| ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py |
| ; RUN: llc -mtriple=riscv32 -mattr=+experimental-v -verify-machineinstrs < %s | FileCheck %s |
| |
| declare i8 @llvm.vector.reduce.add.nxv1i8(<vscale x 1 x i8>) |
| |
| define signext i8 @vreduce_add_nxv1i8(<vscale x 1 x i8> %v) { |
| ; CHECK-LABEL: vreduce_add_nxv1i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf8,ta,mu |
| ; CHECK-NEXT: vredsum.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.add.nxv1i8(<vscale x 1 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.umax.nxv1i8(<vscale x 1 x i8>) |
| |
| define signext i8 @vreduce_umax_nxv1i8(<vscale x 1 x i8> %v) { |
| ; CHECK-LABEL: vreduce_umax_nxv1i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf8,ta,mu |
| ; CHECK-NEXT: vredmaxu.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.umax.nxv1i8(<vscale x 1 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.smax.nxv1i8(<vscale x 1 x i8>) |
| |
| define signext i8 @vreduce_smax_nxv1i8(<vscale x 1 x i8> %v) { |
| ; CHECK-LABEL: vreduce_smax_nxv1i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: addi a0, zero, -128 |
| ; CHECK-NEXT: vsetvli a1, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf8,ta,mu |
| ; CHECK-NEXT: vredmax.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.smax.nxv1i8(<vscale x 1 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.umin.nxv1i8(<vscale x 1 x i8>) |
| |
| define signext i8 @vreduce_umin_nxv1i8(<vscale x 1 x i8> %v) { |
| ; CHECK-LABEL: vreduce_umin_nxv1i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf8,ta,mu |
| ; CHECK-NEXT: vredminu.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.umin.nxv1i8(<vscale x 1 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.smin.nxv1i8(<vscale x 1 x i8>) |
| |
| define signext i8 @vreduce_smin_nxv1i8(<vscale x 1 x i8> %v) { |
| ; CHECK-LABEL: vreduce_smin_nxv1i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: addi a0, zero, 127 |
| ; CHECK-NEXT: vsetvli a1, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf8,ta,mu |
| ; CHECK-NEXT: vredmin.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.smin.nxv1i8(<vscale x 1 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.and.nxv1i8(<vscale x 1 x i8>) |
| |
| define signext i8 @vreduce_and_nxv1i8(<vscale x 1 x i8> %v) { |
| ; CHECK-LABEL: vreduce_and_nxv1i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf8,ta,mu |
| ; CHECK-NEXT: vredand.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.and.nxv1i8(<vscale x 1 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.or.nxv1i8(<vscale x 1 x i8>) |
| |
| define signext i8 @vreduce_or_nxv1i8(<vscale x 1 x i8> %v) { |
| ; CHECK-LABEL: vreduce_or_nxv1i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf8,ta,mu |
| ; CHECK-NEXT: vredor.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.or.nxv1i8(<vscale x 1 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.xor.nxv1i8(<vscale x 1 x i8>) |
| |
| define signext i8 @vreduce_xor_nxv1i8(<vscale x 1 x i8> %v) { |
| ; CHECK-LABEL: vreduce_xor_nxv1i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf8,ta,mu |
| ; CHECK-NEXT: vredxor.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.xor.nxv1i8(<vscale x 1 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.add.nxv2i8(<vscale x 2 x i8>) |
| |
| define signext i8 @vreduce_add_nxv2i8(<vscale x 2 x i8> %v) { |
| ; CHECK-LABEL: vreduce_add_nxv2i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf4,ta,mu |
| ; CHECK-NEXT: vredsum.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.add.nxv2i8(<vscale x 2 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.umax.nxv2i8(<vscale x 2 x i8>) |
| |
| define signext i8 @vreduce_umax_nxv2i8(<vscale x 2 x i8> %v) { |
| ; CHECK-LABEL: vreduce_umax_nxv2i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf4,ta,mu |
| ; CHECK-NEXT: vredmaxu.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.umax.nxv2i8(<vscale x 2 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.smax.nxv2i8(<vscale x 2 x i8>) |
| |
| define signext i8 @vreduce_smax_nxv2i8(<vscale x 2 x i8> %v) { |
| ; CHECK-LABEL: vreduce_smax_nxv2i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: addi a0, zero, -128 |
| ; CHECK-NEXT: vsetvli a1, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf4,ta,mu |
| ; CHECK-NEXT: vredmax.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.smax.nxv2i8(<vscale x 2 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.umin.nxv2i8(<vscale x 2 x i8>) |
| |
| define signext i8 @vreduce_umin_nxv2i8(<vscale x 2 x i8> %v) { |
| ; CHECK-LABEL: vreduce_umin_nxv2i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf4,ta,mu |
| ; CHECK-NEXT: vredminu.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.umin.nxv2i8(<vscale x 2 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.smin.nxv2i8(<vscale x 2 x i8>) |
| |
