RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-24 11:42:57 +01:00
Code Issues Projects Releases Wiki Activity

SLPVectorize: Put horizontal reductions feeding a store under separate flag

Browse Source 
Put them under a separate flag for experimentation. They are more likely to
interfere with loop vectorization which happens later in the pass pipeline.

llvm-svn: 191371
This commit is contained in:
Arnold Schwaighofer 2013-09-25 14:02:32 +00:00
parent 43a9297729
commit 9830391a8f
2 changed files with 28 additions and 21 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
lib/Transforms/Vectorize/SLPVectorizer.cpp
View File

@ -54,6 +54,11 @@ static cl::opt<bool>
ShouldVectorizeHor("slp-vectorize-hor", cl::init(false), cl::Hidden,
cl::desc("Attempt to vectorize horizontal reductions"));
static cl::opt<bool> ShouldStartVectorizeHorAtStore(
"slp-vectorize-hor-store", cl::init(false), cl::Hidden,
cl::desc(
"Attempt to vectorize horizontal reductions feeding into a store"));
namespace {
static const unsigned MinVecRegSize = 128;
@ -2336,12 +2341,12 @@ bool SLPVectorizer::vectorizeChainsInBlock(BasicBlock *BB, BoUpSLP &R) {
}
// Try to vectorize horizontal reductions feeding into a store.
if (ShouldStartVectorizeHorAtStore)
if (StoreInst *SI = dyn_cast<StoreInst>(it))
if (BinaryOperator *BinOp =
dyn_cast<BinaryOperator>(SI->getValueOperand())) {
HorizontalReduction HorRdx;
if (ShouldVectorizeHor &&
((HorRdx.matchAssociativeReduction(0, BinOp, DL) &&
if (((HorRdx.matchAssociativeReduction(0, BinOp, DL) &&
HorRdx.tryToReduce(R, TTI)) ||
tryToVectorize(BinOp, R))) {
Changed = true;

18
test/Transforms/SLPVectorizer/X86/horizontal.ll
View File

@ -1,4 +1,4 @@
; RUN: opt -slp-vectorizer -slp-vectorize-hor -S < %s -mtriple=x86_64-apple-macosx -mcpu=corei7-avx | FileCheck %s
; RUN: opt -slp-vectorizer -slp-vectorize-hor -S < %s -mtriple=x86_64-apple-macosx -mcpu=corei7-avx | FileCheck %s --check-prefix=NOSTORE
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"
@ -15,9 +15,9 @@ target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f3
; return sum;
; }
; CHECK-LABEL: add_red
; CHECK: fmul <4 x float>
; CHECK: shufflevector <4 x float>
; NOSTORE-LABEL: add_red
; NOSTORE: fmul <4 x float>
; NOSTORE: shufflevector <4 x float>
define i32 @add_red(float* %A, i32 %n) {
entry:
@ -369,6 +369,8 @@ for.end:
}
; RUN: opt -slp-vectorizer -slp-vectorize-hor -slp-vectorize-hor-store -S < %s -mtriple=x86_64-apple-macosx -mcpu=corei7-avx | FileCheck %s --check-prefix=STORE
; void foo(double * restrict A, double * restrict B, double * restrict C,
; int n) {
; for (intptr_t i=0; i < n; ++i) {
@ -376,10 +378,10 @@ for.end:
; }
; }
; CHECK-LABEL: store_red_double
; CHECK: fmul <2 x double>
; CHECK: extractelement <2 x double>
; CHECK: extractelement <2 x double>
; STORE-LABEL: store_red_double
; STORE: fmul <2 x double>
; STORE: extractelement <2 x double>
; STORE: extractelement <2 x double>
define void @store_red_double(double* noalias %A, double* noalias %B, double* noalias %C, i32 %n) {
entry: