi" + [(set (mem:ANYI (match_operand 0 "absolute_symbolic_operand" "")) + (match_operand:ANYI 1 "reg_or_0_operand" "rJ")) + (clobber (match_scratch:SI 2 "=&r"))] + "!TARGET_64BIT && (flag_pic && SYMBOL_REF_LOCAL_P (operands[0]))" + "\t%z1,%0,%2" + [(set (attr "length") (const_int 8))]) +(define_insn "*local_pic_storedi" + [(set (mem:ANYF (match_operand 0 "absolute_symbolic_operand" "")) + (match_operand:ANYF 1 "register_operand" "f")) + (clobber (match_scratch:DI 2 "=&r"))] + "TARGET_HARD_FLOAT && TARGET_64BIT && (flag_pic && SYMBOL_REF_LOCAL_P (operands[0]))" + "\t%1,%0,%2" + [(set (attr "length") (const_int 8))]) +(define_insn "*local_pic_storesi" + [(set (mem:ANYF (match_operand 0 "absolute_symbolic_operand" "")) + (match_operand:ANYF 1 "register_operand" "f")) + (clobber (match_scratch:SI 2 "=&r"))] + "TARGET_HARD_FLOAT && !TARGET_64BIT && (flag_pic && SYMBOL_REF_LOCAL_P (operands[0]))" + "\t%1,%0,%2" + [(set (attr "length") (const_int 8))]) diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md new file mode 100644 index 0000000..b0df24e --- /dev/null +++ b/gcc/config/riscv/predicates.md @@ -0,0 +1,187 @@ +;; Predicate description for RISC-V target. +;; Copyright (C) 2011-2014 Free Software Foundation, Inc. +;; Contributed by Andrew Waterman (waterman@cs.berkeley.edu) at UC Berkeley. +;; Based on MIPS target for GNU compiler. +;; +;; This file is part of GCC. +;; +;; GCC is free software; you can redistribute it and/or modify +;; it under the terms of the GNU General Public License as published by +;; the Free Software Foundation; either version 3, or (at your option) +;; any later version. +;; +;; GCC is distributed in the hope that it will be useful, +;; but WITHOUT ANY WARRANTY; without even the implied warranty of +;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +;; GNU General Public License for more details. +;; +;; You should have received a copy of the GNU General Public License +;; along with GCC; see the file COPYING3. If not see +;; . + +(define_predicate "const_arith_operand" + (and (match_code "const_int") + (match_test "SMALL_OPERAND (INTVAL (op))"))) + +(define_predicate "arith_operand" + (ior (match_operand 0 "const_arith_operand") + (match_operand 0 "register_operand"))) + +(define_predicate "sle_operand" + (and (match_code "const_int") + (match_test "SMALL_OPERAND (INTVAL (op) + 1)"))) + +(define_predicate "sleu_operand" + (and (match_operand 0 "sle_operand") + (match_test "INTVAL (op) + 1 != 0"))) + +(define_predicate "const_0_operand" + (and (match_code "const_int,const_double,const_vector") + (match_test "op == CONST0_RTX (GET_MODE (op))"))) + +(define_predicate "reg_or_0_operand" + (ior (match_operand 0 "const_0_operand") + (match_operand 0 "register_operand"))) + +(define_predicate "const_1_operand" + (and (match_code "const_int,const_double,const_vector") + (match_test "op == CONST1_RTX (GET_MODE (op))"))) + +(define_predicate "reg_or_1_operand" + (ior (match_operand 0 "const_1_operand") + (match_operand 0 "register_operand"))) + +;; Only use branch-on-bit sequences when the mask is not an ANDI immediate. +(define_predicate "branch_on_bit_operand" + (and (match_code "const_int") + (match_test "INTVAL (op) >= RISCV_IMM_BITS - 1"))) + +;; This is used for indexing into vectors, and hence only accepts const_int. +(define_predicate "const_0_or_1_operand" + (and (match_code "const_int") + (ior (match_test "op == CONST0_RTX (GET_MODE (op))") + (match_test "op == CONST1_RTX (GET_MODE (op))")))) + +(define_special_predicate "pc_or_label_operand" + (match_code "pc,label_ref")) + +;; A legitimate CONST_INT operand that takes more than one instruction +;; to load. +(define_predicate "splittable_const_int_operand" + (match_code "const_int") +{ + /* Don't handle multi-word moves this way; we don't want to introduce + the individual word-mode moves until after reload. */ + if (GET_MODE_SIZE (mode) > UNITS_PER_WORD) + return false; + + /* Otherwise check whether the constant can be loaded in a single + instruction. */ + return !LUI_INT (op) && !SMALL_INT (op); +}) + +(define_predicate "move_operand" + (match_operand 0 "general_operand") +{ + enum riscv_symbol_type symbol_type; + + /* The thinking here is as follows: + + (1) The move expanders should split complex load sequences into + individual instructions. Those individual instructions can + then be optimized by all rtl passes. + + (2) The target of pre-reload load sequences should not be used + to store temporary results. If the target register is only + assigned one value, reload can rematerialize that value + on demand, rather than spill it to the stack. + + (3) If we allowed pre-reload passes like combine and cse to recreate + complex load sequences, we would want to be able to split the + sequences before reload as well, so that the pre-reload scheduler + can see the individual instructions. This falls foul of (2); + the splitter would be forced to reuse the target register for + intermediate results. + + (4) We want to define complex load splitters for combine. These + splitters can request a temporary scratch register, which avoids + the problem in (2). They allow things like: + + (set (reg T1) (high SYM)) + (set (reg T2) (low (reg T1) SYM)) + (set (reg X) (plus (reg T2) (const_int OFFSET))) + + to be combined into: + + (set (reg T3) (high SYM+OFFSET)) + (set (reg X) (lo_sum (reg T3) SYM+OFFSET)) + + if T2 is only used this once. */ + switch (GET_CODE (op)) + { + case CONST_INT: + return !splittable_const_int_operand (op, mode); + + case CONST: + case SYMBOL_REF: + case LABEL_REF: + return (riscv_symbolic_constant_p (op, &symbol_type) + && !riscv_hi_relocs[symbol_type]); + + case HIGH: + op = XEXP (op, 0); + return riscv_symbolic_constant_p (op, &symbol_type); + + default: + return true; + } +}) + +(define_predicate "consttable_operand" + (match_test "CONSTANT_P (op)")) + +(define_predicate "symbolic_operand" + (match_code "const,symbol_ref,label_ref") +{ + enum riscv_symbol_type type; + return riscv_symbolic_constant_p (op, &type); +}) + +(define_predicate "absolute_symbolic_operand" + (match_code "const,symbol_ref,label_ref") +{ + enum riscv_symbol_type type; + return (riscv_symbolic_constant_p (op, &type) + && type == SYMBOL_ABSOLUTE); +}) + +(define_predicate "plt_symbolic_operand" + (match_code "const,symbol_ref,label_ref") +{ + enum riscv_symbol_type type; + return (riscv_symbolic_constant_p (op, &type) + && type == SYMBOL_GOT_DISP && !SYMBOL_REF_WEAK (op) && TARGET_PLT); +}) + +(define_predicate "call_insn_operand" + (ior (match_operand 0 "absolute_symbolic_operand") + (match_operand 0 "plt_symbolic_operand") + (match_operand 0 "register_operand"))) + +(define_predicate "symbol_ref_operand" + (match_code "symbol_ref")) + +(define_predicate "modular_operator" + (match_code "plus,minus,mult,ashift")) + +(define_predicate "equality_operator" + (match_code "eq,ne")) + +(define_predicate "order_operator" + (match_code "eq,ne,lt,ltu,le,leu,ge,geu,gt,gtu")) + +(define_predicate "fp_order_operator" + (match_code "eq,lt,le,gt,ge")) + +(define_predicate "fp_unorder_operator" + (match_code "ordered,unordered")) diff --git a/gcc/config/riscv/riscv-ftypes.def b/gcc/config/riscv/riscv-ftypes.def new file mode 100644 index 0000000..96a38f1 --- /dev/null +++ b/gcc/config/riscv/riscv-ftypes.def @@ -0,0 +1,39 @@ +/* Definitions of prototypes for RISC-V built-in functions. + Copyright (C) 2011-2014 Free Software Foundation, Inc. + Contributed by Andrew Waterman (waterman@cs.berkeley.edu) at UC Berkeley. + Based on MIPS target for GNU compiler. + +This file is part of GCC. + +GCC is free software; you can redistribute it and/or modify +it under the terms of the GNU General Public License as published by +the Free Software Foundation; either version 3, or (at your option) +any later version. + +GCC is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +You should have received a copy of the GNU General Public License +along with GCC; see the file COPYING3. If not see +. */ + +/* Invoke DEF_RISCV_FTYPE (NARGS, LIST) for each prototype used by + MIPS built-in functions, where: + + NARGS is the number of arguments. + LIST contains the return-type code followed by the codes for each + argument type. + + Argument- and return-type codes are either modes or one of the following: + + VOID for void_type_node + INT for integer_type_node + POINTER for ptr_type_node + + (we don't use PTR because that's a ANSI-compatibillity macro). + + Please keep this list lexicographically sorted by the LIST argument. */ + +DEF_RISCV_FTYPE (1, (VOID, VOID)) diff --git a/gcc/config/riscv/riscv-modes.def b/gcc/config/riscv/riscv-modes.def new file mode 100644 index 0000000..bb42344 --- /dev/null +++ b/gcc/config/riscv/riscv-modes.def @@ -0,0 +1,26 @@ +/* Extra machine modes for RISC-V target. + Copyright (C) 2011-2014 Free Software Foundation, Inc. + Contributed by Andrew Waterman (waterman@cs.berkeley.edu) at UC Berkeley. + Based on MIPS target for GNU compiler. + +This file is part of GCC. + +GCC is free software; you can redistribute it and/or modify +it under the terms of the GNU General Public License as published by +the Free Software Foundation; either version 3, or (at your option) +any later version. + +GCC is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +You should have received a copy of the GNU General Public License +along with GCC; see the file COPYING3. If not see +. */ + +FLOAT_MODE (TF, 16, ieee_quad_format); + +/* Vector modes. */ +VECTOR_MODES (INT, 4); /* V8QI V4HI V2SI */ +VECTOR_MODES (FLOAT, 4); /* V4HF V2SF */ diff --git a/gcc/config/riscv/riscv-opc.h b/gcc/config/riscv/riscv-opc.h new file mode 100644 index 0000000..21cabff --- /dev/null +++ b/gcc/config/riscv/riscv-opc.h @@ -0,0 +1,1348 @@ +/* Automatically generated by parse-opcodes */ +#ifndef RISCV_ENCODING_H +#define RISCV_ENCODING_H +#define MATCH_ADD 0x33 +#define MASK_ADD 0xfe00707f +#define MATCH_ADDI 0x13 +#define MASK_ADDI 0x707f +#define MATCH_ADDIW 0x1b +#define MASK_ADDIW 0x707f +#define MATCH_ADDW 0x3b +#define MASK_ADDW 0xfe00707f +#define MATCH_AMOADD_D 0x302f +#define MASK_AMOADD_D 0xf800707f +#define MATCH_AMOADD_W 0x202f +#define MASK_AMOADD_W 0xf800707f +#define MATCH_AMOAND_D 0x6000302f +#define MASK_AMOAND_D 0xf800707f +#define MATCH_AMOAND_W 0x6000202f +#define MASK_AMOAND_W 0xf800707f +#define MATCH_AMOMAX_D 0xa000302f +#define MASK_AMOMAX_D 0xf800707f +#define MATCH_AMOMAX_W 0xa000202f +#define MASK_AMOMAX_W 0xf800707f +#define MATCH_AMOMAXU_D 0xe000302f +#define MASK_AMOMAXU_D 0xf800707f +#define MATCH_AMOMAXU_W 0xe000202f +#define MASK_AMOMAXU_W 0xf800707f +#define MATCH_AMOMIN_D 0x8000302f +#define MASK_AMOMIN_D 0xf800707f +#define MATCH_AMOMIN_W 0x8000202f +#define MASK_AMOMIN_W 0xf800707f +#define MATCH_AMOMINU_D 0xc000302f +#define MASK_AMOMINU_D 0xf800707f +#define MATCH_AMOMINU_W 0xc000202f +#define MASK_AMOMINU_W 0xf800707f +#define MATCH_AMOOR_D 0x4000302f +#define MASK_AMOOR_D 0xf800707f +#define MATCH_AMOOR_W 0x4000202f +#define MASK_AMOOR_W 0xf800707f +#define MATCH_AMOSWAP_D 0x800302f +#define MASK_AMOSWAP_D 0xf800707f +#define MATCH_AMOSWAP_W 0x800202f +#define MASK_AMOSWAP_W 0xf800707f +#define MATCH_AMOXOR_D 0x2000302f +#define MASK_AMOXOR_D 0xf800707f +#define MATCH_AMOXOR_W 0x2000202f +#define MASK_AMOXOR_W 0xf800707f +#define MATCH_AND 0x7033 +#define MASK_AND 0xfe00707f +#define MATCH_ANDI 0x7013 +#define MASK_ANDI 0x707f +#define MATCH_AUIPC 0x17 +#define MASK_AUIPC 0x7f +#define MATCH_BEQ 0x63 +#define MASK_BEQ 0x707f +#define MATCH_BGE 0x5063 +#define MASK_BGE 0x707f +#define MATCH_BGEU 0x7063 +#define MASK_BGEU 0x707f +#define MATCH_BLT 0x4063 +#define MASK_BLT 0x707f +#define MATCH_BLTU 0x6063 +#define MASK_BLTU 0x707f +#define MATCH_BNE 0x1063 +#define MASK_BNE 0x707f +#define MATCH_C_ADD 0x1000 +#define MASK_C_ADD 0xf003 +#define MATCH_C_ADD3 0xa000 +#define MASK_C_ADD3 0xe063 +#define MATCH_C_ADDI 0xc002 +#define MASK_C_ADDI 0xe003 +#define MATCH_C_ADDI16SP 0xc002 +#define MASK_C_ADDI16SP 0xef83 +#define MATCH_C_ADDI4SPN 0xa001 +#define MASK_C_ADDI4SPN 0xe003 +#define MATCH_C_ADDIN 0x8001 +#define MASK_C_ADDIN 0xe063 +#define MATCH_C_ADDIW 0xe002 +#define MASK_C_ADDIW 0xe003 +#define MATCH_C_ADDW 0x9000 +#define MASK_C_ADDW 0xf003 +#define MATCH_C_AND3 0xa060 +#define MASK_C_AND3 0xe063 +#define MATCH_C_ANDI 0xe002 +#define MASK_C_ANDI 0xe003 +#define MATCH_C_ANDIN 0x8061 +#define MASK_C_ANDIN 0xe063 +#define MATCH_C_BEQZ 0x4002 +#define MASK_C_BEQZ 0xe003 +#define MATCH_C_BGEZ 0xe001 +#define MASK_C_BGEZ 0xe003 +#define MATCH_C_BLTZ 0x6001 +#define MASK_C_BLTZ 0xe003 +#define MATCH_C_BNEZ 0x6002 +#define MASK_C_BNEZ 0xe003 +#define MATCH_C_EBREAK 0x1000 +#define MASK_C_EBREAK 0xffff +#define MATCH_C_J 0x2 +#define MASK_C_J 0xe003 +#define MATCH_C_JAL 0x2002 +#define MASK_C_JAL 0xe003 +#define MATCH_C_JALR 0xa002 +#define MASK_C_JALR 0xf07f +#define MATCH_C_JR 0x8002 +#define MASK_C_JR 0xf07f +#define MATCH_C_LD 0xe000 +#define MASK_C_LD 0xe003 +#define MATCH_C_LDSP 0xe001 +#define MASK_C_LDSP 0xe003 +#define MATCH_C_LI 0x8002 +#define MASK_C_LI 0xe003 +#define MATCH_C_LUI 0xa002 +#define MASK_C_LUI 0xe003 +#define MATCH_C_LW 0xc000 +#define MASK_C_LW 0xe003 +#define MATCH_C_LWSP 0xc001 +#define MASK_C_LWSP 0xe003 +#define MATCH_C_MV 0x0 +#define MASK_C_MV 0xf003 +#define MATCH_C_OR3 0xa040 +#define MASK_C_OR3 0xe063 +#define MATCH_C_ORIN 0x8041 +#define MASK_C_ORIN 0xe063 +#define MATCH_C_SD 0x6000 +#define MASK_C_SD 0xe003 +#define MATCH_C_SDSP 0x6001 +#define MASK_C_SDSP 0xe003 +#define MATCH_C_SLL 0x6400 +#define MASK_C_SLL 0xfc63 +#define MATCH_C_SLLI 0x1 +#define MASK_C_SLLI 0xe003 +#define MATCH_C_SLLIW 0x8001 +#define MASK_C_SLLIW 0xe003 +#define MATCH_C_SLLR 0x6c00 +#define MASK_C_SLLR 0xfc63 +#define MATCH_C_SLT 0x6440 +#define MASK_C_SLT 0xfc63 +#define MATCH_C_SLTR 0x6c40 +#define MASK_C_SLTR 0xfc63 +#define MATCH_C_SLTU 0x6460 +#define MASK_C_SLTU 0xfc63 +#define MATCH_C_SLTUR 0x6c60 +#define MASK_C_SLTUR 0xfc63 +#define MATCH_C_SRA 0x6020 +#define MASK_C_SRA 0xfc63 +#define MATCH_C_SRAI 0x2000 +#define MASK_C_SRAI 0xe003 +#define MATCH_C_SRL 0x6420 +#define MASK_C_SRL 0xfc63 +#define MATCH_C_SRLI 0x2001 +#define MASK_C_SRLI 0xe003 +#define MATCH_C_SRLR 0x6c20 +#define MASK_C_SRLR 0xfc63 +#define MATCH_C_SUB 0x8000 +#define MASK_C_SUB 0xf003 +#define MATCH_C_SUB3 0xa020 +#define MASK_C_SUB3 0xe063 +#define MATCH_C_SW 0x4000 +#define MASK_C_SW 0xe003 +#define MATCH_C_SWSP 0x4001 +#define MASK_C_SWSP 0xe003 +#define MATCH_C_XOR 0x6000 +#define MASK_C_XOR 0xfc63 +#define MATCH_C_XORIN 0x8021 +#define MASK_C_XORIN 0xe063 +#define MATCH_CSRRC 0x3073 +#define MASK_CSRRC 0x707f +#define MATCH_CSRRCI 0x7073 +#define MASK_CSRRCI 0x707f +#define MATCH_CSRRS 0x2073 +#define MASK_CSRRS 0x707f +#define MATCH_CSRRSI 0x6073 +#define MASK_CSRRSI 0x707f +#define MATCH_CSRRW 0x1073 +#define MASK_CSRRW 0x707f +#define MATCH_CSRRWI 0x5073 +#define MASK_CSRRWI 0x707f +#define MATCH_CUSTOM0 0xb +#define MASK_CUSTOM0 0x707f +#define MATCH_CUSTOM0_RD 0x400b +#define MASK_CUSTOM0_RD 0x707f +#define MATCH_CUSTOM0_RD_RS1 0x600b +#define MASK_CUSTOM0_RD_RS1 0x707f +#define MATCH_CUSTOM0_RD_RS1_RS2 0x700b +#define MASK_CUSTOM0_RD_RS1_RS2 0x707f +#define MATCH_CUSTOM0_RS1 0x200b +#define MASK_CUSTOM0_RS1 0x707f +#define MATCH_CUSTOM0_RS1_RS2 0x300b +#define MASK_CUSTOM0_RS1_RS2 0x707f +#define MATCH_CUSTOM1 0x2b +#define MASK_CUSTOM1 0x707f +#define MATCH_CUSTOM1_RD 0x402b +#define MASK_CUSTOM1_RD 0x707f +#define MATCH_CUSTOM1_RD_RS1 0x602b +#define MASK_CUSTOM1_RD_RS1 0x707f +#define MATCH_CUSTOM1_RD_RS1_RS2 0x702b +#define MASK_CUSTOM1_RD_RS1_RS2 0x707f +#define MATCH_CUSTOM1_RS1 0x202b +#define MASK_CUSTOM1_RS1 0x707f +#define MATCH_CUSTOM1_RS1_RS2 0x302b +#define MASK_CUSTOM1_RS1_RS2 0x707f +#define MATCH_CUSTOM2 0x5b +#define MASK_CUSTOM2 0x707f +#define MATCH_CUSTOM2_RD 0x405b +#define MASK_CUSTOM2_RD 0x707f +#define MATCH_CUSTOM2_RD_RS1 0x605b +#define MASK_CUSTOM2_RD_RS1 0x707f +#define MATCH_CUSTOM2_RD_RS1_RS2 0x705b +#define MASK_CUSTOM2_RD_RS1_RS2 0x707f +#define MATCH_CUSTOM2_RS1 0x205b +#define MASK_CUSTOM2_RS1 0x707f +#define MATCH_CUSTOM2_RS1_RS2 0x305b +#define MASK_CUSTOM2_RS1_RS2 0x707f +#define MATCH_CUSTOM3 0x7b +#define MASK_CUSTOM3 0x707f +#define MATCH_CUSTOM3_RD 0x407b +#define MASK_CUSTOM3_RD 0x707f +#define MATCH_CUSTOM3_RD_RS1 0x607b +#define MASK_CUSTOM3_RD_RS1 0x707f +#define MATCH_CUSTOM3_RD_RS1_RS2 0x707b +#define MASK_CUSTOM3_RD_RS1_RS2 0x707f +#define MATCH_CUSTOM3_RS1 0x207b +#define MASK_CUSTOM3_RS1 0x707f +#define MATCH_CUSTOM3_RS1_RS2 0x307b +#define MASK_CUSTOM3_RS1_RS2 0x707f +#define MATCH_DIV 0x2004033 +#define MASK_DIV 0xfe00707f +#define MATCH_DIVU 0x2005033 +#define MASK_DIVU 0xfe00707f +#define MATCH_DIVUW 0x200503b +#define MASK_DIVUW 0xfe00707f +#define MATCH_DIVW 0x200403b +#define MASK_DIVW 0xfe00707f +#define MATCH_EBREAK 0x100073 +#define MASK_EBREAK 0xffffffff +#define MATCH_ECALL 0x73 +#define MASK_ECALL 0xffffffff +#define MATCH_ERET 0x10000073 +#define MASK_ERET 0xffffffff +#define MATCH_FADD_D 0x2000053 +#define MASK_FADD_D 0xfe00007f +#define MATCH_FADD_H 0x4000053 +#define MASK_FADD_H 0xfe00007f +#define MATCH_FADD_S 0x53 +#define MASK_FADD_S 0xfe00007f +#define MATCH_FCLASS_D 0xe2001053 +#define MASK_FCLASS_D 0xfff0707f +#define MATCH_FCLASS_S 0xe0001053 +#define MASK_FCLASS_S 0xfff0707f +#define MATCH_FCVT_D_H 0x8c000053 +#define MASK_FCVT_D_H 0xfff0007f +#define MATCH_FCVT_D_L 0xd2200053 +#define MASK_FCVT_D_L 0xfff0007f +#define MATCH_FCVT_D_LU 0xd2300053 +#define MASK_FCVT_D_LU 0xfff0007f +#define MATCH_FCVT_D_S 0x42000053 +#define MASK_FCVT_D_S 0xfff0007f +#define MATCH_FCVT_D_W 0xd2000053 +#define MASK_FCVT_D_W 0xfff0007f +#define MATCH_FCVT_D_WU 0xd2100053 +#define MASK_FCVT_D_WU 0xfff0007f +#define MATCH_FCVT_H_D 0x92000053 +#define MASK_FCVT_H_D 0xfff0007f +#define MATCH_FCVT_H_L 0x64000053 +#define MASK_FCVT_H_L 0xfff0007f +#define MATCH_FCVT_H_LU 0x6c000053 +#define MASK_FCVT_H_LU 0xfff0007f +#define MATCH_FCVT_H_S 0x90000053 +#define MASK_FCVT_H_S 0xfff0007f +#define MATCH_FCVT_H_W 0x74000053 +#define MASK_FCVT_H_W 0xfff0007f +#define MATCH_FCVT_H_WU 0x7c000053 +#define MASK_FCVT_H_WU 0xfff0007f +#define MATCH_FCVT_L_D 0xc2200053 +#define MASK_FCVT_L_D 0xfff0007f +#define MATCH_FCVT_L_H 0x44000053 +#define MASK_FCVT_L_H 0xfff0007f +#define MATCH_FCVT_L_S 0xc0200053 +#define MASK_FCVT_L_S 0xfff0007f +#define MATCH_FCVT_LU_D 0xc2300053 +#define MASK_FCVT_LU_D 0xfff0007f +#define MATCH_FCVT_LU_H 0x4c000053 +#define MASK_FCVT_LU_H 0xfff0007f +#define MATCH_FCVT_LU_S 0xc0300053 +#define MASK_FCVT_LU_S 0xfff0007f +#define MATCH_FCVT_S_D 0x40100053 +#define MASK_FCVT_S_D 0xfff0007f +#define MATCH_FCVT_S_H 0x84000053 +#define MASK_FCVT_S_H 0xfff0007f +#define MATCH_FCVT_S_L 0xd0200053 +#define MASK_FCVT_S_L 0xfff0007f +#define MATCH_FCVT_S_LU 0xd0300053 +#define MASK_FCVT_S_LU 0xfff0007f +#define MATCH_FCVT_S_W 0xd0000053 +#define MASK_FCVT_S_W 0xfff0007f +#define MATCH_FCVT_S_WU 0xd0100053 +#define MASK_FCVT_S_WU 0xfff0007f +#define MATCH_FCVT_W_D 0xc2000053 +#define MASK_FCVT_W_D 0xfff0007f +#define MATCH_FCVT_W_H 0x54000053 +#define MASK_FCVT_W_H 0xfff0007f +#define MATCH_FCVT_W_S 0xc0000053 +#define MASK_FCVT_W_S 0xfff0007f +#define MATCH_FCVT_WU_D 0xc2100053 +#define MASK_FCVT_WU_D 0xfff0007f +#define MATCH_FCVT_WU_H 0x5c000053 +#define MASK_FCVT_WU_H 0xfff0007f +#define MATCH_FCVT_WU_S 0xc0100053 +#define MASK_FCVT_WU_S 0xfff0007f +#define MATCH_FDIV_D 0x1a000053 +#define MASK_FDIV_D 0xfe00007f +#define MATCH_FDIV_H 0x1c000053 +#define MASK_FDIV_H 0xfe00007f +#define MATCH_FDIV_S 0x18000053 +#define MASK_FDIV_S 0xfe00007f +#define MATCH_FENCE 0xf +#define MASK_FENCE 0x707f +#define MATCH_FENCE_I 0x100f +#define MASK_FENCE_I 0x707f +#define MATCH_FEQ_D 0xa2002053 +#define MASK_FEQ_D 0xfe00707f +#define MATCH_FEQ_H 0xac000053 +#define MASK_FEQ_H 0xfe00707f +#define MATCH_FEQ_S 0xa0002053 +#define MASK_FEQ_S 0xfe00707f +#define MATCH_FLD 0x3007 +#define MASK_FLD 0x707f +#define MATCH_FLE_D 0xa2000053 +#define MASK_FLE_D 0xfe00707f +#define MATCH_FLE_H 0xbc000053 +#define MASK_FLE_H 0xfe00707f +#define MATCH_FLE_S 0xa0000053 +#define MASK_FLE_S 0xfe00707f +#define MATCH_FLH 0x1007 +#define MASK_FLH 0x707f +#define MATCH_FLT_D 0xa2001053 +#define MASK_FLT_D 0xfe00707f +#define MATCH_FLT_H 0xb4000053 +#define MASK_FLT_H 0xfe00707f +#define MATCH_FLT_S 0xa0001053 +#define MASK_FLT_S 0xfe00707f +#define MATCH_FLW 0x2007 +#define MASK_FLW 0x707f +#define MATCH_FMADD_D 0x2000043 +#define MASK_FMADD_D 0x600007f +#define MATCH_FMADD_H 0x4000043 +#define MASK_FMADD_H 0x600007f +#define MATCH_FMADD_S 0x43 +#define MASK_FMADD_S 0x600007f +#define MATCH_FMAX_D 0x2a001053 +#define MASK_FMAX_D 0xfe00707f +#define MATCH_FMAX_H 0xcc000053 +#define MASK_FMAX_H 0xfe00707f +#define MATCH_FMAX_S 0x28001053 +#define MASK_FMAX_S 0xfe00707f +#define MATCH_FMIN_D 0x2a000053 +#define MASK_FMIN_D 0xfe00707f +#define MATCH_FMIN_H 0xc4000053 +#define MASK_FMIN_H 0xfe00707f +#define MATCH_FMIN_S 0x28000053 +#define MASK_FMIN_S 0xfe00707f +#define MATCH_FMOVN 0x6007077 +#define MASK_FMOVN 0xfe00707f +#define MATCH_FMOVZ 0x4007077 +#define MASK_FMOVZ 0xfe00707f +#define MATCH_FMSUB_D 0x2000047 +#define MASK_FMSUB_D 0x600007f +#define MATCH_FMSUB_H 0x4000047 +#define MASK_FMSUB_H 0x600007f +#define MATCH_FMSUB_S 0x47 +#define MASK_FMSUB_S 0x600007f +#define MATCH_FMUL_D 0x12000053 +#define MASK_FMUL_D 0xfe00007f +#define MATCH_FMUL_H 0x14000053 +#define MASK_FMUL_H 0xfe00007f +#define MATCH_FMUL_S 0x10000053 +#define MASK_FMUL_S 0xfe00007f +#define MATCH_FMV_D_X 0xf2000053 +#define MASK_FMV_D_X 0xfff0707f +#define MATCH_FMV_H_X 0xf4000053 +#define MASK_FMV_H_X 0xfff0707f +#define MATCH_FMV_S_X 0xf0000053 +#define MASK_FMV_S_X 0xfff0707f +#define MATCH_FMV_X_D 0xe2000053 +#define MASK_FMV_X_D 0xfff0707f +#define MATCH_FMV_X_H 0xe4000053 +#define MASK_FMV_X_H 0xfff0707f +#define MATCH_FMV_X_S 0xe0000053 +#define MASK_FMV_X_S 0xfff0707f +#define MATCH_FNMADD_D 0x200004f +#define MASK_FNMADD_D 0x600007f +#define MATCH_FNMADD_H 0x400004f +#define MASK_FNMADD_H 0x600007f +#define MATCH_FNMADD_S 0x4f +#define MASK_FNMADD_S 0x600007f +#define MATCH_FNMSUB_D 0x200004b +#define MASK_FNMSUB_D 0x600007f +#define MATCH_FNMSUB_H 0x400004b +#define MASK_FNMSUB_H 0x600007f +#define MATCH_FNMSUB_S 0x4b +#define MASK_FNMSUB_S 0x600007f +#define MATCH_FRCSR 0x302073 +#define MASK_FRCSR 0xfffff07f +#define MATCH_FRFLAGS 0x102073 +#define MASK_FRFLAGS 0xfffff07f +#define MATCH_FRRM 0x202073 +#define MASK_FRRM 0xfffff07f +#define MATCH_FSCSR 0x301073 +#define MASK_FSCSR 0xfff0707f +#define MATCH_FSD 0x3027 +#define MASK_FSD 0x707f +#define MATCH_FSFLAGS 0x101073 +#define MASK_FSFLAGS 0xfff0707f +#define MATCH_FSFLAGSI 0x105073 +#define MASK_FSFLAGSI 0xfff0707f +#define MATCH_FSGNJ_D 0x22000053 +#define MASK_FSGNJ_D 0xfe00707f +#define MATCH_FSGNJ_H 0x2c000053 +#define MASK_FSGNJ_H 0xfe00707f +#define MATCH_FSGNJ_S 0x20000053 +#define MASK_FSGNJ_S 0xfe00707f +#define MATCH_FSGNJN_D 0x22001053 +#define MASK_FSGNJN_D 0xfe00707f +#define MATCH_FSGNJN_H 0x34000053 +#define MASK_FSGNJN_H 0xfe00707f +#define MATCH_FSGNJN_S 0x20001053 +#define MASK_FSGNJN_S 0xfe00707f +#define MATCH_FSGNJX_D 0x22002053 +#define MASK_FSGNJX_D 0xfe00707f +#define MATCH_FSGNJX_H 0x3c000053 +#define MASK_FSGNJX_H 0xfe00707f +#define MATCH_FSGNJX_S 0x20002053 +#define MASK_FSGNJX_S 0xfe00707f +#define MATCH_FSH 0x1027 +#define MASK_FSH 0x707f +#define MATCH_FSQRT_D 0x5a000053 +#define MASK_FSQRT_D 0xfff0007f +#define MATCH_FSQRT_H 0x24000053 +#define MASK_FSQRT_H 0xfff0007f +#define MATCH_FSQRT_S 0x58000053 +#define MASK_FSQRT_S 0xfff0007f +#define MATCH_FSRM 0x201073 +#define MASK_FSRM 0xfff0707f +#define MATCH_FSRMI 0x205073 +#define MASK_FSRMI 0xfff0707f +#define MATCH_FSUB_D 0xa000053 +#define MASK_FSUB_D 0xfe00007f +#define MATCH_FSUB_H 0xc000053 +#define MASK_FSUB_H 0xfe00007f +#define MATCH_FSUB_S 0x8000053 +#define MASK_FSUB_S 0xfe00007f +#define MATCH_FSW 0x2027 +#define MASK_FSW 0x707f +#define MATCH_HRTS 0x20500073 +#define MASK_HRTS 0xffffffff +#define MATCH_JAL 0x6f +#define MASK_JAL 0x7f +#define MATCH_JALR 0x67 +#define MASK_JALR 0x707f +#define MATCH_LB 0x3 +#define MASK_LB 0x707f +#define MATCH_LBU 0x4003 +#define MASK_LBU 0x707f +#define MATCH_LD 0x3003 +#define MASK_LD 0x707f +#define MATCH_LH 0x1003 +#define MASK_LH 0x707f +#define MATCH_LHU 0x5003 +#define MASK_LHU 0x707f +#define MATCH_LR_D 0x1000302f +#define MASK_LR_D 0xf9f0707f +#define MATCH_LR_W 0x1000202f +#define MASK_LR_W 0xf9f0707f +#define MATCH_LUI 0x37 +#define MASK_LUI 0x7f +#define MATCH_LW 0x2003 +#define MASK_LW 0x707f +#define MATCH_LWU 0x6003 +#define MASK_LWU 0x707f +#define MATCH_MOVN 0x2007077 +#define MASK_MOVN 0xfe00707f +#define MATCH_MOVZ 0x7077 +#define MASK_MOVZ 0xfe00707f +#define MATCH_MRTH 0x30600073 +#define MASK_MRTH 0xffffffff +#define MATCH_MRTS 0x30500073 +#define MASK_MRTS 0xffffffff +#define MATCH_MUL 0x2000033 +#define MASK_MUL 0xfe00707f +#define MATCH_MULH 0x2001033 +#define MASK_MULH 0xfe00707f +#define MATCH_MULHSU 0x2002033 +#define MASK_MULHSU 0xfe00707f +#define MATCH_MULHU 0x2003033 +#define MASK_MULHU 0xfe00707f +#define MATCH_MULW 0x200003b +#define MASK_MULW 0xfe00707f +#define MATCH_OR 0x6033 +#define MASK_OR 0xfe00707f +#define MATCH_ORI 0x6013 +#define MASK_ORI 0x707f +#define MATCH_RDCYCLE 0xc0002073 +#define MASK_RDCYCLE 0xfffff07f +#define MATCH_RDCYCLEH 0xc8002073 +#define MASK_RDCYCLEH 0xfffff07f +#define MATCH_RDINSTRET 0xc0202073 +#define MASK_RDINSTRET 0xfffff07f +#define MATCH_RDINSTRETH 0xc8202073 +#define MASK_RDINSTRETH 0xfffff07f +#define MATCH_RDTIME 0xc0102073 +#define MASK_RDTIME 0xfffff07f +#define MATCH_RDTIMEH 0xc8102073 +#define MASK_RDTIMEH 0xfffff07f +#define MATCH_REM 0x2006033 +#define MASK_REM 0xfe00707f +#define MATCH_REMU 0x2007033 +#define MASK_REMU 0xfe00707f +#define MATCH_REMUW 0x200703b +#define MASK_REMUW 0xfe00707f +#define MATCH_REMW 0x200603b +#define MASK_REMW 0xfe00707f +#define MATCH_SB 0x23 +#define MASK_SB 0x707f +#define MATCH_SBREAK 0x100073 +#define MASK_SBREAK 0xffffffff +#define MATCH_SC_D 0x1800302f +#define MASK_SC_D 0xf800707f +#define MATCH_SC_W 0x1800202f +#define MASK_SC_W 0xf800707f +#define MATCH_SCALL 0x73 +#define MASK_SCALL 0xffffffff +#define MATCH_SD 0x3023 +#define MASK_SD 0x707f +#define MATCH_SFENCE_VM 0x10100073 +#define MASK_SFENCE_VM 0xfff07fff +#define MATCH_SH 0x1023 +#define MASK_SH 0x707f +#define MATCH_SLL 0x1033 +#define MASK_SLL 0xfe00707f +#define MATCH_SLLI 0x1013 +#define MASK_SLLI 0xfc00707f +#define MATCH_SLLI_RV32 0x1013 +#define MASK_SLLI_RV32 0xfe00707f +#define MATCH_SLLIW 0x101b +#define MASK_SLLIW 0xfe00707f +#define MATCH_SLLW 0x103b +#define MASK_SLLW 0xfe00707f +#define MATCH_SLT 0x2033 +#define MASK_SLT 0xfe00707f +#define MATCH_SLTI 0x2013 +#define MASK_SLTI 0x707f +#define MATCH_SLTIU 0x3013 +#define MASK_SLTIU 0x707f +#define MATCH_SLTU 0x3033 +#define MASK_SLTU 0xfe00707f +#define MATCH_SRA 0x40005033 +#define MASK_SRA 0xfe00707f +#define MATCH_SRAI 0x40005013 +#define MASK_SRAI 0xfc00707f +#define MATCH_SRAI_RV32 0x40005013 +#define MASK_SRAI_RV32 0xfe00707f +#define MATCH_SRAIW 0x4000501b +#define MASK_SRAIW 0xfe00707f +#define MATCH_SRAW 0x4000503b +#define MASK_SRAW 0xfe00707f +#define MATCH_SRET 0x10000073 +#define MASK_SRET 0xffffffff +#define MATCH_SRL 0x5033 +#define MASK_SRL 0xfe00707f +#define MATCH_SRLI 0x5013 +#define MASK_SRLI 0xfc00707f +#define MATCH_SRLI_RV32 0x5013 +#define MASK_SRLI_RV32 0xfe00707f +#define MATCH_SRLIW 0x501b +#define MASK_SRLIW 0xfe00707f +#define MATCH_SRLW 0x503b +#define MASK_SRLW 0xfe00707f +#define MATCH_STOP 0x5077 +#define MASK_STOP 0xffffffff +#define MATCH_SUB 0x40000033 +#define MASK_SUB 0xfe00707f +#define MATCH_SUBW 0x4000003b +#define MASK_SUBW 0xfe00707f +#define MATCH_SW 0x2023 +#define MASK_SW 0x707f +#define MATCH_UTIDX 0x6077 +#define MASK_UTIDX 0xfffff07f +#define MATCH_VENQCMD 0xa00302b +#define MASK_VENQCMD 0xfe007fff +#define MATCH_VENQCNT 0x1000302b +#define MASK_VENQCNT 0xfe007fff +#define MATCH_VENQIMM1 0xc00302b +#define MASK_VENQIMM1 0xfe007fff +#define MATCH_VENQIMM2 0xe00302b +#define MASK_VENQIMM2 0xfe007fff +#define MATCH_VF 0x10202b +#define MASK_VF 0x1f0707f +#define MATCH_VFLD 0x1600205b +#define MASK_VFLD 0xfff0707f +#define MATCH_VFLSEGD 0x1600205b +#define MASK_VFLSEGD 0x1ff0707f +#define MATCH_VFLSEGSTD 0x1600305b +#define MASK_VFLSEGSTD 0x1e00707f +#define MATCH_VFLSEGSTW 0x1400305b +#define MASK_VFLSEGSTW 0x1e00707f +#define MATCH_VFLSEGW 0x1400205b +#define MASK_VFLSEGW 0x1ff0707f +#define MATCH_VFLSTD 0x1600305b +#define MASK_VFLSTD 0xfe00707f +#define MATCH_VFLSTW 0x1400305b +#define MASK_VFLSTW 0xfe00707f +#define MATCH_VFLW 0x1400205b +#define MASK_VFLW 0xfff0707f +#define MATCH_VFMSV_D 0x1200202b +#define MASK_VFMSV_D 0xfff0707f +#define MATCH_VFMSV_S 0x1000202b +#define MASK_VFMSV_S 0xfff0707f +#define MATCH_VFMVV 0x1000002b +#define MASK_VFMVV 0xfff0707f +#define MATCH_VFSD 0x1600207b +#define MASK_VFSD 0xfff0707f +#define MATCH_VFSSEGD 0x1600207b +#define MASK_VFSSEGD 0x1ff0707f +#define MATCH_VFSSEGSTD 0x1600307b +#define MASK_VFSSEGSTD 0x1e00707f +#define MATCH_VFSSEGSTW 0x1400307b +#define MASK_VFSSEGSTW 0x1e00707f +#define MATCH_VFSSEGW 0x1400207b +#define MASK_VFSSEGW 0x1ff0707f +#define MATCH_VFSSTD 0x1600307b +#define MASK_VFSSTD 0xfe00707f +#define MATCH_VFSSTW 0x1400307b +#define MASK_VFSSTW 0xfe00707f +#define MATCH_VFSW 0x1400207b +#define MASK_VFSW 0xfff0707f +#define MATCH_VGETCFG 0x400b +#define MASK_VGETCFG 0xfffff07f +#define MATCH_VGETVL 0x200400b +#define MASK_VGETVL 0xfffff07f +#define MATCH_VLB 0x205b +#define MASK_VLB 0xfff0707f +#define MATCH_VLBU 0x800205b +#define MASK_VLBU 0xfff0707f +#define MATCH_VLD 0x600205b +#define MASK_VLD 0xfff0707f +#define MATCH_VLH 0x200205b +#define MASK_VLH 0xfff0707f +#define MATCH_VLHU 0xa00205b +#define MASK_VLHU 0xfff0707f +#define MATCH_VLSEGB 0x205b +#define MASK_VLSEGB 0x1ff0707f +#define MATCH_VLSEGBU 0x800205b +#define MASK_VLSEGBU 0x1ff0707f +#define MATCH_VLSEGD 0x600205b +#define MASK_VLSEGD 0x1ff0707f +#define MATCH_VLSEGH 0x200205b +#define MASK_VLSEGH 0x1ff0707f +#define MATCH_VLSEGHU 0xa00205b +#define MASK_VLSEGHU 0x1ff0707f +#define MATCH_VLSEGSTB 0x305b +#define MASK_VLSEGSTB 0x1e00707f +#define MATCH_VLSEGSTBU 0x800305b +#define MASK_VLSEGSTBU 0x1e00707f +#define MATCH_VLSEGSTD 0x600305b +#define MASK_VLSEGSTD 0x1e00707f +#define MATCH_VLSEGSTH 0x200305b +#define MASK_VLSEGSTH 0x1e00707f +#define MATCH_VLSEGSTHU 0xa00305b +#define MASK_VLSEGSTHU 0x1e00707f +#define MATCH_VLSEGSTW 0x400305b +#define MASK_VLSEGSTW 0x1e00707f +#define MATCH_VLSEGSTWU 0xc00305b +#define MASK_VLSEGSTWU 0x1e00707f +#define MATCH_VLSEGW 0x400205b +#define MASK_VLSEGW 0x1ff0707f +#define MATCH_VLSEGWU 0xc00205b +#define MASK_VLSEGWU 0x1ff0707f +#define MATCH_VLSTB 0x305b +#define MASK_VLSTB 0xfe00707f +#define MATCH_VLSTBU 0x800305b +#define MASK_VLSTBU 0xfe00707f +#define MATCH_VLSTD 0x600305b +#define MASK_VLSTD 0xfe00707f +#define MATCH_VLSTH 0x200305b +#define MASK_VLSTH 0xfe00707f +#define MATCH_VLSTHU 0xa00305b +#define MASK_VLSTHU 0xfe00707f +#define MATCH_VLSTW 0x400305b +#define MASK_VLSTW 0xfe00707f +#define MATCH_VLSTWU 0xc00305b +#define MASK_VLSTWU 0xfe00707f +#define MATCH_VLW 0x400205b +#define MASK_VLW 0xfff0707f +#define MATCH_VLWU 0xc00205b +#define MASK_VLWU 0xfff0707f +#define MATCH_VMSV 0x200202b +#define MASK_VMSV 0xfff0707f +#define MATCH_VMVV 0x200002b +#define MASK_VMVV 0xfff0707f +#define MATCH_VSB 0x207b +#define MASK_VSB 0xfff0707f +#define MATCH_VSD 0x600207b +#define MASK_VSD 0xfff0707f +#define MATCH_VSETCFG 0x200b +#define MASK_VSETCFG 0x7fff +#define MATCH_VSETVL 0x600b +#define MASK_VSETVL 0xfff0707f +#define MATCH_VSH 0x200207b +#define MASK_VSH 0xfff0707f +#define MATCH_VSSEGB 0x207b +#define MASK_VSSEGB 0x1ff0707f +#define MATCH_VSSEGD 0x600207b +#define MASK_VSSEGD 0x1ff0707f +#define MATCH_VSSEGH 0x200207b +#define MASK_VSSEGH 0x1ff0707f +#define MATCH_VSSEGSTB 0x307b +#define MASK_VSSEGSTB 0x1e00707f +#define MATCH_VSSEGSTD 0x600307b +#define MASK_VSSEGSTD 0x1e00707f +#define MATCH_VSSEGSTH 0x200307b +#define MASK_VSSEGSTH 0x1e00707f +#define MATCH_VSSEGSTW 0x400307b +#define MASK_VSSEGSTW 0x1e00707f +#define MATCH_VSSEGW 0x400207b +#define MASK_VSSEGW 0x1ff0707f +#define MATCH_VSSTB 0x307b +#define MASK_VSSTB 0xfe00707f +#define MATCH_VSSTD 0x600307b +#define MASK_VSSTD 0xfe00707f +#define MATCH_VSSTH 0x200307b +#define MASK_VSSTH 0xfe00707f +#define MATCH_VSSTW 0x400307b +#define MASK_VSSTW 0xfe00707f +#define MATCH_VSW 0x400207b +#define MASK_VSW 0xfff0707f +#define MATCH_VXCPTAUX 0x200402b +#define MASK_VXCPTAUX 0xfffff07f +#define MATCH_VXCPTCAUSE 0x402b +#define MASK_VXCPTCAUSE 0xfffff07f +#define MATCH_VXCPTEVAC 0x600302b +#define MASK_VXCPTEVAC 0xfff07fff +#define MATCH_VXCPTHOLD 0x800302b +#define MASK_VXCPTHOLD 0xfff07fff +#define MATCH_VXCPTKILL 0x400302b +#define MASK_VXCPTKILL 0xffffffff +#define MATCH_VXCPTRESTORE 0x200302b +#define MASK_VXCPTRESTORE 0xfff07fff +#define MATCH_VXCPTSAVE 0x302b +#define MASK_VXCPTSAVE 0xfff07fff +#define MATCH_WFI 0x10200073 +#define MASK_WFI 0xffffffff +#define MATCH_XOR 0x4033 +#define MASK_XOR 0xfe00707f +#define MATCH_XORI 0x4013 +#define MASK_XORI 0x707f +#define CSR_FFLAGS 0x1 +#define CSR_FRM 0x2 +#define CSR_FCSR 0x3 +#define CSR_CYCLE 0xc00 +#define CSR_TIME 0xc01 +#define CSR_INSTRET 0xc02 +#define CSR_STATS 0xc0 +#define CSR_UARCH0 0xcc0 +#define CSR_UARCH1 0xcc1 +#define CSR_UARCH2 0xcc2 +#define CSR_UARCH3 0xcc3 +#define CSR_UARCH4 0xcc4 +#define CSR_UARCH5 0xcc5 +#define CSR_UARCH6 0xcc6 +#define CSR_UARCH7 0xcc7 +#define CSR_UARCH8 0xcc8 +#define CSR_UARCH9 0xcc9 +#define CSR_UARCH10 0xcca +#define CSR_UARCH11 0xccb +#define CSR_UARCH12 0xccc +#define CSR_UARCH13 0xccd +#define CSR_UARCH14 0xcce +#define CSR_UARCH15 0xccf +#define CSR_SSTATUS 0x100 +#define CSR_STVEC 0x101 +#define CSR_SIE 0x104 +#define CSR_SSCRATCH 0x140 +#define CSR_SEPC 0x141 +#define CSR_SIP 0x144 +#define CSR_SPTBR 0x180 +#define CSR_SASID 0x181 +#define CSR_CYCLEW 0x900 +#define CSR_TIMEW 0x901 +#define CSR_INSTRETW 0x902 +#define CSR_STIME 0xd01 +#define CSR_SCAUSE 0xd42 +#define CSR_SBADADDR 0xd43 +#define CSR_STIMEW 0xa01 +#define CSR_MSTATUS 0x300 +#define CSR_MTVEC 0x301 +#define CSR_MTDELEG 0x302 +#define CSR_MIE 0x304 +#define CSR_MTIMECMP 0x321 +#define CSR_MSCRATCH 0x340 +#define CSR_MEPC 0x341 +#define CSR_MCAUSE 0x342 +#define CSR_MBADADDR 0x343 +#define CSR_MIP 0x344 +#define CSR_MTIME 0x701 +#define CSR_MCPUID 0xf00 +#define CSR_MIMPID 0xf01 +#define CSR_MHARTID 0xf10 +#define CSR_MTOHOST 0x780 +#define CSR_MFROMHOST 0x781 +#define CSR_MRESET 0x782 +#define CSR_SEND_IPI 0x783 +#define CSR_CYCLEH 0xc80 +#define CSR_TIMEH 0xc81 +#define CSR_INSTRETH 0xc82 +#define CSR_CYCLEHW 0x980 +#define CSR_TIMEHW 0x981 +#define CSR_INSTRETHW 0x982 +#define CSR_STIMEH 0xd81 +#define CSR_STIMEHW 0xa81 +#define CSR_MTIMECMPH 0x361 +#define CSR_MTIMEH 0x741 +#define CAUSE_MISALIGNED_FETCH 0x0 +#define CAUSE_FAULT_FETCH 0x1 +#define CAUSE_ILLEGAL_INSTRUCTION 0x2 +#define CAUSE_BREAKPOINT 0x3 +#define CAUSE_MISALIGNED_LOAD 0x4 +#define CAUSE_FAULT_LOAD 0x5 +#define CAUSE_MISALIGNED_STORE 0x6 +#define CAUSE_FAULT_STORE 0x7 +#define CAUSE_USER_ECALL 0x8 +#define CAUSE_SUPERVISOR_ECALL 0x9 +#define CAUSE_HYPERVISOR_ECALL 0xa +#define CAUSE_MACHINE_ECALL 0xb +#endif +#ifdef DECLARE_INSN +DECLARE_INSN(add, MATCH_ADD, MASK_ADD) +DECLARE_INSN(addi, MATCH_ADDI, MASK_ADDI) +DECLARE_INSN(addiw, MATCH_ADDIW, MASK_ADDIW) +DECLARE_INSN(addw, MATCH_ADDW, MASK_ADDW) +DECLARE_INSN(amoadd_d, MATCH_AMOADD_D, MASK_AMOADD_D) +DECLARE_INSN(amoadd_w, MATCH_AMOADD_W, MASK_AMOADD_W) +DECLARE_INSN(amoand_d, MATCH_AMOAND_D, MASK_AMOAND_D) +DECLARE_INSN(amoand_w, MATCH_AMOAND_W, MASK_AMOAND_W) +DECLARE_INSN(amomax_d, MATCH_AMOMAX_D, MASK_AMOMAX_D) +DECLARE_INSN(amomax_w, MATCH_AMOMAX_W, MASK_AMOMAX_W) +DECLARE_INSN(amomaxu_d, MATCH_AMOMAXU_D, MASK_AMOMAXU_D) +DECLARE_INSN(amomaxu_w, MATCH_AMOMAXU_W, MASK_AMOMAXU_W) +DECLARE_INSN(amomin_d, MATCH_AMOMIN_D, MASK_AMOMIN_D) +DECLARE_INSN(amomin_w, MATCH_AMOMIN_W, MASK_AMOMIN_W) +DECLARE_INSN(amominu_d, MATCH_AMOMINU_D, MASK_AMOMINU_D) +DECLARE_INSN(amominu_w, MATCH_AMOMINU_W, MASK_AMOMINU_W) +DECLARE_INSN(amoor_d, MATCH_AMOOR_D, MASK_AMOOR_D) +DECLARE_INSN(amoor_w, MATCH_AMOOR_W, MASK_AMOOR_W) +DECLARE_INSN(amoswap_d, MATCH_AMOSWAP_D, MASK_AMOSWAP_D) +DECLARE_INSN(amoswap_w, MATCH_AMOSWAP_W, MASK_AMOSWAP_W) +DECLARE_INSN(amoxor_d, MATCH_AMOXOR_D, MASK_AMOXOR_D) +DECLARE_INSN(amoxor_w, MATCH_AMOXOR_W, MASK_AMOXOR_W) +DECLARE_INSN(and, MATCH_AND, MASK_AND) +DECLARE_INSN(andi, MATCH_ANDI, MASK_ANDI) +DECLARE_INSN(auipc, MATCH_AUIPC, MASK_AUIPC) +DECLARE_INSN(beq, MATCH_BEQ, MASK_BEQ) +DECLARE_INSN(bge, MATCH_BGE, MASK_BGE) +DECLARE_INSN(bgeu, MATCH_BGEU, MASK_BGEU) +DECLARE_INSN(blt, MATCH_BLT, MASK_BLT) +DECLARE_INSN(bltu, MATCH_BLTU, MASK_BLTU) +DECLARE_INSN(bne, MATCH_BNE, MASK_BNE) +DECLARE_INSN(c_add, MATCH_C_ADD, MASK_C_ADD) +DECLARE_INSN(c_add3, MATCH_C_ADD3, MASK_C_ADD3) +DECLARE_INSN(c_addi, MATCH_C_ADDI, MASK_C_ADDI) +DECLARE_INSN(c_addi16sp, MATCH_C_ADDI16SP, MASK_C_ADDI16SP) +DECLARE_INSN(c_addi4spn, MATCH_C_ADDI4SPN, MASK_C_ADDI4SPN) +DECLARE_INSN(c_addin, MATCH_C_ADDIN, MASK_C_ADDIN) +DECLARE_INSN(c_addiw, MATCH_C_ADDIW, MASK_C_ADDIW) +DECLARE_INSN(c_addw, MATCH_C_ADDW, MASK_C_ADDW) +DECLARE_INSN(c_and3, MATCH_C_AND3, MASK_C_AND3) +DECLARE_INSN(c_andi, MATCH_C_ANDI, MASK_C_ANDI) +DECLARE_INSN(c_andin, MATCH_C_ANDIN, MASK_C_ANDIN) +DECLARE_INSN(c_beqz, MATCH_C_BEQZ, MASK_C_BEQZ) +DECLARE_INSN(c_bgez, MATCH_C_BGEZ, MASK_C_BGEZ) +DECLARE_INSN(c_bltz, MATCH_C_BLTZ, MASK_C_BLTZ) +DECLARE_INSN(c_bnez, MATCH_C_BNEZ, MASK_C_BNEZ) +DECLARE_INSN(c_ebreak, MATCH_C_EBREAK, MASK_C_EBREAK) +DECLARE_INSN(c_j, MATCH_C_J, MASK_C_J) +DECLARE_INSN(c_jal, MATCH_C_JAL, MASK_C_JAL) +DECLARE_INSN(c_jalr, MATCH_C_JALR, MASK_C_JALR) +DECLARE_INSN(c_jr, MATCH_C_JR, MASK_C_JR) +DECLARE_INSN(c_ld, MATCH_C_LD, MASK_C_LD) +DECLARE_INSN(c_ldsp, MATCH_C_LDSP, MASK_C_LDSP) +DECLARE_INSN(c_li, MATCH_C_LI, MASK_C_LI) +DECLARE_INSN(c_lui, MATCH_C_LUI, MASK_C_LUI) +DECLARE_INSN(c_lw, MATCH_C_LW, MASK_C_LW) +DECLARE_INSN(c_lwsp, MATCH_C_LWSP, MASK_C_LWSP) +DECLARE_INSN(c_mv, MATCH_C_MV, MASK_C_MV) +DECLARE_INSN(c_or3, MATCH_C_OR3, MASK_C_OR3) +DECLARE_INSN(c_orin, MATCH_C_ORIN, MASK_C_ORIN) +DECLARE_INSN(c_sd, MATCH_C_SD, MASK_C_SD) +DECLARE_INSN(c_sdsp, MATCH_C_SDSP, MASK_C_SDSP) +DECLARE_INSN(c_sll, MATCH_C_SLL, MASK_C_SLL) +DECLARE_INSN(c_slli, MATCH_C_SLLI, MASK_C_SLLI) +DECLARE_INSN(c_slliw, MATCH_C_SLLIW, MASK_C_SLLIW) +DECLARE_INSN(c_sllr, MATCH_C_SLLR, MASK_C_SLLR) +DECLARE_INSN(c_slt, MATCH_C_SLT, MASK_C_SLT) +DECLARE_INSN(c_sltr, MATCH_C_SLTR, MASK_C_SLTR) +DECLARE_INSN(c_sltu, MATCH_C_SLTU, MASK_C_SLTU) +DECLARE_INSN(c_sltur, MATCH_C_SLTUR, MASK_C_SLTUR) +DECLARE_INSN(c_sra, MATCH_C_SRA, MASK_C_SRA) +DECLARE_INSN(c_srai, MATCH_C_SRAI, MASK_C_SRAI) +DECLARE_INSN(c_srl, MATCH_C_SRL, MASK_C_SRL) +DECLARE_INSN(c_srli, MATCH_C_SRLI, MASK_C_SRLI) +DECLARE_INSN(c_srlr, MATCH_C_SRLR, MASK_C_SRLR) +DECLARE_INSN(c_sub, MATCH_C_SUB, MASK_C_SUB) +DECLARE_INSN(c_sub3, MATCH_C_SUB3, MASK_C_SUB3) +DECLARE_INSN(c_sw, MATCH_C_SW, MASK_C_SW) +DECLARE_INSN(c_swsp, MATCH_C_SWSP, MASK_C_SWSP) +DECLARE_INSN(c_xor, MATCH_C_XOR, MASK_C_XOR) +DECLARE_INSN(c_xorin, MATCH_C_XORIN, MASK_C_XORIN) +DECLARE_INSN(csrrc, MATCH_CSRRC, MASK_CSRRC) +DECLARE_INSN(csrrci, MATCH_CSRRCI, MASK_CSRRCI) +DECLARE_INSN(csrrs, MATCH_CSRRS, MASK_CSRRS) +DECLARE_INSN(csrrsi, MATCH_CSRRSI, MASK_CSRRSI) +DECLARE_INSN(csrrw, MATCH_CSRRW, MASK_CSRRW) +DECLARE_INSN(csrrwi, MATCH_CSRRWI, MASK_CSRRWI) +DECLARE_INSN(custom0, MATCH_CUSTOM0, MASK_CUSTOM0) +DECLARE_INSN(custom0_rd, MATCH_CUSTOM0_RD, MASK_CUSTOM0_RD) +DECLARE_INSN(custom0_rd_rs1, MATCH_CUSTOM0_RD_RS1, MASK_CUSTOM0_RD_RS1) +DECLARE_INSN(custom0_rd_rs1_rs2, MATCH_CUSTOM0_RD_RS1_RS2, MASK_CUSTOM0_RD_RS1_RS2) +DECLARE_INSN(custom0_rs1, MATCH_CUSTOM0_RS1, MASK_CUSTOM0_RS1) +DECLARE_INSN(custom0_rs1_rs2, MATCH_CUSTOM0_RS1_RS2, MASK_CUSTOM0_RS1_RS2) +DECLARE_INSN(custom1, MATCH_CUSTOM1, MASK_CUSTOM1) +DECLARE_INSN(custom1_rd, MATCH_CUSTOM1_RD, MASK_CUSTOM1_RD) +DECLARE_INSN(custom1_rd_rs1, MATCH_CUSTOM1_RD_RS1, MASK_CUSTOM1_RD_RS1) +DECLARE_INSN(custom1_rd_rs1_rs2, MATCH_CUSTOM1_RD_RS1_RS2, MASK_CUSTOM1_RD_RS1_RS2) +DECLARE_INSN(custom1_rs1, MATCH_CUSTOM1_RS1, MASK_CUSTOM1_RS1) +DECLARE_INSN(custom1_rs1_rs2, MATCH_CUSTOM1_RS1_RS2, MASK_CUSTOM1_RS1_RS2) +DECLARE_INSN(custom2, MATCH_CUSTOM2, MASK_CUSTOM2) +DECLARE_INSN(custom2_rd, MATCH_CUSTOM2_RD, MASK_CUSTOM2_RD) +DECLARE_INSN(custom2_rd_rs1, MATCH_CUSTOM2_RD_RS1, MASK_CUSTOM2_RD_RS1) +DECLARE_INSN(custom2_rd_rs1_rs2, MATCH_CUSTOM2_RD_RS1_RS2, MASK_CUSTOM2_RD_RS1_RS2) +DECLARE_INSN(custom2_rs1, MATCH_CUSTOM2_RS1, MASK_CUSTOM2_RS1) +DECLARE_INSN(custom2_rs1_rs2, MATCH_CUSTOM2_RS1_RS2, MASK_CUSTOM2_RS1_RS2) +DECLARE_INSN(custom3, MATCH_CUSTOM3, MASK_CUSTOM3) +DECLARE_INSN(custom3_rd, MATCH_CUSTOM3_RD, MASK_CUSTOM3_RD) +DECLARE_INSN(custom3_rd_rs1, MATCH_CUSTOM3_RD_RS1, MASK_CUSTOM3_RD_RS1) +DECLARE_INSN(custom3_rd_rs1_rs2, MATCH_CUSTOM3_RD_RS1_RS2, MASK_CUSTOM3_RD_RS1_RS2) +DECLARE_INSN(custom3_rs1, MATCH_CUSTOM3_RS1, MASK_CUSTOM3_RS1) +DECLARE_INSN(custom3_rs1_rs2, MATCH_CUSTOM3_RS1_RS2, MASK_CUSTOM3_RS1_RS2) +DECLARE_INSN(div, MATCH_DIV, MASK_DIV) +DECLARE_INSN(divu, MATCH_DIVU, MASK_DIVU) +DECLARE_INSN(divuw, MATCH_DIVUW, MASK_DIVUW) +DECLARE_INSN(divw, MATCH_DIVW, MASK_DIVW) +DECLARE_INSN(ebreak, MATCH_EBREAK, MASK_EBREAK) +DECLARE_INSN(ecall, MATCH_ECALL, MASK_ECALL) +DECLARE_INSN(eret, MATCH_ERET, MASK_ERET) +DECLARE_INSN(fadd_d, MATCH_FADD_D, MASK_FADD_D) +DECLARE_INSN(fadd_h, MATCH_FADD_H, MASK_FADD_H) +DECLARE_INSN(fadd_s, MATCH_FADD_S, MASK_FADD_S) +DECLARE_INSN(fclass_d, MATCH_FCLASS_D, MASK_FCLASS_D) +DECLARE_INSN(fclass_s, MATCH_FCLASS_S, MASK_FCLASS_S) +DECLARE_INSN(fcvt_d_h, MATCH_FCVT_D_H, MASK_FCVT_D_H) +DECLARE_INSN(fcvt_d_l, MATCH_FCVT_D_L, MASK_FCVT_D_L) +DECLARE_INSN(fcvt_d_lu, MATCH_FCVT_D_LU, MASK_FCVT_D_LU) +DECLARE_INSN(fcvt_d_s, MATCH_FCVT_D_S, MASK_FCVT_D_S) +DECLARE_INSN(fcvt_d_w, MATCH_FCVT_D_W, MASK_FCVT_D_W) +DECLARE_INSN(fcvt_d_wu, MATCH_FCVT_D_WU, MASK_FCVT_D_WU) +DECLARE_INSN(fcvt_h_d, MATCH_FCVT_H_D, MASK_FCVT_H_D) +DECLARE_INSN(fcvt_h_l, MATCH_FCVT_H_L, MASK_FCVT_H_L) +DECLARE_INSN(fcvt_h_lu, MATCH_FCVT_H_LU, MASK_FCVT_H_LU) +DECLARE_INSN(fcvt_h_s, MATCH_FCVT_H_S, MASK_FCVT_H_S) +DECLARE_INSN(fcvt_h_w, MATCH_FCVT_H_W, MASK_FCVT_H_W) +DECLARE_INSN(fcvt_h_wu, MATCH_FCVT_H_WU, MASK_FCVT_H_WU) +DECLARE_INSN(fcvt_l_d, MATCH_FCVT_L_D, MASK_FCVT_L_D) +DECLARE_INSN(fcvt_l_h, MATCH_FCVT_L_H, MASK_FCVT_L_H) +DECLARE_INSN(fcvt_l_s, MATCH_FCVT_L_S, MASK_FCVT_L_S) +DECLARE_INSN(fcvt_lu_d, MATCH_FCVT_LU_D, MASK_FCVT_LU_D) +DECLARE_INSN(fcvt_lu_h, MATCH_FCVT_LU_H, MASK_FCVT_LU_H) +DECLARE_INSN(fcvt_lu_s, MATCH_FCVT_LU_S, MASK_FCVT_LU_S) +DECLARE_INSN(fcvt_s_d, MATCH_FCVT_S_D, MASK_FCVT_S_D) +DECLARE_INSN(fcvt_s_h, MATCH_FCVT_S_H, MASK_FCVT_S_H) +DECLARE_INSN(fcvt_s_l, MATCH_FCVT_S_L, MASK_FCVT_S_L) +DECLARE_INSN(fcvt_s_lu, MATCH_FCVT_S_LU, MASK_FCVT_S_LU) +DECLARE_INSN(fcvt_s_w, MATCH_FCVT_S_W, MASK_FCVT_S_W) +DECLARE_INSN(fcvt_s_wu, MATCH_FCVT_S_WU, MASK_FCVT_S_WU) +DECLARE_INSN(fcvt_w_d, MATCH_FCVT_W_D, MASK_FCVT_W_D) +DECLARE_INSN(fcvt_w_h, MATCH_FCVT_W_H, MASK_FCVT_W_H) +DECLARE_INSN(fcvt_w_s, MATCH_FCVT_W_S, MASK_FCVT_W_S) +DECLARE_INSN(fcvt_wu_d, MATCH_FCVT_WU_D, MASK_FCVT_WU_D) +DECLARE_INSN(fcvt_wu_h, MATCH_FCVT_WU_H, MASK_FCVT_WU_H) +DECLARE_INSN(fcvt_wu_s, MATCH_FCVT_WU_S, MASK_FCVT_WU_S) +DECLARE_INSN(fdiv_d, MATCH_FDIV_D, MASK_FDIV_D) +DECLARE_INSN(fdiv_h, MATCH_FDIV_H, MASK_FDIV_H) +DECLARE_INSN(fdiv_s, MATCH_FDIV_S, MASK_FDIV_S) +DECLARE_INSN(fence, MATCH_FENCE, MASK_FENCE) +DECLARE_INSN(fence_i, MATCH_FENCE_I, MASK_FENCE_I) +DECLARE_INSN(feq_d, MATCH_FEQ_D, MASK_FEQ_D) +DECLARE_INSN(feq_h, MATCH_FEQ_H, MASK_FEQ_H) +DECLARE_INSN(feq_s, MATCH_FEQ_S, MASK_FEQ_S) +DECLARE_INSN(fld, MATCH_FLD, MASK_FLD) +DECLARE_INSN(fle_d, MATCH_FLE_D, MASK_FLE_D) +DECLARE_INSN(fle_h, MATCH_FLE_H, MASK_FLE_H) +DECLARE_INSN(fle_s, MATCH_FLE_S, MASK_FLE_S) +DECLARE_INSN(flh, MATCH_FLH, MASK_FLH) +DECLARE_INSN(flt_d, MATCH_FLT_D, MASK_FLT_D) +DECLARE_INSN(flt_h, MATCH_FLT_H, MASK_FLT_H) +DECLARE_INSN(flt_s, MATCH_FLT_S, MASK_FLT_S) +DECLARE_INSN(flw, MATCH_FLW, MASK_FLW) +DECLARE_INSN(fmadd_d, MATCH_FMADD_D, MASK_FMADD_D) +DECLARE_INSN(fmadd_h, MATCH_FMADD_H, MASK_FMADD_H) +DECLARE_INSN(fmadd_s, MATCH_FMADD_S, MASK_FMADD_S) +DECLARE_INSN(fmax_d, MATCH_FMAX_D, MASK_FMAX_D) +DECLARE_INSN(fmax_h, MATCH_FMAX_H, MASK_FMAX_H) +DECLARE_INSN(fmax_s, MATCH_FMAX_S, MASK_FMAX_S) +DECLARE_INSN(fmin_d, MATCH_FMIN_D, MASK_FMIN_D) +DECLARE_INSN(fmin_h, MATCH_FMIN_H, MASK_FMIN_H) +DECLARE_INSN(fmin_s, MATCH_FMIN_S, MASK_FMIN_S) +DECLARE_INSN(fmovn, MATCH_FMOVN, MASK_FMOVN) +DECLARE_INSN(fmovz, MATCH_FMOVZ, MASK_FMOVZ) +DECLARE_INSN(fmsub_d, MATCH_FMSUB_D, MASK_FMSUB_D) +DECLARE_INSN(fmsub_h, MATCH_FMSUB_H, MASK_FMSUB_H) +DECLARE_INSN(fmsub_s, MATCH_FMSUB_S, MASK_FMSUB_S) +DECLARE_INSN(fmul_d, MATCH_FMUL_D, MASK_FMUL_D) +DECLARE_INSN(fmul_h, MATCH_FMUL_H, MASK_FMUL_H) +DECLARE_INSN(fmul_s, MATCH_FMUL_S, MASK_FMUL_S) +DECLARE_INSN(fmv_d_x, MATCH_FMV_D_X, MASK_FMV_D_X) +DECLARE_INSN(fmv_h_x, MATCH_FMV_H_X, MASK_FMV_H_X) +DECLARE_INSN(fmv_s_x, MATCH_FMV_S_X, MASK_FMV_S_X) +DECLARE_INSN(fmv_x_d, MATCH_FMV_X_D, MASK_FMV_X_D) +DECLARE_INSN(fmv_x_h, MATCH_FMV_X_H, MASK_FMV_X_H) +DECLARE_INSN(fmv_x_s, MATCH_FMV_X_S, MASK_FMV_X_S) +DECLARE_INSN(fnmadd_d, MATCH_FNMADD_D, MASK_FNMADD_D) +DECLARE_INSN(fnmadd_h, MATCH_FNMADD_H, MASK_FNMADD_H) +DECLARE_INSN(fnmadd_s, MATCH_FNMADD_S, MASK_FNMADD_S) +DECLARE_INSN(fnmsub_d, MATCH_FNMSUB_D, MASK_FNMSUB_D) +DECLARE_INSN(fnmsub_h, MATCH_FNMSUB_H, MASK_FNMSUB_H) +DECLARE_INSN(fnmsub_s, MATCH_FNMSUB_S, MASK_FNMSUB_S) +DECLARE_INSN(frcsr, MATCH_FRCSR, MASK_FRCSR) +DECLARE_INSN(frflags, MATCH_FRFLAGS, MASK_FRFLAGS) +DECLARE_INSN(frrm, MATCH_FRRM, MASK_FRRM) +DECLARE_INSN(fscsr, MATCH_FSCSR, MASK_FSCSR) +DECLARE_INSN(fsd, MATCH_FSD, MASK_FSD) +DECLARE_INSN(fsflags, MATCH_FSFLAGS, MASK_FSFLAGS) +DECLARE_INSN(fsflagsi, MATCH_FSFLAGSI, MASK_FSFLAGSI) +DECLARE_INSN(fsgnj_d, MATCH_FSGNJ_D, MASK_FSGNJ_D) +DECLARE_INSN(fsgnj_h, MATCH_FSGNJ_H, MASK_FSGNJ_H) +DECLARE_INSN(fsgnj_s, MATCH_FSGNJ_S, MASK_FSGNJ_S) +DECLARE_INSN(fsgnjn_d, MATCH_FSGNJN_D, MASK_FSGNJN_D) +DECLARE_INSN(fsgnjn_h, MATCH_FSGNJN_H, MASK_FSGNJN_H) +DECLARE_INSN(fsgnjn_s, MATCH_FSGNJN_S, MASK_FSGNJN_S) +DECLARE_INSN(fsgnjx_d, MATCH_FSGNJX_D, MASK_FSGNJX_D) +DECLARE_INSN(fsgnjx_h, MATCH_FSGNJX_H, MASK_FSGNJX_H) +DECLARE_INSN(fsgnjx_s, MATCH_FSGNJX_S, MASK_FSGNJX_S) +DECLARE_INSN(fsh, MATCH_FSH, MASK_FSH) +DECLARE_INSN(fsqrt_d, MATCH_FSQRT_D, MASK_FSQRT_D) +DECLARE_INSN(fsqrt_h, MATCH_FSQRT_H, MASK_FSQRT_H) +DECLARE_INSN(fsqrt_s, MATCH_FSQRT_S, MASK_FSQRT_S) +DECLARE_INSN(fsrm, MATCH_FSRM, MASK_FSRM) +DECLARE_INSN(fsrmi, MATCH_FSRMI, MASK_FSRMI) +DECLARE_INSN(fsub_d, MATCH_FSUB_D, MASK_FSUB_D) +DECLARE_INSN(fsub_h, MATCH_FSUB_H, MASK_FSUB_H) +DECLARE_INSN(fsub_s, MATCH_FSUB_S, MASK_FSUB_S) +DECLARE_INSN(fsw, MATCH_FSW, MASK_FSW) +DECLARE_INSN(hrts, MATCH_HRTS, MASK_HRTS) +DECLARE_INSN(jal, MATCH_JAL, MASK_JAL) +DECLARE_INSN(jalr, MATCH_JALR, MASK_JALR) +DECLARE_INSN(lb, MATCH_LB, MASK_LB) +DECLARE_INSN(lbu, MATCH_LBU, MASK_LBU) +DECLARE_INSN(ld, MATCH_LD, MASK_LD) +DECLARE_INSN(lh, MATCH_LH, MASK_LH) +DECLARE_INSN(lhu, MATCH_LHU, MASK_LHU) +DECLARE_INSN(lr_d, MATCH_LR_D, MASK_LR_D) +DECLARE_INSN(lr_w, MATCH_LR_W, MASK_LR_W) +DECLARE_INSN(lui, MATCH_LUI, MASK_LUI) +DECLARE_INSN(lw, MATCH_LW, MASK_LW) +DECLARE_INSN(lwu, MATCH_LWU, MASK_LWU) +DECLARE_INSN(movn, MATCH_MOVN, MASK_MOVN) +DECLARE_INSN(movz, MATCH_MOVZ, MASK_MOVZ) +DECLARE_INSN(mrth, MATCH_MRTH, MASK_MRTH) +DECLARE_INSN(mrts, MATCH_MRTS, MASK_MRTS) +DECLARE_INSN(mul, MATCH_MUL, MASK_MUL) +DECLARE_INSN(mulh, MATCH_MULH, MASK_MULH) +DECLARE_INSN(mulhsu, MATCH_MULHSU, MASK_MULHSU) +DECLARE_INSN(mulhu, MATCH_MULHU, MASK_MULHU) +DECLARE_INSN(mulw, MATCH_MULW, MASK_MULW) +DECLARE_INSN(or, MATCH_OR, MASK_OR) +DECLARE_INSN(ori, MATCH_ORI, MASK_ORI) +DECLARE_INSN(rdcycle, MATCH_RDCYCLE, MASK_RDCYCLE) +DECLARE_INSN(rdcycleh, MATCH_RDCYCLEH, MASK_RDCYCLEH) +DECLARE_INSN(rdinstret, MATCH_RDINSTRET, MASK_RDINSTRET) +DECLARE_INSN(rdinstreth, MATCH_RDINSTRETH, MASK_RDINSTRETH) +DECLARE_INSN(rdtime, MATCH_RDTIME, MASK_RDTIME) +DECLARE_INSN(rdtimeh, MATCH_RDTIMEH, MASK_RDTIMEH) +DECLARE_INSN(rem, MATCH_REM, MASK_REM) +DECLARE_INSN(remu, MATCH_REMU, MASK_REMU) +DECLARE_INSN(remuw, MATCH_REMUW, MASK_REMUW) +DECLARE_INSN(remw, MATCH_REMW, MASK_REMW) +DECLARE_INSN(sb, MATCH_SB, MASK_SB) +DECLARE_INSN(sbreak, MATCH_SBREAK, MASK_SBREAK) +DECLARE_INSN(sc_d, MATCH_SC_D, MASK_SC_D) +DECLARE_INSN(sc_w, MATCH_SC_W, MASK_SC_W) +DECLARE_INSN(scall, MATCH_SCALL, MASK_SCALL) +DECLARE_INSN(sd, MATCH_SD, MASK_SD) +DECLARE_INSN(sfence_vm, MATCH_SFENCE_VM, MASK_SFENCE_VM) +DECLARE_INSN(sh, MATCH_SH, MASK_SH) +DECLARE_INSN(sll, MATCH_SLL, MASK_SLL) +DECLARE_INSN(slli, MATCH_SLLI, MASK_SLLI) +DECLARE_INSN(slli_rv32, MATCH_SLLI_RV32, MASK_SLLI_RV32) +DECLARE_INSN(slliw, MATCH_SLLIW, MASK_SLLIW) +DECLARE_INSN(sllw, MATCH_SLLW, MASK_SLLW) +DECLARE_INSN(slt, MATCH_SLT, MASK_SLT) +DECLARE_INSN(slti, MATCH_SLTI, MASK_SLTI) +DECLARE_INSN(sltiu, MATCH_SLTIU, MASK_SLTIU) +DECLARE_INSN(sltu, MATCH_SLTU, MASK_SLTU) +DECLARE_INSN(sra, MATCH_SRA, MASK_SRA) +DECLARE_INSN(srai, MATCH_SRAI, MASK_SRAI) +DECLARE_INSN(srai_rv32, MATCH_SRAI_RV32, MASK_SRAI_RV32) +DECLARE_INSN(sraiw, MATCH_SRAIW, MASK_SRAIW) +DECLARE_INSN(sraw, MATCH_SRAW, MASK_SRAW) +DECLARE_INSN(sret, MATCH_SRET, MASK_SRET) +DECLARE_INSN(srl, MATCH_SRL, MASK_SRL) +DECLARE_INSN(srli, MATCH_SRLI, MASK_SRLI) +DECLARE_INSN(srli_rv32, MATCH_SRLI_RV32, MASK_SRLI_RV32) +DECLARE_INSN(srliw, MATCH_SRLIW, MASK_SRLIW) +DECLARE_INSN(srlw, MATCH_SRLW, MASK_SRLW) +DECLARE_INSN(stop, MATCH_STOP, MASK_STOP) +DECLARE_INSN(sub, MATCH_SUB, MASK_SUB) +DECLARE_INSN(subw, MATCH_SUBW, MASK_SUBW) +DECLARE_INSN(sw, MATCH_SW, MASK_SW) +DECLARE_INSN(utidx, MATCH_UTIDX, MASK_UTIDX) +DECLARE_INSN(venqcmd, MATCH_VENQCMD, MASK_VENQCMD) +DECLARE_INSN(venqcnt, MATCH_VENQCNT, MASK_VENQCNT) +DECLARE_INSN(venqimm1, MATCH_VENQIMM1, MASK_VENQIMM1) +DECLARE_INSN(venqimm2, MATCH_VENQIMM2, MASK_VENQIMM2) +DECLARE_INSN(vf, MATCH_VF, MASK_VF) +DECLARE_INSN(vfld, MATCH_VFLD, MASK_VFLD) +DECLARE_INSN(vflsegd, MATCH_VFLSEGD, MASK_VFLSEGD) +DECLARE_INSN(vflsegstd, MATCH_VFLSEGSTD, MASK_VFLSEGSTD) +DECLARE_INSN(vflsegstw, MATCH_VFLSEGSTW, MASK_VFLSEGSTW) +DECLARE_INSN(vflsegw, MATCH_VFLSEGW, MASK_VFLSEGW) +DECLARE_INSN(vflstd, MATCH_VFLSTD, MASK_VFLSTD) +DECLARE_INSN(vflstw, MATCH_VFLSTW, MASK_VFLSTW) +DECLARE_INSN(vflw, MATCH_VFLW, MASK_VFLW) +DECLARE_INSN(vfmsv_d, MATCH_VFMSV_D, MASK_VFMSV_D) +DECLARE_INSN(vfmsv_s, MATCH_VFMSV_S, MASK_VFMSV_S) +DECLARE_INSN(vfmvv, MATCH_VFMVV, MASK_VFMVV) +DECLARE_INSN(vfsd, MATCH_VFSD, MASK_VFSD) +DECLARE_INSN(vfssegd, MATCH_VFSSEGD, MASK_VFSSEGD) +DECLARE_INSN(vfssegstd, MATCH_VFSSEGSTD, MASK_VFSSEGSTD) +DECLARE_INSN(vfssegstw, MATCH_VFSSEGSTW, MASK_VFSSEGSTW) +DECLARE_INSN(vfssegw, MATCH_VFSSEGW, MASK_VFSSEGW) +DECLARE_INSN(vfsstd, MATCH_VFSSTD, MASK_VFSSTD) +DECLARE_INSN(vfsstw, MATCH_VFSSTW, MASK_VFSSTW) +DECLARE_INSN(vfsw, MATCH_VFSW, MASK_VFSW) +DECLARE_INSN(vgetcfg, MATCH_VGETCFG, MASK_VGETCFG) +DECLARE_INSN(vgetvl, MATCH_VGETVL, MASK_VGETVL) +DECLARE_INSN(vlb, MATCH_VLB, MASK_VLB) +DECLARE_INSN(vlbu, MATCH_VLBU, MASK_VLBU) +DECLARE_INSN(vld, MATCH_VLD, MASK_VLD) +DECLARE_INSN(vlh, MATCH_VLH, MASK_VLH) +DECLARE_INSN(vlhu, MATCH_VLHU, MASK_VLHU) +DECLARE_INSN(vlsegb, MATCH_VLSEGB, MASK_VLSEGB) +DECLARE_INSN(vlsegbu, MATCH_VLSEGBU, MASK_VLSEGBU) +DECLARE_INSN(vlsegd, MATCH_VLSEGD, MASK_VLSEGD) +DECLARE_INSN(vlsegh, MATCH_VLSEGH, MASK_VLSEGH) +DECLARE_INSN(vlseghu, MATCH_VLSEGHU, MASK_VLSEGHU) +DECLARE_INSN(vlsegstb, MATCH_VLSEGSTB, MASK_VLSEGSTB) +DECLARE_INSN(vlsegstbu, MATCH_VLSEGSTBU, MASK_VLSEGSTBU) +DECLARE_INSN(vlsegstd, MATCH_VLSEGSTD, MASK_VLSEGSTD) +DECLARE_INSN(vlsegsth, MATCH_VLSEGSTH, MASK_VLSEGSTH) +DECLARE_INSN(vlsegsthu, MATCH_VLSEGSTHU, MASK_VLSEGSTHU) +DECLARE_INSN(vlsegstw, MATCH_VLSEGSTW, MASK_VLSEGSTW) +DECLARE_INSN(vlsegstwu, MATCH_VLSEGSTWU, MASK_VLSEGSTWU) +DECLARE_INSN(vlsegw, MATCH_VLSEGW, MASK_VLSEGW) +DECLARE_INSN(vlsegwu, MATCH_VLSEGWU, MASK_VLSEGWU) +DECLARE_INSN(vlstb, MATCH_VLSTB, MASK_VLSTB) +DECLARE_INSN(vlstbu, MATCH_VLSTBU, MASK_VLSTBU) +DECLARE_INSN(vlstd, MATCH_VLSTD, MASK_VLSTD) +DECLARE_INSN(vlsth, MATCH_VLSTH, MASK_VLSTH) +DECLARE_INSN(vlsthu, MATCH_VLSTHU, MASK_VLSTHU) +DECLARE_INSN(vlstw, MATCH_VLSTW, MASK_VLSTW) +DECLARE_INSN(vlstwu, MATCH_VLSTWU, MASK_VLSTWU) +DECLARE_INSN(vlw, MATCH_VLW, MASK_VLW) +DECLARE_INSN(vlwu, MATCH_VLWU, MASK_VLWU) +DECLARE_INSN(vmsv, MATCH_VMSV, MASK_VMSV) +DECLARE_INSN(vmvv, MATCH_VMVV, MASK_VMVV) +DECLARE_INSN(vsb, MATCH_VSB, MASK_VSB) +DECLARE_INSN(vsd, MATCH_VSD, MASK_VSD) +DECLARE_INSN(vsetcfg, MATCH_VSETCFG, MASK_VSETCFG) +DECLARE_INSN(vsetvl, MATCH_VSETVL, MASK_VSETVL) +DECLARE_INSN(vsh, MATCH_VSH, MASK_VSH) +DECLARE_INSN(vssegb, MATCH_VSSEGB, MASK_VSSEGB) +DECLARE_INSN(vssegd, MATCH_VSSEGD, MASK_VSSEGD) +DECLARE_INSN(vssegh, MATCH_VSSEGH, MASK_VSSEGH) +DECLARE_INSN(vssegstb, MATCH_VSSEGSTB, MASK_VSSEGSTB) +DECLARE_INSN(vssegstd, MATCH_VSSEGSTD, MASK_VSSEGSTD) +DECLARE_INSN(vssegsth, MATCH_VSSEGSTH, MASK_VSSEGSTH) +DECLARE_INSN(vssegstw, MATCH_VSSEGSTW, MASK_VSSEGSTW) +DECLARE_INSN(vssegw, MATCH_VSSEGW, MASK_VSSEGW) +DECLARE_INSN(vsstb, MATCH_VSSTB, MASK_VSSTB) +DECLARE_INSN(vsstd, MATCH_VSSTD, MASK_VSSTD) +DECLARE_INSN(vssth, MATCH_VSSTH, MASK_VSSTH) +DECLARE_INSN(vsstw, MATCH_VSSTW, MASK_VSSTW) +DECLARE_INSN(vsw, MATCH_VSW, MASK_VSW) +DECLARE_INSN(vxcptaux, MATCH_VXCPTAUX, MASK_VXCPTAUX) +DECLARE_INSN(vxcptcause, MATCH_VXCPTCAUSE, MASK_VXCPTCAUSE) +DECLARE_INSN(vxcptevac, MATCH_VXCPTEVAC, MASK_VXCPTEVAC) +DECLARE_INSN(vxcpthold, MATCH_VXCPTHOLD, MASK_VXCPTHOLD) +DECLARE_INSN(vxcptkill, MATCH_VXCPTKILL, MASK_VXCPTKILL) +DECLARE_INSN(vxcptrestore, MATCH_VXCPTRESTORE, MASK_VXCPTRESTORE) +DECLARE_INSN(vxcptsave, MATCH_VXCPTSAVE, MASK_VXCPTSAVE) +DECLARE_INSN(wfi, MATCH_WFI, MASK_WFI) +DECLARE_INSN(xor, MATCH_XOR, MASK_XOR) +DECLARE_INSN(xori, MATCH_XORI, MASK_XORI) +#endif +#ifdef DECLARE_CSR +DECLARE_CSR(fflags, CSR_FFLAGS) +DECLARE_CSR(frm, CSR_FRM) +DECLARE_CSR(fcsr, CSR_FCSR) +DECLARE_CSR(cycle, CSR_CYCLE) +DECLARE_CSR(time, CSR_TIME) +DECLARE_CSR(instret, CSR_INSTRET) +DECLARE_CSR(stats, CSR_STATS) +DECLARE_CSR(uarch0, CSR_UARCH0) +DECLARE_CSR(uarch1, CSR_UARCH1) +DECLARE_CSR(uarch2, CSR_UARCH2) +DECLARE_CSR(uarch3, CSR_UARCH3) +DECLARE_CSR(uarch4, CSR_UARCH4) +DECLARE_CSR(uarch5, CSR_UARCH5) +DECLARE_CSR(uarch6, CSR_UARCH6) +DECLARE_CSR(uarch7, CSR_UARCH7) +DECLARE_CSR(uarch8, CSR_UARCH8) +DECLARE_CSR(uarch9, CSR_UARCH9) +DECLARE_CSR(uarch10, CSR_UARCH10) +DECLARE_CSR(uarch11, CSR_UARCH11) +DECLARE_CSR(uarch12, CSR_UARCH12) +DECLARE_CSR(uarch13, CSR_UARCH13) +DECLARE_CSR(uarch14, CSR_UARCH14) +DECLARE_CSR(uarch15, CSR_UARCH15) +DECLARE_CSR(sstatus, CSR_SSTATUS) +DECLARE_CSR(stvec, CSR_STVEC) +DECLARE_CSR(sie, CSR_SIE) +DECLARE_CSR(sscratch, CSR_SSCRATCH) +DECLARE_CSR(sepc, CSR_SEPC) +DECLARE_CSR(sip, CSR_SIP) +DECLARE_CSR(sptbr, CSR_SPTBR) +DECLARE_CSR(sasid, CSR_SASID) +DECLARE_CSR(cyclew, CSR_CYCLEW) +DECLARE_CSR(timew, CSR_TIMEW) +DECLARE_CSR(instretw, CSR_INSTRETW) +DECLARE_CSR(stime, CSR_STIME) +DECLARE_CSR(scause, CSR_SCAUSE) +DECLARE_CSR(sbadaddr, CSR_SBADADDR) +DECLARE_CSR(stimew, CSR_STIMEW) +DECLARE_CSR(mstatus, CSR_MSTATUS) +DECLARE_CSR(mtvec, CSR_MTVEC) +DECLARE_CSR(mtdeleg, CSR_MTDELEG) +DECLARE_CSR(mie, CSR_MIE) +DECLARE_CSR(mtimecmp, CSR_MTIMECMP) +DECLARE_CSR(mscratch, CSR_MSCRATCH) +DECLARE_CSR(mepc, CSR_MEPC) +DECLARE_CSR(mcause, CSR_MCAUSE) +DECLARE_CSR(mbadaddr, CSR_MBADADDR) +DECLARE_CSR(mip, CSR_MIP) +DECLARE_CSR(mtime, CSR_MTIME) +DECLARE_CSR(mcpuid, CSR_MCPUID) +DECLARE_CSR(mimpid, CSR_MIMPID) +DECLARE_CSR(mhartid, CSR_MHARTID) +DECLARE_CSR(mtohost, CSR_MTOHOST) +DECLARE_CSR(mfromhost, CSR_MFROMHOST) +DECLARE_CSR(mreset, CSR_MRESET) +DECLARE_CSR(send_ipi, CSR_SEND_IPI) +DECLARE_CSR(cycleh, CSR_CYCLEH) +DECLARE_CSR(timeh, CSR_TIMEH) +DECLARE_CSR(instreth, CSR_INSTRETH) +DECLARE_CSR(cyclehw, CSR_CYCLEHW) +DECLARE_CSR(timehw, CSR_TIMEHW) +DECLARE_CSR(instrethw, CSR_INSTRETHW) +DECLARE_CSR(stimeh, CSR_STIMEH) +DECLARE_CSR(stimehw, CSR_STIMEHW) +DECLARE_CSR(mtimecmph, CSR_MTIMECMPH) +DECLARE_CSR(mtimeh, CSR_MTIMEH) +#endif +#ifdef DECLARE_CAUSE +DECLARE_CAUSE("fflags", CAUSE_FFLAGS) +DECLARE_CAUSE("frm", CAUSE_FRM) +DECLARE_CAUSE("fcsr", CAUSE_FCSR) +DECLARE_CAUSE("cycle", CAUSE_CYCLE) +DECLARE_CAUSE("time", CAUSE_TIME) +DECLARE_CAUSE("instret", CAUSE_INSTRET) +DECLARE_CAUSE("stats", CAUSE_STATS) +DECLARE_CAUSE("uarch0", CAUSE_UARCH0) +DECLARE_CAUSE("uarch1", CAUSE_UARCH1) +DECLARE_CAUSE("uarch2", CAUSE_UARCH2) +DECLARE_CAUSE("uarch3", CAUSE_UARCH3) +DECLARE_CAUSE("uarch4", CAUSE_UARCH4) +DECLARE_CAUSE("uarch5", CAUSE_UARCH5) +DECLARE_CAUSE("uarch6", CAUSE_UARCH6) +DECLARE_CAUSE("uarch7", CAUSE_UARCH7) +DECLARE_CAUSE("uarch8", CAUSE_UARCH8) +DECLARE_CAUSE("uarch9", CAUSE_UARCH9) +DECLARE_CAUSE("uarch10", CAUSE_UARCH10) +DECLARE_CAUSE("uarch11", CAUSE_UARCH11) +DECLARE_CAUSE("uarch12", CAUSE_UARCH12) +DECLARE_CAUSE("uarch13", CAUSE_UARCH13) +DECLARE_CAUSE("uarch14", CAUSE_UARCH14) +DECLARE_CAUSE("uarch15", CAUSE_UARCH15) +DECLARE_CAUSE("sstatus", CAUSE_SSTATUS) +DECLARE_CAUSE("stvec", CAUSE_STVEC) +DECLARE_CAUSE("sie", CAUSE_SIE) +DECLARE_CAUSE("sscratch", CAUSE_SSCRATCH) +DECLARE_CAUSE("sepc", CAUSE_SEPC) +DECLARE_CAUSE("sip", CAUSE_SIP) +DECLARE_CAUSE("sptbr", CAUSE_SPTBR) +DECLARE_CAUSE("sasid", CAUSE_SASID) +DECLARE_CAUSE("cyclew", CAUSE_CYCLEW) +DECLARE_CAUSE("timew", CAUSE_TIMEW) +DECLARE_CAUSE("instretw", CAUSE_INSTRETW) +DECLARE_CAUSE("stime", CAUSE_STIME) +DECLARE_CAUSE("scause", CAUSE_SCAUSE) +DECLARE_CAUSE("sbadaddr", CAUSE_SBADADDR) +DECLARE_CAUSE("stimew", CAUSE_STIMEW) +DECLARE_CAUSE("mstatus", CAUSE_MSTATUS) +DECLARE_CAUSE("mtvec", CAUSE_MTVEC) +DECLARE_CAUSE("mtdeleg", CAUSE_MTDELEG) +DECLARE_CAUSE("mie", CAUSE_MIE) +DECLARE_CAUSE("mtimecmp", CAUSE_MTIMECMP) +DECLARE_CAUSE("mscratch", CAUSE_MSCRATCH) +DECLARE_CAUSE("mepc", CAUSE_MEPC) +DECLARE_CAUSE("mcause", CAUSE_MCAUSE) +DECLARE_CAUSE("mbadaddr", CAUSE_MBADADDR) +DECLARE_CAUSE("mip", CAUSE_MIP) +DECLARE_CAUSE("mtime", CAUSE_MTIME) +DECLARE_CAUSE("mcpuid", CAUSE_MCPUID) +DECLARE_CAUSE("mimpid", CAUSE_MIMPID) +DECLARE_CAUSE("mhartid", CAUSE_MHARTID) +DECLARE_CAUSE("mtohost", CAUSE_MTOHOST) +DECLARE_CAUSE("mfromhost", CAUSE_MFROMHOST) +DECLARE_CAUSE("mreset", CAUSE_MRESET) +DECLARE_CAUSE("send_ipi", CAUSE_SEND_IPI) +DECLARE_CAUSE("cycleh", CAUSE_CYCLEH) +DECLARE_CAUSE("timeh", CAUSE_TIMEH) +DECLARE_CAUSE("instreth", CAUSE_INSTRETH) +DECLARE_CAUSE("cyclehw", CAUSE_CYCLEHW) +DECLARE_CAUSE("timehw", CAUSE_TIMEHW) +DECLARE_CAUSE("instrethw", CAUSE_INSTRETHW) +DECLARE_CAUSE("stimeh", CAUSE_STIMEH) +DECLARE_CAUSE("stimehw", CAUSE_STIMEHW) +DECLARE_CAUSE("mtimecmph", CAUSE_MTIMECMPH) +DECLARE_CAUSE("mtimeh", CAUSE_MTIMEH) +#endif diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h new file mode 100644 index 0000000..504857c --- /dev/null +++ b/gcc/config/riscv/riscv-protos.h @@ -0,0 +1,96 @@ +/* Definition of RISC-V target for GNU compiler. + Copyright (C) 2011-2014 Free Software Foundation, Inc. + Contributed by Andrew Waterman (waterman@cs.berkeley.edu) at UC Berkeley. + Based on MIPS target for GNU compiler. + +This file is part of GCC. + +GCC is free software; you can redistribute it and/or modify +it under the terms of the GNU General Public License as published by +the Free Software Foundation; either version 3, or (at your option) +any later version. + +GCC is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +You should have received a copy of the GNU General Public License +along with GCC; see the file COPYING3. If not see +. */ + +#ifndef GCC_RISCV_PROTOS_H +#define GCC_RISCV_PROTOS_H + +enum riscv_symbol_type { + SYMBOL_ABSOLUTE, + SYMBOL_GOT_DISP, + SYMBOL_TLS, + SYMBOL_TLS_LE, + SYMBOL_TLS_IE, + SYMBOL_TLS_GD +}; +#define NUM_SYMBOL_TYPES (SYMBOL_TLS_GD + 1) + +enum riscv_code_model { + CM_MEDLOW, + CM_MEDANY, + CM_PIC +}; +extern enum riscv_code_model riscv_cmodel; + +extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); +extern int riscv_regno_mode_ok_for_base_p (int, enum machine_mode, bool); +extern int riscv_address_insns (rtx, enum machine_mode, bool); +extern int riscv_const_insns (rtx); +extern int riscv_split_const_insns (rtx); +extern int riscv_load_store_insns (rtx, rtx); +extern rtx riscv_emit_move (rtx, rtx); +extern bool riscv_split_symbol (rtx, rtx, enum machine_mode, rtx *); +extern rtx riscv_unspec_address (rtx, enum riscv_symbol_type); +extern void riscv_move_integer (rtx, rtx, HOST_WIDE_INT); +extern bool riscv_legitimize_move (enum machine_mode, rtx, rtx); +extern bool riscv_legitimize_vector_move (enum machine_mode, rtx, rtx); + +extern rtx riscv_subword (rtx, bool); +extern bool riscv_split_64bit_move_p (rtx, rtx); +extern void riscv_split_doubleword_move (rtx, rtx); +extern const char *riscv_output_move (rtx, rtx); +extern const char *riscv_output_gpr_save (unsigned); +#ifdef RTX_CODE +extern void riscv_expand_scc (rtx *); +extern void riscv_expand_conditional_branch (rtx *); +#endif +extern rtx riscv_expand_call (bool, rtx, rtx, rtx); +extern void riscv_expand_fcc_reload (rtx, rtx, rtx); +extern void riscv_set_return_address (rtx, rtx); +extern bool riscv_expand_block_move (rtx, rtx, rtx); +extern void riscv_expand_synci_loop (rtx, rtx); + +extern bool riscv_expand_ext_as_unaligned_load (rtx, rtx, HOST_WIDE_INT, + HOST_WIDE_INT); +extern bool riscv_expand_ins_as_unaligned_store (rtx, rtx, HOST_WIDE_INT, + HOST_WIDE_INT); +extern void riscv_order_regs_for_local_alloc (void); + +extern rtx riscv_return_addr (int, rtx); +extern HOST_WIDE_INT riscv_initial_elimination_offset (int, int); +extern void riscv_expand_prologue (void); +extern void riscv_expand_epilogue (bool); +extern bool riscv_can_use_return_insn (void); +extern rtx riscv_function_value (const_tree, const_tree, enum machine_mode); + +extern enum reg_class riscv_secondary_reload_class (enum reg_class, + enum machine_mode, + rtx, bool); +extern int riscv_class_max_nregs (enum reg_class, enum machine_mode); + +extern unsigned int riscv_hard_regno_nregs (int, enum machine_mode); + +extern void irix_asm_output_align (FILE *, unsigned); +extern const char *current_section_name (void); +extern unsigned int current_section_flags (void); + +extern void riscv_expand_vector_init (rtx, rtx); + +#endif /* ! GCC_RISCV_PROTOS_H */ diff --git a/gcc/config/riscv/riscv.c b/gcc/config/riscv/riscv.c new file mode 100644 index 0000000..3af039c --- /dev/null +++ b/gcc/config/riscv/riscv.c @@ -0,0 +1,4411 @@ +/* Subroutines used for code generation for RISC-V. + Copyright (C) 2011-2014 Free Software Foundation, Inc. + Contributed by Andrew Waterman (waterman@cs.berkeley.edu) at UC Berkeley. + Based on MIPS target for GNU compiler. + +This file is part of GCC. + +GCC is free software; you can redistribute it and/or modify +it under the terms of the GNU General Public License as published by +the Free Software Foundation; either version 3, or (at your option) +any later version. + +GCC is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +You should have received a copy of the GNU General Public License +along with GCC; see the file COPYING3. If not see +. */ + +#include "config.h" +#include "system.h" +#include "coretypes.h" +#include "tm.h" +#include "rtl.h" +#include "regs.h" +#include "hard-reg-set.h" +#include "insn-config.h" +#include "conditions.h" +#include "insn-attr.h" +#include "recog.h" +#include "output.h" +#include "tree.h" +#include "varasm.h" +#include "stor-layout.h" +#include "calls.h" +#include "function.h" +#include "expr.h" +#include "optabs.h" +#include "libfuncs.h" +#include "flags.h" +#include "reload.h" +#include "tm_p.h" +#include "ggc.h" +#include "gstab.h" +#include "hashtab.h" +#include "debug.h" +#include "target.h" +#include "target-def.h" +#include "langhooks.h" +#include "sched-int.h" +#include "bitmap.h" +#include "diagnostic.h" +#include "target-globals.h" +#include "symcat.h" +#include + +/* True if X is an UNSPEC wrapper around a SYMBOL_REF or LABEL_REF. */ +#define UNSPEC_ADDRESS_P(X) \ + (GET_CODE (X) == UNSPEC \ + && XINT (X, 1) >= UNSPEC_ADDRESS_FIRST \ + && XINT (X, 1) < UNSPEC_ADDRESS_FIRST + NUM_SYMBOL_TYPES) + +/* Extract the symbol or label from UNSPEC wrapper X. */ +#define UNSPEC_ADDRESS(X) \ + XVECEXP (X, 0, 0) + +/* Extract the symbol type from UNSPEC wrapper X. */ +#define UNSPEC_ADDRESS_TYPE(X) \ + ((enum riscv_symbol_type) (XINT (X, 1) - UNSPEC_ADDRESS_FIRST)) + +/* The maximum distance between the top of the stack frame and the + value sp has when we save and restore registers. This is set by the + range of load/store offsets and must also preserve stack alignment. */ +#define RISCV_MAX_FIRST_STACK_STEP (RISCV_IMM_REACH/2 - 16) + +/* True if INSN is a riscv.md pattern or asm statement. */ +#define USEFUL_INSN_P(INSN) \ + (NONDEBUG_INSN_P (INSN) \ + && GET_CODE (PATTERN (INSN)) != USE \ + && GET_CODE (PATTERN (INSN)) != CLOBBER \ + && GET_CODE (PATTERN (INSN)) != ADDR_VEC \ + && GET_CODE (PATTERN (INSN)) != ADDR_DIFF_VEC) + +/* True if bit BIT is set in VALUE. */ +#define BITSET_P(VALUE, BIT) (((VALUE) & (1 << (BIT))) != 0) + +/* Classifies an address. + + ADDRESS_REG + A natural register + offset address. The register satisfies + riscv_valid_base_register_p and the offset is a const_arith_operand. + + ADDRESS_LO_SUM + A LO_SUM rtx. The first operand is a valid base register and + the second operand is a symbolic address. + + ADDRESS_CONST_INT + A signed 16-bit constant address. + + ADDRESS_SYMBOLIC: + A constant symbolic address. */ +enum riscv_address_type { + ADDRESS_REG, + ADDRESS_LO_SUM, + ADDRESS_CONST_INT, + ADDRESS_SYMBOLIC +}; + +enum riscv_code_model riscv_cmodel = TARGET_DEFAULT_CMODEL; + +/* Macros to create an enumeration identifier for a function prototype. */ +#define RISCV_FTYPE_NAME1(A, B) RISCV_##A##_FTYPE_##B +#define RISCV_FTYPE_NAME2(A, B, C) RISCV_##A##_FTYPE_##B##_##C +#define RISCV_FTYPE_NAME3(A, B, C, D) RISCV_##A##_FTYPE_##B##_##C##_##D +#define RISCV_FTYPE_NAME4(A, B, C, D, E) RISCV_##A##_FTYPE_##B##_##C##_##D##_##E + +/* Classifies the prototype of a built-in function. */ +enum riscv_function_type { +#define DEF_RISCV_FTYPE(NARGS, LIST) RISCV_FTYPE_NAME##NARGS LIST, +#include "config/riscv/riscv-ftypes.def" +#undef DEF_RISCV_FTYPE + RISCV_MAX_FTYPE_MAX +}; + +/* Specifies how a built-in function should be converted into rtl. */ +enum riscv_builtin_type { + /* The function corresponds directly to an .md pattern. The return + value is mapped to operand 0 and the arguments are mapped to + operands 1 and above. */ + RISCV_BUILTIN_DIRECT, + + /* The function corresponds directly to an .md pattern. There is no return + value and the arguments are mapped to operands 0 and above. */ + RISCV_BUILTIN_DIRECT_NO_TARGET +}; + +/* Information about a function's frame layout. */ +struct GTY(()) riscv_frame_info { + /* The size of the frame in bytes. */ + HOST_WIDE_INT total_size; + + /* Bit X is set if the function saves or restores GPR X. */ + unsigned int mask; + + /* Likewise FPR X. */ + unsigned int fmask; + + /* How much the GPR save/restore routines adjust sp (or 0 if unused). */ + unsigned save_libcall_adjustment; + + /* Offsets of fixed-point and floating-point save areas from frame bottom */ + HOST_WIDE_INT gp_sp_offset; + HOST_WIDE_INT fp_sp_offset; + + /* Offset of virtual frame pointer from stack pointer/frame bottom */ + HOST_WIDE_INT frame_pointer_offset; + + /* Offset of hard frame pointer from stack pointer/frame bottom */ + HOST_WIDE_INT hard_frame_pointer_offset; + + /* The offset of arg_pointer_rtx from the bottom of the frame. */ + HOST_WIDE_INT arg_pointer_offset; +}; + +struct GTY(()) machine_function { + /* The number of extra stack bytes taken up by register varargs. + This area is allocated by the callee at the very top of the frame. */ + int varargs_size; + + /* Cached return value of leaf_function_p. <0 if false, >0 if true. */ + int is_leaf; + + /* The current frame information, calculated by riscv_compute_frame_info. */ + struct riscv_frame_info frame; +}; + +/* Information about a single argument. */ +struct riscv_arg_info { + /* True if the argument is passed in a floating-point register, or + would have been if we hadn't run out of registers. */ + bool fpr_p; + + /* The number of words passed in registers, rounded up. */ + unsigned int reg_words; + + /* For EABI, the offset of the first register from GP_ARG_FIRST or + FP_ARG_FIRST. For other ABIs, the offset of the first register from + the start of the ABI's argument structure (see the CUMULATIVE_ARGS + comment for details). + + The value is MAX_ARGS_IN_REGISTERS if the argument is passed entirely + on the stack. */ + unsigned int reg_offset; + + /* The number of words that must be passed on the stack, rounded up. */ + unsigned int stack_words; + + /* The offset from the start of the stack overflow area of the argument's + first stack word. Only meaningful when STACK_WORDS is nonzero. */ + unsigned int stack_offset; +}; + +/* Information about an address described by riscv_address_type. + + ADDRESS_CONST_INT + No fields are used. + + ADDRESS_REG + REG is the base register and OFFSET is the constant offset. + + ADDRESS_LO_SUM + REG and OFFSET are the operands to the LO_SUM and SYMBOL_TYPE + is the type of symbol it references. + + ADDRESS_SYMBOLIC + SYMBOL_TYPE is the type of symbol that the address references. */ +struct riscv_address_info { + enum riscv_address_type type; + rtx reg; + rtx offset; + enum riscv_symbol_type symbol_type; +}; + +/* One stage in a constant building sequence. These sequences have + the form: + + A = VALUE[0] + A = A CODE[1] VALUE[1] + A = A CODE[2] VALUE[2] + ... + + where A is an accumulator, each CODE[i] is a binary rtl operation + and each VALUE[i] is a constant integer. CODE[0] is undefined. */ +struct riscv_integer_op { + enum rtx_code code; + unsigned HOST_WIDE_INT value; +}; + +/* The largest number of operations needed to load an integer constant. + The worst case is LUI, ADDI, SLLI, ADDI, SLLI, ADDI, SLLI, ADDI, + but we may attempt and reject even worse sequences. */ +#define RISCV_MAX_INTEGER_OPS 32 + +/* Costs of various operations on the different architectures. */ + +struct riscv_tune_info +{ + unsigned short fp_add[2]; + unsigned short fp_mul[2]; + unsigned short fp_div[2]; + unsigned short int_mul[2]; + unsigned short int_div[2]; + unsigned short issue_rate; + unsigned short branch_cost; + unsigned short fp_to_int_cost; + unsigned short memory_cost; +}; + +/* Information about one CPU we know about. */ +struct riscv_cpu_info { + /* This CPU's canonical name. */ + const char *name; + + /* The RISC-V ISA and extensions supported by this CPU. */ + const char *isa; + + /* Tuning parameters for this CPU. */ + const struct riscv_tune_info *tune_info; +}; + +/* Global variables for machine-dependent things. */ + +/* Which tuning parameters to use. */ +static const struct riscv_tune_info *tune_info; + +/* Index [M][R] is true if register R is allowed to hold a value of mode M. */ +bool riscv_hard_regno_mode_ok[(int) MAX_MACHINE_MODE][FIRST_PSEUDO_REGISTER]; + +/* riscv_lo_relocs[X] is the relocation to use when a symbol of type X + appears in a LO_SUM. It can be null if such LO_SUMs aren't valid or + if they are matched by a special .md file pattern. */ +const char *riscv_lo_relocs[NUM_SYMBOL_TYPES]; + +/* Likewise for HIGHs. */ +const char *riscv_hi_relocs[NUM_SYMBOL_TYPES]; + +/* Index R is the smallest register class that contains register R. */ +const enum reg_class riscv_regno_to_class[FIRST_PSEUDO_REGISTER] = { + GR_REGS, GR_REGS, GR_REGS, GR_REGS, + GR_REGS, T_REGS, T_REGS, T_REGS, + GR_REGS, GR_REGS, GR_REGS, GR_REGS, + GR_REGS, GR_REGS, GR_REGS, GR_REGS, + GR_REGS, GR_REGS, GR_REGS, GR_REGS, + GR_REGS, GR_REGS, GR_REGS, GR_REGS, + GR_REGS, GR_REGS, GR_REGS, GR_REGS, + T_REGS, T_REGS, T_REGS, T_REGS, + FP_REGS, FP_REGS, FP_REGS, FP_REGS, + FP_REGS, FP_REGS, FP_REGS, FP_REGS, + FP_REGS, FP_REGS, FP_REGS, FP_REGS, + FP_REGS, FP_REGS, FP_REGS, FP_REGS, + FP_REGS, FP_REGS, FP_REGS, FP_REGS, + FP_REGS, FP_REGS, FP_REGS, FP_REGS, + FP_REGS, FP_REGS, FP_REGS, FP_REGS, + FP_REGS, FP_REGS, FP_REGS, FP_REGS, + FRAME_REGS, FRAME_REGS, +}; + +/* Costs to use when optimizing for size. */ +static const struct riscv_tune_info rocket_tune_info = { + {COSTS_N_INSNS (4), COSTS_N_INSNS (5)}, /* fp_add */ + {COSTS_N_INSNS (4), COSTS_N_INSNS (5)}, /* fp_mul */ + {COSTS_N_INSNS (20), COSTS_N_INSNS (20)}, /* fp_div */ + {COSTS_N_INSNS (4), COSTS_N_INSNS (4)}, /* int_mul */ + {COSTS_N_INSNS (6), COSTS_N_INSNS (6)}, /* int_div */ + 1, /* issue_rate */ + 3, /* branch_cost */ + COSTS_N_INSNS (2), /* fp_to_int_cost */ + 5 /* memory_cost */ +}; + +/* Costs to use when optimizing for size. */ +static const struct riscv_tune_info optimize_size_tune_info = { + {COSTS_N_INSNS (1), COSTS_N_INSNS (1)}, /* fp_add */ + {COSTS_N_INSNS (1), COSTS_N_INSNS (1)}, /* fp_mul */ + {COSTS_N_INSNS (1), COSTS_N_INSNS (1)}, /* fp_div */ + {COSTS_N_INSNS (1), COSTS_N_INSNS (1)}, /* int_mul */ + {COSTS_N_INSNS (1), COSTS_N_INSNS (1)}, /* int_div */ + 1, /* issue_rate */ + 1, /* branch_cost */ + COSTS_N_INSNS (1), /* fp_to_int_cost */ + 1 /* memory_cost */ +}; + +/* A table describing all the processors GCC knows about. */ +static const struct riscv_cpu_info riscv_cpu_info_table[] = { + /* Entries for generic ISAs. */ + { "rocket", "IMAFD", &rocket_tune_info }, +}; + +/* Return the riscv_cpu_info entry for the given name string. */ + +static const struct riscv_cpu_info * +riscv_parse_cpu (const char *cpu_string) +{ + unsigned int i; + + for (i = 0; i < ARRAY_SIZE (riscv_cpu_info_table); i++) + if (strcmp (riscv_cpu_info_table[i].name, cpu_string) == 0) + return riscv_cpu_info_table + i; + + error ("unknown cpu `%s'", cpu_string); + return riscv_cpu_info_table; +} + +/* Fill CODES with a sequence of rtl operations to load VALUE. + Return the number of operations needed. */ + +static int +riscv_build_integer_1 (struct riscv_integer_op *codes, HOST_WIDE_INT value, + enum machine_mode mode) +{ + HOST_WIDE_INT low_part = RISCV_CONST_LOW_PART (value); + int cost = INT_MAX, alt_cost; + struct riscv_integer_op alt_codes[RISCV_MAX_INTEGER_OPS]; + + if (SMALL_OPERAND (value) || LUI_OPERAND (value)) + { + /* Simply ADDI or LUI */ + codes[0].code = UNKNOWN; + codes[0].value = value; + return 1; + } + + /* End with ADDI */ + if (low_part != 0 + && !