| define signext i8 @vreduce_smin_nxv2i8(<vscale x 2 x i8> %v) { |
| ; CHECK-LABEL: vreduce_smin_nxv2i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: addi a0, zero, 127 |
| ; CHECK-NEXT: vsetvli a1, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf4,ta,mu |
| ; CHECK-NEXT: vredmin.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.smin.nxv2i8(<vscale x 2 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.and.nxv2i8(<vscale x 2 x i8>) |
| |
| define signext i8 @vreduce_and_nxv2i8(<vscale x 2 x i8> %v) { |
| ; CHECK-LABEL: vreduce_and_nxv2i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf4,ta,mu |
| ; CHECK-NEXT: vredand.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.and.nxv2i8(<vscale x 2 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.or.nxv2i8(<vscale x 2 x i8>) |
| |
| define signext i8 @vreduce_or_nxv2i8(<vscale x 2 x i8> %v) { |
| ; CHECK-LABEL: vreduce_or_nxv2i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf4,ta,mu |
| ; CHECK-NEXT: vredor.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.or.nxv2i8(<vscale x 2 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.xor.nxv2i8(<vscale x 2 x i8>) |
| |
| define signext i8 @vreduce_xor_nxv2i8(<vscale x 2 x i8> %v) { |
| ; CHECK-LABEL: vreduce_xor_nxv2i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf4,ta,mu |
| ; CHECK-NEXT: vredxor.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.xor.nxv2i8(<vscale x 2 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.add.nxv4i8(<vscale x 4 x i8>) |
| |
| define signext i8 @vreduce_add_nxv4i8(<vscale x 4 x i8> %v) { |
| ; CHECK-LABEL: vreduce_add_nxv4i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf2,ta,mu |
| ; CHECK-NEXT: vredsum.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.add.nxv4i8(<vscale x 4 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.umax.nxv4i8(<vscale x 4 x i8>) |
| |
| define signext i8 @vreduce_umax_nxv4i8(<vscale x 4 x i8> %v) { |
| ; CHECK-LABEL: vreduce_umax_nxv4i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf2,ta,mu |
| ; CHECK-NEXT: vredmaxu.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.umax.nxv4i8(<vscale x 4 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.smax.nxv4i8(<vscale x 4 x i8>) |
| |
| define signext i8 @vreduce_smax_nxv4i8(<vscale x 4 x i8> %v) { |
| ; CHECK-LABEL: vreduce_smax_nxv4i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: addi a0, zero, -128 |
| ; CHECK-NEXT: vsetvli a1, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf2,ta,mu |
| ; CHECK-NEXT: vredmax.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.smax.nxv4i8(<vscale x 4 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.umin.nxv4i8(<vscale x 4 x i8>) |
| |
| define signext i8 @vreduce_umin_nxv4i8(<vscale x 4 x i8> %v) { |
| ; CHECK-LABEL: vreduce_umin_nxv4i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf2,ta,mu |
| ; CHECK-NEXT: vredminu.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.umin.nxv4i8(<vscale x 4 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.smin.nxv4i8(<vscale x 4 x i8>) |
| |
| define signext i8 @vreduce_smin_nxv4i8(<vscale x 4 x i8> %v) { |
| ; CHECK-LABEL: vreduce_smin_nxv4i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: addi a0, zero, 127 |
| ; CHECK-NEXT: vsetvli a1, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf2,ta,mu |
| ; CHECK-NEXT: vredmin.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.smin.nxv4i8(<vscale x 4 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.and.nxv4i8(<vscale x 4 x i8>) |
| |
| define signext i8 @vreduce_and_nxv4i8(<vscale x 4 x i8> %v) { |
| ; CHECK-LABEL: vreduce_and_nxv4i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf2,ta,mu |
| ; CHECK-NEXT: vredand.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.and.nxv4i8(<vscale x 4 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.or.nxv4i8(<vscale x 4 x i8>) |
| |
| define signext i8 @vreduce_or_nxv4i8(<vscale x 4 x i8> %v) { |
| ; CHECK-LABEL: vreduce_or_nxv4i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf2,ta,mu |
| ; CHECK-NEXT: vredor.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.or.nxv4i8(<vscale x 4 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i8 @llvm.vector.reduce.xor.nxv4i8(<vscale x 4 x i8>) |
| |
| define signext i8 @vreduce_xor_nxv4i8(<vscale x 4 x i8> %v) { |
| ; CHECK-LABEL: vreduce_xor_nxv4i8: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e8,mf2,ta,mu |
| ; CHECK-NEXT: vredxor.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e8,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i8 @llvm.vector.reduce.xor.nxv4i8(<vscale x 4 x i8> %v) |
| ret i8 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.add.nxv1i16(<vscale x 1 x i16>) |
| |
| define signext i16 @vreduce_add_nxv1i16(<vscale x 1 x i16> %v) { |
| ; CHECK-LABEL: vreduce_add_nxv1i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e16,mf4,ta,mu |
