RPCS3/llvm
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm.git synced 2024-11-24 20:30:06 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

CostModel: add support for Vector Insert and Extract.

Browse Source 
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167329 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Nadav Rotem 2012-11-02 22:31:56 +00:00
parent 4c8edb41e5
commit f149567160
3 changed files with 63 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
include/llvm/TargetTransformInfo.h
View File

@ -179,8 +179,9 @@ public:
}
/// Returns the expected cost of vector Insert and Extract.
/// Use -1 to indicate that there is no information on the index value.
virtual unsigned getVectorInstrCost(unsigned Opcode, Type *Val,
unsigned Index = 0) const {
unsigned Index = -1) const {
return 1;
}

18
lib/Analysis/CostModel.cpp
View File

@ -150,6 +150,24 @@ unsigned CostModelAnalysis::getInstructionCost(Instruction *I) const {
Type *SrcTy = I->getOperand(0)->getType();
return VTTI->getCastInstrCost(I->getOpcode(), I->getType(), SrcTy);
}
case Instruction::ExtractElement: {
ExtractElementInst * EEI = cast<ExtractElementInst>(I);
ConstantInt *CI = dyn_cast<ConstantInt>(I->getOperand(1));
unsigned Idx = -1;
if (CI)
Idx = CI->getZExtValue();
return VTTI->getVectorInstrCost(I->getOpcode(),
EEI->getOperand(0)->getType(), Idx);
}
case Instruction::InsertElement: {
InsertElementInst * IE = cast<InsertElementInst>(I);
ConstantInt *CI = dyn_cast<ConstantInt>(IE->getOperand(2));
unsigned Idx = -1;
if (CI)
Idx = CI->getZExtValue();
return VTTI->getVectorInstrCost(I->getOpcode(),
IE->getType(), Idx);
}
default:
// We don't have any information on this instruction.
return -1;

43
test/Analysis/CostModel/X86/loop_v2.ll Normal file
View File

@ -0,0 +1,43 @@
; RUN: opt < %s -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %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 triple = "x86_64-apple-macosx10.9.0"
define i32 @foo(i32* nocapture %A) nounwind uwtable readonly ssp {
vector.ph:
br label %vector.body
vector.body: ; preds = %vector.body, %vector.ph
%index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
%vec.phi = phi <2 x i32> [ zeroinitializer, %vector.ph ], [ %12, %vector.body ]
%0 = getelementptr inbounds i32* %A, i64 %index
%1 = bitcast i32* %0 to <2 x i32>*
%2 = load <2 x i32>* %1, align 4
%3 = sext <2 x i32> %2 to <2 x i64>
;CHECK: cost of 1 {{.*}} extract
%4 = extractelement <2 x i64> %3, i32 0
%5 = getelementptr inbounds i32* %A, i64 %4
;CHECK: cost of 1 {{.*}} extract
%6 = extractelement <2 x i64> %3, i32 1
%7 = getelementptr inbounds i32* %A, i64 %6
%8 = load i32* %5, align 4, !tbaa !0
;CHECK: cost of 1 {{.*}} insert
%9 = insertelement <2 x i32> undef, i32 %8, i32 0
%10 = load i32* %7, align 4, !tbaa !0
;CHECK: cost of 1 {{.*}} insert
%11 = insertelement <2 x i32> %9, i32 %10, i32 1
%12 = add nsw <2 x i32> %11, %vec.phi
%index.next = add i64 %index, 2
%13 = icmp eq i64 %index.next, 192
br i1 %13, label %for.end, label %vector.body
for.end: ; preds = %vector.body
%14 = extractelement <2 x i32> %12, i32 0
%15 = extractelement <2 x i32> %12, i32 1
%16 = add i32 %14, %15
ret i32 %16
}
!0 = metadata !{metadata !"int", metadata !1}
!1 = metadata !{metadata !"omnipotent char", metadata !2}
!2 = metadata !{metadata !"Simple C/C++ TBAA"}