derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
394 lines
13 KiB
LLVM
394 lines
13 KiB
LLVM
; 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> @sdiv_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i1> %m) {
|
|
; NEON-LABEL: sdiv_v4i32:
|
|
; NEON: // %bb.0:
|
|
; NEON-NEXT: ushll v2.4s, v2.4h, #0
|
|
; NEON-NEXT: fmov w8, s0
|
|
; NEON-NEXT: mov w12, 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 w10, v1.s[1]
|
|
; NEON-NEXT: mov w11, v1.s[2]
|
|
; NEON-NEXT: sdiv w8, w8, w9
|
|
; NEON-NEXT: mov w9, v0.s[1]
|
|
; NEON-NEXT: sdiv w9, w9, w10
|
|
; NEON-NEXT: mov w10, v0.s[2]
|
|
; NEON-NEXT: fmov s0, w8
|
|
; NEON-NEXT: sdiv w10, w10, w11
|
|
; NEON-NEXT: mov w11, v1.s[3]
|
|
; NEON-NEXT: mov v0.s[1], w9
|
|
; NEON-NEXT: sdiv w8, w12, w11
|
|
; NEON-NEXT: mov v0.s[2], w10
|
|
; NEON-NEXT: mov v0.s[3], w8
|
|
; NEON-NEXT: ret
|
|
;
|
|
; SVE-LABEL: sdiv_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: // kill: def $q1 killed $q1 def $z1
|
|
; SVE-NEXT: shl v2.4s, v2.4s, #31
|
|
; SVE-NEXT: cmpne p1.s, p0/z, z2.s, #0
|
|
; SVE-NEXT: sdiv z0.s, p1/m, z0.s, z1.s
|
|
; SVE-NEXT: // kill: def $q0 killed $q0 killed $z0
|
|
; SVE-NEXT: ret
|
|
%res = call <4 x i32> @llvm.masked.sdiv(<4 x i32> %x, <4 x i32> %y, <4 x i1> %m)
|
|
ret <4 x i32> %res
|
|
}
|
|
|
|
define <2 x i64> @sdiv_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i1> %m) {
|
|
; NEON-LABEL: sdiv_v2i64:
|
|
; NEON: // %bb.0:
|
|
; NEON-NEXT: ushll v2.2d, v2.2s, #0
|
|
; NEON-NEXT: fmov x8, d0
|
|
; 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 x10, v1.d[1]
|
|
; NEON-NEXT: sdiv x8, x8, x9
|
|
; NEON-NEXT: mov x9, v0.d[1]
|
|
; NEON-NEXT: sdiv x9, x9, x10
|
|
; NEON-NEXT: fmov d0, x8
|
|
; NEON-NEXT: mov v0.d[1], x9
|
|
; NEON-NEXT: ret
|
|
;
|
|
; SVE-LABEL: sdiv_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: // kill: def $q1 killed $q1 def $z1
|
|
; SVE-NEXT: shl v2.2d, v2.2d, #63
|
|
; SVE-NEXT: cmpne p1.d, p0/z, z2.d, #0
|
|
; SVE-NEXT: sdiv z0.d, p1/m, z0.d, z1.d
|
|
; SVE-NEXT: // kill: def $q0 killed $q0 killed $z0
|
|
; SVE-NEXT: ret
|
|
%res = call <2 x i64> @llvm.masked.sdiv(<2 x i64> %x, <2 x i64> %y, <2 x i1> %m)
|
|
ret <2 x i64> %res
|
|
}
|
|
|
|
; Splitting
|
|
define <4 x i64> @sdiv_v4i64(<4 x i64> %x, <4 x i64> %y, <4 x i1> %m) {
|
|
; NEON-LABEL: sdiv_v4i64:
|
|
; NEON: // %bb.0:
|
|
; NEON-NEXT: ushll v4.4s, v4.4h, #0
|
|
; NEON-NEXT: mov x8, v1.d[1]
|
|
; NEON-NEXT: fmov x11, d0
|
|
; NEON-NEXT: mov x12, v0.d[1]
|
|
; NEON-NEXT: ushll2 v5.2d, v4.4s, #0
|
|
; NEON-NEXT: shl v5.2d, v5.2d, #63
|
|
; NEON-NEXT: cmlt v5.2d, v5.2d, #0
|
|
; NEON-NEXT: and v3.16b, v3.16b, v5.16b
|
|
; NEON-NEXT: mvn v5.16b, v5.16b
|
|
; NEON-NEXT: sub v3.2d, v3.2d, v5.2d
|
|
; NEON-NEXT: mov x9, v3.d[1]
|
|
; NEON-NEXT: fmov x10, d3
|
|
; NEON-NEXT: sdiv x8, x8, x9
|
|
; NEON-NEXT: fmov x9, d1
|
|
