diff options
author | Michael Meissner <meissner@gcc.gnu.org> | 2008-07-23 10:28:06 +0000 |
---|---|---|
committer | Michael Meissner <meissner@gcc.gnu.org> | 2008-07-23 10:28:06 +0000 |
commit | ab442df7fb453434d80a779844fe1a10c0c802ab (patch) | |
tree | efd8e61a3d2ff9dcff5eb5bf03e25922191f7df5 /gcc/config | |
parent | 5295185c3150a8d31685dc44248aa058246bbe73 (diff) | |
download | gcc-ab442df7fb453434d80a779844fe1a10c0c802ab.zip gcc-ab442df7fb453434d80a779844fe1a10c0c802ab.tar.gz gcc-ab442df7fb453434d80a779844fe1a10c0c802ab.tar.bz2 |
Add ability to set target options (ix86 only) and optimization options on a function specific basis
From-SVN: r138075
Diffstat (limited to 'gcc/config')
-rw-r--r-- | gcc/config/i386/darwin.h | 4 | ||||
-rw-r--r-- | gcc/config/i386/i386-c.c | 344 | ||||
-rw-r--r-- | gcc/config/i386/i386-protos.h | 7 | ||||
-rw-r--r-- | gcc/config/i386/i386.c | 1188 | ||||
-rw-r--r-- | gcc/config/i386/i386.h | 215 | ||||
-rw-r--r-- | gcc/config/i386/i386.opt | 163 | ||||
-rw-r--r-- | gcc/config/i386/t-i386 | 13 |
7 files changed, 1526 insertions, 408 deletions
diff --git a/gcc/config/i386/darwin.h b/gcc/config/i386/darwin.h index 6001f64..c6ed10d 100644 --- a/gcc/config/i386/darwin.h +++ b/gcc/config/i386/darwin.h @@ -263,8 +263,8 @@ extern void darwin_x86_file_end (void); : (n) >= 11 && (n) <= 18 ? (n) + 1 \ : (n)) -#undef REGISTER_TARGET_PRAGMAS -#define REGISTER_TARGET_PRAGMAS() DARWIN_REGISTER_TARGET_PRAGMAS() +#undef REGISTER_SUBTARGET_PRAGMAS +#define REGISTER_SUBTARGET_PRAGMAS() DARWIN_REGISTER_TARGET_PRAGMAS() #undef TARGET_SET_DEFAULT_TYPE_ATTRIBUTES #define TARGET_SET_DEFAULT_TYPE_ATTRIBUTES darwin_set_default_type_attributes diff --git a/gcc/config/i386/i386-c.c b/gcc/config/i386/i386-c.c new file mode 100644 index 0000000..f0a3a17 --- /dev/null +++ b/gcc/config/i386/i386-c.c @@ -0,0 +1,344 @@ +/* Subroutines used for macro/preprocessor support on the ia-32. + Copyright (C) 2008 + 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 +<http://www.gnu.org/licenses/>. */ + +#include "config.h" +#include "system.h" +#include "coretypes.h" +#include "tm.h" +#include "rtl.h" +#include "tree.h" +#include "tm_p.h" +#include "flags.h" +#include "c-common.h" +#include "ggc.h" +#include "target.h" +#include "target-def.h" +#include "cpplib.h" +#include "c-pragma.h" + +static bool ix86_pragma_option_parse (tree); +static void ix86_target_macros_internal + (int, enum processor_type, enum processor_type, enum fpmath_unit, + void (*def_or_undef) (cpp_reader *, const char *)); + + +/* Internal function to either define or undef the appropriate system + macros. */ +static void +ix86_target_macros_internal (int isa_flag, + enum processor_type arch, + enum processor_type tune, + enum fpmath_unit fpmath, + void (*def_or_undef) (cpp_reader *, + const char *)) +{ + /* For some of the k6/pentium varients there weren't seperate ISA bits to + identify which tune/arch flag was passed, so figure it out here. */ + size_t arch_len = strlen (ix86_arch_string); + size_t tune_len = strlen (ix86_tune_string); + int last_arch_char = ix86_arch_string[arch_len - 1]; + int last_tune_char = ix86_tune_string[tune_len - 1]; + + /* Built-ins based on -march=. */ + switch (arch) + { + case PROCESSOR_I386: + break; + case PROCESSOR_I486: + def_or_undef (parse_in, "__i486"); + def_or_undef (parse_in, "__i486__"); + break; + case PROCESSOR_PENTIUM: + def_or_undef (parse_in, "__i586"); + def_or_undef (parse_in, "__i586__"); + def_or_undef (parse_in, "__pentium"); + def_or_undef (parse_in, "__pentium__"); + if (isa_flag & OPTION_MASK_ISA_MMX) + def_or_undef (parse_in, "__pentium_mmx__"); + break; + case PROCESSOR_PENTIUMPRO: + def_or_undef (parse_in, "__i686"); + def_or_undef (parse_in, "__i686__"); + def_or_undef (parse_in, "__pentiumpro"); + def_or_undef (parse_in, "__pentiumpro__"); + break; + case PROCESSOR_GEODE: + def_or_undef (parse_in, "__geode"); + def_or_undef (parse_in, "__geode__"); + break; + case PROCESSOR_K6: + def_or_undef (parse_in, "__k6"); + def_or_undef (parse_in, "__k6__"); + if (last_arch_char == '2') + def_or_undef (parse_in, "__k6_2__"); + else if (last_arch_char == '3') + def_or_undef (parse_in, "__k6_3__"); + else if (isa_flag & OPTION_MASK_ISA_3DNOW) + def_or_undef (parse_in, "__k6_3__"); + break; + case PROCESSOR_ATHLON: + def_or_undef (parse_in, "__athlon"); + def_or_undef (parse_in, "__athlon__"); + if (isa_flag & OPTION_MASK_ISA_SSE) + def_or_undef (parse_in, "__athlon_sse__"); + break; + case PROCESSOR_K8: + def_or_undef (parse_in, "__k8"); + def_or_undef (parse_in, "__k8__"); + break; + case PROCESSOR_AMDFAM10: + def_or_undef (parse_in, "__amdfam10"); + def_or_undef (parse_in, "__amdfam10__"); + break; + case PROCESSOR_PENTIUM4: + def_or_undef (parse_in, "__pentium4"); + def_or_undef (parse_in, "__pentium4__"); + break; + case PROCESSOR_NOCONA: + def_or_undef (parse_in, "__nocona"); + def_or_undef (parse_in, "__nocona__"); + break; + case PROCESSOR_CORE2: + def_or_undef (parse_in, "__core2"); + def_or_undef (parse_in, "__core2__"); + break; + /* use PROCESSOR_max to not set/unset the arch macro. */ + case PROCESSOR_max: + break; + case PROCESSOR_GENERIC32: + case PROCESSOR_GENERIC64: + gcc_unreachable (); + } + + /* Built-ins based on -mtune=. */ + switch (tune) + { + case PROCESSOR_I386: + def_or_undef (parse_in, "__tune_i386__"); + break; + case PROCESSOR_I486: + def_or_undef (parse_in, "__tune_i486__"); + break; + case PROCESSOR_PENTIUM: + def_or_undef (parse_in, "__tune_i586__"); + def_or_undef (parse_in, "__tune_pentium__"); + if (last_tune_char == 'x') + def_or_undef (parse_in, "__tune_pentium_mmx__"); + break; + case PROCESSOR_PENTIUMPRO: + def_or_undef (parse_in, "__tune_i686__"); + def_or_undef (parse_in, "__tune_pentiumpro__"); + switch (last_tune_char) + { + case '3': + def_or_undef (parse_in, "__tune_pentium3__"); + /* FALLTHRU */ + case '2': + def_or_undef (parse_in, "__tune_pentium2__"); + break; + } + break; + case PROCESSOR_GEODE: + def_or_undef (parse_in, "__tune_geode__"); + break; + case PROCESSOR_K6: + def_or_undef (parse_in, "__tune_k6__"); + if (last_tune_char == '2') + def_or_undef (parse_in, "__tune_k6_2__"); + else if (last_tune_char == '3') + def_or_undef (parse_in, "__tune_k6_3__"); + else if (isa_flag & OPTION_MASK_ISA_3DNOW) + def_or_undef (parse_in, "__tune_k6_3__"); + break; + case PROCESSOR_ATHLON: + def_or_undef (parse_in, "__tune_athlon__"); + if (isa_flag & OPTION_MASK_ISA_SSE) + def_or_undef (parse_in, "__tune_athlon_sse__"); + break; + case PROCESSOR_K8: + def_or_undef (parse_in, "__tune_k8__"); + break; + case PROCESSOR_AMDFAM10: + def_or_undef (parse_in, "__tune_amdfam10__"); + break; + case PROCESSOR_PENTIUM4: + def_or_undef (parse_in, "__tune_pentium4__"); + break; + case PROCESSOR_NOCONA: + def_or_undef (parse_in, "__tune_nocona__"); + break; + case PROCESSOR_CORE2: + def_or_undef (parse_in, "__tune_core2__"); + break; + case PROCESSOR_GENERIC32: + case PROCESSOR_GENERIC64: + break; + /* use PROCESSOR_max to not set/unset the tune macro. */ + case PROCESSOR_max: + break; + } + + if (isa_flag & OPTION_MASK_ISA_MMX) + def_or_undef (parse_in, "__MMX__"); + if (isa_flag & OPTION_MASK_ISA_3DNOW) + def_or_undef (parse_in, "__3dNOW__"); + if (isa_flag & OPTION_MASK_ISA_3DNOW_A) + def_or_undef (parse_in, "__3dNOW_A__"); + if (isa_flag & OPTION_MASK_ISA_SSE) + def_or_undef (parse_in, "__SSE__"); + if (isa_flag & OPTION_MASK_ISA_SSE2) + def_or_undef (parse_in, "__SSE2__"); + if (isa_flag & OPTION_MASK_ISA_SSE3) + def_or_undef (parse_in, "__SSE3__"); + if (isa_flag & OPTION_MASK_ISA_SSSE3) + def_or_undef (parse_in, "__SSSE3__"); + if (isa_flag & OPTION_MASK_ISA_SSE4_1) + def_or_undef (parse_in, "__SSE4_1__"); + if (isa_flag & OPTION_MASK_ISA_SSE4_2) + def_or_undef (parse_in, "__SSE4_2__"); + if (isa_flag & OPTION_MASK_ISA_AES) + def_or_undef (parse_in, "__AES__"); + if (isa_flag & OPTION_MASK_ISA_PCLMUL) + def_or_undef (parse_in, "__PCLMUL__"); + if (isa_flag & OPTION_MASK_ISA_SSE4A) + def_or_undef (parse_in, "__SSE4A__"); + if (isa_flag & OPTION_MASK_ISA_SSE5) + def_or_undef (parse_in, "__SSE5__"); + if ((fpmath & FPMATH_SSE) && (isa_flag & OPTION_MASK_ISA_SSE)) + def_or_undef (parse_in, "__SSE_MATH__"); + if ((fpmath & FPMATH_SSE) && (isa_flag & OPTION_MASK_ISA_SSE2)) + def_or_undef (parse_in, "__SSE2_MATH__"); +} + + +/* Hook to validate the current #pragma option and set the state, and update + the macros based on what was changed. */ + +static bool +ix86_pragma_option_parse (tree args) +{ + tree prev_tree = build_target_option_node (); + tree cur_tree; + struct cl_target_option *prev_opt; + struct cl_target_option *cur_opt; + int prev_isa; + int cur_isa; + int diff_isa; + enum processor_type prev_arch; + enum processor_type prev_tune; + enum processor_type cur_arch; + enum processor_type cur_tune; + + if (! args) + { + cur_tree = target_option_default_node; + cl_target_option_restore (TREE_TARGET_OPTION (cur_tree)); + } + else + { + cur_tree = ix86_valid_option_attribute_tree (args); + if (!cur_tree) + return false; + } + + target_option_current_node = cur_tree; + + /* Figure out the previous/current isa, arch, tune and the differences. */ + prev_opt = TREE_TARGET_OPTION (prev_tree); + cur_opt = TREE_TARGET_OPTION (cur_tree); + prev_isa = prev_opt->ix86_isa_flags; + cur_isa = cur_opt->ix86_isa_flags; + diff_isa = (prev_isa ^ cur_isa); + prev_arch = prev_opt->arch; + prev_tune = prev_opt->tune; + cur_arch = cur_opt->arch; + cur_tune = cur_opt->tune; + + /* If the same processor is used for both previous and current options, don't + change the macros. */ + if (cur_arch == prev_arch) + cur_arch = prev_arch = PROCESSOR_max; + + if (cur_tune == prev_tune) + cur_tune = prev_tune = PROCESSOR_max; + + /* Undef all of the macros for that are no longer current. */ + ix86_target_macros_internal (prev_isa & diff_isa, + prev_arch, + prev_tune, + prev_opt->fpmath, + cpp_undef); + + /* Define all of the macros for new options that were just turned on. */ + ix86_target_macros_internal (cur_isa & diff_isa, + cur_arch, + cur_tune, + cur_opt->fpmath, + cpp_define); + + return true; +} + +/* Function to tell the preprocessor about the defines for the current target. */ + +void +ix86_target_macros (void) +{ + /* 32/64-bit won't change with target specific options, so do the assert and + builtin_define_std calls here. */ + if (TARGET_64BIT) + { + cpp_assert (parse_in, "cpu=x86_64"); + cpp_assert (parse_in, "machine=x86_64"); + cpp_define (parse_in, "__amd64"); + cpp_define (parse_in, "__amd64__"); + cpp_define (parse_in, "__x86_64"); + cpp_define (parse_in, "__x86_64__"); + } + else + { + cpp_assert (parse_in, "cpu=i386"); + cpp_assert (parse_in, "machine=i386"); + builtin_define_std ("i386"); + } + + ix86_target_macros_internal (ix86_isa_flags, + ix86_arch, + ix86_tune, + ix86_fpmath, + cpp_define); +} + + +/* Register target pragmas. We need to add the hook for parsing #pragma GCC + option here rather than in i386.c since it will pull in various preprocessor + functions, and those are not present in languages like fortran without a + preprocessor. */ + +void +ix86_register_pragmas (void) +{ + /* Update pragma hook to allow parsing #pragma GCC option. */ + targetm.target_option.pragma_parse = ix86_pragma_option_parse; + +#ifdef REGISTER_SUBTARGET_PRAGMAS + REGISTER_SUBTARGET_PRAGMAS (); +#endif +} diff --git a/gcc/config/i386/i386-protos.h b/gcc/config/i386/i386-protos.h index 634a425..3ebfd3c 100644 --- a/gcc/config/i386/i386-protos.h +++ b/gcc/config/i386/i386-protos.h @@ -20,7 +20,7 @@ along with GCC; see the file COPYING3. If not see <http://www.gnu.org/licenses/>. */ /* Functions in i386.c */ -extern void override_options (void); +extern void override_options (bool); extern void optimization_options (int, int); extern int ix86_can_use_return_insn_p (void); @@ -202,6 +202,7 @@ extern int ix86_constant_alignment (tree, int); extern tree ix86_handle_shared_attribute (tree *, tree, tree, int, bool *); extern tree ix86_handle_selectany_attribute (tree *, tree, tree, int, bool *); extern int x86_field_alignment (tree, int); +extern tree ix86_valid_option_attribute_tree (tree); #endif extern rtx ix86_tls_get_addr (void); @@ -215,6 +216,10 @@ extern void ix86_expand_reduc_v4sf (rtx (*)(rtx, rtx, rtx), rtx, rtx); extern bool ix86_sse5_valid_op_p (rtx [], rtx, int, bool, int, bool); extern void ix86_expand_sse5_multiple_memory (rtx [], int, enum machine_mode); +/* In i386-c.c */ +extern void ix86_target_macros (void); +extern void ix86_register_pragmas (void); + /* In winnt.c */ extern void i386_pe_unique_section (tree, int); extern void i386_pe_declare_function_type (FILE *, const char *, int); diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c index e68f35a..fbe4bb6 100644 --- a/gcc/config/i386/i386.c +++ b/gcc/config/i386/i386.c @@ -1210,7 +1210,11 @@ const struct processor_costs *ix86_cost = &pentium_cost; #define m_GENERIC (m_GENERIC32 | m_GENERIC64) /* Feature tests against the various tunings. */ -unsigned int ix86_tune_features[X86_TUNE_LAST] = { +unsigned char ix86_tune_features[X86_TUNE_LAST]; + +/* Feature tests against the various tunings used to create ix86_tune_features + based on the processor mask. */ +static unsigned int initial_ix86_tune_features[X86_TUNE_LAST] = { /* X86_TUNE_USE_LEAVE: Leave does not affect Nocona SPEC2000 results negatively, so enabling for Generic64 seems like good code size tradeoff. We can't enable it for 32bit generic because it does not @@ -1443,7 +1447,11 @@ unsigned int ix86_tune_features[X86_TUNE_LAST] = { }; /* Feature tests against the various architecture variations. */ -unsigned int ix86_arch_features[X86_ARCH_LAST] = { +unsigned char ix86_arch_features[X86_ARCH_LAST]; + +/* Feature tests against the various architecture variations, used to create + ix86_arch_features based on the processor mask. */ +static unsigned int initial_ix86_arch_features[X86_ARCH_LAST] = { /* X86_ARCH_CMOVE: Conditional move was added for pentiumpro. */ ~(m_386 | m_486 | m_PENT | m_K6), @@ -1773,6 +1781,26 @@ static void ix86_compute_frame_layout (struct ix86_frame *); static bool ix86_expand_vector_init_one_nonzero (bool, enum machine_mode, rtx, rtx, int); +enum ix86_function_specific_strings +{ + IX86_FUNCTION_SPECIFIC_ARCH, + IX86_FUNCTION_SPECIFIC_TUNE, + IX86_FUNCTION_SPECIFIC_FPMATH, + IX86_FUNCTION_SPECIFIC_MAX +}; + +static char *ix86_target_string (int, int, const char *, const char *, + const char *, bool); +static void ix86_debug_options (void) ATTRIBUTE_UNUSED; +static void ix86_function_specific_save (struct cl_target_option *); +static void ix86_function_specific_restore (struct cl_target_option *); +static void ix86_function_specific_print (FILE *, int, + struct cl_target_option *); +static bool ix86_valid_option_attribute_p (tree, tree, tree, int); +static bool ix86_valid_option_attribute_inner_p (tree, char *[]); +static bool ix86_can_inline_p (tree, tree); +static void ix86_set_current_function (tree); + /* The svr4 ABI for the i386 says that records and unions are returned in memory. */ @@ -1780,6 +1808,10 @@ static bool ix86_expand_vector_init_one_nonzero (bool, enum machine_mode, #define DEFAULT_PCC_STRUCT_RETURN 1 #endif +/* Whether -mtune= or -march= were specified */ +static int ix86_tune_defaulted; +static int ix86_arch_specified; + /* Bit flags that specify the ISA we are compiling for. */ int ix86_isa_flags = TARGET_64BIT_DEFAULT | TARGET_SUBTARGET_ISA_DEFAULT; @@ -1815,6 +1847,18 @@ static int ix86_isa_flags_explicit; #define OPTION_MASK_ISA_SSE5_SET \ (OPTION_MASK_ISA_SSE5 | OPTION_MASK_ISA_SSE4A_SET) +/* AES and PCLMUL need SSE2 because they use xmm registers */ +#define OPTION_MASK_ISA_AES_SET \ + (OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2_SET) +#define OPTION_MASK_ISA_PCLMUL_SET \ + (OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2_SET) + +#define OPTION_MASK_ISA_ABM_SET \ + (OPTION_MASK_ISA_ABM | OPTION_MASK_ISA_POPCNT) +#define OPTION_MASK_ISA_POPCNT_SET OPTION_MASK_ISA_POPCNT +#define OPTION_MASK_ISA_CX16_SET OPTION_MASK_ISA_CX16 +#define OPTION_MASK_ISA_SAHF_SET OPTION_MASK_ISA_SAHF + /* Define a set of ISAs which aren't available when a given ISA is disabled. MMX and SSE ISAs are handled separately. */ @@ -1844,14 +1888,73 @@ static int ix86_isa_flags_explicit; #define OPTION_MASK_ISA_SSE4A_UNSET \ (OPTION_MASK_ISA_SSE4A | OPTION_MASK_ISA_SSE5_UNSET) - #define OPTION_MASK_ISA_SSE5_UNSET OPTION_MASK_ISA_SSE5 +#define OPTION_MASK_ISA_AES_UNSET OPTION_MASK_ISA_AES +#define OPTION_MASK_ISA_PCLMUL_UNSET OPTION_MASK_ISA_PCLMUL +#define OPTION_MASK_ISA_ABM_UNSET OPTION_MASK_ISA_ABM +#define OPTION_MASK_ISA_POPCNT_UNSET OPTION_MASK_ISA_POPCNT +#define OPTION_MASK_ISA_CX16_UNSET OPTION_MASK_ISA_CX16 +#define OPTION_MASK_ISA_SAHF_UNSET OPTION_MASK_ISA_SAHF /* Vectorization library interface and handlers. */ tree (*ix86_veclib_handler)(enum built_in_function, tree, tree) = NULL; static tree ix86_veclibabi_svml (enum built_in_function, tree, tree); static tree ix86_veclibabi_acml (enum built_in_function, tree, tree); +/* Processor target table, indexed by processor number */ +struct ptt +{ + const struct processor_costs *cost; /* Processor costs */ + const int align_loop; /* Default alignments. */ + const int align_loop_max_skip; + const int align_jump; + const int align_jump_max_skip; + const int align_func; +}; + +static const struct ptt processor_target_table[PROCESSOR_max] = +{ + {&i386_cost, 4, 3, 4, 3, 4}, + {&i486_cost, 16, 15, 16, 15, 16}, + {&pentium_cost, 16, 7, 16, 7, 16}, + {&pentiumpro_cost, 16, 15, 16, 10, 16}, + {&geode_cost, 0, 0, 0, 0, 0}, + {&k6_cost, 32, 7, 32, 7, 32}, + {&athlon_cost, 16, 7, 16, 7, 16}, + {&pentium4_cost, 0, 0, 0, 0, 0}, + {&k8_cost, 16, 7, 16, 7, 16}, + {&nocona_cost, 0, 0, 0, 0, 0}, + {&core2_cost, 16, 10, 16, 10, 16}, + {&generic32_cost, 16, 7, 16, 7, 16}, + {&generic64_cost, 16, 10, 16, 10, 16}, + {&amdfam10_cost, 32, 24, 32, 7, 32} +}; + +static const char *const cpu_names[TARGET_CPU_DEFAULT_max] = +{ + "generic", + "i386", + "i486", + "pentium", + "pentium-mmx", + "pentiumpro", + "pentium2", + "pentium3", + "pentium4", + "pentium-m", + "prescott", + "nocona", + "core2", + "geode", + "k6", + "k6-2", + "k6-3", + "athlon", + "athlon-4", + "k8", + "amdfam10" +}; + /* Implement TARGET_HANDLE_OPTION. */ static bool @@ -2002,11 +2105,294 @@ ix86_handle_option (size_t code, const char *arg ATTRIBUTE_UNUSED, int value) } return true; + case OPT_mabm: + if (value) + { + ix86_isa_flags |= OPTION_MASK_ISA_ABM_SET; + ix86_isa_flags_explicit |= OPTION_MASK_ISA_ABM_SET; + } + else + { + ix86_isa_flags &= ~OPTION_MASK_ISA_ABM_UNSET; + ix86_isa_flags_explicit |= OPTION_MASK_ISA_ABM_UNSET; + } + return true; + + case OPT_mpopcnt: + if (value) + { + ix86_isa_flags |= OPTION_MASK_ISA_POPCNT_SET; + ix86_isa_flags_explicit |= OPTION_MASK_ISA_POPCNT_SET; + } + else + { + ix86_isa_flags &= ~OPTION_MASK_ISA_POPCNT_UNSET; + ix86_isa_flags_explicit |= OPTION_MASK_ISA_POPCNT_UNSET; + } + return true; + + case OPT_msahf: + if (value) + { + ix86_isa_flags |= OPTION_MASK_ISA_SAHF_SET; + ix86_isa_flags_explicit |= OPTION_MASK_ISA_SAHF_SET; + } + else + { + ix86_isa_flags &= ~OPTION_MASK_ISA_SAHF_UNSET; + ix86_isa_flags_explicit |= OPTION_MASK_ISA_SAHF_UNSET; + } + return true; + + case OPT_mcx16: + if (value) + { + ix86_isa_flags |= OPTION_MASK_ISA_CX16_SET; + ix86_isa_flags_explicit |= OPTION_MASK_ISA_CX16_SET; + } + else + { + ix86_isa_flags &= ~OPTION_MASK_ISA_CX16_UNSET; + ix86_isa_flags_explicit |= OPTION_MASK_ISA_CX16_UNSET; + } + return true; + + case OPT_maes: + if (value) + { + ix86_isa_flags |= OPTION_MASK_ISA_AES_SET; + ix86_isa_flags_explicit |= OPTION_MASK_ISA_AES_SET; + } + else + { + ix86_isa_flags &= ~OPTION_MASK_ISA_AES_UNSET; + ix86_isa_flags_explicit |= OPTION_MASK_ISA_AES_UNSET; + } + return true; + + case OPT_mpclmul: + if (value) + { + ix86_isa_flags |= OPTION_MASK_ISA_PCLMUL_SET; + ix86_isa_flags_explicit |= OPTION_MASK_ISA_PCLMUL_SET; + } + else + { + ix86_isa_flags &= ~OPTION_MASK_ISA_PCLMUL_UNSET; + ix86_isa_flags_explicit |= OPTION_MASK_ISA_PCLMUL_UNSET; + } + return true; + default: return true; } } + +/* Return a string the documents the current -m options. The caller is + responsible for freeing the string. */ + +static char * +ix86_target_string (int isa, int flags, const char *arch, const char *tune, + const char *fpmath, bool add_nl_p) +{ + struct ix86_target_opts + { + const char *option; /* option string */ + int mask; /* isa mask options */ + }; + + /* This table is ordered so that options like -msse5 or -msse4.2 that imply + preceding options while match those first. */ + static struct ix86_target_opts isa_opts[] = + { + { "-m64", OPTION_MASK_ISA_64BIT }, + { "-msse5", OPTION_MASK_ISA_SSE5 }, + { "-msse4a", OPTION_MASK_ISA_SSE4A }, + { "-msse4.2", OPTION_MASK_ISA_SSE4_2 }, + { "-msse4.1", OPTION_MASK_ISA_SSE4_1 }, + { "-mssse3", OPTION_MASK_ISA_SSSE3 }, + { "-msse3", OPTION_MASK_ISA_SSE3 }, + { "-msse2", OPTION_MASK_ISA_SSE2 }, + { "-msse", OPTION_MASK_ISA_SSE }, + { "-m3dnow", OPTION_MASK_ISA_3DNOW }, + { "-mmmx", OPTION_MASK_ISA_MMX }, + { "-mabm", OPTION_MASK_ISA_ABM }, + { "-mpopcnt", OPTION_MASK_ISA_POPCNT }, + { "-maes", OPTION_MASK_ISA_AES }, + { "-mpclmul", OPTION_MASK_ISA_PCLMUL }, + }; + + /* Flag options. */ + static struct ix86_target_opts flag_opts[] = + { + { "-m128bit-long-double", MASK_128BIT_LONG_DOUBLE }, + { "-m80387", MASK_80387 }, + { "-maccumulate-outgoing-args", MASK_ACCUMULATE_OUTGOING_ARGS }, + { "-malign-double", MASK_ALIGN_DOUBLE }, + { "-mcld", MASK_CLD }, + { "-mfp-ret-in-387", MASK_FLOAT_RETURNS }, + { "-mieee-fp", MASK_IEEE_FP }, + { "-minline-all-stringops", MASK_INLINE_ALL_STRINGOPS }, + { "-minline-stringops-dynamically", MASK_INLINE_STRINGOPS_DYNAMICALLY }, + { "-mms-bitfields", MASK_MS_BITFIELD_LAYOUT }, + { "-mno-align-stringops", MASK_NO_ALIGN_STRINGOPS }, + { "-mno-fancy-math-387", MASK_NO_FANCY_MATH_387 }, + { "-mno-fused-madd", MASK_NO_FUSED_MADD }, + { "-mno-push-args", MASK_NO_PUSH_ARGS }, + { "-mno-red-zone", MASK_NO_RED_ZONE }, + { "-momit-leaf-frame-pointer", MASK_OMIT_LEAF_FRAME_POINTER }, + { "-mrecip", MASK_RECIP }, + { "-mrtd", MASK_RTD }, + { "-msseregparm", MASK_SSEREGPARM }, + { "-mstack-arg-probe", MASK_STACK_PROBE }, + { "-mtls-direct-seg-refs", MASK_TLS_DIRECT_SEG_REFS }, + }; + + const char *opts[ (sizeof (isa_opts) / sizeof (isa_opts[0]) + + sizeof (flag_opts) / sizeof (flag_opts[0]) + + 6)][2]; + + char isa_other[40]; + char target_other[40]; + unsigned num = 0; + unsigned i, j; + char *ret; + char *ptr; + size_t len; + size_t line_len; + size_t sep_len; + + memset (opts, '\0', sizeof (opts)); + + /* Add -march= option. */ + if (arch) + { + opts[num][0] = "-march="; + opts[num++][1] = arch; + } + + /* Add -mtune= option. */ + if (tune) + { + opts[num][0] = "-mtune="; + opts[num++][1] = tune; + } + + /* Pick out the options in isa options. */ + for (i = 0; i < sizeof (isa_opts) / sizeof (isa_opts[0]); i++) + { + if ((isa & isa_opts[i].mask) != 0) + { + opts[num++][0] = isa_opts[i].option; + isa &= ~ isa_opts[i].mask; + } + } + + if (isa && add_nl_p) + { + opts[num++][0] = isa_other; + sprintf (isa_other, "(other isa: 0x%x)", isa); + } + + /* Add flag options. */ + for (i = 0; i < sizeof (flag_opts) / sizeof (flag_opts[0]); i++) + { + if ((flags & flag_opts[i].mask) != 0) + { + opts[num++][0] = flag_opts[i].option; + flags &= ~ flag_opts[i].mask; + } + } + + if (flags && add_nl_p) + { + opts[num++][0] = target_other; + sprintf (target_other, "(other flags: 0x%x)", isa); + } + + /* Add -fpmath= option. */ + if (fpmath) + { + opts[num][0] = "-mfpmath="; + opts[num++][1] = fpmath; + } + + /* Any options? */ + if (num == 0) + return NULL; + + gcc_assert (num < sizeof (opts) / sizeof (opts[0])); + + /* Size the string. */ + len = 0; + sep_len = (add_nl_p) ? 3 : 1; + for (i = 0; i < num; i++) + { + len += sep_len; + for (j = 0; j < 2; j++) + if (opts[i][j]) + len += strlen (opts[i][j]); + } + + /* Build the string. */ + ret = ptr = (char *) xmalloc (len); + line_len = 0; + + for (i = 0; i < num; i++) + { + size_t len2[2]; + + for (j = 0; j < 2; j++) + len2[j] = (opts[i][j]) ? strlen (opts[i][j]) : 0; + + if (i != 0) + { + *ptr++ = ' '; + line_len++; + + if (add_nl_p && line_len + len2[0] + len2[1] > 70) + { + *ptr++ = '\\'; + *ptr++ = '\n'; + line_len = 0; + } + } + + for (j = 0; j < 2; j++) + if (opts[i][j]) + { + memcpy (ptr, opts[i][j], len2[j]); + ptr += len2[j]; + line_len += len2[j]; + } + } + + *ptr = '\0'; + gcc_assert (ret + len >= ptr); + + return ret; +} + +/* Function that is callable from the debugger to print the current + options. */ +void +ix86_debug_options (void) +{ + char *opts = ix86_target_string (ix86_isa_flags, target_flags, + ix86_arch_string, ix86_tune_string, + ix86_fpmath_string, true); + if (opts) + { + fprintf (stderr, "%s\n\n", opts); + free (opts); + } + else + fprintf (stderr, "<no options>\n\n"); + + return; +} + /* Sometimes certain combinations of command options do not make sense on a particular target machine. You can define a macro `OVERRIDE_OPTIONS' to take account of this. This macro, if @@ -2017,68 +2403,17 @@ ix86_handle_option (size_t code, const char *arg ATTRIBUTE_UNUSED, int value) `-O'. That is what `OPTIMIZATION_OPTIONS' is for. */ void -override_options (void) +override_options (bool main_args_p) { int i; - int ix86_tune_defaulted = 0; - int ix86_arch_specified = 0; unsigned int ix86_arch_mask, ix86_tune_mask; + const char *prefix; + const char *suffix; + const char *sw; /* Comes from final.c -- no real reason to change it. */ #define MAX_CODE_ALIGN 16 - static struct ptt - { - const struct processor_costs *cost; /* Processor costs */ - const int align_loop; /* Default alignments. */ - const int align_loop_max_skip; - const int align_jump; - const int align_jump_max_skip; - const int align_func; - } - const processor_target_table[PROCESSOR_max] = - { - {&i386_cost, 4, 3, 4, 3, 4}, - {&i486_cost, 16, 15, 16, 15, 16}, - {&pentium_cost, 16, 7, 16, 7, 16}, - {&pentiumpro_cost, 16, 15, 16, 10, 16}, - {&geode_cost, 0, 0, 0, 0, 0}, - {&k6_cost, 32, 7, 32, 7, 32}, - {&athlon_cost, 16, 7, 16, 7, 16}, - {&pentium4_cost, 0, 0, 0, 0, 0}, - {&k8_cost, 16, 7, 16, 7, 16}, - {&nocona_cost, 0, 0, 0, 0, 0}, - {&core2_cost, 16, 10, 16, 10, 16}, - {&generic32_cost, 16, 7, 16, 7, 16}, - {&generic64_cost, 16, 10, 16, 10, 16}, - {&amdfam10_cost, 32, 24, 32, 7, 32} - }; - - static const char *const cpu_names[TARGET_CPU_DEFAULT_max] = - { - "generic", - "i386", - "i486", - "pentium", - "pentium-mmx", - "pentiumpro", - "pentium2", - "pentium3", - "pentium4", - "pentium-m", - "prescott", - "nocona", - "core2", - "geode", - "k6", - "k6-2", - "k6-3", - "athlon", - "athlon-4", - "k8", - "amdfam10" - }; - enum pta_flags { PTA_SSE = 1 << 0, @@ -2197,6 +2532,21 @@ override_options (void) int const pta_size = ARRAY_SIZE (processor_alias_table); + /* Set up prefix/suffix so the error messages refer to either the command + line argument, or the attribute(option). */ + if (main_args_p) + { + prefix = "-m"; + suffix = ""; + sw = "switch"; + } + else + { + prefix = "option(\""; + suffix = "\")"; + sw = "attribute"; + } + #ifdef SUBTARGET_OVERRIDE_OPTIONS SUBTARGET_OVERRIDE_OPTIONS; #endif @@ -2246,8 +2596,15 @@ override_options (void) else ix86_tune_string = "generic32"; } + /* If this call is for setting the option attribute, allow the + generic32/generic64 that was previously set. */ + else if (!main_args_p + && (!strcmp (ix86_tune_string, "generic32") + || !strcmp (ix86_tune_string, "generic64"))) + ; else if (!strncmp (ix86_tune_string, "generic", 7)) - error ("bad value (%s) for -mtune= switch", ix86_tune_string); + error ("bad value (%s) for %stune=%s %s", + ix86_tune_string, prefix, suffix, sw); } else { @@ -2288,11 +2645,13 @@ override_options (void) else if (!strcmp (ix86_stringop_string, "unrolled_loop")) stringop_alg = unrolled_loop; else - error ("bad value (%s) for -mstringop-strategy= switch", ix86_stringop_string); + error ("bad value (%s) for %sstringop-strategy=%s %s", + ix86_stringop_string, prefix, suffix, sw); } if (!strcmp (ix86_tune_string, "x86-64")) - warning (OPT_Wdeprecated, "-mtune=x86-64 is deprecated. Use -mtune=k8 or " - "-mtune=generic instead as appropriate."); + warning (OPT_Wdeprecated, "%stune=x86-64%s is deprecated. Use " + "%stune=k8%s or %stune=generic%s instead as appropriate.", + prefix, suffix, prefix, suffix, prefix, suffix); if (!ix86_arch_string) ix86_arch_string = TARGET_64BIT ? "x86-64" : "i386"; @@ -2300,9 +2659,11 @@ override_options (void) ix86_arch_specified = 1; if (!strcmp (ix86_arch_string, "generic")) - error ("generic CPU can be used only for -mtune= switch"); + error ("generic CPU can be used only for %stune=%s %s", + prefix, suffix, sw); if (!strncmp (ix86_arch_string, "generic", 7)) - error ("bad value (%s) for -march= switch", ix86_arch_string); + error ("bad value (%s) for %sarch=%s %s", + ix86_arch_string, prefix, suffix, sw); if (ix86_cmodel_string != 0) { @@ -2319,7 +2680,8 @@ override_options (void) else if (!strcmp (ix86_cmodel_string, "kernel") && !flag_pic) ix86_cmodel = CM_KERNEL; else - error ("bad value (%s) for -mcmodel= switch", ix86_cmodel_string); + error ("bad value (%s) for %scmodel=%s %s", + ix86_cmodel_string, prefix, suffix, sw); } else { @@ -2342,7 +2704,8 @@ override_options (void) else if (!strcmp (ix86_asm_string, "att")) ix86_asm_dialect = ASM_ATT; else - error ("bad value (%s) for -masm= switch", ix86_asm_string); + error ("bad value (%s) for %sasm=%s %s", + ix86_asm_string, prefix, suffix, sw); } if ((TARGET_64BIT == 0) != (ix86_cmodel == CM_32)) error ("code model %qs not supported in the %s bit mode", @@ -2395,31 +2758,37 @@ override_options (void) if (processor_alias_table[i].flags & PTA_SSE5 && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_SSE5)) ix86_isa_flags |= OPTION_MASK_ISA_SSE5; - - if (processor_alias_table[i].flags & PTA_ABM) - x86_abm = true; - if (processor_alias_table[i].flags & PTA_CX16) - x86_cmpxchg16b = true; - if (processor_alias_table[i].flags & (PTA_POPCNT | PTA_ABM)) - x86_popcnt = true; + if (processor_alias_table[i].flags & PTA_ABM + && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_ABM)) + ix86_isa_flags |= OPTION_MASK_ISA_ABM; + if (processor_alias_table[i].flags & PTA_CX16 + && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_CX16)) + ix86_isa_flags |= OPTION_MASK_ISA_CX16; + if (processor_alias_table[i].flags & (PTA_POPCNT | PTA_ABM) + && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_POPCNT)) + ix86_isa_flags |= OPTION_MASK_ISA_POPCNT; + if (!(TARGET_64BIT && (processor_alias_table[i].flags & PTA_NO_SAHF)) + && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_SAHF)) + ix86_isa_flags |= OPTION_MASK_ISA_SAHF; + if (processor_alias_table[i].flags & PTA_AES + && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_AES)) + ix86_isa_flags |= OPTION_MASK_ISA_AES; + if (processor_alias_table[i].flags & PTA_PCLMUL + && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_PCLMUL)) + ix86_isa_flags |= OPTION_MASK_ISA_PCLMUL; if (processor_alias_table[i].flags & (PTA_PREFETCH_SSE | PTA_SSE)) x86_prefetch_sse = true; - if (!(TARGET_64BIT && (processor_alias_table[i].flags & PTA_NO_SAHF))) - x86_sahf = true; - if (processor_alias_table[i].flags & PTA_AES) - x86_aes = true; - if (processor_alias_table[i].flags & PTA_PCLMUL) - x86_pclmul = true; break; } if (i == pta_size) - error ("bad value (%s) for -march= switch", ix86_arch_string); + error ("bad value (%s) for %sarch=%s %s", + ix86_arch_string, prefix, suffix, sw); ix86_arch_mask = 1u << ix86_arch; for (i = 0; i < X86_ARCH_LAST; ++i) - ix86_arch_features[i] &= ix86_arch_mask; + ix86_arch_features[i] = !!(initial_ix86_arch_features[i] & ix86_arch_mask); for (i = 0; i < pta_size; i++) if (! strcmp (ix86_tune_string, processor_alias_table[i].name)) @@ -2451,19 +2820,12 @@ override_options (void) break; } if (i == pta_size) - error ("bad value (%s) for -mtune= switch", ix86_tune_string); - - /* Enable SSE2 if AES or PCLMUL is enabled. */ - if ((x86_aes || x86_pclmul) - && !(ix86_isa_flags_explicit & OPTION_MASK_ISA_SSE2)) - { - ix86_isa_flags |= OPTION_MASK_ISA_SSE2_SET; - ix86_isa_flags_explicit |= OPTION_MASK_ISA_SSE2_SET; - } + error ("bad value (%s) for %stune=%s %s", + ix86_tune_string, prefix, suffix, sw); ix86_tune_mask = 1u << ix86_tune; for (i = 0; i < X86_TUNE_LAST; ++i) - ix86_tune_features[i] &= ix86_tune_mask; + ix86_tune_features[i] = !!(initial_ix86_tune_features[i] & ix86_tune_mask); if (optimize_size) ix86_cost = &size_cost; @@ -2477,10 +2839,11 @@ override_options (void) if (ix86_regparm_string) { if (TARGET_64BIT) - warning (0, "-mregparm is ignored in 64-bit mode"); + warning (0, "%sregparm%s is ignored in 64-bit mode", prefix, suffix); i = atoi (ix86_regparm_string); if (i < 0 || i > REGPARM_MAX) - error ("-mregparm=%d is not between 0 and %d", i, REGPARM_MAX); + error ("%sregparm=%d%s is not between 0 and %d", + prefix, i, suffix, REGPARM_MAX); else ix86_regparm = i; } @@ -2492,12 +2855,14 @@ override_options (void) Remove this code in GCC 3.2 or later. */ if (ix86_align_loops_string) { - warning (0, "-malign-loops is obsolete, use -falign-loops"); + warning (0, "%salign-loops%s is obsolete, use %salign-loops%s", + prefix, suffix, prefix, suffix); if (align_loops == 0) { i = atoi (ix86_align_loops_string); if (i < 0 || i > MAX_CODE_ALIGN) - error ("-malign-loops=%d is not between 0 and %d", i, MAX_CODE_ALIGN); + error ("%salign-loops=%d%s is not between 0 and %d", + prefix, i, suffix, MAX_CODE_ALIGN); else align_loops = 1 << i; } @@ -2505,12 +2870,14 @@ override_options (void) if (ix86_align_jumps_string) { - warning (0, "-malign-jumps is obsolete, use -falign-jumps"); + warning (0, "%salign-jumps%s is obsolete, use %salign-jumps%s", + prefix, suffix, prefix, suffix); if (align_jumps == 0) { i = atoi (ix86_align_jumps_string); if (i < 0 || i > MAX_CODE_ALIGN) - error ("-malign-loops=%d is not between 0 and %d", i, MAX_CODE_ALIGN); + error ("%salign-loops=%d%s is not between 0 and %d", + prefix, i, suffix, MAX_CODE_ALIGN); else align_jumps = 1 << i; } @@ -2518,12 +2885,14 @@ override_options (void) if (ix86_align_funcs_string) { - warning (0, "-malign-functions is obsolete, use -falign-functions"); + warning (0, "%salign-functions%s is obsolete, use %salign-functions%s", + prefix, suffix, prefix, suffix); if (align_functions == 0) { i = atoi (ix86_align_funcs_string); if (i < 0 || i > MAX_CODE_ALIGN) - error ("-malign-loops=%d is not between 0 and %d", i, MAX_CODE_ALIGN); + error ("%salign-loops=%d%s is not between 0 and %d", + prefix, i, suffix, MAX_CODE_ALIGN); else align_functions = 1 << i; } @@ -2551,7 +2920,7 @@ override_options (void) { i = atoi (ix86_branch_cost_string); if (i < 0 || i > 5) - error ("-mbranch-cost=%d is not between 0 and 5", i); + error ("%sbranch-cost=%d%s is not between 0 and 5", prefix, i, suffix); else ix86_branch_cost = i; } @@ -2559,7 +2928,7 @@ override_options (void) { i = atoi (ix86_section_threshold_string); if (i < 0) - error ("-mlarge-data-threshold=%d is negative", i); + error ("%slarge-data-threshold=%d%s is negative", prefix, i, suffix); else ix86_section_threshold = i; } @@ -2573,8 +2942,8 @@ override_options (void) else if (strcmp (ix86_tls_dialect_string, "sun") == 0) ix86_tls_dialect = TLS_DIALECT_SUN; else - error ("bad value (%s) for -mtls-dialect= switch", - ix86_tls_dialect_string); + error ("bad value (%s) for %stls-dialect=%s %s", + ix86_tls_dialect_string, prefix, suffix, sw); } if (ix87_precision_string) @@ -2597,7 +2966,7 @@ override_options (void) | TARGET_SUBTARGET64_ISA_DEFAULT) & ~ix86_isa_flags_explicit); if (TARGET_RTD) - warning (0, "-mrtd is ignored in 64bit mode"); + warning (0, "%srtd%s is ignored in 64bit mode", prefix, suffix); } else { @@ -2643,7 +3012,7 @@ override_options (void) /* Turn on popcnt instruction for -msse4.2 or -mabm. */ if (TARGET_SSE4_2 || TARGET_ABM) - x86_popcnt = true; + ix86_isa_flags |= OPTION_MASK_ISA_POPCNT & ~ix86_isa_flags_explicit; /* Validate -mpreferred-stack-boundary= value, or provide default. The default of 128 bits is for Pentium III's SSE __m128. We can't @@ -2654,8 +3023,8 @@ override_options (void) { i = atoi (ix86_preferred_stack_boundary_string); if (i < (TARGET_64BIT ? 4 : 2) || i > 12) - error ("-mpreferred-stack-boundary=%d is not between %d and 12", i, - TARGET_64BIT ? 4 : 2); + error ("%spreferred-stack-boundary=%d%s is not between %d and 12", + prefix, i, suffix, TARGET_64BIT ? 4 : 2); else ix86_preferred_stack_boundary = (1 << i) * BITS_PER_UNIT; } @@ -2663,7 +3032,7 @@ override_options (void) /* Accept -msseregparm only if at least SSE support is enabled. */ if (TARGET_SSEREGPARM && ! TARGET_SSE) - error ("-msseregparm used without SSE enabled"); + error ("%ssseregparm%s used without SSE enabled", prefix, suffix); ix86_fpmath = TARGET_FPMATH_DEFAULT; if (ix86_fpmath_string != 0) @@ -2681,7 +3050,10 @@ override_options (void) ix86_fpmath = FPMATH_SSE; } else if (! strcmp (ix86_fpmath_string, "387,sse") - || ! strcmp (ix86_fpmath_string, "sse,387")) + || ! strcmp (ix86_fpmath_string, "387+sse") + || ! strcmp (ix86_fpmath_string, "sse,387") + || ! strcmp (ix86_fpmath_string, "sse+387") + || ! strcmp (ix86_fpmath_string, "both")) { if (!TARGET_SSE) { @@ -2697,7 +3069,8 @@ override_options (void) ix86_fpmath = (enum fpmath_unit) (FPMATH_SSE | FPMATH_387); } else - error ("bad value (%s) for -mfpmath= switch", ix86_fpmath_string); + error ("bad value (%s) for %sfpmath=%s %s", + ix86_fpmath_string, prefix, suffix, sw); } /* If the i387 is disabled, then do not return values in it. */ @@ -2713,7 +3086,8 @@ override_options (void) ix86_veclib_handler = ix86_veclibabi_acml; else error ("unknown vectorization library ABI type (%s) for " - "-mveclibabi= switch", ix86_veclibabi_string); + "%sveclibabi=%s %s", ix86_veclibabi_string, + prefix, suffix, sw); } if ((x86_accumulate_outgoing_args & ix86_tune_mask) @@ -2732,7 +3106,8 @@ override_options (void) { if (target_flags_explicit & MASK_ACCUMULATE_OUTGOING_ARGS) warning (0, "unwind tables currently require either a frame pointer " - "or -maccumulate-outgoing-args for correctness"); + "or %saccumulate-outgoing-args%s for correctness", + prefix, suffix); target_flags |= MASK_ACCUMULATE_OUTGOING_ARGS; } @@ -2743,11 +3118,13 @@ override_options (void) && !(target_flags & MASK_ACCUMULATE_OUTGOING_ARGS)) { if (target_flags_explicit & MASK_ACCUMULATE_OUTGOING_ARGS) - warning (0, "stack probing requires -maccumulate-outgoing-args " - "for correctness"); + warning (0, "stack probing requires %saccumulate-outgoing-args%s " + "for correctness", prefix, suffix); target_flags |= MASK_ACCUMULATE_OUTGOING_ARGS; } + TARGET_CMOVE = 0; + /* For sane SSE instruction set generation we need fcomi instruction. It is safe to enable all CMOVE instructions. */ if (TARGET_SSE) @@ -2808,8 +3185,500 @@ override_options (void) if (!TARGET_64BIT) target_flags |= MASK_CLD & ~target_flags_explicit; #endif + + /* Save the initial options in case the user does function specific options */ + if (main_args_p) + target_option_default_node = target_option_current_node + = build_target_option_node (); } +/* Save the current options */ + +static void +ix86_function_specific_save (struct cl_target_option *ptr) +{ + gcc_assert (IN_RANGE (ix86_arch, 0, 255)); + gcc_assert (IN_RANGE (ix86_tune, 0, 255)); + gcc_assert (IN_RANGE (ix86_fpmath, 0, 255)); + gcc_assert (IN_RANGE (ix86_branch_cost, 0, 255)); + + ptr->arch = ix86_arch; + ptr->tune = ix86_tune; + ptr->fpmath = ix86_fpmath; + ptr->branch_cost = ix86_branch_cost; + ptr->tune_defaulted = ix86_tune_defaulted; + ptr->arch_specified = ix86_arch_specified; + ptr->ix86_isa_flags_explicit = ix86_isa_flags_explicit; + ptr->target_flags_explicit = target_flags_explicit; +} + +/* Restore the current options */ + +static void +ix86_function_specific_restore (struct cl_target_option *ptr) +{ + enum processor_type old_tune = ix86_tune; + enum processor_type old_arch = ix86_arch; + unsigned int ix86_arch_mask, ix86_tune_mask; + int i; + + ix86_arch = ptr->arch; + ix86_tune = ptr->tune; + ix86_fpmath = ptr->fpmath; + ix86_branch_cost = ptr->branch_cost; + ix86_tune_defaulted = ptr->tune_defaulted; + ix86_arch_specified = ptr->arch_specified; + ix86_isa_flags_explicit = ptr->ix86_isa_flags_explicit; + target_flags_explicit = ptr->target_flags_explicit; + + /* Recreate the arch feature tests if the arch changed */ + if (old_arch != ix86_arch) + { + ix86_arch_mask = 1u << ix86_arch; + for (i = 0; i < X86_ARCH_LAST; ++i) + ix86_arch_features[i] + = !!(initial_ix86_arch_features[i] & ix86_arch_mask); + } + + /* Recreate the tune optimization tests */ + if (old_tune != ix86_tune) + { + ix86_tune_mask = 1u << ix86_tune; + for (i = 0; i < X86_TUNE_LAST; ++i) + ix86_tune_features[i] + = !!(initial_ix86_tune_features[i] & ix86_tune_mask); + } +} + +/* Print the current options */ + +static void +ix86_function_specific_print (FILE *file, int indent, + struct cl_target_option *ptr) +{ + char *target_string + = ix86_target_string (ptr->ix86_isa_flags, ptr->target_flags, + NULL, NULL, NULL, false); + + fprintf (file, "%*sarch = %d (%s)\n", + indent, "", + ptr->arch, + ((ptr->arch < TARGET_CPU_DEFAULT_max) + ? cpu_names[ptr->arch] + : "<unknown>")); + + fprintf (file, "%*stune = %d (%s)\n", + indent, "", + ptr->tune, + ((ptr->tune < TARGET_CPU_DEFAULT_max) + ? cpu_names[ptr->tune] + : "<unknown>")); + + fprintf (file, "%*sfpmath = %d%s%s\n", indent, "", ptr->fpmath, + (ptr->fpmath & FPMATH_387) ? ", 387" : "", + (ptr->fpmath & FPMATH_SSE) ? ", sse" : ""); + fprintf (file, "%*sbranch_cost = %d\n", indent, "", ptr->branch_cost); + + if (target_string) + { + fprintf (file, "%*s%s\n", indent, "", target_string); + free (target_string); + } +} + + +/* Inner function to process the attribute((option(...))), take an argument and + set the current options from the argument. If we have a list, recursively go + over the list. */ + +static bool +ix86_valid_option_attribute_inner_p (tree args, char *p_strings[]) +{ + char *next_optstr; + bool ret = true; + +#define IX86_ATTR_ISA(S,O) { S, sizeof (S)-1, ix86_opt_isa, O, 0 } +#define IX86_ATTR_STR(S,O) { S, sizeof (S)-1, ix86_opt_str, O, 0 } +#define IX86_ATTR_YES(S,O,M) { S, sizeof (S)-1, ix86_opt_yes, O, M } +#define IX86_ATTR_NO(S,O,M) { S, sizeof (S)-1, ix86_opt_no, O, M } + + enum ix86_opt_type + { + ix86_opt_unknown, + ix86_opt_yes, + ix86_opt_no, + ix86_opt_str, + ix86_opt_isa + }; + + static const struct + { + const char *string; + size_t len; + enum ix86_opt_type type; + int opt; + int mask; + } attrs[] = { + /* isa options */ + IX86_ATTR_ISA ("3dnow", OPT_m3dnow), + IX86_ATTR_ISA ("abm", OPT_mabm), + IX86_ATTR_ISA ("aes", OPT_maes), + IX86_ATTR_ISA ("mmx", OPT_mmmx), + IX86_ATTR_ISA ("pclmul", OPT_mpclmul), + IX86_ATTR_ISA ("popcnt", OPT_mpopcnt), + IX86_ATTR_ISA ("sse", OPT_msse), + IX86_ATTR_ISA ("sse2", OPT_msse2), + IX86_ATTR_ISA ("sse3", OPT_msse3), + IX86_ATTR_ISA ("sse4", OPT_msse4), + IX86_ATTR_ISA ("sse4.1", OPT_msse4_1), + IX86_ATTR_ISA ("sse4.2", OPT_msse4_2), + IX86_ATTR_ISA ("sse4a", OPT_msse4a), + IX86_ATTR_ISA ("sse5", OPT_msse5), + IX86_ATTR_ISA ("ssse3", OPT_mssse3), + + /* string options */ + IX86_ATTR_STR ("arch=", IX86_FUNCTION_SPECIFIC_ARCH), + IX86_ATTR_STR ("fpmath=", IX86_FUNCTION_SPECIFIC_FPMATH), + IX86_ATTR_STR ("tune=", IX86_FUNCTION_SPECIFIC_TUNE), + + /* flag options */ + IX86_ATTR_YES ("cld", + OPT_mcld, + MASK_CLD), + + IX86_ATTR_NO ("fancy-math-387", + OPT_mfancy_math_387, + MASK_NO_FANCY_MATH_387), + + IX86_ATTR_NO ("fused-madd", + OPT_mfused_madd, + MASK_NO_FUSED_MADD), + + IX86_ATTR_YES ("ieee-fp", + OPT_mieee_fp, + MASK_IEEE_FP), + + IX86_ATTR_YES ("inline-all-stringops", + OPT_minline_all_stringops, + MASK_INLINE_ALL_STRINGOPS), + + IX86_ATTR_YES ("inline-stringops-dynamically", + OPT_minline_stringops_dynamically, + MASK_INLINE_STRINGOPS_DYNAMICALLY), + + IX86_ATTR_NO ("align-stringops", + OPT_mno_align_stringops, + MASK_NO_ALIGN_STRINGOPS), + + IX86_ATTR_YES ("recip", + OPT_mrecip, + MASK_RECIP), + + }; + + /* If this is a list, recurse to get the options. */ + if (TREE_CODE (args) == TREE_LIST) + { + bool ret = true; + + for (; args; args = TREE_CHAIN (args)) + if (TREE_VALUE (args) + && !ix86_valid_option_attribute_inner_p (TREE_VALUE (args), p_strings)) + ret = false; + + return ret; + } + + else if (TREE_CODE (args) != STRING_CST) + gcc_unreachable (); + + /* Handle multiple arguments separated by commas. */ + next_optstr = ASTRDUP (TREE_STRING_POINTER (args)); + + while (next_optstr && *next_optstr != '\0') + { + char *p = next_optstr; + char *orig_p = p; + char *comma = strchr (next_optstr, ','); + const char *opt_string; + size_t len, opt_len; + int opt; + bool opt_set_p; + char ch; + unsigned i; + enum ix86_opt_type type = ix86_opt_unknown; + int mask = 0; + + if (comma) + { + *comma = '\0'; + len = comma - next_optstr; + next_optstr = comma + 1; + } + else + { + len = strlen (p); + next_optstr = NULL; + } + + /* Recognize no-xxx. */ + if (len > 3 && p[0] == 'n' && p[1] == 'o' && p[2] == '-') + { + opt_set_p = false; + p += 3; + len -= 3; + } + else + opt_set_p = true; + + /* Find the option. */ + ch = *p; + opt = N_OPTS; + for (i = 0; i < sizeof (attrs) / sizeof (attrs[0]); i++) + { + type = attrs[i].type; + opt_len = attrs[i].len; + if (ch == attrs[i].string[0] + && ((type != ix86_opt_str) ? len == opt_len : len > opt_len) + && memcmp (p, attrs[i].string, opt_len) == 0) + { + opt = attrs[i].opt; + mask = attrs[i].mask; + opt_string = attrs[i].string; + break; + } + } + + /* Process the option. */ + if (opt == N_OPTS) + { + error ("attribute(option(\"%s\")) is unknown", orig_p); + ret = false; + } + + else if (type == ix86_opt_isa) + ix86_handle_option (opt, p, opt_set_p); + + else if (type == ix86_opt_yes || type == ix86_opt_no) + { + if (type == ix86_opt_no) + opt_set_p = !opt_set_p; + + if (opt_set_p) + target_flags |= mask; + else + target_flags &= ~mask; + } + + else if (type == ix86_opt_str) + { + if (p_strings[opt]) + { + error ("option(\"%s\") was already specified", opt_string); + ret = false; + } + else + p_strings[opt] = xstrdup (p + opt_len); + } + + else + gcc_unreachable (); + } + + return ret; +} + +/* Return a TARGET_OPTION_NODE tree of the target options listed or NULL. */ + +tree +ix86_valid_option_attribute_tree (tree args) +{ + const char *orig_arch_string = ix86_arch_string; + const char *orig_tune_string = ix86_tune_string; + const char *orig_fpmath_string = ix86_fpmath_string; + int orig_tune_defaulted = ix86_tune_defaulted; + int orig_arch_specified = ix86_arch_specified; + char *option_strings[IX86_FUNCTION_SPECIFIC_MAX] = { NULL, NULL, NULL }; + tree t = NULL_TREE; + int i; + struct cl_target_option *def + = TREE_TARGET_OPTION (target_option_default_node); + + /* Process each of the options on the chain. */ + if (! ix86_valid_option_attribute_inner_p (args, option_strings)) + return NULL_TREE; + + /* If the changed options are different from the default, rerun override_options, + and then save the options away. The string options are are attribute options, + and will be undone when we copy the save structure. */ + if (ix86_isa_flags != def->ix86_isa_flags + || target_flags != def->target_flags + || option_strings[IX86_FUNCTION_SPECIFIC_ARCH] + || option_strings[IX86_FUNCTION_SPECIFIC_TUNE] + || option_strings[IX86_FUNCTION_SPECIFIC_FPMATH]) + { + /* If we are using the default tune= or arch=, undo the string assigned, + and use the default. */ + if (option_strings[IX86_FUNCTION_SPECIFIC_ARCH]) + ix86_arch_string = option_strings[IX86_FUNCTION_SPECIFIC_ARCH]; + else if (!orig_arch_specified) + ix86_arch_string = NULL; + + if (option_strings[IX86_FUNCTION_SPECIFIC_TUNE]) + ix86_tune_string = option_strings[IX86_FUNCTION_SPECIFIC_TUNE]; + else if (orig_tune_defaulted) + ix86_tune_string = NULL; + + /* If fpmath= is not set, and we now have sse2 on 32-bit, use it. */ + if (option_strings[IX86_FUNCTION_SPECIFIC_FPMATH]) + ix86_fpmath_string = option_strings[IX86_FUNCTION_SPECIFIC_FPMATH]; + else if (!TARGET_64BIT && TARGET_SSE) + ix86_fpmath_string = "sse,387"; + + /* Do any overrides, such as arch=xxx, or tune=xxx support. */ + override_options (false); + + /* Save the current options unless we are validating options for + #pragma. */ + t = build_target_option_node (); + + ix86_arch_string = orig_arch_string; + ix86_tune_string = orig_tune_string; + ix86_fpmath_string = orig_fpmath_string; + + /* Free up memory allocated to hold the strings */ + for (i = 0; i < IX86_FUNCTION_SPECIFIC_MAX; i++) + if (option_strings[i]) + free (option_strings[i]); + } + + return t; +} + +/* Hook to validate attribute((option("string"))). */ + +static bool +ix86_valid_option_attribute_p (tree fndecl, + tree ARG_UNUSED (name), + tree args, + int ARG_UNUSED (flags)) +{ + struct cl_target_option cur_opts; + bool ret = true; + tree new_opts; + + cl_target_option_save (&cur_opts); + new_opts = ix86_valid_option_attribute_tree (args); + if (!new_opts) + ret = false; + + else if (fndecl) + DECL_FUNCTION_SPECIFIC_TARGET (fndecl) = new_opts; + + cl_target_option_restore (&cur_opts); + return ret; +} + + +/* Hook to determine if one function can safely inline another. */ + +static bool +ix86_can_inline_p (tree caller, tree callee) +{ + bool ret = false; + tree caller_tree = DECL_FUNCTION_SPECIFIC_TARGET (caller); + tree callee_tree = DECL_FUNCTION_SPECIFIC_TARGET (callee); + + /* If callee has no option attributes, then it is ok to inline. */ + if (!callee_tree) + ret = true; + + /* If caller has no option attributes, but callee does then it is not ok to + inline. */ + else if (!caller_tree) + ret = false; + + else + { + struct cl_target_option *caller_opts = TREE_TARGET_OPTION (caller_tree); + struct cl_target_option *callee_opts = TREE_TARGET_OPTION (callee_tree); + + /* Callee's isa options should a subset of the caller's, i.e. a SSE5 function + can inline a SSE2 function but a SSE2 function can't inline a SSE5 + function. */ + if ((caller_opts->ix86_isa_flags & callee_opts->ix86_isa_flags) + != callee_opts->ix86_isa_flags) + ret = false; + + /* See if we have the same non-isa options. */ + else if (caller_opts->target_flags != callee_opts->target_flags) + ret = false; + + /* See if arch, tune, etc. are the same. */ + else if (caller_opts->arch != callee_opts->arch) + ret = false; + + else if (caller_opts->tune != callee_opts->tune) + ret = false; + + else if (caller_opts->fpmath != callee_opts->fpmath) + ret = false; + + else if (caller_opts->branch_cost != callee_opts->branch_cost) + ret = false; + + else + ret = true; + } + + return ret; +} + + +/* Remember the last target of ix86_set_current_function. */ +static GTY(()) tree ix86_previous_fndecl; + +/* Establish appropriate back-end context for processing the function + FNDECL. The argument might be NULL to indicate processing at top + level, outside of any function scope. */ +static void +ix86_set_current_function (tree fndecl) +{ + /* Only change the context if the function changes. This hook is called + several times in the course of compiling a function, and we don't want to + slow things down too much or call target_reinit when it isn't safe. */ + if (fndecl && fndecl != ix86_previous_fndecl) + { + tree old_tree = (ix86_previous_fndecl + ? DECL_FUNCTION_SPECIFIC_TARGET (ix86_previous_fndecl) + : NULL_TREE); + + tree new_tree = (fndecl + ? DECL_FUNCTION_SPECIFIC_TARGET (fndecl) + : NULL_TREE); + + ix86_previous_fndecl = fndecl; + if (old_tree == new_tree) + ; + + else if (new_tree) + { + cl_target_option_restore (TREE_TARGET_OPTION (new_tree)); + target_reinit (); + } + + else if (old_tree) + { + struct cl_target_option *def + = TREE_TARGET_OPTION (target_option_current_node); + + cl_target_option_restore (def); + target_reinit (); + } + } +} + + /* Return true if this goes in large data/bss. */ static bool @@ -18129,22 +18998,29 @@ enum ix86_builtins /* Table for the ix86 builtin decls. */ static GTY(()) tree ix86_builtins[(int) IX86_BUILTIN_MAX]; -/* Add an ix86 target builtin function with CODE, NAME and TYPE. Do so, - * if the target_flags include one of MASK. Stores the function decl - * in the ix86_builtins array. - * Returns the function decl or NULL_TREE, if the builtin was not added. */ +/* Table to record which ISA options the builtin needs. */ +static int ix86_builtins_isa[(int) IX86_BUILTIN_MAX]; + +/* Add an ix86 target builtin function with CODE, NAME and TYPE. Save the MASK + * of which isa_flags to use in the ix86_builtins_isa array. Stores the + * function decl in the ix86_builtins array. Returns the function decl or + * NULL_TREE, if the builtin was not added. + * + * Record all builtins, even if it isn't an instruction set in the current ISA + * in case the user uses function specific options for a different ISA. When + * the builtin is expanded, check at that time whether it is valid. */ static inline tree def_builtin (int mask, const char *name, tree type, enum ix86_builtins code) { tree decl = NULL_TREE; - if (mask & ix86_isa_flags - && (!(mask & OPTION_MASK_ISA_64BIT) || TARGET_64BIT)) + if (!(mask & OPTION_MASK_ISA_64BIT) || TARGET_64BIT) { decl = add_builtin_function (name, type, code, BUILT_IN_MD, NULL, NULL_TREE); ix86_builtins[(int) code] = decl; + ix86_builtins_isa[(int) code] = mask; } return decl; @@ -19187,9 +20063,10 @@ static const struct builtin_description bdesc_multi_arg[] = { OPTION_MASK_ISA_SSE5, CODE_FOR_sse5_pcom_tfv2di3, "__builtin_ia32_pcomtrueuq", IX86_BUILTIN_PCOMTRUEUQ, PCOM_TRUE, (int)MULTI_ARG_2_DI_TF }, }; -/* Set up all the MMX/SSE builtins. This is not called if TARGET_MMX - is zero. Otherwise, if TARGET_SSE is not set, only expand the MMX - builtins. */ +/* Set up all the MMX/SSE builtins, even builtins for instructions that are not + in the current target ISA to allow the user to compile particular modules + with different target specific options that differ from the command line + options. */ static void ix86_init_mmx_sse_builtins (void) { @@ -20128,23 +21005,15 @@ ix86_init_mmx_sse_builtins (void) def_builtin (OPTION_MASK_ISA_SSE3, "__builtin_ia32_mwait", void_ftype_unsigned_unsigned, IX86_BUILTIN_MWAIT); /* AES */ - if (TARGET_AES) - { - /* Define AES built-in functions only if AES is enabled. */ - def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_aesenc128", v2di_ftype_v2di_v2di, IX86_BUILTIN_AESENC128); - def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_aesenclast128", v2di_ftype_v2di_v2di, IX86_BUILTIN_AESENCLAST128); - def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_aesdec128", v2di_ftype_v2di_v2di, IX86_BUILTIN_AESDEC128); - def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_aesdeclast128", v2di_ftype_v2di_v2di, IX86_BUILTIN_AESDECLAST128); - def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_aesimc128", v2di_ftype_v2di, IX86_BUILTIN_AESIMC128); - def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_aeskeygenassist128", v2di_ftype_v2di_int, IX86_BUILTIN_AESKEYGENASSIST128); - } + def_builtin_const (OPTION_MASK_ISA_AES, "__builtin_ia32_aesenc128", v2di_ftype_v2di_v2di, IX86_BUILTIN_AESENC128); + def_builtin_const (OPTION_MASK_ISA_AES, "__builtin_ia32_aesenclast128", v2di_ftype_v2di_v2di, IX86_BUILTIN_AESENCLAST128); + def_builtin_const (OPTION_MASK_ISA_AES, "__builtin_ia32_aesdec128", v2di_ftype_v2di_v2di, IX86_BUILTIN_AESDEC128); + def_builtin_const (OPTION_MASK_ISA_AES, "__builtin_ia32_aesdeclast128", v2di_ftype_v2di_v2di, IX86_BUILTIN_AESDECLAST128); + def_builtin_const (OPTION_MASK_ISA_AES, "__builtin_ia32_aesimc128", v2di_ftype_v2di, IX86_BUILTIN_AESIMC128); + def_builtin_const (OPTION_MASK_ISA_AES, "__builtin_ia32_aeskeygenassist128", v2di_ftype_v2di_int, IX86_BUILTIN_AESKEYGENASSIST128); /* PCLMUL */ - if (TARGET_PCLMUL) - { - /* Define PCLMUL built-in function only if PCLMUL is enabled. */ - def_builtin_const (OPTION_MASK_ISA_SSE2, "__builtin_ia32_pclmulqdq128", v2di_ftype_v2di_v2di_int, IX86_BUILTIN_PCLMULQDQ128); - } + def_builtin_const (OPTION_MASK_ISA_PCLMUL, "__builtin_ia32_pclmulqdq128", v2di_ftype_v2di_v2di_int, IX86_BUILTIN_PCLMULQDQ128); /* Access to the vec_init patterns. */ ftype = build_function_type_list (V2SI_type_node, integer_type_node, @@ -20399,8 +21268,7 @@ ix86_init_builtins (void) ix86_builtins[(int) IX86_BUILTIN_COPYSIGNQ] = decl; TREE_READONLY (decl) = 1; - if (TARGET_MMX) - ix86_init_mmx_sse_builtins (); + ix86_init_mmx_sse_builtins (); if (TARGET_64BIT) ix86_init_builtins_va_builtins_abi (); } @@ -21616,6 +22484,28 @@ ix86_expand_builtin (tree exp, rtx target, rtx subtarget ATTRIBUTE_UNUSED, enum machine_mode mode0, mode1, mode2; unsigned int fcode = DECL_FUNCTION_CODE (fndecl); + /* Determine whether the builtin function is available under the current ISA. + Originally the builtin was not created if it wasn't applicable to the + current ISA based on the command line switches. With function specific + options, we need to check in the context of the function making the call + whether it is supported. */ + if (ix86_builtins_isa[fcode] + && !