Skip to content

[RISCV] Disable slideup optimization on the inconsistent element type of EVec and ContainerVT #159373

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 4 commits into from
Sep 24, 2025
Merged

[RISCV] Disable slideup optimization on the inconsistent element type of EVec and ContainerVT #159373

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
5c65f0f
[RISCV] Adapt the element type of EVec into ContainerVT in lowerBUILD…
XChy Sep 17, 2025
c5e8854
early exit
XChy Sep 18, 2025
91979ca
format
XChy Sep 18, 2025
b54b8fe
add tests
XChy Sep 19, 2025
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
8 changes: 8 additions & 0 deletions llvm/lib/Target/RISCV/RISCVISelLowering.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4553,6 +4553,14 @@ static SDValue lowerBUILD_VECTOR(SDValue Op, SelectionDAG &DAG,
break;
}

// Do not slideup if the element type of EVec is different.
if (SlideUp) {
MVT EVecEltVT = EVec.getSimpleValueType().getVectorElementType();
MVT ContainerEltVT = ContainerVT.getVectorElementType();
if (EVecEltVT != ContainerEltVT)
SlideUp = false;
}

if (SlideUp) {
MVT EVecContainerVT = EVec.getSimpleValueType();
// Make sure the original vector has scalable vector type.
Expand Down
122 changes: 122 additions & 0 deletions llvm/test/CodeGen/RISCV/rvv/fixed-vectors-int-buildvec.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3597,5 +3597,127 @@ define <4 x i32> @buildvec_vredmax_slideup(<8 x i32> %arg0, <8 x i32> %arg1, <8
ret <4 x i32> %255
}

