mirrors/gcc
0
0
Fork 0
mirror of https://gcc.gnu.org/git/gcc.git synced 2024-12-04 01:04:26 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

i386.c (ix86_rtx_costs): Check op0 for ZERO_EXTEND when computing costs for widening multiplication.

Browse Source 
* config/i386/i386.c (ix86_rtx_costs) [MULT]: Check op0 for
        ZERO_EXTEND when computing costs for widening multiplication.

From-SVN: r130809
This commit is contained in:
Torbjorn Granlund 2007-12-13 10:19:11 +01:00 committed by Uros Bizjak
parent fe9e534218
commit e3dafdf4d0
2 changed files with 69 additions and 83 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

150
gcc/ChangeLog
View File

@ -1,3 +1,8 @@
2007-12-13 Torbjorn Granlund <tege@swox.com>
* config/i386/i386.c (ix86_rtx_costs) [MULT]: Check op0 for
ZERO_EXTEND when computing costs for widening multiplication.
2007-12-13 Richard Earnshaw <rearnsha@arm.com>
PR target/30192
@ -92,8 +97,7 @@
to extract the integer from operands[2].
(vec_setv2di_internal): Likewise.
* config/arm/arm.c (arm_size_return_regs): Make return type
unsigned.
* config/arm/arm.c (arm_size_return_regs): Make return type unsigned.
(arm_init_neon_builtins): Move code after declarations.
2007-12-10 Hans-Peter Nilsson <hp@axis.com>
@ -535,8 +539,7 @@
* tree-vrp.c (adjust_range_with_scev): Clear scev cache.
(record_numbers_of_iterations): New function.
(execute_vrp): Cache the numbers of iterations of loops.
* tree-scalar-evolution.c (scev_reset_except_niters):
New function.
* tree-scalar-evolution.c (scev_reset_except_niters): New function.
(scev_reset): Use scev_reset_except_niters.
* tree-scalar-evolution.h (scev_reset_except_niters): Declare.
@ -1377,9 +1380,9 @@
PR tree-optimization/33953
* tree-vect-transform.c (vectorizable_operation): In case of SLP,
allocate vec_oprnds1 according to the number of created vector
statements. In case of shift with scalar argument, store scalar operand
for every vector statement to be created for the SLP node. Fix a
comment.
statements. In case of shift with scalar argument, store scalar
operand for every vector statement to be created for the SLP node.
Fix a comment.
2007-11-10 Steven Bosscher <stevenb.gcc@gmail.com>
@ -1445,8 +1448,7 @@
2007-11-09 Richard Guenther <rguenther@suse.de>
* tree-flow.h (struct ptr_info_def): Make escape_mask a
ENUM_BITFIELD.
* tree-flow.h (struct ptr_info_def): Make escape_mask a ENUM_BITFIELD.
(struct ptr_info_def): Likewise.
(enum escape_type): Also use bit zero.
@ -1845,8 +1847,7 @@
(in_safe_group_barrier): New variable.
(rws_update): Remove first argument, always set rws_sum array.
(rws_access_regno): Use rws_insn_set macro. Only call rws_update
or update rws_sum if !in_safe_group_barrier, adjust rws_update
args.
or update rws_sum if !in_safe_group_barrier, adjust rws_update args.
(rtx_needs_barrier): Use rws_insn_test macro.
(safe_group_barrier_needed): Don't save/restore rws_sum array,
instead set/clear in_safe_group_barrier flag.
@ -1996,8 +1997,8 @@
* config/rs6000/rs6000.opt: Rename -mdfp option to -mhard-dfp.
* config/s390/s390.c: (override_options): Replace -mhard-dfp/-msoft-dfp
with -mhard-dfp/-mno-hard-dfp.
* config/s390/s390.c: (override_options): Replace
-mhard-dfp/-msoft-dfp with -mhard-dfp/-mno-hard-dfp.
* config/s390/s390.opt: Likewise.
