; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mattr=+sve -force-streaming-compatible < %s | FileCheck %s --check-prefixes=CHECK,SVE
; RUN: llc -mattr=+sve2 -force-streaming-compatible < %s | FileCheck %s --check-prefixes=CHECK,SVE2
; RUN: llc -mattr=+sme -force-streaming < %s | FileCheck %s --check-prefixes=CHECK,SVE2
; RUN: llc -force-streaming-compatible < %s | FileCheck %s --check-prefix=NONEON-NOSVE
target triple = "aarch64-unknown-linux-gnu"
;
; SDIV
;
define <4 x i8> @sdiv_v4i8(<4 x i8> %op1, <4 x i8> %op2) {
; CHECK-LABEL: sdiv_v4i8:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.h, vl4
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT: sxtb z0.h, p0/m, z0.h
; CHECK-NEXT: sxtb z1.h, p0/m, z1.h
; CHECK-NEXT: ptrue p0.s, vl4
; CHECK-NEXT: sunpklo z1.s, z1.h
; CHECK-NEXT: sunpklo z0.s, z0.h
; CHECK-NEXT: sdiv z0.s, p0/m, z0.s, z1.s
; CHECK-NEXT: uzp1 z0.h, z0.h, z0.h
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: sdiv_v4i8:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: shl v0.4h, v0.4h, #8
; NEON-NOSVE-NEXT: shl v1.4h, v1.4h, #8
; NEON-NOSVE-NEXT: ptrue p0.s, vl4
; NEON-NOSVE-NEXT: sshr v0.4h, v0.4h, #8
; NEON-NOSVE-NEXT: sshr v1.4h, v1.4h, #8
; NEON-NOSVE-NEXT: sshll v1.4s, v1.4h, #0
; NEON-NOSVE-NEXT: sshll v0.4s, v0.4h, #0
; NEON-NOSVE-NEXT: sdiv z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: xtn v0.4h, v0.4s
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: sdiv_v4i8:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #32
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #8]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #22]
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #14]
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #12]
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #10]
; NONEON-NOSVE-NEXT: ldrsb w12, [sp, #8]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #20]
; NONEON-NOSVE-NEXT: sdiv w9, w10, w9
; NONEON-NOSVE-NEXT: ldrsb w10, [sp, #18]
; NONEON-NOSVE-NEXT: strh w8, [sp, #30]
; NONEON-NOSVE-NEXT: sdiv w10, w11, w10
; NONEON-NOSVE-NEXT: ldrsb w11, [sp, #16]
; NONEON-NOSVE-NEXT: strh w9, [sp, #28]
; NONEON-NOSVE-NEXT: sdiv w11, w12, w11
; NONEON-NOSVE-NEXT: strh w10, [sp, #26]
; NONEON-NOSVE-NEXT: strh w11, [sp, #24]
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT: add sp, sp, #32
; NONEON-NOSVE-NEXT: ret
%res = sdiv <4 x i8> %op1, %op2
ret <4 x i8> %res
}
define <8 x i8> @sdiv_v8i8(<8 x i8> %op1, <8 x i8> %op2) {
; CHECK-LABEL: sdiv_v8i8:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT: ptrue p0.s, vl4
; CHECK-NEXT: sunpklo z1.h, z1.b
; CHECK-NEXT: sunpklo z0.h, z0.b
; CHECK-NEXT: sunpklo z2.s, z1.h
; CHECK-NEXT: sunpklo z3.s, z0.h
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
; CHECK-NEXT: sunpklo z1.s, z1.h
; CHECK-NEXT: sunpklo z0.s, z0.h
; CHECK-NEXT: sdivr z2.s, p0/m, z2.s, z3.s
; CHECK-NEXT: sdiv z0.s, p0/m, z0.s, z1.s
; CHECK-NEXT: ptrue p0.h, vl4
; CHECK-NEXT: uzp1 z1.h, z2.h, z2.h
; CHECK-NEXT: uzp1 z0.h, z0.h, z0.h
; CHECK-NEXT: splice z1.h, p0, z1.h, z0.h
; CHECK-NEXT: uzp1 z0.b, z1.b, z1.b
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: sdiv_v8i8:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: sshll v1.8h, v1.8b, #0
; NEON-NOSVE-NEXT: sshll v0.8h, v0.8b, #0
; NEON-NOSVE-NEXT: ptrue p0.s, vl4
; NEON-NOSVE-NEXT: sshll2 v2.4s, v1.8h, #0
; NEON-NOSVE-NEXT: sshll2 v3.4s, v0.8h, #0
; NEON-NOSVE-NEXT: sshll v1.4s, v1.4h, #0
; NEON-NOSVE-NEXT: sshll v0.4s, v0.4h, #0
; NEON-NOSVE-NEXT: sdivr z2.s, p0/m, z2.s, z3.s
; NEON-NOSVE-NEXT: sdiv z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: uzp1 v0.8h, v0.8h, v2.8h
; NEON-NOSVE-NEXT: xtn v0.8b, v0.8h
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: sdiv_v8i8:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #32
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #8]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #23]