| ; CHECK-NEXT: vredsum.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.add.nxv1i16(<vscale x 1 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.umax.nxv1i16(<vscale x 1 x i16>) |
| |
| define signext i16 @vreduce_umax_nxv1i16(<vscale x 1 x i16> %v) { |
| ; CHECK-LABEL: vreduce_umax_nxv1i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e16,mf4,ta,mu |
| ; CHECK-NEXT: vredmaxu.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.umax.nxv1i16(<vscale x 1 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.smax.nxv1i16(<vscale x 1 x i16>) |
| |
| define signext i16 @vreduce_smax_nxv1i16(<vscale x 1 x i16> %v) { |
| ; CHECK-LABEL: vreduce_smax_nxv1i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: lui a0, 1048568 |
| ; CHECK-NEXT: vsetvli a1, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: vsetvli a0, zero, e16,mf4,ta,mu |
| ; CHECK-NEXT: vredmax.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.smax.nxv1i16(<vscale x 1 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.umin.nxv1i16(<vscale x 1 x i16>) |
| |
| define signext i16 @vreduce_umin_nxv1i16(<vscale x 1 x i16> %v) { |
| ; CHECK-LABEL: vreduce_umin_nxv1i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vsetvli a0, zero, e16,mf4,ta,mu |
| ; CHECK-NEXT: vredminu.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.umin.nxv1i16(<vscale x 1 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.smin.nxv1i16(<vscale x 1 x i16>) |
| |
| define signext i16 @vreduce_smin_nxv1i16(<vscale x 1 x i16> %v) { |
| ; CHECK-LABEL: vreduce_smin_nxv1i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: lui a0, 8 |
| ; CHECK-NEXT: addi a0, a0, -1 |
| ; CHECK-NEXT: vsetvli a1, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: vsetvli a0, zero, e16,mf4,ta,mu |
| ; CHECK-NEXT: vredmin.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.smin.nxv1i16(<vscale x 1 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.and.nxv1i16(<vscale x 1 x i16>) |
| |
| define signext i16 @vreduce_and_nxv1i16(<vscale x 1 x i16> %v) { |
| ; CHECK-LABEL: vreduce_and_nxv1i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vsetvli a0, zero, e16,mf4,ta,mu |
| ; CHECK-NEXT: vredand.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.and.nxv1i16(<vscale x 1 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.or.nxv1i16(<vscale x 1 x i16>) |
| |
| define signext i16 @vreduce_or_nxv1i16(<vscale x 1 x i16> %v) { |
| ; CHECK-LABEL: vreduce_or_nxv1i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e16,mf4,ta,mu |
| ; CHECK-NEXT: vredor.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.or.nxv1i16(<vscale x 1 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.xor.nxv1i16(<vscale x 1 x i16>) |
| |
| define signext i16 @vreduce_xor_nxv1i16(<vscale x 1 x i16> %v) { |
| ; CHECK-LABEL: vreduce_xor_nxv1i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e16,mf4,ta,mu |
| ; CHECK-NEXT: vredxor.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.xor.nxv1i16(<vscale x 1 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.add.nxv2i16(<vscale x 2 x i16>) |
| |
| define signext i16 @vreduce_add_nxv2i16(<vscale x 2 x i16> %v) { |
| ; CHECK-LABEL: vreduce_add_nxv2i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e16,mf2,ta,mu |
| ; CHECK-NEXT: vredsum.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.add.nxv2i16(<vscale x 2 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.umax.nxv2i16(<vscale x 2 x i16>) |
| |
| define signext i16 @vreduce_umax_nxv2i16(<vscale x 2 x i16> %v) { |
| ; CHECK-LABEL: vreduce_umax_nxv2i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e16,mf2,ta,mu |
| ; CHECK-NEXT: vredmaxu.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.umax.nxv2i16(<vscale x 2 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.smax.nxv2i16(<vscale x 2 x i16>) |
| |
| define signext i16 @vreduce_smax_nxv2i16(<vscale x 2 x i16> %v) { |
| ; CHECK-LABEL: vreduce_smax_nxv2i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: lui a0, 1048568 |
| ; CHECK-NEXT: vsetvli a1, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: vsetvli a0, zero, e16,mf2,ta,mu |
| ; CHECK-NEXT: vredmax.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.smax.nxv2i16(<vscale x 2 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.umin.nxv2i16(<vscale x 2 x i16>) |
| |
| define signext i16 @vreduce_umin_nxv2i16(<vscale x 2 x i16> %v) { |
| ; CHECK-LABEL: vreduce_umin_nxv2i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vsetvli a0, zero, e16,mf2,ta,mu |
| ; CHECK-NEXT: vredminu.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.umin.nxv2i16(<vscale x 2 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.smin.nxv2i16(<vscale x 2 x i16>) |
| |
| define signext i16 @vreduce_smin_nxv2i16(<vscale x 2 x i16> %v) { |
| ; CHECK-LABEL: vreduce_smin_nxv2i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: lui a0, 8 |
