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

[InstCombine] Add test cases to show bad canonicalization of bitcasts between x86_mmx and <1 x i64>. 

As the test cases show, we end up with an insert/extract and a
bitcast to/from i64. x86_mmx is for some purposes conceptually a
vector. We shouldn't be adding scalar conversions around it.

Since _m64 is defined as <1 x i64> and intrinsics use x86_mmx
as their input/output these extra scalar operations prevent
the X86 backend from generating good code especially on 32-bit
targets where i64 gets split.
This commit is contained in:
Craig Topper 2019-11-07 12:37:24 -08:00
parent 2b943c4687
commit 96119586c9
1 changed files with 19 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
llvm/test/Transforms/InstCombine/bitcast-vec-canon.ll
View File

@ -38,3 +38,22 @@ define <1 x i64> @d(i64 %y) {
ret <1 x i64> %c
}
define x86_mmx @e(<1 x i64> %y) {
; CHECK-LABEL: @e(
; CHECK-NEXT: [[TMP1:%.*]] = extractelement <1 x i64> %y, i32 0
; CHECK-NEXT: [[C:%.*]] = bitcast i64 [[TMP1]] to x86_mmx
; CHECK-NEXT: ret x86_mmx [[C]]
;
%c = bitcast <1 x i64> %y to x86_mmx
ret x86_mmx %c
}
define <1 x i64> @f(x86_mmx %y) {
; CHECK-LABEL: @f(
; CHECK-NEXT: [[TMP1:%.*]] = bitcast x86_mmx %y to i64
; CHECK-NEXT: [[C:%.*]] = insertelement <1 x i64> undef, i64 [[TMP1]], i32 0
; CHECK-NEXT: ret <1 x i64> [[C]]
;
%c = bitcast x86_mmx %y to <1 x i64>
ret <1 x i64> %c
}