[PATCH] [1/2] Add PRFCHW feature to a few AMD processors

[Yunzhong Gao]

ygao added you to the CC list for the revision "[1/2] Add PRFCHW feature to a few AMD processors".

Hi,
This patch enables the 3dNow! prefetch instruction for a few AMD processors: bobcat, jaguar, bulldozer and piledriver. Support for the instruction itself seems to have already been added in r178040. Could someone take a look whether this is good to go in?
Thanks,
- Gao.

http://llvm-reviews.chandlerc.com/D1933

Files:
  test/CodeGen/X86/prefetch.ll
  lib/Target/X86/X86.td

Index: test/CodeGen/X86/prefetch.ll
===================================================================
--- test/CodeGen/X86/prefetch.ll
+++ test/CodeGen/X86/prefetch.ll
@@ -2,6 +2,8 @@
 ; RUN: llc < %s -march=x86 -mattr=+avx | FileCheck %s
 ; RUN: llc < %s -march=x86 -mattr=+sse -mattr=+prfchw | FileCheck %s -check-prefix=PRFCHW
 ; RUN: llc < %s -march=x86 -mcpu=slm | FileCheck %s -check-prefix=SLM
+; RUN: llc < %s -march=x86 -mcpu=btver2 | FileCheck %s -check-prefix=PRFCHW
+; RUN: llc < %s -march=x86 -mcpu=btver2 -mattr=-prfchw | FileCheck %s -check-prefix=NOPRFCHW
 
 ; rdar://10538297
 
@@ -12,6 +14,7 @@
 ; CHECK: prefetcht0
 ; CHECK: prefetchnta
 ; PRFCHW: prefetchw
+; NOPRFCHW-NOT: prefetchw
 ; SLM: prefetchw
 	tail call void @llvm.prefetch( i8* %ptr, i32 0, i32 1, i32 1 )
 	tail call void @llvm.prefetch( i8* %ptr, i32 0, i32 2, i32 1 )
Index: lib/Target/X86/X86.td
===================================================================
--- lib/Target/X86/X86.td
+++ lib/Target/X86/X86.td
@@ -292,19 +292,19 @@
                                FeaturePOPCNT, FeatureSlowBTMem]>;
 // Bobcat
 def : Proc<"btver1",          [FeatureSSSE3, FeatureSSE4A, FeatureCMPXCHG16B,
-                               FeatureLZCNT, FeaturePOPCNT]>;
+                               FeaturePRFCHW, FeatureLZCNT, FeaturePOPCNT]>;
 // Jaguar
 def : Proc<"btver2",          [FeatureAVX, FeatureSSE4A, FeatureCMPXCHG16B,
-                               FeatureAES, FeaturePCLMUL, FeatureBMI,
-                               FeatureF16C, FeatureMOVBE, FeatureLZCNT,
-                               FeaturePOPCNT]>;
+                               FeaturePRFCHW, FeatureAES, FeaturePCLMUL,
+                               FeatureBMI, FeatureF16C, FeatureMOVBE,
+                               FeatureLZCNT, FeaturePOPCNT]>;
 // Bulldozer
 def : Proc<"bdver1",          [FeatureXOP, FeatureFMA4, FeatureCMPXCHG16B,
-                               FeatureAES, FeaturePCLMUL,
+                               FeatureAES, FeaturePRFCHW, FeaturePCLMUL,
                                FeatureLZCNT, FeaturePOPCNT]>;
 // Piledriver
 def : Proc<"bdver2",          [FeatureXOP, FeatureFMA4, FeatureCMPXCHG16B,
-                               FeatureAES, FeaturePCLMUL,
+                               FeatureAES, FeaturePRFCHW, FeaturePCLMUL,
                                FeatureF16C, FeatureLZCNT,
                                FeaturePOPCNT, FeatureBMI,  FeatureTBM,
                                FeatureFMA]>;
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D1933.1.patch
Type: text/x-patch
Size: 2499 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20131014/72ed07f2/attachment.bin>