[RISCV][VLOPT] Add vsadd and vsaddu to isSupportedInstr

[michaelmaitland]

Based on #120351.

[llvmbot]

@llvm/pr-subscribers-backend-risc-v

Author: Michael Maitland (michaelmaitland)

Changes

Based on #120351.

---

Full diff: https://github.com/llvm/llvm-project/pull/120358.diff

2 Files Affected:

• (modified) llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp (+10-1)
• (modified) llvm/test/CodeGen/RISCV/rvv/vl-opt-instrs.ll (+120)

diff --git a/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp b/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp
index eefbad7460d1bd..cbde64c4e210d3 100644
--- a/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp
+++ b/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp
@@ -801,6 +801,15 @@ static bool isSupportedInstr(const MachineInstr &MI) {
   case RISCV::VMSOF_M:
   case RISCV::VIOTA_M:
   case RISCV::VID_V:
+
+  // Vector Single-Width Saturating Add and Subtract
+  case RISCV::VSADDU_VI:
+  case RISCV::VSADDU_VV:
+  case RISCV::VSADDU_VX:
+  case RISCV::VSADD_VI:
+  case RISCV::VSADD_VV:
+  case RISCV::VSADD_VX:
+
     return true;
   }
 
@@ -872,7 +881,7 @@ bool RISCVVLOptimizer::isCandidate(const MachineInstr &MI) const {
   const MCInstrDesc &Desc = MI.getDesc();
   if (!RISCVII::hasVLOp(Desc.TSFlags) || !RISCVII::hasSEWOp(Desc.TSFlags))
     return false;
-  if (MI.getNumDefs() != 1)
+  if (MI.getNumExplicitDefs() != 1)
     return false;
 