| ; CHECK-NEXT: addi a0, a0, -1 |
| ; CHECK-NEXT: vsetvli a1, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: vsetvli a0, zero, e16,mf2,ta,mu |
| ; CHECK-NEXT: vredmin.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.smin.nxv2i16(<vscale x 2 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.and.nxv2i16(<vscale x 2 x i16>) |
| |
| define signext i16 @vreduce_and_nxv2i16(<vscale x 2 x i16> %v) { |
| ; CHECK-LABEL: vreduce_and_nxv2i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vsetvli a0, zero, e16,mf2,ta,mu |
| ; CHECK-NEXT: vredand.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.and.nxv2i16(<vscale x 2 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.or.nxv2i16(<vscale x 2 x i16>) |
| |
| define signext i16 @vreduce_or_nxv2i16(<vscale x 2 x i16> %v) { |
| ; CHECK-LABEL: vreduce_or_nxv2i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e16,mf2,ta,mu |
| ; CHECK-NEXT: vredor.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.or.nxv2i16(<vscale x 2 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.xor.nxv2i16(<vscale x 2 x i16>) |
| |
| define signext i16 @vreduce_xor_nxv2i16(<vscale x 2 x i16> %v) { |
| ; CHECK-LABEL: vreduce_xor_nxv2i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e16,mf2,ta,mu |
| ; CHECK-NEXT: vredxor.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.xor.nxv2i16(<vscale x 2 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.add.nxv4i16(<vscale x 4 x i16>) |
| |
| define signext i16 @vreduce_add_nxv4i16(<vscale x 4 x i16> %v) { |
| ; CHECK-LABEL: vreduce_add_nxv4i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vredsum.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.add.nxv4i16(<vscale x 4 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.umax.nxv4i16(<vscale x 4 x i16>) |
| |
| define signext i16 @vreduce_umax_nxv4i16(<vscale x 4 x i16> %v) { |
| ; CHECK-LABEL: vreduce_umax_nxv4i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vredmaxu.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.umax.nxv4i16(<vscale x 4 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.smax.nxv4i16(<vscale x 4 x i16>) |
| |
| define signext i16 @vreduce_smax_nxv4i16(<vscale x 4 x i16> %v) { |
| ; CHECK-LABEL: vreduce_smax_nxv4i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: lui a0, 1048568 |
| ; CHECK-NEXT: vsetvli a1, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: vredmax.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.smax.nxv4i16(<vscale x 4 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.umin.nxv4i16(<vscale x 4 x i16>) |
| |
| define signext i16 @vreduce_umin_nxv4i16(<vscale x 4 x i16> %v) { |
| ; CHECK-LABEL: vreduce_umin_nxv4i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vredminu.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.umin.nxv4i16(<vscale x 4 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.smin.nxv4i16(<vscale x 4 x i16>) |
| |
| define signext i16 @vreduce_smin_nxv4i16(<vscale x 4 x i16> %v) { |
| ; CHECK-LABEL: vreduce_smin_nxv4i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: lui a0, 8 |
| ; CHECK-NEXT: addi a0, a0, -1 |
| ; CHECK-NEXT: vsetvli a1, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: vredmin.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.smin.nxv4i16(<vscale x 4 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.and.nxv4i16(<vscale x 4 x i16>) |
| |
| define signext i16 @vreduce_and_nxv4i16(<vscale x 4 x i16> %v) { |
| ; CHECK-LABEL: vreduce_and_nxv4i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vredand.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.and.nxv4i16(<vscale x 4 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.or.nxv4i16(<vscale x 4 x i16>) |
| |
| define signext i16 @vreduce_or_nxv4i16(<vscale x 4 x i16> %v) { |
| ; CHECK-LABEL: vreduce_or_nxv4i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vredor.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.or.nxv4i16(<vscale x 4 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i16 @llvm.vector.reduce.xor.nxv4i16(<vscale x 4 x i16>) |
| |
| define signext i16 @vreduce_xor_nxv4i16(<vscale x 4 x i16> %v) { |
| ; CHECK-LABEL: vreduce_xor_nxv4i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e16,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vredxor.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i16 @llvm.vector.reduce.xor.nxv4i16(<vscale x 4 x i16> %v) |
| ret i16 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.add.nxv1i32(<vscale x 1 x i32>) |
| |
| define i32 @vreduce_add_nxv1i32(<vscale x 1 x i32> %v) { |
| ; CHECK-LABEL: vreduce_add_nxv1i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e32,mf2,ta,mu |
| ; CHECK-NEXT: vredsum.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.add.nxv1i32(<vscale x 1 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.umax.nxv1i32(<vscale x 1 x i32>) |
| |