; NEON-NEXT: ushll v1.2d, v4.2s, #0
|
|
; NEON-NEXT: shl v1.2d, v1.2d, #63
|
|
; NEON-NEXT: cmlt v1.2d, v1.2d, #0
|
|
; NEON-NEXT: and v2.16b, v2.16b, v1.16b
|
|
; NEON-NEXT: mvn v1.16b, v1.16b
|
|
; NEON-NEXT: sub v1.2d, v2.2d, v1.2d
|
|
; NEON-NEXT: sdiv x9, x9, x10
|
|
; NEON-NEXT: fmov x10, d1
|
|
; NEON-NEXT: sdiv x10, x11, x10
|
|
; NEON-NEXT: mov x11, v1.d[1]
|
|
; NEON-NEXT: fmov d1, x9
|
|
; NEON-NEXT: mov v1.d[1], x8
|
|
; NEON-NEXT: sdiv x11, x12, x11
|
|
; NEON-NEXT: fmov d0, x10
|
|
; NEON-NEXT: mov v0.d[1], x11
|
|
; NEON-NEXT: ret
|
|
;
|
|
; SVE-LABEL: sdiv_v4i64:
|
|
; SVE: // %bb.0:
|
|
; SVE-NEXT: ushll v4.4s, v4.4h, #0
|
|
; SVE-NEXT: ptrue p0.d, vl2
|
|
; SVE-NEXT: // kill: def $q3 killed $q3 def $z3
|
|
; SVE-NEXT: // kill: def $q1 killed $q1 def $z1
|
|
; SVE-NEXT: // kill: def $q0 killed $q0 def $z0
|
|
; SVE-NEXT: // kill: def $q2 killed $q2 def $z2
|
|
; SVE-NEXT: ushll2 v5.2d, v4.4s, #0
|
|
; SVE-NEXT: shl v5.2d, v5.2d, #63
|
|
; SVE-NEXT: cmpne p1.d, p0/z, z5.d, #0
|
|
; SVE-NEXT: sdiv z1.d, p1/m, z1.d, z3.d
|
|
; SVE-NEXT: ushll v3.2d, v4.2s, #0
|
|
; SVE-NEXT: // kill: def $q1 killed $q1 killed $z1
|
|
; SVE-NEXT: shl v3.2d, v3.2d, #63
|
|
; SVE-NEXT: cmpne p1.d, p0/z, z3.d, #0
|
|
; SVE-NEXT: sdiv z0.d, p1/m, z0.d, z2.d
|
|
; SVE-NEXT: // kill: def $q0 killed $q0 killed $z0
|
|
; SVE-NEXT: ret
|
|
%res = call <4 x i64> @llvm.masked.sdiv(<4 x i64> %x, <4 x i64> %y, <4 x i1> %m)
|
|
ret <4 x i64> %res
|
|
}
|
|
|
|
; Widening
|
|
define <2 x i32> @sdiv_v2i32(<2 x i32> %x, <2 x i32> %y, <2 x i1> %m) {
|
|
; NEON-LABEL: sdiv_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: 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 w10, v1.s[1]
|
|
; NEON-NEXT: sdiv w8, w8, w9
|
|
; NEON-NEXT: mov w9, v0.s[1]
|
|
; NEON-NEXT: sdiv w9, w9, w10
|
|
; NEON-NEXT: fmov s0, w8
|
|
; NEON-NEXT: mov v0.s[1], w9
|
|
; NEON-NEXT: // kill: def $d0 killed $d0 killed $q0
|
|
; NEON-NEXT: ret
|
|
;
|
|
; SVE-LABEL: sdiv_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: // kill: def $d1 killed $d1 def $z1
|
|
; SVE-NEXT: cmpne p1.s, p0/z, z2.s, #0
|
|
; SVE-NEXT: sdiv z0.s, p1/m, z0.s, z1.s
|
|
; SVE-NEXT: // kill: def $d0 killed $d0 killed $z0
|
|
; SVE-NEXT: ret
|
|
%res = call <2 x i32> @llvm.masked.sdiv(<2 x i32> %x, <2 x i32> %y, <2 x i1> %m)
|
|
ret <2 x i32> %res
|
|
}
|
|
|
|
; Promotion
|
|
define <4 x i16> @sdiv_v4i16(<4 x i16> %x, <4 x i16> %y, <4 x i1> %m) {
|
|
; NEON-LABEL: sdiv_v4i16:
|
|
; NEON: // %bb.0:
|
|
; NEON-NEXT: shl v2.4h, v2.4h, #15
|
|
; NEON-NEXT: // kill: def $d0 killed $d0 def $q0
|
|
; NEON-NEXT: smov w8, v0.h[1]
|
|
; 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 w9, v1.h[1]
|
|
; NEON-NEXT: smov w10, v1.h[0]
|
|
; NEON-NEXT: smov w11, v1.h[2]
|
|
; NEON-NEXT: smov w12, v1.h[3]
|
|
; NEON-NEXT: sdiv w8, w8, w9
|
|
; NEON-NEXT: smov w9, v0.h[0]
|
|