   // If we're not using VLMAX, then we need to be careful whether we are using
diff --git a/llvm/test/CodeGen/RISCV/rvv/vl-opt-instrs.ll b/llvm/test/CodeGen/RISCV/rvv/vl-opt-instrs.ll
index 55a50a15c788c2..527edfe1055ea5 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vl-opt-instrs.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vl-opt-instrs.ll
@@ -2925,3 +2925,123 @@ define <vscale x 4 x i32> @vid.v(<vscale x 4 x i32> %c, iXLen %vl) {
   %2 = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %c, iXLen %vl)
   ret <vscale x 4 x i32> %2
 }
+
+define <vscale x 4 x i32> @vsaddu_vi(<vscale x 4 x i32> %a, iXLen %vl) {
+; NOVLOPT-LABEL: vsaddu_vi:
+; NOVLOPT:       # %bb.0:
+; NOVLOPT-NEXT:    vsetvli a1, zero, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vsaddu.vi v10, v8, 5
+; NOVLOPT-NEXT:    vsetvli zero, a0, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vadd.vv v8, v10, v8
+; NOVLOPT-NEXT:    ret
+;
+; VLOPT-LABEL: vsaddu_vi:
+; VLOPT:       # %bb.0:
+; VLOPT-NEXT:    vsetvli zero, a0, e32, m2, ta, ma
+; VLOPT-NEXT:    vsaddu.vi v10, v8, 5
+; VLOPT-NEXT:    vadd.vv v8, v10, v8
+; VLOPT-NEXT:    ret
+  %1 = call <vscale x 4 x i32> @llvm.riscv.vsaddu.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, i32 5, iXLen -1)
+  %2 = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %a, iXLen %vl)
+  ret <vscale x 4 x i32> %2
+}
+
+define <vscale x 4 x i32> @vsaddu_vv(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, iXLen %vl) {
+; NOVLOPT-LABEL: vsaddu_vv:
+; NOVLOPT:       # %bb.0:
+; NOVLOPT-NEXT:    vsetvli a1, zero, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vsaddu.vv v8, v8, v10
+; NOVLOPT-NEXT:    vsetvli zero, a0, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vadd.vv v8, v8, v10
+; NOVLOPT-NEXT:    ret
+;
+; VLOPT-LABEL: vsaddu_vv:
+; VLOPT:       # %bb.0:
+; VLOPT-NEXT:    vsetvli zero, a0, e32, m2, ta, ma
+; VLOPT-NEXT:    vsaddu.vv v8, v8, v10
+; VLOPT-NEXT:    vadd.vv v8, v8, v10
+; VLOPT-NEXT:    ret
+  %1 = call <vscale x 4 x i32> @llvm.riscv.vsaddu.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, <vscale x 4 x i32> %b, iXLen -1)
+  %2 = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %b, iXLen %vl)
+  ret <vscale x 4 x i32> %2
+}
+
+define <vscale x 4 x i32> @vsaddu_vx(<vscale x 4 x i32> %a, i32 %b, iXLen %vl) {
+; NOVLOPT-LABEL: vsaddu_vx:
+; NOVLOPT:       # %bb.0:
+; NOVLOPT-NEXT:    vsetvli a2, zero, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vsaddu.vx v10, v8, a0
+; NOVLOPT-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vadd.vv v8, v10, v8
+; NOVLOPT-NEXT:    ret
+;
+; VLOPT-LABEL: vsaddu_vx:
+; VLOPT:       # %bb.0:
+; VLOPT-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; VLOPT-NEXT:    vsaddu.vx v10, v8, a0
+; VLOPT-NEXT:    vadd.vv v8, v10, v8
+; VLOPT-NEXT:    ret
+  %1 = call <vscale x 4 x i32> @llvm.riscv.vsaddu.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, i32 %b, iXLen -1)
+  %2 = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %a, iXLen %vl)
+  ret <vscale x 4 x i32> %2
+}
+
+define <vscale x 4 x i32> @vsadd_vi(<vscale x 4 x i32> %a, iXLen %vl) {
+; NOVLOPT-LABEL: vsadd_vi:
+; NOVLOPT:       # %bb.0:
+; NOVLOPT-NEXT:    vsetvli a1, zero, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vsadd.vi v10, v8, 5
+; NOVLOPT-NEXT:    vsetvli zero, a0, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vadd.vv v8, v10, v8
+; NOVLOPT-NEXT:    ret
+;
+; VLOPT-LABEL: vsadd_vi:
+; VLOPT:       # %bb.0:
+; VLOPT-NEXT:    vsetvli zero, a0, e32, m2, ta, ma
+; VLOPT-NEXT:    vsadd.vi v10, v8, 5
+; VLOPT-NEXT:    vadd.vv v8, v10, v8
+; VLOPT-NEXT:    ret
+  %1 = call <vscale x 4 x i32> @llvm.riscv.vsadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, i32 5, iXLen -1)
+  %2 = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %a, iXLen %vl)
+  ret <vscale x 4 x i32> %2
+}
+
+define <vscale x 4 x i32> @vsadd_vv(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, iXLen %vl) {
+; NOVLOPT-LABEL: vsadd_vv:
+; NOVLOPT:       # %bb.0:
+; NOVLOPT-NEXT:    vsetvli a1, zero, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vsadd.vv v8, v8, v10
+; NOVLOPT-NEXT:    vsetvli zero, a0, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vadd.vv v8, v8, v10
+; NOVLOPT-NEXT:    ret
+;
+; VLOPT-LABEL: vsadd_vv:
+; VLOPT:       # %bb.0:
+; VLOPT-NEXT:    vsetvli zero, a0, e32, m2, ta, ma
+; VLOPT-NEXT:    vsadd.vv v8, v8, v10
+; VLOPT-NEXT:    vadd.vv v8, v8, v10
+; VLOPT-NEXT:    ret
+  %1 = call <vscale x 4 x i32> @llvm.riscv.vsadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, <vscale x 4 x i32> %b, iXLen -1)
+  %2 = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %b, iXLen %vl)
+  ret <vscale x 4 x i32> %2
+}
+
+define <vscale x 4 x i32> @vsadd_vx(<vscale x 4 x i32> %a, i32 %b, iXLen %vl) {
+; NOVLOPT-LABEL: vsadd_vx:
+; NOVLOPT:       # %bb.0:
+; NOVLOPT-NEXT:    vsetvli a2, zero, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vsadd.vx v10, v8, a0
+; NOVLOPT-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vadd.vv v8, v10, v8
+; NOVLOPT-NEXT:    ret
+;
+; VLOPT-LABEL: vsadd_vx:
+; VLOPT:       # %bb.0:
+; VLOPT-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; VLOPT-NEXT:    vsadd.vx v10, v8, a0
+; VLOPT-NEXT:    vadd.vv v8, v10, v8
+; VLOPT-NEXT:    ret
+  %1 = call <vscale x 4 x i32> @llvm.riscv.vsadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, i32 %b, iXLen -1)
+  %2 = call <vscale x 4 x i32> @llvm.riscv.vadd.nxv4i32.nxv4i32(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %a, iXLen %vl)
+  ret <vscale x 4 x i32> %2
+}

[michaelmaitland]

Pinging since its been a week, but I understand if reviewers are OOO for the holidays.

[topperc]

This can cause a difference in the VXSAT flag, but we make it hard for for software to reliably read that register. So I'm not sure if that should block this or not.

I guess we should make sure no other VXSAT writing instructions are already in isSupportedInstr

[michaelmaitland]

Had a chat offline about this with @topperc and @kito-cheng. Going to close this because of VXSAT. No other VXSAT writing instructions are already in isSupportedInstr.