6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2024-12-13 20:32:21 +00:00
Code Issues Projects Releases Wiki Activity

Reverse order of checking SSE level when calculating compare cost, so we check

Browse Source 
AVX2 before AVX.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@170464 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Jakub Staszak 2012-12-18 22:57:56 +00:00
parent 9466bdebb7
commit 270bfbd3d1
2 changed files with 34 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
lib/Target/X86/X86ISelLowering.cpp
View File

@ -17866,10 +17866,10 @@ unsigned X86VectorTargetTransformInfo::getCmpSelInstrCost(unsigned Opcode,
{ ISD::SETCC, MVT::v32i8, 1 }, { ISD::SETCC, MVT::v32i8, 1 },
}; };
if (ST.hasSSE42()) { if (ST.hasAVX2()) {
int Idx = FindInTable(SSE42CostTbl, array_lengthof(SSE42CostTbl), ISD, MTy); int Idx = FindInTable(AVX2CostTbl, array_lengthof(AVX2CostTbl), ISD, MTy);
if (Idx != -1) if (Idx != -1)
return LT.first * SSE42CostTbl[Idx].Cost; return LT.first * AVX2CostTbl[Idx].Cost;
} }
if (ST.hasAVX()) { if (ST.hasAVX()) {
@ -17878,10 +17878,10 @@ unsigned X86VectorTargetTransformInfo::getCmpSelInstrCost(unsigned Opcode,
return LT.first * AVX1CostTbl[Idx].Cost; return LT.first * AVX1CostTbl[Idx].Cost;
} }
if (ST.hasAVX2()) { if (ST.hasSSE42()) {
int Idx = FindInTable(AVX2CostTbl, array_lengthof(AVX2CostTbl), ISD, MTy); int Idx = FindInTable(SSE42CostTbl, array_lengthof(SSE42CostTbl), ISD, MTy);
if (Idx != -1) if (Idx != -1)
return LT.first * AVX2CostTbl[Idx].Cost; return LT.first * SSE42CostTbl[Idx].Cost;
} }
return VectorTargetTransformImpl::getCmpSelInstrCost(Opcode, ValTy, CondTy); return VectorTargetTransformImpl::getCmpSelInstrCost(Opcode, ValTy, CondTy);

42
test/Analysis/CostModel/X86/cmp.ll
View File

@ -1,38 +1,52 @@
; RUN: opt < %s -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s ; RUN: opt < %s -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck --check-prefix=AVX1 %s
; RUN: opt < %s -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mcpu=core-avx2 | FileCheck --check-prefix=AVX2 %s
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128" target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
target triple = "x86_64-apple-macosx10.8.0" target triple = "x86_64-apple-macosx10.8.0"
define i32 @cmp(i32 %arg) { define i32 @cmp(i32 %arg) {
; -- floats -- ; -- floats --
;CHECK: cost of 1 {{.*}} fcmp ;AVX1: cost of 1 {{.*}} fcmp
;AVX2: cost of 1 {{.*}} fcmp
%A = fcmp olt <2 x float> undef, undef %A = fcmp olt <2 x float> undef, undef
;CHECK: cost of 1 {{.*}} fcmp ;AVX1: cost of 1 {{.*}} fcmp
;AVX2: cost of 1 {{.*}} fcmp
%B = fcmp olt <4 x float> undef, undef %B = fcmp olt <4 x float> undef, undef
;CHECK: cost of 1 {{.*}} fcmp ;AVX1: cost of 1 {{.*}} fcmp
;AVX2: cost of 1 {{.*}} fcmp
%C = fcmp olt <8 x float> undef, undef %C = fcmp olt <8 x float> undef, undef
;CHECK: cost of 1 {{.*}} fcmp ;AVX1: cost of 1 {{.*}} fcmp
;AVX2: cost of 1 {{.*}} fcmp
%D = fcmp olt <2 x double> undef, undef %D = fcmp olt <2 x double> undef, undef
;CHECK: cost of 1 {{.*}} fcmp ;AVX1: cost of 1 {{.*}} fcmp
;AVX2: cost of 1 {{.*}} fcmp
%E = fcmp olt <4 x double> undef, undef %E = fcmp olt <4 x double> undef, undef
; -- integers -- ; -- integers --
;CHECK: cost of 1 {{.*}} icmp ;AVX1: cost of 1 {{.*}} icmp
;AVX2: cost of 1 {{.*}} icmp
%F = icmp eq <16 x i8> undef, undef %F = icmp eq <16 x i8> undef, undef
;CHECK: cost of 1 {{.*}} icmp ;AVX1: cost of 1 {{.*}} icmp
;AVX2: cost of 1 {{.*}} icmp
%G = icmp eq <8 x i16> undef, undef %G = icmp eq <8 x i16> undef, undef
;CHECK: cost of 1 {{.*}} icmp ;AVX1: cost of 1 {{.*}} icmp
;AVX2: cost of 1 {{.*}} icmp
%H = icmp eq <4 x i32> undef, undef %H = icmp eq <4 x i32> undef, undef
;CHECK: cost of 1 {{.*}} icmp ;AVX1: cost of 1 {{.*}} icmp
;AVX2: cost of 1 {{.*}} icmp
%I = icmp eq <2 x i64> undef, undef %I = icmp eq <2 x i64> undef, undef
;CHECK: cost of 4 {{.*}} icmp ;AVX1: cost of 4 {{.*}} icmp
;AVX2: cost of 1 {{.*}} icmp
%J = icmp eq <4 x i64> undef, undef %J = icmp eq <4 x i64> undef, undef
;CHECK: cost of 4 {{.*}} icmp ;AVX1: cost of 4 {{.*}} icmp
;AVX2: cost of 1 {{.*}} icmp
%K = icmp eq <8 x i32> undef, undef %K = icmp eq <8 x i32> undef, undef
;CHECK: cost of 4 {{.*}} icmp ;AVX1: cost of 4 {{.*}} icmp
;AVX2: cost of 1 {{.*}} icmp
%L = icmp eq <16 x i16> undef, undef %L = icmp eq <16 x i16> undef, undef
;CHECK: cost of 4 {{.*}} icmp ;AVX1: cost of 4 {{.*}} icmp
;AVX2: cost of 1 {{.*}} icmp
%M = icmp eq <32 x i8> undef, undef %M = icmp eq <32 x i8> undef, undef
;CHECK: cost of 0 {{.*}} ret ;CHECK: cost of 0 {{.*}} ret