• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1; RUN: opt < %s -slp-vectorizer -o - -S -slp-threshold=-1000
2
3target datalayout = "e-p:32:32-i64:64-v16:16-v32:32-n16:32:64"
4target triple = "nvptx--nvidiacl"
5
6; CTLZ cannot be vectorized currently because the second argument is a scalar
7; for both the scalar and vector forms of the intrinsic. In the future it
8; should be possible to vectorize such functions.
9; Test causes an assert if LLVM tries to vectorize CTLZ.
10
11define <2 x i8> @cltz_test(<2 x i8> %x) #0 {
12entry:
13  %0 = extractelement <2 x i8> %x, i32 0
14  %call.i = call i8 @llvm.ctlz.i8(i8 %0, i1 false)
15  %vecinit = insertelement <2 x i8> undef, i8 %call.i, i32 0
16  %1 = extractelement <2 x i8> %x, i32 1
17  %call.i4 = call i8 @llvm.ctlz.i8(i8 %1, i1 false)
18  %vecinit2 = insertelement <2 x i8> %vecinit, i8 %call.i4, i32 1
19  ret <2 x i8> %vecinit2
20}
21
22define <2 x i8> @cltz_test2(<2 x i8> %x) #1 {
23entry:
24  %0 = extractelement <2 x i8> %x, i32 0
25  %1 = extractelement <2 x i8> %x, i32 1
26  %call.i = call i8 @llvm.ctlz.i8(i8 %0, i1 false)
27  %call.i4 = call i8 @llvm.ctlz.i8(i8 %1, i1 false)
28  %vecinit = insertelement <2 x i8> undef, i8 %call.i, i32 0
29  %vecinit2 = insertelement <2 x i8> %vecinit, i8 %call.i4, i32 1
30  ret <2 x i8> %vecinit2
31}
32
33declare i8 @llvm.ctlz.i8(i8, i1) #3
34
35attributes #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" }
36attributes #1 = { nounwind readnone }
37