Skip to content

[AArch64] Fix BE popcount casts. #129879

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Mar 5, 2025
Merged

[AArch64] Fix BE popcount casts. #129879

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 8 additions & 2 deletions llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10788,7 +10788,10 @@ SDValue AArch64TargetLowering::LowerCTPOP_PARITY(SDValue Op,
if (VT == MVT::i32)
AddV = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, MVT::i32, AddV,
DAG.getConstant(0, DL, MVT::i64));
AddV = DAG.getNode(ISD::BITCAST, DL, VT, AddV);
else
AddV = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, VT,
DAG.getNode(AArch64ISD::NVCAST, DL, MVT::v1i64, AddV),
DAG.getConstant(0, DL, MVT::i64));
if (IsParity)
AddV = DAG.getNode(ISD::AND, DL, VT, AddV, DAG.getConstant(1, DL, VT));
return AddV;
Expand All @@ -10797,7 +10800,10 @@ SDValue AArch64TargetLowering::LowerCTPOP_PARITY(SDValue Op,

SDValue CtPop = DAG.getNode(ISD::CTPOP, DL, MVT::v16i8, Val);
SDValue AddV = DAG.getNode(AArch64ISD::UADDV, DL, MVT::v16i8, CtPop);
AddV = DAG.getNode(ISD::BITCAST, DL, VT, AddV);
AddV = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, MVT::i64,
DAG.getNode(AArch64ISD::NVCAST, DL, MVT::v2i64, AddV),
DAG.getConstant(0, DL, MVT::i64));
AddV = DAG.getZExtOrTrunc(AddV, DL, VT);
if (IsParity)
AddV = DAG.getNode(ISD::AND, DL, VT, AddV, DAG.getConstant(1, DL, VT));
return AddV;
Expand Down
8 changes: 3 additions & 5 deletions llvm/test/CodeGen/AArch64/arm64-popcnt.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -129,7 +129,6 @@ define i64 @cnt64_advsimd(i64 %x) nounwind readnone {
; CHECK-BE-NEXT: rev64 v0.8b, v0.8b
; CHECK-BE-NEXT: cnt v0.8b, v0.8b
; CHECK-BE-NEXT: addv b0, v0.8b
; CHECK-BE-NEXT: rev64 v0.8b, v0.8b
; CHECK-BE-NEXT: fmov x0, d0
; CHECK-BE-NEXT: ret
%cnt = tail call i64 @llvm.ctpop.i64(i64 %x)
Expand Down Expand Up @@ -436,9 +435,9 @@ define i128 @cnt128(i128 %x) nounwind readnone {
; CHECK: // %bb.0:
; CHECK-NEXT: fmov d0, x0
; CHECK-NEXT: mov.d v0[1], x1
; CHECK-NEXT: mov x1, xzr
; CHECK-NEXT: cnt.16b v0, v0
; CHECK-NEXT: addv.16b b0, v0
; CHECK-NEXT: mov.d x1, v0[1]
; CHECK-NEXT: fmov x0, d0
; CHECK-NEXT: ret
;
Expand Down Expand Up @@ -481,13 +480,12 @@ define i128 @cnt128(i128 %x) nounwind readnone {
; CHECK-BE-LABEL: cnt128:
; CHECK-BE: // %bb.0:
; CHECK-BE-NEXT: fmov d0, x0
; CHECK-BE-NEXT: mov x0, xzr
; CHECK-BE-NEXT: mov v0.d[1], x1
; CHECK-BE-NEXT: rev64 v0.16b, v0.16b
; CHECK-BE-NEXT: cnt v0.16b, v0.16b
; CHECK-BE-NEXT: addv b0, v0.16b
; CHECK-BE-NEXT: rev64 v0.16b, v0.16b
; CHECK-BE-NEXT: mov x1, v0.d[1]
; CHECK-BE-NEXT: fmov x0, d0
; CHECK-BE-NEXT: fmov x1, d0
; CHECK-BE-NEXT: ret
%cnt = tail call i128 @llvm.ctpop.i128(i128 %x)
ret i128 %cnt
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/CodeGen/AArch64/parity.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -159,7 +159,7 @@ define i32 @parity_64_trunc(i64 %x) {
; CHECK-NEXT: fmov d0, x0
; CHECK-NEXT: cnt v0.8b, v0.8b
; CHECK-NEXT: addv b0, v0.8b
; CHECK-NEXT: fmov x8, d0
; CHECK-NEXT: fmov w8, s0
; CHECK-NEXT: and w0, w8, #0x1
; CHECK-NEXT: ret
;
Expand Down
19 changes: 9 additions & 10 deletions llvm/test/CodeGen/AArch64/popcount.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -41,8 +41,8 @@ define i8 @popcount128(ptr nocapture nonnull readonly %0) {
; BE-NEXT: rev64 v0.16b, v0.16b
; BE-NEXT: cnt v0.16b, v0.16b
; BE-NEXT: addv b0, v0.16b
; BE-NEXT: rev32 v0.16b, v0.16b
; BE-NEXT: mov w0, v0.s[3]
; BE-NEXT: rev64 v0.4s, v0.4s
; BE-NEXT: mov w0, v0.s[1]
; BE-NEXT: ret
;
; GISEL-LABEL: popcount128:
Expand Down Expand Up @@ -138,10 +138,10 @@ define i16 @popcount256(ptr nocapture nonnull readonly %0) {
; BE-NEXT: cnt v1.16b, v1.16b
; BE-NEXT: addv b0, v0.16b
; BE-NEXT: addv b1, v1.16b
; BE-NEXT: rev32 v0.16b, v0.16b
; BE-NEXT: rev32 v1.16b, v1.16b
; BE-NEXT: mov w8, v0.s[3]
; BE-NEXT: mov w9, v1.s[3]
; BE-NEXT: rev64 v0.4s, v0.4s
; BE-NEXT: rev64 v1.4s, v1.4s
; BE-NEXT: mov w8, v0.s[1]
; BE-NEXT: mov w9, v1.s[1]
; BE-NEXT: add w0, w9, w8
; BE-NEXT: ret
;
Expand Down Expand Up @@ -227,22 +227,21 @@ define <1 x i128> @popcount1x128(<1 x i128> %0) {
; CHECK: // %bb.0: // %Entry
; CHECK-NEXT: fmov d0, x0
; CHECK-NEXT: mov v0.d[1], x1
; CHECK-NEXT: mov x1, xzr
; CHECK-NEXT: cnt v0.16b, v0.16b
; CHECK-NEXT: addv b0, v0.16b
; CHECK-NEXT: mov x1, v0.d[1]
; CHECK-NEXT: fmov x0, d0
; CHECK-NEXT: ret
;
; BE-LABEL: popcount1x128:
; BE: // %bb.0: // %Entry
; BE-NEXT: fmov d0, x0
; BE-NEXT: mov x0, xzr
; BE-NEXT: mov v0.d[1], x1
; BE-NEXT: rev64 v0.16b, v0.16b
; BE-NEXT: cnt v0.16b, v0.16b
; BE-NEXT: addv b0, v0.16b
; BE-NEXT: rev64 v0.16b, v0.16b
; BE-NEXT: mov x1, v0.d[1]
; BE-NEXT: fmov x0, d0
; BE-NEXT: fmov x1, d0
; BE-NEXT: ret
;
; GISEL-LABEL: popcount1x128:
Expand Down