[RISCV] Move RISCVInsertVSETVLI to after phi elimination

Split off from #70549, this patch moves RISCVInsertVSETVLI to after phi elimination where we exit SSA and need to move to LiveVariables.

The motivation for splitting this off is to avoid the large scheduling diffs from moving completely to after regalloc, and instead focus on converting the pass to work on LiveIntervals.
This limits the test diff to vsetvlis moving between csr and fsr instructions, due to RISCVInsertVSETVLI now taking place after RISCVInsertReadWriteCSR and RISCVInsertWriteVXRM.

The two main changes required are updating VSETVLIInfo to store VNInfos instead of MachineInstrs, which allows us to still check for PHI defs in needVSETVLIPHI, and fixing up the live intervals of any AVL operands after inserting new instructions.

On O3 the pass is inserted after the register coalescer, otherwise we end up with a bunch of COPYs around eliminated PHIs that trip up needVSETVLIPHI.

Note that this manually fixes up the LiveIntervals instead of recomputing them as is currently done in #70549, since it seems to avoid most of the changes in spills and reloads that we were seeing. Specifically LIS->handleMove seemed to change the SlotIndex of instructions which might have affected regalloc.

After this patch moving to post regalloc should be a matter of moving the pass in RISCVTargetMachines wihout any further changes to RISCVInsertVSETVLI itself.

Co-authored-by: Piyou Chen <[email protected]>
Co-authored-by: Luke Lau <[email protected]>

Author: lukel97

llvm/lib/Target/RISCV/RISCVInsertVSETVLI.cpp

@@ -541,7 +541,14 @@ void RISCVPassConfig::addPreRegAlloc() {
   addPass(createRISCVPreRAExpandPseudoPass());
   if (TM->getOptLevel() != CodeGenOptLevel::None)
     addPass(createRISCVMergeBaseOffsetOptPass());
-  addPass(createRISCVInsertVSETVLIPass());
+
+  // Run RISCVInsertVSETVLI after PHI elimination. On O1 and above do it after
+  // register coalescing so needVSETVLIPHI doesn't need to look through COPYs.
+  if (TM->getOptLevel() == CodeGenOptLevel::None)
+    insertPass(&PHIEliminationID, createRISCVInsertVSETVLIPass());
+  else
+    insertPass(&RegisterCoalescerID, createRISCVInsertVSETVLIPass());
+
   addPass(createRISCVInsertReadWriteCSRPass());
   addPass(createRISCVInsertWriteVXRMPass());
 }

llvm/lib/Target/RISCV/RISCVTargetMachine.cpp

@@ -40,14 +40,16 @@
 ; CHECK-NEXT:       Finalize ISel and expand pseudo-instructions
 ; CHECK-NEXT:       Local Stack Slot Allocation
 ; CHECK-NEXT:       RISC-V Pre-RA pseudo instruction expansion pass
-; CHECK-NEXT:       RISC-V Insert VSETVLI pass
 ; CHECK-NEXT:       RISC-V Insert Read/Write CSR Pass
 ; CHECK-NEXT:       RISC-V Insert Write VXRM Pass
 ; CHECK-NEXT:       Init Undef Pass
 ; CHECK-NEXT:       Eliminate PHI nodes for register allocation
+; CHECK-NEXT:       MachineDominator Tree Construction
+; CHECK-NEXT:       Slot index numbering
+; CHECK-NEXT:       Live Interval Analysis
+; CHECK-NEXT:       RISC-V Insert VSETVLI pass
 ; CHECK-NEXT:       Two-Address instruction pass
 ; CHECK-NEXT:       Fast Register Allocator
-; CHECK-NEXT:       MachineDominator Tree Construction
 ; CHECK-NEXT:       Slot index numbering
 ; CHECK-NEXT:       Live Interval Analysis
 ; CHECK-NEXT:       RISC-V Coalesce VSETVLI pass

llvm/test/CodeGen/RISCV/O0-pipeline.ll

@@ -115,7 +115,6 @@
 ; RV64-NEXT:        RISC-V Optimize W Instructions
 ; CHECK-NEXT:       RISC-V Pre-RA pseudo instruction expansion pass
 ; CHECK-NEXT:       RISC-V Merge Base Offset
-; CHECK-NEXT:       RISC-V Insert VSETVLI pass
 ; CHECK-NEXT:       RISC-V Insert Read/Write CSR Pass
 ; CHECK-NEXT:       RISC-V Insert Write VXRM Pass
 ; CHECK-NEXT:       Detect Dead Lanes
@@ -129,6 +128,7 @@
 ; CHECK-NEXT:       Slot index numbering
 ; CHECK-NEXT:       Live Interval Analysis
 ; CHECK-NEXT:       Register Coalescer
+; CHECK-NEXT:       RISC-V Insert VSETVLI pass
 ; CHECK-NEXT:       Rename Disconnected Subregister Components
 ; CHECK-NEXT:       Machine Instruction Scheduler
 ; CHECK-NEXT:       Machine Block Frequency Analysis

llvm/test/CodeGen/RISCV/O3-pipeline.ll

@@ -15,8 +15,8 @@ define <vscale x 1 x half> @vp_ceil_vv_nxv1f16(<vscale x 1 x half> %va, <vscale
 ; CHECK-NEXT:    vfabs.v v9, v8, v0.t
 ; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
 ; CHECK-NEXT:    vmflt.vf v0, v9, fa5, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
 ; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
 ; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
 ; CHECK-NEXT:    fsrm a0
 ; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
@@ -57,8 +57,8 @@ define <vscale x 2 x half> @vp_ceil_vv_nxv2f16(<vscale x 2 x half> %va, <vscale
 ; CHECK-NEXT:    vfabs.v v9, v8, v0.t
 ; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
 ; CHECK-NEXT:    vmflt.vf v0, v9, fa5, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, ma
 ; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, ma
 ; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
 ; CHECK-NEXT:    fsrm a0
 ; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
@@ -99,8 +99,8 @@ define <vscale x 4 x half> @vp_ceil_vv_nxv4f16(<vscale x 4 x half> %va, <vscale
 ; CHECK-NEXT:    vfabs.v v9, v8, v0.t
 ; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, mu
 ; CHECK-NEXT:    vmflt.vf v0, v9, fa5, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
 ; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
 ; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
 ; CHECK-NEXT:    fsrm a0
 ; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
@@ -142,8 +142,8 @@ define <vscale x 8 x half> @vp_ceil_vv_nxv8f16(<vscale x 8 x half> %va, <vscale
 ; CHECK-NEXT:    vfabs.v v12, v8, v0.t
 ; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, mu
 ; CHECK-NEXT:    vmflt.vf v10, v12, fa5, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
 ; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
 ; CHECK-NEXT:    vmv1r.v v0, v10
 ; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
 ; CHECK-NEXT:    fsrm a0
@@ -186,8 +186,8 @@ define <vscale x 16 x half> @vp_ceil_vv_nxv16f16(<vscale x 16 x half> %va, <vsca
 ; CHECK-NEXT:    vfabs.v v16, v8, v0.t
 ; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, mu
 ; CHECK-NEXT:    vmflt.vf v12, v16, fa5, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
 ; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
 ; CHECK-NEXT:    vmv1r.v v0, v12
 ; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
 ; CHECK-NEXT:    fsrm a0
@@ -230,8 +230,8 @@ define <vscale x 32 x half> @vp_ceil_vv_nxv32f16(<vscale x 32 x half> %va, <vsca
 ; CHECK-NEXT:    vfabs.v v24, v8, v0.t
 ; CHECK-NEXT:    vsetvli zero, zero, e16, m8, ta, mu
 ; CHECK-NEXT:    vmflt.vf v16, v24, fa5, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e16, m8, ta, ma
 ; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e16, m8, ta, ma
 ; CHECK-NEXT:    vmv1r.v v0, v16
 ; CHECK-NEXT:    vfcvt.x.f.v v24, v8, v0.t
 ; CHECK-NEXT:    fsrm a0
@@ -273,8 +273,8 @@ define <vscale x 1 x float> @vp_ceil_vv_nxv1f32(<vscale x 1 x float> %va, <vscal
 ; CHECK-NEXT:    fmv.w.x fa5, a0
 ; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
 ; CHECK-NEXT:    vmflt.vf v0, v9, fa5, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
 ; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
 ; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
 ; CHECK-NEXT:    fsrm a0
 ; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
@@ -315,8 +315,8 @@ define <vscale x 2 x float> @vp_ceil_vv_nxv2f32(<vscale x 2 x float> %va, <vscal
 ; CHECK-NEXT:    fmv.w.x fa5, a0
 ; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
 ; CHECK-NEXT:    vmflt.vf v0, v9, fa5, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
 ; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
 ; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
 ; CHECK-NEXT:    fsrm a0
 ; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
@@ -358,8 +358,8 @@ define <vscale x 4 x float> @vp_ceil_vv_nxv4f32(<vscale x 4 x float> %va, <vscal
 ; CHECK-NEXT:    fmv.w.x fa5, a0
 ; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, mu
 ; CHECK-NEXT:    vmflt.vf v10, v12, fa5, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
 ; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
 ; CHECK-NEXT:    vmv1r.v v0, v10
 ; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
 ; CHECK-NEXT:    fsrm a0
@@ -402,8 +402,8 @@ define <vscale x 8 x float> @vp_ceil_vv_nxv8f32(<vscale x 8 x float> %va, <vscal
 ; CHECK-NEXT:    fmv.w.x fa5, a0
 ; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
 ; CHECK-NEXT:    vmflt.vf v12, v16, fa5, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
 ; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
 ; CHECK-NEXT:    vmv1r.v v0, v12
 ; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
 ; CHECK-NEXT:    fsrm a0
@@ -446,8 +446,8 @@ define <vscale x 16 x float> @vp_ceil_vv_nxv16f32(<vscale x 16 x float> %va, <vs
 ; CHECK-NEXT:    fmv.w.x fa5, a0
 ; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, mu
 ; CHECK-NEXT:    vmflt.vf v16, v24, fa5, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
 ; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
 ; CHECK-NEXT:    vmv1r.v v0, v16
 ; CHECK-NEXT:    vfcvt.x.f.v v24, v8, v0.t
 ; CHECK-NEXT:    fsrm a0
@@ -489,8 +489,8 @@ define <vscale x 1 x double> @vp_ceil_vv_nxv1f64(<vscale x 1 x double> %va, <vsc
 ; CHECK-NEXT:    vfabs.v v9, v8, v0.t
 ; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
 ; CHECK-NEXT:    vmflt.vf v0, v9, fa5, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
 ; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
 ; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
 ; CHECK-NEXT:    fsrm a0
 ; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
@@ -532,8 +532,8 @@ define <vscale x 2 x double> @vp_ceil_vv_nxv2f64(<vscale x 2 x double> %va, <vsc
 ; CHECK-NEXT:    vfabs.v v12, v8, v0.t
 ; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, mu
 ; CHECK-NEXT:    vmflt.vf v10, v12, fa5, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, ma
 ; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, ma
 ; CHECK-NEXT:    vmv1r.v v0, v10
 ; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
 ; CHECK-NEXT:    fsrm a0
@@ -576,8 +576,8 @@ define <vscale x 4 x double> @vp_ceil_vv_nxv4f64(<vscale x 4 x double> %va, <vsc
 ; CHECK-NEXT:    vfabs.v v16, v8, v0.t
 ; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, mu
 ; CHECK-NEXT:    vmflt.vf v12, v16, fa5, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, ma
 ; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, ma
 ; CHECK-NEXT:    vmv1r.v v0, v12
 ; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
 ; CHECK-NEXT:    fsrm a0
@@ -620,8 +620,8 @@ define <vscale x 7 x double> @vp_ceil_vv_nxv7f64(<vscale x 7 x double> %va, <vsc
 ; CHECK-NEXT:    vfabs.v v24, v8, v0.t
 ; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, mu
 ; CHECK-NEXT:    vmflt.vf v16, v24, fa5, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
 ; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
 ; CHECK-NEXT:    vmv1r.v v0, v16
 ; CHECK-NEXT:    vfcvt.x.f.v v24, v8, v0.t
 ; CHECK-NEXT:    fsrm a0
@@ -664,8 +664,8 @@ define <vscale x 8 x double> @vp_ceil_vv_nxv8f64(<vscale x 8 x double> %va, <vsc
 ; CHECK-NEXT:    vfabs.v v24, v8, v0.t
 ; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, mu
 ; CHECK-NEXT:    vmflt.vf v16, v24, fa5, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
 ; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
 ; CHECK-NEXT:    vmv1r.v v0, v16
 ; CHECK-NEXT:    vfcvt.x.f.v v24, v8, v0.t
 ; CHECK-NEXT:    fsrm a0
@@ -726,8 +726,8 @@ define <vscale x 16 x double> @vp_ceil_vv_nxv16f64(<vscale x 16 x double> %va, <
 ; CHECK-NEXT:    vfabs.v v8, v16, v0.t
 ; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, mu
 ; CHECK-NEXT:    vmflt.vf v25, v8, fa5, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
 ; CHECK-NEXT:    fsrmi a2, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
 ; CHECK-NEXT:    vmv1r.v v0, v25
 ; CHECK-NEXT:    vfcvt.x.f.v v8, v16, v0.t
 ; CHECK-NEXT:    fsrm a2
@@ -750,8 +750,8 @@ define <vscale x 16 x double> @vp_ceil_vv_nxv16f64(<vscale x 16 x double> %va, <
 ; CHECK-NEXT:    vfabs.v v16, v8, v0.t
 ; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, mu
 ; CHECK-NEXT:    vmflt.vf v24, v16, fa5, v0.t
-; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
 ; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
 ; CHECK-NEXT:    vmv1r.v v0, v24
 ; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
 ; CHECK-NEXT:    fsrm a0

