RPCS3/llvm
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm.git synced 2024-11-25 04:39:51 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

[LSR] Call canonicalize after we generate a new Formula in GenerateTruncates. Fix PR33077.

Browse Source 
The testcase in PR33077 generates a LSR Use Formula with two SCEVAddRecExprs for the same
loop. Such uncommon formula will become non-canonical after GenerateTruncates adds sign
extension to the ScaledReg of the Formula, and it will break the assertion that every
Formula to be inserted is canonical.

The fix is to call canonicalize for the raw Formula generated by GenerateTruncates
before inserting it.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@303361 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Wei Mi 2017-05-18 17:21:22 +00:00
parent ae1c853358
commit 7db2df6d96
2 changed files with 37 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
lib/Transforms/Scalar/LoopStrengthReduce.cpp
View File

@ -3805,6 +3805,7 @@ void LSRInstance::GenerateTruncates(LSRUse &LU, unsigned LUIdx, Formula Base) {
if (!F.hasRegsUsedByUsesOtherThan(LUIdx, RegUses))
continue;
F.canonicalize(*L);
(void)InsertFormula(LU, LUIdx, F);
}
}

36
test/Transforms/LoopStrengthReduce/X86/canonical-2.ll Normal file
View File

@ -0,0 +1,36 @@
; REQUIRES: asserts
; RUN: opt -mtriple=x86_64-unknown-linux-gnu -loop-reduce -S < %s
; PR33077. Check the LSR Use formula to be inserted is already canonicalized and
; will not trigger assertion.
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
; Function Attrs: uwtable
define void @foo() {
cHeapLvb.exit:
br label %not_zero48.us
not_zero48.us: ; preds = %not_zero48.us, %cHeapLvb.exit
%indvars.iv.us = phi i64 [ %indvars.iv.next.us.7, %not_zero48.us ], [ undef, %cHeapLvb.exit ]
%0 = phi i32 [ %13, %not_zero48.us ], [ undef, %cHeapLvb.exit ]
%indvars.iv.next.us = add nuw nsw i64 %indvars.iv.us, 1
%1 = add i32 %0, 2
%2 = getelementptr inbounds i32, i32 addrspace(1)* undef, i64 %indvars.iv.next.us
%3 = load i32, i32 addrspace(1)* %2, align 4
%4 = add i32 %0, 3
%5 = load i32, i32 addrspace(1)* undef, align 4
%6 = sub i32 undef, %5
%factor.us.2 = shl i32 %6, 1
%7 = add i32 %factor.us.2, %1
%8 = load i32, i32 addrspace(1)* undef, align 4
%9 = sub i32 %7, %8
%factor.us.3 = shl i32 %9, 1
%10 = add i32 %factor.us.3, %4
%11 = load i32, i32 addrspace(1)* undef, align 4
%12 = sub i32 %10, %11
%factor.us.4 = shl i32 %12, 1
%13 = add i32 %0, 8
%indvars.iv.next.us.7 = add nsw i64 %indvars.iv.us, 8
br label %not_zero48.us
}