diff mbox

[PR,49089] Don't split AVX256 unaligned loads by default on bdver1 and generic

Message ID D4C76825A6780047854A11E93CDE84D005980DC6FA@SAUSEXMBP01.amd.com
State New
Headers show

Commit Message

Fang, Changpeng June 17, 2011, 10:18 p.m. UTC 
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

Comments

H.J. Lu June 17, 2011, 10:44 p.m. UTC | #1 
On Fri, Jun 17, 2011 at 3:18 PM, Fang, Changpeng <Changpeng.Fang@amd.com> wrote:
> 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?

Can you do something similar to how MASK_ACCUMULATE_OUTGOING_ARGS
is handled?

Thanks.

H.J.
diff mbox

Patch

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