(mode == HImode && (int16_t)(value - low_part) != (value - low_part))) + { + cost = 1 + riscv_build_integer_1 (codes, value - low_part, mode); + codes[cost-1].code = PLUS; + codes[cost-1].value = low_part; + } + + /* End with XORI */ + if (cost > 2 && (low_part < 0 || mode == HImode)) + { + alt_cost = 1 + riscv_build_integer_1 (alt_codes, value ^ low_part, mode); + alt_codes[alt_cost-1].code = XOR; + alt_codes[alt_cost-1].value = low_part; + if (alt_cost < cost) + cost = alt_cost, memcpy (codes, alt_codes, sizeof(alt_codes)); + } + + /* Eliminate trailing zeros and end with SLLI */ + if (cost > 2 && (value & 1) == 0) + { + int shift = 0; + while ((value & 1) == 0) + shift++, value >>= 1; + alt_cost = 1 + riscv_build_integer_1 (alt_codes, value, mode); + alt_codes[alt_cost-1].code = ASHIFT; + alt_codes[alt_cost-1].value = shift; + if (alt_cost < cost) + cost = alt_cost, memcpy (codes, alt_codes, sizeof(alt_codes)); + } + + gcc_assert (cost <= RISCV_MAX_INTEGER_OPS); + return cost; +} + +static int +riscv_build_integer (struct riscv_integer_op *codes, HOST_WIDE_INT value, + enum machine_mode mode) +{ + int cost = riscv_build_integer_1 (codes, value, mode); + + /* Eliminate leading zeros and end with SRLI */ + if (value > 0 && cost > 2) + { + struct riscv_integer_op alt_codes[RISCV_MAX_INTEGER_OPS]; + int alt_cost, shift = 0; + HOST_WIDE_INT shifted_val; + + /* Try filling trailing bits with 1s */ + while ((value << shift) >= 0) + shift++; + shifted_val = (value << shift) | ((((HOST_WIDE_INT) 1) << shift) - 1); + alt_cost = 1 + riscv_build_integer_1 (alt_codes, shifted_val, mode); + alt_codes[alt_cost-1].code = LSHIFTRT; + alt_codes[alt_cost-1].value = shift; + if (alt_cost < cost) + cost = alt_cost, memcpy (codes, alt_codes, sizeof (alt_codes)); + + /* Try filling trailing bits with 0s */ + shifted_val = value << shift; + alt_cost = 1 + riscv_build_integer_1 (alt_codes, shifted_val, mode); + alt_codes[alt_cost-1].code = LSHIFTRT; + alt_codes[alt_cost-1].value = shift; + if (alt_cost < cost) + cost = alt_cost, memcpy (codes, alt_codes, sizeof (alt_codes)); + } + + return cost; +} + +static int +riscv_split_integer_cost (HOST_WIDE_INT val) +{ + int cost; + int32_t loval = val, hival = (val - (int32_t)val) >> 32; + struct riscv_integer_op codes[RISCV_MAX_INTEGER_OPS]; + + cost = 2 + riscv_build_integer (codes, loval, VOIDmode); + if (loval != hival) + cost += riscv_build_integer (codes, hival, VOIDmode); + + return cost; +} + +static int +riscv_integer_cost (HOST_WIDE_INT val) +{ + struct riscv_integer_op codes[RISCV_MAX_INTEGER_OPS]; + return MIN (riscv_build_integer (codes, val, VOIDmode), + riscv_split_integer_cost (val)); +} + +/* Try to split a 64b integer into 32b parts, then reassemble. */ + +static rtx +riscv_split_integer (HOST_WIDE_INT val, enum machine_mode mode) +{ + int32_t loval = val, hival = (val - (int32_t)val) >> 32; + rtx hi = gen_reg_rtx (mode), lo = gen_reg_rtx (mode); + + riscv_move_integer (hi, hi, hival); + riscv_move_integer (lo, lo, loval); + + hi = gen_rtx_fmt_ee (ASHIFT, mode, hi, GEN_INT (32)); + hi = force_reg (mode, hi); + + return gen_rtx_fmt_ee (PLUS, mode, hi, lo); +} + +/* Return true if X is a thread-local symbol. */ + +static bool +riscv_tls_symbol_p (const_rtx x) +{ + return GET_CODE (x) == SYMBOL_REF && SYMBOL_REF_TLS_MODEL (x) != 0; +} + +static bool +riscv_symbol_binds_local_p (const_rtx x) +{ + return (SYMBOL_REF_DECL (x) + ? targetm.binds_local_p (SYMBOL_REF_DECL (x)) + : SYMBOL_REF_LOCAL_P (x)); +} + +/* Return the method that should be used to access SYMBOL_REF or + LABEL_REF X in context CONTEXT. */ + +static enum riscv_symbol_type +riscv_classify_symbol (const_rtx x) +{ + if (riscv_tls_symbol_p (x)) + return SYMBOL_TLS; + + if (GET_CODE (x) == LABEL_REF) + { + if (LABEL_REF_NONLOCAL_P (x)) + return SYMBOL_GOT_DISP; + return SYMBOL_ABSOLUTE; + } + + gcc_assert (GET_CODE (x) == SYMBOL_REF); + + if (flag_pic && !riscv_symbol_binds_local_p (x)) + return SYMBOL_GOT_DISP; + + return SYMBOL_ABSOLUTE; +} + +/* Classify the base of symbolic expression X, given that X appears in + context CONTEXT. */ + +static enum riscv_symbol_type +riscv_classify_symbolic_expression (rtx x) +{ + rtx offset; + + split_const (x, &x, &offset); + if (UNSPEC_ADDRESS_P (x)) + return UNSPEC_ADDRESS_TYPE (x); + + return riscv_classify_symbol (x); +} + +/* Return true if X is a symbolic constant that can be used in context + CONTEXT. If it is, store the type of the symbol in *SYMBOL_TYPE. */ + +bool +riscv_symbolic_constant_p (rtx x, enum riscv_symbol_type *symbol_type) +{ + rtx offset; + + split_const (x, &x, &offset); + if (UNSPEC_ADDRESS_P (x)) + { + *symbol_type = UNSPEC_ADDRESS_TYPE (x); + x = UNSPEC_ADDRESS (x); + } + else if (GET_CODE (x) == SYMBOL_REF || GET_CODE (x) == LABEL_REF) + *symbol_type = riscv_classify_symbol (x); + else + return false; + + if (offset == const0_rtx) + return true; + + /* Check whether a nonzero offset is valid for the underlying + relocations. */ + switch (*symbol_type) + { + case SYMBOL_ABSOLUTE: + case SYMBOL_TLS_LE: + return (int32_t) INTVAL (offset) == INTVAL (offset); + + default: + return false; + } + gcc_unreachable (); +} + +/* Returns the number of instructions necessary to reference a symbol. */ + +static int riscv_symbol_insns (enum riscv_symbol_type type) +{ + switch (type) + { + case SYMBOL_TLS: return 0; /* Depends on the TLS model. */ + case SYMBOL_ABSOLUTE: return 2; /* LUI + the reference itself */ + case SYMBOL_TLS_LE: return 3; /* LUI + ADD TP + the reference itself */ + case SYMBOL_GOT_DISP: return 3; /* AUIPC + LD GOT + the reference itself */ + default: gcc_unreachable(); + } +} + +/* A for_each_rtx callback. Stop the search if *X references a + thread-local symbol. */ + +static int +riscv_tls_symbol_ref_1 (rtx *x, void *data ATTRIBUTE_UNUSED) +{ + return riscv_tls_symbol_p (*x); +} + +/* Implement TARGET_LEGITIMATE_CONSTANT_P. */ + +static bool +riscv_legitimate_constant_p (enum machine_mode mode ATTRIBUTE_UNUSED, rtx x) +{ + return riscv_const_insns (x) > 0; +} + +/* Implement TARGET_CANNOT_FORCE_CONST_MEM. */ + +static bool +riscv_cannot_force_const_mem (enum machine_mode mode ATTRIBUTE_UNUSED, rtx x) +{ + enum riscv_symbol_type type; + rtx base, offset; + + /* There is no assembler syntax for expressing an address-sized + high part. */ + if (GET_CODE (x) == HIGH) + return true; + + /* As an optimization, reject constants that riscv_legitimize_move + can expand inline. + + Suppose we have a multi-instruction sequence that loads constant C + into register R. If R does not get allocated a hard register, and + R is used in an operand that allows both registers and memory + references, reload will consider forcing C into memory and using + one of the instruction's memory alternatives. Returning false + here will force it to use an input reload instead. */ + if (CONST_INT_P (x) && riscv_legitimate_constant_p (mode, x)) + return true; + + split_const (x, &base, &offset); + if (riscv_symbolic_constant_p (base, &type)) + { + /* The same optimization as for CONST_INT. */ + if (SMALL_INT (offset) && riscv_symbol_insns (type) > 0) + return true; + + /* It's not worth creating additional dynamic relocations. */ + if (flag_pic) + return true; + } + + /* TLS symbols must be computed by riscv_legitimize_move. */ + if (for_each_rtx (&x, &riscv_tls_symbol_ref_1, NULL)) + return true; + + return false; +} + +/* Return true if register REGNO is a valid base register for mode MODE. + STRICT_P is true if REG_OK_STRICT is in effect. */ + +int +riscv_regno_mode_ok_for_base_p (int regno, enum machine_mode mode ATTRIBUTE_UNUSED, + bool strict_p) +{ + if (!HARD_REGISTER_NUM_P (regno)) + { + if (!strict_p) + return true; + regno = reg_renumber[regno]; + } + + /* These fake registers will be eliminated to either the stack or + hard frame pointer, both of which are usually valid base registers. + Reload deals with the cases where the eliminated form isn't valid. */ + if (regno == ARG_POINTER_REGNUM || regno == FRAME_POINTER_REGNUM) + return true; + + return GP_REG_P (regno); +} + +/* Return true if X is a valid base register for mode MODE. + STRICT_P is true if REG_OK_STRICT is in effect. */ + +static bool +riscv_valid_base_register_p (rtx x, enum machine_mode mode, bool strict_p) +{ + if (!strict_p && GET_CODE (x) == SUBREG) + x = SUBREG_REG (x); + + return (REG_P (x) + && riscv_regno_mode_ok_for_base_p (REGNO (x), mode, strict_p)); +} + +/* Return true if, for every base register BASE_REG, (plus BASE_REG X) + can address a value of mode MODE. */ + +static bool +riscv_valid_offset_p (rtx x, enum machine_mode mode) +{ + /* Check that X is a signed 12-bit number. */ + if (!const_arith_operand (x, Pmode)) + return false; + + /* We may need to split multiword moves, so make sure that every word + is accessible. */ + if (GET_MODE_SIZE (mode) > UNITS_PER_WORD + && !SMALL_OPERAND (INTVAL (x) + GET_MODE_SIZE (mode) - UNITS_PER_WORD)) + return false; + + return true; +} + +/* Return true if a LO_SUM can address a value of mode MODE when the + LO_SUM symbol has type SYMBOL_TYPE. */ + +static bool +riscv_valid_lo_sum_p (enum riscv_symbol_type symbol_type, enum machine_mode mode) +{ + /* Check that symbols of type SYMBOL_TYPE can be used to access values + of mode MODE. */ + if (riscv_symbol_insns (symbol_type) == 0) + return false; + + /* Check that there is a known low-part relocation. */ + if (riscv_lo_relocs[symbol_type] == NULL) + return false; + + /* We may need to split multiword moves, so make sure that each word + can be accessed without inducing a carry. This is mainly needed + for o64, which has historically only guaranteed 64-bit alignment + for 128-bit types. */ + if (GET_MODE_SIZE (mode) > UNITS_PER_WORD + && GET_MODE_BITSIZE (mode) > GET_MODE_ALIGNMENT (mode)) + return false; + + return true; +} + +/* Return true if X is a valid address for machine mode MODE. If it is, + fill in INFO appropriately. STRICT_P is true if REG_OK_STRICT is in + effect. */ + +static bool +riscv_classify_address (struct riscv_address_info *info, rtx x, + enum machine_mode mode, bool strict_p) +{ + switch (GET_CODE (x)) + { + case REG: + case SUBREG: + info->type = ADDRESS_REG; + info->reg = x; + info->offset = const0_rtx; + return riscv_valid_base_register_p (info->reg, mode, strict_p); + + case PLUS: + info->type = ADDRESS_REG; + info->reg = XEXP (x, 0); + info->offset = XEXP (x, 1); + return (riscv_valid_base_register_p (info->reg, mode, strict_p) + && riscv_valid_offset_p (info->offset, mode)); + + case LO_SUM: + info->type = ADDRESS_LO_SUM; + info->reg = XEXP (x, 0); + info->offset = XEXP (x, 1); + /* We have to trust the creator of the LO_SUM to do something vaguely + sane. Target-independent code that creates a LO_SUM should also + create and verify the matching HIGH. Target-independent code that + adds an offset to a LO_SUM must prove that the offset will not + induce a carry. Failure to do either of these things would be + a bug, and we are not required to check for it here. The RISCV + backend itself should only create LO_SUMs for valid symbolic + constants, with the high part being either a HIGH or a copy + of _gp. */ + info->symbol_type + = riscv_classify_symbolic_expression (info->offset); + return (riscv_valid_base_register_p (info->reg, mode, strict_p) + && riscv_valid_lo_sum_p (info->symbol_type, mode)); + + case CONST_INT: + /* Small-integer addresses don't occur very often, but they + are legitimate if $0 is a valid base register. */ + info->type = ADDRESS_CONST_INT; + return SMALL_INT (x); + + default: + return false; + } +} + +/* Implement TARGET_LEGITIMATE_ADDRESS_P. */ + +static bool +riscv_legitimate_address_p (enum machine_mode mode, rtx x, bool strict_p) +{ + struct riscv_address_info addr; + + return riscv_classify_address (&addr, x, mode, strict_p); +} + +/* Return the number of instructions needed to load or store a value + of mode MODE at address X. Return 0 if X isn't valid for MODE. + Assume that multiword moves may need to be split into word moves + if MIGHT_SPLIT_P, otherwise assume that a single load or store is + enough. */ + +int +riscv_address_insns (rtx x, enum machine_mode mode, bool might_split_p) +{ + struct riscv_address_info addr; + int n = 1; + + if (!riscv_classify_address (&addr, x, mode, false)) + return 0; + + /* BLKmode is used for single unaligned loads and stores and should + not count as a multiword mode. */ + if (mode != BLKmode && might_split_p) + n += (GET_MODE_SIZE (mode) + UNITS_PER_WORD - 1) / UNITS_PER_WORD; + + if (addr.type == ADDRESS_LO_SUM) + n += riscv_symbol_insns (addr.symbol_type) - 1; + + return n; +} + +/* Return the number of instructions needed to load constant X. + Return 0 if X isn't a valid constant. */ + +int +riscv_const_insns (rtx x) +{ + enum riscv_symbol_type symbol_type; + rtx offset; + + switch (GET_CODE (x)) + { + case HIGH: + if (!riscv_symbolic_constant_p (XEXP (x, 0), &symbol_type) + || !riscv_hi_relocs[symbol_type]) + return 0; + + /* This is simply an LUI. */ + return 1; + + case CONST_INT: + { + int cost = riscv_integer_cost (INTVAL (x)); + /* Force complicated constants to memory. */ + return cost < 4 ? cost : 0; + } + + case CONST_DOUBLE: + case CONST_VECTOR: + /* Allow zeros for normal mode, where we can use x0. */ + return x == CONST0_RTX (GET_MODE (x)) ? 1 : 0; + + case CONST: + /* See if we can refer to X directly. */ + if (riscv_symbolic_constant_p (x, &symbol_type)) + return riscv_symbol_insns (symbol_type); + + /* Otherwise try splitting the constant into a base and offset. + If the offset is a 16-bit value, we can load the base address + into a register and then use (D)ADDIU to add in the offset. + If the offset is larger, we can load the base and offset + into separate registers and add them together with (D)ADDU. + However, the latter is only possible before reload; during + and after reload, we must have the option of forcing the + constant into the pool instead. */ + split_const (x, &x, &offset); + if (offset != 0) + { + int n = riscv_const_insns (x); + if (n != 0) + { + if (SMALL_INT (offset)) + return n + 1; + else if (!targetm.cannot_force_const_mem (GET_MODE (x), x)) + return n + 1 + riscv_integer_cost (INTVAL (offset)); + } + } + return 0; + + case SYMBOL_REF: + case LABEL_REF: + return riscv_symbol_insns (riscv_classify_symbol (x)); + + default: + return 0; + } +} + +/* X is a doubleword constant that can be handled by splitting it into + two words and loading each word separately. Return the number of + instructions required to do this. */ + +int +riscv_split_const_insns (rtx x) +{ + unsigned int low, high; + + low = riscv_const_insns (riscv_subword (x, false)); + high = riscv_const_insns (riscv_subword (x, true)); + gcc_assert (low > 0 && high > 0); + return low + high; +} + +/* Return the number of instructions needed to implement INSN, + given that it loads from or stores to MEM. */ + +int +riscv_load_store_insns (rtx mem, rtx insn) +{ + enum machine_mode mode; + bool might_split_p; + rtx set; + + gcc_assert (MEM_P (mem)); + mode = GET_MODE (mem); + + /* Try to prove that INSN does not need to be split. */ + might_split_p = true; + if (GET_MODE_BITSIZE (mode) == 64) + { + set = single_set (insn); + if (set && !riscv_split_64bit_move_p (SET_DEST (set), SET_SRC (set))) + might_split_p = false; + } + + return riscv_address_insns (XEXP (mem, 0), mode, might_split_p); +} + +/* Emit a move from SRC to DEST. Assume that the move expanders can + handle all moves if !can_create_pseudo_p (). The distinction is + important because, unlike emit_move_insn, the move expanders know + how to force Pmode objects into the constant pool even when the + constant pool address is not itself legitimate. */ + +rtx +riscv_emit_move (rtx dest, rtx src) +{ + return (can_create_pseudo_p () + ? emit_move_insn (dest, src) + : emit_move_insn_1 (dest, src)); +} + +/* Emit an instruction of the form (set TARGET (CODE OP0 OP1)). */ + +static void +riscv_emit_binary (enum rtx_code code, rtx target, rtx op0, rtx op1) +{ + emit_insn (gen_rtx_SET (VOIDmode, target, + gen_rtx_fmt_ee (code, GET_MODE (target), op0, op1))); +} + +/* Compute (CODE OP0 OP1) and store the result in a new register + of mode MODE. Return that new register. */ + +static rtx +riscv_force_binary (enum machine_mode mode, enum rtx_code code, rtx op0, rtx op1) +{ + rtx reg; + + reg = gen_reg_rtx (mode); + riscv_emit_binary (code, reg, op0, op1); + return reg; +} + +/* Copy VALUE to a register and return that register. If new pseudos + are allowed, copy it into a new register, otherwise use DEST. */ + +static rtx +riscv_force_temporary (rtx dest, rtx value) +{ + if (can_create_pseudo_p ()) + return force_reg (Pmode, value); + else + { + riscv_emit_move (dest, value); + return dest; + } +} + +/* Wrap symbol or label BASE in an UNSPEC address of type SYMBOL_TYPE, + then add CONST_INT OFFSET to the result. */ + +static rtx +riscv_unspec_address_offset (rtx base, rtx offset, + enum riscv_symbol_type symbol_type) +{ + base = gen_rtx_UNSPEC (Pmode, gen_rtvec (1, base), + UNSPEC_ADDRESS_FIRST + symbol_type); + if (offset != const0_rtx) + base = gen_rtx_PLUS (Pmode, base, offset); + return gen_rtx_CONST (Pmode, base); +} + +/* Return an UNSPEC address with underlying address ADDRESS and symbol + type SYMBOL_TYPE. */ + +rtx +riscv_unspec_address (rtx address, enum riscv_symbol_type symbol_type) +{ + rtx base, offset; + + split_const (address, &base, &offset); + return riscv_unspec_address_offset (base, offset, symbol_type); +} + +/* If OP is an UNSPEC address, return the address to which it refers, + otherwise return OP itself. */ + +static rtx +riscv_strip_unspec_address (rtx op) +{ + rtx base, offset; + + split_const (op, &base, &offset); + if (UNSPEC_ADDRESS_P (base)) + op = plus_constant (Pmode, UNSPEC_ADDRESS (base), INTVAL (offset)); + return op; +} + +/* If riscv_unspec_address (ADDR, SYMBOL_TYPE) is a 32-bit value, add the + high part to BASE and return the result. Just return BASE otherwise. + TEMP is as for riscv_force_temporary. + + The returned expression can be used as the first operand to a LO_SUM. */ + +static rtx +riscv_unspec_offset_high (rtx temp, rtx addr, enum riscv_symbol_type symbol_type) +{ + addr = gen_rtx_HIGH (Pmode, riscv_unspec_address (addr, symbol_type)); + return riscv_force_temporary (temp, addr); +} + +/* Load an entry from the GOT. */ +static rtx riscv_got_load_tls_gd(rtx dest, rtx sym) +{ + return (Pmode == DImode ? gen_got_load_tls_gddi(dest, sym) : gen_got_load_tls_gdsi(dest, sym)); +} + +static rtx riscv_got_load_tls_ie(rtx dest, rtx sym) +{ + return (Pmode == DImode ? gen_got_load_tls_iedi(dest, sym) : gen_got_load_tls_iesi(dest, sym)); +} + +static rtx riscv_tls_add_tp_le(rtx dest, rtx base, rtx sym) +{ + rtx tp = gen_rtx_REG (Pmode, THREAD_POINTER_REGNUM); + return (Pmode == DImode ? gen_tls_add_tp_ledi(dest, base, tp, sym) : gen_tls_add_tp_lesi(dest, base, tp, sym)); +} + +/* If MODE is MAX_MACHINE_MODE, ADDR appears as a move operand, otherwise + it appears in a MEM of that mode. Return true if ADDR is a legitimate + constant in that context and can be split into high and low parts. + If so, and if LOW_OUT is nonnull, emit the high part and store the + low part in *LOW_OUT. Leave *LOW_OUT unchanged otherwise. + + TEMP is as for riscv_force_temporary and is used to load the high + part into a register. + + When MODE is MAX_MACHINE_MODE, the low part is guaranteed to be + a legitimize SET_SRC for an .md pattern, otherwise the low part + is guaranteed to be a legitimate address for mode MODE. */ + +bool +riscv_split_symbol (rtx temp, rtx addr, enum machine_mode mode, rtx *low_out) +{ + enum riscv_symbol_type symbol_type; + rtx high; + + if ((GET_CODE (addr) == HIGH && mode == MAX_MACHINE_MODE) + || !riscv_symbolic_constant_p (addr, &symbol_type) + || riscv_symbol_insns (symbol_type) == 0 + || !riscv_hi_relocs[symbol_type]) + return false; + + if (low_out) + { + switch (symbol_type) + { + case SYMBOL_ABSOLUTE: + high = gen_rtx_HIGH (Pmode, copy_rtx (addr)); + high = riscv_force_temporary (temp, high); + *low_out = gen_rtx_LO_SUM (Pmode, high, addr); + break; + + default: + gcc_unreachable (); + } + } + + return true; +} + +/* Return a legitimate address for REG + OFFSET. TEMP is as for + riscv_force_temporary; it is only needed when OFFSET is not a + SMALL_OPERAND. */ + +static rtx +riscv_add_offset (rtx temp, rtx reg, HOST_WIDE_INT offset) +{ + if (!SMALL_OPERAND (offset)) + { + rtx high; + + /* Leave OFFSET as a 16-bit offset and put the excess in HIGH. + The addition inside the macro CONST_HIGH_PART may cause an + overflow, so we need to force a sign-extension check. */ + high = gen_int_mode (RISCV_CONST_HIGH_PART (offset), Pmode); + offset = RISCV_CONST_LOW_PART (offset); + high = riscv_force_temporary (temp, high); + reg = riscv_force_temporary (temp, gen_rtx_PLUS (Pmode, high, reg)); + } + return plus_constant (Pmode, reg, offset); +} + +/* The __tls_get_attr symbol. */ +static GTY(()) rtx riscv_tls_symbol; + +/* Return an instruction sequence that calls __tls_get_addr. SYM is + the TLS symbol we are referencing and TYPE is the symbol type to use + (either global dynamic or local dynamic). RESULT is an RTX for the + return value location. */ + +static rtx +riscv_call_tls_get_addr (rtx sym, rtx result) +{ + rtx insn, a0 = gen_rtx_REG (Pmode, GP_ARG_FIRST); + + if (!riscv_tls_symbol) + riscv_tls_symbol = init_one_libfunc ("__tls_get_addr"); + + start_sequence (); + + emit_insn (riscv_got_load_tls_gd (a0, sym)); + insn = riscv_expand_call (false, result, riscv_tls_symbol, const0_rtx); + RTL_CONST_CALL_P (insn) = 1; + use_reg (&CALL_INSN_FUNCTION_USAGE (insn), a0); + insn = get_insns (); + + end_sequence (); + + return insn; +} + +/* Generate the code to access LOC, a thread-local SYMBOL_REF, and return + its address. The return value will be both a valid address and a valid + SET_SRC (either a REG or a LO_SUM). */ + +static rtx +riscv_legitimize_tls_address (rtx loc) +{ + rtx dest, insn, tp, tmp1; + enum tls_model model = SYMBOL_REF_TLS_MODEL (loc); + + /* Since we support TLS copy relocs, non-PIC TLS accesses may all use LE. */ + if (!flag_pic) + model = TLS_MODEL_LOCAL_EXEC; + + switch (model) + { + case TLS_MODEL_LOCAL_DYNAMIC: + /* Rely on section anchors for the optimization that LDM TLS + provides. The anchor's address is loaded with GD TLS. */ + case TLS_MODEL_GLOBAL_DYNAMIC: + tmp1 = gen_rtx_REG (Pmode, GP_RETURN); + insn = riscv_call_tls_get_addr (loc, tmp1); + dest = gen_reg_rtx (Pmode); + emit_libcall_block (insn, dest, tmp1, loc); + break; + + case TLS_MODEL_INITIAL_EXEC: + /* la.tls.ie; tp-relative add */ + tp = gen_rtx_REG (Pmode, THREAD_POINTER_REGNUM); + tmp1 = gen_reg_rtx (Pmode); + emit_insn (riscv_got_load_tls_ie (tmp1, loc)); + dest = gen_reg_rtx (Pmode); + emit_insn (gen_add3_insn (dest, tmp1, tp)); + break; + + case TLS_MODEL_LOCAL_EXEC: + tmp1 = riscv_unspec_offset_high (NULL, loc, SYMBOL_TLS_LE); + dest = gen_reg_rtx (Pmode); + emit_insn (riscv_tls_add_tp_le (dest, tmp1, loc)); + dest = gen_rtx_LO_SUM (Pmode, dest, + riscv_unspec_address (loc, SYMBOL_TLS_LE)); + break; + + default: + gcc_unreachable (); + } + return dest; +} + +/* If X is not a valid address for mode MODE, force it into a register. */ + +static rtx +riscv_force_address (rtx x, enum machine_mode mode) +{ + if (!riscv_legitimate_address_p (mode, x, false)) + x = force_reg (Pmode, x); + return x; +} + +/* This function is used to implement LEGITIMIZE_ADDRESS. If X can + be legitimized in a way that the generic machinery might not expect, + return a new address, otherwise return NULL. MODE is the mode of + the memory being accessed. */ + +static rtx +riscv_legitimize_address (rtx x, rtx oldx ATTRIBUTE_UNUSED, + enum machine_mode mode) +{ + rtx addr; + + if (riscv_tls_symbol_p (x)) + return riscv_legitimize_tls_address (x); + + /* See if the address can split into a high part and a LO_SUM. */ + if (riscv_split_symbol (NULL, x, mode, &addr)) + return riscv_force_address (addr, mode); + + /* Handle BASE + OFFSET using riscv_add_offset. */ + if (GET_CODE (x) == PLUS && CONST_INT_P (XEXP (x, 1)) + && INTVAL (XEXP (x, 1)) != 0) + { + rtx base = XEXP (x, 0); + HOST_WIDE_INT offset = INTVAL (XEXP (x, 1)); + + if (!riscv_valid_base_register_p (base, mode, false)) + base = copy_to_mode_reg (Pmode, base); + addr = riscv_add_offset (NULL, base, offset); + return riscv_force_address (addr, mode); + } + + return x; +} + +/* Load VALUE into DEST. TEMP is as for riscv_force_temporary. */ + +void +riscv_move_integer (rtx temp, rtx dest, HOST_WIDE_INT value) +{ + struct riscv_integer_op codes[RISCV_MAX_INTEGER_OPS]; + enum machine_mode mode; + int i, num_ops; + rtx x; + + mode = GET_MODE (dest); + num_ops = riscv_build_integer (codes, value, mode); + + if (can_create_pseudo_p () && num_ops > 2 /* not a simple constant */ + && num_ops >= riscv_split_integer_cost (value)) + x = riscv_split_integer (value, mode); + else + { + /* Apply each binary operation to X. */ + x = GEN_INT (codes[0].value); + + for (i = 1; i < num_ops; i++) + { + if (!can_create_pseudo_p ()) + { + emit_insn (gen_rtx_SET (VOIDmode, temp, x)); + x = temp; + } + else + x = force_reg (mode, x); + + x = gen_rtx_fmt_ee (codes[i].code, mode, x, GEN_INT (codes[i].value)); + } + } + + emit_insn (gen_rtx_SET (VOIDmode, dest, x)); +} + +/* Subroutine of riscv_legitimize_move. Move constant SRC into register + DEST given that SRC satisfies immediate_operand but doesn't satisfy + move_operand. */ + +static void +riscv_legitimize_const_move (enum machine_mode mode, rtx dest, rtx src) +{ + rtx base, offset; + + /* Split moves of big integers into smaller pieces. */ + if (splittable_const_int_operand (src, mode)) + { + riscv_move_integer (dest, dest, INTVAL (src)); + return; + } + + /* Split moves of symbolic constants into high/low pairs. */ + if (riscv_split_symbol (dest, src, MAX_MACHINE_MODE, &src)) + { + emit_insn (gen_rtx_SET (VOIDmode, dest, src)); + return; + } + + /* Generate the appropriate access sequences for TLS symbols. */ + if (riscv_tls_symbol_p (src)) + { + riscv_emit_move (dest, riscv_legitimize_tls_address (src)); + return; + } + + /* If we have (const (plus symbol offset)), and that expression cannot + be forced into memory, load the symbol first and add in the offset. Also + prefer to do this even if the constant _can_ be forced into memory, as it + usually produces better code. */ + split_const (src, &base, &offset); + if (offset != const0_rtx + && (targetm.cannot_force_const_mem (mode, src) || can_create_pseudo_p ())) + { + base = riscv_force_temporary (dest, base); + riscv_emit_move (dest, riscv_add_offset (NULL, base, INTVAL (offset))); + return; + } + + src = force_const_mem (mode, src); + + /* When using explicit relocs, constant pool references are sometimes + not legitimate addresses. */ + riscv_split_symbol (dest, XEXP (src, 0), mode, &XEXP (src, 0)); + riscv_emit_move (dest, src); +} + +/* If (set DEST SRC) is not a valid move instruction, emit an equivalent + sequence that is valid. */ + +bool +riscv_legitimize_move (enum machine_mode mode, rtx dest, rtx src) +{ + if (!register_operand (dest, mode) && !reg_or_0_operand (src, mode)) + { + riscv_emit_move (dest, force_reg (mode, src)); + return true; + } + + /* We need to deal with constants that would be legitimate + immediate_operands but aren't legitimate move_operands. */ + if (CONSTANT_P (src) && !move_operand (src, mode)) + { + riscv_legitimize_const_move (mode, dest, src); + set_unique_reg_note (get_last_insn (), REG_EQUAL, copy_rtx (src)); + return true; + } + return false; +} + +/* Return true if there is an instruction that implements CODE and accepts + X as an immediate operand. */ + +static int +riscv_immediate_operand_p (int code, HOST_WIDE_INT x) +{ + switch (code) + { + case ASHIFT: + case ASHIFTRT: + case LSHIFTRT: + /* All shift counts are truncated to a valid constant. */ + return true; + + case AND: + case IOR: + case XOR: + case PLUS: + case LT: + case LTU: + /* These instructions take 12-bit signed immediates. */ + return SMALL_OPERAND (x); + + case LE: + /* We add 1 to the immediate and use SLT. */ + return SMALL_OPERAND (x + 1); + + case LEU: + /* Likewise SLTU, but reject the always-true case. */ + return SMALL_OPERAND (x + 1) && x + 1 != 0; + + case GE: + case GEU: + /* We can emulate an immediate of 1 by using GT/GTU against x0. */ + return x == 1; + + default: + /* By default assume that x0 can be used for 0. */ + return x == 0; + } +} + +/* Return the cost of binary operation X, given that the instruction + sequence for a word-sized or smaller operation takes SIGNLE_INSNS + instructions and that the sequence of a double-word operation takes + DOUBLE_INSNS instructions. */ + +static int +riscv_binary_cost (rtx x, int single_insns, int double_insns) +{ + if (GET_MODE_SIZE (GET_MODE (x)) == UNITS_PER_WORD * 2) + return COSTS_N_INSNS (double_insns); + return COSTS_N_INSNS (single_insns); +} + +/* Return the cost of sign-extending OP to mode MODE, not including the + cost of OP itself. */ + +static int +riscv_sign_extend_cost (enum machine_mode mode, rtx op) +{ + if (MEM_P (op)) + /* Extended loads are as cheap as unextended ones. */ + return 0; + + if (TARGET_64BIT && mode == DImode && GET_MODE (op) == SImode) + /* A sign extension from SImode to DImode in 64-bit mode is free. */ + return 0; + + /* We need to use a shift left and a shift right. */ + return COSTS_N_INSNS (2); +} + +/* Return the cost of zero-extending OP to mode MODE, not including the + cost of OP itself. */ + +static int +riscv_zero_extend_cost (enum machine_mode mode, rtx op) +{ + if (MEM_P (op)) + /* Extended loads are as cheap as unextended ones. */ + return 0; + + if ((TARGET_64BIT && mode == DImode && GET_MODE (op) == SImode) || + ((mode == DImode || mode == SImode) && GET_MODE (op) == HImode)) + /* We need a shift left by 32 bits and a shift right by 32 bits. */ + return COSTS_N_INSNS (2); + + /* We can use ANDI. */ + return COSTS_N_INSNS (1); +} + +/* Implement TARGET_RTX_COSTS. */ + +static bool +riscv_rtx_costs (rtx x, int code, int outer_code, int opno ATTRIBUTE_UNUSED, + int *total, bool speed) +{ + enum machine_mode mode = GET_MODE (x); + bool float_mode_p = FLOAT_MODE_P (mode); + int cost; + + switch (code) + { + case CONST_INT: + if (riscv_immediate_operand_p (outer_code, INTVAL (x))) + { + *total = 0; + return true; + } + /* Fall through. */ + + case SYMBOL_REF: + case LABEL_REF: + case CONST_DOUBLE: + case CONST: + if (speed) + *total = 1; + else if ((cost = riscv_const_insns (x)) > 0) + *total = COSTS_N_INSNS (cost); + else /* The instruction will be fetched from the constant pool. */ + *total = COSTS_N_INSNS (riscv_symbol_insns (SYMBOL_ABSOLUTE)); + return true; + + case MEM: + /* If the address is legitimate, return the number of + instructions it needs. */ + if ((cost = riscv_address_insns (XEXP (x, 0), mode, true)) > 0) + { + *total = COSTS_N_INSNS (cost + tune_info->memory_cost); + return true; + } + /* Otherwise use the default handling. */ + return false; + + case NOT: + *total = COSTS_N_INSNS (GET_MODE_SIZE (mode) > UNITS_PER_WORD ? 2 : 1); + return false; + + case AND: + case IOR: + case XOR: + /* Double-word operations use two single-word operations. */ + *total = riscv_binary_cost (x, 1, 2); + return false; + + case ASHIFT: + case ASHIFTRT: + case LSHIFTRT: + *total = riscv_binary_cost (x, 1, CONSTANT_P (XEXP (x, 1)) ? 4 : 9); + return false; + + case ABS: + *total = COSTS_N_INSNS (float_mode_p ? 1 : 3); + return false; + + case LO_SUM: + *total = set_src_cost (XEXP (x, 0), speed); + return true; + + case LT: + case LTU: + case LE: + case LEU: + case GT: + case GTU: + case GE: + case GEU: + case EQ: + case NE: + case UNORDERED: + case LTGT: + /* Branch comparisons have VOIDmode, so use the first operand's + mode instead. */ + mode = GET_MODE (XEXP (x, 0)); + if (float_mode_p) + *total = tune_info->fp_add[mode == DFmode]; + else + *total = riscv_binary_cost (x, 1, 3); + return false; + + case MINUS: + if (float_mode_p + && !HONOR_NANS (mode) + && !HONOR_SIGNED_ZEROS (mode)) + { + /* See if we can use NMADD or NMSUB. See riscv.md for the + associated patterns. */ + rtx op0 = XEXP (x, 0); + rtx op1 = XEXP (x, 1); + if (GET_CODE (op0) == MULT && GET_CODE (XEXP (op0, 0)) == NEG) + { + *total = (tune_info->fp_mul[mode == DFmode] + + set_src_cost (XEXP (XEXP (op0, 0), 0), speed) + + set_src_cost (XEXP (op0, 1), speed) + + set_src_cost (op1, speed)); + return true; + } + if (GET_CODE (op1) == MULT) + { + *total = (tune_info->fp_mul[mode == DFmode] + + set_src_cost (op0, speed) + + set_src_cost (XEXP (op1, 0), speed) + + set_src_cost (XEXP (op1, 1), speed)); + return true; + } + } + /* Fall through. */ + + case PLUS: + if (float_mode_p) + *total = tune_info->fp_add[mode == DFmode]; + else + *total = riscv_binary_cost (x, 1, 4); + return false; + + case NEG: + if (float_mode_p + && !HONOR_NANS (mode) + && HONOR_SIGNED_ZEROS (mode)) + { + /* See if we can use NMADD or NMSUB. See riscv.md for the + associated patterns. */ + rtx op = XEXP (x, 0); + if ((GET_CODE (op) == PLUS || GET_CODE (op) == MINUS) + && GET_CODE (XEXP (op, 0)) == MULT) + { + *total = (tune_info->fp_mul[mode == DFmode] + + set_src_cost (XEXP (XEXP (op, 0), 0), speed) + + set_src_cost (XEXP (XEXP (op, 0), 1), speed) + + set_src_cost (XEXP (op, 1), speed)); + return true; + } + } + + if (float_mode_p) + *total = tune_info->fp_add[mode == DFmode]; + else + *total = COSTS_N_INSNS (GET_MODE_SIZE (mode) > UNITS_PER_WORD ? 4 : 1); + return false; + + case MULT: + if (float_mode_p) + *total = tune_info->fp_mul[mode == DFmode]; + else if (GET_MODE_SIZE (mode) > UNITS_PER_WORD) + *total = 3 * tune_info->int_mul[0] + COSTS_N_INSNS (2); + else if (!speed) + *total = COSTS_N_INSNS (1); + else + *total = tune_info->int_mul[mode == DImode]; + return false; + + case DIV: + case SQRT: + case MOD: + if (float_mode_p) + { + *total = tune_info->fp_div[mode == DFmode]; + return false; + } + /* Fall through. */ + + case UDIV: + case UMOD: + if (speed) + *total = tune_info->int_div[mode == DImode]; + else + *total = COSTS_N_INSNS (1); + return false; + + case SIGN_EXTEND: + *total = riscv_sign_extend_cost (mode, XEXP (x, 0)); + return false; + + case ZERO_EXTEND: + *total = riscv_zero_extend_cost (mode, XEXP (x, 0)); + return false; + + case FLOAT: + case UNSIGNED_FLOAT: + case FIX: + case FLOAT_EXTEND: + case FLOAT_TRUNCATE: + *total = tune_info->fp_add[mode == DFmode]; + return false; + + default: + return false; + } +} + +/* Implement TARGET_ADDRESS_COST. */ + +static int +riscv_address_cost (rtx addr, enum machine_mode mode, + addr_space_t as ATTRIBUTE_UNUSED, + bool speed ATTRIBUTE_UNUSED) +{ + return riscv_address_insns (addr, mode, false); +} + +/* Return one word of double-word value OP. HIGH_P is true to select the + high part or false to select the low part. */ + +rtx +riscv_subword (rtx op, bool high_p) +{ + unsigned int byte; + enum machine_mode mode; + + mode = GET_MODE (op); + if (mode == VOIDmode) + mode = TARGET_64BIT ? TImode : DImode; + + byte = high_p ? UNITS_PER_WORD : 0; + + if (FP_REG_RTX_P (op)) + return gen_rtx_REG (word_mode, REGNO (op) + high_p); + + if (MEM_P (op)) + return adjust_address (op, word_mode, byte); + + return simplify_gen_subreg (word_mode, op, mode, byte); +} + +/* Return true if a 64-bit move from SRC to DEST should be split into two. */ + +bool +riscv_split_64bit_move_p (rtx dest, rtx src) +{ + /* All 64b moves are legal in 64b mode. All 64b FPR <-> FPR and + FPR <-> MEM moves are legal in 32b mode, too. Although + FPR <-> GPR moves are not available in general in 32b mode, + we can at least load 0 into an FPR with fcvt.d.w fpr, x0. */ + return !(TARGET_64BIT + || (FP_REG_RTX_P (src) && FP_REG_RTX_P (dest)) + || (FP_REG_RTX_P (dest) && MEM_P (src)) + || (FP_REG_RTX_P (src) && MEM_P (dest)) + || (FP_REG_RTX_P(dest) && src == CONST0_RTX(GET_MODE(src)))); +} + +/* Split a doubleword move from SRC to DEST. On 32-bit targets, + this function handles 64-bit moves for which riscv_split_64bit_move_p + holds. For 64-bit targets, this function handles 128-bit moves. */ + +void +riscv_split_doubleword_move (rtx dest, rtx src) +{ + rtx low_dest; + + /* The operation can be split into two normal moves. Decide in + which order to do them. */ + low_dest = riscv_subword (dest, false); + if (REG_P (low_dest) && reg_overlap_mentioned_p (low_dest, src)) + { + riscv_emit_move (riscv_subword (dest, true), riscv_subword (src, true)); + riscv_emit_move (low_dest, riscv_subword (src, false)); + } + else + { + riscv_emit_move (low_dest, riscv_subword (src, false)); + riscv_emit_move (riscv_subword (dest, true), riscv_subword (src, true)); + } +} + +/* Return the appropriate instructions to move SRC into DEST. Assume + that SRC is operand 1 and DEST is operand 0. */ + +const char * +riscv_output_move (rtx dest, rtx src) +{ + enum rtx_code dest_code, src_code; + enum machine_mode mode; + bool dbl_p; + + dest_code = GET_CODE (dest); + src_code = GET_CODE (src); + mode = GET_MODE (dest); + dbl_p = (GET_MODE_SIZE (mode) == 8); + + if (dbl_p && riscv_split_64bit_move_p (dest, src)) + return "#"; + + if ((src_code == REG && GP_REG_P (REGNO (src))) + || (src == CONST0_RTX (mode))) + { + if (dest_code == REG) + { + if (GP_REG_P (REGNO (dest))) + return "mv\t%0,%z1"; + + if (FP_REG_P (REGNO (dest))) + { + if (!dbl_p) + return "fmv.s.x\t%0,%z1"; + if (TARGET_64BIT) + return "fmv.d.x\t%0,%z1"; + /* in RV32, we can emulate fmv.d.x %0, x0 using fcvt.d.w */ + gcc_assert (src == CONST0_RTX (mode)); + return "fcvt.d.w\t%0,x0"; + } + } + if (dest_code == MEM) + switch (GET_MODE_SIZE (mode)) + { + case 1: return "sb\t%z1,%0"; + case 2: return "sh\t%z1,%0"; + case 4: return "sw\t%z1,%0"; + case 8: return "sd\t%z1,%0"; + } + } + if (dest_code == REG && GP_REG_P (REGNO (dest))) + { + if (src_code == REG) + { + if (FP_REG_P (REGNO (src))) + return dbl_p ? "fmv.x.d\t%0,%1" : "fmv.x.s\t%0,%1"; + } + + if (src_code == MEM) + switch (GET_MODE_SIZE (mode)) + { + case 1: return "lbu\t%0,%1"; + case 2: return "lhu\t%0,%1"; + case 4: return "lw\t%0,%1"; + case 8: return "ld\t%0,%1"; + } + + if (src_code == CONST_INT) + return "li\t%0,%1"; + + if (src_code == HIGH) + return "lui\t%0,%h1"; + + if (symbolic_operand (src, VOIDmode)) + switch (riscv_classify_symbolic_expression (src)) + { + case SYMBOL_GOT_DISP: return "la\t%0,%1"; + case SYMBOL_ABSOLUTE: return "lla\t%0,%1"; + default: gcc_unreachable(); + } + } + if (src_code == REG && FP_REG_P (REGNO (src))) + { + if (dest_code == REG && FP_REG_P (REGNO (dest))) + return dbl_p ? "fmv.d\t%0,%1" : "fmv.s\t%0,%1"; + + if (dest_code == MEM) + return dbl_p ? "fsd\t%1,%0" : "fsw\t%1,%0"; + } + if (dest_code == REG && FP_REG_P (REGNO (dest))) + { + if (src_code == MEM) + return dbl_p ? "fld\t%0,%1" : "flw\t%0,%1"; + } + gcc_unreachable (); +} + +/* Return true if CMP1 is a suitable second operand for integer ordering + test CODE. See also the *sCC patterns in riscv.md. */ + +static bool +riscv_int_order_operand_ok_p (enum rtx_code code, rtx cmp1) +{ + switch (code) + { + case GT: + case GTU: + return reg_or_0_operand (cmp1, VOIDmode); + + case GE: + case GEU: + return cmp1 == const1_rtx; + + case LT: + case LTU: + return arith_operand (cmp1, VOIDmode); + + case LE: + return sle_operand (cmp1, VOIDmode); + + case LEU: + return sleu_operand (cmp1, VOIDmode); + + default: + gcc_unreachable (); + } +} + +/* Return true if *CMP1 (of mode MODE) is a valid second operand for + integer ordering test *CODE, or if an equivalent combination can + be formed by adjusting *CODE and *CMP1. When returning true, update + *CODE and *CMP1 with the chosen code and operand, otherwise leave + them alone. */ + +static bool +riscv_canonicalize_int_order_test (enum rtx_code *code, rtx *cmp1, + enum machine_mode mode) +{ + HOST_WIDE_INT plus_one; + + if (riscv_int_order_operand_ok_p (*code, *cmp1)) + return true; + + if (CONST_INT_P (*cmp1)) + switch (*code) + { + case LE: + plus_one = trunc_int_for_mode (UINTVAL (*cmp1) + 1, mode); + if (INTVAL (*cmp1) < plus_one) + { + *code = LT; + *cmp1 = force_reg (mode, GEN_INT (plus_one)); + return true; + } + break; + + case LEU: + plus_one = trunc_int_for_mode (UINTVAL (*cmp1) + 1, mode); + if (plus_one != 0) + { + *code = LTU; + *cmp1 = force_reg (mode, GEN_INT (plus_one)); + return true; + } + break; + + default: + break; + } + return false; +} + +/* Compare CMP0 and CMP1 using ordering test CODE and store the result + in TARGET. CMP0 and TARGET are register_operands. If INVERT_PTR + is nonnull, it's OK to set TARGET to the inverse of the result and + flip *INVERT_PTR instead. */ + +static void +riscv_emit_int_order_test (enum rtx_code code, bool *invert_ptr, + rtx target, rtx cmp0, rtx cmp1) +{ + enum machine_mode mode; + + /* First see if there is a RISCV instruction that can do this operation. + If not, try doing the same for the inverse operation. If that also + fails, force CMP1 into a register and try again. */ + mode = GET_MODE (cmp0); + if (riscv_canonicalize_int_order_test (&code, &cmp1, mode)) + riscv_emit_binary (code, target, cmp0, cmp1); + else + { + enum rtx_code inv_code = reverse_condition (code); + if (!riscv_canonicalize_int_order_test (&inv_code, &cmp1, mode)) + { + cmp1 = force_reg (mode, cmp1); + riscv_emit_int_order_test (code, invert_ptr, target, cmp0, cmp1); + } + else if (invert_ptr == 0) + { + rtx inv_target; + + inv_target = riscv_force_binary (GET_MODE (target), + inv_code, cmp0, cmp1); + riscv_emit_binary (XOR, target, inv_target, const1_rtx); + } + else + { + *invert_ptr = !