Skip to content

[RISCV] Adjust LMUL if not used to avoid VL toggle #69259

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

Closed
wants to merge 3 commits into from
Closed

[RISCV] Adjust LMUL if not used to avoid VL toggle #69259

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
31f3692
[RISCV] Adjust LMUL if not used to avoid VL toggle
lukel97 Oct 16, 2023
c131942
Check that new ratio is valid
lukel97 Oct 17, 2023
0ddba02
Update test outside of rvv tree
lukel97 Oct 17, 2023
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
60 changes: 60 additions & 0 deletions llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1463,6 +1463,57 @@ static bool canMutatePriorConfig(const MachineInstr &PrevMI,
return areCompatibleVTYPEs(PriorVType, VType, Used);
}

// If LMUL or the SEW/LMUL ratio aren't demanded and MI and NextMI have the same
// AVL, then we can try and change MI's LMUL so that we can avoid setting VL in
// NextMI, e.g:
//
// vsetivli zero, 4, e32, m1, ta, ma
// vsetivli zero, 4, e16, mf4, ta, ma
//
// vsetivli zero, 4, e32, mf2, ta, ma
// vsetvli zero, zero, e16, mf4, ta, ma
//
// If possible, returns the new VTYPE that should be used for MI.
static std::optional<unsigned>
canAdjustSEWLMULRatio(const MachineInstr &MI, const MachineInstr &NextMI,
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Function turns the VTYPE, not a boolean. So a better name might be: getAdjustedVTYPE

const DemandedFields &Used) {
if (Used.LMUL || Used.SEWLMULRatio)
return std::nullopt;
if (!NextMI.getOperand(0).isDead())
return std::nullopt;
// If we end up increasing the SEW/LMUL ratio, then we will decrease VLMAX,
// which means we might end up changing VL in the case that AVL > VLMAX. So
// bail if the exact VL value is needed.
//
// TODO: We could potentially relax this when we know we're increasing VLMAX.
if (Used.VLAny)
return std::nullopt;

// If NextMI is already zero, zero then bail. If MI is zero, zero then we
// won't be able to tell if it has the same AVL as NextMI, so also bail.
if (isVLPreservingConfig(MI) || isVLPreservingConfig(NextMI))
return std::nullopt;

VSETVLIInfo NextMIInfo = getInfoForVSETVLI(NextMI);
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't remember if it's safe to call getInfoForVSETVLI here on the post-rewrite MI. However, the existence of isNonZeroAVL seems to indicate I thought it wasn't previously?

We should either write this in terms of the raw MI, or insert assertions sufficient to prove this is safe during insertion, and rewrite the existing post pass logic in terms of the VSETVLIInfo object.

Copy link
Contributor Author

@lukel97 lukel97 Oct 20, 2023 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, in my local branch I've rewritten this to just operate on the MI, it's not too bad and we don't need getInfoForVSETVLI here.

VSETVLIInfo MIInfo = getInfoForVSETVLI(MI);
if (!MIInfo.hasSameAVL(NextMIInfo))
return std::nullopt;

unsigned SEW = MIInfo.getSEW() * 8;
// Fixed point value with 3 fractional bits.
unsigned NewRatio = SEW / NextMIInfo.getSEWLMULRatio();
if (NewRatio < 1 || NewRatio > 64)
return std::nullopt;
bool Fractional = NewRatio < 8;
RISCVII::VLMUL NewVLMul = RISCVVType::encodeLMUL(
Fractional ? 8 / NewRatio : NewRatio / 8, Fractional);

unsigned VType = MIInfo.encodeVTYPE();
return RISCVVType::encodeVTYPE(NewVLMul, SEW / 8,
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

SEW/8 appears to just be MIInfo.getSEW()

RISCVVType::isTailAgnostic(VType),
RISCVVType::isMaskAgnostic(VType));
}

void RISCVInsertVSETVLI::doLocalPostpass(MachineBasicBlock &MBB) {
MachineInstr *NextMI = nullptr;
// We can have arbitrary code in successors, so VL and VTYPE
Expand All @@ -1484,6 +1535,15 @@ void RISCVInsertVSETVLI::doLocalPostpass(MachineBasicBlock &MBB) {
Used.demandVL();

if (NextMI) {
if (auto NewVType = canAdjustSEWLMULRatio(MI, *NextMI, Used)) {
MI.getOperand(2).setImm(*NewVType);
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Have we properly guarded for the save/restore form which has a vreg as operand 2?

// Convert NextMI to vsetvli zero, zero
NextMI->setDesc(TII->get(RISCV::PseudoVSETVLIX0));
NextMI->getOperand(0).setReg(RISCV::X0);
NextMI->getOperand(0).setIsDead(true);
NextMI->getOperand(1).ChangeToRegister(RISCV::X0, false, false, true);
}

if (!Used.usedVL() && !Used.usedVTYPE()) {
ToDelete.push_back(&MI);
// Leave NextMI unchanged
Expand Down
4 changes: 2 additions & 2 deletions llvm/test/CodeGen/RISCV/double_reduct.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -106,9 +106,9 @@ define i16 @add_ext_i16(<16 x i8> %a, <16 x i8> %b) {
define i16 @add_ext_v32i16(<32 x i8> %a, <16 x i8> %b) {
; CHECK-LABEL: add_ext_v32i16:
; CHECK: # %bb.0:
; CHECK-NEXT: vsetivli zero, 16, e16, m1, ta, ma
; CHECK-NEXT: vsetivli zero, 16, e16, m2, ta, ma
; CHECK-NEXT: vmv.s.x v11, zero
; CHECK-NEXT: vsetivli zero, 16, e8, m1, ta, ma
; CHECK-NEXT: vsetvli zero, zero, e8, m1, ta, ma
; CHECK-NEXT: vwredsumu.vs v10, v10, v11
; CHECK-NEXT: li a0, 32
; CHECK-NEXT: vsetvli zero, a0, e8, m2, ta, ma
Expand Down
10 changes: 4 additions & 6 deletions llvm/test/CodeGen/RISCV/rvv/fixed-vectors-insert.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -65,9 +65,8 @@ define <32 x i32> @insertelt_v32i32_31(<32 x i32> %a, i32 %y) {
; CHECK-LABEL: insertelt_v32i32_31:
; CHECK: # %bb.0:
; CHECK-NEXT: li a1, 32
; CHECK-NEXT: vsetvli zero, a1, e32, m1, ta, ma
; CHECK-NEXT: vmv.s.x v16, a0
; CHECK-NEXT: vsetvli zero, a1, e32, m8, ta, ma
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is an interesting sub-case as we shouldn't need to be reasoning about ratios here. Why didn't the existing logic get this case given only LMUL is changing? I think this might just be the lack of LI handling in isNonZeroAVL?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Good point, I'm not quite sure why this part wasn't already caught since it should just be an entire VTYPE substitution.
Note to self though, this is one of the cases the forward-pass-ratio-mutation in #69788 won't catch because it's the first instruction in the BB

; CHECK-NEXT: vmv.s.x v16, a0
; CHECK-NEXT: vslideup.vi v8, v16, 31
; CHECK-NEXT: ret
%b = insertelement <32 x i32> %a, i32 %y, i32 31
Expand Down Expand Up @@ -103,9 +102,8 @@ define <64 x i32> @insertelt_v64i32_63(<64 x i32> %a, i32 %y) {
; CHECK-LABEL: insertelt_v64i32_63:
; CHECK: # %bb.0:
; CHECK-NEXT: li a1, 32
; CHECK-NEXT: vsetvli zero, a1, e32, m1, ta, ma
; CHECK-NEXT: vmv.s.x v24, a0
; CHECK-NEXT: vsetvli zero, a1, e32, m8, ta, ma
; CHECK-NEXT: vmv.s.x v24, a0
; CHECK-NEXT: vslideup.vi v16, v24, 31
; CHECK-NEXT: ret
%b = insertelement <64 x i32> %a, i32 %y, i32 63
Expand Down Expand Up @@ -550,9 +548,9 @@ define void @insertelt_c6_v8i64_0_add(ptr %x, ptr %y) {
; CHECK-NEXT: vsetivli zero, 8, e64, m4, ta, ma
; CHECK-NEXT: vle64.v v8, (a0)
; CHECK-NEXT: li a2, 6
; CHECK-NEXT: vsetivli zero, 8, e64, m1, tu, ma
; CHECK-NEXT: vsetivli zero, 8, e64, m4, tu, ma
; CHECK-NEXT: vmv.s.x v8, a2
; CHECK-NEXT: vsetivli zero, 8, e64, m4, ta, ma
; CHECK-NEXT: vsetvli zero, zero, e64, m4, ta, ma
; CHECK-NEXT: vle64.v v12, (a1)
; CHECK-NEXT: vadd.vv v8, v8, v12
; CHECK-NEXT: vse64.v v8, (a0)
Expand Down
72 changes: 36 additions & 36 deletions llvm/test/CodeGen/RISCV/rvv/fixed-vectors-masked-scatter.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1138,11 +1138,11 @@ define void @mscatter_baseidx_v8i8_v8i16(<8 x i16> %val, ptr %base, <8 x i8> %id
; RV64ZVE32F-NEXT: andi a2, a1, 16
; RV64ZVE32F-NEXT: beqz a2, .LBB18_7
; RV64ZVE32F-NEXT: .LBB18_14: # %cond.store7
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
; RV64ZVE32F-NEXT: slli a2, a2, 1
; RV64ZVE32F-NEXT: add a2, a0, a2
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 4
; RV64ZVE32F-NEXT: vse16.v v9, (a2)
; RV64ZVE32F-NEXT: andi a2, a1, 32
Expand Down Expand Up @@ -1271,11 +1271,11 @@ define void @mscatter_baseidx_sext_v8i8_v8i16(<8 x i16> %val, ptr %base, <8 x i8
; RV64ZVE32F-NEXT: andi a2, a1, 16
; RV64ZVE32F-NEXT: beqz a2, .LBB19_7
; RV64ZVE32F-NEXT: .LBB19_14: # %cond.store7
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
; RV64ZVE32F-NEXT: slli a2, a2, 1
; RV64ZVE32F-NEXT: add a2, a0, a2
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why isn't this already a VTYPE toggle from the forward pass? Having the same literal VL and knowing that VL is less than VLMAX for both should be sufficient?

; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 4
; RV64ZVE32F-NEXT: vse16.v v9, (a2)
; RV64ZVE32F-NEXT: andi a2, a1, 32
Expand Down Expand Up @@ -1408,12 +1408,12 @@ define void @mscatter_baseidx_zext_v8i8_v8i16(<8 x i16> %val, ptr %base, <8 x i8
; RV64ZVE32F-NEXT: andi a2, a1, 16
; RV64ZVE32F-NEXT: beqz a2, .LBB20_7
; RV64ZVE32F-NEXT: .LBB20_14: # %cond.store7
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
; RV64ZVE32F-NEXT: andi a2, a2, 255
; RV64ZVE32F-NEXT: slli a2, a2, 1
; RV64ZVE32F-NEXT: add a2, a0, a2
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 4
; RV64ZVE32F-NEXT: vse16.v v9, (a2)
; RV64ZVE32F-NEXT: andi a2, a1, 32
Expand Down Expand Up @@ -2043,11 +2043,11 @@ define void @mscatter_baseidx_v8i8_v8i32(<8 x i32> %val, ptr %base, <8 x i8> %id
; RV64ZVE32F-NEXT: andi a2, a1, 16
; RV64ZVE32F-NEXT: beqz a2, .LBB29_7
; RV64ZVE32F-NEXT: .LBB29_14: # %cond.store7
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
; RV64ZVE32F-NEXT: slli a2, a2, 2
; RV64ZVE32F-NEXT: add a2, a0, a2
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
; RV64ZVE32F-NEXT: andi a2, a1, 32
Expand Down Expand Up @@ -2175,11 +2175,11 @@ define void @mscatter_baseidx_sext_v8i8_v8i32(<8 x i32> %val, ptr %base, <8 x i8
; RV64ZVE32F-NEXT: andi a2, a1, 16
; RV64ZVE32F-NEXT: beqz a2, .LBB30_7
; RV64ZVE32F-NEXT: .LBB30_14: # %cond.store7
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
; RV64ZVE32F-NEXT: slli a2, a2, 2
; RV64ZVE32F-NEXT: add a2, a0, a2
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
; RV64ZVE32F-NEXT: andi a2, a1, 32
Expand Down Expand Up @@ -2314,12 +2314,12 @@ define void @mscatter_baseidx_zext_v8i8_v8i32(<8 x i32> %val, ptr %base, <8 x i8
; RV64ZVE32F-NEXT: andi a2, a1, 16
; RV64ZVE32F-NEXT: beqz a2, .LBB31_7
; RV64ZVE32F-NEXT: .LBB31_14: # %cond.store7
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
; RV64ZVE32F-NEXT: andi a2, a2, 255
; RV64ZVE32F-NEXT: slli a2, a2, 2
; RV64ZVE32F-NEXT: add a2, a0, a2
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
; RV64ZVE32F-NEXT: andi a2, a1, 32
Expand Down Expand Up @@ -2451,11 +2451,11 @@ define void @mscatter_baseidx_v8i16_v8i32(<8 x i32> %val, ptr %base, <8 x i16> %
; RV64ZVE32F-NEXT: andi a2, a1, 16
; RV64ZVE32F-NEXT: beqz a2, .LBB32_7
; RV64ZVE32F-NEXT: .LBB32_14: # %cond.store7
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
; RV64ZVE32F-NEXT: slli a2, a2, 2
; RV64ZVE32F-NEXT: add a2, a0, a2
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
; RV64ZVE32F-NEXT: andi a2, a1, 32
Expand Down Expand Up @@ -2584,11 +2584,11 @@ define void @mscatter_baseidx_sext_v8i16_v8i32(<8 x i32> %val, ptr %base, <8 x i
; RV64ZVE32F-NEXT: andi a2, a1, 16
; RV64ZVE32F-NEXT: beqz a2, .LBB33_7
; RV64ZVE32F-NEXT: .LBB33_14: # %cond.store7
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
; RV64ZVE32F-NEXT: slli a2, a2, 2
; RV64ZVE32F-NEXT: add a2, a0, a2
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
; RV64ZVE32F-NEXT: andi a2, a1, 32
Expand Down Expand Up @@ -2724,12 +2724,12 @@ define void @mscatter_baseidx_zext_v8i16_v8i32(<8 x i32> %val, ptr %base, <8 x i
; RV64ZVE32F-NEXT: andi a3, a2, 16
; RV64ZVE32F-NEXT: beqz a3, .LBB34_7
; RV64ZVE32F-NEXT: .LBB34_14: # %cond.store7
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
; RV64ZVE32F-NEXT: vmv.x.s a3, v11
; RV64ZVE32F-NEXT: and a3, a3, a1
; RV64ZVE32F-NEXT: slli a3, a3, 2
; RV64ZVE32F-NEXT: add a3, a0, a3
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
; RV64ZVE32F-NEXT: vse32.v v12, (a3)
; RV64ZVE32F-NEXT: andi a3, a2, 32
Expand Down Expand Up @@ -6393,11 +6393,11 @@ define void @mscatter_baseidx_v8i8_v8f16(<8 x half> %val, ptr %base, <8 x i8> %i
; RV64ZVE32F-NEXT: andi a2, a1, 16
; RV64ZVE32F-NEXT: beqz a2, .LBB58_7
; RV64ZVE32F-NEXT: .LBB58_14: # %cond.store7
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
; RV64ZVE32F-NEXT: slli a2, a2, 1
; RV64ZVE32F-NEXT: add a2, a0, a2
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 4
; RV64ZVE32F-NEXT: vse16.v v9, (a2)
; RV64ZVE32F-NEXT: andi a2, a1, 32
Expand Down Expand Up @@ -6526,11 +6526,11 @@ define void @mscatter_baseidx_sext_v8i8_v8f16(<8 x half> %val, ptr %base, <8 x i
; RV64ZVE32F-NEXT: andi a2, a1, 16
; RV64ZVE32F-NEXT: beqz a2, .LBB59_7
; RV64ZVE32F-NEXT: .LBB59_14: # %cond.store7
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
; RV64ZVE32F-NEXT: slli a2, a2, 1
; RV64ZVE32F-NEXT: add a2, a0, a2
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 4
; RV64ZVE32F-NEXT: vse16.v v9, (a2)
; RV64ZVE32F-NEXT: andi a2, a1, 32
Expand Down Expand Up @@ -6663,12 +6663,12 @@ define void @mscatter_baseidx_zext_v8i8_v8f16(<8 x half> %val, ptr %base, <8 x i
; RV64ZVE32F-NEXT: andi a2, a1, 16
; RV64ZVE32F-NEXT: beqz a2, .LBB60_7
; RV64ZVE32F-NEXT: .LBB60_14: # %cond.store7
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
; RV64ZVE32F-NEXT: vmv.x.s a2, v10
; RV64ZVE32F-NEXT: andi a2, a2, 255
; RV64ZVE32F-NEXT: slli a2, a2, 1
; RV64ZVE32F-NEXT: add a2, a0, a2
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
; RV64ZVE32F-NEXT: vsetvli zero, zero, e16, m1, ta, ma
; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 4
; RV64ZVE32F-NEXT: vse16.v v9, (a2)
; RV64ZVE32F-NEXT: andi a2, a1, 32
Expand Down Expand Up @@ -7249,11 +7249,11 @@ define void @mscatter_baseidx_v8i8_v8f32(<8 x float> %val, ptr %base, <8 x i8> %
; RV64ZVE32F-NEXT: andi a2, a1, 16
; RV64ZVE32F-NEXT: beqz a2, .LBB68_7
; RV64ZVE32F-NEXT: .LBB68_14: # %cond.store7
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
; RV64ZVE32F-NEXT: slli a2, a2, 2
; RV64ZVE32F-NEXT: add a2, a0, a2
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
Expand Down Expand Up @@ -7385,11 +7385,11 @@ define void @mscatter_baseidx_sext_v8i8_v8f32(<8 x float> %val, ptr %base, <8 x
; RV64ZVE32F-NEXT: andi a2, a1, 16
; RV64ZVE32F-NEXT: beqz a2, .LBB69_7
; RV64ZVE32F-NEXT: .LBB69_14: # %cond.store7
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
; RV64ZVE32F-NEXT: slli a2, a2, 2
; RV64ZVE32F-NEXT: add a2, a0, a2
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
Expand Down Expand Up @@ -7528,12 +7528,12 @@ define void @mscatter_baseidx_zext_v8i8_v8f32(<8 x float> %val, ptr %base, <8 x
; RV64ZVE32F-NEXT: andi a2, a1, 16
; RV64ZVE32F-NEXT: beqz a2, .LBB70_7
; RV64ZVE32F-NEXT: .LBB70_14: # %cond.store7
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf2, ta, ma
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
; RV64ZVE32F-NEXT: andi a2, a2, 255
; RV64ZVE32F-NEXT: slli a2, a2, 2
; RV64ZVE32F-NEXT: add a2, a0, a2
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
Expand Down Expand Up @@ -7669,11 +7669,11 @@ define void @mscatter_baseidx_v8i16_v8f32(<8 x float> %val, ptr %base, <8 x i16>
; RV64ZVE32F-NEXT: andi a2, a1, 16
; RV64ZVE32F-NEXT: beqz a2, .LBB71_7
; RV64ZVE32F-NEXT: .LBB71_14: # %cond.store7
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
; RV64ZVE32F-NEXT: slli a2, a2, 2
; RV64ZVE32F-NEXT: add a2, a0, a2
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
Expand Down Expand Up @@ -7806,11 +7806,11 @@ define void @mscatter_baseidx_sext_v8i16_v8f32(<8 x float> %val, ptr %base, <8 x
; RV64ZVE32F-NEXT: andi a2, a1, 16
; RV64ZVE32F-NEXT: beqz a2, .LBB72_7
; RV64ZVE32F-NEXT: .LBB72_14: # %cond.store7
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
; RV64ZVE32F-NEXT: vmv.x.s a2, v11
; RV64ZVE32F-NEXT: slli a2, a2, 2
; RV64ZVE32F-NEXT: add a2, a0, a2
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
; RV64ZVE32F-NEXT: vse32.v v12, (a2)
Expand Down Expand Up @@ -7950,12 +7950,12 @@ define void @mscatter_baseidx_zext_v8i16_v8f32(<8 x float> %val, ptr %base, <8 x
; RV64ZVE32F-NEXT: andi a3, a2, 16
; RV64ZVE32F-NEXT: beqz a3, .LBB73_7
; RV64ZVE32F-NEXT: .LBB73_14: # %cond.store7
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, mf2, ta, ma
; RV64ZVE32F-NEXT: vsetivli zero, 1, e16, m1, ta, ma
; RV64ZVE32F-NEXT: vmv.x.s a3, v11
; RV64ZVE32F-NEXT: and a3, a3, a1
; RV64ZVE32F-NEXT: slli a3, a3, 2
; RV64ZVE32F-NEXT: add a3, a0, a3
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
; RV64ZVE32F-NEXT: vslidedown.vi v12, v8, 4
; RV64ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
; RV64ZVE32F-NEXT: vse32.v v12, (a3)
Expand Down
Loading