| ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 |
| ; RUN: llc -mtriple aarch64 < %s | FileCheck %s --check-prefix=NEON |
| ; RUN: llc -mtriple aarch64 -mattr=+sve < %s | FileCheck %s --check-prefix=SVE |
| |
| ; Legal |
| define <4 x i32> @srem_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i1> %m) { |
| ; NEON-LABEL: srem_v4i32: |
| ; NEON: // %bb.0: |
| ; NEON-NEXT: ushll v2.4s, v2.4h, #0 |
| ; NEON-NEXT: fmov w8, s0 |
| ; NEON-NEXT: mov w11, v0.s[1] |
| ; NEON-NEXT: mov w14, v0.s[2] |
| ; NEON-NEXT: mov w18, v0.s[3] |
| ; NEON-NEXT: shl v2.4s, v2.4s, #31 |
| ; NEON-NEXT: cmlt v2.4s, v2.4s, #0 |
| ; NEON-NEXT: and v1.16b, v1.16b, v2.16b |
| ; NEON-NEXT: mvn v2.16b, v2.16b |
| ; NEON-NEXT: sub v1.4s, v1.4s, v2.4s |
| ; NEON-NEXT: fmov w9, s1 |
| ; NEON-NEXT: mov w12, v1.s[1] |
| ; NEON-NEXT: mov w15, v1.s[2] |
| ; NEON-NEXT: mov w17, v1.s[3] |
| ; NEON-NEXT: sdiv w10, w8, w9 |
| ; NEON-NEXT: sdiv w13, w11, w12 |
| ; NEON-NEXT: msub w8, w10, w9, w8 |
| ; NEON-NEXT: fmov s0, w8 |
| ; NEON-NEXT: sdiv w16, w14, w15 |
| ; NEON-NEXT: msub w9, w13, w12, w11 |
| ; NEON-NEXT: mov v0.s[1], w9 |
| ; NEON-NEXT: sdiv w10, w18, w17 |
| ; NEON-NEXT: msub w8, w16, w15, w14 |
| ; NEON-NEXT: mov v0.s[2], w8 |
| ; NEON-NEXT: msub w8, w10, w17, w18 |
| ; NEON-NEXT: mov v0.s[3], w8 |
| ; NEON-NEXT: ret |
| ; |
| ; SVE-LABEL: srem_v4i32: |
| ; SVE: // %bb.0: |
| ; SVE-NEXT: ushll v2.4s, v2.4h, #0 |
| ; SVE-NEXT: ptrue p0.s, vl4 |
| ; SVE-NEXT: // kill: def $q0 killed $q0 def $z0 |
| ; SVE-NEXT: shl v2.4s, v2.4s, #31 |
| ; SVE-NEXT: cmlt v2.4s, v2.4s, #0 |
| ; SVE-NEXT: and v1.16b, v1.16b, v2.16b |
| ; SVE-NEXT: mvn v2.16b, v2.16b |
| ; SVE-NEXT: sub v1.4s, v1.4s, v2.4s |
| ; SVE-NEXT: movprfx z2, z0 |
| ; SVE-NEXT: sdiv z2.s, p0/m, z2.s, z1.s |
| ; SVE-NEXT: mls v0.4s, v2.4s, v1.4s |
| ; SVE-NEXT: // kill: def $q0 killed $q0 killed $z0 |
| ; SVE-NEXT: ret |
| %res = call <4 x i32> @llvm.masked.srem(<4 x i32> %x, <4 x i32> %y, <4 x i1> %m) |
| ret <4 x i32> %res |
| } |
| |
| define <2 x i64> @srem_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i1> %m) { |
| ; NEON-LABEL: srem_v2i64: |
| ; NEON: // %bb.0: |
| ; NEON-NEXT: ushll v2.2d, v2.2s, #0 |
| ; NEON-NEXT: fmov x8, d0 |
| ; NEON-NEXT: mov x11, v0.d[1] |