llvm/test/CodeGen/RISCV/rvv/ceil-vp.ll

@@ -36,8 +36,8 @@ define <vscale x 4 x i32> @vadd_undef(<vscale x 4 x i32> %a, <vscale x 4 x i32>
 define <vscale x 4 x i32> @vadd_same_passthru(<vscale x 4 x i32> %passthru, <vscale x 4 x i32> %a, <vscale x 4 x i32> %b, iXLen %vl1, iXLen %vl2) {
 ; CHECK-LABEL: vadd_same_passthru:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetvli zero, a0, e32, m2, tu, ma
 ; CHECK-NEXT:    vmv2r.v v14, v8
+; CHECK-NEXT:    vsetvli zero, a0, e32, m2, tu, ma
 ; CHECK-NEXT:    vadd.vv v14, v10, v12
 ; CHECK-NEXT:    vsetvli zero, a1, e32, m2, tu, ma
 ; CHECK-NEXT:    vmv.v.v v8, v14

llvm/test/CodeGen/RISCV/rvv/combine-vmv.ll

@@ -720,8 +720,8 @@ declare <vscale x 1 x i64> @llvm.riscv.vaadd.nxv1i64.nxv1i64(<vscale x 1 x i64>,
 define <vscale x 1 x i64> @commutable_vaadd_vv(<vscale x 1 x i64> %0, <vscale x 1 x i64> %1, iXLen %2) nounwind {
 ; CHECK-LABEL: commutable_vaadd_vv:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    vsetvli zero, a0, e64, m1, ta, ma
 ; CHECK-NEXT:    csrwi vxrm, 0
+; CHECK-NEXT:    vsetvli zero, a0, e64, m1, ta, ma
 ; CHECK-NEXT:    vaadd.vv v8, v8, v9
 ; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, ma
 ; CHECK-NEXT:    vadd.vv v8, v8, v8
@@ -737,8 +737,8 @@ declare <vscale x 1 x i64> @llvm.riscv.vaadd.mask.nxv1i64.nxv1i64(<vscale x 1 x
 define <vscale x 1 x i64> @commutable_vaadd_vv_masked(<vscale x 1 x i64> %0, <vscale x 1 x i64> %1, <vscale x 1 x i1> %mask, iXLen %2) {
 ; CHECK-LABEL: commutable_vaadd_vv_masked:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetvli zero, a0, e64, m1, ta, ma
 ; CHECK-NEXT:    csrwi vxrm, 0
+; CHECK-NEXT:    vsetvli zero, a0, e64, m1, ta, ma
 ; CHECK-NEXT:    vaadd.vv v10, v8, v9, v0.t
 ; CHECK-NEXT:    vaadd.vv v8, v8, v9, v0.t
 ; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, ma
@@ -755,8 +755,8 @@ declare <vscale x 1 x i64> @llvm.riscv.vaaddu.nxv1i64.nxv1i64(<vscale x 1 x i64>
 define <vscale x 1 x i64> @commutable_vaaddu_vv(<vscale x 1 x i64> %0, <vscale x 1 x i64> %1, iXLen %2) nounwind {
 ; CHECK-LABEL: commutable_vaaddu_vv:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    vsetvli zero, a0, e64, m1, ta, ma
 ; CHECK-NEXT:    csrwi vxrm, 0
+; CHECK-NEXT:    vsetvli zero, a0, e64, m1, ta, ma
 ; CHECK-NEXT:    vaaddu.vv v8, v8, v9
 ; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, ma
 ; CHECK-NEXT:    vadd.vv v8, v8, v8
@@ -772,8 +772,8 @@ declare <vscale x 1 x i64> @llvm.riscv.vaaddu.mask.nxv1i64.nxv1i64(<vscale x 1 x
 define <vscale x 1 x i64> @commutable_vaaddu_vv_masked(<vscale x 1 x i64> %0, <vscale x 1 x i64> %1, <vscale x 1 x i1> %mask, iXLen %2) {
 ; CHECK-LABEL: commutable_vaaddu_vv_masked:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetvli zero, a0, e64, m1, ta, ma
 ; CHECK-NEXT:    csrwi vxrm, 0
+; CHECK-NEXT:    vsetvli zero, a0, e64, m1, ta, ma
 ; CHECK-NEXT:    vaaddu.vv v10, v8, v9, v0.t
 ; CHECK-NEXT:    vaaddu.vv v8, v8, v9, v0.t
 ; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, ma
@@ -790,8 +790,8 @@ declare <vscale x 1 x i64> @llvm.riscv.vsmul.nxv1i64.nxv1i64(<vscale x 1 x i64>,
 define <vscale x 1 x i64> @commutable_vsmul_vv(<vscale x 1 x i64> %0, <vscale x 1 x i64> %1, iXLen %2) nounwind {
 ; CHECK-LABEL: commutable_vsmul_vv:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    vsetvli zero, a0, e64, m1, ta, ma
 ; CHECK-NEXT:    csrwi vxrm, 0
+; CHECK-NEXT:    vsetvli zero, a0, e64, m1, ta, ma
 ; CHECK-NEXT:    vsmul.vv v8, v8, v9
 ; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, ma
 ; CHECK-NEXT:    vadd.vv v8, v8, v8
@@ -807,8 +807,8 @@ declare <vscale x 1 x i64> @llvm.riscv.vsmul.mask.nxv1i64.nxv1i64(<vscale x 1 x
 define <vscale x 1 x i64> @commutable_vsmul_vv_masked(<vscale x 1 x i64> %0, <vscale x 1 x i64> %1, <vscale x 1 x i1> %mask, iXLen %2) {
 ; CHECK-LABEL: commutable_vsmul_vv_masked:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vsetvli zero, a0, e64, m1, ta, ma
 ; CHECK-NEXT:    csrwi vxrm, 0
+; CHECK-NEXT:    vsetvli zero, a0, e64, m1, ta, ma
 ; CHECK-NEXT:    vsmul.vv v10, v8, v9, v0.t
 ; CHECK-NEXT:    vsmul.vv v8, v8, v9, v0.t
 ; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, ma

llvm/test/CodeGen/RISCV/rvv/commutable.ll

@@ -149,8 +149,8 @@ define void @constant_zero_stride(ptr %s, ptr %d) {
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetivli zero, 2, e8, mf8, ta, ma
 ; CHECK-NEXT:    vle8.v v8, (a0)
-; CHECK-NEXT:    vsetivli zero, 4, e8, mf4, ta, ma
 ; CHECK-NEXT:    vmv1r.v v9, v8
+; CHECK-NEXT:    vsetivli zero, 4, e8, mf4, ta, ma
 ; CHECK-NEXT:    vslideup.vi v9, v8, 2
 ; CHECK-NEXT:    vse8.v v9, (a1)
 ; CHECK-NEXT:    ret