This is an archive of the discontinued LLVM Phabricator instance.

Differential D153630

[RISCV] Add test cases for vmerge peephole with vslides
ClosedPublic

Authored by luke on Jun 23 2023, 7:04 AM.

Details

Reviewers
craig.topper
fakepaper56
nitinjohnraj
reames
frasercrmck
Commits
rG0d0bfa8a14dc: [RISCV] Add test cases for vmerge peephole with vslides
Summary

Currently vslide1{up,down}s can have vmerges folded into them, but not
vslide{up,down}s.

Diff Detail

Event Timeline

luke created this revision.Jun 23 2023, 7:04 AM
Herald added a project: Restricted Project. · View Herald TranscriptJun 23 2023, 7:04 AM
Herald added subscribers: jobnoorman, asb, pmatos and 26 others. · View Herald Transcript
luke requested review of this revision.Jun 23 2023, 7:04 AM
Herald added a project: Restricted Project. · View Herald TranscriptJun 23 2023, 7:04 AM
Herald added subscribers: llvm-commits, wangpc, eopXD, MaskRay. · View Herald Transcript
luke added a child revision: D153631: [RISCV] Teach doPeepholeMaskedRVV to handle vslide{up,down}.Jun 23 2023, 7:37 AM
Harbormaster completed remote builds in B240752: Diff 533950.Jun 23 2023, 9:11 AM
craig.topper accepted this revision.Jun 24 2023, 2:02 PM

LGTM

This revision is now accepted and ready to land.Jun 24 2023, 2:02 PM
Closed by commit rG0d0bfa8a14dc: [RISCV] Add test cases for vmerge peephole with vslides (authored by luke). · Explain WhyJun 26 2023, 1:36 AM
This revision was automatically updated to reflect the committed changes.
luke added a commit: rG0d0bfa8a14dc: [RISCV] Add test cases for vmerge peephole with vslides.

