procyberian/gcc
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI, [...]): Move to right spot in the file.

Browse source 
* x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI, X86_TUNE_ADJUST_UNROLL,
	X86_TUNE_ONE_IF_CONV_INSN): Move to right spot in the file.

From-SVN: r253598
This commit is contained in:
Jan Hubicka 2017-10-10 19:47:35 +02:00 • committed by Jan Hubicka
parent f1e2a00403
commit 3652a4d243
2 changed files with 22 additions and 16 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

6
gcc/ChangeLog
View file

@ -1,3 +1,8 @@
2017-10-10 Jan Hubicka <hubicka@ucw.cz>
* x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI, X86_TUNE_ADJUST_UNROLL,
X86_TUNE_ONE_IF_CONV_INSN): Move to right spot in the file.
2017-10-10 Richard Sandiford <richard.sandiford@linaro.org>
* wide-int.h (wide_int_ref_storage): Make host_dependent_precision
@ -359,6 +364,7 @@
* builtins.def (BUILT_IN_SETJMP): Revert latest change.
2017-10-08 Jan Hubicka <hubicka@ucw.cz>
2017-10-08 Jan Hubicka <hubicka@ucw.cz>
* i386.c (ix86_expand_set_or_movmem): Disable 512bit loops for targets

32
gcc/config/i386/x86-tune.def
View file

@ -284,6 +284,22 @@ DEF_TUNE (X86_TUNE_USE_BT, "use_bt",
m_CORE_ALL | m_BONNELL | m_SILVERMONT | m_KNL | m_KNM | m_INTEL
| m_LAKEMONT | m_AMD_MULTIPLE | m_GENERIC)
/* X86_TUNE_AVOID_FALSE_DEP_FOR_BMI: Avoid false dependency
for bit-manipulation instructions. */
DEF_TUNE (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI, "avoid_false_dep_for_bmi",
m_SANDYBRIDGE | m_HASWELL | m_GENERIC)
/* X86_TUNE_ADJUST_UNROLL: This enables adjusting the unroll factor based
on hardware capabilities. Bdver3 hardware has a loop buffer which makes
unrolling small loop less important. For, such architectures we adjust
the unroll factor so that the unrolled loop fits the loop buffer. */
DEF_TUNE (X86_TUNE_ADJUST_UNROLL, "adjust_unroll_factor", m_BDVER3 | m_BDVER4)
/* X86_TUNE_ONE_IF_CONV_INSNS: Restrict a number of cmov insns in
if-converted sequence to one. */
DEF_TUNE (X86_TUNE_ONE_IF_CONV_INSN, "one_if_conv_insn",
m_SILVERMONT | m_KNL | m_KNM | m_INTEL | m_CORE_ALL | m_GENERIC)
/*****************************************************************************/
/* 387 instruction selection tuning */
/*****************************************************************************/
@ -503,11 +519,6 @@ DEF_TUNE (X86_TUNE_NOT_VECTORMODE, "not_vectormode", m_K6)
DEF_TUNE (X86_TUNE_AVOID_VECTOR_DECODE, "avoid_vector_decode",
m_K8)
/* X86_TUNE_AVOID_FALSE_DEP_FOR_BMI: Avoid false dependency
for bit-manipulation instructions. */
DEF_TUNE (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI, "avoid_false_dep_for_bmi",
m_SANDYBRIDGE | m_HASWELL | m_GENERIC)
/*****************************************************************************/
/* This never worked well before. */
/*****************************************************************************/
@ -525,14 +536,3 @@ DEF_TUNE (X86_TUNE_QIMODE_MATH, "qimode_math", ~0U)
arithmetic to 32bit via PROMOTE_MODE macro. This code generation scheme
is usually used for RISC targets. */
DEF_TUNE (X86_TUNE_PROMOTE_QI_REGS, "promote_qi_regs", 0U)
/* X86_TUNE_ADJUST_UNROLL: This enables adjusting the unroll factor based
on hardware capabilities. Bdver3 hardware has a loop buffer which makes
unrolling small loop less important. For, such architectures we adjust
the unroll factor so that the unrolled loop fits the loop buffer. */
DEF_TUNE (X86_TUNE_ADJUST_UNROLL, "adjust_unroll_factor", m_BDVER3 | m_BDVER4)
/* X86_TUNE_ONE_IF_CONV_INSNS: Restrict a number of cmov insns in
if-converted sequence to one. */
DEF_TUNE (X86_TUNE_ONE_IF_CONV_INSN, "one_if_conv_insn",
m_SILVERMONT | m_KNL | m_KNM | m_INTEL | m_CORE_ALL | m_GENERIC)