define <16 x i16> @PR159294(<2 x i32> %a, <2 x i32> %b, <2 x i32> %c) {
; RV32-ONLY-LABEL: PR159294:
; RV32-ONLY: # %bb.0: # %entry
; RV32-ONLY-NEXT: vsetivli zero, 1, e32, m1, ta, ma
; RV32-ONLY-NEXT: vmv.x.s a0, v8
; RV32-ONLY-NEXT: vmv.x.s a1, v9
; RV32-ONLY-NEXT: vmv.x.s a2, v10
; RV32-ONLY-NEXT: vsetivli zero, 16, e16, m2, ta, ma
; RV32-ONLY-NEXT: vmv.v.x v8, a2
; RV32-ONLY-NEXT: vslide1down.vx v8, v8, a0
; RV32-ONLY-NEXT: vslide1down.vx v8, v8, a1
; RV32-ONLY-NEXT: vslidedown.vi v8, v8, 13
; RV32-ONLY-NEXT: ret
;
; RV32VB-LABEL: PR159294:
; RV32VB: # %bb.0: # %entry
; RV32VB-NEXT: vsetivli zero, 8, e32, m2, ta, ma
; RV32VB-NEXT: vmv.x.s a0, v8
; RV32VB-NEXT: vmv.x.s a1, v10
; RV32VB-NEXT: slli a0, a0, 16
; RV32VB-NEXT: zext.h a1, a1
; RV32VB-NEXT: or a0, a1, a0
; RV32VB-NEXT: vmv.x.s a1, v9
; RV32VB-NEXT: vmv.v.i v8, 0
; RV32VB-NEXT: zext.h a1, a1
; RV32VB-NEXT: vsetvli zero, zero, e32, m2, tu, ma
; RV32VB-NEXT: vmv.s.x v8, a0
; RV32VB-NEXT: vmv.s.x v10, a1
; RV32VB-NEXT: vsetivli zero, 2, e32, m1, tu, ma
; RV32VB-NEXT: vslideup.vi v8, v10, 1
; RV32VB-NEXT: ret
;
; RV32VB-PACK-LABEL: PR159294:
; RV32VB-PACK: # %bb.0: # %entry
; RV32VB-PACK-NEXT: vsetivli zero, 8, e32, m2, ta, ma
; RV32VB-PACK-NEXT: vmv.x.s a0, v8
; RV32VB-PACK-NEXT: vmv.x.s a1, v10
; RV32VB-PACK-NEXT: vmv.x.s a2, v9
; RV32VB-PACK-NEXT: pack a0, a1, a0
; RV32VB-PACK-NEXT: pack a1, a0, a0
; RV32VB-PACK-NEXT: vmv.v.x v8, a1
; RV32VB-PACK-NEXT: pack a1, a2, a0
; RV32VB-PACK-NEXT: vsetvli zero, zero, e32, m2, tu, ma
; RV32VB-PACK-NEXT: vmv.s.x v8, a0
; RV32VB-PACK-NEXT: vmv.s.x v10, a1
; RV32VB-PACK-NEXT: vsetivli zero, 2, e32, m1, tu, ma
; RV32VB-PACK-NEXT: vslideup.vi v8, v10, 1
; RV32VB-PACK-NEXT: ret
;
; RV64V-ONLY-LABEL: PR159294:
; RV64V-ONLY: # %bb.0: # %entry
; RV64V-ONLY-NEXT: vsetivli zero, 1, e32, m1, ta, ma
; RV64V-ONLY-NEXT: vmv.x.s a0, v8
; RV64V-ONLY-NEXT: vmv.x.s a1, v9
; RV64V-ONLY-NEXT: vmv.x.s a2, v10
; RV64V-ONLY-NEXT: vsetivli zero, 16, e16, m2, ta, ma
; RV64V-ONLY-NEXT: vmv.v.x v8, a2
; RV64V-ONLY-NEXT: vslide1down.vx v8, v8, a0
; RV64V-ONLY-NEXT: vslide1down.vx v8, v8, a1
; RV64V-ONLY-NEXT: vslidedown.vi v8, v8, 13
; RV64V-ONLY-NEXT: ret
;
; RVA22U64-LABEL: PR159294:
; RVA22U64: # %bb.0: # %entry
; RVA22U64-NEXT: vsetivli zero, 8, e32, m2, ta, ma
; RVA22U64-NEXT: vmv.x.s a0, v8
; RVA22U64-NEXT: vmv.x.s a1, v10
; RVA22U64-NEXT: slli a0, a0, 16
; RVA22U64-NEXT: zext.h a1, a1
; RVA22U64-NEXT: or a0, a0, a1
; RVA22U64-NEXT: vmv.x.s a1, v9
; RVA22U64-NEXT: vmv.v.i v8, 0
; RVA22U64-NEXT: zext.h a1, a1
; RVA22U64-NEXT: vsetvli zero, zero, e32, m2, tu, ma
; RVA22U64-NEXT: vmv.s.x v8, a0
; RVA22U64-NEXT: vmv.s.x v10, a1
; RVA22U64-NEXT: vsetivli zero, 2, e32, m1, tu, ma
; RVA22U64-NEXT: vslideup.vi v8, v10, 1
; RVA22U64-NEXT: ret
;
; RVA22U64-PACK-LABEL: PR159294:
; RVA22U64-PACK: # %bb.0: # %entry
; RVA22U64-PACK-NEXT: vsetivli zero, 8, e32, m2, ta, ma
; RVA22U64-PACK-NEXT: vmv.x.s a0, v8
; RVA22U64-PACK-NEXT: vmv.x.s a1, v10
; RVA22U64-PACK-NEXT: vmv.x.s a2, v9
; RVA22U64-PACK-NEXT: packw a0, a1, a0
; RVA22U64-PACK-NEXT: packw a1, a0, a0
; RVA22U64-PACK-NEXT: vmv.v.x v8, a1
; RVA22U64-PACK-NEXT: packw a1, a2, a0
; RVA22U64-PACK-NEXT: vsetvli zero, zero, e32, m2, tu, ma
; RVA22U64-PACK-NEXT: vmv.s.x v8, a0
; RVA22U64-PACK-NEXT: vmv.s.x v10, a1
; RVA22U64-PACK-NEXT: vsetivli zero, 2, e32, m1, tu, ma
; RVA22U64-PACK-NEXT: vslideup.vi v8, v10, 1
; RVA22U64-PACK-NEXT: ret
;
; RV64ZVE32-LABEL: PR159294:
; RV64ZVE32: # %bb.0: # %entry
; RV64ZVE32-NEXT: vsetivli zero, 1, e32, m1, ta, ma
; RV64ZVE32-NEXT: vmv.x.s a0, v8
; RV64ZVE32-NEXT: vmv.x.s a1, v9
; RV64ZVE32-NEXT: vmv.x.s a2, v10
; RV64ZVE32-NEXT: vsetivli zero, 16, e16, m2, ta, ma
; RV64ZVE32-NEXT: vmv.v.x v8, a2
; RV64ZVE32-NEXT: vslide1down.vx v8, v8, a0
; RV64ZVE32-NEXT: vslide1down.vx v8, v8, a1
; RV64ZVE32-NEXT: vslidedown.vi v8, v8, 13
; RV64ZVE32-NEXT: ret
entry:
%vecext3 = extractelement <2 x i32> %a, i32 0
%conv4 = trunc i32 %vecext3 to i16
%vecinit5 = insertelement <16 x i16> <i16 0, i16 poison, i16 poison, i16 poison, i16 poison, i16 poison, i16 poison, i16 poison, i16 poison, i16 poison, i16 poison, i16 poison, i16 poison, i16 poison, i16 poison, i16 poison>, i16 %conv4, i32 1
%vecext7 = extractelement <2 x i32> %b, i32 0
%conv8 = trunc i32 %vecext7 to i16
%vecinit9 = insertelement <16 x i16> %vecinit5, i16 %conv8, i32 2
%vecext59 = extractelement <2 x i32> %c, i32 0
%conv60 = trunc i32 %vecext59 to i16
%vecinit61 = insertelement <16 x i16> %vecinit9, i16 %conv60, i32 0
ret <16 x i16> %vecinit61
}

;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; RV64: {{.*}}