replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[SLPVectorizer] Don't ignore scalar extraction instructions of aggregate value 

In SLPVectorizer, the vector build instructions (insertvalue for aggregate type) is passed to BoUpSLP.buildTree, it is treated as UserIgnoreList, so later in cost estimation, the cost of these instructions are not counted. 
For aggregate value, later usage are more likely to be done in scalar registers, either used as individual scalars or used as a whole for function call or return value. Ignore scalar extraction instructions may cause too aggressive vectorization for aggregate values, and slow down performance. So for vectorization of aggregate value, the scalar extraction instructions are required in cost estimation.

Differential Revision: https://reviews.llvm.org/D41139

llvm-svn: 320736
This commit is contained in:
Guozhi Wei 2017-12-14 19:35:43 +00:00
parent 2cfa11a524
commit d22d1b953d
4 changed files with 49 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
llvm/include/llvm/Transforms/Vectorize/SLPVectorizer.h
View File

@ -96,11 +96,13 @@ private:
/// \brief Try to vectorize a list of operands.
/// \@param BuildVector A list of users to ignore for the purpose of
/// scheduling and that don't need extracting.
/// scheduling and cost estimation when NeedExtraction
/// is false.
/// \returns true if a value was vectorized.
bool tryToVectorizeList(ArrayRef<Value *> VL, slpvectorizer::BoUpSLP &R,
ArrayRef<Value *> BuildVector = None,
bool AllowReorder = false);
bool AllowReorder = false,
bool NeedExtraction = false);
/// \brief Try to vectorize a chain that may start at the operands of \p I.
bool tryToVectorize(Instruction *I, slpvectorizer::BoUpSLP &R);

10
llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
View File

@ -4533,7 +4533,8 @@ bool SLPVectorizerPass::tryToVectorizePair(Value *A, Value *B, BoUpSLP &R) {
bool SLPVectorizerPass::tryToVectorizeList(ArrayRef<Value *> VL, BoUpSLP &R,
ArrayRef<Value *> BuildVector,
bool AllowReorder) {
bool AllowReorder,
bool NeedExtraction) {
if (VL.size() < 2)
return false;
@ -4627,11 +4628,12 @@ bool SLPVectorizerPass::tryToVectorizeList(ArrayRef<Value *> VL, BoUpSLP &R,
<< "\n");
ArrayRef<Value *> Ops = VL.slice(I, OpsWidth);
ArrayRef<Value *> EmptyArray;
ArrayRef<Value *> BuildVectorSlice;
if (!BuildVector.empty())
BuildVectorSlice = BuildVector.slice(I, OpsWidth);
R.buildTree(Ops, BuildVectorSlice);
R.buildTree(Ops, NeedExtraction ? EmptyArray : BuildVectorSlice);
// TODO: check if we can allow reordering for more cases.
if (AllowReorder && R.shouldReorder()) {
// Conceptually, there is nothing actually preventing us from trying to
@ -5821,7 +5823,9 @@ bool SLPVectorizerPass::vectorizeInsertValueInst(InsertValueInst *IVI,
return false;
DEBUG(dbgs() << "SLP: array mappable to vector: " << *IVI << "\n");
return tryToVectorizeList(BuildVectorOpds, R, BuildVector, false);
// Aggregate value is unlikely to be processed in vector register, we need to
// extract scalars into scalar registers, so NeedExtraction is set true.
return tryToVectorizeList(BuildVectorOpds, R, BuildVector, false, true);
}
bool SLPVectorizerPass::vectorizeInsertElementInst(InsertElementInst *IEI,

19
llvm/test/Transforms/SLPVectorizer/PowerPC/aggregate.ll Normal file
View File

@ -0,0 +1,19 @@
; RUN: opt -S -mtriple=powerpc64-linux-gnu -mcpu=pwr9 -mattr=+vsx -slp-vectorizer < %s | FileCheck %s
%struct.S = type { i8*, i8* }
@kS0 = common global %struct.S zeroinitializer, align 8
define { i64, i64 } @getS() {
entry:
%0 = load i64, i64* bitcast (%struct.S* @kS0 to i64*), align 8
%1 = load i64, i64* bitcast (i8** getelementptr inbounds (%struct.S, %struct.S* @kS0, i64 0, i32 1) to i64*), align 8
%2 = insertvalue { i64, i64 } undef, i64 %0, 0
%3 = insertvalue { i64, i64 } %2, i64 %1, 1
ret { i64, i64 } %3
}
; CHECK: load i64
; CHECK-NOT: load <2 x i64>
; CHECK-NOT: extractelement

19
llvm/test/Transforms/SLPVectorizer/X86/aggregate.ll Normal file
View File

@ -0,0 +1,19 @@
; RUN: opt -S -mtriple=x86_64-unknown-linux -mcpu=corei7 -slp-vectorizer < %s | FileCheck %s
%struct.S = type { i8*, i8* }
@kS0 = common global %struct.S zeroinitializer, align 8
define { i64, i64 } @getS() {
entry:
%0 = load i64, i64* bitcast (%struct.S* @kS0 to i64*), align 8
%1 = load i64, i64* bitcast (i8** getelementptr inbounds (%struct.S, %struct.S* @kS0, i64 0, i32 1) to i64*), align 8
%2 = insertvalue { i64, i64 } undef, i64 %0, 0
%3 = insertvalue { i64, i64 } %2, i64 %1, 1
ret { i64, i64 } %3
}
; CHECK: load i64
; CHECK-NOT: load <2 x i64>
; CHECK-NOT: extractelement