Hi,
I added AVX256_SPLIT_UNALIGNED_STORE to ix86_tune_indices
and put m_COREI7, m_BDVER1 and m_GENERIC as the targets that
enable it.
Is this OK?
Thanks,
Changpeng
________________________________________
From: H.J. Lu [hjl.to...@gmail.com]
Sent: Friday, June 17, 2011 1:08 PM
To: Fang, Changpeng
Cc: Richard Guenther; gcc-patches@gcc.gnu.org
Subject: Re: [PATCH, PR 49089] Don't split AVX256 unaligned loads by default on
bdver1 and generic
On Fri, Jun 17, 2011 at 10:45 AM, Fang, Changpeng
<changpeng.f...@amd.com> wrote:
>>Why not just move AVX256_SPLIT_UNALIGNED_STORE
>>and AVX256_SPLIT_UNALIGNED_LOAD to ix86_tune_indices?
>
> I would like to keep the -m option so that at least we can explicitly turn
> off the splittings when regressions are found!
I prefer to implement it the same way as:
x86_accumulate_outgoing_args
x86_arch_always_fancy_math_387
> By the way, I can add an index for store splitting, if you want.
>
Yes, please.
--
H.J.
From 91e715213bb37d089cb490e769b115d1d131918f Mon Sep 17 00:00:00 2001
From: Changpeng Fang <chfang@huainan.(none)>
Date: Mon, 13 Jun 2011 13:13:32 -0700
Subject: [PATCH 2/2] pr49089: enable avx256 splitting unaligned load/store only when beneficial
* config/i386/i386.h (ix86_tune_indices): Introduce
X86_TUNE_AVX256_SPLIT_UNALIGNED_LOAD_OPTIMAL and
X86_TUNE_AVX256_SPLIT_UNALIGNED_STORE_OPTIMAL.
(TARGET_AVX256_SPLIT_UNALIGNED_LOAD_OPTIMAL): New definition.
(TARGET_AVX256_SPLIT_UNALIGNED_STORE_OPTIMAL): New definition.
* config/i386/i386.c (ix86_tune_features): Add entries for
X86_TUNE_AVX256_SPLIT_UNALIGNED_LOAD_OPTIMAL and
X86_TUNE_AVX256_SPLIT_UNALIGNED_STORE_OPTIMAL.
(ix86_option_override_internal): Enable avx256 unaligned load(store)
splitting when TARGET_AVX256_SPLIT_UNALIGNED_LOAD(STORE)_OPTIMAL
are set.
---
gcc/config/i386/i386.c | 17 ++++++++++++++---
gcc/config/i386/i386.h | 4 ++++
2 files changed, 18 insertions(+), 3 deletions(-)
diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
index 7b266b9..b50d349 100644
--- a/gcc/config/i386/i386.c
+++ b/gcc/config/i386/i386.c
@@ -2088,7 +2088,16 @@ static unsigned int initial_ix86_tune_features[X86_TUNE_LAST] = {
/* X86_SOFTARE_PREFETCHING_BENEFICIAL: Enable software prefetching
at -O3. For the moment, the prefetching seems badly tuned for Intel
chips. */
- m_K6_GEODE | m_AMD_MULTIPLE
+ m_K6_GEODE | m_AMD_MULTIPLE,
+
+ /* X86_TUNE_AVX256_SPLIT_UNALIGNED_LOAD_OPTIMAL: Enable splitting 256-bit
+ unaligned load. It hurts the performance on Bulldozer. We need to
+ re-tune the generic options for current cpus! */
+ m_COREI7 | m_GENERIC,
+
+ /* X86_TUNE_AVX256_SPLIT_UNALIGNED_STORE_OPTIMAL: Enable splitting 256-bit
+ unaligned store. */
+ m_COREI7 | m_BDVER1 | m_GENERIC
};
/* Feature tests against the various architecture variations. */
@@ -4194,9 +4203,11 @@ ix86_option_override_internal (bool main_args_p)
if (flag_expensive_optimizations
&& !(target_flags_explicit & MASK_VZEROUPPER))
target_flags |= MASK_VZEROUPPER;
- if (!(target_flags_explicit & MASK_AVX256_SPLIT_UNALIGNED_LOAD))
+ if (TARGET_AVX256_SPLIT_UNALIGNED_LOAD_OPTIMAL
+ && !(target_flags_explicit & MASK_AVX256_SPLIT_UNALIGNED_LOAD))
target_flags |= MASK_AVX256_SPLIT_UNALIGNED_LOAD;
- if (!(target_flags_explicit & MASK_AVX256_SPLIT_UNALIGNED_STORE))
+ if (TARGET_AVX256_SPLIT_UNALIGNED_STORE_OPTIMAL
+ && !(target_flags_explicit & MASK_AVX256_SPLIT_UNALIGNED_STORE))
target_flags |= MASK_AVX256_SPLIT_UNALIGNED_STORE;
}
}
diff --git a/gcc/config/i386/i386.h b/gcc/config/i386/i386.h
index 8badcbb..b6e5570 100644
--- a/gcc/config/i386/i386.h
+++ b/gcc/config/i386/i386.h
@@ -312,6 +312,8 @@ enum ix86_tune_indices {
X86_TUNE_OPT_AGU,
X86_TUNE_VECTORIZE_DOUBLE,
X86_TUNE_SOFTWARE_PREFETCHING_BENEFICIAL,
+ X86_TUNE_AVX256_SPLIT_UNALIGNED_LOAD_OPTIMAL,
+ X86_TUNE_AVX256_SPLIT_UNALIGNED_STORE_OPTIMAL,
X86_TUNE_LAST
};
@@ -410,6 +412,8 @@ extern unsigned char ix86_tune_features[X86_TUNE_LAST];
ix86_tune_features[X86_TUNE_VECTORIZE_DOUBLE]
#define TARGET_SOFTWARE_PREFETCHING_BENEFICIAL \
ix86_tune_features[X86_TUNE_SOFTWARE_PREFETCHING_BENEFICIAL]
+#define TARGET_AVX256_SPLIT_UNALIGNED_LOAD_OPTIMAL \
+ ix86_tune_features[X86_TUNE_AVX256_SPLIT_UNALIGNED_LOAD_OPTIMAL]
/* Feature tests against the various architecture variations. */
enum ix86_arch_indices {
--
1.7.0.4