*invert_ptr; + riscv_emit_binary (inv_code, target, cmp0, cmp1); + } + } +} + +/* Return a register that is zero iff CMP0 and CMP1 are equal. + The register will have the same mode as CMP0. */ + +static rtx +riscv_zero_if_equal (rtx cmp0, rtx cmp1) +{ + if (cmp1 == const0_rtx) + return cmp0; + + return expand_binop (GET_MODE (cmp0), sub_optab, + cmp0, cmp1, 0, 0, OPTAB_DIRECT); +} + +/* Return false if we can easily emit code for the FP comparison specified + by *CODE. If not, set *CODE to its inverse and return true. */ + +static bool +riscv_reversed_fp_cond (enum rtx_code *code) +{ + switch (*code) + { + case EQ: + case LT: + case LE: + case GT: + case GE: + case LTGT: + case ORDERED: + /* We know how to emit code for these cases... */ + return false; + + default: + /* ...but we must invert these and rely on the others. */ + *code = reverse_condition_maybe_unordered (*code); + return true; + } +} + +/* Convert a comparison into something that can be used in a branch or + conditional move. On entry, *OP0 and *OP1 are the values being + compared and *CODE is the code used to compare them. + + Update *CODE, *OP0 and *OP1 so that they describe the final comparison. */ + +static void +riscv_emit_compare (enum rtx_code *code, rtx *op0, rtx *op1) +{ + rtx cmp_op0 = *op0; + rtx cmp_op1 = *op1; + + if (GET_MODE_CLASS (GET_MODE (*op0)) == MODE_INT) + { + if (splittable_const_int_operand (cmp_op1, VOIDmode)) + { + HOST_WIDE_INT rhs = INTVAL (cmp_op1), new_rhs; + enum rtx_code new_code; + + switch (*code) + { + case LTU: new_rhs = rhs - 1; new_code = LEU; goto try_new_rhs; + case LEU: new_rhs = rhs + 1; new_code = LTU; goto try_new_rhs; + case GTU: new_rhs = rhs + 1; new_code = GEU; goto try_new_rhs; + case GEU: new_rhs = rhs - 1; new_code = GTU; goto try_new_rhs; + case LT: new_rhs = rhs - 1; new_code = LE; goto try_new_rhs; + case LE: new_rhs = rhs + 1; new_code = LT; goto try_new_rhs; + case GT: new_rhs = rhs + 1; new_code = GE; goto try_new_rhs; + case GE: new_rhs = rhs - 1; new_code = GT; + try_new_rhs: + /* Convert e.g. OP0 > 4095 into OP0 >= 4096. */ + if ((rhs < 0) == (new_rhs < 0) + && riscv_integer_cost (new_rhs) < riscv_integer_cost (rhs)) + { + *op1 = GEN_INT (new_rhs); + *code = new_code; + } + break; + + case EQ: + case NE: + /* Convert e.g. OP0 == 2048 into OP0 - 2048 == 0. */ + if (SMALL_OPERAND (-rhs)) + { + *op0 = gen_reg_rtx (GET_MODE (cmp_op0)); + riscv_emit_binary (PLUS, *op0, cmp_op0, GEN_INT (-rhs)); + *op1 = const0_rtx; + } + default: + break; + } + } + + if (*op1 != const0_rtx) + *op1 = force_reg (GET_MODE (cmp_op0), *op1); + } + else + { + /* For FP comparisons, set an integer register with the result of the + comparison, then branch on it. */ + rtx tmp0, tmp1, final_op; + enum rtx_code fp_code = *code; + *code = riscv_reversed_fp_cond (&fp_code) ? EQ : NE; + + switch (fp_code) + { + case ORDERED: + /* a == a && b == b */ + tmp0 = gen_reg_rtx (SImode); + riscv_emit_binary (EQ, tmp0, cmp_op0, cmp_op0); + tmp1 = gen_reg_rtx (SImode); + riscv_emit_binary (EQ, tmp1, cmp_op1, cmp_op1); + final_op = gen_reg_rtx (SImode); + riscv_emit_binary (AND, final_op, tmp0, tmp1); + break; + + case LTGT: + /* a < b || a > b */ + tmp0 = gen_reg_rtx (SImode); + riscv_emit_binary (LT, tmp0, cmp_op0, cmp_op1); + tmp1 = gen_reg_rtx (SImode); + riscv_emit_binary (GT, tmp1, cmp_op0, cmp_op1); + final_op = gen_reg_rtx (SImode); + riscv_emit_binary (IOR, final_op, tmp0, tmp1); + break; + + case EQ: + case LE: + case LT: + case GE: + case GT: + /* We have instructions for these cases. */ + final_op = gen_reg_rtx (SImode); + riscv_emit_binary (fp_code, final_op, cmp_op0, cmp_op1); + break; + + default: + gcc_unreachable (); + } + + /* Compare the binary result against 0. */ + *op0 = final_op; + *op1 = const0_rtx; + } +} + +/* Try performing the comparison in OPERANDS[1], whose arms are OPERANDS[2] + and OPERAND[3]. Store the result in OPERANDS[0]. + + On 64-bit targets, the mode of the comparison and target will always be + SImode, thus possibly narrower than that of the comparison's operands. */ + +void +riscv_expand_scc (rtx operands[]) +{ + rtx target = operands[0]; + enum rtx_code code = GET_CODE (operands[1]); + rtx op0 = operands[2]; + rtx op1 = operands[3]; + + gcc_assert (GET_MODE_CLASS (GET_MODE (op0)) == MODE_INT); + + if (code == EQ || code == NE) + { + rtx zie = riscv_zero_if_equal (op0, op1); + riscv_emit_binary (code, target, zie, const0_rtx); + } + else + riscv_emit_int_order_test (code, 0, target, op0, op1); +} + +/* Compare OPERANDS[1] with OPERANDS[2] using comparison code + CODE and jump to OPERANDS[3] if the condition holds. */ + +void +riscv_expand_conditional_branch (rtx *operands) +{ + enum rtx_code code = GET_CODE (operands[0]); + rtx op0 = operands[1]; + rtx op1 = operands[2]; + rtx condition; + + riscv_emit_compare (&code, &op0, &op1); + condition = gen_rtx_fmt_ee (code, VOIDmode, op0, op1); + emit_jump_insn (gen_condjump (condition, operands[3])); +} + +/* Implement TARGET_FUNCTION_ARG_BOUNDARY. Every parameter gets at + least PARM_BOUNDARY bits of alignment, but will be given anything up + to STACK_BOUNDARY bits if the type requires it. */ + +static unsigned int +riscv_function_arg_boundary (enum machine_mode mode, const_tree type) +{ + unsigned int alignment; + + alignment = type ? TYPE_ALIGN (type) : GET_MODE_ALIGNMENT (mode); + if (alignment < PARM_BOUNDARY) + alignment = PARM_BOUNDARY; + if (alignment > STACK_BOUNDARY) + alignment = STACK_BOUNDARY; + return alignment; +} + +/* Fill INFO with information about a single argument. CUM is the + cumulative state for earlier arguments. MODE is the mode of this + argument and TYPE is its type (if known). NAMED is true if this + is a named (fixed) argument rather than a variable one. */ + +static void +riscv_get_arg_info (struct riscv_arg_info *info, const CUMULATIVE_ARGS *cum, + enum machine_mode mode, const_tree type, bool named) +{ + bool doubleword_aligned_p; + unsigned int num_bytes, num_words, max_regs; + + /* Work out the size of the argument. */ + num_bytes = type ? int_size_in_bytes (type) : GET_MODE_SIZE (mode); + num_words = (num_bytes + UNITS_PER_WORD - 1) / UNITS_PER_WORD; + + /* Scalar, complex and vector floating-point types are passed in + floating-point registers, as long as this is a named rather + than a variable argument. */ + info->fpr_p = (named + && (type == 0 || FLOAT_TYPE_P (type)) + && (GET_MODE_CLASS (mode) == MODE_FLOAT + || GET_MODE_CLASS (mode) == MODE_COMPLEX_FLOAT + || GET_MODE_CLASS (mode) == MODE_VECTOR_FLOAT) + && GET_MODE_UNIT_SIZE (mode) <= UNITS_PER_FPVALUE); + + /* Complex floats should only go into FPRs if there are two FPRs free, + otherwise they should be passed in the same way as a struct + containing two floats. */ + if (info->fpr_p + && GET_MODE_CLASS (mode) == MODE_COMPLEX_FLOAT + && GET_MODE_UNIT_SIZE (mode) < UNITS_PER_FPVALUE) + { + if (cum->num_gprs >= MAX_ARGS_IN_REGISTERS - 1) + info->fpr_p = false; + else + num_words = 2; + } + + /* See whether the argument has doubleword alignment. */ + doubleword_aligned_p = (riscv_function_arg_boundary (mode, type) + > BITS_PER_WORD); + + /* Set REG_OFFSET to the register count we're interested in. + The EABI allocates the floating-point registers separately, + but the other ABIs allocate them like integer registers. */ + info->reg_offset = cum->num_gprs; + + /* Advance to an even register if the argument is doubleword-aligned. */ + if (doubleword_aligned_p) + info->reg_offset += info->reg_offset & 1; + + /* Work out the offset of a stack argument. */ + info->stack_offset = cum->stack_words; + if (doubleword_aligned_p) + info->stack_offset += info->stack_offset & 1; + + max_regs = MAX_ARGS_IN_REGISTERS - info->reg_offset; + + /* Partition the argument between registers and stack. */ + info->reg_words = MIN (num_words, max_regs); + info->stack_words = num_words - info->reg_words; +} + +/* INFO describes a register argument that has the normal format for the + argument's mode. Return the register it uses, assuming that FPRs are + available if HARD_FLOAT_P. */ + +static unsigned int +riscv_arg_regno (const struct riscv_arg_info *info, bool hard_float_p) +{ + if (!info->fpr_p || !hard_float_p) + return GP_ARG_FIRST + info->reg_offset; + else + return FP_ARG_FIRST + info->reg_offset; +} + +/* Implement TARGET_FUNCTION_ARG. */ + +static rtx +riscv_function_arg (cumulative_args_t cum_v, enum machine_mode mode, + const_tree type, bool named) +{ + CUMULATIVE_ARGS *cum = get_cumulative_args (cum_v); + struct riscv_arg_info info; + + if (mode == VOIDmode) + return NULL; + + riscv_get_arg_info (&info, cum, mode, type, named); + + /* Return straight away if the whole argument is passed on the stack. */ + if (info.reg_offset == MAX_ARGS_IN_REGISTERS) + return NULL; + + /* The n32 and n64 ABIs say that if any 64-bit chunk of the structure + contains a double in its entirety, then that 64-bit chunk is passed + in a floating-point register. */ + if (TARGET_HARD_FLOAT + && named + && type != 0 + && TREE_CODE (type) == RECORD_TYPE + && TYPE_SIZE_UNIT (type) + && tree_fits_uhwi_p (TYPE_SIZE_UNIT (type))) + { + tree field; + + /* First check to see if there is any such field. */ + for (field = TYPE_FIELDS (type); field; field = DECL_CHAIN (field)) + if (TREE_CODE (field) == FIELD_DECL + && SCALAR_FLOAT_TYPE_P (TREE_TYPE (field)) + && TYPE_PRECISION (TREE_TYPE (field)) == BITS_PER_WORD + && tree_fits_shwi_p (bit_position (field)) + && int_bit_position (field) % BITS_PER_WORD == 0) + break; + + if (field != 0) + { + /* Now handle the special case by returning a PARALLEL + indicating where each 64-bit chunk goes. INFO.REG_WORDS + chunks are passed in registers. */ + unsigned int i; + HOST_WIDE_INT bitpos; + rtx ret; + + /* assign_parms checks the mode of ENTRY_PARM, so we must + use the actual mode here. */ + ret = gen_rtx_PARALLEL (mode, rtvec_alloc (info.reg_words)); + + bitpos = 0; + field = TYPE_FIELDS (type); + for (i = 0; i < info.reg_words; i++) + { + rtx reg; + + for (; field; field = DECL_CHAIN (field)) + if (TREE_CODE (field) == FIELD_DECL + && int_bit_position (field) >= bitpos) + break; + + if (field + && int_bit_position (field) == bitpos + && SCALAR_FLOAT_TYPE_P (TREE_TYPE (field)) + && TYPE_PRECISION (TREE_TYPE (field)) == BITS_PER_WORD) + reg = gen_rtx_REG (DFmode, FP_ARG_FIRST + info.reg_offset + i); + else + reg = gen_rtx_REG (DImode, GP_ARG_FIRST + info.reg_offset + i); + + XVECEXP (ret, 0, i) + = gen_rtx_EXPR_LIST (VOIDmode, reg, + GEN_INT (bitpos / BITS_PER_UNIT)); + + bitpos += BITS_PER_WORD; + } + return ret; + } + } + + /* Handle the n32/n64 conventions for passing complex floating-point + arguments in FPR pairs. The real part goes in the lower register + and the imaginary part goes in the upper register. */ + if (info.fpr_p + && GET_MODE_CLASS (mode) == MODE_COMPLEX_FLOAT) + { + rtx real, imag; + enum machine_mode inner; + unsigned int regno; + + inner = GET_MODE_INNER (mode); + regno = FP_ARG_FIRST + info.reg_offset; + if (info.reg_words * UNITS_PER_WORD == GET_MODE_SIZE (inner)) + { + /* Real part in registers, imaginary part on stack. */ + gcc_assert (info.stack_words == info.reg_words); + return gen_rtx_REG (inner, regno); + } + else + { + gcc_assert (info.stack_words == 0); + real = gen_rtx_EXPR_LIST (VOIDmode, + gen_rtx_REG (inner, regno), + const0_rtx); + imag = gen_rtx_EXPR_LIST (VOIDmode, + gen_rtx_REG (inner, + regno + info.reg_words / 2), + GEN_INT (GET_MODE_SIZE (inner))); + return gen_rtx_PARALLEL (mode, gen_rtvec (2, real, imag)); + } + } + + return gen_rtx_REG (mode, riscv_arg_regno (&info, TARGET_HARD_FLOAT)); +} + +/* Implement TARGET_FUNCTION_ARG_ADVANCE. */ + +static void +riscv_function_arg_advance (cumulative_args_t cum_v, enum machine_mode mode, + const_tree type, bool named) +{ + CUMULATIVE_ARGS *cum = get_cumulative_args (cum_v); + struct riscv_arg_info info; + + riscv_get_arg_info (&info, cum, mode, type, named); + + /* Advance the register count. This has the effect of setting + num_gprs to MAX_ARGS_IN_REGISTERS if a doubleword-aligned + argument required us to skip the final GPR and pass the whole + argument on the stack. */ + cum->num_gprs = info.reg_offset + info.reg_words; + + /* Advance the stack word count. */ + if (info.stack_words > 0) + cum->stack_words = info.stack_offset + info.stack_words; +} + +/* Implement TARGET_ARG_PARTIAL_BYTES. */ + +static int +riscv_arg_partial_bytes (cumulative_args_t cum, + enum machine_mode mode, tree type, bool named) +{ + struct riscv_arg_info info; + + riscv_get_arg_info (&info, get_cumulative_args (cum), mode, type, named); + return info.stack_words > 0 ? info.reg_words * UNITS_PER_WORD : 0; +} + +/* See whether VALTYPE is a record whose fields should be returned in + floating-point registers. If so, return the number of fields and + list them in FIELDS (which should have two elements). Return 0 + otherwise. + + For n32 & n64, a structure with one or two fields is returned in + floating-point registers as long as every field has a floating-point + type. */ + +static int +riscv_fpr_return_fields (const_tree valtype, tree *fields) +{ + tree field; + int i; + + if (TREE_CODE (valtype) != RECORD_TYPE) + return 0; + + i = 0; + for (field = TYPE_FIELDS (valtype); field != 0; field = DECL_CHAIN (field)) + { + if (TREE_CODE (field) != FIELD_DECL) + continue; + + if (!SCALAR_FLOAT_TYPE_P (TREE_TYPE (field))) + return 0; + + if (i == 2) + return 0; + + fields[i++] = field; + } + return i; +} + +/* Return true if the function return value MODE will get returned in a + floating-point register. */ + +static bool +riscv_return_mode_in_fpr_p (enum machine_mode mode) +{ + return ((GET_MODE_CLASS (mode) == MODE_FLOAT + || GET_MODE_CLASS (mode) == MODE_VECTOR_FLOAT + || GET_MODE_CLASS (mode) == MODE_COMPLEX_FLOAT) + && GET_MODE_UNIT_SIZE (mode) <= UNITS_PER_HWFPVALUE); +} + +/* Return the representation of an FPR return register when the + value being returned in FP_RETURN has mode VALUE_MODE and the + return type itself has mode TYPE_MODE. On NewABI targets, + the two modes may be different for structures like: + + struct __attribute__((packed)) foo { float f; } + + where we return the SFmode value of "f" in FP_RETURN, but where + the structure itself has mode BLKmode. */ + +static rtx +riscv_return_fpr_single (enum machine_mode type_mode, + enum machine_mode value_mode) +{ + rtx x; + + x = gen_rtx_REG (value_mode, FP_RETURN); + if (type_mode != value_mode) + { + x = gen_rtx_EXPR_LIST (VOIDmode, x, const0_rtx); + x = gen_rtx_PARALLEL (type_mode, gen_rtvec (1, x)); + } + return x; +} + +/* Return a composite value in a pair of floating-point registers. + MODE1 and OFFSET1 are the mode and byte offset for the first value, + likewise MODE2 and OFFSET2 for the second. MODE is the mode of the + complete value. + + For n32 & n64, $f0 always holds the first value and $f2 the second. + Otherwise the values are packed together as closely as possible. */ + +static rtx +riscv_return_fpr_pair (enum machine_mode mode, + enum machine_mode mode1, HOST_WIDE_INT offset1, + enum machine_mode mode2, HOST_WIDE_INT offset2) +{ + return gen_rtx_PARALLEL + (mode, + gen_rtvec (2, + gen_rtx_EXPR_LIST (VOIDmode, + gen_rtx_REG (mode1, FP_RETURN), + GEN_INT (offset1)), + gen_rtx_EXPR_LIST (VOIDmode, + gen_rtx_REG (mode2, FP_RETURN + 1), + GEN_INT (offset2)))); + +} + +/* Implement FUNCTION_VALUE and LIBCALL_VALUE. For normal calls, + VALTYPE is the return type and MODE is VOIDmode. For libcalls, + VALTYPE is null and MODE is the mode of the return value. */ + +rtx +riscv_function_value (const_tree valtype, const_tree func, enum machine_mode mode) +{ + if (valtype) + { + tree fields[2]; + int unsigned_p; + + mode = TYPE_MODE (valtype); + unsigned_p = TYPE_UNSIGNED (valtype); + + /* Since TARGET_PROMOTE_FUNCTION_MODE unconditionally promotes, + return values, promote the mode here too. */ + mode = promote_function_mode (valtype, mode, &unsigned_p, func, 1); + + /* Handle structures whose fields are returned in $f0/$f2. */ + switch (riscv_fpr_return_fields (valtype, fields)) + { + case 1: + return riscv_return_fpr_single (mode, + TYPE_MODE (TREE_TYPE (fields[0]))); + + case 2: + return riscv_return_fpr_pair (mode, + TYPE_MODE (TREE_TYPE (fields[0])), + int_byte_position (fields[0]), + TYPE_MODE (TREE_TYPE (fields[1])), + int_byte_position (fields[1])); + } + + /* Only use FPRs for scalar, complex or vector types. */ + if (!FLOAT_TYPE_P (valtype)) + return gen_rtx_REG (mode, GP_RETURN); + } + + /* Handle long doubles for n32 & n64. */ + if (mode == TFmode) + return riscv_return_fpr_pair (mode, + DImode, 0, + DImode, GET_MODE_SIZE (mode) / 2); + + if (riscv_return_mode_in_fpr_p (mode)) + { + if (GET_MODE_CLASS (mode) == MODE_COMPLEX_FLOAT) + return riscv_return_fpr_pair (mode, + GET_MODE_INNER (mode), 0, + GET_MODE_INNER (mode), + GET_MODE_SIZE (mode) / 2); + else + return gen_rtx_REG (mode, FP_RETURN); + } + + return gen_rtx_REG (mode, GP_RETURN); +} + +/* Implement TARGET_RETURN_IN_MEMORY. Scalars and small structures + that fit in two registers are returned in a0/a1. */ + +static bool +riscv_return_in_memory (const_tree type, const_tree fndecl ATTRIBUTE_UNUSED) +{ + return !IN_RANGE (int_size_in_bytes (type), 0, 2 * UNITS_PER_WORD); +} + +/* Implement TARGET_PASS_BY_REFERENCE. */ + +static bool +riscv_pass_by_reference (cumulative_args_t cum ATTRIBUTE_UNUSED, + enum machine_mode mode, const_tree type, + bool named ATTRIBUTE_UNUSED) +{ + if (type && riscv_return_in_memory (type, NULL_TREE)) + return true; + return targetm.calls.must_pass_in_stack (mode, type); +} + +/* Implement TARGET_SETUP_INCOMING_VARARGS. */ + +static void +riscv_setup_incoming_varargs (cumulative_args_t cum, enum machine_mode mode, + tree type, int *pretend_size ATTRIBUTE_UNUSED, + int no_rtl) +{ + CUMULATIVE_ARGS local_cum; + int gp_saved; + + /* The caller has advanced CUM up to, but not beyond, the last named + argument. Advance a local copy of CUM past the last "real" named + argument, to find out how many registers are left over. */ + local_cum = *get_cumulative_args (cum); + riscv_function_arg_advance (pack_cumulative_args (&local_cum), mode, type, 1); + + /* Found out how many registers we need to save. */ + gp_saved = MAX_ARGS_IN_REGISTERS - local_cum.num_gprs; + + if (!no_rtl && gp_saved > 0) + { + rtx ptr, mem; + + ptr = plus_constant (Pmode, virtual_incoming_args_rtx, + REG_PARM_STACK_SPACE (cfun->decl) + - gp_saved * UNITS_PER_WORD); + mem = gen_frame_mem (BLKmode, ptr); + set_mem_alias_set (mem, get_varargs_alias_set ()); + + move_block_from_reg (local_cum.num_gprs + GP_ARG_FIRST, + mem, gp_saved); + } + if (REG_PARM_STACK_SPACE (cfun->decl) == 0) + cfun->machine->varargs_size = gp_saved * UNITS_PER_WORD; +} + +/* Implement TARGET_EXPAND_BUILTIN_VA_START. */ + +static void +riscv_va_start (tree valist, rtx nextarg) +{ + nextarg = plus_constant (Pmode, nextarg, -cfun->machine->varargs_size); + std_expand_builtin_va_start (valist, nextarg); +} + +/* Expand a call of type TYPE. RESULT is where the result will go (null + for "call"s and "sibcall"s), ADDR is the address of the function, + ARGS_SIZE is the size of the arguments and AUX is the value passed + to us by riscv_function_arg. Return the call itself. */ + +rtx +riscv_expand_call (bool sibcall_p, rtx result, rtx addr, rtx args_size) +{ + rtx pattern; + + if (!call_insn_operand (addr, VOIDmode)) + { + rtx reg = RISCV_EPILOGUE_TEMP (Pmode); + riscv_emit_move (reg, addr); + addr = reg; + } + + if (result == 0) + { + rtx (*fn) (rtx, rtx); + + if (sibcall_p) + fn = gen_sibcall_internal; + else + fn = gen_call_internal; + + pattern = fn (addr, args_size); + } + else if (GET_CODE (result) == PARALLEL && XVECLEN (result, 0) == 2) + { + /* Handle return values created by riscv_return_fpr_pair. */ + rtx (*fn) (rtx, rtx, rtx, rtx); + rtx reg1, reg2; + + if (sibcall_p) + fn = gen_sibcall_value_multiple_internal; + else + fn = gen_call_value_multiple_internal; + + reg1 = XEXP (XVECEXP (result, 0, 0), 0); + reg2 = XEXP (XVECEXP (result, 0, 1), 0); + pattern = fn (reg1, addr, args_size, reg2); + } + else + { + rtx (*fn) (rtx, rtx, rtx); + + if (sibcall_p) + fn = gen_sibcall_value_internal; + else + fn = gen_call_value_internal; + + /* Handle return values created by riscv_return_fpr_single. */ + if (GET_CODE (result) == PARALLEL && XVECLEN (result, 0) == 1) + result = XEXP (XVECEXP (result, 0, 0), 0); + pattern = fn (result, addr, args_size); + } + + return emit_call_insn (pattern); +} + +/* Emit straight-line code to move LENGTH bytes from SRC to DEST. + Assume that the areas do not overlap. */ + +static void +riscv_block_move_straight (rtx dest, rtx src, HOST_WIDE_INT length) +{ + HOST_WIDE_INT offset, delta; + unsigned HOST_WIDE_INT bits; + int i; + enum machine_mode mode; + rtx *regs; + + bits = MAX( BITS_PER_UNIT, + MIN( BITS_PER_WORD, MIN( MEM_ALIGN(src),MEM_ALIGN(dest) ) ) ); + + mode = mode_for_size (bits, MODE_INT, 0); + delta = bits / BITS_PER_UNIT; + + /* Allocate a buffer for the temporary registers. */ + regs = XALLOCAVEC (rtx, length / delta); + + /* Load as many BITS-sized chunks as possible. Use a normal load if + the source has enough alignment, otherwise use left/right pairs. */ + for (offset = 0, i = 0; offset + delta <= length; offset += delta, i++) + { + regs[i] = gen_reg_rtx (mode); + riscv_emit_move (regs[i], adjust_address (src, mode, offset)); + } + + /* Copy the chunks to the destination. */ + for (offset = 0, i = 0; offset + delta <= length; offset += delta, i++) + riscv_emit_move (adjust_address (dest, mode, offset), regs[i]); + + /* Mop up any left-over bytes. */ + if (offset < length) + { + src = adjust_address (src, BLKmode, offset); + dest = adjust_address (dest, BLKmode, offset); + move_by_pieces (dest, src, length - offset, + MIN (MEM_ALIGN (src), MEM_ALIGN (dest)), 0); + } +} + +/* Helper function for doing a loop-based block operation on memory + reference MEM. Each iteration of the loop will operate on LENGTH + bytes of MEM. + + Create a new base register for use within the loop and point it to + the start of MEM. Create a new memory reference that uses this + register. Store them in *LOOP_REG and *LOOP_MEM respectively. */ + +static void +riscv_adjust_block_mem (rtx mem, HOST_WIDE_INT length, + rtx *loop_reg, rtx *loop_mem) +{ + *loop_reg = copy_addr_to_reg (XEXP (mem, 0)); + + /* Although the new mem does not refer to a known location, + it does keep up to LENGTH bytes of alignment. */ + *loop_mem = change_address (mem, BLKmode, *loop_reg); + set_mem_align (*loop_mem, MIN (MEM_ALIGN (mem), length * BITS_PER_UNIT)); +} + +/* Move LENGTH bytes from SRC to DEST using a loop that moves BYTES_PER_ITER + bytes at a time. LENGTH must be at least BYTES_PER_ITER. Assume that + the memory regions do not overlap. */ + +static void +riscv_block_move_loop (rtx dest, rtx src, HOST_WIDE_INT length, + HOST_WIDE_INT bytes_per_iter) +{ + rtx label, src_reg, dest_reg, final_src, test; + HOST_WIDE_INT leftover; + + leftover = length % bytes_per_iter; + length -= leftover; + + /* Create registers and memory references for use within the loop. */ + riscv_adjust_block_mem (src, bytes_per_iter, &src_reg, &src); + riscv_adjust_block_mem (dest, bytes_per_iter, &dest_reg, &dest); + + /* Calculate the value that SRC_REG should have after the last iteration + of the loop. */ + final_src = expand_simple_binop (Pmode, PLUS, src_reg, GEN_INT (length), + 0, 0, OPTAB_WIDEN); + + /* Emit the start of the loop. */ + label = gen_label_rtx (); + emit_label (label); + + /* Emit the loop body. */ + riscv_block_move_straight (dest, src, bytes_per_iter); + + /* Move on to the next block. */ + riscv_emit_move (src_reg, plus_constant (Pmode, src_reg, bytes_per_iter)); + riscv_emit_move (dest_reg, plus_constant (Pmode, dest_reg, bytes_per_iter)); + + /* Emit the loop condition. */ + test = gen_rtx_NE (VOIDmode, src_reg, final_src); + if (Pmode == DImode) + emit_jump_insn (gen_cbranchdi4 (test, src_reg, final_src, label)); + else + emit_jump_insn (gen_cbranchsi4 (test, src_reg, final_src, label)); + + /* Mop up any left-over bytes. */ + if (leftover) + riscv_block_move_straight (dest, src, leftover); +} + +/* Expand a movmemsi instruction, which copies LENGTH bytes from + memory reference SRC to memory reference DEST. */ + +bool +riscv_expand_block_move (rtx dest, rtx src, rtx length) +{ + if (CONST_INT_P (length)) + { + HOST_WIDE_INT factor, align; + + align = MIN (MIN (MEM_ALIGN (src), MEM_ALIGN (dest)), BITS_PER_WORD); + factor = BITS_PER_WORD / align; + + if (INTVAL (length) <= RISCV_MAX_MOVE_BYTES_STRAIGHT / factor) + { + riscv_block_move_straight (dest, src, INTVAL (length)); + return true; + } + else if (optimize && align >= BITS_PER_WORD) + { + riscv_block_move_loop (dest, src, INTVAL (length), + RISCV_MAX_MOVE_BYTES_PER_LOOP_ITER / factor); + return true; + } + } + return false; +} + +/* (Re-)Initialize riscv_lo_relocs and riscv_hi_relocs. */ + +static void +riscv_init_relocs (void) +{ + memset (riscv_hi_relocs, '\0', sizeof (riscv_hi_relocs)); + memset (riscv_lo_relocs, '\0', sizeof (riscv_lo_relocs)); + + if (!flag_pic && riscv_cmodel == CM_MEDLOW) + { + riscv_hi_relocs[SYMBOL_ABSOLUTE] = "%hi("; + riscv_lo_relocs[SYMBOL_ABSOLUTE] = "%lo("; + } + + if (!flag_pic || flag_pie) + { + riscv_hi_relocs[SYMBOL_TLS_LE] = "%tprel_hi("; + riscv_lo_relocs[SYMBOL_TLS_LE] = "%tprel_lo("; + } +} + +/* Print symbolic operand OP, which is part of a HIGH or LO_SUM + in context CONTEXT. RELOCS is the array of relocations to use. */ + +static void +riscv_print_operand_reloc (FILE *file, rtx op, const char **relocs) +{ + enum riscv_symbol_type symbol_type; + const char *p; + + symbol_type = riscv_classify_symbolic_expression (op); + gcc_assert (relocs[symbol_type]); + + fputs (relocs[symbol_type], file); + output_addr_const (file, riscv_strip_unspec_address (op)); + for (p = relocs[symbol_type]; *p != 0; p++) + if (*p == '(') + fputc (')', file); +} + +static const char * +riscv_memory_model_suffix (enum memmodel model) +{ + switch (model) + { + case MEMMODEL_ACQ_REL: + case MEMMODEL_SEQ_CST: + return ".sc"; + case MEMMODEL_ACQUIRE: + case MEMMODEL_CONSUME: + return ".aq"; + case MEMMODEL_RELEASE: + return ".rl"; + case MEMMODEL_RELAXED: + return ""; + default: gcc_unreachable(); + } +} + +/* Implement TARGET_PRINT_OPERAND. The RISCV-specific operand codes are: + + 'h' Print the high-part relocation associated with OP, after stripping + any outermost HIGH. + 'R' Print the low-part relocation associated with OP. + 'C' Print the integer branch condition for comparison OP. + 'A' Print the atomic operation suffix for memory model OP. + 'z' Print $0 if OP is zero, otherwise print OP normally. */ + +static void +riscv_print_operand (FILE *file, rtx op, int letter) +{ + enum rtx_code code; + + gcc_assert (op); + code = GET_CODE (op); + + switch (letter) + { + case 'h': + if (code == HIGH) + op = XEXP (op, 0); + riscv_print_operand_reloc (file, op, riscv_hi_relocs); + break; + + case 'R': + riscv_print_operand_reloc (file, op, riscv_lo_relocs); + break; + + case 'C': + /* The RTL names match the instruction names. */ + fputs (GET_RTX_NAME (code), file); + break; + + case 'A': + fputs (riscv_memory_model_suffix ((enum memmodel)INTVAL (op)), file); + break; + + default: + switch (code) + { + case REG: + if (letter && letter != 'z') + output_operand_lossage ("invalid use of '%%%c'", letter); + fprintf (file, "%s", reg_names[REGNO (op)]); + break; + + case MEM: + if (letter == 'y') + fprintf (file, "%s", reg_names[REGNO(XEXP(op, 0))]); + else if (letter && letter != 'z') + output_operand_lossage ("invalid use of '%%%c'", letter); + else + output_address (XEXP (op, 0)); + break; + + default: + if (letter == 'z' && op == CONST0_RTX (GET_MODE (op))) + fputs (reg_names[GP_REG_FIRST], file); + else if (letter && letter != 'z') + output_operand_lossage ("invalid use of '%%%c'", letter); + else + output_addr_const (file, riscv_strip_unspec_address (op)); + break; + } + } +} + +/* Implement TARGET_PRINT_OPERAND_ADDRESS. */ + +static void +riscv_print_operand_address (FILE *file, rtx x) +{ + struct riscv_address_info addr; + + if (riscv_classify_address (&addr, x, word_mode, true)) + switch (addr.type) + { + case ADDRESS_REG: + riscv_print_operand (file, addr.offset, 0); + fprintf (file, "(%s)", reg_names[REGNO (addr.reg)]); + return; + + case ADDRESS_LO_SUM: + riscv_print_operand_reloc (file, addr.offset, riscv_lo_relocs); + fprintf (file, "(%s)", reg_names[REGNO (addr.reg)]); + return; + + case ADDRESS_CONST_INT: + output_addr_const (file, x); + fprintf (file, "(%s)", reg_names[GP_REG_FIRST]); + return; + + case ADDRESS_SYMBOLIC: + output_addr_const (file, riscv_strip_unspec_address (x)); + return; + } + gcc_unreachable (); +} + +static bool +riscv_size_ok_for_small_data_p (int size) +{ + return g_switch_value && IN_RANGE (size, 1, g_switch_value); +} + +/* Return true if EXP should be placed in the small data section. */ + +static bool +riscv_in_small_data_p (const_tree x) +{ + if (TREE_CODE (x) == STRING_CST || TREE_CODE (x) == FUNCTION_DECL) + return false; + + if (TREE_CODE (x) == VAR_DECL && DECL_SECTION_NAME (x)) + { + const char *sec = TREE_STRING_POINTER (DECL_SECTION_NAME (x)); + return strcmp (sec, ".sdata") == 0 || strcmp (sec, ".sbss") == 0; + } + + return riscv_size_ok_for_small_data_p (int_size_in_bytes (TREE_TYPE (x))); +} + +/* Return a section for X, handling small data. */ + +static section * +riscv_elf_select_rtx_section (enum machine_mode mode, rtx x, + unsigned HOST_WIDE_INT align) +{ + section *s = default_elf_select_rtx_section (mode, x, align); + + if (riscv_size_ok_for_small_data_p (GET_MODE_SIZE (mode))) + { + if (strncmp (s->named.name, ".rodata.cst", strlen (".rodata.cst")) == 0) + { + /* Rename .rodata.cst* to .srodata.cst*. */ + char name[32]; + sprintf (name, ".s%s", s->named.name + 1); + return get_section (name, s->named.common.flags, NULL); + } + + if (s == data_section) + return sdata_section; + } + + return s; +} + +/* Implement TARGET_ASM_OUTPUT_DWARF_DTPREL. */ + +static void ATTRIBUTE_UNUSED +riscv_output_dwarf_dtprel (FILE *file, int size, rtx x) +{ + switch (size) + { + case 4: + fputs ("\t.dtprelword\t", file); + break; + + case 8: + fputs ("\t.dtpreldword\t", file); + break; + + default: + gcc_unreachable (); + } + output_addr_const (file, x); + fputs ("+0x800", file); +} + +/* Make the last instruction frame-related and note that it performs + the operation described by FRAME_PATTERN. */ + +static void +riscv_set_frame_expr (rtx frame_pattern) +{ + rtx insn; + + insn = get_last_insn (); + RTX_FRAME_RELATED_P (insn) = 1; + REG_NOTES (insn) = alloc_EXPR_LIST (REG_FRAME_RELATED_EXPR, + frame_pattern, + REG_NOTES (insn)); +} + +/* Return a frame-related rtx that stores REG at MEM. + REG must be a single register. */ + +static rtx +riscv_frame_set (rtx mem, rtx reg) +{ + rtx set; + + set = gen_rtx_SET (VOIDmode, mem, reg); + RTX_FRAME_RELATED_P (set) = 1; + + return set; +} + +/* Return true if the current function must save register REGNO. */ + +static bool +riscv_save_reg_p (unsigned int regno) +{ + bool call_saved = !global_regs[regno] && !call_really_used_regs[regno]; + bool might_clobber = crtl->saves_all_registers + || df_regs_ever_live_p (regno) + || (regno == HARD_FRAME_POINTER_REGNUM + && frame_pointer_needed); + + return (call_saved && might_clobber) + || (regno == RETURN_ADDR_REGNUM && crtl->calls_eh_return); +} + +/* Determine whether to call GPR save/restore routines. */ +static bool +riscv_use_save_libcall (const struct riscv_frame_info *frame) +{ + if (!TARGET_SAVE_RESTORE || crtl->calls_eh_return || frame_pointer_needed) + return false; + + return frame->save_libcall_adjustment != 0; +} + +/* Determine which GPR save/restore routine to call. */ + +static unsigned +riscv_save_libcall_count (unsigned mask) +{ + for (unsigned n = GP_REG_LAST; n > GP_REG_FIRST; n--) + if (BITSET_P (mask, n)) + return CALLEE_SAVED_REG_NUMBER (n) + 1; + abort (); +} + +/* Populate the current function's riscv_frame_info structure. + + RISC-V stack frames grown downward. High addresses are at the top. + + +-------------------------------+ + | | + | incoming stack arguments | + | | + +-------------------------------+ <-- incoming stack pointer + | | + | callee-allocated save area | + | for arguments that are | + | split between registers and | + | the stack | + | | + +-------------------------------+ <-- arg_pointer_rtx + | | + | callee-allocated save area | + | for register varargs | + | | + +-------------------------------+ <-- hard_frame_pointer_rtx; + | | stack_pointer_rtx + gp_sp_offset + | GPR save area | + UNITS_PER_WORD + | | + +-------------------------------+ <-- stack_pointer_rtx + fp_sp_offset + | | + UNITS_PER_HWVALUE + | FPR save area | + | | + +-------------------------------+ <-- frame_pointer_rtx (virtual) + | | + | local variables | + | | + P +-------------------------------+ + | | + | outgoing stack arguments | + | | + +-------------------------------+ <-- stack_pointer_rtx + + Dynamic stack allocations such as alloca insert data at point P. + They decrease stack_pointer_rtx but leave frame_pointer_rtx and + hard_frame_pointer_rtx unchanged. */ + +static void +riscv_compute_frame_info (void) +{ + struct riscv_frame_info *frame; + HOST_WIDE_INT offset; + unsigned int regno, i, num_x_saved = 0, num_f_saved = 0; + + frame = &cfun->machine->frame; + memset (frame, 0, sizeof (*frame)); + + /* Find out which GPRs we need to save. */ + for (regno = GP_REG_FIRST; regno <= GP_REG_LAST; regno++) + if (riscv_save_reg_p (regno)) + frame->mask |= 1 << (regno - GP_REG_FIRST), num_x_saved++; + + /* If this function calls eh_return, we must also save and restore the + EH data registers. */ + if (crtl->calls_eh_return) + for (i = 0; (regno = EH_RETURN_DATA_REGNO (i)) != INVALID_REGNUM; i++) + frame->mask |= 1 << (regno - GP_REG_FIRST), num_x_saved++; + + /* Find out which FPRs we need to save. This loop must iterate over + the same space as its companion in riscv_for_each_saved_gpr_and_fpr. */ + if (TARGET_HARD_FLOAT) + for (regno = FP_REG_FIRST; regno <= FP_REG_LAST; regno++) + if (riscv_save_reg_p (regno)) + frame->fmask |= 1 << (regno - FP_REG_FIRST), num_f_saved++; + + /* At the bottom of the frame are any outgoing stack arguments. */ + offset = crtl->outgoing_args_size; + /* Next are local stack variables. */ + offset += RISCV_STACK_ALIGN (get_frame_size ()); + /* The virtual frame pointer points above the local variables. */ + frame->frame_pointer_offset = offset; + /* Next are the callee-saved FPRs. */ + if (frame->fmask) + { + offset += RISCV_STACK_ALIGN (num_f_saved * UNITS_PER_FPREG); + frame->fp_sp_offset = offset - UNITS_PER_HWFPVALUE; + } + /* Next are the callee-saved GPRs. */ + if (frame->mask) + { + unsigned x_save_size = RISCV_STACK_ALIGN (num_x_saved * UNITS_PER_WORD); + unsigned num_save_restore = 1 + riscv_save_libcall_count (frame->mask); + + /* Only use save/restore routines if they don't alter the stack size. */ + if (RISCV_STACK_ALIGN (num_save_restore * UNITS_PER_WORD) == x_save_size) + frame->save_libcall_adjustment = x_save_size; + + offset += x_save_size; + frame->gp_sp_offset = offset - UNITS_PER_WORD; + } + /* The hard frame pointer points above the callee-saved GPRs. */ + frame->hard_frame_pointer_offset = offset; + /* Above the hard frame pointer is the callee-allocated varags save area. */ + offset += RISCV_STACK_ALIGN (cfun->machine->varargs_size); + frame->arg_pointer_offset = offset; + /* Next is the callee-allocated area for pretend stack arguments. */ + offset += crtl->args.pretend_args_size; + frame->total_size = offset; + /* Next points the incoming stack pointer and any incoming arguments. */ + + /* Only use save/restore routines when the GPRs are atop the frame. */ + if (frame->hard_frame_pointer_offset != frame->total_size) + frame->save_libcall_adjustment = 0; +} + +/* Make sure that we're not trying to eliminate to the wrong hard frame + pointer. */ + +static bool +riscv_can_eliminate (const int from ATTRIBUTE_UNUSED, const int to) +{ + return (to == HARD_FRAME_POINTER_REGNUM || to == STACK_POINTER_REGNUM); +} + +/* Implement INITIAL_ELIMINATION_OFFSET. FROM is either the frame pointer + or argument pointer. TO is either the stack pointer or hard frame + pointer. */ + +HOST_WIDE_INT +riscv_initial_elimination_offset (int from, int to) +{ + HOST_WIDE_INT src, dest; + + riscv_compute_frame_info (); + + if (to == HARD_FRAME_POINTER_REGNUM) + dest = cfun->machine->frame.hard_frame_pointer_offset; + else if (to == STACK_POINTER_REGNUM) + dest = 0; /* this is the base of all offsets */ + else + gcc_unreachable (); + + if (from == FRAME_POINTER_REGNUM) + src = cfun->machine->frame.frame_pointer_offset; + else if (from == ARG_POINTER_REGNUM) + src = cfun->machine->frame.arg_pointer_offset; + else + gcc_unreachable (); + + return src - dest; +} + +/* Implement RETURN_ADDR_RTX. We do not support moving back to a + previous frame. */ + +rtx +riscv_return_addr (int count, rtx frame ATTRIBUTE_UNUSED) +{ + if (count != 0) + return const0_rtx; + + return get_hard_reg_initial_val (Pmode, RETURN_ADDR_REGNUM); +} + +/* Emit code to change the current function's return address to + ADDRESS. SCRATCH is available as a scratch register, if needed. + ADDRESS and SCRATCH are both word-mode GPRs. */ + +void +riscv_set_return_address (rtx address, rtx scratch) +{ + rtx slot_address; + + gcc_assert (BITSET_P (cfun->machine->frame.mask, RETURN_ADDR_REGNUM)); + slot_address = riscv_add_offset (scratch, stack_pointer_rtx, + cfun->machine->frame.gp_sp_offset); + riscv_emit_move (gen_frame_mem (GET_MODE (address), slot_address), address); +} + +/* A function to save or store a register. The first argument is the + register and the second is the stack slot. */ +typedef void (*riscv_save_restore_fn) (rtx, rtx); + +/* Use FN to save or restore register REGNO. MODE is the register's + mode and OFFSET is the offset of its save slot from the current + stack pointer. */ + +static void +riscv_save_restore_reg (enum machine_mode mode, int regno, + HOST_WIDE_INT offset, riscv_save_restore_fn fn) +{ + rtx mem; + + mem = gen_frame_mem (mode, plus_constant (Pmode, stack_pointer_rtx, offset)); + fn (gen_rtx_REG (mode, regno), mem); +} + +/* Call FN for each register that is saved by the current function. + SP_OFFSET is the offset of the current stack pointer from the start + of the frame. */ + +static void +riscv_for_each_saved_gpr_and_fpr (HOST_WIDE_INT sp_offset, + riscv_save_restore_fn fn) +{ + HOST_WIDE_INT offset; + int regno; + + /* Save the link register and s-registers. */ + offset = cfun->machine->frame.gp_sp_offset - sp_offset; + for (regno = GP_REG_FIRST; regno <= GP_REG_LAST-1; regno++) + if (BITSET_P (cfun->machine->frame.mask, regno - GP_REG_FIRST)) + { + riscv_save_restore_reg (word_mode, regno, offset, fn); + offset -= UNITS_PER_WORD; + } + + /* This loop must iterate over the same space as its companion in + riscv_compute_frame_info. */ + offset = cfun->machine->frame.fp_sp_offset - sp_offset; + for (regno = FP_REG_FIRST; regno <= FP_REG_LAST; regno++) + if (BITSET_P (cfun->machine->frame.fmask, regno - FP_REG_FIRST)) + { + riscv_save_restore_reg (DFmode, regno, offset, fn); + offset -= GET_MODE_SIZE (DFmode); + } +} + +/* Emit a move from SRC to DEST, given that one of them is a register + save slot and that the other is a register. TEMP is a temporary + GPR of the same mode that is available if need be. */ + +static void +riscv_emit_save_slot_move (rtx dest, rtx src, rtx temp) +{ + unsigned int regno; + rtx mem; + enum reg_class rclass; + + if (REG_P (src)) + { + regno = REGNO (src); + mem = dest; + } + else + { + regno = REGNO (dest); + mem = src; + } + + rclass = riscv_secondary_reload_class (REGNO_REG_CLASS (regno), + GET_MODE (mem), mem, mem == src); + + if (rclass == NO_REGS) + riscv_emit_move (dest, src); + else + { + gcc_assert (!reg_overlap_mentioned_p (dest, temp)); + riscv_emit_move (temp, src); + riscv_emit_move (dest, temp); + } + if (MEM_P (dest)) + riscv_set_frame_expr (riscv_frame_set (dest, src)); +} + +/* Save register REG to MEM. Make the instruction frame-related. */ + +static void +riscv_save_reg (rtx reg, rtx mem) +{ + riscv_emit_save_slot_move (mem, reg, RISCV_PROLOGUE_TEMP (GET_MODE (reg))); +} + +/* Return the code to invoke the GPR save routine. */ + +const char * +riscv_output_gpr_save (unsigned mask) +{ + static char buf[GP_REG_NUM * 32]; + size_t len = 0; + unsigned n = riscv_save_libcall_count (mask), i; + unsigned frame_size = RISCV_STACK_ALIGN ((n + 1) * UNITS_PER_WORD); + + len += sprintf (buf + len, "call\tt0,__riscv_save_%u", n); + +#ifdef DWARF2_UNWIND_INFO + /* Describe the effect of the call to __riscv_save_X. */ + if (dwarf2out_do_cfi_asm ()) + { + len += sprintf (buf + len, "\n\t.cfi_def_cfa_offset %u", frame_size); + + for (i = GP_REG_FIRST; i <= GP_REG_LAST; i++) + if (BITSET_P (cfun->machine->frame.mask, i)) + len += sprintf (buf + len, "\n\t.cfi_offset %u,%d", i, + (CALLEE_SAVED_REG_NUMBER (i) + 2) * -UNITS_PER_WORD); + } +#endif + + return buf; +} + +/* Expand the "prologue" pattern. */ + +void +riscv_expand_prologue (void) +{ + struct riscv_frame_info *frame = &cfun->machine->frame; + HOST_WIDE_INT size = frame->total_size; + unsigned mask = frame->mask; + rtx insn; + + if (flag_stack_usage_info) + current_function_static_stack_size = size; + + /* When optimizing for size, call a subroutine to save the registers. */ + if (riscv_use_save_libcall (frame)) + { + frame->mask = 0; /* Temporarily fib that we need not save GPRs. */ + size -= frame->save_libcall_adjustment; + emit_insn (gen_gpr_save (GEN_INT (mask))); + } + + /* Save the registers. Allocate up to RISCV_MAX_FIRST_STACK_STEP + bytes beforehand; this is enough to cover the register save area + without going out of range. */ + if ((frame->mask | frame->fmask) != 0) + { + HOST_WIDE_INT step1; + + step1 = MIN (size, RISCV_MAX_FIRST_STACK_STEP); + insn = gen_add3_insn (stack_pointer_rtx, + stack_pointer_rtx, + GEN_INT (-step1)); + RTX_FRAME_RELATED_P (emit_insn (insn)) = 1; + size -= step1; + riscv_for_each_saved_gpr_and_fpr (size, riscv_save_reg); + } + + frame->mask = mask; /* Undo the above fib. */ + + /* Set up the frame pointer, if we're using one. */ + if (frame_pointer_needed) + { + insn = gen_add3_insn (hard_frame_pointer_rtx, stack_pointer_rtx, + GEN_INT (frame->hard_frame_pointer_offset - size)); + RTX_FRAME_RELATED_P (emit_insn (insn)) = 1; + } + + /* Allocate the rest of the frame. */ + if (size > 0) + { + if (SMALL_OPERAND (-size)) + { + insn = gen_add3_insn (stack_pointer_rtx, stack_pointer_rtx, + GEN_INT (-size)); + RTX_FRAME_RELATED_P (emit_insn (insn)) = 1; + } + else + { + riscv_emit_move (RISCV_PROLOGUE_TEMP (Pmode), GEN_INT (-size)); + emit_insn (gen_add3_insn (stack_pointer_rtx, + stack_pointer_rtx, + RISCV_PROLOGUE_TEMP (Pmode))); + + /* Describe the effect of the previous instructions. */ + insn = plus_constant (Pmode, stack_pointer_rtx, -size); + insn = gen_rtx_SET (VOIDmode, stack_pointer_rtx, insn); + riscv_set_frame_expr (insn); + } + } +} + +/* Emit instructions to restore register REG from slot MEM. */ + +static void +riscv_restore_reg (rtx reg, rtx mem) +{ + riscv_emit_save_slot_move (reg, mem, RISCV_EPILOGUE_TEMP (GET_MODE (reg))); +} + +/* Expand an "epilogue" or "sibcall_epilogue" pattern; SIBCALL_P + says which. */ + +void +riscv_expand_epilogue (bool sibcall_p) +{ + /* Split the frame into two. STEP1 is the amount of stack we should + deallocate before restoring the registers. STEP2 is the amount we + should deallocate afterwards. + + Start off by assuming that no registers need to be restored. */ + struct riscv_frame_info *frame = &cfun->machine->frame; + unsigned mask = frame->mask; + HOST_WIDE_INT step1 = frame->total_size; + HOST_WIDE_INT step2 = 0; + bool use_restore_libcall = !sibcall_p && riscv_use_save_libcall (frame); + rtx ra = gen_rtx_REG (Pmode, RETURN_ADDR_REGNUM); + + if (!sibcall_p && riscv_can_use_return_insn ()) + { + emit_jump_insn (gen_return ()); + return; + } + + /* Move past any dynamic stack allocations. */ + if (cfun->calls_alloca) + { + rtx adjust = GEN_INT (-frame->hard_frame_pointer_offset); + if (!SMALL_INT (adjust)) + { + riscv_emit_move (RISCV_EPILOGUE_TEMP (Pmode), adjust); + adjust = RISCV_EPILOGUE_TEMP (Pmode); + } + + emit_insn (gen_add3_insn (stack_pointer_rtx, hard_frame_pointer_rtx, adjust)); + } + + /* If we need to restore registers, deallocate as much stack as + possible in the second step without going out of range. */ + if ((frame->mask | frame->fmask) != 0) + { + step2 = MIN (step1, RISCV_MAX_FIRST_STACK_STEP); + step1 -= step2; + } + + /* Set TARGET to BASE + STEP1. */ + if (step1 > 0) + { + /* Get an rtx for STEP1 that we can add to BASE. */ + rtx adjust = GEN_INT (step1); + if (!SMALL_OPERAND (step1)) + { + riscv_emit_move (RISCV_EPILOGUE_TEMP (Pmode), adjust); + adjust = RISCV_EPILOGUE_TEMP (Pmode); + } + + emit_insn (gen_add3_insn (stack_pointer_rtx, stack_pointer_rtx, adjust)); + } + + if (use_restore_libcall) + frame->mask = 0; /* Temporarily fib that we need not save GPRs. */ + + /* Restore the registers. */ + riscv_for_each_saved_gpr_and_fpr (frame->total_size - step2, + riscv_restore_reg); + + if (use_restore_libcall) + { + frame->mask = mask; /* Undo the above fib. */ + gcc_assert (step2 >= frame->save_libcall_adjustment); + step2 -= frame->save_libcall_adjustment; + } + + /* Deallocate the final bit of the frame. */ + if (step2 > 0) + emit_insn (gen_add3_insn (stack_pointer_rtx, stack_pointer_rtx, + GEN_INT (step2))); + + if (use_restore_libcall) + { + emit_insn (gen_gpr_restore (GEN_INT (riscv_save_libcall_count (mask)))); + emit_jump_insn (gen_gpr_restore_return (ra)); + return; + } + + /* Add in the __builtin_eh_return stack adjustment. */ + if (crtl->calls_eh_return) + emit_insn (gen_add3_insn (stack_pointer_rtx, stack_pointer_rtx, + EH_RETURN_STACKADJ_RTX)); + + if (!sibcall_p) + emit_jump_insn (gen_simple_return_internal (ra)); +} + +/* Return nonzero if this function is known to have a null epilogue. + This allows the optimizer to omit jumps to jumps if no stack + was created. */ + +bool +riscv_can_use_return_insn (void) +{ + return reload_completed && cfun->machine->frame.total_size == 0; +} + +/* Return true if register REGNO can store a value of mode MODE. + The result of this function is cached in riscv_hard_regno_mode_ok. */ + +static bool +riscv_hard_regno_mode_ok_p (unsigned int regno, enum machine_mode mode) +{ + unsigned int size = GET_MODE_SIZE (mode); + enum mode_class mclass = GET_MODE_CLASS (mode); + + /* This is hella bogus but ira_build segfaults on RV32 without it. */ + if (VECTOR_MODE_P (mode)) + return true; + + if (GP_REG_P (regno)) + { + if (size <= UNITS_PER_WORD) + return true; + + /* Double-word values must be even-register-aligned. */ + if (size <= 2 * UNITS_PER_WORD) + return regno % 2 == 0; + } + + if (FP_REG_P (regno)) + { + if (mclass == MODE_FLOAT + || mclass == MODE_COMPLEX_FLOAT + || mclass == MODE_VECTOR_FLOAT) + return size <= UNITS_PER_FPVALUE; + } + + return false; +} + +/* Implement HARD_REGNO_NREGS. */ + +unsigned int +riscv_hard_regno_nregs (int regno, enum machine_mode mode) +{ + if (FP_REG_P (regno)) + return (GET_MODE_SIZE (mode) + UNITS_PER_FPREG - 1) / UNITS_PER_FPREG; + + /* All other registers are word-sized. */ + return (GET_MODE_SIZE (mode) + UNITS_PER_WORD - 1) / UNITS_PER_WORD; +} + +/* Implement CLASS_MAX_NREGS, taking the maximum of the cases + in riscv_hard_regno_nregs. */ + +int +riscv_class_max_nregs (enum reg_class rclass, enum machine_mode mode) +{ + int size; + HARD_REG_SET left; + + size = 0x8000; + COPY_HARD_REG_SET (left, reg_class_contents[(int) rclass]); + if (hard_reg_set_intersect_p (left, reg_class_contents[(int) FP_REGS])) + { + size = MIN (size, UNITS_PER_FPREG); + AND_COMPL_HARD_REG_SET (left, reg_class_contents[(int) FP_REGS]); + } + if (!hard_reg_set_empty_p (left)) + size = MIN (size, UNITS_PER_WORD); + return (GET_MODE_SIZE (mode) + size - 1) / size; +} + +/* Implement TARGET_PREFERRED_RELOAD_CLASS. */ + +static reg_class_t +riscv_preferred_reload_class (rtx x ATTRIBUTE_UNUSED, reg_class_t rclass) +{ + return reg_class_subset_p (FP_REGS, rclass) ? FP_REGS : + reg_class_subset_p (GR_REGS, rclass) ? GR_REGS : + rclass; +} + +/* RCLASS is a class involved in a REGISTER_MOVE_COST calculation. + Return a "canonical" class to represent it in later calculations. */ + +static reg_class_t +riscv_canonicalize_move_class (reg_class_t rclass) +{ + if (reg_class_subset_p (rclass, GENERAL_REGS)) + rclass = GENERAL_REGS; + + return rclass; +} + +/* Implement TARGET_REGISTER_MOVE_COST. Return 0 for classes that are the + maximum of the move costs for subclasses; regclass will work out + the maximum for us. */ + +static int +riscv_register_move_cost (enum machine_mode mode ATTRIBUTE_UNUSED, + reg_class_t from, reg_class_t to) +{ + from = riscv_canonicalize_move_class (from); + to = riscv_canonicalize_move_class (to); + + if ((from == GENERAL_REGS && to == GENERAL_REGS) + || (from == GENERAL_REGS && to == FP_REGS) + || (from == FP_REGS && to == FP_REGS)) + return COSTS_N_INSNS (1); + + if (from == FP_REGS && to == GENERAL_REGS) + return tune_info->fp_to_int_cost; + + return 0; +} + +/* Implement TARGET_MEMORY_MOVE_COST. */ + +static int +riscv_memory_move_cost (enum machine_mode mode, reg_class_t rclass, bool in) +{ + return (tune_info->memory_cost + + memory_move_secondary_cost (mode, rclass, in)); +} + +/* Return the register class required for a secondary register when + copying between one of the registers in RCLASS and value X, which + has mode MODE. X is the source of the move if IN_P, otherwise it + is the destination. Return NO_REGS if no secondary register is + needed. */ + +enum reg_class +riscv_secondary_reload_class (enum reg_class rclass, + enum machine_mode mode, rtx x, + bool in_p ATTRIBUTE_UNUSED) +{ + int regno; + + regno = true_regnum (x); + + if (reg_class_subset_p (rclass, FP_REGS)) + { + if (MEM_P (x) && (GET_MODE_SIZE (mode) == 4 || GET_MODE_SIZE (mode) == 8)) + /* We can use flw/fld/fsw/fsd. */ + return NO_REGS; + + if (GP_REG_P (regno) || x == CONST0_RTX (mode)) + /* We can use fmv or go through memory when mode > Pmode. */ + return NO_REGS; + + if (CONSTANT_P (x) && !targetm.cannot_force_const_mem (mode, x)) + /* We can force the constant to memory and use flw/fld. */ + return NO_REGS; + + if (FP_REG_P (regno)) + /* We can use fmv.fmt. */ + return NO_REGS; + + /* Otherwise, we need to reload through an integer register. */ + return GR_REGS; + } + if (FP_REG_P (regno)) + return reg_class_subset_p (rclass, GR_REGS) ? NO_REGS : GR_REGS; + + return NO_REGS; +} + +/* Implement TARGET_MODE_REP_EXTENDED. */ + +static int +riscv_mode_rep_extended (enum machine_mode mode, enum machine_mode mode_rep) +{ + /* On 64-bit targets, SImode register values are sign-extended to DImode. */ + if (TARGET_64BIT && mode == SImode && mode_rep == DImode) + return SIGN_EXTEND; + + return UNKNOWN; +} + +/* Implement TARGET_SCALAR_MODE_SUPPORTED_P. */ + +static bool +riscv_scalar_mode_supported_p (enum machine_mode mode) +{ + if (ALL_FIXED_POINT_MODE_P (mode) + && GET_MODE_PRECISION (mode) <= 2 * BITS_PER_WORD) + return true; + + return default_scalar_mode_supported_p (mode); +} + +/* Implement TARGET_SCHED_ADJUST_COST. We assume that anti and output + dependencies have no cost. */ + +static int +riscv_adjust_cost (rtx insn ATTRIBUTE_UNUSED, rtx link, + rtx dep ATTRIBUTE_UNUSED, int cost) +{ + if (REG_NOTE_KIND (link) != 0) + return 0; + return cost; +} + +/* Return the number of instructions that can be issued per cycle. */ + +static int +riscv_issue_rate (void) +{ + return tune_info->issue_rate; +} + +/* This structure describes a single built-in function. */ +struct riscv_builtin_description { + /* The code of the main .md file instruction. See riscv_builtin_type + for more information. */ + enum insn_code icode; + + /* The name of the built-in function. */ + const char *name; + + /* Specifies how the function should be expanded. */ + enum riscv_builtin_type builtin_type; + + /* The function's prototype. */ + enum riscv_function_type function_type; + + /* Whether the function is available. */ + unsigned int (*avail) (void); +}; + +static unsigned int +riscv_builtin_avail_riscv (void) +{ + return 1; +} + +/* Construct a riscv_builtin_description from the given arguments. + + INSN is the name of the associated instruction pattern, without the + leading CODE_FOR_riscv_. + + CODE is the floating-point condition code associated with the + function. It can be 'f' if the field is not applicable. + + NAME is the name of the function itself, without the leading + "__builtin_riscv_". + + BUILTIN_TYPE and FUNCTION_TYPE are riscv_builtin_description fields. + + AVAIL is the name of the availability predicate, without the leading + riscv_builtin_avail_. */ +#define RISCV_BUILTIN(INSN, NAME, BUILTIN_TYPE, FUNCTION_TYPE, AVAIL) \ + { CODE_FOR_ ## INSN, "__builtin_riscv_" NAME, \ + BUILTIN_TYPE, FUNCTION_TYPE, riscv_builtin_avail_ ## AVAIL } + +/* Define __builtin_riscv_, which is a RISCV_BUILTIN_DIRECT function + mapped to instruction CODE_FOR_, FUNCTION_TYPE and AVAIL + are as for RISCV_BUILTIN. */ +#define DIRECT_BUILTIN(INSN, FUNCTION_TYPE, AVAIL) \ + RISCV_BUILTIN (INSN, #INSN, RISCV_BUILTIN_DIRECT, FUNCTION_TYPE, AVAIL) + +/* Define __builtin_riscv_, which is a RISCV_BUILTIN_DIRECT_NO_TARGET + function mapped to instruction CODE_FOR_, FUNCTION_TYPE + and AVAIL are as for RISCV_BUILTIN. */ +#define DIRECT_NO_TARGET_BUILTIN(INSN, FUNCTION_TYPE, AVAIL) \ + RISCV_BUILTIN (INSN, #INSN, RISCV_BUILTIN_DIRECT_NO_TARGET, \ + FUNCTION_TYPE, AVAIL) + +static const struct riscv_builtin_description riscv_builtins[] = { + DIRECT_NO_TARGET_BUILTIN (nop, RISCV_VOID_FTYPE_VOID, riscv), +}; + +/* Index I is the function declaration for riscv_builtins[I], or null if the + function isn't defined on this target. */ +static GTY(()) tree riscv_builtin_decls[ARRAY_SIZE (riscv_builtins)]; + + +/* Source-level argument types. */ +#define RISCV_ATYPE_VOID void_type_node +#define RISCV_ATYPE_INT integer_type_node +#define RISCV_ATYPE_POINTER ptr_type_node +#define RISCV_ATYPE_CPOINTER const_ptr_type_node + +/* Standard mode-based argument types. */ +#define RISCV_ATYPE_UQI unsigned_intQI_type_node +#define RISCV_ATYPE_SI intSI_type_node +#define RISCV_ATYPE_USI unsigned_intSI_type_node +#define RISCV_ATYPE_DI intDI_type_node +#define RISCV_ATYPE_UDI unsigned_intDI_type_node +#define RISCV_ATYPE_SF float_type_node +#define RISCV_ATYPE_DF double_type_node + +/* RISCV_FTYPE_ATYPESN takes N RISCV_FTYPES-like type codes and lists + their associated RISCV_ATYPEs. */ +#define RISCV_FTYPE_ATYPES1(A, B) \ + RISCV_ATYPE_##A, RISCV_ATYPE_##B + +#define RISCV_FTYPE_ATYPES2(A, B, C) \ + RISCV_ATYPE_##A, RISCV_ATYPE_##B, RISCV_ATYPE_##C + +#define RISCV_FTYPE_ATYPES3(A, B, C, D) \ + RISCV_ATYPE_##A, RISCV_ATYPE_##B, RISCV_ATYPE_##C, RISCV_ATYPE_##D + +#define RISCV_FTYPE_ATYPES4(A, B, C, D, E) \ + RISCV_ATYPE_##A, RISCV_ATYPE_##B, RISCV_ATYPE_##C, RISCV_ATYPE_##D, \ + RISCV_ATYPE_##E + +/* Return the function type associated with function prototype TYPE. */ + +static tree +riscv_build_function_type (enum riscv_function_type type) +{ + static tree types[(int) RISCV_MAX_FTYPE_MAX]; + + if (types[(int) type] == NULL_TREE) + switch (type) + { +#define DEF_RISCV_FTYPE(NUM, ARGS) \ + case RISCV_FTYPE_NAME##NUM ARGS: \ + types[(int) type] \ + = build_function_type_list (RISCV_FTYPE_ATYPES##NUM ARGS, \ + NULL_TREE); \ + break; +#include "config/riscv/riscv-ftypes.def" +#undef DEF_RISCV_FTYPE + default: + gcc_unreachable (); + } + + return types[(int) type]; +} + +/* Implement TARGET_INIT_BUILTINS. */ + +static void +riscv_init_builtins (void) +{ + const struct riscv_builtin_description *d; + unsigned int i; + + /* Iterate through all of the bdesc arrays, initializing all of the + builtin functions. */ + for (i = 0; i < ARRAY_SIZE (riscv_builtins); i++) + { + d = &riscv_builtins[i]; + if (d->avail ()) + riscv_builtin_decls[i] + = add_builtin_function (d->name, + riscv_build_function_type (d->function_type), + i, BUILT_IN_MD, NULL, NULL); + } +} + +/* Implement TARGET_BUILTIN_DECL. */ + +static tree +riscv_builtin_decl (unsigned int code, bool initialize_p ATTRIBUTE_UNUSED) +{ + if (code >= ARRAY_SIZE (riscv_builtins)) + return error_mark_node; + return riscv_builtin_decls[code]; +} + +/* Take argument ARGNO from EXP's argument list and convert it into a + form suitable for input operand OPNO of instruction ICODE. Return the + value. */ + +static rtx +riscv_prepare_builtin_arg (enum insn_code icode, + unsigned int opno, tree exp, unsigned int argno) +{ + tree arg; + rtx value; + enum machine_mode mode; + + arg = CALL_EXPR_ARG (exp, argno); + value = expand_normal (arg); + mode = insn_data[icode].operand[opno].mode; + if (!insn_data[icode].operand[opno].predicate (value, mode)) + { + /* We need to get the mode from ARG for two reasons: + + - to cope with address operands, where MODE is the mode of the + memory, rather than of VALUE itself. + + - to cope with special predicates like pmode_register_operand, + where MODE is VOIDmode. */ + value = copy_to_mode_reg (TYPE_MODE (TREE_TYPE (arg)), value); + + /* Check the predicate again. */ + if (!insn_data[icode].operand[opno].predicate (value, mode)) + { + error ("invalid argument to built-in function"); + return const0_rtx; + } + } + + return value; +} + +/* Return an rtx suitable for output operand OP of instruction ICODE. + If TARGET is non-null, try to use it where possible. */ + +static rtx +riscv_prepare_builtin_target (enum insn_code icode, unsigned int op, rtx target) +{ + enum machine_mode mode; + + mode = insn_data[icode].operand[op].mode; + if (target == 0 || !insn_data[icode].operand[op].predicate (target, mode)) + target = gen_reg_rtx (mode); + + return target; +} + +/* Expand a RISCV_BUILTIN_DIRECT or RISCV_BUILTIN_DIRECT_NO_TARGET function; + HAS_TARGET_P says which. EXP is the CALL_EXPR that calls the function + and ICODE is the code of the associated .md pattern. TARGET, if nonnull, + suggests a good place to put the result. */ + +static rtx +riscv_expand_builtin_direct (enum insn_code icode, rtx target, tree exp, + bool has_target_p) +{ + rtx ops[MAX_RECOG_OPERANDS]; + int opno, argno; + + /* Map any target to operand 0. */ + opno = 0; + if (has_target_p) + { + target = riscv_prepare_builtin_target (icode, opno, target); + ops[opno] = target; + opno++; + } + + /* Map the arguments to the other operands. The n_operands value + for an expander includes match_dups and match_scratches as well as + match_operands, so n_operands is only an upper bound on the number + of arguments to the expander function. */ + gcc_assert (opno + call_expr_nargs (exp) <= insn_data[icode].n_operands); + for (argno = 0; argno < call_expr_nargs (exp); argno++, opno++) + ops[opno] = riscv_prepare_builtin_arg (icode, opno, exp, argno); + + switch (opno) + { + case 2: + emit_insn (GEN_FCN (icode) (ops[0], ops[1])); + break; + + case 3: + emit_insn (GEN_FCN (icode) (ops[0], ops[1], ops[2])); + break; + + case 4: + emit_insn (GEN_FCN (icode) (ops[0], ops[1], ops[2], ops[3])); + break; + + default: + gcc_unreachable (); + } + return target; +} + +/* Implement TARGET_EXPAND_BUILTIN. */ + +static rtx +riscv_expand_builtin (tree exp, rtx target, rtx subtarget ATTRIBUTE_UNUSED, + enum machine_mode mode ATTRIBUTE_UNUSED, + int ignore ATTRIBUTE_UNUSED) +{ + tree fndecl; + unsigned int fcode, avail; + const struct riscv_builtin_description *d; + + fndecl = TREE_OPERAND (CALL_EXPR_FN (exp), 0); + fcode = DECL_FUNCTION_CODE (fndecl); + gcc_assert (fcode < ARRAY_SIZE (riscv_builtins)); + d = &riscv_builtins[fcode]; + avail = d->avail (); + gcc_assert (avail != 0); + switch (d->builtin_type) + { + case RISCV_BUILTIN_DIRECT: + return riscv_expand_builtin_direct (d->icode, target, exp, true); + + case RISCV_BUILTIN_DIRECT_NO_TARGET: + return riscv_expand_builtin_direct (d->icode, target, exp, false); + } + gcc_unreachable (); +} + +/* Implement TARGET_ASM_OUTPUT_MI_THUNK. Generate rtl rather than asm text + in order to avoid duplicating too much logic from elsewhere. */ + +static void +riscv_output_mi_thunk (FILE *file, tree thunk_fndecl ATTRIBUTE_UNUSED, + HOST_WIDE_INT delta, HOST_WIDE_INT vcall_offset, + tree function) +{ + rtx this_rtx, temp1, temp2, insn, fnaddr; + bool use_sibcall_p; + + /* Pretend to be a post-reload pass while generating rtl. */ + reload_completed = 1; + + /* Mark the end of the (empty) prologue. */ + emit_note (NOTE_INSN_PROLOGUE_END); + + /* Determine if we can use a sibcall to call FUNCTION directly. */ + fnaddr = XEXP (DECL_RTL (function), 0); + use_sibcall_p = absolute_symbolic_operand (fnaddr, Pmode); + + /* We need two temporary registers in some cases. */ + temp1 = gen_rtx_REG (Pmode, GP_TEMP_FIRST); + temp2 = gen_rtx_REG (Pmode, GP_TEMP_FIRST + 1); + + /* Find out which register contains the "this" pointer. */ + if (aggregate_value_p (TREE_TYPE (TREE_TYPE (function)), function)) + this_rtx = gen_rtx_REG (Pmode, GP_ARG_FIRST + 1); + else + this_rtx = gen_rtx_REG (Pmode, GP_ARG_FIRST); + + /* Add DELTA to THIS_RTX. */ + if (delta != 0) + { + rtx offset = GEN_INT (delta); + if (!SMALL_OPERAND (delta)) + { + riscv_emit_move (temp1, offset); + offset = temp1; + } + emit_insn (gen_add3_insn (this_rtx, this_rtx, offset)); + } + + /* If needed, add *(*THIS_RTX + VCALL_OFFSET) to THIS_RTX. */ + if (vcall_offset != 0) + { + rtx addr; + + /* Set TEMP1 to *THIS_RTX. */ + riscv_emit_move (temp1, gen_rtx_MEM (Pmode, this_rtx)); + + /* Set ADDR to a legitimate address for *THIS_RTX + VCALL_OFFSET. */ + addr = riscv_add_offset (temp2, temp1, vcall_offset); + + /* Load the offset and add it to THIS_RTX. */ + riscv_emit_move (temp1, gen_rtx_MEM (Pmode, addr)); + emit_insn (gen_add3_insn (this_rtx, this_rtx, temp1)); + } + + /* Jump to the target function. Use a sibcall if direct jumps are + allowed, otherwise load the address into a register first. */ + if (use_sibcall_p) + { + insn = emit_call_insn (gen_sibcall_internal (fnaddr, const0_rtx)); + SIBLING_CALL_P (insn) = 1; + } + else + { + riscv_emit_move(temp1, fnaddr); + emit_jump_insn (gen_indirect_jump (temp1)); + } + + /* Run just enough of rest_of_compilation. This sequence was + "borrowed" from alpha.c. */ + insn = get_insns (); + split_all_insns_noflow (); + shorten_branches (insn); + final_start_function (insn, file, 1); + final (insn, file, 1); + final_end_function (); + + /* Clean up the vars set above. Note that final_end_function resets + the global pointer for us. */ + reload_completed = 0; +} + +/* Allocate a chunk of memory for per-function machine-dependent data. */ + +static struct machine_function * +riscv_init_machine_status (void) +{ + return ggc_alloc_cleared_machine_function (); +} + +/* Implement TARGET_OPTION_OVERRIDE. */ + +static void +riscv_option_override (void) +{ + int regno, mode; + const struct riscv_cpu_info *cpu; + +#ifdef SUBTARGET_OVERRIDE_OPTIONS + SUBTARGET_OVERRIDE_OPTIONS; +#endif + + flag_pcc_struct_return = 0; + + if (flag_pic) + g_switch_value = 0; + + /* Prefer a call to memcpy over inline code when optimizing for size, + though see MOVE_RATIO in riscv.h. */ + if (optimize_size && (target_flags_explicit & MASK_MEMCPY) == 0) + target_flags |= MASK_MEMCPY; + + /* Handle -mtune. */ + cpu = riscv_parse_cpu (riscv_tune_string ? riscv_tune_string : + RISCV_TUNE_STRING_DEFAULT); + tune_info = optimize_size ? &optimize_size_tune_info : cpu->tune_info; + + /* If the user hasn't specified a branch cost, use the processor's + default. */ + if (riscv_branch_cost == 0) + riscv_branch_cost = tune_info->branch_cost; + + /* Set up riscv_hard_regno_mode_ok. */ + for (mode = 0; mode < MAX_MACHINE_MODE; mode++) + for (regno = 0; regno < FIRST_PSEUDO_REGISTER; regno++) + riscv_hard_regno_mode_ok[mode][regno] + = riscv_hard_regno_mode_ok_p (regno, (enum machine_mode) mode); + + /* Function to allocate machine-dependent function status. */ + init_machine_status = &riscv_init_machine_status; + + if (riscv_cmodel_string) + { + if (strcmp (riscv_cmodel_string, "medlow") == 0) + riscv_cmodel = CM_MEDLOW; + else if (strcmp (riscv_cmodel_string, "medany") == 0) + riscv_cmodel = CM_MEDANY; + else + error ("unsupported code model: %s", riscv_cmodel_string); + } + + if (flag_pic) + riscv_cmodel = CM_PIC; + + riscv_init_relocs (); +} + +/* Implement TARGET_CONDITIONAL_REGISTER_USAGE. */ + +static void +riscv_conditional_register_usage (void) +{ + int regno; + + if (!TARGET_HARD_FLOAT) + { + for (regno = FP_REG_FIRST; regno <= FP_REG_LAST; regno++) + fixed_regs[regno] = call_used_regs[regno] = 1; + } +} + +/* Implement TARGET_TRAMPOLINE_INIT. */ + +static void +riscv_trampoline_init (rtx m_tramp, tree fndecl, rtx chain_value) +{ + rtx addr, end_addr, mem; + rtx trampoline[4]; + unsigned int i; + HOST_WIDE_INT static_chain_offset, target_function_offset; + + /* Work out the offsets of the pointers from the start of the + trampoline code. */ + gcc_assert (ARRAY_SIZE (trampoline) * 4 == TRAMPOLINE_CODE_SIZE); + static_chain_offset = TRAMPOLINE_CODE_SIZE; + target_function_offset = static_chain_offset + GET_MODE_SIZE (ptr_mode); + + /* Get pointers to the beginning and end of the code block. */ + addr = force_reg (Pmode, XEXP (m_tramp, 0)); + end_addr = riscv_force_binary (Pmode, PLUS, addr, GEN_INT (TRAMPOLINE_CODE_SIZE)); + +#define OP(X) gen_int_mode (X, SImode) +#define MATCH_LREG ((Pmode) == DImode ? MATCH_LD : MATCH_LW) + + /* auipc t0, 0 + l[wd] t1, target_function_offset(t0) + l[wd] $static_chain, static_chain_offset(t0) + jr t1 + */ + + trampoline[0] = OP (RISCV_UTYPE (AUIPC, STATIC_CHAIN_REGNUM, 0)); + trampoline[1] = OP (RISCV_ITYPE (LREG, RISCV_PROLOGUE_TEMP_REGNUM, + STATIC_CHAIN_REGNUM, target_function_offset)); + trampoline[2] = OP (RISCV_ITYPE (LREG, STATIC_CHAIN_REGNUM, + STATIC_CHAIN_REGNUM, static_chain_offset)); + trampoline[3] = OP (RISCV_ITYPE (JALR, 0, RISCV_PROLOGUE_TEMP_REGNUM, 0)); + +#undef MATCH_LREG +#undef OP + + /* Copy the trampoline code. Leave any padding uninitialized. */ + for (i = 0; i < ARRAY_SIZE (trampoline); i++) + { + mem = adjust_address (m_tramp, SImode, i * GET_MODE_SIZE (SImode)); + riscv_emit_move (mem, trampoline[i]); + } + + /* Set up the static chain pointer field. */ + mem = adjust_address (m_tramp, ptr_mode, static_chain_offset); + riscv_emit_move (mem, chain_value); + + /* Set up the target function field. */ + mem = adjust_address (m_tramp, ptr_mode, target_function_offset); + riscv_emit_move (mem, XEXP (DECL_RTL (fndecl), 0)); + + /* Flush the code part of the trampoline. */ + emit_insn (gen_add3_insn (end_addr, addr, GEN_INT (TRAMPOLINE_SIZE))); + emit_insn (gen_clear_cache (addr, end_addr)); +} + +/* Implement TARGET_FUNCTION_OK_FOR_SIBCALL. */ + +static bool +riscv_function_ok_for_sibcall (tree decl, tree exp ATTRIBUTE_UNUSED) +{ + if (TARGET_SAVE_RESTORE) + { + /* When optimzing for size, don't use sibcalls in non-leaf routines */ + if (cfun->machine->is_leaf == 0) + cfun->machine->is_leaf = leaf_function_p () ? 1 : -1; + + return cfun->machine->is_leaf > 0; + } + + return true; +} + +static bool +riscv_lra_p (void) +{ + return riscv_lra_flag; +} + +/* Initialize the GCC target structure. */ +#undef TARGET_ASM_ALIGNED_HI_OP +#define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" +#undef TARGET_ASM_ALIGNED_SI_OP +#define TARGET_ASM_ALIGNED_SI_OP "\t.word\t" +#undef TARGET_ASM_ALIGNED_DI_OP +#define TARGET_ASM_ALIGNED_DI_OP "\t.dword\t" + +#undef TARGET_OPTION_OVERRIDE +#define TARGET_OPTION_OVERRIDE riscv_option_override + +#undef TARGET_LEGITIMIZE_ADDRESS +#define TARGET_LEGITIMIZE_ADDRESS riscv_legitimize_address + +#undef TARGET_SCHED_ADJUST_COST +#define TARGET_SCHED_ADJUST_COST riscv_adjust_cost +#undef TARGET_SCHED_ISSUE_RATE +#define TARGET_SCHED_ISSUE_RATE riscv_issue_rate + +#undef TARGET_FUNCTION_OK_FOR_SIBCALL +#define TARGET_FUNCTION_OK_FOR_SIBCALL riscv_function_ok_for_sibcall + +#undef TARGET_REGISTER_MOVE_COST +#define TARGET_REGISTER_MOVE_COST riscv_register_move_cost +#undef TARGET_MEMORY_MOVE_COST +#define TARGET_MEMORY_MOVE_COST riscv_memory_move_cost +#undef TARGET_RTX_COSTS +#define TARGET_RTX_COSTS riscv_rtx_costs +#undef TARGET_ADDRESS_COST +#define TARGET_ADDRESS_COST riscv_address_cost + +#undef TARGET_PREFERRED_RELOAD_CLASS +#define TARGET_PREFERRED_RELOAD_CLASS riscv_preferred_reload_class + +#undef TARGET_ASM_FILE_START_FILE_DIRECTIVE +#define TARGET_ASM_FILE_START_FILE_DIRECTIVE true + +#undef TARGET_EXPAND_BUILTIN_VA_START +#define TARGET_EXPAND_BUILTIN_VA_START riscv_va_start + +#undef TARGET_PROMOTE_FUNCTION_MODE +#define TARGET_PROMOTE_FUNCTION_MODE default_promote_function_mode_always_promote + +#undef TARGET_RETURN_IN_MEMORY +#define TARGET_RETURN_IN_MEMORY riscv_return_in_memory + +#undef TARGET_ASM_OUTPUT_MI_THUNK +#define TARGET_ASM_OUTPUT_MI_THUNK riscv_output_mi_thunk +#undef TARGET_ASM_CAN_OUTPUT_MI_THUNK +#define TARGET_ASM_CAN_OUTPUT_MI_THUNK hook_bool_const_tree_hwi_hwi_const_tree_true + +#undef TARGET_PRINT_OPERAND +#define TARGET_PRINT_OPERAND riscv_print_operand +#undef TARGET_PRINT_OPERAND_ADDRESS +#define TARGET_PRINT_OPERAND_ADDRESS riscv_print_operand_address + +#undef TARGET_SETUP_INCOMING_VARARGS +#define TARGET_SETUP_INCOMING_VARARGS riscv_setup_incoming_varargs +#undef TARGET_STRICT_ARGUMENT_NAMING +#define TARGET_STRICT_ARGUMENT_NAMING hook_bool_CUMULATIVE_ARGS_true +#undef TARGET_MUST_PASS_IN_STACK +#define TARGET_MUST_PASS_IN_STACK must_pass_in_stack_var_size +#undef TARGET_PASS_BY_REFERENCE +#define TARGET_PASS_BY_REFERENCE riscv_pass_by_reference +#undef TARGET_ARG_PARTIAL_BYTES +#define TARGET_ARG_PARTIAL_BYTES riscv_arg_partial_bytes +#undef TARGET_FUNCTION_ARG +#define TARGET_FUNCTION_ARG riscv_function_arg +#undef TARGET_FUNCTION_ARG_ADVANCE +#define TARGET_FUNCTION_ARG_ADVANCE riscv_function_arg_advance +#undef TARGET_FUNCTION_ARG_BOUNDARY +#define TARGET_FUNCTION_ARG_BOUNDARY riscv_function_arg_boundary + +#undef TARGET_MODE_REP_EXTENDED +#define TARGET_MODE_REP_EXTENDED riscv_mode_rep_extended + +#undef TARGET_SCALAR_MODE_SUPPORTED_P +#define TARGET_SCALAR_MODE_SUPPORTED_P riscv_scalar_mode_supported_p + +#undef TARGET_INIT_BUILTINS +#define TARGET_INIT_BUILTINS riscv_init_builtins +#undef TARGET_BUILTIN_DECL +#define TARGET_BUILTIN_DECL riscv_builtin_decl +#undef TARGET_EXPAND_BUILTIN +#define TARGET_EXPAND_BUILTIN riscv_expand_builtin + +#undef TARGET_HAVE_TLS +#define TARGET_HAVE_TLS HAVE_AS_TLS + +#undef TARGET_CANNOT_FORCE_CONST_MEM +#define TARGET_CANNOT_FORCE_CONST_MEM riscv_cannot_force_const_mem + +#undef TARGET_LEGITIMATE_CONSTANT_P +#define TARGET_LEGITIMATE_CONSTANT_P riscv_legitimate_constant_p + +#undef TARGET_USE_BLOCKS_FOR_CONSTANT_P +#define TARGET_USE_BLOCKS_FOR_CONSTANT_P hook_bool_mode_const_rtx_true + +#ifdef HAVE_AS_DTPRELWORD +#undef TARGET_ASM_OUTPUT_DWARF_DTPREL +#define TARGET_ASM_OUTPUT_DWARF_DTPREL riscv_output_dwarf_dtprel +#endif + +#undef TARGET_LEGITIMATE_ADDRESS_P +#define TARGET_LEGITIMATE_ADDRESS_P riscv_legitimate_address_p + +#undef TARGET_CAN_ELIMINATE +#define TARGET_CAN_ELIMINATE riscv_can_eliminate + +#undef TARGET_CONDITIONAL_REGISTER_USAGE +#define TARGET_CONDITIONAL_REGISTER_USAGE riscv_conditional_register_usage + +#undef TARGET_TRAMPOLINE_INIT +#define TARGET_TRAMPOLINE_INIT riscv_trampoline_init + +#undef TARGET_IN_SMALL_DATA_P +#define TARGET_IN_SMALL_DATA_P riscv_in_small_data_p + +#undef TARGET_ASM_SELECT_RTX_SECTION +#define TARGET_ASM_SELECT_RTX_SECTION riscv_elf_select_rtx_section + +#undef TARGET_MIN_ANCHOR_OFFSET +#define TARGET_MIN_ANCHOR_OFFSET (-RISCV_IMM_REACH/2) + +#undef TARGET_MAX_ANCHOR_OFFSET +#define TARGET_MAX_ANCHOR_OFFSET (RISCV_IMM_REACH/2-1) + +#undef TARGET_LRA_P +#define TARGET_LRA_P riscv_lra_p + +struct gcc_target targetm = TARGET_INITIALIZER; + +#include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv.h b/gcc/config/riscv/riscv.h new file mode 100644 index 0000000..562369c --- /dev/null +++ b/gcc/config/riscv/riscv.h @@ -0,0 +1,1141 @@ +/* Definition of RISC-V target for GNU compiler. + Copyright (C) 2011-2014 Free Software Foundation, Inc. + Contributed by Andrew Waterman (waterman@cs.berkeley.edu) at UC Berkeley. + Based on MIPS target for GNU compiler. + +This file is part of GCC. + +GCC is free software; you can redistribute it and/or modify +it under the terms of the GNU General Public License as published by +the Free Software Foundation; either version 3, or (at your option) +any later version. + +GCC is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +You should have received a copy of the GNU General Public License +along with GCC; see the file COPYING3. If not see +. */ + +/* TARGET_HARD_FLOAT and TARGET_SOFT_FLOAT reflect whether the FPU is + directly accessible, while the command-line options select + TARGET_HARD_FLOAT_ABI and TARGET_SOFT_FLOAT_ABI to reflect the ABI + in use. */ +#define TARGET_HARD_FLOAT TARGET_HARD_FLOAT_ABI +#define TARGET_SOFT_FLOAT TARGET_SOFT_FLOAT_ABI + +/* Target CPU builtins. */ +#define TARGET_CPU_CPP_BUILTINS() \ + do \ + { \ + builtin_assert ("machine=riscv"); \ + \ + builtin_assert ("cpu=riscv"); \ + builtin_define ("__riscv__"); \ + builtin_define ("__riscv"); \ + builtin_define ("_riscv"); \ + \ + if (TARGET_64BIT) \ + { \ + builtin_define ("__riscv64"); \ + builtin_define ("_RISCV_SIM=_ABI64"); \ + } \ + else \ + builtin_define ("_RISCV_SIM=_ABI32"); \ + \ + builtin_define ("_ABI32=1"); \ + builtin_define ("_ABI64=3"); \ + \ + \ + builtin_define_with_int_value ("_RISCV_SZINT", INT_TYPE_SIZE); \ + builtin_define_with_int_value ("_RISCV_SZLONG", LONG_TYPE_SIZE); \ + builtin_define_with_int_value ("_RISCV_SZPTR", POINTER_SIZE); \ + builtin_define_with_int_value ("_RISCV_FPSET", 32); \ + \ + if (TARGET_ATOMIC) { \ + builtin_define ("__riscv_atomic"); \ + } \ + \ + /* These defines reflect the ABI in use, not whether the \ + FPU is directly accessible. */ \ + if (TARGET_HARD_FLOAT_ABI) { \ + builtin_define ("__riscv_hard_float"); \ + if (TARGET_FDIV) { \ + builtin_define ("__riscv_fdiv"); \ + builtin_define ("__riscv_fsqrt"); \ + } \ + } else \ + builtin_define ("__riscv_soft_float"); \ + \ + /* The base RISC-V ISA is always little-endian. */ \ + builtin_define_std ("RISCVEL"); \ + builtin_define ("_RISCVEL"); \ + \ + /* Macros dependent on the C dialect. */ \ + if (preprocessing_asm_p ()) \ + { \ + builtin_define_std ("LANGUAGE_ASSEMBLY"); \ + builtin_define ("_LANGUAGE_ASSEMBLY"); \ + } \ + else if (c_dialect_cxx ()) \ + { \ + builtin_define ("_LANGUAGE_C_PLUS_PLUS"); \ + builtin_define ("__LANGUAGE_C_PLUS_PLUS"); \ + builtin_define ("__LANGUAGE_C_PLUS_PLUS__"); \ + } \ + else \ + { \ + builtin_define_std ("LANGUAGE_C"); \ + builtin_define ("_LANGUAGE_C"); \ + } \ + if (c_dialect_objc ()) \ + { \ + builtin_define ("_LANGUAGE_OBJECTIVE_C"); \ + builtin_define ("__LANGUAGE_OBJECTIVE_C"); \ + /* Bizarre, but needed at least for Irix. */ \ + builtin_define_std ("LANGUAGE_C"); \ + builtin_define ("_LANGUAGE_C"); \ + } \ + if (riscv_cmodel == CM_MEDANY) \ + builtin_define ("_RISCV_CMODEL_MEDANY"); \ + } \ + while (0) + +/* Default target_flags if no switches are specified */ + +#ifndef TARGET_DEFAULT +#define TARGET_DEFAULT 0 +#endif + +#ifndef RISCV_ARCH_STRING_DEFAULT +#define RISCV_ARCH_STRING_DEFAULT "IMAFD" +#endif + +#ifndef RISCV_TUNE_STRING_DEFAULT +#define RISCV_TUNE_STRING_DEFAULT "rocket" +#endif + +#ifndef TARGET_64BIT_DEFAULT +#define TARGET_64BIT_DEFAULT 1 +#endif + +#if TARGET_64BIT_DEFAULT +# define MULTILIB_ARCH_DEFAULT "m64" +# define OPT_ARCH64 "!m32" +# define OPT_ARCH32 "m32" +#else +# define MULTILIB_ARCH_DEFAULT "m32" +# define OPT_ARCH64 "m64" +# define OPT_ARCH32 "!m64" +#endif + +#ifndef MULTILIB_DEFAULTS +#define MULTILIB_DEFAULTS \ + { MULTILIB_ARCH_DEFAULT } +#endif + + +/* Support for a compile-time default CPU, et cetera. The rules are: + --with-arch is ignored if -march is specified. + --with-tune is ignored if -mtune is specified. + --with-float is ignored if -mhard-float or -msoft-float are specified. */ +#define OPTION_DEFAULT_SPECS \ + {"arch", "%{!march=*:-march=%(VALUE)}"}, \ + {"arch_32", "%{" OPT_ARCH32 ":%{m32}}" }, \ + {"arch_64", "%{" OPT_ARCH64 ":%{m64}}" }, \ + {"tune", "%{!mtune=*:-mtune=%(VALUE)}" }, \ + {"float", "%{!msoft-float:%{!mhard-float:-m%(VALUE)-float}}" }, \ + +#define DRIVER_SELF_SPECS "" + +#ifdef IN_LIBGCC2 +#undef TARGET_64BIT +/* Make this compile time constant for libgcc2 */ +#ifdef __riscv64 +#define TARGET_64BIT 1 +#else +#define TARGET_64BIT 0 +#endif +#endif /* IN_LIBGCC2 */ + +/* Tell collect what flags to pass to nm. */ +#ifndef NM_FLAGS +#define NM_FLAGS "-Bn" +#endif + +#undef ASM_SPEC +#define ASM_SPEC "\ +%(subtarget_asm_debugging_spec) \ +%{m32} %{m64} %{!m32:%{!m64: %(asm_abi_default_spec)}} \ +%{mrvc} \ +%{fPIC|fpic|fPIE|fpie:-fpic} \ +%{march=*} \ +%(subtarget_asm_spec)" + +/* Extra switches sometimes passed to the linker. */ + +#ifndef LINK_SPEC +#define LINK_SPEC "\ +%{!T:-dT riscv.ld} \ +%{m64:-melf64lriscv} \ +%{m32:-melf32lriscv} \ +%{shared}" +#endif /* LINK_SPEC defined */ + +/* This macro defines names of additional specifications to put in the specs + that can be used in various specifications like CC1_SPEC. Its definition + is an initializer with a subgrouping for each command option. + + Each subgrouping contains a string constant, that defines the + specification name, and a string constant that used by the GCC driver + program. + + Do not define this macro if it does not need to do anything. */ + +#define EXTRA_SPECS \ + { "asm_abi_default_spec", "-" MULTILIB_ARCH_DEFAULT }, \ + SUBTARGET_EXTRA_SPECS + +#ifndef SUBTARGET_EXTRA_SPECS +#define SUBTARGET_EXTRA_SPECS +#endif + +#define TARGET_DEFAULT_CMODEL CM_MEDLOW + +/* By default, turn on GDB extensions. */ +#define DEFAULT_GDB_EXTENSIONS 1 + +#define LOCAL_LABEL_PREFIX "." +#define USER_LABEL_PREFIX "" + +#define DWARF2_DEBUGGING_INFO 1 +#define DWARF2_ASM_LINE_DEBUG_INFO 0 + +/* The mapping from gcc register number to DWARF 2 CFA column number. */ +#define DWARF_FRAME_REGNUM(REGNO) \ + (GP_REG_P (REGNO) || FP_REG_P (REGNO) ? REGNO : INVALID_REGNUM) + +/* The DWARF 2 CFA column which tracks the return address. */ +#define DWARF_FRAME_RETURN_COLUMN RETURN_ADDR_REGNUM + +/* Don't emit .cfi_sections, as it does not work */ +#undef HAVE_GAS_CFI_SECTIONS_DIRECTIVE +#define HAVE_GAS_CFI_SECTIONS_DIRECTIVE 0 + +/* Before the prologue, RA lives in r31. */ +#define INCOMING_RETURN_ADDR_RTX gen_rtx_REG (VOIDmode, RETURN_ADDR_REGNUM) + +/* Describe how we implement __builtin_eh_return. */ +#define EH_RETURN_DATA_REGNO(N) \ + ((N) < 4 ? (N) + GP_ARG_FIRST : INVALID_REGNUM) + +#define EH_RETURN_STACKADJ_RTX gen_rtx_REG (Pmode, GP_ARG_FIRST + 4) + +/* Target machine storage layout */ + +#define BITS_BIG_ENDIAN 0 +#define BYTES_BIG_ENDIAN 0 +#define WORDS_BIG_ENDIAN 0 + +#define MAX_BITS_PER_WORD 64 + +/* Width of a word, in units (bytes). */ +#define UNITS_PER_WORD (TARGET_64BIT ? 8 : 4) +#ifndef IN_LIBGCC2 +#define MIN_UNITS_PER_WORD 4 +#endif + +/* We currently require both or neither of the `F' and `D' extensions. */ +#define UNITS_PER_FPREG 8 + +/* If FP regs aren't wide enough for a given FP argument, it is passed in + integer registers. */ +#define MIN_FPRS_PER_FMT 1 + +/* The largest size of value that can be held in floating-point + registers and moved with a single instruction. */ +#define UNITS_PER_HWFPVALUE \ + (TARGET_SOFT_FLOAT_ABI ? 0 : UNITS_PER_FPREG) + +/* The largest size of value that can be held in floating-point + registers. */ +#define UNITS_PER_FPVALUE \ + (TARGET_SOFT_FLOAT_ABI ? 0 \ + : LONG_DOUBLE_TYPE_SIZE / BITS_PER_UNIT) + +/* The number of bytes in a double. */ +#define UNITS_PER_DOUBLE (TYPE_PRECISION (double_type_node) / BITS_PER_UNIT) + +/* Set the sizes of the core types. */ +#define SHORT_TYPE_SIZE 16 +#define INT_TYPE_SIZE 32 +#define LONG_TYPE_SIZE (TARGET_64BIT ? 64 : 32) +#define LONG_LONG_TYPE_SIZE 64 + +#define FLOAT_TYPE_SIZE 32 +#define DOUBLE_TYPE_SIZE 64 +/* XXX The ABI says long doubles are IEEE-754-2008 float128s. */ +#define LONG_DOUBLE_TYPE_SIZE 64 + +#ifdef IN_LIBGCC2 +# define LIBGCC2_LONG_DOUBLE_TYPE_SIZE LONG_DOUBLE_TYPE_SIZE +#endif + +/* Allocation boundary (in *bits*) for storing arguments in argument list. */ +#define PARM_BOUNDARY BITS_PER_WORD + +/* Allocation boundary (in *bits*) for the code of a function. */ +#define FUNCTION_BOUNDARY (TARGET_RVC ? 