| ; NEON-NEXT: shl v2.2d, v2.2d, #63 |
| ; NEON-NEXT: cmlt v2.2d, v2.2d, #0 |
| ; NEON-NEXT: and v1.16b, v1.16b, v2.16b |
| ; NEON-NEXT: mvn v2.16b, v2.16b |
| ; NEON-NEXT: sub v1.2d, v1.2d, v2.2d |
| ; NEON-NEXT: fmov x9, d1 |
| ; NEON-NEXT: mov x12, v1.d[1] |
| ; NEON-NEXT: sdiv x10, x8, x9 |
| ; NEON-NEXT: sdiv x13, x11, x12 |
| ; NEON-NEXT: msub x8, x10, x9, x8 |
| ; NEON-NEXT: fmov d0, x8 |
| ; NEON-NEXT: msub x9, x13, x12, x11 |
| ; NEON-NEXT: mov v0.d[1], x9 |
| ; NEON-NEXT: ret |
| ; |
| ; SVE-LABEL: srem_v2i64: |
| ; SVE: // %bb.0: |
| ; SVE-NEXT: ushll v2.2d, v2.2s, #0 |
| ; SVE-NEXT: ptrue p0.d, vl2 |
| ; SVE-NEXT: // kill: def $q0 killed $q0 def $z0 |
| ; SVE-NEXT: shl v2.2d, v2.2d, #63 |
| ; SVE-NEXT: cmlt v2.2d, v2.2d, #0 |
| ; SVE-NEXT: and v1.16b, v1.16b, v2.16b |
| ; SVE-NEXT: mvn v2.16b, v2.16b |
| ; SVE-NEXT: sub v1.2d, v1.2d, v2.2d |
| ; SVE-NEXT: movprfx z2, z0 |
| ; SVE-NEXT: sdiv z2.d, p0/m, z2.d, z1.d |
| ; SVE-NEXT: mls z0.d, p0/m, z2.d, z1.d |
| ; SVE-NEXT: // kill: def $q0 killed $q0 killed $z0 |
| ; SVE-NEXT: ret |
| %res = call <2 x i64> @llvm.masked.srem(<2 x i64> %x, <2 x i64> %y, <2 x i1> %m) |
| ret <2 x i64> %res |
| } |
| |
| ; Splitting |
| define <4 x i64> @srem_v4i64(<4 x i64> %x, <4 x i64> %y, <4 x i1> %m) { |
| ; NEON-LABEL: srem_v4i64: |
| ; NEON: // %bb.0: |
| ; NEON-NEXT: ushll v4.4s, v4.4h, #0 |
| ; NEON-NEXT: mov x8, v1.d[1] |
| ; NEON-NEXT: fmov x11, d1 |
| ; NEON-NEXT: fmov x15, d0 |
| ; NEON-NEXT: mov x18, v0.d[1] |
| ; NEON-NEXT: ushll2 v5.2d, v4.4s, #0 |
| ; NEON-NEXT: ushll v1.2d, v4.2s, #0 |
| ; NEON-NEXT: shl v5.2d, v5.2d, #63 |
| ; NEON-NEXT: shl v1.2d, v1.2d, #63 |
| ; NEON-NEXT: cmlt v5.2d, v5.2d, #0 |
| ; NEON-NEXT: cmlt v1.2d, v1.2d, #0 |
| ; NEON-NEXT: and v3.16b, v3.16b, v5.16b |
| ; NEON-NEXT: mvn v5.16b, v5.16b |
| ; NEON-NEXT: and v2.16b, v2.16b, v1.16b |
| ; NEON-NEXT: mvn v1.16b, v1.16b |
| ; NEON-NEXT: sub v3.2d, v3.2d, v5.2d |
| ; NEON-NEXT: sub v1.2d, v2.2d, v1.2d |
| ; NEON-NEXT: mov x9, v3.d[1] |
| ; NEON-NEXT: fmov x12, d3 |
| ; NEON-NEXT: fmov x14, d1 |
| ; NEON-NEXT: mov x17, v1.d[1] |
| ; NEON-NEXT: sdiv x13, x11, x12 |
| ; NEON-NEXT: sdiv x10, x8, x9 |
| ; NEON-NEXT: sdiv x16, x15, x14 |