| define i32 @vreduce_umax_nxv1i32(<vscale x 1 x i32> %v) { |
| ; CHECK-LABEL: vreduce_umax_nxv1i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e32,mf2,ta,mu |
| ; CHECK-NEXT: vredmaxu.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.umax.nxv1i32(<vscale x 1 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.smax.nxv1i32(<vscale x 1 x i32>) |
| |
| define i32 @vreduce_smax_nxv1i32(<vscale x 1 x i32> %v) { |
| ; CHECK-LABEL: vreduce_smax_nxv1i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: lui a0, 524288 |
| ; CHECK-NEXT: vsetvli a1, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: vsetvli a0, zero, e32,mf2,ta,mu |
| ; CHECK-NEXT: vredmax.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.smax.nxv1i32(<vscale x 1 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.umin.nxv1i32(<vscale x 1 x i32>) |
| |
| define i32 @vreduce_umin_nxv1i32(<vscale x 1 x i32> %v) { |
| ; CHECK-LABEL: vreduce_umin_nxv1i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vsetvli a0, zero, e32,mf2,ta,mu |
| ; CHECK-NEXT: vredminu.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.umin.nxv1i32(<vscale x 1 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.smin.nxv1i32(<vscale x 1 x i32>) |
| |
| define i32 @vreduce_smin_nxv1i32(<vscale x 1 x i32> %v) { |
| ; CHECK-LABEL: vreduce_smin_nxv1i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: lui a0, 524288 |
| ; CHECK-NEXT: addi a0, a0, -1 |
| ; CHECK-NEXT: vsetvli a1, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: vsetvli a0, zero, e32,mf2,ta,mu |
| ; CHECK-NEXT: vredmin.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.smin.nxv1i32(<vscale x 1 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.and.nxv1i32(<vscale x 1 x i32>) |
| |
| define i32 @vreduce_and_nxv1i32(<vscale x 1 x i32> %v) { |
| ; CHECK-LABEL: vreduce_and_nxv1i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vsetvli a0, zero, e32,mf2,ta,mu |
| ; CHECK-NEXT: vredand.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.and.nxv1i32(<vscale x 1 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.or.nxv1i32(<vscale x 1 x i32>) |
| |
| define i32 @vreduce_or_nxv1i32(<vscale x 1 x i32> %v) { |
| ; CHECK-LABEL: vreduce_or_nxv1i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e32,mf2,ta,mu |
| ; CHECK-NEXT: vredor.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.or.nxv1i32(<vscale x 1 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.xor.nxv1i32(<vscale x 1 x i32>) |
| |
| define i32 @vreduce_xor_nxv1i32(<vscale x 1 x i32> %v) { |
| ; CHECK-LABEL: vreduce_xor_nxv1i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e32,mf2,ta,mu |
| ; CHECK-NEXT: vredxor.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.xor.nxv1i32(<vscale x 1 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.add.nxv2i32(<vscale x 2 x i32>) |
| |
| define i32 @vreduce_add_nxv2i32(<vscale x 2 x i32> %v) { |
| ; CHECK-LABEL: vreduce_add_nxv2i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vredsum.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.add.nxv2i32(<vscale x 2 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.umax.nxv2i32(<vscale x 2 x i32>) |
| |
| define i32 @vreduce_umax_nxv2i32(<vscale x 2 x i32> %v) { |
| ; CHECK-LABEL: vreduce_umax_nxv2i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vredmaxu.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.umax.nxv2i32(<vscale x 2 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.smax.nxv2i32(<vscale x 2 x i32>) |
| |
| define i32 @vreduce_smax_nxv2i32(<vscale x 2 x i32> %v) { |
| ; CHECK-LABEL: vreduce_smax_nxv2i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: lui a0, 524288 |
| ; CHECK-NEXT: vsetvli a1, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: vredmax.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.smax.nxv2i32(<vscale x 2 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.umin.nxv2i32(<vscale x 2 x i32>) |
| |
| define i32 @vreduce_umin_nxv2i32(<vscale x 2 x i32> %v) { |
| ; CHECK-LABEL: vreduce_umin_nxv2i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vredminu.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.umin.nxv2i32(<vscale x 2 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.smin.nxv2i32(<vscale x 2 x i32>) |
| |
| define i32 @vreduce_smin_nxv2i32(<vscale x 2 x i32> %v) { |
| ; CHECK-LABEL: vreduce_smin_nxv2i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: lui a0, 524288 |
| ; CHECK-NEXT: addi a0, a0, -1 |
| ; CHECK-NEXT: vsetvli a1, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: vredmin.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.smin.nxv2i32(<vscale x 2 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.and.nxv2i32(<vscale x 2 x i32>) |
| |
| define i32 @vreduce_and_nxv2i32(<vscale x 2 x i32> %v) { |
