[llvm] [RISCV] Sink instructions so AVL dominates in RISCVVLOptimizer (PR #184155)
via llvm-commits
llvm-commits at lists.llvm.org
Mon Mar 2 07:34:01 PST 2026
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-backend-risc-v
Author: Luke Lau (lukel97)
<details>
<summary>Changes</summary>
The last feature that RISCVVectorPeephole::tryToReduceVL has that RISCVVLOptimizer doesn't is that if the vl to reduce to doesn't dominate a vector pseudo, it can sink the pseudo below the vl definition if it's safe to do so.
This PR shares the logic to check for physical register clobbering in RISCVInstrInfo, but there may be a better place for it.
The DemandedVLs DenseMap needs to be switched to a MapVector to get deterministic ordering, since the order in which we sink instructions can affect the resulting codegen.
This helps removes a few vsetvli toggles in llvm-test-suite.
---
Patch is 217.84 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/184155.diff
36 Files Affected:
- (modified) llvm/lib/Target/RISCV/RISCVInstrInfo.cpp (+27)
- (modified) llvm/lib/Target/RISCV/RISCVInstrInfo.h (+5)
- (modified) llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp (+15-5)
- (modified) llvm/lib/Target/RISCV/RISCVVectorPeephole.cpp (+1-29)
- (modified) llvm/test/CodeGen/RISCV/GlobalISel/rvv/vadd.ll (+4-4)
- (modified) llvm/test/CodeGen/RISCV/GlobalISel/rvv/vfadd.ll (+6-6)
- (modified) llvm/test/CodeGen/RISCV/regalloc-last-chance-recoloring-failure.ll (+2-4)
- (modified) llvm/test/CodeGen/RISCV/rvv/access-fixed-objects-by-rvv.ll (+12-11)
- (modified) llvm/test/CodeGen/RISCV/rvv/combine-reduce-add-to-vcpop.ll (+1-2)
- (modified) llvm/test/CodeGen/RISCV/rvv/fmaximum-vp.ll (+13-13)
- (modified) llvm/test/CodeGen/RISCV/rvv/fminimum-vp.ll (+13-13)
- (modified) llvm/test/CodeGen/RISCV/rvv/fshr-fshl-vp.ll (+28-24)
- (modified) llvm/test/CodeGen/RISCV/rvv/memory-args.ll (+4-3)
- (modified) llvm/test/CodeGen/RISCV/rvv/nontemporal-vp-scalable.ll (+730-700)
- (modified) llvm/test/CodeGen/RISCV/rvv/setcc-fp-vp.ll (+13-11)
- (modified) llvm/test/CodeGen/RISCV/rvv/setcc-int-vp.ll (+41-39)
- (modified) llvm/test/CodeGen/RISCV/rvv/vadd-vp.ll (+10-10)
- (modified) llvm/test/CodeGen/RISCV/rvv/vector-splice.ll (+6-6)
- (modified) llvm/test/CodeGen/RISCV/rvv/vfma-vp.ll (+62-86)
- (modified) llvm/test/CodeGen/RISCV/rvv/vfmuladd-vp.ll (+58-78)
- (modified) llvm/test/CodeGen/RISCV/rvv/vfptrunc-vp.ll (+6-3)
- (modified) llvm/test/CodeGen/RISCV/rvv/vl-opt.mir (+128-1)
- (modified) llvm/test/CodeGen/RISCV/rvv/vmax-vp.ll (+10-10)
- (modified) llvm/test/CodeGen/RISCV/rvv/vmaxu-vp.ll (+10-10)
- (modified) llvm/test/CodeGen/RISCV/rvv/vmin-vp.ll (+10-10)
- (modified) llvm/test/CodeGen/RISCV/rvv/vminu-vp.ll (+10-10)
- (modified) llvm/test/CodeGen/RISCV/rvv/vp-vector-interleaved-access.ll (+5-5)
- (modified) llvm/test/CodeGen/RISCV/rvv/vpgather-sdnode.ll (+14-12)
- (modified) llvm/test/CodeGen/RISCV/rvv/vpmerge-sdnode.ll (+30-27)
- (modified) llvm/test/CodeGen/RISCV/rvv/vsadd-vp.ll (+10-10)
- (modified) llvm/test/CodeGen/RISCV/rvv/vsaddu-vp.ll (+10-10)
- (modified) llvm/test/CodeGen/RISCV/rvv/vselect-vp.ll (+46-39)
- (modified) llvm/test/CodeGen/RISCV/rvv/vsetvli-insert.ll (+3-3)
- (modified) llvm/test/CodeGen/RISCV/rvv/vssub-vp.ll (+13-13)
- (modified) llvm/test/CodeGen/RISCV/rvv/vssubu-vp.ll (+13-13)
- (modified) llvm/test/CodeGen/RISCV/rvv/vtrunc-vp.ll (+6-3)
``````````diff
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
index fa388a3ec64d8..e7fc3aed6ea7e 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
@@ -5334,3 +5334,30 @@ bool RISCVInstrInfo::requiresNTLHint(const MachineInstr &MI) const {
return true;
}
+
+bool RISCVInstrInfo::isSafeToMove(const MachineInstr &From,
+ const MachineInstr &To) {
+ assert(From.getParent() == To.getParent());
+ SmallVector<Register> PhysUses, PhysDefs;
+ for (const MachineOperand &MO : From.all_uses())
+ if (MO.getReg().isPhysical())
+ PhysUses.push_back(MO.getReg());
+ for (const MachineOperand &MO : From.all_defs())
+ if (MO.getReg().isPhysical())
+ PhysDefs.push_back(MO.getReg());
+ bool SawStore = false;
+ for (auto II = std::next(From.getIterator()); II != To.getIterator(); II++) {
+ for (Register PhysReg : PhysUses)
+ if (II->definesRegister(PhysReg, nullptr))
+ return false;
+ for (Register PhysReg : PhysDefs)
+ if (II->definesRegister(PhysReg, nullptr) ||
+ II->readsRegister(PhysReg, nullptr))
+ return false;
+ if (II->mayStore()) {
+ SawStore = true;
+ break;
+ }
+ }
+ return From.isSafeToMove(SawStore);
+}
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.h b/llvm/lib/Target/RISCV/RISCVInstrInfo.h
index 15d0e4d3e7d52..d3ad5bf0a0763 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.h
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.h
@@ -333,6 +333,11 @@ class RISCVInstrInfo : public RISCVGenInstrInfo {
/// Return true if the instruction requires an NTL hint to be emitted.
bool requiresNTLHint(const MachineInstr &MI) const;
+ /// Return true if moving \p From down to \p To won't cause any physical
+ /// register reads or writes to be clobbered and no visible side effects are
+ /// affected. From and To must be in the same block.
+ static bool isSafeToMove(const MachineInstr &From, const MachineInstr &To);
+
/// Return true if pairing the given load or store may be paired with another.
static bool isPairableLdStInstOpc(unsigned Opc);
diff --git a/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp b/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp
index 99eaf2683e1cc..14f10f9b11128 100644
--- a/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp
+++ b/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp
@@ -92,7 +92,7 @@ class RISCVVLOptimizer : public MachineFunctionPass {
/// For a given instruction, records what elements of it are demanded by
/// downstream users.
- DenseMap<const MachineInstr *, DemandedVL> DemandedVLs;
+ MapVector<const MachineInstr *, DemandedVL> DemandedVLs;
SetVector<const MachineInstr *> Worklist;
/// \returns all vector virtual registers that \p MI uses.
@@ -1255,10 +1255,20 @@ bool RISCVVLOptimizer::tryReduceVL(MachineInstr &MI,
VLOp.ChangeToImmediate(CommonVL.getImm());
return true;
}
- const MachineInstr *VLMI = MRI->getVRegDef(CommonVL.getReg());
- if (!MDT->dominates(VLMI, &MI)) {
- LLVM_DEBUG(dbgs() << " Abort due to VL not dominating.\n");
- return false;
+ MachineInstr *VLMI = MRI->getVRegDef(CommonVL.getReg());
+ auto VLDominates = [this, &VLMI](MachineInstr &MI) {
+ return MDT->dominates(VLMI, &MI);
+ };
+ if (!VLDominates(MI)) {
+ assert(MI.getNumExplicitDefs() == 1);
+ if (VLMI->getParent() == MI.getParent() &&
+ all_of(MRI->use_instructions(MI.getOperand(0).getReg()), VLDominates) &&
+ RISCVInstrInfo::isSafeToMove(MI, *VLMI->getNextNode())) {
+ MI.moveBefore(VLMI->getNextNode());
+ } else {
+ LLVM_DEBUG(dbgs() << " Abort due to VL not dominating.\n");
+ return false;
+ }
}
LLVM_DEBUG(dbgs() << " Reduce VL from " << VLOp << " to "
<< printReg(CommonVL.getReg(), MRI->getTargetRegisterInfo())
diff --git a/llvm/lib/Target/RISCV/RISCVVectorPeephole.cpp b/llvm/lib/Target/RISCV/RISCVVectorPeephole.cpp
index c7c8eeed02aeb..07110c4b7f1ad 100644
--- a/llvm/lib/Target/RISCV/RISCVVectorPeephole.cpp
+++ b/llvm/lib/Target/RISCV/RISCVVectorPeephole.cpp
@@ -530,34 +530,6 @@ bool RISCVVectorPeephole::convertToUnmasked(MachineInstr &MI) const {
return true;
}
-/// Check if it's safe to move From down to To, checking that no physical
-/// registers are clobbered.
-static bool isSafeToMove(const MachineInstr &From, const MachineInstr &To) {
- assert(From.getParent() == To.getParent());
- SmallVector<Register> PhysUses, PhysDefs;
- for (const MachineOperand &MO : From.all_uses())
- if (MO.getReg().isPhysical())
- PhysUses.push_back(MO.getReg());
- for (const MachineOperand &MO : From.all_defs())
- if (MO.getReg().isPhysical())
- PhysDefs.push_back(MO.getReg());
- bool SawStore = false;
- for (auto II = std::next(From.getIterator()); II != To.getIterator(); II++) {
- for (Register PhysReg : PhysUses)
- if (II->definesRegister(PhysReg, nullptr))
- return false;
- for (Register PhysReg : PhysDefs)
- if (II->definesRegister(PhysReg, nullptr) ||
- II->readsRegister(PhysReg, nullptr))
- return false;
- if (II->mayStore()) {
- SawStore = true;
- break;
- }
- }
- return From.isSafeToMove(SawStore);
-}
-
/// Given A and B are in the same MBB, returns true if A comes before B.
static bool dominates(MachineBasicBlock::const_iterator A,
MachineBasicBlock::const_iterator B) {
@@ -585,7 +557,7 @@ bool RISCVVectorPeephole::ensureDominates(const MachineOperand &MO,
MachineInstr *Def = MRI->getVRegDef(MO.getReg());
if (Def->getParent() == Src.getParent() && !dominates(Def, Src)) {
- if (!isSafeToMove(Src, *Def->getNextNode()))
+ if (!RISCVInstrInfo::isSafeToMove(Src, *Def->getNextNode()))
return false;
Src.moveBefore(Def->getNextNode());
}
diff --git a/llvm/test/CodeGen/RISCV/GlobalISel/rvv/vadd.ll b/llvm/test/CodeGen/RISCV/GlobalISel/rvv/vadd.ll
index c3cc472c4706f..e07548f45492a 100644
--- a/llvm/test/CodeGen/RISCV/GlobalISel/rvv/vadd.ll
+++ b/llvm/test/CodeGen/RISCV/GlobalISel/rvv/vadd.ll
@@ -221,8 +221,8 @@ entry:
define <vscale x 64 x i8> @intrinsic_vadd_mask_vv_nxv64i8_nxv64i8_nxv64i8(<vscale x 64 x i8> %0, <vscale x 64 x i8> %1, <vscale x 64 x i8> %2, <vscale x 64 x i1> %3, iXLen %4) nounwind {
; CHECK-LABEL: intrinsic_vadd_mask_vv_nxv64i8_nxv64i8_nxv64i8:
; CHECK: # %bb.0: # %entry
-; CHECK-NEXT: vl8r.v v24, (a0)
; CHECK-NEXT: vsetvli zero, a1, e8, m8, ta, mu
+; CHECK-NEXT: vle8.v v24, (a0)
; CHECK-NEXT: vadd.vv v8, v16, v24, v0.t
; CHECK-NEXT: ret
entry:
@@ -420,8 +420,8 @@ entry:
define <vscale x 32 x i16> @intrinsic_vadd_mask_vv_nxv32i16_nxv32i16_nxv32i16(<vscale x 32 x i16> %0, <vscale x 32 x i16> %1, <vscale x 32 x i16> %2, <vscale x 32 x i1> %3, iXLen %4) nounwind {
; CHECK-LABEL: intrinsic_vadd_mask_vv_nxv32i16_nxv32i16_nxv32i16:
; CHECK: # %bb.0: # %entry
-; CHECK-NEXT: vl8re16.v v24, (a0)
; CHECK-NEXT: vsetvli zero, a1, e16, m8, ta, mu
+; CHECK-NEXT: vle16.v v24, (a0)
; CHECK-NEXT: vadd.vv v8, v16, v24, v0.t