Revision Contents

PathSize
llvm/
test/
CodeGen/
RISCV/
rvv/
59 lines
55 lines

Diff 534458

llvm/test/CodeGen/RISCV/rvv/rvv-peephole-vmerge-masked-vops.ll

Show First 20 Lines • Show All 87 Lines • ▼ Show 20 Lines %a = call <vscale x 2 x i32> @llvm.riscv.vle.mask.nxv2i32(
i64 %vl, i64 1) i64 %vl, i64 1)
%splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0 %splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
%mask = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer %mask = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
%b = call <vscale x 2 x i32> @llvm.riscv.vmerge.nxv2i32.nxv2i32(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %passthru, <vscale x 2 x i32> %a, <vscale x 2 x i1> %mask, i64 %vl) %b = call <vscale x 2 x i32> @llvm.riscv.vmerge.nxv2i32.nxv2i32(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %passthru, <vscale x 2 x i32> %a, <vscale x 2 x i1> %mask, i64 %vl)
ret <vscale x 2 x i32> %b ret <vscale x 2 x i32> %b
} }
declare <vscale x 2 x i32> @llvm.riscv.vle.mask.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i32>*, <vscale x 2 x i1>, i64, i64) declare <vscale x 2 x i32> @llvm.riscv.vle.mask.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i32>*, <vscale x 2 x i1>, i64, i64)
declare <vscale x 2 x i32> @llvm.riscv.vslideup.mask.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i32>, i64, <vscale x 2 x i1>, i64, i64)
define <vscale x 2 x i32> @vpmerge_vslideup(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %v, i64 %x, <vscale x 2 x i1> %m, i64 %vl) {
; CHECK-LABEL: vpmerge_vslideup:
; CHECK: # %bb.0:
; CHECK-NEXT: vsetvli zero, a1, e32, m1, tu, mu
; CHECK-NEXT: vmv1r.v v10, v8
; CHECK-NEXT: vslideup.vx v10, v9, a0, v0.t
; CHECK-NEXT: vmv.v.v v8, v10
; CHECK-NEXT: ret
%a = call <vscale x 2 x i32> @llvm.riscv.vslideup.mask.nxv2i32(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %v, i64 %x, <vscale x 2 x i1> %m, i64 %vl, i64 0)
%splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
%mask = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
%b = call <vscale x 2 x i32> @llvm.riscv.vmerge.nxv2i32.nxv2i32(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %passthru, <vscale x 2 x i32> %a, <vscale x 2 x i1> %mask, i64 %vl)
ret <vscale x 2 x i32> %b
}
declare <vscale x 2 x i32> @llvm.riscv.vslidedown.mask.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i32>, i64, <vscale x 2 x i1>, i64, i64)
define <vscale x 2 x i32> @vpmerge_vslidedown(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %v, i64 %x, <vscale x 2 x i1> %m, i64 %vl) {
; CHECK-LABEL: vpmerge_vslidedown:
; CHECK: # %bb.0:
; CHECK-NEXT: vsetvli zero, a1, e32, m1, tu, mu
; CHECK-NEXT: vmv1r.v v10, v8
; CHECK-NEXT: vslidedown.vx v10, v9, a0, v0.t
; CHECK-NEXT: vmv.v.v v8, v10
; CHECK-NEXT: ret
%a = call <vscale x 2 x i32> @llvm.riscv.vslidedown.mask.nxv2i32(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %v, i64 %x, <vscale x 2 x i1> %m, i64 %vl, i64 0)
%splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
%mask = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
%b = call <vscale x 2 x i32> @llvm.riscv.vmerge.nxv2i32.nxv2i32(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %passthru, <vscale x 2 x i32> %a, <vscale x 2 x i1> %mask, i64 %vl)
ret <vscale x 2 x i32> %b
}
declare <vscale x 2 x i32> @llvm.riscv.vslide1up.mask.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i32>, i32, <vscale x 2 x i1>, i64, i64)
define <vscale x 2 x i32> @vpmerge_vslide1up(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %v, i32 %x, <vscale x 2 x i1> %m, i64 %vl) {
; CHECK-LABEL: vpmerge_vslide1up:
; CHECK: # %bb.0:
; CHECK-NEXT: vsetvli zero, a1, e32, m1, tu, mu
; CHECK-NEXT: vslide1up.vx v8, v9, a0, v0.t
; CHECK-NEXT: ret
%a = call <vscale x 2 x i32> @llvm.riscv.vslide1up.mask.nxv2i32(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %v, i32 %x, <vscale x 2 x i1> %m, i64 %vl, i64 0)
%splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
%mask = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
%b = call <vscale x 2 x i32> @llvm.riscv.vmerge.nxv2i32.nxv2i32(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %passthru, <vscale x 2 x i32> %a, <vscale x 2 x i1> %mask, i64 %vl)
ret <vscale x 2 x i32> %b
}
declare <vscale x 2 x i32> @llvm.riscv.vslide1down.mask.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i32>, i32, <vscale x 2 x i1>, i64, i64)
define <vscale x 2 x i32> @vpmerge_vslide1down(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %v, i32 %x, <vscale x 2 x i1> %m, i64 %vl) {
; CHECK-LABEL: vpmerge_vslide1down:
; CHECK: # %bb.0:
; CHECK-NEXT: vsetvli zero, a1, e32, m1, tu, mu
; CHECK-NEXT: vslide1down.vx v8, v9, a0, v0.t
; CHECK-NEXT: ret
%a = call <vscale x 2 x i32> @llvm.riscv.vslide1down.mask.nxv2i32(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %v, i32 %x, <vscale x 2 x i1> %m, i64 %vl, i64 0)
%splat = insertelement <vscale x 2 x i1> poison, i1 -1, i32 0
%mask = shufflevector <vscale x 2 x i1> %splat, <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer
%b = call <vscale x 2 x i32> @llvm.riscv.vmerge.nxv2i32.nxv2i32(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %passthru, <vscale x 2 x i32> %a, <vscale x 2 x i1> %mask, i64 %vl)
ret <vscale x 2 x i32> %b
}

llvm/test/CodeGen/RISCV/rvv/rvv-peephole-vmerge-vops.ll

Show First 20 Lines • Show All 740 Lines • ▼ Show 20 Lines
; CHECK-NEXT: vfrec7.v v8, v9, v0.t ; CHECK-NEXT: vfrec7.v v8, v9, v0.t
; CHECK-NEXT: ret ; CHECK-NEXT: ret
%1 = zext i32 %vl to i64 %1 = zext i32 %vl to i64
%a = call <vscale x 2 x float> @llvm.riscv.vfrec7.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> %vf, i64 %1) %a = call <vscale x 2 x float> @llvm.riscv.vfrec7.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> %vf, i64 %1)
%b = call <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %a, <vscale x 2 x float> %passthru, i32 %vl) %b = call <vscale x 2 x float> @llvm.vp.select.nxv2f32(<vscale x 2 x i1> %m, <vscale x 2 x float> %a, <vscale x 2 x float> %passthru, i32 %vl)
ret <vscale x 2 x float> %b ret <vscale x 2 x float> %b
} }
; Test slides
declare <vscale x 2 x i32> @llvm.riscv.vslideup.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i32>, i64, i64, i64)
define <vscale x 2 x i32> @vpselect_vslideup(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %v, i64 %x, <vscale x 2 x i1> %m, i32 zeroext %vl) {
; CHECK-LABEL: vpselect_vslideup:
; CHECK: # %bb.0:
; CHECK-NEXT: vsetvli zero, a1, e32, m1, ta, ma
; CHECK-NEXT: vslideup.vx v10, v9, a0
; CHECK-NEXT: vmerge.vvm v8, v8, v10, v0
; CHECK-NEXT: ret
%1 = zext i32 %vl to i64
%a = call <vscale x 2 x i32> @llvm.riscv.vslideup.nxv2i32(<vscale x 2 x i32> undef, <vscale x 2 x i32> %v, i64 %x, i64 %1, i64 0)
%b = call <vscale x 2 x i32> @llvm.vp.select.nxv2i32(<vscale x 2 x i1> %m, <vscale x 2 x i32> %a, <vscale x 2 x i32> %passthru, i32 %vl)
ret <vscale x 2 x i32> %b
}
declare <vscale x 2 x i32> @llvm.riscv.vslidedown.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i32>, i64, i64, i64)
define <vscale x 2 x i32> @vpselect_vslidedown(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %v, i64 %x, <vscale x 2 x i1> %m, i32 zeroext %vl) {
; CHECK-LABEL: vpselect_vslidedown:
; CHECK: # %bb.0:
; CHECK-NEXT: vsetvli zero, a1, e32, m1, ta, ma
; CHECK-NEXT: vslidedown.vx v9, v9, a0
; CHECK-NEXT: vmerge.vvm v8, v8, v9, v0
; CHECK-NEXT: ret
%1 = zext i32 %vl to i64
%a = call <vscale x 2 x i32> @llvm.riscv.vslidedown.nxv2i32(<vscale x 2 x i32> undef, <vscale x 2 x i32> %v, i64 %x, i64 %1, i64 0)
%b = call <vscale x 2 x i32> @llvm.vp.select.nxv2i32(<vscale x 2 x i1> %m, <vscale x 2 x i32> %a, <vscale x 2 x i32> %passthru, i32 %vl)
ret <vscale x 2 x i32> %b
}
declare <vscale x 2 x i32> @llvm.riscv.vslide1up.nxv2i32.i32(<vscale x 2 x i32>, <vscale x 2 x i32>, i32, i64)
define <vscale x 2 x i32> @vpselect_vslide1up(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %v, i32 %x, <vscale x 2 x i1> %m, i32 zeroext %vl) {
; CHECK-LABEL: vpselect_vslide1up:
; CHECK: # %bb.0:
; CHECK-NEXT: vsetvli zero, a1, e32, m1, ta, mu
; CHECK-NEXT: vslide1up.vx v8, v9, a0, v0.t
; CHECK-NEXT: ret
%1 = zext i32 %vl to i64
%a = call <vscale x 2 x i32> @llvm.riscv.vslide1up.nxv2i32.i32(<vscale x 2 x i32> undef, <vscale x 2 x i32> %v, i32 %x, i64 %1)
%b = call <vscale x 2 x i32> @llvm.vp.select.nxv2i32(<vscale x 2 x i1> %m, <vscale x 2 x i32> %a, <vscale x 2 x i32> %passthru, i32 %vl)
ret <vscale x 2 x i32> %b
}
declare <vscale x 2 x i32> @llvm.riscv.vslide1down.nxv2i32.i32(<vscale x 2 x i32>, <vscale x 2 x i32>, i32, i64)
define <vscale x 2 x i32> @vpselect_vslide1down(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %v, i32 %x, <vscale x 2 x i1> %m, i32 zeroext %vl) {
; CHECK-LABEL: vpselect_vslide1down:
; CHECK: # %bb.0:
; CHECK-NEXT: vsetvli zero, a1, e32, m1, ta, mu
; CHECK-NEXT: vslide1down.vx v8, v9, a0, v0.t
; CHECK-NEXT: ret
%1 = zext i32 %vl to i64
%a = call <vscale x 2 x i32> @llvm.riscv.vslide1down.nxv2i32.i32(<vscale x 2 x i32> undef, <vscale x 2 x i32> %v, i32 %x, i64 %1)
%b = call <vscale x 2 x i32> @llvm.vp.select.nxv2i32(<vscale x 2 x i1> %m, <vscale x 2 x i32> %a, <vscale x 2 x i32> %passthru, i32 %vl)
ret <vscale x 2 x i32> %b
}
; Test vector operations with VLMAX vector length. ; Test vector operations with VLMAX vector length.
; Test binary operator with vp.select and add. ; Test binary operator with vp.select and add.
define <vscale x 2 x i32> @vpselect_add(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %x, <vscale x 2 x i32> %y, <vscale x 2 x i1> %m, i32 zeroext %vl) { define <vscale x 2 x i32> @vpselect_add(<vscale x 2 x i32> %passthru, <vscale x 2 x i32> %x, <vscale x 2 x i32> %y, <vscale x 2 x i1> %m, i32 zeroext %vl) {
; CHECK-LABEL: vpselect_add: ; CHECK-LABEL: vpselect_add:
; CHECK: # %bb.0: ; CHECK: # %bb.0:
; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu ; CHECK-NEXT: vsetvli zero, a0, e32, m1, ta, mu
; CHECK-NEXT: vadd.vv v8, v9, v10, v0.t ; CHECK-NEXT: vadd.vv v8, v9, v10, v0.t
▲ Show 20 Lines • Show All 126 Lines • Show Last 20 Lines