| ; CHECK-LABEL: vreduce_and_nxv2i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vredand.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.and.nxv2i32(<vscale x 2 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.or.nxv2i32(<vscale x 2 x i32>) |
| |
| define i32 @vreduce_or_nxv2i32(<vscale x 2 x i32> %v) { |
| ; CHECK-LABEL: vreduce_or_nxv2i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vredor.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.or.nxv2i32(<vscale x 2 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.xor.nxv2i32(<vscale x 2 x i32>) |
| |
| define i32 @vreduce_xor_nxv2i32(<vscale x 2 x i32> %v) { |
| ; CHECK-LABEL: vreduce_xor_nxv2i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vredxor.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.xor.nxv2i32(<vscale x 2 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.add.nxv4i32(<vscale x 4 x i32>) |
| |
| define i32 @vreduce_add_nxv4i32(<vscale x 4 x i32> %v) { |
| ; CHECK-LABEL: vreduce_add_nxv4i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m2,ta,mu |
| ; CHECK-NEXT: vredsum.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.add.nxv4i32(<vscale x 4 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.umax.nxv4i32(<vscale x 4 x i32>) |
| |
| define i32 @vreduce_umax_nxv4i32(<vscale x 4 x i32> %v) { |
| ; CHECK-LABEL: vreduce_umax_nxv4i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m2,ta,mu |
| ; CHECK-NEXT: vredmaxu.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.umax.nxv4i32(<vscale x 4 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.smax.nxv4i32(<vscale x 4 x i32>) |
| |
| define i32 @vreduce_smax_nxv4i32(<vscale x 4 x i32> %v) { |
| ; CHECK-LABEL: vreduce_smax_nxv4i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: lui a0, 524288 |
| ; CHECK-NEXT: vsetvli a1, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m2,ta,mu |
| ; CHECK-NEXT: vredmax.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.smax.nxv4i32(<vscale x 4 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.umin.nxv4i32(<vscale x 4 x i32>) |
| |
| define i32 @vreduce_umin_nxv4i32(<vscale x 4 x i32> %v) { |
| ; CHECK-LABEL: vreduce_umin_nxv4i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m2,ta,mu |
| ; CHECK-NEXT: vredminu.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.umin.nxv4i32(<vscale x 4 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.smin.nxv4i32(<vscale x 4 x i32>) |
| |
| define i32 @vreduce_smin_nxv4i32(<vscale x 4 x i32> %v) { |
| ; CHECK-LABEL: vreduce_smin_nxv4i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: lui a0, 524288 |
| ; CHECK-NEXT: addi a0, a0, -1 |
| ; CHECK-NEXT: vsetvli a1, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m2,ta,mu |
| ; CHECK-NEXT: vredmin.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.smin.nxv4i32(<vscale x 4 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.and.nxv4i32(<vscale x 4 x i32>) |
| |
| define i32 @vreduce_and_nxv4i32(<vscale x 4 x i32> %v) { |
| ; CHECK-LABEL: vreduce_and_nxv4i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m2,ta,mu |
| ; CHECK-NEXT: vredand.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.and.nxv4i32(<vscale x 4 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.or.nxv4i32(<vscale x 4 x i32>) |
| |
| define i32 @vreduce_or_nxv4i32(<vscale x 4 x i32> %v) { |
| ; CHECK-LABEL: vreduce_or_nxv4i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m2,ta,mu |
| ; CHECK-NEXT: vredor.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.or.nxv4i32(<vscale x 4 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i32 @llvm.vector.reduce.xor.nxv4i32(<vscale x 4 x i32>) |
| |
| define i32 @vreduce_xor_nxv4i32(<vscale x 4 x i32> %v) { |
| ; CHECK-LABEL: vreduce_xor_nxv4i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e32,m2,ta,mu |
| ; CHECK-NEXT: vredxor.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e32,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: ret |
| %red = call i32 @llvm.vector.reduce.xor.nxv4i32(<vscale x 4 x i32> %v) |
| ret i32 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.add.nxv1i64(<vscale x 1 x i64>) |
| |
| define i64 @vreduce_add_nxv1i64(<vscale x 1 x i64> %v) { |
| ; CHECK-LABEL: vreduce_add_nxv1i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vredsum.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.add.nxv1i64(<vscale x 1 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.umax.nxv1i64(<vscale x 1 x i64>) |
| |
| define i64 @vreduce_umax_nxv1i64(<vscale x 1 x i64> %v) { |
| ; CHECK-LABEL: vreduce_umax_nxv1i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vredmaxu.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.umax.nxv1i64(<vscale x 1 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.smax.nxv1i64(<vscale x 1 x i64>) |
| |
| define i64 @vreduce_smax_nxv1i64(<vscale x 1 x i64> %v) { |
| ; CHECK-LABEL: vreduce_smax_nxv1i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: lui a0, 524288 |