; CHECK-NEXT: ret
entry:
@@ -586,8 +586,8 @@ entry:
define <vscale x 16 x i32> @intrinsic_vadd_mask_vv_nxv16i32_nxv16i32_nxv16i32(<vscale x 16 x i32> %0, <vscale x 16 x i32> %1, <vscale x 16 x i32> %2, <vscale x 16 x i1> %3, iXLen %4) nounwind {
; CHECK-LABEL: intrinsic_vadd_mask_vv_nxv16i32_nxv16i32_nxv16i32:
; CHECK: # %bb.0: # %entry
-; CHECK-NEXT: vl8re32.v v24, (a0)
; CHECK-NEXT: vsetvli zero, a1, e32, m8, ta, mu
+; CHECK-NEXT: vle32.v v24, (a0)
; CHECK-NEXT: vadd.vv v8, v16, v24, v0.t
; CHECK-NEXT: ret
entry:
@@ -719,8 +719,8 @@ entry:
define <vscale x 8 x i64> @intrinsic_vadd_mask_vv_nxv8i64_nxv8i64_nxv8i64(<vscale x 8 x i64> %0, <vscale x 8 x i64> %1, <vscale x 8 x i64> %2, <vscale x 8 x i1> %3, iXLen %4) nounwind {
; CHECK-LABEL: intrinsic_vadd_mask_vv_nxv8i64_nxv8i64_nxv8i64:
; CHECK: # %bb.0: # %entry
-; CHECK-NEXT: vl8re64.v v24, (a0)
; CHECK-NEXT: vsetvli zero, a1, e64, m8, ta, mu
+; CHECK-NEXT: vle64.v v24, (a0)
; CHECK-NEXT: vadd.vv v8, v16, v24, v0.t
; CHECK-NEXT: ret
entry:
diff --git a/llvm/test/CodeGen/RISCV/GlobalISel/rvv/vfadd.ll b/llvm/test/CodeGen/RISCV/GlobalISel/rvv/vfadd.ll
index 1dd3a831903b5..d53c3609f88ef 100644
--- a/llvm/test/CodeGen/RISCV/GlobalISel/rvv/vfadd.ll
+++ b/llvm/test/CodeGen/RISCV/GlobalISel/rvv/vfadd.ll
@@ -216,9 +216,9 @@ entry:
define <vscale x 32 x half> @intrinsic_vfadd_mask_vv_nxv32f16_nxv32f16_nxv32f16(<vscale x 32 x half> %0, <vscale x 32 x half> %1, <vscale x 32 x half> %2, <vscale x 32 x i1> %3, iXLen %4) nounwind {
; CHECK-LABEL: intrinsic_vfadd_mask_vv_nxv32f16_nxv32f16_nxv32f16:
; CHECK: # %bb.0: # %entry
-; CHECK-NEXT: vl8re16.v v24, (a0)
-; CHECK-NEXT: fsrmi a0, 0
; CHECK-NEXT: vsetvli zero, a1, e16, m8, ta, mu
+; CHECK-NEXT: vle16.v v24, (a0)
+; CHECK-NEXT: fsrmi a0, 0
; CHECK-NEXT: vfadd.vv v8, v16, v24, v0.t
; CHECK-NEXT: fsrm a0
; CHECK-NEXT: ret
@@ -402,9 +402,9 @@ entry:
define <vscale x 16 x float> @intrinsic_vfadd_mask_vv_nxv16f32_nxv16f32_nxv16f32(<vscale x 16 x float> %0, <vscale x 16 x float> %1, <vscale x 16 x float> %2, <vscale x 16 x i1> %3, iXLen %4) nounwind {
; CHECK-LABEL: intrinsic_vfadd_mask_vv_nxv16f32_nxv16f32_nxv16f32:
; CHECK: # %bb.0: # %entry
-; CHECK-NEXT: vl8re32.v v24, (a0)
-; CHECK-NEXT: fsrmi a0, 0
; CHECK-NEXT: vsetvli zero, a1, e32, m8, ta, mu
+; CHECK-NEXT: vle32.v v24, (a0)
+; CHECK-NEXT: fsrmi a0, 0
; CHECK-NEXT: vfadd.vv v8, v16, v24, v0.t
; CHECK-NEXT: fsrm a0
; CHECK-NEXT: ret
@@ -551,9 +551,9 @@ entry:
define <vscale x 8 x double> @intrinsic_vfadd_mask_vv_nxv8f64_nxv8f64_nxv8f64(<vscale x 8 x double> %0, <vscale x 8 x double> %1, <vscale x 8 x double> %2, <vscale x 8 x i1> %3, iXLen %4) nounwind {
; CHECK-LABEL: intrinsic_vfadd_mask_vv_nxv8f64_nxv8f64_nxv8f64:
; CHECK: # %bb.0: # %entry
-; CHECK-NEXT: vl8re64.v v24, (a0)
-; CHECK-NEXT: fsrmi a0, 0
; CHECK-NEXT: vsetvli zero, a1, e64, m8, ta, mu
+; CHECK-NEXT: vle64.v v24, (a0)
+; CHECK-NEXT: fsrmi a0, 0
; CHECK-NEXT: vfadd.vv v8, v16, v24, v0.t
; CHECK-NEXT: fsrm a0
; CHECK-NEXT: ret
diff --git a/llvm/test/CodeGen/RISCV/regalloc-last-chance-recoloring-failure.ll b/llvm/test/CodeGen/RISCV/regalloc-last-chance-recoloring-failure.ll
index ae797de91b857..c399a1c3ca07f 100644
--- a/llvm/test/CodeGen/RISCV/regalloc-last-chance-recoloring-failure.ll
+++ b/llvm/test/CodeGen/RISCV/regalloc-last-chance-recoloring-failure.ll
@@ -24,8 +24,6 @@ define void @last_chance_recoloring_failure() {
; CHECK-NEXT: sub sp, sp, a0
; CHECK-NEXT: .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x20, 0x22, 0x11, 0x10, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 32 + 16 * vlenb
; CHECK-NEXT: li a0, 55
-; CHECK-NEXT: vsetvli a2, zero, e8, m2, ta, ma
-; CHECK-NEXT: vmclr.m v0
; CHECK-NEXT: vsetvli zero, a0, e16, m4, ta, ma
; CHECK-NEXT: vloxseg2ei32.v v16, (a1), v8
; CHECK-NEXT: csrr a0, vlenb
@@ -35,6 +33,7 @@ define void @last_chance_recoloring_failure() {
; CHECK-NEXT: vs8r.v v16, (a0) # vscale x 64-byte Folded Spill
; CHECK-NEXT: li s0, 36
; CHECK-NEXT: vsetvli zero, s0, e16, m4, ta, ma
+; CHECK-NEXT: vmclr.m v0
; CHECK-NEXT: vfwadd.vv v16, v8, v12, v0.t
; CHECK-NEXT: addi a0, sp, 16
; CHECK-NEXT: vs8r.v v16, (a0) # vscale x 64-byte Folded Spill
@@ -76,8 +75,6 @@ define void @last_chance_recoloring_failure() {
; SUBREGLIVENESS-NEXT: sub sp, sp, a0
; SUBREGLIVENESS-NEXT: .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x20, 0x22, 0x11, 0x10, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 32 + 16 * vlenb
; SUBREGLIVENESS-NEXT: li a0, 55
-; SUBREGLIVENESS-NEXT: vsetvli a2, zero, e8, m2, ta, ma
-; SUBREGLIVENESS-NEXT: vmclr.m v0
; SUBREGLIVENESS-NEXT: vsetvli zero, a0, e16, m4, ta, ma
; SUBREGLIVENESS-NEXT: vloxseg2ei32.v v16, (a1), v8
; SUBREGLIVENESS-NEXT: csrr a0, vlenb
@@ -87,6 +84,7 @@ define void @last_chance_recoloring_failure() {
; SUBREGLIVENESS-NEXT: vs8r.v v16, (a0) # vscale x 64-byte Folded Spill
; SUBREGLIVENESS-NEXT: li s0, 36
; SUBREGLIVENESS-NEXT: vsetvli zero, s0, e16, m4, ta, ma
+; SUBREGLIVENESS-NEXT: vmclr.m v0
; SUBREGLIVENESS-NEXT: vfwadd.vv v16, v8, v12, v0.t
; SUBREGLIVENESS-NEXT: addi a0, sp, 16
; SUBREGLIVENESS-NEXT: vs8r.v v16, (a0) # vscale x 64-byte Folded Spill
diff --git a/llvm/test/CodeGen/RISCV/rvv/access-fixed-objects-by-rvv.ll b/llvm/test/CodeGen/RISCV/rvv/access-fixed-objects-by-rvv.ll
index 4aaaa88db8b6e..dfde5607eb542 100644
--- a/llvm/test/CodeGen/RISCV/rvv/access-fixed-objects-by-rvv.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/access-fixed-objects-by-rvv.ll
@@ -30,13 +30,13 @@ define <vscale x 1 x i64> @access_fixed_and_vector_objects(ptr %val) {
; RV64IV-NEXT: csrr a0, vlenb
; RV64IV-NEXT: sub sp, sp, a0
; RV64IV-NEXT: .cfi_escape 0x0f, 0x0e, 0x72, 0x00, 0x11, 0x90, 0x04, 0x22, 0x11, 0x01, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 528 + 1 * vlenb
-; RV64IV-NEXT: addi a0, sp, 8
-; RV64IV-NEXT: vl1re64.v v8, (a0)
-; RV64IV-NEXT: addi a0, sp, 528
-; RV64IV-NEXT: vl1re64.v v9, (a0)
; RV64IV-NEXT: ld a0, 520(sp)
+; RV64IV-NEXT: addi a1, sp, 8
+; RV64IV-NEXT: addi a2, sp, 528
; RV64IV-NEXT: vsetvli zero, a0, e64, m1, ta, ma
-; RV64IV-NEXT: vadd.vv v8, v8, v9
+; RV64IV-NEXT: vle64.v v8, (a2)
+; RV64IV-NEXT: vle64.v v9, (a1)
+; RV64IV-NEXT: vadd.vv v8, v9, v8
; RV64IV-NEXT: csrr a0, vlenb
; RV64IV-NEXT: add sp, sp, a0
; RV64IV-NEXT: .cfi_def_cfa sp, 528
@@ -76,13 +76,14 @@ define <vscale x 1 x i64> @probe_fixed_and_vector_objects(ptr %val, <vscale x 1
; RV64IV-NEXT: .cfi_def_cfa_register sp
; RV64IV-NEXT: sub sp, sp, t1
; RV64IV-NEXT: .cfi_escape 0x0f, 0x0e, 0x72, 0x00, 0x11, 0x90, 0x04, 0x22, 0x11, 0x01, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 528 + 1 * vlenb
-; RV64IV-NEXT: addi a0, sp, 8
-; RV64IV-NEXT: vl1re64.v v9, (a0)
-; RV64IV-NEXT: addi a0, sp, 528
-; RV64IV-NEXT: vl1re64.v v10, (a0)
; RV64IV-NEXT: ld a0, 520(sp)
-; RV64IV-NEXT: vsetvli zero, a0, e64, m1, tu, ma
-; RV64IV-NEXT: vadd.vv v8, v9, v10
+; RV64IV-NEXT: addi a1, sp, 8
+; RV64IV-NEXT: addi a2, sp, 528
+; RV64IV-NEXT: vsetvli zero, a0, e64, m1, ta, ma
+; RV64IV-NEXT: vle64.v v9, (a2)
+; RV64IV-NEXT: vle64.v v10, (a1)
+; RV64IV-NEXT: vsetvli zero, zero, e64, m1, tu, ma
+; RV64IV-NEXT: vadd.vv v8, v10, v9
; RV64IV-NEXT: csrr a0, vlenb
; RV64IV-NEXT: add sp, sp, a0
; RV64IV-NEXT: .cfi_def_cfa sp, 528
diff --git a/llvm/test/CodeGen/RISCV/rvv/combine-reduce-add-to-vcpop.ll b/llvm/test/CodeGen/RISCV/rvv/combine-reduce-add-to-vcpop.ll
index 96252f070a580..550869b10d3f3 100644
--- a/llvm/test/CodeGen/RISCV/rvv/combine-reduce-add-to-vcpop.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/combine-reduce-add-to-vcpop.ll
@@ -191,11 +191,10 @@ define i32 @test_nxv1i1(<vscale x 1 x i1> %x) {
;
; ZVE-LABEL: test_nxv1i1:
; ZVE: # %bb.0: # %entry
-; ZVE-NEXT: vsetvli a0, zero, e32, m1, ta, ma