16 : 32) + +/* There is no point aligning anything to a rounder boundary than this. */ +#define BIGGEST_ALIGNMENT 128 + +/* All accesses must be aligned. */ +#define STRICT_ALIGNMENT 1 + +/* Define this if you wish to imitate the way many other C compilers + handle alignment of bitfields and the structures that contain + them. + + The behavior is that the type written for a bit-field (`int', + `short', or other integer type) imposes an alignment for the + entire structure, as if the structure really did contain an + ordinary field of that type. In addition, the bit-field is placed + within the structure so that it would fit within such a field, + not crossing a boundary for it. + + Thus, on most machines, a bit-field whose type is written as `int' + would not cross a four-byte boundary, and would force four-byte + alignment for the whole structure. (The alignment used may not + be four bytes; it is controlled by the other alignment + parameters.) + + If the macro is defined, its definition should be a C expression; + a nonzero value for the expression enables this behavior. */ + +#define PCC_BITFIELD_TYPE_MATTERS 1 + +/* If defined, a C expression to compute the alignment given to a + constant that is being placed in memory. CONSTANT is the constant + and ALIGN is the alignment that the object would ordinarily have. + The value of this macro is used instead of that alignment to align + the object. + + If this macro is not defined, then ALIGN is used. + + The typical use of this macro is to increase alignment for string + constants to be word aligned so that `strcpy' calls that copy + constants can be done inline. */ + +#define CONSTANT_ALIGNMENT(EXP, ALIGN) \ + ((TREE_CODE (EXP) == STRING_CST || TREE_CODE (EXP) == CONSTRUCTOR) \ + && (ALIGN) < BITS_PER_WORD ? BITS_PER_WORD : (ALIGN)) + +/* If defined, a C expression to compute the alignment for a static + variable. TYPE is the data type, and ALIGN is the alignment that + the object would ordinarily have. The value of this macro is used + instead of that alignment to align the object. + + If this macro is not defined, then ALIGN is used. + + One use of this macro is to increase alignment of medium-size + data to make it all fit in fewer cache lines. Another is to + cause character arrays to be word-aligned so that `strcpy' calls + that copy constants to character arrays can be done inline. */ + +#undef DATA_ALIGNMENT +#define DATA_ALIGNMENT(TYPE, ALIGN) \ + ((((ALIGN) < BITS_PER_WORD) \ + && (TREE_CODE (TYPE) == ARRAY_TYPE \ + || TREE_CODE (TYPE) == UNION_TYPE \ + || TREE_CODE (TYPE) == RECORD_TYPE)) ? BITS_PER_WORD : (ALIGN)) + +/* We need this for the same reason as DATA_ALIGNMENT, namely to cause + character arrays to be word-aligned so that `strcpy' calls that copy + constants to character arrays can be done inline, and 'strcmp' can be + optimised to use word loads. */ +#define LOCAL_ALIGNMENT(TYPE, ALIGN) \ + DATA_ALIGNMENT (TYPE, ALIGN) + +/* Define if operations between registers always perform the operation + on the full register even if a narrower mode is specified. */ +#define WORD_REGISTER_OPERATIONS + +/* When in 64-bit mode, move insns will sign extend SImode and CCmode + moves. All other references are zero extended. */ +#define LOAD_EXTEND_OP(MODE) \ + (TARGET_64BIT && ((MODE) == SImode || (MODE) == CCmode) \ + ? SIGN_EXTEND : ZERO_EXTEND) + +/* Define this macro if it is advisable to hold scalars in registers + in a wider mode than that declared by the program. In such cases, + the value is constrained to be within the bounds of the declared + type, but kept valid in the wider mode. The signedness of the + extension may differ from that of the type. */ + +#define PROMOTE_MODE(MODE, UNSIGNEDP, TYPE) \ + if (GET_MODE_CLASS (MODE) == MODE_INT \ + && GET_MODE_SIZE (MODE) < 4) \ + { \ + (MODE) = Pmode; \ + } + +/* Pmode is always the same as ptr_mode, but not always the same as word_mode. + Extensions of pointers to word_mode must be signed. */ +#define POINTERS_EXTEND_UNSIGNED false + +/* RV32 double-precision FP <-> integer moves go through memory */ +#define SECONDARY_MEMORY_NEEDED(CLASS1,CLASS2,MODE) \ + (!TARGET_64BIT && GET_MODE_SIZE (MODE) == 8 && \ + (((CLASS1) == FP_REGS && (CLASS2) != FP_REGS) \ + || ((CLASS2) == FP_REGS && (CLASS1) != FP_REGS))) + +/* Define if loading short immediate values into registers sign extends. */ +#define SHORT_IMMEDIATES_SIGN_EXTEND + +/* Standard register usage. */ + +/* Number of hardware registers. We have: + + - 32 integer registers + - 32 floating point registers + - 32 vector integer registers + - 32 vector floating point registers + - 2 fake registers: + - ARG_POINTER_REGNUM + - FRAME_POINTER_REGNUM */ + +#define FIRST_PSEUDO_REGISTER 66 + +/* x0, sp, gp, and tp are fixed. */ + +#define FIXED_REGISTERS \ +{ /* General registers. */ \ + 1, 0, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, \ + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, \ + /* Floating-point registers. */ \ + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, \ + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, \ + /* Others. */ \ + 1, 1 \ +} + + +/* a0-a7, t0-a6, fa0-fa7, and ft0-ft11 are volatile across calls. + The call RTLs themselves clobber ra. */ + +#define CALL_USED_REGISTERS \ +{ /* General registers. */ \ + 1, 0, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, \ + 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, \ + /* Floating-point registers. */ \ + 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, \ + 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, \ + /* Others. */ \ + 1, 1 \ +} + +#define CALL_REALLY_USED_REGISTERS \ +{ /* General registers. */ \ + 1, 0, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, \ + 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, \ + /* Floating-point registers. */ \ + 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, \ + 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, \ + /* Others. */ \ + 1, 1 \ +} + +/* Internal macros to classify an ISA register's type. */ + +#define GP_REG_FIRST 0 +#define GP_REG_LAST 31 +#define GP_REG_NUM (GP_REG_LAST - GP_REG_FIRST + 1) + +#define FP_REG_FIRST 32 +#define FP_REG_LAST 63 +#define FP_REG_NUM (FP_REG_LAST - FP_REG_FIRST + 1) + +/* The DWARF 2 CFA column which tracks the return address from a + signal handler context. This means that to maintain backwards + compatibility, no hard register can be assigned this column if it + would need to be handled by the DWARF unwinder. */ +#define DWARF_ALT_FRAME_RETURN_COLUMN 64 + +#define GP_REG_P(REGNO) \ + ((unsigned int) ((int) (REGNO) - GP_REG_FIRST) < GP_REG_NUM) +#define FP_REG_P(REGNO) \ + ((unsigned int) ((int) (REGNO) - FP_REG_FIRST) < FP_REG_NUM) + +#define FP_REG_RTX_P(X) (REG_P (X) && FP_REG_P (REGNO (X))) + +/* Return coprocessor number from register number. */ + +#define COPNUM_AS_CHAR_FROM_REGNUM(REGNO) \ + (COP0_REG_P (REGNO) ? '0' : COP2_REG_P (REGNO) ? '2' \ + : COP3_REG_P (REGNO) ? '3' : '?') + + +#define HARD_REGNO_NREGS(REGNO, MODE) riscv_hard_regno_nregs (REGNO, MODE) + +#define HARD_REGNO_MODE_OK(REGNO, MODE) \ + riscv_hard_regno_mode_ok[ (int)(MODE) ][ (REGNO) ] + +#define MODES_TIEABLE_P(MODE1, MODE2) \ + ((MODE1) == (MODE2) || (GET_MODE_CLASS (MODE1) == MODE_INT \ + && GET_MODE_CLASS (MODE2) == MODE_INT)) + +/* Use s0 as the frame pointer if it is so requested. */ +#define HARD_FRAME_POINTER_REGNUM 8 +#define STACK_POINTER_REGNUM 2 +#define THREAD_POINTER_REGNUM 4 + +/* These two registers don't really exist: they get eliminated to either + the stack or hard frame pointer. */ +#define ARG_POINTER_REGNUM 64 +#define FRAME_POINTER_REGNUM 65 + +#define HARD_FRAME_POINTER_IS_FRAME_POINTER 0 +#define HARD_FRAME_POINTER_IS_ARG_POINTER 0 + +/* Register in which static-chain is passed to a function. */ +#define STATIC_CHAIN_REGNUM GP_TEMP_FIRST + +/* Registers used as temporaries in prologue/epilogue code. + + The prologue registers mustn't conflict with any + incoming arguments, the static chain pointer, or the frame pointer. + The epilogue temporary mustn't conflict with the return registers, + the frame pointer, the EH stack adjustment, or the EH data registers. */ + +#define RISCV_PROLOGUE_TEMP_REGNUM (GP_TEMP_FIRST + 1) +#define RISCV_EPILOGUE_TEMP_REGNUM RISCV_PROLOGUE_TEMP_REGNUM + +#define RISCV_PROLOGUE_TEMP(MODE) gen_rtx_REG (MODE, RISCV_PROLOGUE_TEMP_REGNUM) +#define RISCV_EPILOGUE_TEMP(MODE) gen_rtx_REG (MODE, RISCV_EPILOGUE_TEMP_REGNUM) + +#define FUNCTION_PROFILER(STREAM, LABELNO) \ +{ \ + sorry ("profiler support for RISC-V"); \ +} + +/* Define this macro if it is as good or better to call a constant + function address than to call an address kept in a register. */ +#define NO_FUNCTION_CSE 1 + +/* Define the classes of registers for register constraints in the + machine description. Also define ranges of constants. + + One of the classes must always be named ALL_REGS and include all hard regs. + If there is more than one class, another class must be named NO_REGS + and contain no registers. + + The name GENERAL_REGS must be the name of a class (or an alias for + another name such as ALL_REGS). This is the class of registers + that is allowed by "g" or "r" in a register constraint. + Also, registers outside this class are allocated only when + instructions express preferences for them. + + The classes must be numbered in nondecreasing order; that is, + a larger-numbered class must never be contained completely + in a smaller-numbered class. + + For any two classes, it is very desirable that there be another + class that represents their union. */ + +enum reg_class +{ + NO_REGS, /* no registers in set */ + T_REGS, /* registers used by indirect sibcalls */ + JALR_REGS, /* registers used by indirect calls */ + GR_REGS, /* integer registers */ + FP_REGS, /* floating point registers */ + FRAME_REGS, /* $arg and $frame */ + ALL_REGS, /* all registers */ + LIM_REG_CLASSES /* max value + 1 */ +}; + +#define N_REG_CLASSES (int) LIM_REG_CLASSES + +#define GENERAL_REGS GR_REGS + +/* An initializer containing the names of the register classes as C + string constants. These names are used in writing some of the + debugging dumps. */ + +#define REG_CLASS_NAMES \ +{ \ + "NO_REGS", \ + "T_REGS", \ + "JALR_REGS", \ + "GR_REGS", \ + "FP_REGS", \ + "FRAME_REGS", \ + "ALL_REGS" \ +} + +/* An initializer containing the contents of the register classes, + as integers which are bit masks. The Nth integer specifies the + contents of class N. The way the integer MASK is interpreted is + that register R is in the class if `MASK & (1 << R)' is 1. + + When the machine has more than 32 registers, an integer does not + suffice. Then the integers are replaced by sub-initializers, + braced groupings containing several integers. Each + sub-initializer must be suitable as an initializer for the type + `HARD_REG_SET' which is defined in `hard-reg-set.h'. */ + +#define REG_CLASS_CONTENTS \ +{ \ + { 0x00000000, 0x00000000, 0x00000000 }, /* NO_REGS */ \ + { 0xf0000040, 0x00000000, 0x00000000 }, /* T_REGS */ \ + { 0xffffff40, 0x00000000, 0x00000000 }, /* JALR_REGS */ \ + { 0xffffffff, 0x00000000, 0x00000000 }, /* GR_REGS */ \ + { 0x00000000, 0xffffffff, 0x00000000 }, /* FP_REGS */ \ + { 0x00000000, 0x00000000, 0x00000003 }, /* FRAME_REGS */ \ + { 0xffffffff, 0xffffffff, 0x00000003 } /* ALL_REGS */ \ +} + +/* A C expression whose value is a register class containing hard + register REGNO. In general there is more that one such class; + choose a class which is "minimal", meaning that no smaller class + also contains the register. */ + +#define REGNO_REG_CLASS(REGNO) riscv_regno_to_class[ (REGNO) ] + +/* A macro whose definition is the name of the class to which a + valid base register must belong. A base register is one used in + an address which is the register value plus a displacement. */ + +#define BASE_REG_CLASS GR_REGS + +/* A macro whose definition is the name of the class to which a + valid index register must belong. An index register is one used + in an address where its value is either multiplied by a scale + factor or added to another register (as well as added to a + displacement). */ + +#define INDEX_REG_CLASS NO_REGS + +/* We generally want to put call-clobbered registers ahead of + call-saved ones. (IRA expects this.) */ + +#define REG_ALLOC_ORDER \ +{ \ + /* Call-clobbered GPRs. */ \ + 15, 14, 13, 12, 11, 10, 16, 17, 5, 6, 7, 28, 29, 30, 31, 1, \ + /* Call-saved GPRs. */ \ + 8, 9, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, \ + /* GPRs that can never be exposed to the register allocator. */ \ + 0, 2, 3, 4, \ + /* Call-clobbered FPRs. */ \ + 32, 33, 34, 35, 36, 37, 38, 39, 42, 43, 44, 45, 46, 47, 48, 49, \ + 60, 61, 62, 63, \ + /* Call-saved FPRs. */ \ + 40, 41, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, \ + /* None of the remaining classes have defined call-saved \ + registers. */ \ + 64, 65 \ +} + +/* True if VALUE is a signed 16-bit number. */ + +#include "opcode-riscv.h" +#define SMALL_OPERAND(VALUE) \ + ((unsigned HOST_WIDE_INT) (VALUE) + RISCV_IMM_REACH/2 < RISCV_IMM_REACH) + +/* True if VALUE can be loaded into a register using LUI. */ + +#define LUI_OPERAND(VALUE) \ + (((VALUE) | ((1UL<<31) - RISCV_IMM_REACH)) == ((1UL<<31) - RISCV_IMM_REACH) \ + || ((VALUE) | ((1UL<<31) - RISCV_IMM_REACH)) + RISCV_IMM_REACH == 0) + +/* Return a value X with the low 16 bits clear, and such that + VALUE - X is a signed 16-bit value. */ + +#define SMALL_INT(X) SMALL_OPERAND (INTVAL (X)) +#define LUI_INT(X) LUI_OPERAND (INTVAL (X)) + +/* The HI and LO registers can only be reloaded via the general + registers. Condition code registers can only be loaded to the + general registers, and from the floating point registers. */ + +#define SECONDARY_INPUT_RELOAD_CLASS(CLASS, MODE, X) \ + riscv_secondary_reload_class (CLASS, MODE, X, true) +#define SECONDARY_OUTPUT_RELOAD_CLASS(CLASS, MODE, X) \ + riscv_secondary_reload_class (CLASS, MODE, X, false) + +/* Return the maximum number of consecutive registers + needed to represent mode MODE in a register of class CLASS. */ + +#define CLASS_MAX_NREGS(CLASS, MODE) riscv_class_max_nregs (CLASS, MODE) + +/* It is undefined to interpret an FP register in a different format than + that which it was created to be. */ + +#define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \ + reg_classes_intersect_p (FP_REGS, CLASS) + +/* Stack layout; function entry, exit and calling. */ + +#define STACK_GROWS_DOWNWARD + +#define FRAME_GROWS_DOWNWARD 1 + +#define STARTING_FRAME_OFFSET 0 + +#define RETURN_ADDR_RTX riscv_return_addr + +#define ELIMINABLE_REGS \ +{{ ARG_POINTER_REGNUM, STACK_POINTER_REGNUM}, \ + { ARG_POINTER_REGNUM, HARD_FRAME_POINTER_REGNUM}, \ + { FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM}, \ + { FRAME_POINTER_REGNUM, HARD_FRAME_POINTER_REGNUM}} \ + +#define INITIAL_ELIMINATION_OFFSET(FROM, TO, OFFSET) \ + (OFFSET) = riscv_initial_elimination_offset (FROM, TO) + +/* Allocate stack space for arguments at the beginning of each function. */ +#define ACCUMULATE_OUTGOING_ARGS 1 + +/* The argument pointer always points to the first argument. */ +#define FIRST_PARM_OFFSET(FNDECL) 0 + +#define REG_PARM_STACK_SPACE(FNDECL) 0 + +/* Define this if it is the responsibility of the caller to + allocate the area reserved for arguments passed in registers. + If `ACCUMULATE_OUTGOING_ARGS' is also defined, the only effect + of this macro is to determine whether the space is included in + `crtl->outgoing_args_size'. */ +#define OUTGOING_REG_PARM_STACK_SPACE(FNTYPE) 1 + +#define STACK_BOUNDARY 128 + +/* Symbolic macros for the registers used to return integer and floating + point values. */ + +#define GP_RETURN GP_ARG_FIRST +#define FP_RETURN ((TARGET_SOFT_FLOAT) ? GP_RETURN : FP_ARG_FIRST) + +#define MAX_ARGS_IN_REGISTERS 8 + +/* Symbolic macros for the first/last argument registers. */ + +#define GP_ARG_FIRST (GP_REG_FIRST + 10) +#define GP_ARG_LAST (GP_ARG_FIRST + MAX_ARGS_IN_REGISTERS - 1) +#define GP_TEMP_FIRST (GP_REG_FIRST + 5) +#define FP_ARG_FIRST (FP_REG_FIRST + 10) +#define FP_ARG_LAST (FP_ARG_FIRST + MAX_ARGS_IN_REGISTERS - 1) + +#define CALLEE_SAVED_REG_NUMBER(REGNO) \ + ((REGNO) >= 8 && (REGNO) <= 9 ? (REGNO) - 8 : \ + (REGNO) >= 18 && (REGNO) <= 27 ? (REGNO) - 16 : -1) + +#define LIBCALL_VALUE(MODE) \ + riscv_function_value (NULL_TREE, NULL_TREE, MODE) + +#define FUNCTION_VALUE(VALTYPE, FUNC) \ + riscv_function_value (VALTYPE, FUNC, VOIDmode) + +#define FUNCTION_VALUE_REGNO_P(N) ((N) == GP_RETURN || (N) == FP_RETURN) + +/* 1 if N is a possible register number for function argument passing. + We have no FP argument registers when soft-float. When FP registers + are 32 bits, we can't directly reference the odd numbered ones. */ + +/* Accept arguments in a0-a7 and/or fa0-fa7. */ +#define FUNCTION_ARG_REGNO_P(N) \ + (IN_RANGE((N), GP_ARG_FIRST, GP_ARG_LAST) \ + || IN_RANGE((N), FP_ARG_FIRST, FP_ARG_LAST)) + +/* The ABI views the arguments as a structure, of which the first 8 + words go in registers and the rest go on the stack. If I < 8, N, the Ith + word might go in the Ith integer argument register or the Ith + floating-point argument register. */ + +typedef struct { + /* Number of integer registers used so far, up to MAX_ARGS_IN_REGISTERS. */ + unsigned int num_gprs; + + /* Number of words passed on the stack. */ + unsigned int stack_words; +} CUMULATIVE_ARGS; + +/* Initialize a variable CUM of type CUMULATIVE_ARGS + for a call to a function whose data type is FNTYPE. + For a library call, FNTYPE is 0. */ + +#define INIT_CUMULATIVE_ARGS(CUM, FNTYPE, LIBNAME, INDIRECT, N_NAMED_ARGS) \ + memset (&(CUM), 0, sizeof (CUM)) + +#define EPILOGUE_USES(REGNO) ((REGNO) == RETURN_ADDR_REGNUM) + +/* ABI requires 16-byte alignment, even on ven on RV32. */ +#define RISCV_STACK_ALIGN(LOC) (((LOC) + 15) & -16) + +#define NO_PROFILE_COUNTERS 1 + +/* Define this macro if the code for function profiling should come + before the function prologue. Normally, the profiling code comes + after. */ + +/* #define PROFILE_BEFORE_PROLOGUE */ + +/* EXIT_IGNORE_STACK should be nonzero if, when returning from a function, + the stack pointer does not matter. The value is tested only in + functions that have frame pointers. + No definition is equivalent to always zero. */ + +#define EXIT_IGNORE_STACK 1 + + +/* Trampolines are a block of code followed by two pointers. */ + +#define TRAMPOLINE_CODE_SIZE 16 +#define TRAMPOLINE_SIZE (TRAMPOLINE_CODE_SIZE + POINTER_SIZE * 2) +#define TRAMPOLINE_ALIGNMENT POINTER_SIZE + +/* Addressing modes, and classification of registers for them. */ + +#define REGNO_OK_FOR_INDEX_P(REGNO) 0 +#define REGNO_MODE_OK_FOR_BASE_P(REGNO, MODE) \ + riscv_regno_mode_ok_for_base_p (REGNO, MODE, 1) + +/* The macros REG_OK_FOR..._P assume that the arg is a REG rtx + and check its validity for a certain class. + We have two alternate definitions for each of them. + The usual definition accepts all pseudo regs; the other rejects them all. + The symbol REG_OK_STRICT causes the latter definition to be used. + + Most source files want to accept pseudo regs in the hope that + they will get allocated to the class that the insn wants them to be in. + Some source files that are used after register allocation + need to be strict. */ + +#ifndef REG_OK_STRICT +#define REG_MODE_OK_FOR_BASE_P(X, MODE) \ + riscv_regno_mode_ok_for_base_p (REGNO (X), MODE, 0) +#else +#define REG_MODE_OK_FOR_BASE_P(X, MODE) \ + riscv_regno_mode_ok_for_base_p (REGNO (X), MODE, 1) +#endif + +#define REG_OK_FOR_INDEX_P(X) 0 + + +/* Maximum number of registers that can appear in a valid memory address. */ + +#define MAX_REGS_PER_ADDRESS 1 + +#define CONSTANT_ADDRESS_P(X) \ + (CONSTANT_P (X) && memory_address_p (SImode, X)) + +/* This handles the magic '..CURRENT_FUNCTION' symbol, which means + 'the start of the function that this code is output in'. */ + +#define ASM_OUTPUT_LABELREF(FILE,NAME) \ + if (strcmp (NAME, "..CURRENT_FUNCTION") == 0) \ + asm_fprintf ((FILE), "%U%s", \ + XSTR (XEXP (DECL_RTL (current_function_decl), 0), 0)); \ + else \ + asm_fprintf ((FILE), "%U%s", (NAME)) + +/* This flag marks functions that cannot be lazily bound. */ +#define SYMBOL_FLAG_BIND_NOW (SYMBOL_FLAG_MACH_DEP << 1) +#define SYMBOL_REF_BIND_NOW_P(RTX) \ + ((SYMBOL_REF_FLAGS (RTX) & SYMBOL_FLAG_BIND_NOW) != 0) + +#define JUMP_TABLES_IN_TEXT_SECTION 0 +#define CASE_VECTOR_MODE SImode +#define CASE_VECTOR_PC_RELATIVE (riscv_cmodel != CM_MEDLOW) + +/* Define this as 1 if `char' should by default be signed; else as 0. */ +#define DEFAULT_SIGNED_CHAR 0 + +/* Consider using fld/fsd to move 8 bytes at a time for RV32IFD. */ +#define MOVE_MAX UNITS_PER_WORD +#define MAX_MOVE_MAX 8 + +#define SLOW_BYTE_ACCESS 0 + +#define SHIFT_COUNT_TRUNCATED 1 + +/* Value is 1 if truncating an integer of INPREC bits to OUTPREC bits + is done just by pretending it is already truncated. */ +#define TRULY_NOOP_TRUNCATION(OUTPREC, INPREC) \ + (TARGET_64BIT ? ((INPREC) <= 32 || (OUTPREC) < 32) : 1) + +/* Specify the machine mode that pointers have. + After generation of rtl, the compiler makes no further distinction + between pointers and any other objects of this machine mode. */ + +#ifndef Pmode +#define Pmode (TARGET_64BIT ? DImode : SImode) +#endif + +/* Give call MEMs SImode since it is the "most permissive" mode + for both 32-bit and 64-bit targets. */ + +#define FUNCTION_MODE SImode + +/* A C expression for the cost of a branch instruction. A value of 2 + seems to minimize code size. */ + +#define BRANCH_COST(speed_p, predictable_p) \ + ((!(speed_p) || (predictable_p)) ? 2 : riscv_branch_cost) + +#define LOGICAL_OP_NON_SHORT_CIRCUIT 0 + +/* Control the assembler format that we output. */ + +/* Output to assembler file text saying following lines + may contain character constants, extra white space, comments, etc. */ + +#ifndef ASM_APP_ON +#define ASM_APP_ON " #APP\n" +#endif + +/* Output to assembler file text saying following lines + no longer contain unusual constructs. */ + +#ifndef ASM_APP_OFF +#define ASM_APP_OFF " #NO_APP\n" +#endif + +#define REGISTER_NAMES \ +{ "zero","ra", "sp", "gp", "tp", "t0", "t1", "t2", \ + "s0", "s1", "a0", "a1", "a2", "a3", "a4", "a5", \ + "a6", "a7", "s2", "s3", "s4", "s5", "s6", "s7", \ + "s8", "s9", "s10", "s11", "t3", "t4", "t5", "t6", \ + "ft0", "ft1", "ft2", "ft3", "ft4", "ft5", "ft6", "ft7", \ + "fs0", "fs1", "fa0", "fa1", "fa2", "fa3", "fa4", "fa5", \ + "fa6", "fa7", "fs2", "fs3", "fs4", "fs5", "fs6", "fs7", \ + "fs8", "fs9", "fs10","fs11","ft8", "ft9", "ft10","ft11", \ + "arg", "frame", } + +#define ADDITIONAL_REGISTER_NAMES \ +{ \ + { "x0", 0 + GP_REG_FIRST }, \ + { "x1", 1 + GP_REG_FIRST }, \ + { "x2", 2 + GP_REG_FIRST }, \ + { "x3", 3 + GP_REG_FIRST }, \ + { "x4", 4 + GP_REG_FIRST }, \ + { "x5", 5 + GP_REG_FIRST }, \ + { "x6", 6 + GP_REG_FIRST }, \ + { "x7", 7 + GP_REG_FIRST }, \ + { "x8", 8 + GP_REG_FIRST }, \ + { "x9", 9 + GP_REG_FIRST }, \ + { "x10", 10 + GP_REG_FIRST }, \ + { "x11", 11 + GP_REG_FIRST }, \ + { "x12", 12 + GP_REG_FIRST }, \ + { "x13", 13 + GP_REG_FIRST }, \ + { "x14", 14 + GP_REG_FIRST }, \ + { "x15", 15 + GP_REG_FIRST }, \ + { "x16", 16 + GP_REG_FIRST }, \ + { "x17", 17 + GP_REG_FIRST }, \ + { "x18", 18 + GP_REG_FIRST }, \ + { "x19", 19 + GP_REG_FIRST }, \ + { "x20", 20 + GP_REG_FIRST }, \ + { "x21", 21 + GP_REG_FIRST }, \ + { "x22", 22 + GP_REG_FIRST }, \ + { "x23", 23 + GP_REG_FIRST }, \ + { "x24", 24 + GP_REG_FIRST }, \ + { "x25", 25 + GP_REG_FIRST }, \ + { "x26", 26 + GP_REG_FIRST }, \ + { "x27", 27 + GP_REG_FIRST }, \ + { "x28", 28 + GP_REG_FIRST }, \ + { "x29", 29 + GP_REG_FIRST }, \ + { "x30", 30 + GP_REG_FIRST }, \ + { "x31", 31 + GP_REG_FIRST }, \ + { "f0", 0 + FP_REG_FIRST }, \ + { "f1", 1 + FP_REG_FIRST }, \ + { "f2", 2 + FP_REG_FIRST }, \ + { "f3", 3 + FP_REG_FIRST }, \ + { "f4", 4 + FP_REG_FIRST }, \ + { "f5", 5 + FP_REG_FIRST }, \ + { "f6", 6 + FP_REG_FIRST }, \ + { "f7", 7 + FP_REG_FIRST }, \ + { "f8", 8 + FP_REG_FIRST }, \ + { "f9", 9 + FP_REG_FIRST }, \ + { "f10", 10 + FP_REG_FIRST }, \ + { "f11", 11 + FP_REG_FIRST }, \ + { "f12", 12 + FP_REG_FIRST }, \ + { "f13", 13 + FP_REG_FIRST }, \ + { "f14", 14 + FP_REG_FIRST }, \ + { "f15", 15 + FP_REG_FIRST }, \ + { "f16", 16 + FP_REG_FIRST }, \ + { "f17", 17 + FP_REG_FIRST }, \ + { "f18", 18 + FP_REG_FIRST }, \ + { "f19", 19 + FP_REG_FIRST }, \ + { "f20", 20 + FP_REG_FIRST }, \ + { "f21", 21 + FP_REG_FIRST }, \ + { "f22", 22 + FP_REG_FIRST }, \ + { "f23", 23 + FP_REG_FIRST }, \ + { "f24", 24 + FP_REG_FIRST }, \ + { "f25", 25 + FP_REG_FIRST }, \ + { "f26", 26 + FP_REG_FIRST }, \ + { "f27", 27 + FP_REG_FIRST }, \ + { "f28", 28 + FP_REG_FIRST }, \ + { "f29", 29 + FP_REG_FIRST }, \ + { "f30", 30 + FP_REG_FIRST }, \ + { "f31", 31 + FP_REG_FIRST }, \ +} + +/* Globalizing directive for a label. */ +#define GLOBAL_ASM_OP "\t.globl\t" + +/* This is how to store into the string LABEL + the symbol_ref name of an internal numbered label where + PREFIX is the class of label and NUM is the number within the class. + This is suitable for output with `assemble_name'. */ + +#undef ASM_GENERATE_INTERNAL_LABEL +#define ASM_GENERATE_INTERNAL_LABEL(LABEL,PREFIX,NUM) \ + sprintf ((LABEL), "*%s%s%ld", (LOCAL_LABEL_PREFIX), (PREFIX), (long)(NUM)) + +/* This is how to output an element of a case-vector that is absolute. */ + +#define ASM_OUTPUT_ADDR_VEC_ELT(STREAM, VALUE) \ + fprintf (STREAM, "\t.word\t%sL%d\n", LOCAL_LABEL_PREFIX, VALUE) + +/* This is how to output an element of a PIC case-vector. */ + +#define ASM_OUTPUT_ADDR_DIFF_ELT(STREAM, BODY, VALUE, REL) \ + fprintf (STREAM, "\t.word\t%sL%d-%sL%d\n", \ + LOCAL_LABEL_PREFIX, VALUE, LOCAL_LABEL_PREFIX, REL) + +/* This is how to output an assembler line + that says to advance the location counter + to a multiple of 2**LOG bytes. */ + +#define ASM_OUTPUT_ALIGN(STREAM,LOG) \ + fprintf (STREAM, "\t.align\t%d\n", (LOG)) + +/* Define the strings to put out for each section in the object file. */ +#define TEXT_SECTION_ASM_OP "\t.text" /* instructions */ +#define DATA_SECTION_ASM_OP "\t.data" /* large data */ +#define READONLY_DATA_SECTION_ASM_OP "\t.section\t.rodata" +#define BSS_SECTION_ASM_OP "\t.bss" +#define SBSS_SECTION_ASM_OP "\t.section\t.sbss,\"aw\",@nobits" +#define SDATA_SECTION_ASM_OP "\t.section\t.sdata,\"aw\",@progbits" + +#define ASM_OUTPUT_REG_PUSH(STREAM,REGNO) \ +do \ + { \ + fprintf (STREAM, "\taddi\t%s,%s,-8\n\t%s\t%s,0(%s)\n", \ + reg_names[STACK_POINTER_REGNUM], \ + reg_names[STACK_POINTER_REGNUM], \ + TARGET_64BIT ? "sd" : "sw", \ + reg_names[REGNO], \ + reg_names[STACK_POINTER_REGNUM]); \ + } \ +while (0) + +#define ASM_OUTPUT_REG_POP(STREAM,REGNO) \ +do \ + { \ + fprintf (STREAM, "\t%s\t%s,0(%s)\n\taddi\t%s,%s,8\n", \ + TARGET_64BIT ? "ld" : "lw", \ + reg_names[REGNO], \ + reg_names[STACK_POINTER_REGNUM], \ + reg_names[STACK_POINTER_REGNUM], \ + reg_names[STACK_POINTER_REGNUM]); \ + } \ +while (0) + +#define ASM_COMMENT_START "#" + +#undef SIZE_TYPE +#define SIZE_TYPE (POINTER_SIZE == 64 ? "long unsigned int" : "unsigned int") + +#undef PTRDIFF_TYPE +#define PTRDIFF_TYPE (POINTER_SIZE == 64 ? "long int" : "int") + +/* The maximum number of bytes that can be copied by one iteration of + a movmemsi loop; see riscv_block_move_loop. */ +#define RISCV_MAX_MOVE_BYTES_PER_LOOP_ITER (UNITS_PER_WORD * 4) + +/* The maximum number of bytes that can be copied by a straight-line + implementation of movmemsi; see riscv_block_move_straight. We want + to make sure that any loop-based implementation will iterate at + least twice. */ +#define RISCV_MAX_MOVE_BYTES_STRAIGHT (RISCV_MAX_MOVE_BYTES_PER_LOOP_ITER * 2) + +/* The base cost of a memcpy call, for MOVE_RATIO and friends. */ + +#define RISCV_CALL_RATIO 6 + +/* Any loop-based implementation of movmemsi will have at least + RISCV_MAX_MOVE_BYTES_STRAIGHT / UNITS_PER_WORD memory-to-memory + moves, so allow individual copies of fewer elements. + + When movmemsi is not available, use a value approximating + the length of a memcpy call sequence, so that move_by_pieces + will generate inline code if it is shorter than a function call. + Since move_by_pieces_ninsns counts memory-to-memory moves, but + we'll have to generate a load/store pair for each, halve the + value of RISCV_CALL_RATIO to take that into account. */ + +#define MOVE_RATIO(speed) \ + (HAVE_movmemsi \ + ? RISCV_MAX_MOVE_BYTES_STRAIGHT / MOVE_MAX \ + : RISCV_CALL_RATIO / 2) + +/* movmemsi is meant to generate code that is at least as good as + move_by_pieces. However, movmemsi effectively uses a by-pieces + implementation both for moves smaller than a word and for word-aligned + moves of no more than RISCV_MAX_MOVE_BYTES_STRAIGHT bytes. We should + allow the tree-level optimisers to do such moves by pieces, as it + often exposes other optimization opportunities. We might as well + continue to use movmemsi at the rtl level though, as it produces + better code when scheduling is disabled (such as at -O). */ + +#define MOVE_BY_PIECES_P(SIZE, ALIGN) \ + (HAVE_movmemsi \ + ? (!currently_expanding_to_rtl \ + && ((ALIGN) < BITS_PER_WORD \ + ? (SIZE) < UNITS_PER_WORD \ + : (SIZE) <= RISCV_MAX_MOVE_BYTES_STRAIGHT)) \ + : (move_by_pieces_ninsns (SIZE, ALIGN, MOVE_MAX_PIECES + 1) \ + < (unsigned int) MOVE_RATIO (false))) + +/* For CLEAR_RATIO, when optimizing for size, give a better estimate + of the length of a memset call, but use the default otherwise. */ + +#define CLEAR_RATIO(speed)\ + ((speed) ? 15 : RISCV_CALL_RATIO) + +/* This is similar to CLEAR_RATIO, but for a non-zero constant, so when + optimizing for size adjust the ratio to account for the overhead of + loading the constant and replicating it across the word. */ + +#define SET_RATIO(speed) \ + ((speed) ? 15 : RISCV_CALL_RATIO - 2) + +/* STORE_BY_PIECES_P can be used when copying a constant string, but + in that case each word takes 3 insns (lui, ori, sw), or more in + 64-bit mode, instead of 2 (lw, sw). For now we always fail this + and let the move_by_pieces code copy the string from read-only + memory. In the future, this could be tuned further for multi-issue + CPUs that can issue stores down one pipe and arithmetic instructions + down another; in that case, the lui/ori/sw combination would be a + win for long enough strings. */ + +#define STORE_BY_PIECES_P(SIZE, ALIGN) 0 + +#ifndef HAVE_AS_TLS +#define HAVE_AS_TLS 0 +#endif + +#ifndef USED_FOR_TARGET + +extern const enum reg_class riscv_regno_to_class[]; +extern bool riscv_hard_regno_mode_ok[][FIRST_PSEUDO_REGISTER]; +extern const char* riscv_hi_relocs[]; +#endif + +#define ASM_PREFERRED_EH_DATA_FORMAT(CODE,GLOBAL) \ + (((GLOBAL) ? DW_EH_PE_indirect : 0) | DW_EH_PE_pcrel | DW_EH_PE_sdata4) diff --git a/gcc/config/riscv/riscv.md b/gcc/config/riscv/riscv.md new file mode 100644 index 0000000..ee2582a --- /dev/null +++ b/gcc/config/riscv/riscv.md @@ -0,0 +1,2451 @@ +;; Machine description for RISC-V for GNU compiler. +;; Copyright (C) 2011-2014 Free Software Foundation, Inc. +;; Contributed by Andrew Waterman (waterman@cs.berkeley.edu) at UC Berkeley. +;; Based on MIPS target for GNU compiler. + +;; This file is part of GCC. + +;; GCC is free software; you can redistribute it and/or modify +;; it under the terms of the GNU General Public License as published by +;; the Free Software Foundation; either version 3, or (at your option) +;; any later version. + +;; GCC is distributed in the hope that it will be useful, +;; but WITHOUT ANY WARRANTY; without even the implied warranty of +;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +;; GNU General Public License for more details. + +;; You should have received a copy of the GNU General Public License +;; along with GCC; see the file COPYING3. If not see +;; . + +(define_c_enum "unspec" [ + ;; Floating-point moves. + UNSPEC_LOAD_LOW + UNSPEC_LOAD_HIGH + UNSPEC_STORE_WORD + + ;; GP manipulation. + UNSPEC_EH_RETURN + + ;; Symbolic accesses. + UNSPEC_ADDRESS_FIRST + UNSPEC_LOAD_GOT + UNSPEC_TLS + UNSPEC_TLS_LE + UNSPEC_TLS_IE + UNSPEC_TLS_GD + + ;; Register save and restore. + UNSPEC_GPR_SAVE + UNSPEC_GPR_RESTORE + + ;; Blockage and synchronisation. + UNSPEC_BLOCKAGE + UNSPEC_FENCE + UNSPEC_FENCE_I +]) + +(define_constants + [(RETURN_ADDR_REGNUM 1) + (T0_REGNUM 5) + (T1_REGNUM 6) +]) + +(include "predicates.md") +(include "constraints.md") + +;; .................... +;; +;; Attributes +;; +;; .................... + +(define_attr "got" "unset,xgot_high,load" + (const_string "unset")) + +;; For jal instructions, this attribute is DIRECT when the target address +;; is symbolic and INDIRECT when it is a register. +(define_attr "jal" "unset,direct,indirect" + (const_string "unset")) + +;; Classification of moves, extensions and truncations. Most values +;; are as for "type" (see below) but there are also the following +;; move-specific values: +;; +;; andi a single ANDI instruction +;; shift_shift a shift left followed by a shift right +;; +;; This attribute is used to determine the instruction's length and +;; scheduling type. For doubleword moves, the attribute always describes +;; the split instructions; in some cases, it is more appropriate for the +;; scheduling type to be "multi" instead. +(define_attr "move_type" + "unknown,load,fpload,store,fpstore,mtc,mfc,move,fmove, + const,logical,arith,andi,shift_shift" + (const_string "unknown")) + +(define_attr "alu_type" "unknown,add,sub,and,or,xor" + (const_string "unknown")) + +;; Main data type used by the insn +(define_attr "mode" "unknown,none,QI,HI,SI,DI,TI,SF,DF,TF,FPSW" + (const_string "unknown")) + +;; True if the main data type is twice the size of a word. +(define_attr "dword_mode" "no,yes" + (cond [(and (eq_attr "mode" "DI,DF") + (eq (symbol_ref "TARGET_64BIT") (const_int 0))) + (const_string "yes") + + (and (eq_attr "mode" "TI,TF") + (ne (symbol_ref "TARGET_64BIT") (const_int 0))) + (const_string "yes")] + (const_string "no"))) + +;; Classification of each insn. +;; branch conditional branch +;; jump unconditional jump +;; call unconditional call +;; load load instruction(s) +;; fpload floating point load +;; fpidxload floating point indexed load +;; store store instruction(s) +;; fpstore floating point store +;; fpidxstore floating point indexed store +;; mtc transfer to coprocessor +;; mfc transfer from coprocessor +;; const load constant +;; arith integer arithmetic instructions +;; logical integer logical instructions +;; shift integer shift instructions +;; slt set less than instructions +;; imul integer multiply +;; idiv integer divide +;; move integer register move (addi rd, rs1, 0) +;; fmove floating point register move +;; fadd floating point add/subtract +;; fmul floating point multiply +;; fmadd floating point multiply-add +;; fdiv floating point divide +;; fcmp floating point compare +;; fcvt floating point convert +;; fsqrt floating point square root +;; multi multiword sequence (or user asm statements) +;; nop no operation +;; ghost an instruction that produces no real code +(define_attr "type" + "unknown,branch,jump,call,load,fpload,fpidxload,store,fpstore,fpidxstore, + mtc,mfc,const,arith,logical,shift,slt,imul,idiv,move,fmove,fadd,fmul, + fmadd,fdiv,fcmp,fcvt,fsqrt,multi,nop,ghost" + (cond [(eq_attr "jal" "!unset") (const_string "call") + (eq_attr "got" "load") (const_string "load") + + (eq_attr "alu_type" "add,sub") (const_string "arith") + + (eq_attr "alu_type" "and,or,xor") (const_string "logical") + + ;; If a doubleword move uses these expensive instructions, + ;; it is usually better to schedule them in the same way + ;; as the singleword form, rather than as "multi". + (eq_attr "move_type" "load") (const_string "load") + (eq_attr "move_type" "fpload") (const_string "fpload") + (eq_attr "move_type" "store") (const_string "store") + (eq_attr "move_type" "fpstore") (const_string "fpstore") + (eq_attr "move_type" "mtc") (const_string "mtc") + (eq_attr "move_type" "mfc") (const_string "mfc") + + ;; These types of move are always single insns. + (eq_attr "move_type" "fmove") (const_string "fmove") + (eq_attr "move_type" "arith") (const_string "arith") + (eq_attr "move_type" "logical") (const_string "logical") + (eq_attr "move_type" "andi") (const_string "logical") + + ;; These types of move are always split. + (eq_attr "move_type" "shift_shift") + (const_string "multi") + + ;; These types of move are split for doubleword modes only. + (and (eq_attr "move_type" "move,const") + (eq_attr "dword_mode" "yes")) + (const_string "multi") + (eq_attr "move_type" "move") (const_string "move") + (eq_attr "move_type" "const") (const_string "const")] + (const_string "unknown"))) + +;; Mode for conversion types (fcvt) +;; I2S integer to float single (SI/DI to SF) +;; I2D integer to float double (SI/DI to DF) +;; S2I float to integer (SF to SI/DI) +;; D2I float to integer (DF to SI/DI) +;; D2S double to float single +;; S2D float single to double + +(define_attr "cnv_mode" "unknown,I2S,I2D,S2I,D2I,D2S,S2D" + (const_string "unknown")) + +;; Length of instruction in bytes. +(define_attr "length" "" + (cond [ + ;; Direct branch instructions have a range of [-0x1000,0xffc], + ;; relative to the address of the delay slot. If a branch is + ;; outside this range, convert a branch like: + ;; + ;; bne r1,r2,target + ;; + ;; to: + ;; + ;; beq r1,r2,1f + ;; j target + ;; 1: + ;; + (eq_attr "type" "branch") + (if_then_else (and (le (minus (match_dup 0) (pc)) (const_int 4088)) + (le (minus (pc) (match_dup 0)) (const_int 4092))) + (const_int 4) + (const_int 8)) + + ;; Conservatively assume calls take two instructions, as in: + ;; auipc t0, %pcrel_hi(target) + ;; jalr ra, t0, %lo(target) + ;; The linker will relax these into JAL when appropriate. + (eq_attr "type" "call") + (const_int 8) + + ;; "Ghost" instructions occupy no space. + (eq_attr "type" "ghost") + (const_int 0) + + (eq_attr "got" "load") (const_int 8) + + ;; SHIFT_SHIFTs are decomposed into two separate instructions. + (eq_attr "move_type" "shift_shift") + (const_int 8) + + ;; Check for doubleword moves that are decomposed into two + ;; instructions. + (and (eq_attr "move_type" "mtc,mfc,move") + (eq_attr "dword_mode" "yes")) + (const_int 8) + + ;; Doubleword CONST{,N} moves are split into two word + ;; CONST{,N} moves. + (and (eq_attr "move_type" "const") + (eq_attr "dword_mode" "yes")) + (symbol_ref "riscv_split_const_insns (operands[1]) * 4") + + ;; Otherwise, constants, loads and stores are handled by external + ;; routines. + (eq_attr "move_type" "load,fpload") + (symbol_ref "riscv_load_store_insns (operands[1], insn) * 4") + (eq_attr "move_type" "store,fpstore") + (symbol_ref "riscv_load_store_insns (operands[0], insn) * 4") + ] (const_int 4))) + +;; Describe a user's asm statement. +(define_asm_attributes + [(set_attr "type" "multi")]) + +;; This mode iterator allows 32-bit and 64-bit GPR patterns to be generated +;; from the same template. +(define_mode_iterator GPR [SI (DI "TARGET_64BIT")]) +(define_mode_iterator SUPERQI [HI SI (DI "TARGET_64BIT")]) + +;; A copy of GPR that can be used when a pattern has two independent +;; modes. +(define_mode_iterator GPR2 [SI (DI "TARGET_64BIT")]) + +;; This mode iterator allows :P to be used for patterns that operate on +;; pointer-sized quantities. Exactly one of the two alternatives will match. +(define_mode_iterator P [(SI "Pmode == SImode") (DI "Pmode == DImode")]) + +;; 32-bit integer moves for which we provide move patterns. +(define_mode_iterator IMOVE32 [SI]) + +;; 64-bit modes for which we provide move patterns. +(define_mode_iterator MOVE64 [DI DF]) + +;; 128-bit modes for which we provide move patterns on 64-bit targets. +(define_mode_iterator MOVE128 [TI TF]) + +;; This mode iterator allows the QI and HI extension patterns to be +;; defined from the same template. +(define_mode_iterator SHORT [QI HI]) + +;; Likewise the 64-bit truncate-and-shift patterns. +(define_mode_iterator SUBDI [QI HI SI]) +(define_mode_iterator HISI [HI SI]) +(define_mode_iterator ANYI [QI HI SI (DI "TARGET_64BIT")]) + +;; This mode iterator allows :ANYF to be used wherever a scalar or vector +;; floating-point mode is allowed. +(define_mode_iterator ANYF [(SF "TARGET_HARD_FLOAT") + (DF "TARGET_HARD_FLOAT")]) +(define_mode_iterator ANYIF [QI HI SI (DI "TARGET_64BIT") + (SF "TARGET_HARD_FLOAT") + (DF "TARGET_HARD_FLOAT")]) + +;; Like ANYF, but only applies to scalar modes. +(define_mode_iterator SCALARF [(SF "TARGET_HARD_FLOAT") + (DF "TARGET_HARD_FLOAT")]) + +;; A floating-point mode for which moves involving FPRs may need to be split. +(define_mode_iterator SPLITF + [(DF "!TARGET_64BIT") + (DI "!TARGET_64BIT") + (TF "TARGET_64BIT")]) + +;; This attribute gives the length suffix for a sign- or zero-extension +;; instruction. +(define_mode_attr size [(QI "b") (HI "h")]) + +;; Mode attributes for loads. +(define_mode_attr load [(QI "lb") (HI "lh") (SI "lw") (DI "ld") (SF "flw") (DF "fld")]) + +;; Instruction names for stores. +(define_mode_attr store [(QI "sb") (HI "sh") (SI "sw") (DI "sd") (SF "fsw") (DF "fsd")]) + +;; This attribute gives the best constraint to use for registers of +;; a given mode. +(define_mode_attr reg [(SI "d") (DI "d") (CC "d")]) + +;; This attribute gives the format suffix for floating-point operations. +(define_mode_attr fmt [(SF "s") (DF "d")]) + +;; This attribute gives the format suffix for atomic memory operations. +(define_mode_attr amo [(SI "w") (DI "d")]) + +;; This attribute gives the upper-case mode name for one unit of a +;; floating-point mode. +(define_mode_attr UNITMODE [(SF "SF") (DF "DF")]) + +;; This attribute gives the integer mode that has half the size of +;; the controlling mode. +(define_mode_attr HALFMODE [(DF "SI") (DI "SI") (TF "DI")]) + +;; This code iterator allows signed and unsigned widening multiplications +;; to use the same template. +(define_code_iterator any_extend [sign_extend zero_extend]) + +;; This code iterator allows the two right shift instructions to be +;; generated from the same template. +(define_code_iterator any_shiftrt [ashiftrt lshiftrt]) + +;; This code iterator allows the three shift instructions to be generated +;; from the same template. +(define_code_iterator any_shift [ashift ashiftrt lshiftrt]) + +;; This code iterator allows unsigned and signed division to be generated +;; from the same template. +(define_code_iterator any_div [div udiv]) + +;; This code iterator allows unsigned and signed modulus to be generated +;; from the same template. +(define_code_iterator any_mod [mod umod]) + +;; These code iterators allow the signed and unsigned scc operations to use +;; the same template. +(define_code_iterator any_gt [gt gtu]) +(define_code_iterator any_ge [ge geu]) +(define_code_iterator any_lt [lt ltu]) +(define_code_iterator any_le [le leu]) + +;; expands to an empty string when doing a signed operation and +;; "u" when doing an unsigned operation. +(define_code_attr u [(sign_extend "") (zero_extend "u") + (div "") (udiv "u") + (mod "") (umod "u") + (gt "") (gtu "u") + (ge "") (geu "u") + (lt "") (ltu "u") + (le "") (leu "u")]) + +;; is like , but the signed form expands to "s" rather than "". +(define_code_attr su [(sign_extend "s") (zero_extend "u")]) + +;; expands to the name of the optab for a particular code. +(define_code_attr optab [(ashift "ashl") + (ashiftrt "ashr") + (lshiftrt "lshr") + (ior "ior") + (xor "xor") + (and "and") + (plus "add") + (minus "sub")]) + +;; expands to the name of the insn that implements a particular code. +(define_code_attr insn [(ashift "sll") + (ashiftrt "sra") + (lshiftrt "srl") + (ior "or") + (xor "xor") + (and "and") + (plus "add") + (minus "sub")]) + +;; Pipeline descriptions. +;; +;; generic.md provides a fallback for processors without a specific +;; pipeline description. It is derived from the old define_function_unit +;; version and uses the "alu" and "imuldiv" units declared below. +;; +;; Some of the processor-specific files are also derived from old +;; define_function_unit descriptions and simply override the parts of +;; generic.md that don't apply. The other processor-specific files +;; are self-contained. +(define_automaton "alu,imuldiv") + +(define_cpu_unit "alu" "alu") +(define_cpu_unit "imuldiv" "imuldiv") + +;; Ghost instructions produce no real code and introduce no hazards. +;; They exist purely to express an effect on dataflow. +(define_insn_reservation "ghost" 0 + (eq_attr "type" "ghost") + "nothing") + +(include "generic.md") + +;; +;; .................... +;; +;; ADDITION +;; +;; .................... +;; + +(define_insn "add3" + [(set (match_operand:ANYF 0 "register_operand" "=f") + (plus:ANYF (match_operand:ANYF 1 "register_operand" "f") + (match_operand:ANYF 2 "register_operand" "f")))] + "" + "fadd.