| ; CHECK-NEXT: vsetvli a1, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsll.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.v.i v26, 0 |
| ; CHECK-NEXT: vsll.vx v26, v26, a1 |
| ; CHECK-NEXT: vsrl.vx v26, v26, a1 |
| ; CHECK-NEXT: vor.vv v25, v26, v25 |
| ; CHECK-NEXT: vredmax.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.smax.nxv1i64(<vscale x 1 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.umin.nxv1i64(<vscale x 1 x i64>) |
| |
| define i64 @vreduce_umin_nxv1i64(<vscale x 1 x i64> %v) { |
| ; CHECK-LABEL: vreduce_umin_nxv1i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vredminu.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.umin.nxv1i64(<vscale x 1 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.smin.nxv1i64(<vscale x 1 x i64>) |
| |
| define i64 @vreduce_smin_nxv1i64(<vscale x 1 x i64> %v) { |
| ; CHECK-LABEL: vreduce_smin_nxv1i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsll.vx v25, v25, a1 |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: lui a0, 524288 |
| ; CHECK-NEXT: addi a0, a0, -1 |
| ; CHECK-NEXT: vmv.v.x v26, a0 |
| ; CHECK-NEXT: vsll.vx v26, v26, a1 |
| ; CHECK-NEXT: vor.vv v25, v25, v26 |
| ; CHECK-NEXT: vredmin.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.smin.nxv1i64(<vscale x 1 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.and.nxv1i64(<vscale x 1 x i64>) |
| |
| define i64 @vreduce_and_nxv1i64(<vscale x 1 x i64> %v) { |
| ; CHECK-LABEL: vreduce_and_nxv1i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vredand.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.and.nxv1i64(<vscale x 1 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.or.nxv1i64(<vscale x 1 x i64>) |
| |
| define i64 @vreduce_or_nxv1i64(<vscale x 1 x i64> %v) { |
| ; CHECK-LABEL: vreduce_or_nxv1i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vredor.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.or.nxv1i64(<vscale x 1 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.xor.nxv1i64(<vscale x 1 x i64>) |
| |
| define i64 @vreduce_xor_nxv1i64(<vscale x 1 x i64> %v) { |
| ; CHECK-LABEL: vreduce_xor_nxv1i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vredxor.vs v25, v8, v25 |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.xor.nxv1i64(<vscale x 1 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.add.nxv2i64(<vscale x 2 x i64>) |
| |
| define i64 @vreduce_add_nxv2i64(<vscale x 2 x i64> %v) { |
| ; CHECK-LABEL: vreduce_add_nxv2i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m2,ta,mu |
| ; CHECK-NEXT: vredsum.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.add.nxv2i64(<vscale x 2 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.umax.nxv2i64(<vscale x 2 x i64>) |
| |
| define i64 @vreduce_umax_nxv2i64(<vscale x 2 x i64> %v) { |
| ; CHECK-LABEL: vreduce_umax_nxv2i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m2,ta,mu |
| ; CHECK-NEXT: vredmaxu.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.umax.nxv2i64(<vscale x 2 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.smax.nxv2i64(<vscale x 2 x i64>) |
| |
| define i64 @vreduce_smax_nxv2i64(<vscale x 2 x i64> %v) { |
| ; CHECK-LABEL: vreduce_smax_nxv2i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: lui a0, 524288 |
| ; CHECK-NEXT: vsetvli a1, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsll.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.v.i v26, 0 |
| ; CHECK-NEXT: vsll.vx v26, v26, a1 |
| ; CHECK-NEXT: vsrl.vx v26, v26, a1 |
| ; CHECK-NEXT: vor.vv v25, v26, v25 |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m2,ta,mu |
| ; CHECK-NEXT: vredmax.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.smax.nxv2i64(<vscale x 2 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.umin.nxv2i64(<vscale x 2 x i64>) |
| |
| define i64 @vreduce_umin_nxv2i64(<vscale x 2 x i64> %v) { |
| ; CHECK-LABEL: vreduce_umin_nxv2i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m2,ta,mu |
| ; CHECK-NEXT: vredminu.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.umin.nxv2i64(<vscale x 2 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.smin.nxv2i64(<vscale x 2 x i64>) |
| |
| define i64 @vreduce_smin_nxv2i64(<vscale x 2 x i64> %v) { |
| ; CHECK-LABEL: vreduce_smin_nxv2i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsll.vx v25, v25, a1 |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: lui a0, 524288 |
| ; CHECK-NEXT: addi a0, a0, -1 |
| ; CHECK-NEXT: vmv.v.x v26, a0 |
| ; CHECK-NEXT: vsll.vx v26, v26, a1 |
| ; CHECK-NEXT: vor.vv v25, v25, v26 |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m2,ta,mu |
| ; CHECK-NEXT: vredmin.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.smin.nxv2i64(<vscale x 2 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.and.nxv2i64(<vscale x 2 x i64>) |
| |
| define i64 @vreduce_and_nxv2i64(<vscale x 2 x i64> %v) { |