-; ZVE-NEXT: vmv.v.i v8, 0
; ZVE-NEXT: csrr a0, vlenb
; ZVE-NEXT: srli a0, a0, 3
; ZVE-NEXT: vsetvli zero, a0, e32, m1, ta, ma
+; ZVE-NEXT: vmv.v.i v8, 0
; ZVE-NEXT: vmerge.vim v8, v8, 1, v0
; ZVE-NEXT: vsetivli zero, 1, e32, m1, ta, ma
; ZVE-NEXT: vmv.s.x v9, zero
diff --git a/llvm/test/CodeGen/RISCV/rvv/fmaximum-vp.ll b/llvm/test/CodeGen/RISCV/rvv/fmaximum-vp.ll
index c432204b078f5..b95e138dfd3da 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fmaximum-vp.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fmaximum-vp.ll
@@ -1301,17 +1301,17 @@ define <vscale x 16 x double> @vfmax_vv_nxv16f64(<vscale x 16 x double> %va, <vs
; CHECK-NEXT: addi a1, a1, 16
; CHECK-NEXT: vs8r.v v8, (a1) # vscale x 64-byte Folded Spill
; CHECK-NEXT: csrr a1, vlenb
-; CHECK-NEXT: slli a3, a1, 3
-; CHECK-NEXT: srli a4, a1, 3
-; CHECK-NEXT: vslidedown.vx v6, v0, a4
+; CHECK-NEXT: srli a3, a1, 3
; CHECK-NEXT: sub a4, a2, a1
-; CHECK-NEXT: add a3, a0, a3
-; CHECK-NEXT: vl8re64.v v24, (a3)
+; CHECK-NEXT: vslidedown.vx v6, v0, a3
; CHECK-NEXT: sltu a3, a2, a4
; CHECK-NEXT: addi a3, a3, -1
; CHECK-NEXT: and a3, a3, a4
-; CHECK-NEXT: vmv1r.v v0, v6
+; CHECK-NEXT: slli a4, a1, 3
+; CHECK-NEXT: add a4, a0, a4
; CHECK-NEXT: vsetvli zero, a3, e64, m8, ta, ma
+; CHECK-NEXT: vle64.v v24, (a4)
+; CHECK-NEXT: vmv1r.v v0, v6
; CHECK-NEXT: vmfeq.vv v8, v16, v16, v0.t
; CHECK-NEXT: vmv1r.v v0, v8
; CHECK-NEXT: vmv8r.v v8, v16
@@ -1377,14 +1377,14 @@ define <vscale x 16 x double> @vfmax_vv_nxv16f64_unmasked(<vscale x 16 x double>
; CHECK-NEXT: addi a1, a1, 16
; CHECK-NEXT: vs8r.v v8, (a1) # vscale x 64-byte Folded Spill
; CHECK-NEXT: csrr a1, vlenb
-; CHECK-NEXT: slli a3, a1, 3
-; CHECK-NEXT: sub a4, a2, a1
-; CHECK-NEXT: add a3, a0, a3
-; CHECK-NEXT: vl8re64.v v24, (a3)
-; CHECK-NEXT: sltu a3, a2, a4
-; CHECK-NEXT: addi a3, a3, -1
-; CHECK-NEXT: and a3, a3, a4
+; CHECK-NEXT: sub a3, a2, a1
+; CHECK-NEXT: sltu a4, a2, a3
+; CHECK-NEXT: addi a4, a4, -1
+; CHECK-NEXT: and a3, a4, a3
+; CHECK-NEXT: slli a4, a1, 3
+; CHECK-NEXT: add a4, a0, a4
; CHECK-NEXT: vsetvli zero, a3, e64, m8, ta, ma
+; CHECK-NEXT: vle64.v v24, (a4)
; CHECK-NEXT: vmfeq.vv v0, v16, v16
; CHECK-NEXT: vmfeq.vv v7, v24, v24
; CHECK-NEXT: vl8re64.v v8, (a0)
diff --git a/llvm/test/CodeGen/RISCV/rvv/fminimum-vp.ll b/llvm/test/CodeGen/RISCV/rvv/fminimum-vp.ll
index 27816a862cc1b..06c9cffa4eb1e 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fminimum-vp.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fminimum-vp.ll
@@ -1301,17 +1301,17 @@ define <vscale x 16 x double> @vfmin_vv_nxv16f64(<vscale x 16 x double> %va, <vs
