6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2024-12-30 17:33:24 +00:00
Code Issues Projects Releases Wiki Activity

InstCombine now optimizes vector udiv by power of 2 to shifts

Browse Source 
Fixes r8429


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@144036 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Pete Cooper 2011-11-07 23:04:49 +00:00
parent 1b4f6f2532
commit a29fc806fe
2 changed files with 24 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
View File

@ -442,18 +442,22 @@ Instruction *InstCombiner::visitUDiv(BinaryOperator &I) {
if (Instruction *Common = commonIDivTransforms(I))
return Common;
if (ConstantInt *C = dyn_cast<ConstantInt>(Op1)) {
{
// X udiv 2^C -> X >> C
// Check to see if this is an unsigned division with an exact power of 2,
// if so, convert to a right shift.
if (C->getValue().isPowerOf2()) { // 0 not included in isPowerOf2
const APInt *C;
if (match(Op1, m_Power2(C))) {
BinaryOperator *LShr =
BinaryOperator::CreateLShr(Op0,
ConstantInt::get(Op0->getType(), C->getValue().logBase2()));
BinaryOperator::CreateLShr(Op0,
ConstantInt::get(Op0->getType(),
C->logBase2()));
if (I.isExact()) LShr->setIsExact();
return LShr;
}
}
if (ConstantInt *C = dyn_cast<ConstantInt>(Op1)) {
// X udiv C, where C >= signbit
if (C->getValue().isNegative()) {
Value *IC = Builder->CreateICmpULT(Op0, C);

15
test/CodeGen/X86/vec_udiv_to_shift.ll Normal file
View File

@ -0,0 +1,15 @@
; RUN: opt < %s -instcombine -S | FileCheck %s
define <8 x i16> @udiv_vec8x16(<8 x i16> %var) {
entry:
; CHECK: lshr <8 x i16> %var, <i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5>
%0 = udiv <8 x i16> %var, <i16 32, i16 32, i16 32, i16 32, i16 32, i16 32, i16 32, i16 32>
ret <8 x i16> %0
}
define <4 x i32> @udiv_vec4x32(<4 x i32> %var) {
entry:
; CHECK: lshr <4 x i32> %var, <i32 4, i32 4, i32 4, i32 4>
%0 = udiv <4 x i32> %var, <i32 16, i32 16, i32 16, i32 16>
ret <4 x i32> %0
}