| ; CHECK-LABEL: vreduce_and_nxv2i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m2,ta,mu |
| ; CHECK-NEXT: vredand.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.and.nxv2i64(<vscale x 2 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.or.nxv2i64(<vscale x 2 x i64>) |
| |
| define i64 @vreduce_or_nxv2i64(<vscale x 2 x i64> %v) { |
| ; CHECK-LABEL: vreduce_or_nxv2i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m2,ta,mu |
| ; CHECK-NEXT: vredor.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.or.nxv2i64(<vscale x 2 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.xor.nxv2i64(<vscale x 2 x i64>) |
| |
| define i64 @vreduce_xor_nxv2i64(<vscale x 2 x i64> %v) { |
| ; CHECK-LABEL: vreduce_xor_nxv2i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m2,ta,mu |
| ; CHECK-NEXT: vredxor.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.xor.nxv2i64(<vscale x 2 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.add.nxv4i64(<vscale x 4 x i64>) |
| |
| define i64 @vreduce_add_nxv4i64(<vscale x 4 x i64> %v) { |
| ; CHECK-LABEL: vreduce_add_nxv4i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m4,ta,mu |
| ; CHECK-NEXT: vredsum.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.add.nxv4i64(<vscale x 4 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.umax.nxv4i64(<vscale x 4 x i64>) |
| |
| define i64 @vreduce_umax_nxv4i64(<vscale x 4 x i64> %v) { |
| ; CHECK-LABEL: vreduce_umax_nxv4i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m4,ta,mu |
| ; CHECK-NEXT: vredmaxu.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.umax.nxv4i64(<vscale x 4 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.smax.nxv4i64(<vscale x 4 x i64>) |
| |
| define i64 @vreduce_smax_nxv4i64(<vscale x 4 x i64> %v) { |
| ; CHECK-LABEL: vreduce_smax_nxv4i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: lui a0, 524288 |
| ; CHECK-NEXT: vsetvli a1, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.x v25, a0 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsll.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.v.i v26, 0 |
| ; CHECK-NEXT: vsll.vx v26, v26, a1 |
| ; CHECK-NEXT: vsrl.vx v26, v26, a1 |
| ; CHECK-NEXT: vor.vv v25, v26, v25 |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m4,ta,mu |
| ; CHECK-NEXT: vredmax.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.smax.nxv4i64(<vscale x 4 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.umin.nxv4i64(<vscale x 4 x i64>) |
| |
| define i64 @vreduce_umin_nxv4i64(<vscale x 4 x i64> %v) { |
| ; CHECK-LABEL: vreduce_umin_nxv4i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m4,ta,mu |
| ; CHECK-NEXT: vredminu.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.umin.nxv4i64(<vscale x 4 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.smin.nxv4i64(<vscale x 4 x i64>) |
| |
| define i64 @vreduce_smin_nxv4i64(<vscale x 4 x i64> %v) { |
| ; CHECK-LABEL: vreduce_smin_nxv4i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsll.vx v25, v25, a1 |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: lui a0, 524288 |
| ; CHECK-NEXT: addi a0, a0, -1 |
| ; CHECK-NEXT: vmv.v.x v26, a0 |
| ; CHECK-NEXT: vsll.vx v26, v26, a1 |
| ; CHECK-NEXT: vor.vv v25, v25, v26 |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m4,ta,mu |
| ; CHECK-NEXT: vredmin.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.smin.nxv4i64(<vscale x 4 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.and.nxv4i64(<vscale x 4 x i64>) |
| |
| define i64 @vreduce_and_nxv4i64(<vscale x 4 x i64> %v) { |
| ; CHECK-LABEL: vreduce_and_nxv4i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, -1 |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m4,ta,mu |
| ; CHECK-NEXT: vredand.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.and.nxv4i64(<vscale x 4 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.or.nxv4i64(<vscale x 4 x i64>) |
| |
| define i64 @vreduce_or_nxv4i64(<vscale x 4 x i64> %v) { |
| ; CHECK-LABEL: vreduce_or_nxv4i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m4,ta,mu |
| ; CHECK-NEXT: vredor.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.or.nxv4i64(<vscale x 4 x i64> %v) |
| ret i64 %red |
| } |
| |
| declare i64 @llvm.vector.reduce.xor.nxv4i64(<vscale x 4 x i64>) |
| |
| define i64 @vreduce_xor_nxv4i64(<vscale x 4 x i64> %v) { |
| ; CHECK-LABEL: vreduce_xor_nxv4i64: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.v.i v25, 0 |
| ; CHECK-NEXT: vsetvli a0, zero, e64,m4,ta,mu |
| ; CHECK-NEXT: vredxor.vs v25, v8, v25 |
| ; CHECK-NEXT: vsetvli zero, zero, e64,m1,ta,mu |
| ; CHECK-NEXT: vmv.x.s a0, v25 |
| ; CHECK-NEXT: addi a1, zero, 32 |
| ; CHECK-NEXT: vsetivli a2, 1, e64,m1,ta,mu |
| ; CHECK-NEXT: vsrl.vx v25, v25, a1 |
| ; CHECK-NEXT: vmv.x.s a1, v25 |
| ; CHECK-NEXT: ret |
| %red = call i64 @llvm.vector.reduce.xor.nxv4i64(<vscale x 4 x i64> %v) |
| ret i64 %red |
| } |