; NEON-NEXT: sdiv w9, w9, w10
|
|
; NEON-NEXT: smov w10, v0.h[2]
|
|
; NEON-NEXT: sdiv w10, w10, w11
|
|
; NEON-NEXT: smov w11, v0.h[3]
|
|
; NEON-NEXT: fmov s0, w9
|
|
; NEON-NEXT: mov v0.h[1], w8
|
|
; NEON-NEXT: sdiv w8, w11, w12
|
|
; NEON-NEXT: mov v0.h[2], w10
|
|
; NEON-NEXT: mov v0.h[3], w8
|
|
; NEON-NEXT: // kill: def $d0 killed $d0 killed $q0
|
|
; NEON-NEXT: ret
|
|
;
|
|
; SVE-LABEL: sdiv_v4i16:
|
|
; SVE: // %bb.0:
|
|
; SVE-NEXT: shl v2.4h, v2.4h, #15
|
|
; SVE-NEXT: sshll v0.4s, v0.4h, #0
|
|
; 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 v1.4s, v1.4h, #0
|
|
; SVE-NEXT: sdiv z0.s, p0/m, z0.s, z1.s
|
|
; SVE-NEXT: xtn v0.4h, v0.4s
|
|
; SVE-NEXT: ret
|
|
%res = call <4 x i16> @llvm.masked.sdiv(<4 x i16> %x, <4 x i16> %y, <4 x i1> %m)
|
|
ret <4 x i16> %res
|
|
}
|
|
|
|
; Scalarization
|
|
define <1 x i64> @sdiv_v1i164(<1 x i64> %x, <1 x i64> %y, <1 x i1> %m) {
|
|
; NEON-LABEL: sdiv_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 x8, x9, x8
|
|
; NEON-NEXT: fmov d0, x8
|
|
; NEON-NEXT: ret
|
|
;
|
|
; SVE-LABEL: sdiv_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 x8, x9, x8
|
|
; SVE-NEXT: fmov d0, x8
|
|
; SVE-NEXT: ret
|
|
%res = call <1 x i64> @llvm.masked.sdiv(<1 x i64> %x, <1 x i64> %y, <1 x i1> %m)
|
|
ret <1 x i64> %res
|
|
}
|
|
|
|
; Expansion
|
|
define <2 x i128> @sdiv_v2i128(<2 x i128> %x, <2 x i128> %y, <2 x i1> %m) nounwind {
|
|
; NEON-LABEL: sdiv_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 __divti3
|
|
; 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 __divti3
|
|
; 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: sdiv_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 __divti3
|
|
; 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 __divti3
|
|
; 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.sdiv(<2 x i128> %x, <2 x i128> %y, <2 x i1> %m)
|
|
ret <2 x i128> %res
|
|
}
|
|
|
|
; Promotion and widening
|
|
define <3 x i10> @sdiv_v3i10(<3 x i10> %x, <3 x i10> %y, <3 x i1> %m) {
|
|
; NEON-LABEL: sdiv_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: 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: sdiv w0, w8, w9
|
|
; NEON-NEXT: smov w8, v2.h[1]
|
|
; NEON-NEXT: smov w9, v0.h[1]
|
|
; NEON-NEXT: sdiv w1, w8, w9
|
|
; NEON-NEXT: smov w8, v2.h[2]
|
|
; NEON-NEXT: smov w9, v0.h[2]
|
|
; NEON-NEXT: sdiv w2, w8, w9
|
|
; NEON-NEXT: ret
|
|
;
|
|
; SVE-LABEL: sdiv_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: 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, v2.4h, #0
|
|
; SVE-NEXT: sshll v0.4s, v0.4h, #0
|
|
; SVE-NEXT: sdivr z0.s, p0/m, z0.s, z1.s
|
|
; SVE-NEXT: xtn v0.4h, v0.4s
|
|
; SVE-NEXT: umov w0, v0.h[0]
|
|
; SVE-NEXT: umov w1, v0.h[1]
|
|
; SVE-NEXT: umov w2, v0.h[2]
|
|
; SVE-NEXT: ret
|
|
%res = call <3 x i10> @llvm.masked.sdiv(<3 x i10> %x, <3 x i10> %y, <3 x i1> %m)
|
|
ret <3 x i10> %res
|
|
}
|