6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2025-06-23 17:24:48 +00:00
Code Issues Projects Releases Wiki Activity

LoopVectorize: We don't need an identity element for min/max reductions

Browse Source 
We can just use the initial element that feeds the reduction.

  max(max(x, y), z) == max(max(x,y), max(x,z))

radar://13723044

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181141 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Arnold Schwaighofer
2013-05-05 01:54:42 +00:00
parent 5c332dbd30
commit c1738fdadd
2 changed files with 22 additions and 32 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
test/Transforms/LoopVectorize/minmax_reduction.ll
View File

@ -6,8 +6,11 @@ target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f3
; Signed tests.
; Turn this into a max reduction.
; Turn this into a max reduction. Make sure we use a splat to initialize the
; vector for the reduction.
; CHECK: @max_red
; CHECK: %[[VAR:.*]] = insertelement <2 x i32> undef, i32 %max, i32 0
; CHECK: {{.*}} = shufflevector <2 x i32> %[[VAR]], <2 x i32> undef, <2 x i32> zeroinitializer
; CHECK: icmp sgt <2 x i32>
; CHECK: select <2 x i1>
; CHECK: middle.block