| ; NEON-NEXT: msub x8, x10, x9, x8 |
| ; NEON-NEXT: msub x9, x13, x12, x11 |
| ; NEON-NEXT: fmov d1, x9 |
| ; NEON-NEXT: mov v1.d[1], x8 |
| ; NEON-NEXT: sdiv x0, x18, x17 |
| ; NEON-NEXT: msub x10, x16, x14, x15 |
| ; NEON-NEXT: fmov d0, x10 |
| ; NEON-NEXT: msub x11, x0, x17, x18 |
| ; NEON-NEXT: mov v0.d[1], x11 |
| ; NEON-NEXT: ret |
| ; |
| ; SVE-LABEL: srem_v4i64: |
| ; SVE: // %bb.0: |
| ; SVE-NEXT: ushll v4.4s, v4.4h, #0 |
| ; SVE-NEXT: ptrue p0.d, vl2 |
| ; SVE-NEXT: // kill: def $q1 killed $q1 def $z1 |
| ; SVE-NEXT: // kill: def $q0 killed $q0 def $z0 |
| ; SVE-NEXT: ushll2 v5.2d, v4.4s, #0 |
| ; SVE-NEXT: ushll v4.2d, v4.2s, #0 |
| ; SVE-NEXT: shl v5.2d, v5.2d, #63 |
| ; SVE-NEXT: shl v4.2d, v4.2d, #63 |
| ; SVE-NEXT: cmlt v5.2d, v5.2d, #0 |
| ; SVE-NEXT: cmlt v4.2d, v4.2d, #0 |
| ; SVE-NEXT: and v3.16b, v3.16b, v5.16b |
| ; SVE-NEXT: mvn v5.16b, v5.16b |
| ; SVE-NEXT: and v2.16b, v2.16b, v4.16b |
| ; SVE-NEXT: mvn v4.16b, v4.16b |
| ; SVE-NEXT: sub v3.2d, v3.2d, v5.2d |
| ; SVE-NEXT: sub v2.2d, v2.2d, v4.2d |
| ; SVE-NEXT: movprfx z5, z1 |
| ; SVE-NEXT: sdiv z5.d, p0/m, z5.d, z3.d |
| ; SVE-NEXT: movprfx z4, z0 |
| ; SVE-NEXT: sdiv z4.d, p0/m, z4.d, z2.d |
| ; SVE-NEXT: mls z1.d, p0/m, z5.d, z3.d |
| ; SVE-NEXT: mls z0.d, p0/m, z4.d, z2.d |
| ; SVE-NEXT: // kill: def $q1 killed $q1 killed $z1 |
| ; SVE-NEXT: // kill: def $q0 killed $q0 killed $z0 |
| ; SVE-NEXT: ret |
| %res = call <4 x i64> @llvm.masked.srem(<4 x i64> %x, <4 x i64> %y, <4 x i1> %m) |
| ret <4 x i64> %res |
| } |
| |
| ; Widening |
| define <2 x i32> @srem_v2i32(<2 x i32> %x, <2 x i32> %y, <2 x i1> %m) { |
| ; NEON-LABEL: srem_v2i32: |
| ; NEON: // %bb.0: |
| ; NEON-NEXT: shl v2.2s, v2.2s, #31 |
| ; NEON-NEXT: // kill: def $d0 killed $d0 def $q0 |
| ; NEON-NEXT: fmov w8, s0 |
| ; NEON-NEXT: mov w11, v0.s[1] |
| ; NEON-NEXT: cmlt v2.2s, v2.2s, #0 |
| ; NEON-NEXT: and v1.8b, v1.8b, v2.8b |
| ; NEON-NEXT: mvn v2.8b, v2.8b |
| ; NEON-NEXT: sub v1.2s, v1.2s, v2.2s |
| ; NEON-NEXT: fmov w9, s1 |
| ; NEON-NEXT: mov w12, v1.s[1] |
| ; NEON-NEXT: sdiv w10, w8, w9 |
| ; NEON-NEXT: sdiv w13, w11, w12 |
| ; NEON-NEXT: msub w8, w10, w9, w8 |
| ; NEON-NEXT: fmov s0, w8 |
| ; NEON-NEXT: msub w9, w13, w12, w11 |
| ; NEON-NEXT: mov v0.s[1], w9 |