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #15]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #14]
; NONEON-NOSVE-NEXT: strb w8, [sp, #31]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #22]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #13]
; NONEON-NOSVE-NEXT: strb w8, [sp, #30]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #21]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #12]
; NONEON-NOSVE-NEXT: strb w8, [sp, #29]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #20]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #11]
; NONEON-NOSVE-NEXT: strb w8, [sp, #28]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #19]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #10]
; NONEON-NOSVE-NEXT: strb w8, [sp, #27]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #18]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #9]
; NONEON-NOSVE-NEXT: strb w8, [sp, #26]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #17]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #8]
; NONEON-NOSVE-NEXT: strb w8, [sp, #25]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #16]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: strb w8, [sp, #24]
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT: add sp, sp, #32
; NONEON-NOSVE-NEXT: ret
%res = sdiv <8 x i8> %op1, %op2
ret <8 x i8> %res
}
define <16 x i8> @sdiv_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
; CHECK-LABEL: sdiv_v16i8:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
; CHECK-NEXT: sunpklo z2.h, z1.b
; CHECK-NEXT: sunpklo z3.h, z0.b
; CHECK-NEXT: ptrue p0.s, vl4
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
; CHECK-NEXT: sunpklo z4.s, z2.h
; CHECK-NEXT: sunpklo z5.s, z3.h
; CHECK-NEXT: ext z2.b, z2.b, z2.b, #8
; CHECK-NEXT: ext z3.b, z3.b, z3.b, #8
; CHECK-NEXT: sunpklo z1.h, z1.b
; CHECK-NEXT: sunpklo z0.h, z0.b
; CHECK-NEXT: sunpklo z2.s, z2.h
; CHECK-NEXT: sunpklo z3.s, z3.h
; CHECK-NEXT: sdivr z4.s, p0/m, z4.s, z5.s
; CHECK-NEXT: sunpklo z5.s, z0.h
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
; CHECK-NEXT: sunpklo z0.s, z0.h
; CHECK-NEXT: sdivr z2.s, p0/m, z2.s, z3.s
; CHECK-NEXT: sunpklo z3.s, z1.h
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
; CHECK-NEXT: sunpklo z1.s, z1.h
; CHECK-NEXT: sdivr z3.s, p0/m, z3.s, z5.s
; CHECK-NEXT: uzp1 z2.h, z2.h, z2.h
; CHECK-NEXT: sdiv z0.s, p0/m, z0.s, z1.s
; CHECK-NEXT: uzp1 z1.h, z4.h, z4.h
; CHECK-NEXT: ptrue p0.h, vl4
; CHECK-NEXT: uzp1 z3.h, z3.h, z3.h
; CHECK-NEXT: splice z1.h, p0, z1.h, z2.h
; CHECK-NEXT: uzp1 z0.h, z0.h, z0.h
; CHECK-NEXT: splice z3.h, p0, z3.h, z0.h
; CHECK-NEXT: uzp1 z0.b, z1.b, z1.b
; CHECK-NEXT: ptrue p0.b, vl8
; CHECK-NEXT: uzp1 z1.b, z3.b, z3.b
; CHECK-NEXT: splice z0.b, p0, z0.b, z1.b
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: sdiv_v16i8:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: sshll2 v2.8h, v1.16b, #0
; NEON-NOSVE-NEXT: sshll2 v3.8h, v0.16b, #0
; NEON-NOSVE-NEXT: sshll v1.8h, v1.8b, #0
; NEON-NOSVE-NEXT: sshll v0.8h, v0.8b, #0
; NEON-NOSVE-NEXT: ptrue p0.s, vl4
; NEON-NOSVE-NEXT: sshll2 v4.4s, v2.8h, #0
; NEON-NOSVE-NEXT: sshll2 v5.4s, v3.8h, #0
; NEON-NOSVE-NEXT: sshll v2.4s, v2.4h, #0
; NEON-NOSVE-NEXT: sshll v3.4s, v3.4h, #0
; NEON-NOSVE-NEXT: sdivr z4.s, p0/m, z4.s, z5.s
; NEON-NOSVE-NEXT: sshll2 v5.4s, v0.8h, #0
; NEON-NOSVE-NEXT: sshll v0.4s, v0.4h, #0
; NEON-NOSVE-NEXT: sdivr z2.s, p0/m, z2.s, z3.s
; NEON-NOSVE-NEXT: sshll2 v3.4s, v1.8h, #0
; NEON-NOSVE-NEXT: sshll v1.4s, v1.4h, #0
; NEON-NOSVE-NEXT: sdivr z3.s, p0/m, z3.s, z5.s
; NEON-NOSVE-NEXT: sdiv z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: uzp1 v1.8h, v2.8h, v4.8h
; NEON-NOSVE-NEXT: uzp1 v0.8h, v0.8h, v3.8h
; NEON-NOSVE-NEXT: uzp1 v0.16b, v0.16b, v1.16b
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: sdiv_v16i8:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: stp q0, q1, [sp, #-48]!
