6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2025-10-28 17:23:27 +00:00
Code Issues Projects Releases Wiki Activity

BBVectorize: Correctly merge SubclassOptionalData

Browse Source 
When two instructions are combined into a vector instruction,
the resulting instruction must have the most-conservative flags.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@168765 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Hal Finkel
2012-11-28 03:04:10 +00:00
parent 6acb53152c
commit 430b9079c6
2 changed files with 25 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
lib/Transforms/Vectorize/BBVectorize.cpp
View File

@@ -2903,6 +2903,7 @@ namespace {
K->mutateType(getVecTypeForPair(L->getType(), H->getType())); K->mutateType(getVecTypeForPair(L->getType(), H->getType()));
combineMetadata(K, H); combineMetadata(K, H);
K->intersectOptionalDataWith(H);
for (unsigned o = 0; o < NumOperands; ++o) for (unsigned o = 0; o < NumOperands; ++o)
K->setOperand(o, ReplacedOperands[o]); K->setOperand(o, ReplacedOperands[o]);

24
test/Transforms/BBVectorize/simple.ll
View File

@@ -173,3 +173,27 @@ define double @test7(double %A1, double %A2, double %B1, double %B2) {
; CHECK: ret double %R ; CHECK: ret double %R
} }
; Basic depth-3 chain (subclass data)
define i64 @test8(i64 %A1, i64 %A2, i64 %B1, i64 %B2) {
; CHECK: @test8
; CHECK: %X1.v.i1.1 = insertelement <2 x i64> undef, i64 %B1, i32 0
; CHECK: %X1.v.i1.2 = insertelement <2 x i64> %X1.v.i1.1, i64 %B2, i32 1
; CHECK: %X1.v.i0.1 = insertelement <2 x i64> undef, i64 %A1, i32 0
; CHECK: %X1.v.i0.2 = insertelement <2 x i64> %X1.v.i0.1, i64 %A2, i32 1
%X1 = sub nsw i64 %A1, %B1
%X2 = sub i64 %A2, %B2
; CHECK: %X1 = sub <2 x i64> %X1.v.i0.2, %X1.v.i1.2
%Y1 = mul i64 %X1, %A1
%Y2 = mul i64 %X2, %A2
; CHECK: %Y1 = mul <2 x i64> %X1, %X1.v.i0.2
%Z1 = add i64 %Y1, %B1
%Z2 = add i64 %Y2, %B2
; CHECK: %Z1 = add <2 x i64> %Y1, %X1.v.i1.2
%R = mul i64 %Z1, %Z2
; CHECK: %Z1.v.r1 = extractelement <2 x i64> %Z1, i32 0
; CHECK: %Z1.v.r2 = extractelement <2 x i64> %Z1, i32 1
; CHECK: %R = mul i64 %Z1.v.r1, %Z1.v.r2
ret i64 %R
; CHECK: ret i64 %R
}