| ; NEON-NEXT: // kill: def $d0 killed $d0 killed $q0 |
| ; NEON-NEXT: ret |
| ; |
| ; SVE-LABEL: srem_v2i32: |
| ; SVE: // %bb.0: |
| ; SVE-NEXT: shl v2.2s, v2.2s, #31 |
| ; SVE-NEXT: ptrue p0.s, vl2 |
| ; SVE-NEXT: // kill: def $d0 killed $d0 def $z0 |
| ; SVE-NEXT: cmlt v2.2s, v2.2s, #0 |
| ; SVE-NEXT: and v1.8b, v1.8b, v2.8b |
| ; SVE-NEXT: mvn v2.8b, v2.8b |
| ; SVE-NEXT: sub v1.2s, v1.2s, v2.2s |
| ; SVE-NEXT: movprfx z2, z0 |
| ; SVE-NEXT: sdiv z2.s, p0/m, z2.s, z1.s |
| ; SVE-NEXT: mls v0.2s, v2.2s, v1.2s |
| ; SVE-NEXT: // kill: def $d0 killed $d0 killed $z0 |
| ; SVE-NEXT: ret |
| %res = call <2 x i32> @llvm.masked.srem(<2 x i32> %x, <2 x i32> %y, <2 x i1> %m) |
| ret <2 x i32> %res |
| } |
| |
| ; Promotion |
| define <4 x i16> @srem_v4i16(<4 x i16> %x, <4 x i16> %y, <4 x i1> %m) { |
| ; NEON-LABEL: srem_v4i16: |
| ; NEON: // %bb.0: |
| ; NEON-NEXT: shl v2.4h, v2.4h, #15 |
| ; NEON-NEXT: // kill: def $d0 killed $d0 def $q0 |
| ; NEON-NEXT: smov w11, v0.h[0] |
| ; NEON-NEXT: smov w8, v0.h[1] |
| ; NEON-NEXT: smov w14, v0.h[2] |
| ; NEON-NEXT: smov w17, v0.h[3] |
| ; NEON-NEXT: cmlt v2.4h, v2.4h, #0 |
| ; NEON-NEXT: and v1.8b, v1.8b, v2.8b |
| ; NEON-NEXT: mvn v2.8b, v2.8b |
| ; NEON-NEXT: sub v1.4h, v1.4h, v2.4h |
| ; NEON-NEXT: smov w12, v1.h[0] |
| ; NEON-NEXT: smov w9, v1.h[1] |
| ; NEON-NEXT: smov w15, v1.h[2] |
| ; NEON-NEXT: smov w18, v1.h[3] |
| ; NEON-NEXT: sdiv w13, w11, w12 |
| ; NEON-NEXT: sdiv w10, w8, w9 |
| ; NEON-NEXT: msub w11, w13, w12, w11 |
| ; NEON-NEXT: fmov s0, w11 |
| ; NEON-NEXT: sdiv w16, w14, w15 |
| ; NEON-NEXT: msub w8, w10, w9, w8 |
| ; NEON-NEXT: mov v0.h[1], w8 |
| ; NEON-NEXT: sdiv w9, w17, w18 |
| ; NEON-NEXT: msub w8, w16, w15, w14 |
| ; NEON-NEXT: mov v0.h[2], w8 |
| ; NEON-NEXT: msub w8, w9, w18, w17 |
| ; NEON-NEXT: mov v0.h[3], w8 |
| ; NEON-NEXT: // kill: def $d0 killed $d0 killed $q0 |
| ; NEON-NEXT: ret |
| ; |
| ; SVE-LABEL: srem_v4i16: |
| ; SVE: // %bb.0: |
| ; SVE-NEXT: shl v2.4h, v2.4h, #15 |
| ; SVE-NEXT: ptrue p0.s, vl4 |
| ; SVE-NEXT: cmlt v2.4h, v2.4h, #0 |
| ; SVE-NEXT: and v1.8b, v1.8b, v2.8b |
| ; SVE-NEXT: mvn v2.8b, v2.8b |
| ; SVE-NEXT: sub v1.4h, v1.4h, v2.4h |
| ; SVE-NEXT: sshll v2.4s, v0.4h, #0 |
| ; SVE-NEXT: sshll v3.4s, v1.4h, #0 |
