[RISCV][VLOPT] Support v[f]slide1up.v{x,f} (#146716)

Similarly to #146710, for vslide1ups vl only determines the destination
elements written to so we can safely reduce their AVL.

We cannot do this for vslide1downs as the vl determines which lane the
new element is to be inserted in, so some negative tests have been
added.

Author: lukel97

llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp

@@ -1045,7 +1045,8 @@ static bool isSupportedInstr(const MachineInstr &MI) {
   case RISCV::VSLIDEUP_VI:
   case RISCV::VSLIDEDOWN_VX:
   case RISCV::VSLIDEDOWN_VI:
-  // TODO: Handle v[f]slide1up, but not v[f]slide1down.
+  case RISCV::VSLIDE1UP_VX:
+  case RISCV::VFSLIDE1UP_VF:
   // Vector Single-Width Floating-Point Add/Subtract Instructions
   case RISCV::VFADD_VF:
   case RISCV::VFADD_VV:

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

@@ -3514,6 +3514,92 @@ define <vscale x 4 x i32> @vslidedown_vi(<vscale x 4 x i32> %a, iXLen %vl) {
   ret <vscale x 4 x i32> %2
 }
 
+define <vscale x 4 x i32> @vslide1up_vx(<vscale x 4 x i32> %a, iXLen %b, iXLen %vl) {
+; NOVLOPT-LABEL: vslide1up_vx:
+; NOVLOPT:       # %bb.0:
+; NOVLOPT-NEXT:    vsetvli a2, zero, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vslide1up.vx v10, v8, a0
+; NOVLOPT-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vadd.vv v8, v10, v10
+; NOVLOPT-NEXT:    ret
+;
+; VLOPT-LABEL: vslide1up_vx:
+; VLOPT:       # %bb.0:
+; VLOPT-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; VLOPT-NEXT:    vslide1up.vx v10, v8, a0
+; VLOPT-NEXT:    vadd.vv v8, v10, v10
+; VLOPT-NEXT:    ret
+  %1 = call <vscale x 4 x i32> @llvm.riscv.vslide1up(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, iXLen %b, iXLen -1)
+  %2 = call <vscale x 4 x i32> @llvm.riscv.vadd(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %1, iXLen %vl)
+  ret <vscale x 4 x i32> %2
+}
+
+define <vscale x 4 x float> @vfslide1up_vf(<vscale x 4 x float> %a, float %b, iXLen %vl) {
+; NOVLOPT-LABEL: vfslide1up_vf:
+; NOVLOPT:       # %bb.0:
+; NOVLOPT-NEXT:    vsetvli a1, zero, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vfslide1up.vf v10, v8, fa0
+; NOVLOPT-NEXT:    vsetvli zero, a0, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vfadd.vv v8, v10, v10
+; NOVLOPT-NEXT:    ret
+;
+; VLOPT-LABEL: vfslide1up_vf:
+; VLOPT:       # %bb.0:
+; VLOPT-NEXT:    vsetvli zero, a0, e32, m2, ta, ma
+; VLOPT-NEXT:    vfslide1up.vf v10, v8, fa0
+; VLOPT-NEXT:    vfadd.vv v8, v10, v10
+; VLOPT-NEXT:    ret
+  %1 = call <vscale x 4 x float> @llvm.riscv.vfslide1up(<vscale x 4 x float> poison, <vscale x 4 x float> %a, float %b, iXLen -1)
+  %2 = call <vscale x 4 x float> @llvm.riscv.vfadd(<vscale x 4 x float> poison, <vscale x 4 x float> %1, <vscale x 4 x float> %1, iXLen 7, iXLen %vl)
+  ret <vscale x 4 x float> %2
+}
+
+; Negative test – not safe to reduce vl
+
+define <vscale x 4 x i32> @vslide1down_vx(<vscale x 4 x i32> %a, iXLen %b, iXLen %vl) {
+; NOVLOPT-LABEL: vslide1down_vx:
+; NOVLOPT:       # %bb.0:
+; NOVLOPT-NEXT:    vsetvli a2, zero, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vslide1down.vx v8, v8, a0
+; NOVLOPT-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vadd.vv v8, v8, v8
+; NOVLOPT-NEXT:    ret
+;
+; VLOPT-LABEL: vslide1down_vx:
+; VLOPT:       # %bb.0:
+; VLOPT-NEXT:    vsetvli a2, zero, e32, m2, ta, ma
+; VLOPT-NEXT:    vslide1down.vx v8, v8, a0
+; VLOPT-NEXT:    vsetvli zero, a1, e32, m2, ta, ma
+; VLOPT-NEXT:    vadd.vv v8, v8, v8
+; VLOPT-NEXT:    ret
+  %1 = call <vscale x 4 x i32> @llvm.riscv.vslide1down(<vscale x 4 x i32> poison, <vscale x 4 x i32> %a, iXLen %b, iXLen -1)
+  %2 = call <vscale x 4 x i32> @llvm.riscv.vadd(<vscale x 4 x i32> poison, <vscale x 4 x i32> %1, <vscale x 4 x i32> %1, iXLen %vl)
+  ret <vscale x 4 x i32> %2
+}
+
+; Negative test – not safe to reduce vl
+
+define <vscale x 4 x float> @vfslide1down_vf(<vscale x 4 x float> %a, float %b, iXLen %vl) {
+; NOVLOPT-LABEL: vfslide1down_vf:
+; NOVLOPT:       # %bb.0:
+; NOVLOPT-NEXT:    vsetvli a1, zero, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vfslide1down.vf v8, v8, fa0
+; NOVLOPT-NEXT:    vsetvli zero, a0, e32, m2, ta, ma
+; NOVLOPT-NEXT:    vfadd.vv v8, v8, v8
+; NOVLOPT-NEXT:    ret
+;
+; VLOPT-LABEL: vfslide1down_vf:
+; VLOPT:       # %bb.0:
+; VLOPT-NEXT:    vsetvli a1, zero, e32, m2, ta, ma
+; VLOPT-NEXT:    vfslide1down.vf v8, v8, fa0
+; VLOPT-NEXT:    vsetvli zero, a0, e32, m2, ta, ma
+; VLOPT-NEXT:    vfadd.vv v8, v8, v8
+; VLOPT-NEXT:    ret
+  %1 = call <vscale x 4 x float> @llvm.riscv.vfslide1down(<vscale x 4 x float> poison, <vscale x 4 x float> %a, float %b, iXLen -1)
+  %2 = call <vscale x 4 x float> @llvm.riscv.vfadd(<vscale x 4 x float> poison, <vscale x 4 x float> %1, <vscale x 4 x float> %1, iXLen 7, iXLen %vl)
+  ret <vscale x 4 x float> %2
+}
+
 define <vscale x 4 x float> @vfadd_vv(<vscale x 4 x float> %a, <vscale x 4 x float> %b, iXLen %vl) {
 ; NOVLOPT-LABEL: vfadd_vv:
 ; NOVLOPT:       # %bb.0: