replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[Hexagon] preserve test intent by removing undef 

We need to clean up the DAG floating-point undef logic.
This process is similar to how we handled integer undef
logic in D43141.

And as we did there, I'm trying to reduce the patch by
changing tests that would probably become meaningless
once we correct FP undef folding.

llvm-svn: 332550
This commit is contained in:
Sanjay Patel 2018-05-16 22:49:08 +00:00
parent 354842abc5
commit ae83159530
1 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
llvm/test/CodeGen/Hexagon/fp_latency.ll
View File

@ -40,15 +40,15 @@ b2: ; preds = %b2, %b1
%v15 = add nsw i32 %v14, %v11
%v16 = getelementptr inbounds [1000 x float], [1000 x float]* %v1, i32 0, i32 %v15
%v17 = load float, float* %v16, align 4, !tbaa !0
%v18 = fmul float %v17, undef
%v18 = fmul float %v17, %v17
%v19 = mul nsw i32 %v13, 25
%v20 = add nsw i32 %v19, %v11
%v21 = getelementptr inbounds [1000 x float], [1000 x float]* %v2, i32 0, i32 %v20
%v22 = load float, float* %v21, align 4, !tbaa !0
%v23 = fmul float %v22, undef
%v23 = fmul float %v22, %v22
%v24 = fadd float %v18, %v23
%v25 = load float, float* %v12, align 4, !tbaa !0
%v26 = fmul float %v25, undef
%v26 = fmul float %v25, %v25
%v27 = fadd float %v24, %v26
%v28 = getelementptr inbounds [1000 x float], [1000 x float]* %v0, i32 0, i32 %v20
%v29 = load float, float* %v28, align 4, !tbaa !0