| ; SVE-NEXT: sdiv z2.s, p0/m, z2.s, z3.s |
| ; SVE-NEXT: xtn v2.4h, v2.4s |
| ; SVE-NEXT: mls v0.4h, v2.4h, v1.4h |
| ; SVE-NEXT: ret |
| %res = call <4 x i16> @llvm.masked.srem(<4 x i16> %x, <4 x i16> %y, <4 x i1> %m) |
| ret <4 x i16> %res |
| } |
| |
| ; Scalarization |
| define <1 x i64> @srem_v1i164(<1 x i64> %x, <1 x i64> %y, <1 x i1> %m) { |
| ; NEON-LABEL: srem_v1i164: |
| ; NEON: // %bb.0: |
| ; NEON-NEXT: // kill: def $d1 killed $d1 def $q1 |
| ; NEON-NEXT: fmov x8, d1 |
| ; NEON-NEXT: // kill: def $d0 killed $d0 def $q0 |
| ; NEON-NEXT: fmov x9, d0 |
| ; NEON-NEXT: tst w0, #0x1 |
| ; NEON-NEXT: csinc x8, x8, xzr, ne |
| ; NEON-NEXT: sdiv x10, x9, x8 |
| ; NEON-NEXT: msub x8, x10, x8, x9 |
| ; NEON-NEXT: fmov d0, x8 |
| ; NEON-NEXT: ret |
| ; |
| ; SVE-LABEL: srem_v1i164: |
| ; SVE: // %bb.0: |
| ; SVE-NEXT: // kill: def $d1 killed $d1 def $q1 |
| ; SVE-NEXT: fmov x8, d1 |
| ; SVE-NEXT: // kill: def $d0 killed $d0 def $q0 |
| ; SVE-NEXT: fmov x9, d0 |
| ; SVE-NEXT: tst w0, #0x1 |
| ; SVE-NEXT: csinc x8, x8, xzr, ne |
| ; SVE-NEXT: sdiv x10, x9, x8 |
| ; SVE-NEXT: msub x8, x10, x8, x9 |
| ; SVE-NEXT: fmov d0, x8 |
| ; SVE-NEXT: ret |
| %res = call <1 x i64> @llvm.masked.srem(<1 x i64> %x, <1 x i64> %y, <1 x i1> %m) |
| ret <1 x i64> %res |
| } |
| |
| ; Expansion |
| define <2 x i128> @srem_v2i128(<2 x i128> %x, <2 x i128> %y, <2 x i1> %m) nounwind { |
| ; NEON-LABEL: srem_v2i128: |
| ; NEON: // %bb.0: |
| ; NEON-NEXT: stp x30, x25, [sp, #-64]! // 16-byte Folded Spill |
| ; NEON-NEXT: // kill: def $d0 killed $d0 def $q0 |
| ; NEON-NEXT: fmov w8, s0 |
| ; NEON-NEXT: stp x22, x21, [sp, #32] // 16-byte Folded Spill |
| ; NEON-NEXT: mov x21, x3 |
| ; NEON-NEXT: mov x22, x2 |
| ; NEON-NEXT: stp x24, x23, [sp, #16] // 16-byte Folded Spill |
| ; NEON-NEXT: mov w25, v0.s[1] |
| ; NEON-NEXT: stp x20, x19, [sp, #48] // 16-byte Folded Spill |
| ; NEON-NEXT: mov x19, x7 |
| ; NEON-NEXT: mov x20, x6 |
| ; NEON-NEXT: tst w8, #0x1 |
| ; NEON-NEXT: csel x3, x5, xzr, ne |
| ; NEON-NEXT: csinc x2, x4, xzr, ne |
| ; NEON-NEXT: bl __modti3 |
| ; NEON-NEXT: tst w25, #0x1 |
| ; NEON-NEXT: mov x23, x0 |
| ; NEON-NEXT: mov x24, x1 |
| ; NEON-NEXT: csel x3, x19, xzr, ne |
| ; NEON-NEXT: csinc x2, x20, xzr, ne |
