RPCS3/llvm
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm.git synced 2026-07-19 23:23:38 -04:00
Code Issues Packages Projects Releases 10 Wiki Activity

[LSV] Insert stores at the right point.

Browse Source 
Summary:
Previously, the insertion point for stores was the last instruction in
Chain *before calling getVectorizablePrefixEndIdx*.  Thus if
getVectorizablePrefixEndIdx didn't return Chain.size(), we still would
insert at the last instruction in Chain.

This patch changes our internal API a bit in an attempt to make it less
prone to this sort of error.  As a result, we end up recalculating the
Chain's boundary instructions, but I think worrying about the speed hit
of this is a premature optimization right now.

Reviewers: asbirlea, tstellarAMD

Subscribers: mzolotukhin, arsenm, llvm-commits

Differential Revision: https://reviews.llvm.org/D22534

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@276056 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Justin Lebar
2016-07-19 23:19:20 +00:00
parent abcf9144a5
commit 85ee7e9487
2 changed files with 57 additions and 32 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/Transforms/LoadStoreVectorizer/AMDGPU/insertion-point.ll
+29 -2
View File
@@ -2,8 +2,9 @@
target datalayout = "e-p:32:32-p1:64:64-p2:64:64-p3:32:32-p4:64:64-p5:32:32-p24:64:64-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64"
; Check relative position of the inserted vector load relative to the existing
; adds. Vectorized loads should be inserted at the position of the first load.
; Check position of the inserted vector load/store. Vectorized loads should be
; inserted at the position of the first load in the chain, and stores should be
; inserted at the position of the last store.
; CHECK-LABEL: @insert_load_point(
; CHECK: %z = add i32 %x, 4
@@ -59,4 +60,30 @@ entry:
ret void
}
; Here we have four stores, with an aliasing load before the last one. We can
; vectorize the first two stores as <2 x float>, but this vectorized store must
; be inserted at the location of the second scalar store, not the fourth one.
;
; CHECK-LABEL: @insert_store_point_alias
; CHECK: store <2 x float>
; CHECK: store float
; CHECK-SAME: %a.idx.2
; CHECK: load float, float addrspace(1)* %a.idx.2
; CHECK: store float
; CHECK-SAME: %a.idx.3
define float @insert_store_point_alias(float addrspace(1)* nocapture %a, i64 %idx) {
%a.idx = getelementptr inbounds float, float addrspace(1)* %a, i64 %idx
%a.idx.1 = getelementptr inbounds float, float addrspace(1)* %a.idx, i64 1
%a.idx.2 = getelementptr inbounds float, float addrspace(1)* %a.idx.1, i64 1
%a.idx.3 = getelementptr inbounds float, float addrspace(1)* %a.idx.2, i64 1
store float 0.0, float addrspace(1)* %a.idx, align 4
store float 0.0, float addrspace(1)* %a.idx.1, align 4
store float 0.0, float addrspace(1)* %a.idx.2, align 4
%x = load float, float addrspace(1)* %a.idx.2, align 4
store float 0.0, float addrspace(1)* %a.idx.3, align 4
ret float %x
}
attributes #0 = { nounwind }