6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2025-06-14 14:24:05 +00:00
Code Issues Projects Releases Wiki Activity

SLPVectorizer: Only vectorize intrinsics whose operands are widened equally

Browse Source 
The vectorizer only knows how to vectorize intrinics by widening all operands by
the same factor.

Patch by Tyler Nowicki!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205855 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Arnold Schwaighofer
2014-04-09 14:20:47 +00:00
parent 0d5d656524
commit b0ee2374ce
4 changed files with 100 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files

36
test/Transforms/SLPVectorizer/X86/non-vectorizable-intrinsic.ll Normal file
View File

@ -0,0 +1,36 @@
; RUN: opt < %s -slp-vectorizer -o - -S -slp-threshold=-1000
target datalayout = "e-p:32:32-i64:64-v16:16-v32:32-n16:32:64"
target triple = "nvptx--nvidiacl"
; CTLZ cannot be vectorized currently because the second argument is a scalar
; for both the scalar and vector forms of the intrinsic. In the future it
; should be possible to vectorize such functions.
; Test causes an assert if LLVM tries to vectorize CTLZ.
define <2 x i8> @cltz_test(<2 x i8> %x) #0 {
entry:
%0 = extractelement <2 x i8> %x, i32 0
%call.i = call i8 @llvm.ctlz.i8(i8 %0, i1 false)
%vecinit = insertelement <2 x i8> undef, i8 %call.i, i32 0
%1 = extractelement <2 x i8> %x, i32 1
%call.i4 = call i8 @llvm.ctlz.i8(i8 %1, i1 false)
%vecinit2 = insertelement <2 x i8> %vecinit, i8 %call.i4, i32 1
ret <2 x i8> %vecinit2
}
define <2 x i8> @cltz_test2(<2 x i8> %x) #1 {
entry:
%0 = extractelement <2 x i8> %x, i32 0
%1 = extractelement <2 x i8> %x, i32 1
%call.i = call i8 @llvm.ctlz.i8(i8 %0, i1 false)
%call.i4 = call i8 @llvm.ctlz.i8(i8 %1, i1 false)
%vecinit = insertelement <2 x i8> undef, i8 %call.i, i32 0
%vecinit2 = insertelement <2 x i8> %vecinit, i8 %call.i4, i32 1
ret <2 x i8> %vecinit2
}
declare i8 @llvm.ctlz.i8(i8, i1) #3
attributes #0 = { alwaysinline nounwind "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #1 = { nounwind readnone }