; CHECK-NEXT: addi a1, a1, 16
; CHECK-NEXT: vs8r.v v8, (a1) # vscale x 64-byte Folded Spill
; CHECK-NEXT: csrr a1, vlenb
-; CHECK-NEXT: slli a3, a1, 3
-; CHECK-NEXT: srli a4, a1, 3
-; CHECK-NEXT: vslidedown.vx v6, v0, a4
+; CHECK-NEXT: srli a3, a1, 3
; CHECK-NEXT: sub a4, a2, a1
-; CHECK-NEXT: add a3, a0, a3
-; CHECK-NEXT: vl8re64.v v24, (a3)
+; CHECK-NEXT: vslidedown.vx v6, v0, a3
; CHECK-NEXT: sltu a3, a2, a4
; CHECK-NEXT: addi a3, a3, -1
; CHECK-NEXT: and a3, a3, a4
-; CHECK-NEXT: vmv1r.v v0, v6
+; CHECK-NEXT: slli a4, a1, 3
+; CHECK-NEXT: add a4, a0, a4
; CHECK-NEXT: vsetvli zero, a3, e64, m8, ta, ma
+; CHECK-NEXT: vle64.v v24, (a4)
+; CHECK-NEXT: vmv1r.v v0, v6
; CHECK-NEXT: vmfeq.vv v8, v16, v16, v0.t
; CHECK-NEXT: vmv1r.v v0, v8
; CHECK-NEXT: vmv8r.v v8, v16
@@ -1377,14 +1377,14 @@ define <vscale x 16 x double> @vfmin_vv_nxv16f64_unmasked(<vscale x 16 x double>
; CHECK-NEXT: addi a1, a1, 16
; CHECK-NEXT: vs8r.v v8, (a1) # vscale x 64-byte Folded Spill
; CHECK-NEXT: csrr a1, vlenb
-; CHECK-NEXT: slli a3, a1, 3
-; CHECK-NEXT: sub a4, a2, a1
-; CHECK-NEXT: add a3, a0, a3
-; CHECK-NEXT: vl8re64.v v24, (a3)
-; CHECK-NEXT: sltu a3, a2, a4
-; CHECK-NEXT: addi a3, a3, -1
-; CHECK-NEXT: and a3, a3, a4
+; CHECK-NEXT: sub a3, a2, a1
+; CHECK-NEXT: sltu a4, a2, a3
+; CHECK-NEXT: addi a4, a4, -1
+; CHECK-NEXT: and a3, a4, a3
+; CHECK-NEXT: slli a4, a1, 3
+; CHECK-NEXT: add a4, a0, a4
; CHECK-NEXT: vsetvli zero, a3, e64, m8, ta, ma
+; CHECK-NEXT: vle64.v v24, (a4)
; CHECK-NEXT: vmfeq.vv v0, v16, v16
; CHECK-NEXT: vmfeq.vv v7, v24, v24
; CHECK-NEXT: vl8re64.v v8, (a0)
diff --git a/llvm/test/CodeGen/RISCV/rvv/fshr-fshl-vp.ll b/llvm/test/CodeGen/RISCV/rvv/fshr-fshl-vp.ll
index 576753d1d7821..6303dccbdfb91 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fshr-fshl-vp.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fshr-fshl-vp.ll
@@ -940,7 +940,7 @@ define <vscale x 16 x i64> @fshr_v16i64(<vscale x 16 x i64> %a, <vscale x 16 x i
; CHECK-NEXT: slli a1, a1, 5
; CHECK-NEXT: sub sp, sp, a1
; CHECK-NEXT: .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x20, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 32 * vlenb
-; CHECK-NEXT: vsetvli a1, zero, e8, mf4, ta, ma
+; CHECK-NEXT: vsetivli zero, 1, e8, m1, ta, ma
; CHECK-NEXT: vmv1r.v v7, v0
; CHECK-NEXT: csrr a1, vlenb
; CHECK-NEXT: slli a1, a1, 4
@@ -956,23 +956,25 @@ define <vscale x 16 x i64> @fshr_v16i64(<vscale x 16 x i64> %a, <vscale x 16 x i
;...
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/184155
More information about the llvm-commits
mailing list