\t%0,%1,%2" + [(set_attr "type" "fadd") + (set_attr "mode" "")]) + +(define_expand "add3" + [(set (match_operand:GPR 0 "register_operand") + (plus:GPR (match_operand:GPR 1 "register_operand") + (match_operand:GPR 2 "arith_operand")))] + "") + +(define_insn "*addsi3" + [(set (match_operand:SI 0 "register_operand" "=r,r") + (plus:SI (match_operand:GPR 1 "register_operand" "r,r") + (match_operand:GPR2 2 "arith_operand" "r,Q")))] + "" + { return TARGET_64BIT ? "addw\t%0,%1,%2" : "add\t%0,%1,%2"; } + [(set_attr "type" "arith") + (set_attr "mode" "SI")]) + +(define_insn "*adddi3" + [(set (match_operand:DI 0 "register_operand" "=r,r") + (plus:DI (match_operand:DI 1 "register_operand" "r,r") + (match_operand:DI 2 "arith_operand" "r,Q")))] + "TARGET_64BIT" + "add\t%0,%1,%2" + [(set_attr "type" "arith") + (set_attr "mode" "DI")]) + +(define_insn "*addsi3_extended" + [(set (match_operand:DI 0 "register_operand" "=r,r") + (sign_extend:DI + (plus:SI (match_operand:SI 1 "register_operand" "r,r") + (match_operand:SI 2 "arith_operand" "r,Q"))))] + "TARGET_64BIT" + "addw\t%0,%1,%2" + [(set_attr "type" "arith") + (set_attr "mode" "SI")]) + +(define_insn "*adddisi3" + [(set (match_operand:SI 0 "register_operand" "=r,r") + (plus:SI (truncate:SI (match_operand:DI 1 "register_operand" "r,r")) + (truncate:SI (match_operand:DI 2 "arith_operand" "r,Q"))))] + "TARGET_64BIT" + "addw\t%0,%1,%2" + [(set_attr "type" "arith") + (set_attr "mode" "SI")]) + +(define_insn "*adddisisi3" + [(set (match_operand:SI 0 "register_operand" "=r,r") + (plus:SI (truncate:SI (match_operand:DI 1 "register_operand" "r,r")) + (match_operand:SI 2 "arith_operand" "r,Q")))] + "TARGET_64BIT" + "addw\t%0,%1,%2" + [(set_attr "type" "arith") + (set_attr "mode" "SI")]) + +(define_insn "*adddi3_truncsi" + [(set (match_operand:SI 0 "register_operand" "=r,r") + (truncate:SI + (plus:DI (match_operand:DI 1 "register_operand" "r,r") + (match_operand:DI 2 "arith_operand" "r,Q"))))] + "TARGET_64BIT" + "addw\t%0,%1,%2" + [(set_attr "type" "arith") + (set_attr "mode" "SI")]) + +;; +;; .................... +;; +;; SUBTRACTION +;; +;; .................... +;; + +(define_insn "sub3" + [(set (match_operand:ANYF 0 "register_operand" "=f") + (minus:ANYF (match_operand:ANYF 1 "register_operand" "f") + (match_operand:ANYF 2 "register_operand" "f")))] + "" + "fsub.\t%0,%1,%2" + [(set_attr "type" "fadd") + (set_attr "mode" "")]) + +(define_expand "sub3" + [(set (match_operand:GPR 0 "register_operand") + (minus:GPR (match_operand:GPR 1 "reg_or_0_operand") + (match_operand:GPR 2 "register_operand")))] + "") + +(define_insn "*subdi3" + [(set (match_operand:DI 0 "register_operand" "=r") + (minus:DI (match_operand:DI 1 "reg_or_0_operand" "rJ") + (match_operand:DI 2 "register_operand" "r")))] + "TARGET_64BIT" + "sub\t%0,%z1,%2" + [(set_attr "type" "arith") + (set_attr "mode" "DI")]) + +(define_insn "*subsi3" + [(set (match_operand:SI 0 "register_operand" "=r") + (minus:SI (match_operand:GPR 1 "reg_or_0_operand" "rJ") + (match_operand:GPR2 2 "register_operand" "r")))] + "" + { return TARGET_64BIT ? "subw\t%0,%z1,%2" : "sub\t%0,%z1,%2"; } + [(set_attr "type" "arith") + (set_attr "mode" "SI")]) + +(define_insn "*subsi3_extended" + [(set (match_operand:DI 0 "register_operand" "=r") + (sign_extend:DI + (minus:SI (match_operand:SI 1 "reg_or_0_operand" "rJ") + (match_operand:SI 2 "register_operand" "r"))))] + "TARGET_64BIT" + "subw\t%0,%z1,%2" + [(set_attr "type" "arith") + (set_attr "mode" "DI")]) + +(define_insn "*subdisi3" + [(set (match_operand:SI 0 "register_operand" "=r") + (minus:SI (truncate:SI (match_operand:DI 1 "reg_or_0_operand" "rJ")) + (truncate:SI (match_operand:DI 2 "register_operand" "r"))))] + "TARGET_64BIT" + "subw\t%0,%z1,%2" + [(set_attr "type" "arith") + (set_attr "mode" "SI")]) + +(define_insn "*subdisisi3" + [(set (match_operand:SI 0 "register_operand" "=r") + (minus:SI (truncate:SI (match_operand:DI 1 "reg_or_0_operand" "rJ")) + (match_operand:SI 2 "register_operand" "r")))] + "TARGET_64BIT" + "subw\t%0,%z1,%2" + [(set_attr "type" "arith") + (set_attr "mode" "SI")]) + +(define_insn "*subsidisi3" + [(set (match_operand:SI 0 "register_operand" "=r") + (minus:SI (match_operand:SI 1 "reg_or_0_operand" "rJ") + (truncate:SI (match_operand:DI 2 "register_operand" "r"))))] + "TARGET_64BIT" + "subw\t%0,%z1,%2" + [(set_attr "type" "arith") + (set_attr "mode" "SI")]) + +(define_insn "*subdi3_truncsi" + [(set (match_operand:SI 0 "register_operand" "=r,r") + (truncate:SI + (minus:DI (match_operand:DI 1 "reg_or_0_operand" "rJ,r") + (match_operand:DI 2 "arith_operand" "r,Q"))))] + "TARGET_64BIT" + "subw\t%0,%z1,%2" + [(set_attr "type" "arith") + (set_attr "mode" "SI")]) + +;; +;; .................... +;; +;; MULTIPLICATION +;; +;; .................... +;; + +(define_insn "mul3" + [(set (match_operand:SCALARF 0 "register_operand" "=f") + (mult:SCALARF (match_operand:SCALARF 1 "register_operand" "f") + (match_operand:SCALARF 2 "register_operand" "f")))] + "" + "fmul.\t%0,%1,%2" + [(set_attr "type" "fmul") + (set_attr "mode" "")]) + +(define_expand "mul3" + [(set (match_operand:GPR 0 "register_operand") + (mult:GPR (match_operand:GPR 1 "reg_or_0_operand") + (match_operand:GPR 2 "register_operand")))] + "TARGET_MULDIV") + +(define_insn "*mulsi3" + [(set (match_operand:SI 0 "register_operand" "=r") + (mult:SI (match_operand:GPR 1 "register_operand" "r") + (match_operand:GPR2 2 "register_operand" "r")))] + "TARGET_MULDIV" + { return TARGET_64BIT ? "mulw\t%0,%1,%2" : "mul\t%0,%1,%2"; } + [(set_attr "type" "imul") + (set_attr "mode" "SI")]) + +(define_insn "*muldisi3" + [(set (match_operand:SI 0 "register_operand" "=r") + (mult:SI (truncate:SI (match_operand:DI 1 "register_operand" "r")) + (truncate:SI (match_operand:DI 2 "register_operand" "r"))))] + "TARGET_MULDIV && TARGET_64BIT" + "mulw\t%0,%1,%2" + [(set_attr "type" "imul") + (set_attr "mode" "SI")]) + +(define_insn "*muldi3_truncsi" + [(set (match_operand:SI 0 "register_operand" "=r") + (truncate:SI + (mult:DI (match_operand:DI 1 "register_operand" "r") + (match_operand:DI 2 "register_operand" "r"))))] + "TARGET_MULDIV && TARGET_64BIT" + "mulw\t%0,%1,%2" + [(set_attr "type" "imul") + (set_attr "mode" "SI")]) + +(define_insn "*muldi3" + [(set (match_operand:DI 0 "register_operand" "=r") + (mult:DI (match_operand:DI 1 "register_operand" "r") + (match_operand:DI 2 "register_operand" "r")))] + "TARGET_MULDIV && TARGET_64BIT" + "mul\t%0,%1,%2" + [(set_attr "type" "imul") + (set_attr "mode" "DI")]) + +;; +;; ........................ +;; +;; MULTIPLICATION HIGH-PART +;; +;; ........................ +;; + + +;; Using a clobber here is ghetto, but I'm not smart enough to do better. ' +(define_insn_and_split "mulditi3" + [(set (match_operand:TI 0 "register_operand" "=r") + (mult:TI (any_extend:TI + (match_operand:DI 1 "register_operand" "r")) + (any_extend:TI + (match_operand:DI 2 "register_operand" "r")))) + (clobber (match_scratch:DI 3 "=r"))] + "TARGET_MULDIV && TARGET_64BIT" + "#" + "reload_completed" + [ + (set (match_dup 3) (mult:DI (match_dup 1) (match_dup 2))) + (set (match_dup 4) (truncate:DI + (lshiftrt:TI + (mult:TI (any_extend:TI (match_dup 1)) + (any_extend:TI (match_dup 2))) + (const_int 64)))) + (set (match_dup 5) (match_dup 3)) + ] +{ + operands[4] = riscv_subword (operands[0], true); + operands[5] = riscv_subword (operands[0], false); +} + ) + +(define_insn "muldi3_highpart" + [(set (match_operand:DI 0 "register_operand" "=r") + (truncate:DI + (lshiftrt:TI + (mult:TI (any_extend:TI + (match_operand:DI 1 "register_operand" "r")) + (any_extend:TI + (match_operand:DI 2 "register_operand" "r"))) + (const_int 64))))] + "TARGET_MULDIV && TARGET_64BIT" + "mulh\t%0,%1,%2" + [(set_attr "type" "imul") + (set_attr "mode" "DI")]) + + +(define_insn_and_split "usmulditi3" + [(set (match_operand:TI 0 "register_operand" "=r") + (mult:TI (zero_extend:TI + (match_operand:DI 1 "register_operand" "r")) + (sign_extend:TI + (match_operand:DI 2 "register_operand" "r")))) + (clobber (match_scratch:DI 3 "=r"))] + "TARGET_MULDIV && TARGET_64BIT" + "#" + "reload_completed" + [ + (set (match_dup 3) (mult:DI (match_dup 1) (match_dup 2))) + (set (match_dup 4) (truncate:DI + (lshiftrt:TI + (mult:TI (zero_extend:TI (match_dup 1)) + (sign_extend:TI (match_dup 2))) + (const_int 64)))) + (set (match_dup 5) (match_dup 3)) + ] +{ + operands[4] = riscv_subword (operands[0], true); + operands[5] = riscv_subword (operands[0], false); +} + ) + +(define_insn "usmuldi3_highpart" + [(set (match_operand:DI 0 "register_operand" "=r") + (truncate:DI + (lshiftrt:TI + (mult:TI (zero_extend:TI + (match_operand:DI 1 "register_operand" "r")) + (sign_extend:TI + (match_operand:DI 2 "register_operand" "r"))) + (const_int 64))))] + "TARGET_MULDIV && TARGET_64BIT" + "mulhsu\t%0,%2,%1" + [(set_attr "type" "imul") + (set_attr "mode" "DI")]) + +(define_expand "mulsidi3" + [(set (match_operand:DI 0 "register_operand" "=r") + (mult:DI (any_extend:DI + (match_operand:SI 1 "register_operand" "r")) + (any_extend:DI + (match_operand:SI 2 "register_operand" "r")))) + (clobber (match_scratch:SI 3 "=r"))] + "TARGET_MULDIV && !TARGET_64BIT" +{ + rtx temp = gen_reg_rtx (SImode); + emit_insn (gen_mulsi3 (temp, operands[1], operands[2])); + emit_insn (gen_mulsi3_highpart (riscv_subword (operands[0], true), + operands[1], operands[2])); + emit_insn (gen_movsi (riscv_subword (operands[0], false), temp)); + DONE; +} + ) + +(define_insn "mulsi3_highpart" + [(set (match_operand:SI 0 "register_operand" "=r") + (truncate:SI + (lshiftrt:DI + (mult:DI (any_extend:DI + (match_operand:SI 1 "register_operand" "r")) + (any_extend:DI + (match_operand:SI 2 "register_operand" "r"))) + (const_int 32))))] + "TARGET_MULDIV && !TARGET_64BIT" + "mulh\t%0,%1,%2" + [(set_attr "type" "imul") + (set_attr "mode" "SI")]) + + +(define_expand "usmulsidi3" + [(set (match_operand:DI 0 "register_operand" "=r") + (mult:DI (zero_extend:DI + (match_operand:SI 1 "register_operand" "r")) + (sign_extend:DI + (match_operand:SI 2 "register_operand" "r")))) + (clobber (match_scratch:SI 3 "=r"))] + "TARGET_MULDIV && !TARGET_64BIT" +{ + rtx temp = gen_reg_rtx (SImode); + emit_insn (gen_mulsi3 (temp, operands[1], operands[2])); + emit_insn (gen_usmulsi3_highpart (riscv_subword (operands[0], true), + operands[1], operands[2])); + emit_insn (gen_movsi (riscv_subword (operands[0], false), temp)); + DONE; +} + ) + +(define_insn "usmulsi3_highpart" + [(set (match_operand:SI 0 "register_operand" "=r") + (truncate:SI + (lshiftrt:DI + (mult:DI (zero_extend:DI + (match_operand:SI 1 "register_operand" "r")) + (sign_extend:DI + (match_operand:SI 2 "register_operand" "r"))) + (const_int 32))))] + "TARGET_MULDIV && !TARGET_64BIT" + "mulhsu\t%0,%2,%1" + [(set_attr "type" "imul") + (set_attr "mode" "SI")]) + +;; +;; .................... +;; +;; DIVISION and REMAINDER +;; +;; .................... +;; + +(define_insn "divsi3" + [(set (match_operand:SI 0 "register_operand" "=r") + (any_div:SI (match_operand:SI 1 "register_operand" "r") + (match_operand:SI 2 "register_operand" "r")))] + "TARGET_MULDIV" + { return TARGET_64BIT ? "divw\t%0,%1,%2" : "div\t%0,%1,%2"; } + [(set_attr "type" "idiv") + (set_attr "mode" "SI")]) + +(define_insn "divdi3" + [(set (match_operand:DI 0 "register_operand" "=r") + (any_div:DI (match_operand:DI 1 "register_operand" "r") + (match_operand:DI 2 "register_operand" "r")))] + "TARGET_MULDIV && TARGET_64BIT" + "div\t%0,%1,%2" + [(set_attr "type" "idiv") + (set_attr "mode" "DI")]) + +(define_insn "modsi3" + [(set (match_operand:SI 0 "register_operand" "=r") + (any_mod:SI (match_operand:SI 1 "register_operand" "r") + (match_operand:SI 2 "register_operand" "r")))] + "TARGET_MULDIV" + { return TARGET_64BIT ? "remw\t%0,%1,%2" : "rem\t%0,%1,%2"; } + [(set_attr "type" "idiv") + (set_attr "mode" "SI")]) + +(define_insn "moddi3" + [(set (match_operand:DI 0 "register_operand" "=r") + (any_mod:DI (match_operand:DI 1 "register_operand" "r") + (match_operand:DI 2 "register_operand" "r")))] + "TARGET_MULDIV && TARGET_64BIT" + "rem\t%0,%1,%2" + [(set_attr "type" "idiv") + (set_attr "mode" "DI")]) + +(define_insn "div3" + [(set (match_operand:ANYF 0 "register_operand" "=f") + (div:ANYF (match_operand:ANYF 1 "register_operand" "f") + (match_operand:ANYF 2 "register_operand" "f")))] + "TARGET_HARD_FLOAT && TARGET_FDIV" + "fdiv.\t%0,%1,%2" + [(set_attr "type" "fdiv") + (set_attr "mode" "")]) + +;; +;; .................... +;; +;; SQUARE ROOT +;; +;; .................... + +(define_insn "sqrt2" + [(set (match_operand:ANYF 0 "register_operand" "=f") + (sqrt:ANYF (match_operand:ANYF 1 "register_operand" "f")))] + "TARGET_HARD_FLOAT && TARGET_FDIV" +{ + return "fsqrt.\t%0,%1"; +} + [(set_attr "type" "fsqrt") + (set_attr "mode" "")]) + +;; Floating point multiply accumulate instructions. + +(define_insn "fma4" + [(set (match_operand:ANYF 0 "register_operand" "=f") + (fma:ANYF + (match_operand:ANYF 1 "register_operand" "f") + (match_operand:ANYF 2 "register_operand" "f") + (match_operand:ANYF 3 "register_operand" "f")))] + "TARGET_HARD_FLOAT" + "fmadd.\t%0,%1,%2,%3" + [(set_attr "type" "fmadd") + (set_attr "mode" "")]) + +(define_insn "fms4" + [(set (match_operand:ANYF 0 "register_operand" "=f") + (fma:ANYF + (match_operand:ANYF 1 "register_operand" "f") + (match_operand:ANYF 2 "register_operand" "f") + (neg:ANYF (match_operand:ANYF 3 "register_operand" "f"))))] + "TARGET_HARD_FLOAT" + "fmsub.\t%0,%1,%2,%3" + [(set_attr "type" "fmadd") + (set_attr "mode" "")]) + +(define_insn "nfma4" + [(set (match_operand:ANYF 0 "register_operand" "=f") + (neg:ANYF + (fma:ANYF + (match_operand:ANYF 1 "register_operand" "f") + (match_operand:ANYF 2 "register_operand" "f") + (match_operand:ANYF 3 "register_operand" "f"))))] + "TARGET_HARD_FLOAT" + "fnmadd.\t%0,%1,%2,%3" + [(set_attr "type" "fmadd") + (set_attr "mode" "")]) + +(define_insn "nfms4" + [(set (match_operand:ANYF 0 "register_operand" "=f") + (neg:ANYF + (fma:ANYF + (match_operand:ANYF 1 "register_operand" "f") + (match_operand:ANYF 2 "register_operand" "f") + (neg:ANYF (match_operand:ANYF 3 "register_operand" "f")))))] + "TARGET_HARD_FLOAT" + "fnmsub.\t%0,%1,%2,%3" + [(set_attr "type" "fmadd") + (set_attr "mode" "")]) + +;; modulo signed zeros, -(a*b+c) == -c-a*b +(define_insn "*nfma4_fastmath" + [(set (match_operand:ANYF 0 "register_operand" "=f") + (minus:ANYF + (match_operand:ANYF 3 "register_operand" "f") + (mult:ANYF + (neg:ANYF (match_operand:ANYF 1 "register_operand" "f")) + (match_operand:ANYF 2 "register_operand" "f"))))] + "TARGET_HARD_FLOAT && !HONOR_SIGNED_ZEROS (mode)" + "fnmadd.\t%0,%1,%2,%3" + [(set_attr "type" "fmadd") + (set_attr "mode" "")]) + +;; modulo signed zeros, -(a*b-c) == c-a*b +(define_insn "*nfms4_fastmath" + [(set (match_operand:ANYF 0 "register_operand" "=f") + (minus:ANYF + (match_operand:ANYF 3 "register_operand" "f") + (mult:ANYF + (match_operand:ANYF 1 "register_operand" "f") + (match_operand:ANYF 2 "register_operand" "f"))))] + "TARGET_HARD_FLOAT && !HONOR_SIGNED_ZEROS (mode)" + "fnmsub.\t%0,%1,%2,%3" + [(set_attr "type" "fmadd") + (set_attr "mode" "")]) + +;; +;; .................... +;; +;; ABSOLUTE VALUE +;; +;; .................... + +(define_insn "abs2" + [(set (match_operand:ANYF 0 "register_operand" "=f") + (abs:ANYF (match_operand:ANYF 1 "register_operand" "f")))] + "TARGET_HARD_FLOAT" + "fabs.\t%0,%1" + [(set_attr "type" "fmove") + (set_attr "mode" "")]) + + +;; +;; .................... +;; +;; MIN/MAX +;; +;; .................... + +(define_insn "smin3" + [(set (match_operand:ANYF 0 "register_operand" "=f") + (smin:ANYF (match_operand:ANYF 1 "register_operand" "f") + (match_operand:ANYF 2 "register_operand" "f")))] + "TARGET_HARD_FLOAT" + "fmin.\t%0,%1,%2" + [(set_attr "type" "fmove") + (set_attr "mode" "")]) + +(define_insn "smax3" + [(set (match_operand:ANYF 0 "register_operand" "=f") + (smax:ANYF (match_operand:ANYF 1 "register_operand" "f") + (match_operand:ANYF 2 "register_operand" "f")))] + "TARGET_HARD_FLOAT" + "fmax.\t%0,%1,%2" + [(set_attr "type" "fmove") + (set_attr "mode" "")]) + + +;; +;; .................... +;; +;; NEGATION and ONE'S COMPLEMENT ' +;; +;; .................... + +(define_insn "neg2" + [(set (match_operand:ANYF 0 "register_operand" "=f") + (neg:ANYF (match_operand:ANYF 1 "register_operand" "f")))] + "TARGET_HARD_FLOAT" + "fneg.\t%0,%1" + [(set_attr "type" "fmove") + (set_attr "mode" "")]) + +(define_insn "one_cmpl2" + [(set (match_operand:GPR 0 "register_operand" "=r") + (not:GPR (match_operand:GPR 1 "register_operand" "r")))] + "" + "not\t%0,%1" + [(set_attr "type" "logical") + (set_attr "mode" "")]) + +;; +;; .................... +;; +;; LOGICAL +;; +;; .................... +;; + +(define_insn "and3" + [(set (match_operand:GPR 0 "register_operand" "=r,r") + (and:GPR (match_operand:GPR 1 "register_operand" "%r,r") + (match_operand:GPR 2 "arith_operand" "r,Q")))] + "" + "and\t%0,%1,%2" + [(set_attr "type" "logical") + (set_attr "mode" "")]) + +(define_insn "ior3" + [(set (match_operand:GPR 0 "register_operand" "=r,r") + (ior:GPR (match_operand:GPR 1 "register_operand" "%r,r") + (match_operand:GPR 2 "arith_operand" "r,Q")))] + "" + "or\t%0,%1,%2" + [(set_attr "type" "logical") + (set_attr "mode" "")]) + +(define_insn "xor3" + [(set (match_operand:GPR 0 "register_operand" "=r,r") + (xor:GPR (match_operand:GPR 1 "register_operand" "%r,r") + (match_operand:GPR 2 "arith_operand" "r,Q")))] + "" + "xor\t%0,%1,%2" + [(set_attr "type" "logical") + (set_attr "mode" "")]) + +;; +;; .................... +;; +;; TRUNCATION +;; +;; .................... + +(define_insn "truncdfsf2" + [(set (match_operand:SF 0 "register_operand" "=f") + (float_truncate:SF (match_operand:DF 1 "register_operand" "f")))] + "TARGET_HARD_FLOAT" + "fcvt.s.d\t%0,%1" + [(set_attr "type" "fcvt") + (set_attr "cnv_mode" "D2S") + (set_attr "mode" "SF")]) + +;; Integer truncation patterns. Truncating to HImode/QImode is a no-op. +;; Truncating from DImode to SImode is not, because we always keep SImode +;; values sign-extended in a register so we can safely use DImode branches +;; and comparisons on SImode values. + +(define_insn "truncdisi2" + [(set (match_operand:SI 0 "nonimmediate_operand" "=r,m") + (truncate:SI (match_operand:DI 1 "register_operand" "r,r")))] + "TARGET_64BIT" + "@ + sext.w\t%0,%1 + sw\t%1,%0" + [(set_attr "move_type" "arith,store") + (set_attr "mode" "SI")]) + +;; Combiner patterns to optimize shift/truncate combinations. + +(define_insn "*ashr_trunc" + [(set (match_operand:SUBDI 0 "register_operand" "=r") + (truncate:SUBDI + (ashiftrt:DI (match_operand:DI 1 "register_operand" "r") + (match_operand:DI 2 "const_arith_operand" ""))))] + "TARGET_64BIT && IN_RANGE (INTVAL (operands[2]), 32, 63)" + "sra\t%0,%1,%2" + [(set_attr "type" "shift") + (set_attr "mode" "")]) + +(define_insn "*lshr32_trunc" + [(set (match_operand:SUBDI 0 "register_operand" "=r") + (truncate:SUBDI + (lshiftrt:DI (match_operand:DI 1 "register_operand" "r") + (const_int 32))))] + "TARGET_64BIT" + "sra\t%0,%1,32" + [(set_attr "type" "shift") + (set_attr "mode" "")]) + +;; +;; .................... +;; +;; ZERO EXTENSION +;; +;; .................... + +;; Extension insns. + +(define_insn_and_split "zero_extendsidi2" + [(set (match_operand:DI 0 "register_operand" "=r,r") + (zero_extend:DI (match_operand:SI 1 "nonimmediate_operand" "r,W")))] + "TARGET_64BIT" + "@ + # + lwu\t%0,%1" + "&& reload_completed && REG_P (operands[1])" + [(set (match_dup 0) + (ashift:DI (match_dup 1) (const_int 32))) + (set (match_dup 0) + (lshiftrt:DI (match_dup 0) (const_int 32)))] + { operands[1] = gen_lowpart (DImode, operands[1]); } + [(set_attr "move_type" "shift_shift,load") + (set_attr "mode" "DI")]) + +;; Combine is not allowed to convert this insn into a zero_extendsidi2 +;; because of TRULY_NOOP_TRUNCATION. + +(define_insn_and_split "*clear_upper32" + [(set (match_operand:DI 0 "register_operand" "=r,r") + (and:DI (match_operand:DI 1 "nonimmediate_operand" "r,W") + (const_int 4294967295)))] + "TARGET_64BIT" +{ + if (which_alternative == 0) + return "#"; + + operands[1] = gen_lowpart (SImode, operands[1]); + return "lwu\t%0,%1"; +} + "&& reload_completed && REG_P (operands[1])" + [(set (match_dup 0) + (ashift:DI (match_dup 1) (const_int 32))) + (set (match_dup 0) + (lshiftrt:DI (match_dup 0) (const_int 32)))] + "" + [(set_attr "move_type" "shift_shift,load") + (set_attr "mode" "DI")]) + +(define_insn_and_split "zero_extendhi2" + [(set (match_operand:GPR 0 "register_operand" "=r,r") + (zero_extend:GPR (match_operand:HI 1 "nonimmediate_operand" "r,m")))] + "" + "@ + # + lhu\t%0,%1" + "&& reload_completed && REG_P (operands[1])" + [(set (match_dup 0) + (ashift:GPR (match_dup 1) (match_dup 2))) + (set (match_dup 0) + (lshiftrt:GPR (match_dup 0) (match_dup 2)))] + { + operands[1] = gen_lowpart (mode, operands[1]); + operands[2] = GEN_INT(GET_MODE_BITSIZE(mode) - 16); + } + [(set_attr "move_type" "shift_shift,load") + (set_attr "mode" "")]) + +(define_insn "zero_extendqi2" + [(set (match_operand:SUPERQI 0 "register_operand" "=r,r") + (zero_extend:SUPERQI + (match_operand:QI 1 "nonimmediate_operand" "r,m")))] + "" + "@ + and\t%0,%1,0xff + lbu\t%0,%1" + [(set_attr "move_type" "andi,load") + (set_attr "mode" "")]) + +;; +;; .................... +;; +;; SIGN EXTENSION +;; +;; .................... + +;; Extension insns. +;; Those for integer source operand are ordered widest source type first. + +;; When TARGET_64BIT, all SImode integer registers should already be in +;; sign-extended form (see TRULY_NOOP_TRUNCATION and truncdisi2). We can +;; therefore get rid of register->register instructions if we constrain +;; the source to be in the same register as the destination. +;; +;; The register alternative has type "arith" so that the pre-reload +;; scheduler will treat it as a move. This reflects what happens if +;; the register alternative needs a reload. +(define_insn_and_split "extendsidi2" + [(set (match_operand:DI 0 "register_operand" "=r,r") + (sign_extend:DI (match_operand:SI 1 "nonimmediate_operand" "r,m")))] + "TARGET_64BIT" + "@ + # + lw\t%0,%1" + "&& reload_completed && register_operand (operands[1], VOIDmode)" + [(set (match_dup 0) (match_dup 1))] +{ + if (REGNO (operands[0]) == REGNO (operands[1])) + { + emit_note (NOTE_INSN_DELETED); + DONE; + } + operands[1] = gen_rtx_REG (DImode, REGNO (operands[1])); +} + [(set_attr "move_type" "move,load") + (set_attr "mode" "DI")]) + +(define_insn_and_split "extend2" + [(set (match_operand:SUPERQI 0 "register_operand" "=r,r") + (sign_extend:SUPERQI + (match_operand:SHORT 1 "nonimmediate_operand" "r,m")))] + "" + "@ + # + l\t%0,%1" + "&& reload_completed && REG_P (operands[1])" + [(set (match_dup 0) (ashift:SI (match_dup 1) (match_dup 2))) + (set (match_dup 0) (ashiftrt:SI (match_dup 0) (match_dup 2)))] +{ + operands[0] = gen_lowpart (SImode, operands[0]); + operands[1] = gen_lowpart (SImode, operands[1]); + operands[2] = GEN_INT (GET_MODE_BITSIZE (SImode) + - GET_MODE_BITSIZE (mode)); +} + [(set_attr "move_type" "shift_shift,load") + (set_attr "mode" "SI")]) + +(define_insn "extendsfdf2" + [(set (match_operand:DF 0 "register_operand" "=f") + (float_extend:DF (match_operand:SF 1 "register_operand" "f")))] + "TARGET_HARD_FLOAT" + "fcvt.d.s\t%0,%1" + [(set_attr "type" "fcvt") + (set_attr "cnv_mode" "S2D") + (set_attr "mode" "DF")]) + +;; +;; .................... +;; +;; CONVERSIONS +;; +;; .................... + +(define_insn "fix_truncdfsi2" + [(set (match_operand:SI 0 "register_operand" "=r") + (fix:SI (match_operand:DF 1 "register_operand" "f")))] + "TARGET_HARD_FLOAT" + "fcvt.w.d %0,%1,rtz" + [(set_attr "type" "fcvt") + (set_attr "mode" "DF") + (set_attr "cnv_mode" "D2I")]) + + +(define_insn "fix_truncsfsi2" + [(set (match_operand:SI 0 "register_operand" "=r") + (fix:SI (match_operand:SF 1 "register_operand" "f")))] + "TARGET_HARD_FLOAT" + "fcvt.w.s %0,%1,rtz" + [(set_attr "type" "fcvt") + (set_attr "mode" "SF") + (set_attr "cnv_mode" "S2I")]) + + +(define_insn "fix_truncdfdi2" + [(set (match_operand:DI 0 "register_operand" "=r") + (fix:DI (match_operand:DF 1 "register_operand" "f")))] + "TARGET_HARD_FLOAT && TARGET_64BIT" + "fcvt.l.d %0,%1,rtz" + [(set_attr "type" "fcvt") + (set_attr "mode" "DF") + (set_attr "cnv_mode" "D2I")]) + + +(define_insn "fix_truncsfdi2" + [(set (match_operand:DI 0 "register_operand" "=r") + (fix:DI (match_operand:SF 1 "register_operand" "f")))] + "TARGET_HARD_FLOAT && TARGET_64BIT" + "fcvt.l.s %0,%1,rtz" + [(set_attr "type" "fcvt") + (set_attr "mode" "SF") + (set_attr "cnv_mode" "S2I")]) + + +(define_insn "floatsidf2" + [(set (match_operand:DF 0 "register_operand" "=f") + (float:DF (match_operand:SI 1 "reg_or_0_operand" "rJ")))] + "TARGET_HARD_FLOAT" + "fcvt.d.w\t%0,%z1" + [(set_attr "type" "fcvt") + (set_attr "mode" "DF") + (set_attr "cnv_mode" "I2D")]) + + +(define_insn "floatdidf2" + [(set (match_operand:DF 0 "register_operand" "=f") + (float:DF (match_operand:DI 1 "reg_or_0_operand" "rJ")))] + "TARGET_HARD_FLOAT && TARGET_64BIT" + "fcvt.d.l\t%0,%z1" + [(set_attr "type" "fcvt") + (set_attr "mode" "DF") + (set_attr "cnv_mode" "I2D")]) + + +(define_insn "floatsisf2" + [(set (match_operand:SF 0 "register_operand" "=f") + (float:SF (match_operand:SI 1 "reg_or_0_operand" "rJ")))] + "TARGET_HARD_FLOAT" + "fcvt.s.w\t%0,%z1" + [(set_attr "type" "fcvt") + (set_attr "mode" "SF") + (set_attr "cnv_mode" "I2S")]) + + +(define_insn "floatdisf2" + [(set (match_operand:SF 0 "register_operand" "=f") + (float:SF (match_operand:DI 1 "reg_or_0_operand" "rJ")))] + "TARGET_HARD_FLOAT && TARGET_64BIT" + "fcvt.s.l\t%0,%z1" + [(set_attr "type" "fcvt") + (set_attr "mode" "SF") + (set_attr "cnv_mode" "I2S")]) + + +(define_insn "floatunssidf2" + [(set (match_operand:DF 0 "register_operand" "=f") + (unsigned_float:DF (match_operand:SI 1 "reg_or_0_operand" "rJ")))] + "TARGET_HARD_FLOAT" + "fcvt.d.wu\t%0,%z1" + [(set_attr "type" "fcvt") + (set_attr "mode" "DF") + (set_attr "cnv_mode" "I2D")]) + + +(define_insn "floatunsdidf2" + [(set (match_operand:DF 0 "register_operand" "=f") + (unsigned_float:DF (match_operand:DI 1 "reg_or_0_operand" "rJ")))] + "TARGET_HARD_FLOAT && TARGET_64BIT" + "fcvt.d.lu\t%0,%z1" + [(set_attr "type" "fcvt") + (set_attr "mode" "DF") + (set_attr "cnv_mode" "I2D")]) + + +(define_insn "floatunssisf2" + [(set (match_operand:SF 0 "register_operand" "=f") + (unsigned_float:SF (match_operand:SI 1 "reg_or_0_operand" "rJ")))] + "TARGET_HARD_FLOAT" + "fcvt.s.wu\t%0,%z1" + [(set_attr "type" "fcvt") + (set_attr "mode" "SF") + (set_attr "cnv_mode" "I2S")]) + + +(define_insn "floatunsdisf2" + [(set (match_operand:SF 0 "register_operand" "=f") + (unsigned_float:SF (match_operand:DI 1 "reg_or_0_operand" "rJ")))] + "TARGET_HARD_FLOAT && TARGET_64BIT" + "fcvt.s.lu\t%0,%z1" + [(set_attr "type" "fcvt") + (set_attr "mode" "SF") + (set_attr "cnv_mode" "I2S")]) + + +(define_insn "fixuns_truncdfsi2" + [(set (match_operand:SI 0 "register_operand" "=r") + (unsigned_fix:SI (match_operand:DF 1 "register_operand" "f")))] + "TARGET_HARD_FLOAT" + "fcvt.wu.d %0,%1,rtz" + [(set_attr "type" "fcvt") + (set_attr "mode" "DF") + (set_attr "cnv_mode" "D2I")]) + + +(define_insn "fixuns_truncsfsi2" + [(set (match_operand:SI 0 "register_operand" "=r") + (unsigned_fix:SI (match_operand:SF 1 "register_operand" "f")))] + "TARGET_HARD_FLOAT" + "fcvt.wu.s %0,%1,rtz" + [(set_attr "type" "fcvt") + (set_attr "mode" "SF") + (set_attr "cnv_mode" "S2I")]) + + +(define_insn "fixuns_truncdfdi2" + [(set (match_operand:DI 0 "register_operand" "=r") + (unsigned_fix:DI (match_operand:DF 1 "register_operand" "f")))] + "TARGET_HARD_FLOAT && TARGET_64BIT" + "fcvt.lu.d %0,%1,rtz" + [(set_attr "type" "fcvt") + (set_attr "mode" "DF") + (set_attr "cnv_mode" "D2I")]) + + +(define_insn "fixuns_truncsfdi2" + [(set (match_operand:DI 0 "register_operand" "=r") + (unsigned_fix:DI (match_operand:SF 1 "register_operand" "f")))] + "TARGET_HARD_FLOAT && TARGET_64BIT" + "fcvt.lu.s %0,%1,rtz" + [(set_attr "type" "fcvt") + (set_attr "mode" "SF") + (set_attr "cnv_mode" "S2I")]) + +;; +;; .................... +;; +;; DATA MOVEMENT +;; +;; .................... + +;; Lower-level instructions for loading an address from the GOT. +;; We could use MEMs, but an unspec gives more optimization +;; opportunities. + +(define_insn "got_load" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "symbolic_operand" "")] + UNSPEC_LOAD_GOT))] + "flag_pic" + "la\t%0,%1" + [(set_attr "got" "load") + (set_attr "mode" "")]) + +(define_insn "tls_add_tp_le" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "register_operand" "r") + (match_operand:P 2 "register_operand" "r") + (match_operand:P 3 "symbolic_operand" "")] + UNSPEC_TLS_LE))] + "!flag_pic || flag_pie" + "add\t%0,%1,%2,%%tprel_add(%3)" + [(set_attr "type" "arith") + (set_attr "mode" "")]) + +(define_insn "got_load_tls_gd" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "symbolic_operand" "")] + UNSPEC_TLS_GD))] + "flag_pic" + "la.tls.gd\t%0,%1" + [(set_attr "got" "load") + (set_attr "mode" "")]) + +(define_insn "got_load_tls_ie" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "symbolic_operand" "")] + UNSPEC_TLS_IE))] + "flag_pic" + "la.tls.ie\t%0,%1" + [(set_attr "got" "load") + (set_attr "mode" "")]) + +;; Instructions for adding the low 16 bits of an address to a register. +;; Operand 2 is the address: riscv_print_operand works out which relocation +;; should be applied. + +(define_insn "*low" + [(set (match_operand:P 0 "register_operand" "=r") + (lo_sum:P (match_operand:P 1 "register_operand" "r") + (match_operand:P 2 "immediate_operand" "")))] + "" + "add\t%0,%1,%R2" + [(set_attr "alu_type" "add") + (set_attr "mode" "")]) + +;; Allow combine to split complex const_int load sequences, using operand 2 +;; to store the intermediate results. See move_operand for details. +(define_split + [(set (match_operand:GPR 0 "register_operand") + (match_operand:GPR 1 "splittable_const_int_operand")) + (clobber (match_operand:GPR 2 "register_operand"))] + "" + [(const_int 0)] +{ + riscv_move_integer (operands[2], operands[0], INTVAL (operands[1])); + DONE; +}) + +;; Likewise, for symbolic operands. +(define_split + [(set (match_operand:P 0 "register_operand") + (match_operand:P 1)) + (clobber (match_operand:P 2 "register_operand"))] + "riscv_split_symbol (operands[2], operands[1], MAX_MACHINE_MODE, NULL)" + [(set (match_dup 0) (match_dup 3))] +{ + riscv_split_symbol (operands[2], operands[1], + MAX_MACHINE_MODE, &operands[3]); +}) + +;; 64-bit integer moves + +;; Unlike most other insns, the move insns can't be split with ' +;; different predicates, because register spilling and other parts of +;; the compiler, have memoized the insn number already. + +(define_expand "movdi" + [(set (match_operand:DI 0 "") + (match_operand:DI 1 ""))] + "" +{ + if (riscv_legitimize_move (DImode, operands[0], operands[1])) + DONE; +}) + +(define_insn "*movdi_32bit" + [(set (match_operand:DI 0 "nonimmediate_operand" "=r,r,r,m,*f,*f,*r,*m") + (match_operand:DI 1 "move_operand" "r,i,m,r,*J*r,*m,*f,*f"))] + "!TARGET_64BIT + && (register_operand (operands[0], DImode) + || reg_or_0_operand (operands[1], DImode))" + { return riscv_output_move (operands[0], operands[1]); } + [(set_attr "move_type" "move,const,load,store,mtc,fpload,mfc,fpstore") + (set_attr "mode" "DI")]) + +(define_insn "*movdi_64bit" + [(set (match_operand:DI 0 "nonimmediate_operand" "=r,r,r,m,*f,*f,*r,*m") + (match_operand:DI 1 "move_operand" "r,T,m,rJ,*r*J,*m,*f,*f"))] + "TARGET_64BIT + && (register_operand (operands[0], DImode) + || reg_or_0_operand (operands[1], DImode))" + { return riscv_output_move (operands[0], operands[1]); } + [(set_attr "move_type" "move,const,load,store,mtc,fpload,mfc,fpstore") + (set_attr "mode" "DI")]) + +;; 32-bit Integer moves + +;; Unlike most other insns, the move insns can't be split with +;; different predicates, because register spilling and other parts of +;; the compiler, have memoized the insn number already. + +(define_expand "mov" + [(set (match_operand:IMOVE32 0 "") + (match_operand:IMOVE32 1 ""))] + "" +{ + if (riscv_legitimize_move (mode, operands[0], operands[1])) + DONE; +}) + +;; The difference between these two is whether or not ints are allowed +;; in FP registers (off by default, use -mdebugh to enable). + +(define_insn "*mov_internal" + [(set (match_operand:IMOVE32 0 "nonimmediate_operand" "=r,r,r,m,*f,*f,*r,*m") + (match_operand:IMOVE32 1 "move_operand" "r,T,m,rJ,*r*J,*m,*f,*f"))] + "(register_operand (operands[0], mode) + || reg_or_0_operand (operands[1], mode))" + { return riscv_output_move (operands[0], operands[1]); } + [(set_attr "move_type" "move,const,load,store,mtc,fpload,mfc,fpstore") + (set_attr "mode" "SI")]) + +;; 16-bit Integer moves + +;; Unlike most other insns, the move insns can't be split with +;; different predicates, because register spilling and other parts of +;; the compiler, have memoized the insn number already. +;; Unsigned loads are used because LOAD_EXTEND_OP returns ZERO_EXTEND. + +(define_expand "movhi" + [(set (match_operand:HI 0 "") + (match_operand:HI 1 ""))] + "" +{ + if (riscv_legitimize_move (HImode, operands[0], operands[1])) + DONE; +}) + +(define_insn "*movhi_internal" + [(set (match_operand:HI 0 "nonimmediate_operand" "=r,r,r,m,*f,*r") + (match_operand:HI 1 "move_operand" "r,T,m,rJ,*r*J,*f"))] + "(register_operand (operands[0], HImode) + || reg_or_0_operand (operands[1], HImode))" + { return riscv_output_move (operands[0], operands[1]); } + [(set_attr "move_type" "move,const,load,store,mtc,mfc") + (set_attr "mode" "HI")]) + +;; HImode constant generation; see riscv_move_integer for details. +;; si+si->hi without truncation is legal because of TRULY_NOOP_TRUNCATION. + +(define_insn "addhi3" + [(set (match_operand:HI 0 "register_operand" "=r,r") + (plus:HI (match_operand:HISI 1 "register_operand" "r,r") + (match_operand:HISI 2 "arith_operand" "r,Q")))] + "" + { return TARGET_64BIT ? "addw\t%0,%1,%2" : "add\t%0,%1,%2"; } + [(set_attr "type" "arith") + (set_attr "mode" "HI")]) + +(define_insn "xorhi3" + [(set (match_operand:HI 0 "register_operand" "=r,r") + (xor:HI (match_operand:HISI 1 "register_operand" "r,r") + (match_operand:HISI 2 "arith_operand" "r,Q")))] + "" + "xor\t%0,%1,%2" + [(set_attr "type" "logical") + (set_attr "mode" "HI")]) + +;; 8-bit Integer moves + +(define_expand "movqi" + [(set (match_operand:QI 0 "") + (match_operand:QI 1 ""))] + "" +{ + if (riscv_legitimize_move (QImode, operands[0], operands[1])) + DONE; +}) + +(define_insn "*movqi_internal" + [(set (match_operand:QI 0 "nonimmediate_operand" "=r,r,r,m,*f,*r") + (match_operand:QI 1 "move_operand" "r,I,m,rJ,*r*J,*f"))] + "(register_operand (operands[0], QImode) + || reg_or_0_operand (operands[1], QImode))" + { return riscv_output_move (operands[0], operands[1]); } + [(set_attr "move_type" "move,const,load,store,mtc,mfc") + (set_attr "mode" "QI")]) + +;; 32-bit floating point moves + +(define_expand "movsf" + [(set (match_operand:SF 0 "") + (match_operand:SF 1 ""))] + "" +{ + if (riscv_legitimize_move (SFmode, operands[0], operands[1])) + DONE; +}) + +(define_insn "*movsf_hardfloat" + [(set (match_operand:SF 0 "nonimmediate_operand" "=f,f,f,m,m,*f,*r,*r,*r,*m") + (match_operand:SF 1 "move_operand" "f,G,m,f,G,*r,*f,*G*r,*m,*r"))] + "TARGET_HARD_FLOAT + && (register_operand (operands[0], SFmode) + || reg_or_0_operand (operands[1], SFmode))" + { return riscv_output_move (operands[0], operands[1]); } + [(set_attr "move_type" "fmove,mtc,fpload,fpstore,store,mtc,mfc,move,load,store") + (set_attr "mode" "SF")]) + +(define_insn "*movsf_softfloat" + [(set (match_operand:SF 0 "nonimmediate_operand" "=r,r,m") + (match_operand:SF 1 "move_operand" "Gr,m,r"))] + "TARGET_SOFT_FLOAT + && (register_operand (operands[0], SFmode) + || reg_or_0_operand (operands[1], SFmode))" + { return riscv_output_move (operands[0], operands[1]); } + [(set_attr "move_type" "move,load,store") + (set_attr "mode" "SF")]) + +;; 64-bit floating point moves + +(define_expand "movdf" + [(set (match_operand:DF 0 "") + (match_operand:DF 1 ""))] + "" +{ + if (riscv_legitimize_move (DFmode, operands[0], operands[1])) + DONE; +}) + +;; In RV32, we lack mtf.d/mff.d. Go through memory instead. +;; (except for moving a constant 0 to an FPR. for that we use fcvt.d.w.) +(define_insn "*movdf_hardfloat_rv32" + [(set (match_operand:DF 0 "nonimmediate_operand" "=f,f,f,m,m,*r,*r,*m") + (match_operand:DF 1 "move_operand" "f,G,m,f,G,*r*G,*m,*r"))] + "!TARGET_64BIT && TARGET_HARD_FLOAT + && (register_operand (operands[0], DFmode) + || reg_or_0_operand (operands[1], DFmode))" + { return riscv_output_move (operands[0], operands[1]); } + [(set_attr "move_type" "fmove,mtc,fpload,fpstore,store,move,load,store") + (set_attr "mode" "DF")]) + +(define_insn "*movdf_hardfloat_rv64" + [(set (match_operand:DF 0 "nonimmediate_operand" "=f,f,f,m,m,*f,*r,*r,*r,*m") + (match_operand:DF 1 "move_operand" "f,G,m,f,G,*r,*f,*r*G,*m,*r"))] + "TARGET_64BIT && TARGET_HARD_FLOAT + && (register_operand (operands[0], DFmode) + || reg_or_0_operand (operands[1], DFmode))" + { return riscv_output_move (operands[0], operands[1]); } + [(set_attr "move_type" "fmove,mtc,fpload,fpstore,store,mtc,mfc,move,load,store") + (set_attr "mode" "DF")]) + +(define_insn "*movdf_softfloat" + [(set (match_operand:DF 0 "nonimmediate_operand" "=r,r,m") + (match_operand:DF 1 "move_operand" "rG,m,rG"))] + "TARGET_SOFT_FLOAT + && (register_operand (operands[0], DFmode) + || reg_or_0_operand (operands[1], DFmode))" + { return riscv_output_move (operands[0], operands[1]); } + [(set_attr "move_type" "move,load,store") + (set_attr "mode" "DF")]) + +;; 128-bit integer moves + +(define_expand "movti" + [(set (match_operand:TI 0) + (match_operand:TI 1))] + "TARGET_64BIT" +{ + if (riscv_legitimize_move (TImode, operands[0], operands[1])) + DONE; +}) + +(define_insn "*movti" + [(set (match_operand:TI 0 "nonimmediate_operand" "=r,r,r,m") + (match_operand:TI 1 "move_operand" "r,i,m,rJ"))] + "TARGET_64BIT + && (register_operand (operands[0], TImode) + || reg_or_0_operand (operands[1], TImode))" + "#" + [(set_attr "move_type" "move,const,load,store") + (set_attr "mode" "TI")]) + +(define_split + [(set (match_operand:MOVE64 0 "nonimmediate_operand") + (match_operand:MOVE64 1 "move_operand"))] + "reload_completed && !TARGET_64BIT + && riscv_split_64bit_move_p (operands[0], operands[1])" + [(const_int 0)] +{ + riscv_split_doubleword_move (operands[0], operands[1]); + DONE; +}) + +(define_split + [(set (match_operand:MOVE128 0 "nonimmediate_operand") + (match_operand:MOVE128 1 "move_operand"))] + "TARGET_64BIT && reload_completed" + [(const_int 0)] +{ + riscv_split_doubleword_move (operands[0], operands[1]); + DONE; +}) + +;; 64-bit paired-single floating point moves + +;; Load the low word of operand 0 with operand 1. +(define_insn "load_low" + [(set (match_operand:SPLITF 0 "register_operand" "=f,f") + (unspec:SPLITF [(match_operand: 1 "general_operand" "rJ,m")] + UNSPEC_LOAD_LOW))] + "TARGET_HARD_FLOAT" +{ + operands[0] = riscv_subword (operands[0], 0); + return riscv_output_move (operands[0], operands[1]); +} + [(set_attr "move_type" "mtc,fpload") + (set_attr "mode" "")]) + +;; Load the high word of operand 0 from operand 1, preserving the value +;; in the low word. +(define_insn "load_high" + [(set (match_operand:SPLITF 0 "register_operand" "=f,f") + (unspec:SPLITF [(match_operand: 1 "general_operand" "rJ,m") + (match_operand:SPLITF 2 "register_operand" "0,0")] + UNSPEC_LOAD_HIGH))] + "TARGET_HARD_FLOAT" +{ + operands[0] = riscv_subword (operands[0], 1); + return riscv_output_move (operands[0], operands[1]); +} + [(set_attr "move_type" "mtc,fpload") + (set_attr "mode" "")]) + +;; Store one word of operand 1 in operand 0. Operand 2 is 1 to store the +;; high word and 0 to store the low word. +(define_insn "store_word" + [(set (match_operand: 0 "nonimmediate_operand" "=r,m") + (unspec: [(match_operand:SPLITF 1 "register_operand" "f,f") + (match_operand 2 "const_int_operand")] + UNSPEC_STORE_WORD))] + "TARGET_HARD_FLOAT" +{ + operands[1] = riscv_subword (operands[1], INTVAL (operands[2])); + return riscv_output_move (operands[0], operands[1]); +} + [(set_attr "move_type" "mfc,fpstore") + (set_attr "mode" "")]) + +;; Expand in-line code to clear the instruction cache between operand[0] and +;; operand[1]. +(define_expand "clear_cache" + [(match_operand 0 "pmode_register_operand") + (match_operand 1 "pmode_register_operand")] + "" + " +{ + emit_insn(gen_fence_i()); + DONE; +}") + +(define_insn "fence" + [(unspec_volatile [(const_int 0)] UNSPEC_FENCE)] + "" + "%|fence%-") + +(define_insn "fence_i" + [(unspec_volatile [(const_int 0)] UNSPEC_FENCE_I)] + "" + "fence.i") + +;; Block moves, see riscv.c for more details. +;; Argument 0 is the destination +;; Argument 1 is the source +;; Argument 2 is the length +;; Argument 3 is the alignment + +(define_expand "movmemsi" + [(parallel [(set (match_operand:BLK 0 "general_operand") + (match_operand:BLK 1 "general_operand")) + (use (match_operand:SI 2 "")) + (use (match_operand:SI 3 "const_int_operand"))])] + "!TARGET_MEMCPY" +{ + if (riscv_expand_block_move (operands[0], operands[1], operands[2])) + DONE; + else + FAIL; +}) + +;; +;; .................... +;; +;; SHIFTS +;; +;; .................... + +(define_insn "si3" + [(set (match_operand:SI 0 "register_operand" "=r") + (any_shift:SI (match_operand:SI 1 "register_operand" "r") + (match_operand:SI 2 "arith_operand" "rI")))] + "" +{ + if (GET_CODE (operands[2]) == CONST_INT) + operands[2] = GEN_INT (INTVAL (operands[2]) + & (GET_MODE_BITSIZE (SImode) - 1)); + + return TARGET_64BIT ? "w\t%0,%1,%2" : "\t%0,%1,%2"; +} + [(set_attr "type" "shift") + (set_attr "mode" "SI")]) + +(define_insn "*disi3" + [(set (match_operand:SI 0 "register_operand" "=r") + (any_shift:SI (truncate:SI (match_operand:DI 1 "register_operand" "r")) + (truncate:SI (match_operand:DI 2 "arith_operand" "rI"))))] + "TARGET_64BIT" + "w\t%0,%1,%2" + [(set_attr "type" "shift") + (set_attr "mode" "SI")]) + +(define_insn "*ashldi3_truncsi" + [(set (match_operand:SI 0 "register_operand" "=r") + (truncate:SI + (ashift:DI (match_operand:DI 1 "register_operand" "r") + (match_operand:DI 2 "const_arith_operand" "I"))))] + "TARGET_64BIT && INTVAL (operands[2]) < 32" + "sllw\t%0,%1,%2" + [(set_attr "type" "shift") + (set_attr "mode" "SI")]) + +(define_insn "*ashldisi3" + [(set (match_operand:SI 0 "register_operand" "=r") + (ashift:SI (match_operand:GPR 1 "register_operand" "r") + (match_operand:GPR2 2 "arith_operand" "rI")))] + "TARGET_64BIT && (GET_CODE (operands[2]) == CONST_INT ? INTVAL (operands[2]) < 32 : 1)" + "sllw\t%0,%1,%2" + [(set_attr "type" "shift") + (set_attr "mode" "SI")]) + +(define_insn "di3" + [(set (match_operand:DI 0 "register_operand" "=r") + (any_shift:DI (match_operand:DI 1 "register_operand" "r") + (match_operand:DI 2 "arith_operand" "rI")))] + "TARGET_64BIT" +{ + if (GET_CODE (operands[2]) == CONST_INT) + operands[2] = GEN_INT (INTVAL (operands[2]) + & (GET_MODE_BITSIZE (DImode) - 1)); + + return "\t%0,%1,%2"; +} + [(set_attr "type" "shift") + (set_attr "mode" "DI")]) + +(define_insn "si3_extend" + [(set (match_operand:DI 0 "register_operand" "=r") + (sign_extend:DI + (any_shift:SI (match_operand:SI 1 "register_operand" "r") + (match_operand:SI 2 "arith_operand" "rI"))))] + "TARGET_64BIT" +{ + if (GET_CODE (operands[2]) == CONST_INT) + operands[2] = GEN_INT (INTVAL (operands[2]) & 0x1f); + + return "w\t%0,%1,%2"; +} + [(set_attr "type" "shift") + (set_attr "mode" "SI")]) + +;; +;; .................... +;; +;; CONDITIONAL BRANCHES +;; +;; .................... + +;; Conditional branches + +(define_insn "*branch_order" + [(set (pc) + (if_then_else + (match_operator 1 "order_operator" + [(match_operand:GPR 2 "register_operand" "r") + (match_operand:GPR 3 "reg_or_0_operand" "rJ")]) + (label_ref (match_operand 0 "" "")) + (pc)))] + "" +{ + if (GET_CODE (operands[3]) == CONST_INT) + return "b%C1z\t%2,%0"; + return "b%C1\t%2,%3,%0"; +} + [(set_attr "type" "branch") + (set_attr "mode" "none")]) + +;; Used to implement built-in functions. +(define_expand "condjump" + [(set (pc) + (if_then_else (match_operand 0) + (label_ref (match_operand 1)) + (pc)))]) + +(define_expand "cbranch4" + [(set (pc) + (if_then_else (match_operator 0 "comparison_operator" + [(match_operand:GPR 1 "register_operand") + (match_operand:GPR 2 "nonmemory_operand")]) + (label_ref (match_operand 3 "")) + (pc)))] + "" +{ + riscv_expand_conditional_branch (operands); + DONE; +}) + +(define_expand "cbranch4" + [(set (pc) + (if_then_else (match_operator 0 "comparison_operator" + [(match_operand:SCALARF 1 "register_operand") + (match_operand:SCALARF 2 "register_operand")]) + (label_ref (match_operand 3 "")) + (pc)))] + "" +{ + riscv_expand_conditional_branch (operands); + DONE; +}) + +(define_insn_and_split "*branch_on_bit" + [(set (pc) + (if_then_else + (match_operator 0 "equality_operator" + [(zero_extract:GPR (match_operand:GPR 2 "register_operand" "r") + (const_int 1) + (match_operand 3 "branch_on_bit_operand")) + (const_int 0)]) + (label_ref (match_operand 1)) + (pc))) + (clobber (match_scratch:GPR 4 "=&r"))] + "" + "#" + "reload_completed" + [(set (match_dup 4) + (ashift:GPR (match_dup 2) (match_dup 3))) + (set (pc) + (if_then_else + (match_op_dup 0 [(match_dup 4) (const_int 0)]) + (label_ref (match_operand 1)) + (pc)))] +{ + int shift = GET_MODE_BITSIZE (mode) - 1 - INTVAL (operands[3]); + operands[3] = GEN_INT (shift); + + if (GET_CODE (operands[0]) == EQ) + operands[0] = gen_rtx_GE (mode, operands[4], const0_rtx); + else + operands[0] = gen_rtx_LT (mode, operands[4], const0_rtx); +}) + +(define_insn_and_split "*branch_on_bit_range" + [(set (pc) + (if_then_else + (match_operator 0 "equality_operator" + [(zero_extract:GPR (match_operand:GPR 2 "register_operand" "r") + (match_operand 3 "branch_on_bit_operand") + (const_int 0)) + (const_int 0)]) + (label_ref (match_operand 1)) + (pc))) + (clobber (match_scratch:GPR 4 "=&r"))] + "" + "#" + "reload_completed" + [(set (match_dup 4) + (ashift:GPR (match_dup 2) (match_dup 3))) + (set (pc) + (if_then_else + (match_op_dup 0 [(match_dup 4) (const_int 0)]) + (label_ref (match_operand 1)) + (pc)))] +{ + operands[3] = GEN_INT (GET_MODE_BITSIZE (mode) - INTVAL (operands[3])); +}) + +;; +;; .................... +;; +;; SETTING A REGISTER FROM A COMPARISON +;; +;; .................... + +;; Destination is always set in SI mode. + +(define_expand "cstore4" + [(set (match_operand:SI 0 "register_operand") + (match_operator:SI 1 "order_operator" + [(match_operand:GPR 2 "register_operand") + (match_operand:GPR 3 "nonmemory_operand")]))] + "" +{ + riscv_expand_scc (operands); + DONE; +}) + +(define_insn "cstore4" + [(set (match_operand:SI 0 "register_operand" "=r") + (match_operator:SI 1 "fp_order_operator" + [(match_operand:SCALARF 2 "register_operand" "f") + (match_operand:SCALARF 3 "register_operand" "f")]))] + "TARGET_HARD_FLOAT" + "f%C1.\t%0,%2,%3" + [(set_attr "type" "fcmp") + (set_attr "mode" "")]) + +(define_insn "*seq_zero_" + [(set (match_operand:GPR2 0 "register_operand" "=r") + (eq:GPR2 (match_operand:GPR 1 "register_operand" "r") + (const_int 0)))] + "" + "seqz\t%0,%1" + [(set_attr "type" "slt") + (set_attr "mode" "")]) + +(define_insn "*sne_zero_" + [(set (match_operand:GPR2 0 "register_operand" "=r") + (ne:GPR2 (match_operand:GPR 1 "register_operand" "r") + (const_int 0)))] + "" + "snez\t%0,%1" + [(set_attr "type" "slt") + (set_attr "mode" "")]) + +(define_insn "*sgt_" + [(set (match_operand:GPR2 0 "register_operand" "=r") + (any_gt:GPR2 (match_operand:GPR 1 "register_operand" "r") + (match_operand:GPR 2 "reg_or_0_operand" "rJ")))] + "" + "slt\t%0,%z2,%1" + [(set_attr "type" "slt") + (set_attr "mode" "")]) + +(define_insn "*sge_" + [(set (match_operand:GPR2 0 "register_operand" "=r") + (any_ge:GPR2 (match_operand:GPR 1 "register_operand" "r") + (const_int 1)))] + "" + "slt\t%0,zero,%1" + [(set_attr "type" "slt") + (set_attr "mode" "")]) + +(define_insn "*slt_" + [(set (match_operand:GPR2 0 "register_operand" "=r") + (any_lt:GPR2 (match_operand:GPR 1 "register_operand" "r") + (match_operand:GPR 2 "arith_operand" "rI")))] + "" + "slt\t%0,%1,%2" + [(set_attr "type" "slt") + (set_attr "mode" "")]) + +(define_insn "*sle_" + [(set (match_operand:GPR2 0 "register_operand" "=r") + (any_le:GPR2 (match_operand:GPR 1 "register_operand" "r") + (match_operand:GPR 2 "sle_operand" "")))] + "" +{ + operands[2] = GEN_INT (INTVAL (operands[2]) + 1); + return "slt\t%0,%1,%2"; +} + [(set_attr "type" "slt") + (set_attr "mode" "")]) + +;; +;; .................... +;; +;; UNCONDITIONAL BRANCHES +;; +;; .................... + +;; Unconditional branches. + +(define_insn "jump" + [(set (pc) + (label_ref (match_operand 0 "" "")))] + "" + "j\t%l0" + [(set_attr "type" "jump") + (set_attr "mode" "none")]) + +(define_expand "indirect_jump" + [(set (pc) (match_operand 0 "register_operand"))] + "" +{ + operands[0] = force_reg (Pmode, operands[0]); + if (Pmode == SImode) + emit_jump_insn (gen_indirect_jumpsi (operands[0])); + else + emit_jump_insn (gen_indirect_jumpdi (operands[0])); + DONE; +}) + +(define_insn "indirect_jump" + [(set (pc) (match_operand:P 0 "register_operand" "l"))] + "" + "jr\t%0" + [(set_attr "type" "jump") + (set_attr "mode" "none")]) + +(define_expand "tablejump" + [(set (pc) (match_operand 0 "register_operand" "")) + (use (label_ref (match_operand 1 "" "")))] + "" +{ + if (CASE_VECTOR_PC_RELATIVE) + operands[0] = expand_simple_binop (Pmode, PLUS, operands[0], + gen_rtx_LABEL_REF (Pmode, operands[1]), + NULL_RTX, 0, OPTAB_DIRECT); + + if (CASE_VECTOR_PC_RELATIVE && Pmode == DImode) + emit_jump_insn (gen_tablejumpdi (operands[0], operands[1])); + else + emit_jump_insn (gen_tablejumpsi (operands[0], operands[1])); + DONE; +}) + +(define_insn "tablejump" + [(set (pc) (match_operand:GPR 0 "register_operand" "l")) + (use (label_ref (match_operand 1 "" "")))] + "" + "jr\t%0" + [(set_attr "type" "jump") + (set_attr "mode" "none")]) + +;; +;; .................... +;; +;; Function prologue/epilogue +;; +;; .................... +;; + +(define_expand "prologue" + [(const_int 1)] + "" +{ + riscv_expand_prologue (); + DONE; +}) + +;; Block any insns from being moved before this point, since the +;; profiling call to mcount can use various registers that aren't +;; saved or used to pass arguments. + +(define_insn "blockage" + [(unspec_volatile [(const_int 0)] UNSPEC_BLOCKAGE)] + "" + "" + [(set_attr "type" "ghost") + (set_attr "mode" "none")]) + +(define_expand "epilogue" + [(const_int 2)] + "" +{ + riscv_expand_epilogue (false); + DONE; +}) + +(define_expand "sibcall_epilogue" + [(const_int 2)] + "" +{ + riscv_expand_epilogue (true); + DONE; +}) + +;; Trivial return. Make it look like a normal return insn as that +;; allows jump optimizations to work better. + +(define_expand "return" + [(simple_return)] + "riscv_can_use_return_insn ()" + "") + +(define_insn "simple_return" + [(simple_return)] + "" + "ret" + [(set_attr "type" "jump") + (set_attr "mode" "none")]) + +;; Normal return. + +(define_insn "simple_return_internal" + [(simple_return) + (use (match_operand 0 "pmode_register_operand" ""))] + "" + "jr\t%0" + [(set_attr "type" "jump") + (set_attr "mode" "none")]) + +;; This is used in compiling the unwind routines. +(define_expand "eh_return" + [(use (match_operand 0 "general_operand"))] + "" +{ + if (GET_MODE (operands[0]) != word_mode) + operands[0] = convert_to_mode (word_mode, operands[0], 0); + if (TARGET_64BIT) + emit_insn (gen_eh_set_lr_di (operands[0])); + else + emit_insn (gen_eh_set_lr_si (operands[0])); + DONE; +}) + +;; Clobber the return address on the stack. We can't expand this +;; until we know where it will be put in the stack frame. + +(define_insn "eh_set_lr_si" + [(unspec [(match_operand:SI 0 "register_operand" "r")] UNSPEC_EH_RETURN) + (clobber (match_scratch:SI 1 "=&r"))] + "! TARGET_64BIT" + "#") + +(define_insn "eh_set_lr_di" + [(unspec [(match_operand:DI 0 "register_operand" "r")] UNSPEC_EH_RETURN) + (clobber (match_scratch:DI 1 "=&r"))] + "TARGET_64BIT" + "#") + +(define_split + [(unspec [(match_operand 0 "register_operand")] UNSPEC_EH_RETURN) + (clobber (match_scratch 1))] + "reload_completed" + [(const_int 0)] +{ + riscv_set_return_address (operands[0], operands[1]); + DONE; +}) + +;; +;; .................... +;; +;; FUNCTION CALLS +;; +;; .................... + +;; Sibling calls. All these patterns use jump instructions. + +;; call_insn_operand will only accept constant +;; addresses if a direct jump is acceptable. Since the 'S' constraint +;; is defined in terms of call_insn_operand, the same is true of the +;; constraints. + +;; When we use an indirect jump, we need a register that will be +;; preserved by the epilogue (constraint j). + +(define_expand "sibcall" + [(parallel [(call (match_operand 0 "") + (match_operand 1 "")) + (use (match_operand 2 "")) ;; next_arg_reg + (use (match_operand 3 ""))])] ;; struct_value_size_rtx + "" +{ + riscv_expand_call (true, NULL_RTX, XEXP (operands[0], 0), operands[1]); + DONE; +}) + +(define_insn "sibcall_internal" + [(call (mem:SI (match_operand 0 "call_insn_operand" "j,S")) + (match_operand 1 "" ""))] + "SIBLING_CALL_P (insn)" + { return REG_P (operands[0]) ? "jr\t%0" + : absolute_symbolic_operand (operands[0], VOIDmode) ? "tail\t%0" + : "tail\t%0@"; } + [(set_attr "type" "call")]) + +(define_expand "sibcall_value" + [(parallel [(set (match_operand 0 "") + (call (match_operand 1 "") + (match_operand 2 ""))) + (use (match_operand 3 ""))])] ;; next_arg_reg + "" +{ + riscv_expand_call (true, operands[0], XEXP (operands[1], 0), operands[2]); + DONE; +}) + +(define_insn "sibcall_value_internal" + [(set (match_operand 0 "register_operand" "") + (call (mem:SI (match_operand 1 "call_insn_operand" "j,S")) + (match_operand 2 "" "")))] + "SIBLING_CALL_P (insn)" + { return REG_P (operands[1]) ? "jr\t%1" + : absolute_symbolic_operand (operands[1], VOIDmode) ? "tail\t%1" + : "tail\t%1@"; } + [(set_attr "type" "call")]) + +(define_insn "sibcall_value_multiple_internal" + [(set (match_operand 0 "register_operand" "") + (call (mem:SI (match_operand 1 "call_insn_operand" "j,S")) + (match_operand 2 "" ""))) + (set (match_operand 3 "register_operand" "") + (call (mem:SI (match_dup 1)) + (match_dup 2)))] + "SIBLING_CALL_P (insn)" + { return REG_P (operands[1]) ? "jr\t%1" + : absolute_symbolic_operand (operands[1], VOIDmode) ? "tail\t%1" + : "tail\t%1@"; } + [(set_attr "type" "call")]) + +(define_expand "call" + [(parallel [(call (match_operand 0 "") + (match_operand 1 "")) + (use (match_operand 2 "")) ;; next_arg_reg + (use (match_operand 3 ""))])] ;; struct_value_size_rtx + "" +{ + riscv_expand_call (false, NULL_RTX, XEXP (operands[0], 0), operands[1]); + DONE; +}) + +(define_insn "call_internal" + [(call (mem:SI (match_operand 0 "call_insn_operand" "l,S")) + (match_operand 1 "" "")) + (clobber (reg:SI RETURN_ADDR_REGNUM))] + "" + { return REG_P (operands[0]) ? "jalr\t%0" + : absolute_symbolic_operand (operands[0], VOIDmode) ? "call\t%0" + : "call\t%0@"; } + [(set_attr "jal" "indirect,direct")]) + +(define_expand "call_value" + [(parallel [(set (match_operand 0 "") + (call (match_operand 1 "") + (match_operand 2 ""))) + (use (match_operand 3 ""))])] ;; next_arg_reg + "" +{ + riscv_expand_call (false, operands[0], XEXP (operands[1], 0), operands[2]); + DONE; +}) + +;; See comment for call_internal. +(define_insn "call_value_internal" + [(set (match_operand 0 "register_operand" "") + (call (mem:SI (match_operand 1 "call_insn_operand" "l,S")) + (match_operand 2 "" ""))) + (clobber (reg:SI RETURN_ADDR_REGNUM))] + "" + { return REG_P (operands[1]) ? "jalr\t%1" + : absolute_symbolic_operand (operands[1], VOIDmode) ? "call\t%1" + : "call\t%1@"; } + [(set_attr "jal" "indirect,direct")]) + +;; See comment for call_internal. +(define_insn "call_value_multiple_internal" + [(set (match_operand 0 "register_operand" "") + (call (mem:SI (match_operand 1 "call_insn_operand" "l,S")) + (match_operand 2 "" ""))) + (set (match_operand 3 "register_operand" "") + (call (mem:SI (match_dup 1)) + (match_dup 2))) + (clobber (reg:SI RETURN_ADDR_REGNUM))] + "" + { return REG_P (operands[1]) ? "jalr\t%1" + : absolute_symbolic_operand (operands[1], VOIDmode) ? "call\t%1" + : "call\t%1@"; } + [(set_attr "jal" "indirect,direct")]) + +;; Call subroutine returning any type. + +(define_expand "untyped_call" + [(parallel [(call (match_operand 0 "") + (const_int 0)) + (match_operand 1 "") + (match_operand 2 "")])] + "" +{ + int i; + + emit_call_insn (GEN_CALL (operands[0], const0_rtx, NULL, const0_rtx)); + + for (i = 0; i < XVECLEN (operands[2], 0); i++) + { + rtx set = XVECEXP (operands[2], 0, i); + riscv_emit_move (SET_DEST (set), SET_SRC (set)); + } + + emit_insn (gen_blockage ()); + DONE; +}) + +(define_insn "nop" + [(const_int 0)] + "" + "nop" + [(set_attr "type" "nop") + (set_attr "mode" "none")]) + +(define_insn "trap" + [(trap_if (const_int 1) (const_int 0))] + "" + "sbreak") + +(define_insn "gpr_save" + [(unspec_volatile [(match_operand 0 "const_int_operand")] UNSPEC_GPR_SAVE) + (clobber (reg:SI T0_REGNUM)) + (clobber (reg:SI T1_REGNUM))] + "" + { return riscv_output_gpr_save (INTVAL (operands[0])); }) + +(define_insn "gpr_restore" + [(unspec_volatile [(match_operand 0 "const_int_operand")] UNSPEC_GPR_RESTORE)] + "" + "tail\t__riscv_restore_%0") + +(define_insn "gpr_restore_return" + [(return) + (use (match_operand 0 "pmode_register_operand" "")) + (const_int 0)] + "" + "") + +(include "sync.md") +(include "peephole.md") diff --git a/gcc/config/riscv/riscv.opt b/gcc/config/riscv/riscv.opt new file mode 100644 index 0000000..d31105d --- /dev/null +++ b/gcc/config/riscv/riscv.opt @@ -0,0 +1,87 @@ +; Options for the MIPS port of the compiler +; +; Copyright (C) 2005, 2007, 2008, 2010, 2011 Free Software Foundation, Inc. +; +; This file is part of GCC. +; +; GCC is free software; you can redistribute it and/or modify it under +; the terms of the GNU General Public License as published by the Free +; Software Foundation; either version 3, or (at your option) any later +; version. +; +; GCC is distributed in the hope that it will be useful, but WITHOUT +; ANY WARRANTY; without even the implied warranty of MERCHANTABILITY +; or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public +; License for more details. +; +; You should have received a copy of the GNU General Public License +; along with GCC; see the file COPYING3. If not see +; . + +m32 +Target RejectNegative Mask(32BIT) +Generate RV32 code + +m64 +Target RejectNegative InverseMask(32BIT, 64BIT) +Generate RV64 code + +mbranch-cost= +Target RejectNegative Joined UInteger Var(riscv_branch_cost) +-mbranch-cost=COST Set the cost of branches to roughly COST instructions + +mhard-float +Target Report RejectNegative InverseMask(SOFT_FLOAT_ABI, HARD_FLOAT_ABI) +Allow the use of hardware floating-point ABI and instructions + +mmemcpy +Target Report Mask(MEMCPY) +Don't optimize block moves + +mplt +Target Report Var(TARGET_PLT) Init(1) +When generating -fpic code, allow the use of PLTs. Ignored for fno-pic. + +msoft-float +Target Report RejectNegative Mask(SOFT_FLOAT_ABI) +Prevent the use of all hardware floating-point instructions + +mfdiv +Target Report RejectNegative Mask(FDIV) +Use hardware floating-point divide and square root instructions + +march= +Target RejectNegative Joined Var(riscv_arch_string) +-march= Generate code for given RISC-V ISA (e.g. RV64IM) + +mtune= +Target RejectNegative Joined Var(riscv_tune_string) +-mtune=PROCESSOR Optimize the output for PROCESSOR + +msmall-data-limit= +Target Joined Separate UInteger Var(g_switch_value) Init(8) +-msmall-data-limit= Put global and static data smaller than bytes into a special section (on some targets) + +matomic +Target Report Mask(ATOMIC) +Use hardware atomic memory instructions. + +mmuldiv +Target Report Mask(MULDIV) +Use hardware instructions for integer multiplication and division. + +mrvc +Target Report Mask(RVC) +Use compressed instruction encoding + +msave-restore +Target Report Mask(SAVE_RESTORE) +Use smaller but slower prologue and epilogue code + +mlra +Target Report Var(riscv_lra_flag) Init(0) Save +Use LRA instead of reload + +mcmodel= +Target RejectNegative Joined Var(riscv_cmodel_string) +Use given RISC-V code model (medlow or medany) diff --git a/gcc/config/riscv/sync.md b/gcc/config/riscv/sync.md new file mode 100644 index 0000000..c4bd8d9 --- /dev/null +++ b/gcc/config/riscv/sync.md @@ -0,0 +1,198 @@ +;; Machine description for RISC-V atomic operations. +;; Copyright (C) 2011-2014 Free Software Foundation, Inc. +;; Contributed by Andrew Waterman (waterman@cs.berkeley.edu) at UC Berkeley. +;; Based on MIPS target for GNU compiler. + +;; This file is part of GCC. + +;; GCC is free software; you can redistribute it and/or modify +;; it under the terms of the GNU General Public License as published by +;; the Free Software Foundation; either version 3, or (at your option) +;; any later version. + +;; GCC is distributed in the hope that it will be useful, +;; but WITHOUT ANY WARRANTY; without even the implied warranty of +;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +;; GNU General Public License for more details. + +;; You should have received a copy of the GNU General Public License +;; along with GCC; see the file COPYING3. If not see +;; . + +(define_c_enum "unspec" [ + UNSPEC_COMPARE_AND_SWAP + UNSPEC_SYNC_OLD_OP + UNSPEC_SYNC_EXCHANGE + UNSPEC_ATOMIC_STORE + UNSPEC_MEMORY_BARRIER +]) + +(define_code_iterator any_atomic [plus ior xor and]) +(define_code_attr atomic_optab + [(plus "add") (ior "or") (xor "xor") (and "and")]) + +;; Memory barriers. + +(define_expand "mem_thread_fence" + [(match_operand:SI 0 "const_int_operand" "")] ;; model + "" +{ + if (INTVAL (operands[0]) != MEMMODEL_RELAXED) + { + rtx mem = gen_rtx_MEM (BLKmode, gen_rtx_SCRATCH (Pmode)); + MEM_VOLATILE_P (mem) = 1; + emit_insn (gen_mem_thread_fence_1 (mem, operands[0])); + } + DONE; +}) + +(define_insn "mem_thread_fence_1" + [(set (match_operand:BLK 0 "" "") + (unspec:BLK [(match_dup 0)] UNSPEC_MEMORY_BARRIER)) + (match_operand:SI 1 "const_int_operand" "")] ;; model + "" +{ + switch (INTVAL (operands[1])) + { + case MEMMODEL_SEQ_CST: + case MEMMODEL_ACQ_REL: + return "fence rw,rw"; + case MEMMODEL_ACQUIRE: + case MEMMODEL_CONSUME: + return "fence r,rw"; + case MEMMODEL_RELEASE: + return "fence rw,w"; + default: + gcc_unreachable(); + } +}) + +;; Atomic memory operations. + +;; Implement atomic stores with amoswap. Fall back to fences for atomic loads. +(define_insn "atomic_store" + [(set (match_operand:GPR 0 "memory_operand" "=A") + (unspec_volatile:GPR + [(match_operand:GPR 1 "reg_or_0_operand" "rJ") + (match_operand:SI 2 "const_int_operand")] ;; model + UNSPEC_ATOMIC_STORE))] + "TARGET_ATOMIC" + "amoswap.%A2 zero,%z1,%0") + +(define_insn "atomic_" + [(set (match_operand:GPR 0 "memory_operand" "+A") + (unspec_volatile:GPR + [(any_atomic:GPR (match_dup 0) + (match_operand:GPR 1 "reg_or_0_operand" "rJ")) + (match_operand:SI 2 "const_int_operand")] ;; model + UNSPEC_SYNC_OLD_OP))] + "TARGET_ATOMIC" + "amo.%A2 zero,%z1,%0") + +(define_insn "atomic_fetch_" + [(set (match_operand:GPR 0 "register_operand" "=&r") + (match_operand:GPR 1 "memory_operand" "+A")) + (set (match_dup 1) + (unspec_volatile:GPR + [(any_atomic:GPR (match_dup 1) + (match_operand:GPR 2 "reg_or_0_operand" "rJ")) + (match_operand:SI 3 "const_int_operand")] ;; model + UNSPEC_SYNC_OLD_OP))] + "TARGET_ATOMIC" + "amo.%A3 %0,%z2,%1") + +(define_insn "atomic_exchange" + [(set (match_operand:GPR 0 "register_operand" "=&r") + (unspec_volatile:GPR + [(match_operand:GPR 1 "memory_operand" "+A") + (match_operand:SI 3 "const_int_operand")] ;; model + UNSPEC_SYNC_EXCHANGE)) + (set (match_dup 1) + (match_operand:GPR 2 "register_operand" "0"))] + "TARGET_ATOMIC" + "amoswap.%A3 %0,%z2,%1") + +(define_insn "atomic_cas_value_strong" + [(set (match_operand:GPR 0 "register_operand" "=&r") + (match_operand:GPR 1 "memory_operand" "+A")) + (set (match_dup 1) + (unspec_volatile:GPR [(match_operand:GPR 2 "reg_or_0_operand" "rJ") + (match_operand:GPR 3 "reg_or_0_operand" "rJ") + (match_operand:SI 4 "const_int_operand") ;; mod_s + (match_operand:SI 5 "const_int_operand")] ;; mod_f + UNSPEC_COMPARE_AND_SWAP)) + (clobber (match_scratch:GPR 6 "=&r"))] + "TARGET_ATOMIC" + "1: lr.%A5 %0,%1; bne %0,%z2,1f; sc.%A4 %6,%z3,%1; bnez %6,1b; 1:" + [(set (attr "length") (const_int 16))]) + +(define_expand "atomic_compare_and_swap" + [(match_operand:SI 0 "register_operand" "") ;; bool output + (match_operand:GPR 1 "register_operand" "") ;; val output + (match_operand:GPR 2 "memory_operand" "") ;; memory + (match_operand:GPR 3 "reg_or_0_operand" "") ;; expected value + (match_operand:GPR 4 "reg_or_0_operand" "") ;; desired value + (match_operand:SI 5 "const_int_operand" "") ;; is_weak + (match_operand:SI 6 "const_int_operand" "") ;; mod_s + (match_operand:SI 7 "const_int_operand" "")] ;; mod_f + "TARGET_ATOMIC" +{ + emit_insn (gen_atomic_cas_value_strong (operands[1], operands[2], + operands[3], operands[4], + operands[6], operands[7])); + + rtx compare = operands[1]; + if (operands[3] != const0_rtx) + { + rtx difference = gen_rtx_MINUS (mode, operands[1], operands[3]); + compare = gen_reg_rtx (mode); + emit_insn (gen_rtx_SET (VOIDmode, compare, difference)); + } + + rtx eq = gen_rtx_EQ (mode, compare, const0_rtx); + rtx result = gen_reg_rtx (mode); + emit_insn (gen_rtx_SET (VOIDmode, result, eq)); + emit_insn (gen_rtx_SET (VOIDmode, operands[0], gen_lowpart (SImode, result))); + DONE; +}) + +(define_expand "atomic_test_and_set" + [(match_operand:QI 0 "register_operand" "") ;; bool output + (match_operand:QI 1 "memory_operand" "+A") ;; memory + (match_operand:SI 2 "const_int_operand" "")] ;; model + "TARGET_ATOMIC" +{ + /* We have no QImode atomics, so use the address LSBs to form a mask, + then use an aligned SImode atomic. */ + rtx result = operands[0]; + rtx mem = operands[1]; + rtx model = operands[2]; + rtx addr = force_reg (Pmode, XEXP (mem, 0)); + + rtx aligned_addr = gen_reg_rtx (Pmode); + emit_move_insn (aligned_addr, gen_rtx_AND (Pmode, addr, GEN_INT (-4))); + + rtx aligned_mem = change_address (mem, SImode, aligned_addr); + set_mem_alias_set (aligned_mem, 0); + + rtx offset = gen_reg_rtx (SImode); + emit_move_insn (offset, gen_rtx_AND (SImode, gen_lowpart (SImode, addr), + GEN_INT (3))); + + rtx tmp = gen_reg_rtx (SImode); + emit_move_insn (tmp, GEN_INT (1)); + + rtx shmt = gen_reg_rtx (SImode); + emit_move_insn (shmt, gen_rtx_ASHIFT (SImode, offset, GEN_INT (3))); + + rtx word = gen_reg_rtx (SImode); + emit_move_insn (word, gen_rtx_ASHIFT (SImode, tmp, shmt)); + + tmp = gen_reg_rtx (SImode); + emit_insn (gen_atomic_fetch_orsi (tmp, aligned_mem, word, model)); + + emit_move_insn (gen_lowpart (SImode, result), + gen_rtx_LSHIFTRT (SImode, tmp, + gen_lowpart (SImode, shmt))); + DONE; +}) diff --git a/gcc/config/riscv/t-elf b/gcc/config/riscv/t-elf new file mode 100644 index 0000000..8339026 --- /dev/null +++ b/gcc/config/riscv/t-elf @@ -0,0 +1,4 @@ +# Build the libraries for both hard and soft floating point + +MULTILIB_OPTIONS = msoft-float m64/m32 mno-atomic +MULTILIB_DIRNAMES = soft-float 64 32 no-atomic diff --git a/gcc/config/riscv/t-linux64 b/gcc/config/riscv/t-linux64 new file mode 100644 index 0000000..71eeeb2 --- /dev/null +++ b/gcc/config/riscv/t-linux64 @@ -0,0 +1,5 @@ +# Build the libraries for both hard and soft floating point + +MULTILIB_OPTIONS = m64/m32 msoft-float mno-atomic +MULTILIB_DIRNAMES = 64 32 soft-float no-atomic +MULTILIB_OSDIRNAMES = ../lib ../lib32 diff --git a/libgcc/config/riscv/crti.S b/libgcc/config/riscv/crti.S new file mode 100644 index 0000000..89bac70 --- /dev/null +++ b/libgcc/config/riscv/crti.S @@ -0,0 +1 @@ +/* crti.S is empty because .init_array/.fini_array are used exclusively. */ diff --git a/libgcc/config/riscv/crtn.S b/libgcc/config/riscv/crtn.S new file mode 100644 index 0000000..ca6ee7b --- /dev/null +++ b/libgcc/config/riscv/crtn.S @@ -0,0 +1 @@ +/* crtn.S is empty because .init_array/.fini_array are used exclusively. */ diff --git a/libgcc/config/riscv/div.S b/libgcc/config/riscv/div.S new file mode 100644 index 0000000..385634a --- /dev/null +++ b/libgcc/config/riscv/div.S @@ -0,0 +1,121 @@ + .text + .align 2 + +#ifndef __riscv64 +/* Our RV64 64-bit routines are equivalent to our RV32 32-bit routines. */ +# define __udivdi3 __udivsi3 +# define __umoddi3 __umodsi3 +# define __divdi3 __divsi3 +# define __moddi3 __modsi3 +#else + .globl __udivsi3 +__udivsi3: + /* Compute __udivdi3(a0 << 32, a1 << 32); cast result to uint32_t. */ + sll a0, a0, 32 + sll a1, a1, 32 + move t0, ra + jal __udivdi3 + sext.w a0, a0 + jr t0 + + .globl __umodsi3 +__umodsi3: + /* Compute __udivdi3((uint32_t)a0, (uint32_t)a1); cast a1 to uint32_t. */ + sll a0, a0, 32 + sll a1, a1, 32 + srl a0, a0, 32 + srl a1, a1, 32 + move t0, ra + jal __udivdi3 + sext.w a0, a1 + jr t0 + + .globl __modsi3 + __modsi3 = __moddi3 + + .globl __divsi3 +__divsi3: + /* Check for special case of INT_MIN/-1. Otherwise, fall into __divdi3. */ + li t0, -1 + beq a1, t0, .L20 +#endif + + .globl __divdi3 +__divdi3: + bltz a0, .L10 + bltz a1, .L11 + /* Since the quotient is positive, fall into __udivdi3. */ + + .globl __udivdi3 +__udivdi3: + mv a2, a1 + mv a1, a0 + li a0, -1 + beqz a2, .L5 + li a3, 1 + bgeu a2, a1, .L2 +.L1: + blez a2, .L2 + slli a2, a2, 1 + slli a3, a3, 1 + bgtu a1, a2, .L1 +.L2: + li a0, 0 +.L3: + bltu a1, a2, .L4 + sub a1, a1, a2 + or a0, a0, a3 +.L4: + srli a3, a3, 1 + srli a2, a2, 1 + bnez a3, .L3 +.L5: + ret + + .globl __umoddi3 +__umoddi3: + /* Call __udivdi3(a0, a1), then return the remainder, which is in a1. */ + move t0, ra + jal __udivdi3 + move a0, a1 + jr t0 + + /* Handle negative arguments to __divdi3. */ +.L10: + neg a0, a0 + bgez a1, .L12 /* Compute __udivdi3(-a0, a1), then negate the result. */ + neg a1, a1 + j __divdi3 /* Compute __udivdi3(-a0, -a1). */ +.L11: /* Compute __udivdi3(a0, -a1), then negate the result. */ + neg a1, a1 +.L12: + move t0, ra + jal __divdi3 + neg a0, a0 + jr t0 + + .globl __moddi3 +__moddi3: + move t0, ra + bltz a1, .L31 + bltz a0, .L32 +.L30: + jal __udivdi3 /* The dividend is not negative. */ + move a0, a1 + jr t0 +.L31: + neg a1, a1 + bgez a0, .L30 +.L32: + neg a0, a0 + jal __udivdi3 /* The dividend is hella negative. */ + neg a0, a1 + jr t0 + +#ifdef __riscv64 + /* continuation of __divsi3 */ +.L20: + sll t0, t0, 31 + bne a0, t0, __divdi3 + ret +#endif diff --git a/libgcc/config/riscv/mul.S b/libgcc/config/riscv/mul.S new file mode 100644 index 0000000..5c93541 --- /dev/null +++ b/libgcc/config/riscv/mul.S @@ -0,0 +1,21 @@ + .text + .align 2 + +#ifndef __riscv64 +/* Our RV64 64-bit routine is equivalent to our RV32 32-bit routine. */ +# define __muldi3 __mulsi3 +#endif + + .globl __muldi3 +__muldi3: + mv a2, a0 + li a0, 0 +.L1: + slli a3, a1, _RISCV_SZPTR-1 + bgez a3, .L2 + add a0, a0, a2 +.L2: + srli a1, a1, 1 + slli a2, a2, 1 + bnez a1, .L1 + ret diff --git a/libgcc/config/riscv/riscv-fp.c b/libgcc/config/riscv/riscv-fp.c new file mode 100644 index 0000000..6d7a1f1 --- /dev/null +++ b/libgcc/config/riscv/riscv-fp.c @@ -0,0 +1,178 @@ +/* Functions needed for soft-float on riscv-linux. Based on + rs6000/ppc64-fp.c with TF types removed. + + Copyright (C) 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, + 2000, 2001, 2002, 2003, 2004, 2006, 2009 Free Software Foundation, + Inc. + +This file is part of GCC. + +GCC is free software; you can redistribute it and/or modify it under +the terms of the GNU General Public License as published by the Free +Software Foundation; either version 3, or (at your option) any later +version. + +GCC is distributed in the hope that it will be useful, but WITHOUT ANY +WARRANTY; without even the implied warranty of MERCHANTABILITY or +FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License +for more details. + +Under Section 7 of GPL version 3, you are granted additional +permissions described in the GCC Runtime Library Exception, version +3.1, as published by the Free Software Foundation. + +You should have received a copy of the GNU General Public License and +a copy of the GCC Runtime Library Exception along with this program; +see the files COPYING3 and COPYING.RUNTIME respectively. If not, see +. */ + +#if defined(__riscv64) +#include "fp-bit.h" + +extern DItype __fixdfdi (DFtype); +extern DItype __fixsfdi (SFtype); +extern USItype __fixunsdfsi (DFtype); +extern USItype __fixunssfsi (SFtype); +extern DFtype __floatdidf (DItype); +extern DFtype __floatundidf (UDItype); +extern SFtype __floatdisf (DItype); +extern SFtype __floatundisf (UDItype); + +static DItype local_fixunssfdi (SFtype); +static DItype local_fixunsdfdi (DFtype); + +DItype +__fixdfdi (DFtype a) +{ + if (a < 0) + return - local_fixunsdfdi (-a); + return local_fixunsdfdi (a); +} + +DItype +__fixsfdi (SFtype a) +{ + if (a < 0) + return - local_fixunssfdi (-a); + return local_fixunssfdi (a); +} + +USItype +__fixunsdfsi (DFtype a) +{ + if (a >= - (DFtype) (- ((SItype)(((USItype)1 << ((4 * 8) - 1)) - 1)) - 1)) + return (SItype) (a + (- ((SItype)(((USItype)1 << ((4 * 8) - 1)) - 1)) - 1)) + - (- ((SItype)(((USItype)1 << ((4 * 8) - 1)) - 1)) - 1); + return (SItype) a; +} + +USItype +__fixunssfsi (SFtype a) +{ + if (a >= - (SFtype) (- ((SItype)(((USItype)1 << ((4 * 8) - 1)) - 1)) - 1)) + return (SItype) (a + (- ((SItype)(((USItype)1 << ((4 * 8) - 1)) - 1)) - 1)) + - (- ((SItype)(((USItype)1 << ((4 * 8) - 1)) - 1)) - 1); + return (SItype) a; +} + +DFtype +__floatdidf (DItype u) +{ + DFtype d; + + d = (SItype) (u >> (sizeof (SItype) * 8)); + d *= 2.0 * (((UDItype) 1) << ((sizeof (SItype) * 8) - 1)); + d += (USItype) (u & ((((UDItype) 1) << (sizeof (SItype) * 8)) - 1)); + + return d; +} + +DFtype +__floatundidf (UDItype u) +{ + DFtype d; + + d = (USItype) (u >> (sizeof (SItype) * 8)); + d *= 2.0 * (((UDItype) 1) << ((sizeof (SItype) * 8) - 1)); + d += (USItype) (u & ((((UDItype) 1) << (sizeof (SItype) * 8)) - 1)); + + return d; +} + +SFtype +__floatdisf (DItype u) +{ + DFtype f; + + if (53 < (sizeof (DItype) * 8) + && 53 > ((sizeof (DItype) * 8) - 53 + 24)) + { + if (! (- ((DItype) 1 << 53) < u + && u < ((DItype) 1 << 53))) + { + if ((UDItype) u & (((UDItype) 1 << ((sizeof (DItype) * 8) - 53)) - 1)) + { + u &= ~ (((UDItype) 1 << ((sizeof (DItype) * 8) - 53)) - 1); + u |= ((UDItype) 1 << ((sizeof (DItype) * 8) - 53)); + } + } + } + f = (SItype) (u >> (sizeof (SItype) * 8)); + f *= 2.0 * (((UDItype) 1) << ((sizeof (SItype) * 8) - 1)); + f += (USItype) (u & ((((UDItype) 1) << (sizeof (SItype) * 8)) - 1)); + + return (SFtype) f; +} + +SFtype +__floatundisf (UDItype u) +{ + DFtype f; + + if (53 < (sizeof (DItype) * 8) + && 53 > ((sizeof (DItype) * 8) - 53 + 24)) + { + if (u >= ((UDItype) 1 << 53)) + { + if ((UDItype) u & (((UDItype) 1 << ((sizeof (DItype) * 8) - 53)) - 1)) + { + u &= ~ (((UDItype) 1 << ((sizeof (DItype) * 8) - 53)) - 1); + u |= ((UDItype) 1 << ((sizeof (DItype) * 8) - 53)); + } + } + } + f = (USItype) (u >> (sizeof (SItype) * 8)); + f *= 2.0 * (((UDItype) 1) << ((sizeof (SItype) * 8) - 1)); + f += (USItype) (u & ((((UDItype) 1) << (sizeof (SItype) * 8)) - 1)); + + return (SFtype) f; +} + +/* This version is needed to prevent recursion; fixunsdfdi in libgcc + calls fixdfdi, which in turn calls calls fixunsdfdi. */ + +static DItype +local_fixunsdfdi (DFtype a) +{ + USItype hi, lo; + + hi = a / (((UDItype) 1) << (sizeof (SItype) * 8)); + lo = (a - ((DFtype) hi) * (((UDItype) 1) << (sizeof (SItype) * 8))); + return ((UDItype) hi << (sizeof (SItype) * 8)) | lo; +} + +/* This version is needed to prevent recursion; fixunssfdi in libgcc + calls fixsfdi, which in turn calls calls fixunssfdi. */ + +static DItype +local_fixunssfdi (SFtype original_a) +{ + DFtype a = original_a; + USItype hi, lo; + + hi = a / (((UDItype) 1) << (sizeof (SItype) * 8)); + lo = (a - ((DFtype) hi) * (((UDItype) 1) << (sizeof (SItype) * 8))); + return ((UDItype) hi << (sizeof (SItype) * 8)) | lo; +} + +#endif diff --git a/libgcc/config/riscv/save-restore.S b/libgcc/config/riscv/save-restore.S new file mode 100644 index 0000000..bbf0e33 --- /dev/null +++ b/libgcc/config/riscv/save-restore.S @@ -0,0 +1,220 @@ + .text + + .globl __riscv_save_12 + .globl __riscv_save_11 + .globl __riscv_save_10 + .globl __riscv_save_9 + .globl __riscv_save_8 + .globl __riscv_save_7 + .globl __riscv_save_6 + .globl __riscv_save_5 + .globl __riscv_save_4 + .globl __riscv_save_3 + .globl __riscv_save_2 + .globl __riscv_save_1 + .globl __riscv_save_0 + + .globl __riscv_restore_12 + .globl __riscv_restore_11 + .globl __riscv_restore_10 + .globl __riscv_restore_9 + .globl __riscv_restore_8 + .globl __riscv_restore_7 + .globl __riscv_restore_6 + .globl __riscv_restore_5 + .globl __riscv_restore_4 + .globl __riscv_restore_3 + .globl __riscv_restore_2 + .globl __riscv_restore_1 + .globl __riscv_restore_0 + +#ifdef __riscv64 + +__riscv_save_12: + addi sp, sp, -112 + li t1, 0 + sd s11, 8(sp) + j .Ls10 + +__riscv_save_11: +__riscv_save_10: + addi sp, sp, -112 + li t1, -16 +.Ls10: + sd s10, 16(sp) + sd s9, 24(sp) + j .Ls8 + +__riscv_save_9: +__riscv_save_8: + addi sp, sp, -112 + li t1, -32 +.Ls8: + sd s8, 32(sp) + sd s7, 40(sp) + j .Ls6 + +__riscv_save_7: +__riscv_save_6: + addi sp, sp, -112 + li t1, -48 +.Ls6: + sd s6, 48(sp) + sd s5, 56(sp) + j .Ls4 + +__riscv_save_5: +__riscv_save_4: + addi sp, sp, -112 + li t1, -64 +.Ls4: + sd s4, 64(sp) + sd s3, 72(sp) + j .Ls2 + +__riscv_save_3: +__riscv_save_2: + addi sp, sp, -112 + li t1, -80 +.Ls2: + sd s2, 80(sp) + sd s1, 88(sp) + sd s0, 96(sp) + sd ra, 104(sp) + sub sp, sp, t1 + jr t0 + +__riscv_save_1: +__riscv_save_0: + addi sp, sp, -16 + sd s0, 0(sp) + sd ra, 8(sp) + jr t0 + +__riscv_restore_12: + ld s11, 8(sp) + addi sp, sp, 16 + +__riscv_restore_11: +__riscv_restore_10: + ld s10, 0(sp) + ld s9, 8(sp) + addi sp, sp, 16 + +__riscv_restore_9: +__riscv_restore_8: + ld s8, 0(sp) + ld s7, 8(sp) + addi sp, sp, 16 + +__riscv_restore_7: +__riscv_restore_6: + ld s6, 0(sp) + ld s5, 8(sp) + addi sp, sp, 16 + +__riscv_restore_5: +__riscv_restore_4: + ld s4, 0(sp) + ld s3, 8(sp) + addi sp, sp, 16 + +__riscv_restore_3: +__riscv_restore_2: + ld s2, 0(sp) + ld s1, 8(sp) + addi sp, sp, 16 + +__riscv_restore_1: +__riscv_restore_0: + ld s0, 0(sp) + ld ra, 8(sp) + addi sp, sp, 16 + ret + +#else + +__riscv_save_12: + addi sp, sp, -64 + li t1, 0 + sw s11, 12(sp) + j .Ls10 + +__riscv_save_11: +__riscv_save_10: +__riscv_save_9: +__riscv_save_8: + addi sp, sp, -64 + li t1, -16 +.Ls10: + sw s10, 16(sp) + sw s9, 20(sp) + sw s8, 24(sp) + sw s7, 28(sp) + j .Ls6 + +__riscv_save_7: +__riscv_save_6: +__riscv_save_5: +__riscv_save_4: + addi sp, sp, -64 + li t1, -32 +.Ls6: + sw s6, 32(sp) + sw s5, 36(sp) + sw s4, 40(sp) + sw s3, 44(sp) + sw s2, 48(sp) + sw s1, 52(sp) + sw s0, 56(sp) + sw ra, 60(sp) + sub sp, sp, t1 + jr t0 + +__riscv_save_3: +__riscv_save_2: +__riscv_save_1: +__riscv_save_0: + addi sp, sp, -16 + sw s2, 0(sp) + sw s1, 4(sp) + sw s0, 8(sp) + sw ra, 12(sp) + jr t0 + +__riscv_restore_12: + lw s11, 12(sp) + addi sp, sp, 16 + +__riscv_restore_11: +__riscv_restore_10: +__riscv_restore_9: +__riscv_restore_8: + lw s10, 0(sp) + lw s9, 4(sp) + lw s8, 8(sp) + lw s7, 12(sp) + addi sp, sp, 16 + +__riscv_restore_7: +__riscv_restore_6: +__riscv_restore_5: +__riscv_restore_4: + lw s6, 0(sp) + lw s5, 4(sp) + lw s4, 8(sp) + lw s3, 12(sp) + addi sp, sp, 16 + +__riscv_restore_3: +__riscv_restore_2: +__riscv_restore_1: +__riscv_restore_0: + lw s2, 0(sp) + lw s1, 4(sp) + lw s0, 8(sp) + lw ra, 12(sp) + addi sp, sp, 16 + ret + +#endif diff --git a/libgcc/config/riscv/t-dpbit b/libgcc/config/riscv/t-dpbit new file mode 100644 index 0000000..b2f2c17 --- /dev/null +++ b/libgcc/config/riscv/t-dpbit @@ -0,0 +1,4 @@ +LIB2ADD += dp-bit.c + +dp-bit.c: $(srcdir)/fp-bit.c + cat $(srcdir)/fp-bit.c > dp-bit.c diff --git a/libgcc/config/riscv/t-elf b/libgcc/config/riscv/t-elf new file mode 100644 index 0000000..7b5800f --- /dev/null +++ b/libgcc/config/riscv/t-elf @@ -0,0 +1,4 @@ +LIB2ADD += $(srcdir)/config/riscv/riscv-fp.c \ + $(srcdir)/config/riscv/save-restore.S \ + $(srcdir)/config/riscv/mul.S \ + $(srcdir)/config/riscv/div.S diff --git a/libgcc/config/riscv/t-elf32 b/libgcc/config/riscv/t-elf32 new file mode 100644 index 0000000..3eed9ac --- /dev/null +++ b/libgcc/config/riscv/t-elf32 @@ -0,0 +1,2 @@ +HOST_LIBGCC2_CFLAGS += -m32 +CRTSTUFF_CFLAGS += -m32 diff --git a/libgcc/config/riscv/t-fpbit b/libgcc/config/riscv/t-fpbit new file mode 100644 index 0000000..772add5 --- /dev/null +++ b/libgcc/config/riscv/t-fpbit @@ -0,0 +1,5 @@ +LIB2ADD += fp-bit.c + +fp-bit.c: $(srcdir)/fp-bit.c + echo '#define FLOAT' > fp-bit.c + cat $(srcdir)/fp-bit.c >> fp-bit.c diff --git a/libgcc/config/riscv/t-tpbit b/libgcc/config/riscv/t-tpbit new file mode 100644 index 0000000..d5df179 --- /dev/null +++ b/libgcc/config/riscv/t-tpbit @@ -0,0 +1,10 @@ +LIB2ADD += tp-bit.c + +tp-bit.c: $(srcdir)/fp-bit.c + echo '#ifdef _RISCVEL' > tp-bit.c + echo '# define FLOAT_BIT_ORDER_MISMATCH' >> tp-bit.c + echo '#endif' >> tp-bit.c + echo '#if __LDBL_MANT_DIG__ == 113' >> tp-bit.c + echo '# define TFLOAT' >> tp-bit.c + cat $(srcdir)/fp-bit.c >> tp-bit.c + echo '#endif' >> tp-bit.c