[RISCV][VLOPT] Add widening floating point multiply to isSupportedInstr

Author: michaelmaitland

llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp

@@ -1004,6 +1004,9 @@ static bool isSupportedInstr(const MachineInstr &MI) {
   case RISCV::VFDIV_VF:
   case RISCV::VFDIV_VV:
   case RISCV::VFRDIV_VF:
+  // Vector Widening Floating-Point Multiply
+  case RISCV::VFWMUL_VF:
+  case RISCV::VFWMUL_VV:
   // Single-Width Floating-Point/Integer Type-Convert Instructions
   case RISCV::VFCVT_XU_F_V:
   case RISCV::VFCVT_X_F_V:

llvm/test/CodeGen/RISCV/rvv/vl-opt-instrs.ll

@@ -3293,3 +3293,45 @@ define <vscale x 4 x float> @vfrdiv_vf(<vscale x 4 x float> %a, float %b, iXLen
   %2 = call <vscale x 4 x float> @llvm.riscv.vfadd.nxv4f32.nxv4f32(<vscale x 4 x float> poison, <vscale x 4 x float> %1, <vscale x 4 x float> %a, iXLen 7, iXLen %vl)
   ret <vscale x 4 x float> %2
 }
+
+define <vscale x 4 x double> @vfwmul_vv(<vscale x 4 x float> %a, <vscale x 4 x float> %b, iXLen %vl) {
+; NOVLOPT-LABEL: vfwmul_vv:
+; NOVLOPT:       # %bb.0:
+; NOVLOPT-NEXT:    vsetvli a1, zero, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vfwmul.vv v12, v8, v10
+; NOVLOPT-NEXT:    vsetvli zero, a0, e64, m4, ta, ma
+; NOVLOPT-NEXT:    vfadd.vv v8, v12, v12
+; NOVLOPT-NEXT:    ret
+;
+; VLOPT-LABEL: vfwmul_vv:
+; VLOPT:       # %bb.0:
+; VLOPT-NEXT:    vsetvli zero, a0, e32, m2, ta, ma
+; VLOPT-NEXT:    vfwmul.vv v12, v8, v10
+; VLOPT-NEXT:    vsetvli zero, zero, e64, m4, ta, ma
+; VLOPT-NEXT:    vfadd.vv v8, v12, v12
+; VLOPT-NEXT:    ret
+  %1 = call <vscale x 4 x double> @llvm.riscv.vfwmul.nxv4f64.nxv4f32.nxv4f32(<vscale x 4 x double> poison, <vscale x 4 x float> %a, <vscale x 4 x float> %b, iXLen 7, iXLen -1)
+  %2 = call <vscale x 4 x double> @llvm.riscv.vfadd.nxv4f64.nxv4f64(<vscale x 4 x double> poison, <vscale x 4 x double> %1, <vscale x 4 x double> %1, iXLen 7, iXLen %vl)
+  ret <vscale x 4 x double> %2
+}
+
+define <vscale x 4 x double> @vfwmul_vf(<vscale x 4 x float> %a, float %b, iXLen %vl) {
+; NOVLOPT-LABEL: vfwmul_vf:
+; NOVLOPT:       # %bb.0:
+; NOVLOPT-NEXT:    vsetvli a1, zero, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vfwmul.vf v12, v8, fa0
+; NOVLOPT-NEXT:    vsetvli zero, a0, e64, m4, ta, ma
+; NOVLOPT-NEXT:    vfadd.vv v8, v12, v12
+; NOVLOPT-NEXT:    ret
+;
+; VLOPT-LABEL: vfwmul_vf:
+; VLOPT:       # %bb.0:
+; VLOPT-NEXT:    vsetvli zero, a0, e32, m2, ta, ma
+; VLOPT-NEXT:    vfwmul.vf v12, v8, fa0
+; VLOPT-NEXT:    vsetvli zero, zero, e64, m4, ta, ma
+; VLOPT-NEXT:    vfadd.vv v8, v12, v12
+; VLOPT-NEXT:    ret
+  %1 = call <vscale x 4 x double> @llvm.riscv.vfwmul.nxv4f64.nxv4f32.f32(<vscale x 4 x double> poison, <vscale x 4 x float> %a, float %b, iXLen 7, iXLen -1)
+  %2 = call <vscale x 4 x double> @llvm.riscv.vfadd.nxv4f64.nxv4f64(<vscale x 4 x double> poison, <vscale x 4 x double> %1, <vscale x 4 x double> %1, iXLen 7, iXLen %vl)
+  ret <vscale x 4 x double> %2
+}