* config/s390/s390.h: Add MASK_HARD_DFP to the TARGET_DEFAULT
definition.
@ -2493,7 +2494,7 @@
SSA_NAME pointers.
2007-10-22 Seongbae Park <seongbae.park@gmail.com>
David S. Miller <davem@davemloft.net>
David S. Miller <davem@davemloft.net>
* df-scan.c (df_get_call_refs): Mark global registers as both a
DF_REF_REG_USE and a non-clobber DF_REF_REG_DEF.
@ -3010,8 +3011,7 @@
2007-10-19 Richard Guenther <rguenther@suse.de>
PR middle-end/32921
* tree.c (build_array_type): Do not re-layout unbound array
types.
* tree.c (build_array_type): Do not re-layout unbound array types.
2007-10-19 Richard Sandiford <rsandifo@nildram.co.uk>
@ -3765,8 +3765,7 @@
PR middle-end/33669
* ra-conflict.c (record_one_conflict_between_regnos,
set_conflicts_for_earlyclobber, global_conflicts): Improved
logging.
set_conflicts_for_earlyclobber, global_conflicts): Improved logging.
(global_conflicts): Removed incorrect check.
2007-10-09 Richard Sandiford <rsandifo@nildram.co.uk>
@ -3909,8 +3908,7 @@
PR tree-optimization/33655
PR middle-end/22156
* tree-sra.c (bitfield_overlaps_p): Handle array and complex
elements.
* tree-sra.c (bitfield_overlaps_p): Handle array and complex elements.
2007-10-06 Alexandre Oliva <aoliva@redhat.com>
@ -4024,8 +4022,7 @@
2007-10-04 Geoffrey Keating <geoffk@apple.com>
* cgraphunit.c (cgraph_build_static_cdtor): Don't set
DECL_IGNORED_P.
* cgraphunit.c (cgraph_build_static_cdtor): Don't set DECL_IGNORED_P.
2007-10-04 Anatoly Sokolov <aesok@post.ru>
@ -4150,8 +4147,7 @@
2007-10-02 Richard Sandiford <rsandifo@nildram.co.uk>
PR middle-end/33617
* expr.c (expand_expr_addr_expr_1): Pass CONSTRUCTORs to
expand_expr.
* expr.c (expand_expr_addr_expr_1): Pass CONSTRUCTORs to expand_expr.
2007-10-02 David Daney <ddaney@avtrex.com>
@ -4263,14 +4259,12 @@
(canon_type_for_field): New.
(try_instantiate_multiple_fields): New. Infer widest possible
access mode from decl or member type, but clip it at word
size, and only widen it if a field crosses an alignment
boundary.
size, and only widen it if a field crosses an alignment boundary.
(instantiate_missing_elements): Use them.
(generate_one_element_ref): Handle BIT_FIELD_REFs.
(scalar_bitfield_p): New.
(sra_build_assignment): Optimize assignments from scalarizable
BIT_FIELD_REFs. Use BITS_BIG_ENDIAN to determine shift
counts.
BIT_FIELD_REFs. Use BITS_BIG_ENDIAN to determine shift counts.
(REPLDUP): New.
(sra_build_bf_assignment): New. Optimize assignments to
scalarizable BIT_FIELD_REFs.
@ -4292,8 +4286,7 @@
BIT_FIELD_REFs. Explode or sync needed members for
BIT_FIELD_REFs accesses or assignments. Use REPLDUP.
(scalarize_copy): Use REPLDUP.
(scalarize_ldst): Move assert before dereference. Adjust EH
handling.
(scalarize_ldst): Move assert before dereference. Adjust EH handling.
(dump_sra_elt_name): Handle BIT_FIELD_REFs.
2007-10-01 Paolo Bonzini <bonzini@gnu.org>
@ -4579,8 +4572,7 @@
2007-09-25 DJ Delorie <dj@redhat.com>
PR target/33551
* config/m32c/m32c.c (m32c_immd_dbl_mov): Use INTVAL instead of
XINT.
* config/m32c/m32c.c (m32c_immd_dbl_mov): Use INTVAL instead of XINT.
2007-09-25 Michael Meissner <michael.meissner@amd.com>
@ -4888,8 +4880,7 @@
2007-09-22 Richard Guenther <rguenther@suse.de>
PR tree-optimization/33146
* fold-const.c (fold_binary): Use the original tree
for negating.
* fold-const.c (fold_binary): Use the original tree for negating.
* tree.h (STRIP_SIGN_NOPS): Converting from or to pointer
also changes "sign".
@ -5352,8 +5343,7 @@
* doc/tm.texi (LIBGCC2_UNWIND_ATTRIBUTE): Document.
* unwind-generic.h (LIBGCC2_UNWIND_ATTRIBUTE): Define.
(_Unwind_RaiseException): Add LIBGCC2_UNWIND_ATTRIBUTE to
declaration.
(_Unwind_RaiseException): Add LIBGCC2_UNWIND_ATTRIBUTE to declaration.
(_Unwind_ForcedUnwind): Likewise.
(_Unwind_Resume): Likewise.
(_Unwind_Resume_or_Rethrow): Likewise.
@ -5417,8 +5407,7 @@
PR target/32337
* config/ia64/ia64.c (find_gr_spill): Don't decrement
current_frame_info.n_local_regs. Don't return emitted local
regs.
current_frame_info.n_local_regs. Don't return emitted local regs.
(ia64_compute_frame_size): Improve unwind hack to put
RP, PFS, FP in that order by allowing some of the registers
been already emitted, as long as they are emitted to the
@ -5563,19 +5552,15 @@
pointer to hold the argument.
(function_args_iter_cond): Inline function to return the next
argument.
(function_args_iter_cond_next): Advance the function args
iterator.
(stdarg_p): New function, return true if variable argument
function.
(prototype_p): New function, return true if function is
prototyped.
(function_args_iter_cond_next): Advance the function args iterator.
(stdarg_p): New function, return true if variable argument function.
(prototype_p): New function, return true if function is prototyped.
(function_args_count): New function, count the number of arguments
of a function.
* tree.c (stdarg_p): New function, return true if variable
argument function.
(prototype_p): New function, return true if function is
prototyped.
(prototype_p): New function, return true if function is prototyped.
* config/i386/i386.h (TARGET_SSE5): New macro for SSE5.
(TARGET_ROUND): New macro for the round/ptest instructions which
@ -6067,8 +6052,8 @@
2007-09-12 Jan Hubicka <jh@suse.cz>
PR target/33393
* i386.md (floatsisf2_mixed_memory, floatsisf2_sse_memory): Disable
for !SSE_MATH
* config/i386/i386.md (floatsisf2_mixed_memory, floatsisf2_sse_memory):
Disable for !SSE_MATH
2007-09-12 Christian Bruel <christian.bruel@st.com>
@ -6084,8 +6069,8 @@
2007-09-12 Ira Rosen <irar@il.ibm.com>
* params.def (PARAM_MIN_VECT_LOOP_BOUND): Change default and minimum
to 1.
* params.def (PARAM_MIN_VECT_LOOP_BOUND): Change default and
minimum to 1.
2007-09-11 James E. Wilson <wilson@specifix.com>
@ -6285,15 +6270,15 @@
2007-09-11 Zdenek Dvorak <ook@ucw.cz>
* fold-const.c (extract_muldiv_1): Do not simplify
var * c * c to var.
* fold-const.c (extract_muldiv_1): Do not simplify var * c * c to var.
2007-09-11 Jan Hubicka <jh@suse.cz>
* i386.h (ix86_tune_indices): Add X86_TUNE_INTER_UNIT_CONVERSIONS.
* config/i386/i386.h (ix86_tune_indices): Add
X86_TUNE_INTER_UNIT_CONVERSIONS.
(TARGET_INTER_UNIT_CONVERSIONS): New.
* i386.md (floatsi expanders): Remove redundant check for SImode
source; offload to memory when asked for.
* config/i386/i386.md (floatsi expanders): Remove redundant check
for SImode source; offload to memory when asked for.
(floatsisf2_mixed, floatsisf2_sse, floatsidf2_mixed, floatsidf2_sse
floatdisf2_mixed, floatsisf2_sse, floatsidf2_mixed, floatsidf2_sse):
Update conditions;
@ -6581,7 +6566,7 @@
2007-09-09 H.J. Lu <hongjiu.lu@intel.com>
* i386.md (*floatsisf2_mixed_vector): Use cvtdq2ps instead
* config/i386/i386.md (*floatsisf2_mixed_vector): Use cvtdq2ps instead
of cvtpq2ps.
(*floatsisf2_sse_vector): Likewise.
@ -6615,10 +6600,10 @@
2007-09-09 Jan Hubicka <jh@suse.cz>
Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
* i386.h (ix86_tune_indices): Add X86_USE_VECTOR_CONVERTS.
* config/i386/i386.h (ix86_tune_indices): Add X86_USE_VECTOR_CONVERTS.
(TARGET_USE_VECTOR_CONVERTS): New.
* i386.md: New post-reload splitters for converting SF to DF and DF to
SF.
* config/i386/i386.md: New post-reload splitters for converting
SF to DF and DF to SF.
(floatsi* expander): Special case vector conversions.
(floatsisf2_mixed_vector, floatsisf2_sse_vector_nointernunit,
floatsisf2_sse_vector_internunit, floatsisf2_sse_vector,
@ -6626,8 +6611,8 @@
(floatsisf2_mixed, floatsisf2_sse, floatsidf2_mixed, floatsidf2_sse):
Disable when doing vector converts.
(floatsi<mode>2_i387): Disable when
* sse.md (vec_dupv2df): Export.
* i386.c (ix86_tune_features): Enable SSE conversions.
* config/i386/sse.md (vec_dupv2df): Export.
* config/i386/i386.c (ix86_tune_features): Enable SSE conversions.
2007-09-09 Richard Guenther <rguenther@suse.de>
@ -7356,8 +7341,8 @@
2007-09-06 Jan Hubicka <jh@suse.cz>
* i386.c (ix86_expand_lround, ix86_expand_round): Update call of
real_2expN.
* config/i386/i386.c (ix86_expand_lround, ix86_expand_round):
Update call of real_2expN.
2007-09-06 Richard Sandiford <richard@codesourcery.com>
@ -7514,8 +7499,8 @@
2007-09-06 Jan Hubicka <jh@suse.cz>
* i386.c (ix86_expand_lround, ix86_expand_round): Update call of
real_2expN.
* config/i386/i386.c (ix86_expand_lround, ix86_expand_round):
Update call of real_2expN.
2007-09-06 Jan Hubicka <jh@suse.cz>
@ -10095,7 +10080,7 @@
2007-08-19 Jan Hubicka <jh@suse.cz>
* i386.md: Replace "rim" and "mri" constraints by "g".
* config/i386/i386.md: Replace "rim" and "mri" constraints by "g".
2007-08-19 Joseph Myers <joseph@codesourcery.com>
@ -10547,8 +10532,8 @@
2007-08-14 Kai Tietz <kai.tietz@onevision.com>
* i386.c: (legitimize_address): Move dllimported variable check
infront of legitimizing pic address of CONST symbols.
* config/i386/i386.c (legitimize_address): Move dllimported variable
check infront of legitimizing pic address of CONST symbols.
2007-08-14 Steve Ellcey <sje@cup.hp.com>
@ -11924,8 +11909,8 @@
2007-08-03 Jan Hubicka <jh@suse.cz>
* i386.c (setup_incoming_varargs_64): Tolerate 64bit preferred
stack boundary.
* config/i386/i386.c (setup_incoming_varargs_64): Tolerate 64bit
preferred stack boundary.