| ; NEON-NEXT: mov x0, x22 |
| ; NEON-NEXT: mov x1, x21 |
| ; NEON-NEXT: bl __modti3 |
| ; NEON-NEXT: mov x2, x0 |
| ; NEON-NEXT: mov x3, x1 |
| ; NEON-NEXT: mov x0, x23 |
| ; NEON-NEXT: mov x1, x24 |
| ; NEON-NEXT: ldp x20, x19, [sp, #48] // 16-byte Folded Reload |
| ; NEON-NEXT: ldp x22, x21, [sp, #32] // 16-byte Folded Reload |
| ; NEON-NEXT: ldp x24, x23, [sp, #16] // 16-byte Folded Reload |
| ; NEON-NEXT: ldp x30, x25, [sp], #64 // 16-byte Folded Reload |
| ; NEON-NEXT: ret |
| ; |
| ; SVE-LABEL: srem_v2i128: |
| ; SVE: // %bb.0: |
| ; SVE-NEXT: stp x30, x25, [sp, #-64]! // 16-byte Folded Spill |
| ; SVE-NEXT: // kill: def $d0 killed $d0 def $q0 |
| ; SVE-NEXT: fmov w8, s0 |
| ; SVE-NEXT: stp x22, x21, [sp, #32] // 16-byte Folded Spill |
| ; SVE-NEXT: mov x21, x3 |
| ; SVE-NEXT: mov x22, x2 |
| ; SVE-NEXT: stp x24, x23, [sp, #16] // 16-byte Folded Spill |
| ; SVE-NEXT: mov w25, v0.s[1] |
| ; SVE-NEXT: stp x20, x19, [sp, #48] // 16-byte Folded Spill |
| ; SVE-NEXT: mov x19, x7 |
| ; SVE-NEXT: mov x20, x6 |
| ; SVE-NEXT: tst w8, #0x1 |
| ; SVE-NEXT: csel x3, x5, xzr, ne |
| ; SVE-NEXT: csinc x2, x4, xzr, ne |
| ; SVE-NEXT: bl __modti3 |
| ; SVE-NEXT: tst w25, #0x1 |
| ; SVE-NEXT: mov x23, x0 |
| ; SVE-NEXT: mov x24, x1 |
| ; SVE-NEXT: csel x3, x19, xzr, ne |
| ; SVE-NEXT: csinc x2, x20, xzr, ne |
| ; SVE-NEXT: mov x0, x22 |
| ; SVE-NEXT: mov x1, x21 |
| ; SVE-NEXT: bl __modti3 |
| ; SVE-NEXT: mov x2, x0 |
| ; SVE-NEXT: mov x3, x1 |
| ; SVE-NEXT: mov x0, x23 |
| ; SVE-NEXT: mov x1, x24 |
| ; SVE-NEXT: ldp x20, x19, [sp, #48] // 16-byte Folded Reload |
| ; SVE-NEXT: ldp x22, x21, [sp, #32] // 16-byte Folded Reload |
| ; SVE-NEXT: ldp x24, x23, [sp, #16] // 16-byte Folded Reload |
| ; SVE-NEXT: ldp x30, x25, [sp], #64 // 16-byte Folded Reload |
| ; SVE-NEXT: ret |
| %res = call <2 x i128> @llvm.masked.srem(<2 x i128> %x, <2 x i128> %y, <2 x i1> %m) |
| ret <2 x i128> %res |
| } |
| |
| ; Promotion and widening |
| define <3 x i10> @srem_v3i10(<3 x i10> %x, <3 x i10> %y, <3 x i1> %m) { |
| ; NEON-LABEL: srem_v3i10: |
| ; NEON: // %bb.0: |
| ; NEON-NEXT: movi v0.2d, #0000000000000000 |
| ; NEON-NEXT: fmov s1, w3 |
| ; NEON-NEXT: ldr w8, [sp] |
| ; NEON-NEXT: fmov s2, w0 |