(ix86_builtins_isa[fcode] & ix86_isa_flags)) + { + char *opts = ix86_target_string (ix86_builtins_isa[fcode], 0, NULL, + NULL, NULL, false); + + if (!opts) + error ("%qE needs unknown isa option", fndecl); + else + { + gcc_assert (opts != NULL); + error ("%qE needs isa option %s", fndecl, opts); + free (opts); + } + return const0_rtx; + } + switch (fcode) { case IX86_BUILTIN_MASKMOVQ: @@ -26436,6 +27326,24 @@ ix86_enum_va_list (int idx, const char **pname, tree *ptree) #undef TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST #define TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST x86_builtin_vectorization_cost +#undef TARGET_SET_CURRENT_FUNCTION +#define TARGET_SET_CURRENT_FUNCTION ix86_set_current_function + +#undef TARGET_OPTION_VALID_ATTRIBUTE_P +#define TARGET_OPTION_VALID_ATTRIBUTE_P ix86_valid_option_attribute_p + +#undef TARGET_OPTION_SAVE +#define TARGET_OPTION_SAVE ix86_function_specific_save + +#undef TARGET_OPTION_RESTORE +#define TARGET_OPTION_RESTORE ix86_function_specific_restore + +#undef TARGET_OPTION_PRINT +#define TARGET_OPTION_PRINT ix86_function_specific_print + +#undef TARGET_OPTION_CAN_INLINE_P +#define TARGET_OPTION_CAN_INLINE_P ix86_can_inline_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-i386.h" diff --git a/gcc/config/i386/i386.h b/gcc/config/i386/i386.h index d17e414..c7d33c7 100644 --- a/gcc/config/i386/i386.h +++ b/gcc/config/i386/i386.h @@ -49,6 +49,13 @@ along with GCC; see the file COPYING3. If not see #define TARGET_SSE4A OPTION_ISA_SSE4A #define TARGET_SSE5 OPTION_ISA_SSE5 #define TARGET_ROUND OPTION_ISA_ROUND +#define TARGET_ABM OPTION_ISA_ABM +#define TARGET_POPCNT OPTION_ISA_POPCNT +#define TARGET_SAHF OPTION_ISA_SAHF +#define TARGET_AES OPTION_ISA_AES +#define TARGET_PCLMUL OPTION_ISA_PCLMUL +#define TARGET_CMPXCHG16B OPTION_ISA_CX16 + /* SSE5 and SSE4.1 define the same round instructions */ #define OPTION_MASK_ISA_ROUND (OPTION_MASK_ISA_SSE4_1 | OPTION_MASK_ISA_SSE5) @@ -286,7 +293,7 @@ enum ix86_tune_indices { X86_TUNE_LAST }; -extern unsigned int ix86_tune_features[X86_TUNE_LAST]; +extern unsigned char ix86_tune_features[X86_TUNE_LAST]; #define TARGET_USE_LEAVE ix86_tune_features[X86_TUNE_USE_LEAVE] #define TARGET_PUSH_MEMORY ix86_tune_features[X86_TUNE_PUSH_MEMORY] @@ -380,7 +387,7 @@ enum ix86_arch_indices { X86_ARCH_LAST }; -extern unsigned int ix86_arch_features[X86_ARCH_LAST]; +extern unsigned char ix86_arch_features[X86_ARCH_LAST]; #define TARGET_CMOVE ix86_arch_features[X86_ARCH_CMOVE] #define TARGET_CMPXCHG ix86_arch_features[X86_ARCH_CMPXCHG] @@ -392,15 +399,7 @@ extern unsigned int ix86_arch_features[X86_ARCH_LAST]; extern int x86_prefetch_sse; -#define TARGET_ABM x86_abm -#define TARGET_CMPXCHG16B x86_cmpxchg16b -#define TARGET_POPCNT x86_popcnt #define TARGET_PREFETCH_SSE x86_prefetch_sse -#define TARGET_SAHF x86_sahf -#define TARGET_RECIP x86_recip -#define TARGET_FUSED_MADD x86_fused_muladd -#define TARGET_AES (TARGET_SSE2 && x86_aes) -#define TARGET_PCLMUL (TARGET_SSE2 && x86_pclmul) #define ASSEMBLER_DIALECT (ix86_asm_dialect) @@ -475,7 +474,7 @@ enum calling_abi Don't use this macro to turn on various extra optimizations for `-O'. That is what `OPTIMIZATION_OPTIONS' is for. */ -#define OVERRIDE_OPTIONS override_options () +#define OVERRIDE_OPTIONS override_options (true) /* Define this to change the optimizations performed by default. */ #define OPTIMIZATION_OPTIONS(LEVEL, SIZE) \ @@ -537,196 +536,10 @@ extern const char *host_detect_local_cpu (int argc, const char **argv); #endif /* Target CPU builtins. */ -#define TARGET_CPU_CPP_BUILTINS() \ - do \ - { \ - size_t arch_len = strlen (ix86_arch_string); \ - size_t tune_len = strlen (ix86_tune_string); \ - int last_arch_char = ix86_arch_string[arch_len - 1]; \ - int last_tune_char = ix86_tune_string[tune_len - 1]; \ - \ - if (TARGET_64BIT) \ - { \ - builtin_assert ("cpu=x86_64"); \ - builtin_assert ("machine=x86_64"); \ - builtin_define ("__amd64"); \ - builtin_define ("__amd64__"); \ - builtin_define ("__x86_64"); \ - builtin_define ("__x86_64__"); \ - } \ - else \ - { \ - builtin_assert ("cpu=i386"); \ - builtin_assert ("machine=i386"); \ - builtin_define_std ("i386"); \ - } \ - \ - /* Built-ins based on -march=. */ \ - switch (ix86_arch) \ - { \ - case PROCESSOR_I386: \ - break; \ - case PROCESSOR_I486: \ - builtin_define ("__i486"); \ - builtin_define ("__i486__"); \ - break; \ - case PROCESSOR_PENTIUM: \ - builtin_define ("__i586"); \ - builtin_define ("__i586__"); \ - builtin_define ("__pentium"); \ - builtin_define ("__pentium__"); \ - if (last_arch_char == 'x') \ - builtin_define ("__pentium_mmx__"); \ - break; \ - case PROCESSOR_PENTIUMPRO: \ - builtin_define ("__i686"); \ - builtin_define ("__i686__"); \ - builtin_define ("__pentiumpro"); \ - builtin_define ("__pentiumpro__"); \ - break; \ - case PROCESSOR_GEODE: \ - builtin_define ("__geode"); \ - builtin_define ("__geode__"); \ - break; \ - case PROCESSOR_K6: \ - builtin_define ("__k6"); \ - builtin_define ("__k6__"); \ - if (last_arch_char == '2') \ - builtin_define ("__k6_2__"); \ - else if (last_arch_char == '3') \ - builtin_define ("__k6_3__"); \ - break; \ - case PROCESSOR_ATHLON: \ - builtin_define ("__athlon"); \ - builtin_define ("__athlon__"); \ - /* Only plain "athlon" lacks SSE. */ \ - if (last_arch_char != 'n') \ - builtin_define ("__athlon_sse__"); \ - break; \ - case PROCESSOR_K8: \ - builtin_define ("__k8"); \ - builtin_define ("__k8__"); \ - break; \ - case PROCESSOR_AMDFAM10: \ - builtin_define ("__amdfam10"); \ - builtin_define ("__amdfam10__"); \ - break; \ - case PROCESSOR_PENTIUM4: \ - builtin_define ("__pentium4"); \ - builtin_define ("__pentium4__"); \ - break; \ - case PROCESSOR_NOCONA: \ - builtin_define ("__nocona"); \ - builtin_define ("__nocona__"); \ - break; \ - case PROCESSOR_CORE2: \ - builtin_define ("__core2"); \ - builtin_define ("__core2__"); \ - break; \ - case PROCESSOR_GENERIC32: \ - case PROCESSOR_GENERIC64: \ - case PROCESSOR_max: \ - gcc_unreachable (); \ - } \ - \ - /* Built-ins based on -mtune=. */ \ - switch (ix86_tune) \ - { \ - case PROCESSOR_I386: \ - builtin_define ("__tune_i386__"); \ - break; \ - case PROCESSOR_I486: \ - builtin_define ("__tune_i486__"); \ - break; \ - case PROCESSOR_PENTIUM: \ - builtin_define ("__tune_i586__"); \ - builtin_define ("__tune_pentium__"); \ - if (last_tune_char == 'x') \ - builtin_define ("__tune_pentium_mmx__"); \ - break; \ - case PROCESSOR_PENTIUMPRO: \ - builtin_define ("__tune_i686__"); \ - builtin_define ("__tune_pentiumpro__"); \ - switch (last_tune_char) \ - { \ - case '3': \ - builtin_define ("__tune_pentium3__"); \ - /* FALLTHRU */ \ - case '2': \ - builtin_define ("__tune_pentium2__"); \ - break; \ - } \ - break; \ - case PROCESSOR_GEODE: \ - builtin_define ("__tune_geode__"); \ - break; \ - case PROCESSOR_K6: \ - builtin_define ("__tune_k6__"); \ - if (last_tune_char == '2') \ - builtin_define ("__tune_k6_2__"); \ - else if (last_tune_char == '3') \ - builtin_define ("__tune_k6_3__"); \ - break; \ - case PROCESSOR_ATHLON: \ - builtin_define ("__tune_athlon__"); \ - /* Only plain "athlon" lacks SSE. */ \ - if (last_tune_char != 'n') \ - builtin_define ("__tune_athlon_sse__"); \ - break; \ - case PROCESSOR_K8: \ - builtin_define ("__tune_k8__"); \ - break; \ - case PROCESSOR_AMDFAM10: \ - builtin_define ("__tune_amdfam10__"); \ - break; \ - case PROCESSOR_PENTIUM4: \ - builtin_define ("__tune_pentium4__"); \ - break; \ - case PROCESSOR_NOCONA: \ - builtin_define ("__tune_nocona__"); \ - break; \ - case PROCESSOR_CORE2: \ - builtin_define ("__tune_core2__"); \ - break; \ - case PROCESSOR_GENERIC32: \ - case PROCESSOR_GENERIC64: \ - break; \ - case PROCESSOR_max: \ - gcc_unreachable (); \ - } \ - \ - if (TARGET_MMX) \ - builtin_define ("__MMX__"); \ - if (TARGET_3DNOW) \ - builtin_define ("__3dNOW__"); \ - if (TARGET_3DNOW_A) \ - builtin_define ("__3dNOW_A__"); \ - if (TARGET_SSE) \ - builtin_define ("__SSE__"); \ - if (TARGET_SSE2) \ - builtin_define ("__SSE2__"); \ - if (TARGET_SSE3) \ - builtin_define ("__SSE3__"); \ - if (TARGET_SSSE3) \ - builtin_define ("__SSSE3__"); \ - if (TARGET_SSE4_1) \ - builtin_define ("__SSE4_1__"); \ - if (TARGET_SSE4_2) \ - builtin_define ("__SSE4_2__"); \ - if (TARGET_AES) \ - builtin_define ("__AES__"); \ - if (TARGET_PCLMUL) \ - builtin_define ("__PCLMUL__"); \ - if (TARGET_SSE4A) \ - builtin_define ("__SSE4A__"); \ - if (TARGET_SSE5) \ - builtin_define ("__SSE5__"); \ - if (TARGET_SSE_MATH && TARGET_SSE) \ - builtin_define ("__SSE_MATH__"); \ - if (TARGET_SSE_MATH && TARGET_SSE2) \ - builtin_define ("__SSE2_MATH__"); \ - } \ - while (0) +#define TARGET_CPU_CPP_BUILTINS() ix86_target_macros () + +/* Target Pragmas. */ +#define REGISTER_TARGET_PRAGMAS() ix86_register_pragmas () enum target_cpu_default { diff --git a/gcc/config/i386/i386.opt b/gcc/config/i386/i386.opt index 75c94ba..d5c0978 100644 --- a/gcc/config/i386/i386.opt +++ b/gcc/config/i386/i386.opt @@ -18,24 +18,58 @@ ; along with GCC; see the file COPYING3. If not see ; <http://www.gnu.org/licenses/>. +;; Definitions to add to the cl_target_option structure +;; -march= processor +TargetSave +unsigned char arch + +;; -mtune= processor +TargetSave +unsigned char tune + +;; -mfpath= +TargetSave +unsigned char fpmath + +;; branch cost +TargetSave +unsigned char branch_cost + +;; which flags were passed by the user +TargetSave +int ix86_isa_flags_explicit + +;; which flags were passed by the user +TargetSave +int target_flags_explicit + +;; whether -mtune was not specified +TargetSave +unsigned char tune_defaulted + +;; whether -march was specified +TargetSave +unsigned char arch_specified + +;; x86 options m128bit-long-double -Target RejectNegative Report Mask(128BIT_LONG_DOUBLE) +Target RejectNegative Report Mask(128BIT_LONG_DOUBLE) Save sizeof(long double) is 16 m80387 -Target Report Mask(80387) +Target Report Mask(80387) Save Use hardware fp m96bit-long-double -Target RejectNegative Report InverseMask(128BIT_LONG_DOUBLE) +Target RejectNegative Report InverseMask(128BIT_LONG_DOUBLE) Save sizeof(long double) is 12 maccumulate-outgoing-args -Target Report Mask(ACCUMULATE_OUTGOING_ARGS) +Target Report Mask(ACCUMULATE_OUTGOING_ARGS) Save Reserve space for outgoing arguments in the function prologue malign-double -Target Report Mask(ALIGN_DOUBLE) +Target Report Mask(ALIGN_DOUBLE) Save Align some doubles on dword boundary malign-functions= @@ -51,7 +85,7 @@ Target RejectNegative Joined Var(ix86_align_loops_string) Loop code aligned to this power of 2 malign-stringops -Target RejectNegative Report InverseMask(NO_ALIGN_STRINGOPS, ALIGN_STRINGOPS) +Target RejectNegative Report InverseMask(NO_ALIGN_STRINGOPS, ALIGN_STRINGOPS) Save Align destination of the string operations march= @@ -75,11 +109,11 @@ Target RejectNegative Joined Var(ix86_cmodel_string) Use given x86-64 code model mfancy-math-387 -Target RejectNegative Report InverseMask(NO_FANCY_MATH_387, USE_FANCY_MATH_387) +Target RejectNegative Report InverseMask(NO_FANCY_MATH_387, USE_FANCY_MATH_387) Save Generate sin, cos, sqrt for FPU mfp-ret-in-387 -Target Report Mask(FLOAT_RETURNS) +Target Report Mask(FLOAT_RETURNS) Save Return values of functions in FPU registers mfpmath= @@ -87,19 +121,19 @@ Target RejectNegative Joined Var(ix86_fpmath_string) Generate floating point mathematics using given instruction set mhard-float -Target RejectNegative Mask(80387) MaskExists +Target RejectNegative Mask(80387) MaskExists Save Use hardware fp mieee-fp -Target Report Mask(IEEE_FP) +Target Report Mask(IEEE_FP) Save Use IEEE math for fp comparisons minline-all-stringops -Target Report Mask(INLINE_ALL_STRINGOPS) +Target Report Mask(INLINE_ALL_STRINGOPS) Save Inline all known string operations minline-stringops-dynamically -Target Report Mask(INLINE_STRINGOPS_DYNAMICALLY) +Target Report Mask(INLINE_STRINGOPS_DYNAMICALLY) Save Inline memset/memcpy string operations, but perform inline version only for small blocks mintel-syntax @@ -107,23 +141,23 @@ Target Undocumented ;; Deprecated mms-bitfields -Target Report Mask(MS_BITFIELD_LAYOUT) +Target Report Mask(MS_BITFIELD_LAYOUT) Save Use native (MS) bitfield layout mno-align-stringops -Target RejectNegative Report Mask(NO_ALIGN_STRINGOPS) Undocumented +Target RejectNegative Report Mask(NO_ALIGN_STRINGOPS) Undocumented Save mno-fancy-math-387 -Target RejectNegative Report Mask(NO_FANCY_MATH_387) Undocumented +Target RejectNegative Report Mask(NO_FANCY_MATH_387) Undocumented Save mno-push-args -Target RejectNegative Report Mask(NO_PUSH_ARGS) Undocumented +Target RejectNegative Report Mask(NO_PUSH_ARGS) Undocumented Save mno-red-zone -Target RejectNegative Report Mask(NO_RED_ZONE) Undocumented +Target RejectNegative Report Mask(NO_RED_ZONE) Undocumented Save momit-leaf-frame-pointer -Target Report Mask(OMIT_LEAF_FRAME_POINTER) +Target Report Mask(OMIT_LEAF_FRAME_POINTER) Save Omit the frame pointer in leaf functions mpc @@ -135,11 +169,11 @@ Target RejectNegative Joined Var(ix86_preferred_stack_boundary_string) Attempt to keep stack aligned to this power of 2 mpush-args -Target Report InverseMask(NO_PUSH_ARGS, PUSH_ARGS) +Target Report InverseMask(NO_PUSH_ARGS, PUSH_ARGS) Save Use push instructions to save outgoing arguments mred-zone -Target RejectNegative Report InverseMask(NO_RED_ZONE, RED_ZONE) +Target RejectNegative Report InverseMask(NO_RED_ZONE, RED_ZONE) Save Use red-zone in the x86-64 code mregparm= @@ -147,15 +181,15 @@ Target RejectNegative Joined Var(ix86_regparm_string) Number of registers used to pass integer arguments mrtd -Target Report Mask(RTD) +Target Report Mask(RTD) Save Alternate calling convention msoft-float -Target InverseMask(80387) +Target InverseMask(80387) Save Do not use hardware fp msseregparm -Target RejectNegative Mask(SSEREGPARM) +Target RejectNegative Mask(SSEREGPARM) Save Use SSE register passing conventions for SF and DF mode mstackrealign @@ -163,7 +197,7 @@ Target Report Var(ix86_force_align_arg_pointer) Realign stack in prologue mstack-arg-probe -Target Report Mask(STACK_PROBE) +Target Report Mask(STACK_PROBE) Save Enable stack probing mstringop-strategy= @@ -186,104 +220,105 @@ mveclibabi= Target RejectNegative Joined Var(ix86_veclibabi_string) Vector library ABI to use +mrecip +Target Report Mask(RECIP) Save +Generate reciprocals instead of divss and sqrtss. + +mcld +Target Report Mask(CLD) Save +Generate cld instruction in the function prologue. + +mno-fused-madd +Target RejectNegative Report Mask(NO_FUSED_MADD) Undocumented Save + +mfused-madd +Target Report InverseMask(NO_FUSED_MADD, FUSED_MADD) Save +Enable automatic generation of fused floating point multiply-add instructions +if the ISA supports such instructions. The -mfused-madd option is on by +default. + ;; ISA support m32 -Target RejectNegative Negative(m64) Report InverseMask(ISA_64BIT) Var(ix86_isa_flags) VarExists +Target RejectNegative Negative(m64) Report InverseMask(ISA_64BIT) Var(ix86_isa_flags) VarExists Save Generate 32bit i386 code m64 -Target RejectNegative Negative(m32) Report Mask(ISA_64BIT) Var(ix86_isa_flags) VarExists +Target RejectNegative Negative(m32) Report Mask(ISA_64BIT) Var(ix86_isa_flags) VarExists Save Generate 64bit x86-64 code mmmx -Target Report Mask(ISA_MMX) Var(ix86_isa_flags) VarExists +Target Report Mask(ISA_MMX) Var(ix86_isa_flags) VarExists Save Support MMX built-in functions m3dnow -Target Report Mask(ISA_3DNOW) Var(ix86_isa_flags) VarExists +Target Report Mask(ISA_3DNOW) Var(ix86_isa_flags) VarExists Save Support 3DNow! built-in functions m3dnowa -Target Undocumented Mask(ISA_3DNOW_A) Var(ix86_isa_flags) VarExists +Target Undocumented Mask(ISA_3DNOW_A) Var(ix86_isa_flags) VarExists Save Support Athlon 3Dnow! built-in functions msse -Target Report Mask(ISA_SSE) Var(ix86_isa_flags) VarExists +Target Report Mask(ISA_SSE) Var(ix86_isa_flags) VarExists Save Support MMX and SSE built-in functions and code generation msse2 -Target Report Mask(ISA_SSE2) Var(ix86_isa_flags) VarExists +Target Report Mask(ISA_SSE2) Var(ix86_isa_flags) VarExists Save Support MMX, SSE and SSE2 built-in functions and code generation msse3 -Target Report Mask(ISA_SSE3) Var(ix86_isa_flags) VarExists +Target Report Mask(ISA_SSE3) Var(ix86_isa_flags) VarExists Save Support MMX, SSE, SSE2 and SSE3 built-in functions and code generation mssse3 -Target Report Mask(ISA_SSSE3) Var(ix86_isa_flags) VarExists +Target Report Mask(ISA_SSSE3) Var(ix86_isa_flags) VarExists Save Support MMX, SSE, SSE2, SSE3 and SSSE3 built-in functions and code generation msse4.1 -Target Report Mask(ISA_SSE4_1) Var(ix86_isa_flags) VarExists +Target Report Mask(ISA_SSE4_1) Var(ix86_isa_flags) VarExists Save Support MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1 built-in functions and code generation msse4.2 -Target Report Mask(ISA_SSE4_2) Var(ix86_isa_flags) VarExists +Target Report Mask(ISA_SSE4_2) Var(ix86_isa_flags) VarExists Save Support MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1 and SSE4.2 built-in functions and code generation msse4 -Target RejectNegative Report Mask(ISA_SSE4_2) MaskExists Var(ix86_isa_flags) VarExists +Target RejectNegative Report Mask(ISA_SSE4_2) MaskExists Var(ix86_isa_flags) VarExists Save Support MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1 and SSE4.2 built-in functions and code generation mno-sse4 -Target RejectNegative Report InverseMask(ISA_SSE4_1) MaskExists Var(ix86_isa_flags) VarExists +Target RejectNegative Report InverseMask(ISA_SSE4_1) MaskExists Var(ix86_isa_flags) VarExists Save Do not support SSE4.1 and SSE4.2 built-in functions and code generation msse4a -Target Report Mask(ISA_SSE4A) Var(ix86_isa_flags) VarExists +Target Report Mask(ISA_SSE4A) Var(ix86_isa_flags) VarExists Save Support MMX, SSE, SSE2, SSE3 and SSE4A built-in functions and code generation msse5 -Target Report Mask(ISA_SSE5) Var(ix86_isa_flags) VarExists +Target Report Mask(ISA_SSE5) Var(ix86_isa_flags) VarExists Save Support SSE5 built-in functions and code generation -;; Instruction support - -mcld -Target Report Mask(CLD) -Generate cld instruction in the function prologue. - mabm -Target Report RejectNegative Var(x86_abm) +Target Report Mask(ISA_ABM) Var(ix86_isa_flags) VarExists Save Support code generation of Advanced Bit Manipulation (ABM) instructions. -mcx16 -Target Report RejectNegative Var(x86_cmpxchg16b) -Support code generation of cmpxchg16b instruction. - mpopcnt -Target Report RejectNegative Var(x86_popcnt) +Target Report Mask(ISA_POPCNT) Var(ix86_isa_flags) VarExists Save Support code generation of popcnt instruction. +mcx16 +Target Report Mask(ISA_CX16) Var(ix86_isa_flags) VarExists Save +Support code generation of cmpxchg16b instruction. + msahf -Target Report RejectNegative Var(x86_sahf) +Target Report Mask(ISA_SAHF) Var(ix86_isa_flags) VarExists Save Support code generation of sahf instruction in 64bit x86-64 code. -mrecip -Target Report RejectNegative Var(x86_recip) -Generate reciprocals instead of divss and sqrtss. - -mfused-madd -Target Report Var(x86_fused_muladd) Init(1) -Enable automatic generation of fused floating point multiply-add instructions -if the ISA supports such instructions. The -mfused-madd option is on by -default. - maes -Target Report RejectNegative Var(x86_aes) +Target Report Mask(ISA_AES) Var(ix86_isa_flags) VarExists Save Support AES built-in functions and code generation mpclmul -Target Report RejectNegative Var(x86_pclmul) +Target Report Mask(ISA_PCLMUL) Var(ix86_isa_flags) VarExists Save Support PCLMUL built-in functions and code generation diff --git a/gcc/config/i386/t-i386 b/gcc/config/i386/t-i386 new file mode 100644 index 0000000..4c0c046 --- /dev/null +++ b/gcc/config/i386/t-i386 @@ -0,0 +1,13 @@ +i386.o: $(CONFIG_H) $(SYSTEM_H) coretypes.h $(TM_H) \ + $(RTL_H) $(TREE_H) $(TM_P_H) $(REGS_H) hard-reg-set.h \ + $(REAL_H) insn-config.h conditions.h output.h insn-codes.h \ + $(INSN_ATTR_H) $(FLAGS_H) $(C_COMMON_H) except.h $(FUNCTION_H) \ + $(RECOG_H) $(EXPR_H) $(OPTABS_H) toplev.h $(BASIC_BLOCK_H) \ + $(GGC_H) $(TARGET_H) $(TARGET_DEF_H) langhooks.h $(CGRAPH_H) \ + $(TREE_GIMPLE_H) dwarf2.h $(DF_H) tm-constrs.h $(PARAMS_H) + +i386-c.o: $(srcdir)/config/i386/i386-c.c \ + $(srcdir)/config/i386/i386-protos.h $(CONFIG_H) $(SYSTEM_H) coretypes.h \ + $(TM_H) $(RTL_H) $(TREE_H) $(TM_P_H) $(FLAGS_H) $(C_COMMON_H) $(GGC_H) \ + $(TARGET_H) $(TARGET_DEF_H) $(CPPLIB_H) $(C_PRAGMA_H) + $(CC) -c $(ALL_CFLAGS) $(ALL_CPPFLAGS) $(INCLUDES) $(srcdir)/config/i386/i386-c.c |