2007-08-03 David Edelsohn <edelsohn@gnu.org>
@ -13673,7 +13658,7 @@
2007-07-23 Jan Hubicka <jH@suse.cz>
* i386.c (ix86_secondary_memory_needed): Break out to...
* config/i386/i386.c (ix86_secondary_memory_needed): Break out to...
(inline_secondary_memory_needed): ... here.
(ix86_memory_move_cost): Break out to ...
(inline_memory_move_cost): ... here; add support for IN value of 2 for
@ -19411,7 +19396,7 @@
extra_headers.
(x86_64-*-*): Likewise.
* i386/i386-modes.def (V2QI): New.
* config/i386/i386-modes.def (V2QI): New.
* config/i386/i386.c (ix86_handle_option): Handle SSE4.1 and
SSE4A.
@ -25875,9 +25860,9 @@
* reg-stack.c (reg_to_stack): Large models don't allow NAN to be
loaded for constant large models. Non-large 64bit PIC can do.
* i386.h (CASE_VECTOR_MODE): Large PIC cases are 64bit.
* config/i386/i386.h (CASE_VECTOR_MODE): Large PIC cases are 64bit.
* cmodel.h: Add LARGE PIC.
* i386.md (UNSPEC_PLTOFF): New.
* config/i386/i386.md (UNSPEC_PLTOFF): New.
(UNSPEC_SET_RIP, UNSPEC_SET_GOT_OFFSET): New; renumber other
unspecs as needed.
(*call_1_rex64): Disable for large models.
@ -25886,10 +25871,10 @@
(*call_value_1_rex64_large): New.
(set_rip_rex4): New.
(set_got_offset_rex64): New.
* predicates.md (constant_call_address_operand): For large model
constant calls are not possible.
* i386-protos.h (construct_plt_address): Declare.
* i386.c (override_options): Accept large models.
* config/i386/predicates.md (constant_call_address_operand): For
large model constant calls are not possible.
* config/i386/i386-protos.h (construct_plt_address): Declare.
* config/i386/i386.c (override_options): Accept large models.
(ix86_expand_prologue): Expand large PIC GOT pointer load.
(legitimate_constant_p): Add new UNSPECs.
(legitimate_pic_operand_p): Likewise.
@ -26883,7 +26868,7 @@
Jan Hubicka <jh@suse.cz>
PR target/30778
* i386.c (counter_mode): New function.
* cofnig/i386/i386.c (counter_mode): New function.
(expand_set_or_movmem_via_loop): Use it.
(expand_movmem_epilogue): Likewise; fix pasto.
(ix86_expand_movmem): Do emit guard even for constant counts.
@ -30316,7 +30301,8 @@
* Makefile.in (LIBGCOV): Add new constants.
* libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
__gcov_ior_profiler): New.
* i386.md (movmem/setmem expanders): Add new optional arguments.
* config/i386/i386.md (movmem/setmem expanders): Add new optional
arguments.
2007-01-28 David Edelsohn <edelsohn@gnu.org>
@ -31184,8 +31170,8 @@
2007-01-18 Michael Meissner <michael.meissner@amd.com>
* i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
type correct.
* config/i386/i386.c (ix86_compute_frame_layout): Make fprintf's
in #if 0 code type correct.
2007-01-18 Jan Hubicka <jh@suse.cz>

2
gcc/config/i386/i386.c
View File

@ -22236,7 +22236,7 @@ ix86_rtx_costs (rtx x, int code, int outer_code_i, int *total)
nbits = 7;
/* Compute costs correctly for widening multiplication. */
if ((GET_CODE (op0) == SIGN_EXTEND || GET_CODE (op1) == ZERO_EXTEND)
if ((GET_CODE (op0) == SIGN_EXTEND || GET_CODE (op0) == ZERO_EXTEND)
&& GET_MODE_SIZE (GET_MODE (XEXP (op0, 0))) * 2
== GET_MODE_SIZE (mode))
{