| ; NEON-NEXT: mov v1.h[1], w4 |
| ; NEON-NEXT: mov v0.h[0], w6 |
| ; NEON-NEXT: mov v2.h[1], w1 |
| ; NEON-NEXT: mov v1.h[2], w5 |
| ; NEON-NEXT: mov v0.h[1], w7 |
| ; NEON-NEXT: mov v2.h[2], w2 |
| ; NEON-NEXT: shl v1.4h, v1.4h, #6 |
| ; NEON-NEXT: mov v0.h[2], w8 |
| ; NEON-NEXT: shl v2.4h, v2.4h, #6 |
| ; NEON-NEXT: sshr v1.4h, v1.4h, #6 |
| ; NEON-NEXT: sshr v2.4h, v2.4h, #6 |
| ; NEON-NEXT: shl v0.4h, v0.4h, #15 |
| ; NEON-NEXT: smov w8, v2.h[0] |
| ; NEON-NEXT: smov w11, v2.h[1] |
| ; NEON-NEXT: smov w14, v2.h[2] |
| ; NEON-NEXT: cmlt v0.4h, v0.4h, #0 |
| ; NEON-NEXT: and v1.8b, v1.8b, v0.8b |
| ; NEON-NEXT: mvn v0.8b, v0.8b |
| ; NEON-NEXT: sub v0.4h, v1.4h, v0.4h |
| ; NEON-NEXT: smov w9, v0.h[0] |
| ; NEON-NEXT: smov w12, v0.h[1] |
| ; NEON-NEXT: smov w15, v0.h[2] |
| ; NEON-NEXT: sdiv w10, w8, w9 |
| ; NEON-NEXT: sdiv w13, w11, w12 |
| ; NEON-NEXT: msub w0, w10, w9, w8 |
| ; NEON-NEXT: sdiv w16, w14, w15 |
| ; NEON-NEXT: msub w1, w13, w12, w11 |
| ; NEON-NEXT: msub w2, w16, w15, w14 |
| ; NEON-NEXT: ret |
| ; |
| ; SVE-LABEL: srem_v3i10: |
| ; SVE: // %bb.0: |
| ; SVE-NEXT: movi v0.2d, #0000000000000000 |
| ; SVE-NEXT: fmov s1, w3 |
| ; SVE-NEXT: ldr w8, [sp] |
| ; SVE-NEXT: fmov s2, w0 |
| ; SVE-NEXT: ptrue p0.s, vl4 |
| ; SVE-NEXT: mov v1.h[1], w4 |
| ; SVE-NEXT: mov v0.h[0], w6 |
| ; SVE-NEXT: mov v2.h[1], w1 |
| ; SVE-NEXT: mov v1.h[2], w5 |
| ; SVE-NEXT: mov v0.h[1], w7 |
| ; SVE-NEXT: mov v2.h[2], w2 |
| ; SVE-NEXT: shl v1.4h, v1.4h, #6 |
| ; SVE-NEXT: mov v0.h[2], w8 |
| ; SVE-NEXT: shl v2.4h, v2.4h, #6 |
| ; SVE-NEXT: sshr v1.4h, v1.4h, #6 |
| ; SVE-NEXT: sshr v2.4h, v2.4h, #6 |
| ; SVE-NEXT: shl v0.4h, v0.4h, #15 |
| ; SVE-NEXT: sshll v3.4s, v2.4h, #0 |
| ; SVE-NEXT: cmlt v0.4h, v0.4h, #0 |
| ; SVE-NEXT: and v1.8b, v1.8b, v0.8b |
| ; SVE-NEXT: mvn v0.8b, v0.8b |
| ; SVE-NEXT: sub v0.4h, v1.4h, v0.4h |
| ; SVE-NEXT: sshll v1.4s, v0.4h, #0 |
| ; SVE-NEXT: sdivr z1.s, p0/m, z1.s, z3.s |
| ; SVE-NEXT: xtn v1.4h, v1.4s |
| ; SVE-NEXT: mls v2.4h, v1.4h, v0.4h |
| ; SVE-NEXT: umov w0, v2.h[0] |
| ; SVE-NEXT: umov w1, v2.h[1] |
| ; SVE-NEXT: umov w2, v2.h[2] |
| ; SVE-NEXT: ret |
| %res = call <3 x i10> @llvm.masked.srem(<3 x i10> %x, <3 x i10> %y, <3 x i1> %m) |
| ret <3 x i10> %res |
| } |