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 48
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #31]
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #15]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #14]
; NONEON-NOSVE-NEXT: strb w8, [sp, #47]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #30]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #13]
; NONEON-NOSVE-NEXT: strb w8, [sp, #46]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #29]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #12]
; NONEON-NOSVE-NEXT: strb w8, [sp, #45]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #28]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #11]
; NONEON-NOSVE-NEXT: strb w8, [sp, #44]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #27]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #10]
; NONEON-NOSVE-NEXT: strb w8, [sp, #43]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #26]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #9]
; NONEON-NOSVE-NEXT: strb w8, [sp, #42]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #25]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #8]
; NONEON-NOSVE-NEXT: strb w8, [sp, #41]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #24]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #7]
; NONEON-NOSVE-NEXT: strb w8, [sp, #40]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #23]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #6]
; NONEON-NOSVE-NEXT: strb w8, [sp, #39]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #22]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #5]
; NONEON-NOSVE-NEXT: strb w8, [sp, #38]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #21]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #4]
; NONEON-NOSVE-NEXT: strb w8, [sp, #37]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #20]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #3]
; NONEON-NOSVE-NEXT: strb w8, [sp, #36]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #19]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #2]
; NONEON-NOSVE-NEXT: strb w8, [sp, #35]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #18]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #1]
; NONEON-NOSVE-NEXT: strb w8, [sp, #34]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #17]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp]
; NONEON-NOSVE-NEXT: strb w8, [sp, #33]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #16]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: strb w8, [sp, #32]
; NONEON-NOSVE-NEXT: ldr q0, [sp, #32]
; NONEON-NOSVE-NEXT: add sp, sp, #48
; NONEON-NOSVE-NEXT: ret
%res = sdiv <16 x i8> %op1, %op2
ret <16 x i8> %res
}
define void @sdiv_v32i8(ptr %a, ptr %b) {
; CHECK-LABEL: sdiv_v32i8:
; CHECK: // %bb.0:
; CHECK-NEXT: ldp q6, q3, [x1]
; CHECK-NEXT: ptrue p0.s, vl4
; CHECK-NEXT: ldr q2, [x0, #16]
; CHECK-NEXT: sunpklo z1.h, z3.b
; CHECK-NEXT: sunpklo z4.h, z2.b
; CHECK-NEXT: ext z3.b, z3.b, z3.b, #8
; CHECK-NEXT: ext z2.b, z2.b, z2.b, #8
; CHECK-NEXT: sunpklo z7.h, z6.b
; CHECK-NEXT: ext z6.b, z6.b, z6.b, #8
; CHECK-NEXT: sunpklo z3.h, z3.b
; CHECK-NEXT: sunpklo z0.s, z1.h
; CHECK-NEXT: sunpklo z5.s, z4.h
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
; CHECK-NEXT: ext z4.b, z4.b, z4.b, #8
; CHECK-NEXT: sunpklo z17.s, z7.h
; CHECK-NEXT: ext z7.b, z7.b, z7.b, #8
; CHECK-NEXT: sunpklo z6.h, z6.b
; CHECK-NEXT: sunpklo z1.s, z1.h
; CHECK-NEXT: sunpklo z4.s, z4.h
; CHECK-NEXT: sunpklo z7.s, z7.h
; CHECK-NEXT: sdivr z0.s, p0/m, z0.s, z5.s
; CHECK-NEXT: sdivr z1.s, p0/m, z1.s, z4.s
; CHECK-NEXT: sunpklo z4.h, z2.b
; CHECK-NEXT: sunpklo z2.s, z3.h
; CHECK-NEXT: ext z3.b, z3.b, z3.b, #8
; CHECK-NEXT: uzp1 z0.h, z0.h, z0.h
; CHECK-NEXT: sunpklo z5.s, z4.h
; CHECK-NEXT: ext z4.b, z4.b, z4.b, #8
; CHECK-NEXT: sunpklo z3.s, z3.h
; CHECK-NEXT: sunpklo z4.s, z4.h
; CHECK-NEXT: sdivr z2.s, p0/m, z2.s, z5.s
; CHECK-NEXT: ldr q5, [x0]
; CHECK-NEXT: sunpklo z16.h, z5.b
; CHECK-NEXT: ext z5.b, z5.b, z5.b, #8
; CHECK-NEXT: uzp1 z1.h, z1.h, z1.h
; CHECK-NEXT: sunpklo z5.h, z5.b
; CHECK-NEXT: sunpklo z18.s, z16.h
; CHECK-NEXT: ext z16.b, z16.b, z16.b, #8
; CHECK-NEXT: sunpklo z16.s, z16.h
; CHECK-NEXT: sdivr z17.s, p0/m, z17.s, z18.s
; CHECK-NEXT: sunpklo z18.s, z5.h
; CHECK-NEXT: ext z5.b, z5.b, z5.b, #8
; CHECK-NEXT: sunpklo z5.s, z5.h
; CHECK-NEXT: uzp1 z2.h, z2.h, z2.h
; CHECK-NEXT: sdivr z7.s, p0/m, z7.s, z16.s
; CHECK-NEXT: sunpklo z16.s, z6.h
; CHECK-NEXT: ext z6.b, z6.b, z6.b, #8
; CHECK-NEXT: sunpklo z6.s, z6.h
; CHECK-NEXT: sdivr z16.s, p0/m, z16.s, z18.s
; CHECK-NEXT: sdiv z5.s, p0/m, z5.s, z6.s
; CHECK-NEXT: uzp1 z6.h, z7.h, z7.h
; CHECK-NEXT: uzp1 z7.h, z16.h, z16.h
; CHECK-NEXT: sdivr z3.s, p0/m, z3.s, z4.s
; CHECK-NEXT: uzp1 z4.h, z17.h, z17.h
; CHECK-NEXT: ptrue p0.h, vl4
; CHECK-NEXT: splice z0.h, p0, z0.h, z1.h
; CHECK-NEXT: uzp1 z5.h, z5.h, z5.h
; CHECK-NEXT: splice z4.h, p0, z4.h, z6.h
; CHECK-NEXT: splice z7.h, p0, z7.h, z5.h
; CHECK-NEXT: uzp1 z0.b, z0.b, z0.b
; CHECK-NEXT: uzp1 z1.b, z4.b, z4.b
; CHECK-NEXT: uzp1 z3.h, z3.h, z3.h
; CHECK-NEXT: splice z2.h, p0, z2.h, z3.h
; CHECK-NEXT: uzp1 z3.b, z7.b, z7.b
; CHECK-NEXT: ptrue p0.b, vl8
; CHECK-NEXT: splice z1.b, p0, z1.b, z3.b
; CHECK-NEXT: uzp1 z2.b, z2.b, z2.b
; CHECK-NEXT: splice z0.b, p0, z0.b, z2.b
; CHECK-NEXT: stp q1, q0, [x0]
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: sdiv_v32i8:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ldp q6, q3, [x1]
; NEON-NOSVE-NEXT: ptrue p0.s, vl4
; NEON-NOSVE-NEXT: ldr q2, [x0, #16]
; NEON-NOSVE-NEXT: sshll2 v1.8h, v3.16b, #0
; NEON-NOSVE-NEXT: sshll2 v4.8h, v2.16b, #0
; NEON-NOSVE-NEXT: sshll v3.8h, v3.8b, #0
; NEON-NOSVE-NEXT: sshll v2.8h, v2.8b, #0
; NEON-NOSVE-NEXT: sshll2 v7.8h, v6.16b, #0
; NEON-NOSVE-NEXT: sshll v6.8h, v6.8b, #0
; NEON-NOSVE-NEXT: sshll2 v0.4s, v1.8h, #0
; NEON-NOSVE-NEXT: sshll2 v5.4s, v4.8h, #0
; NEON-NOSVE-NEXT: sshll v1.4s, v1.4h, #0
; NEON-NOSVE-NEXT: sshll v4.4s, v4.4h, #0
; NEON-NOSVE-NEXT: sshll2 v17.4s, v7.8h, #0
; NEON-NOSVE-NEXT: sshll v7.4s, v7.4h, #0
; NEON-NOSVE-NEXT: sdivr z0.s, p0/m, z0.s, z5.s
; NEON-NOSVE-NEXT: sshll2 v5.4s, v2.8h, #0
; NEON-NOSVE-NEXT: sshll v2.4s, v2.4h, #0
; NEON-NOSVE-NEXT: sdivr z1.s, p0/m, z1.s, z4.s
; NEON-NOSVE-NEXT: sshll2 v4.4s, v3.8h, #0
; NEON-NOSVE-NEXT: sshll v3.4s, v3.4h, #0
; NEON-NOSVE-NEXT: sdivr z4.s, p0/m, z4.s, z5.s
; NEON-NOSVE-NEXT: ldr q5, [x0]
; NEON-NOSVE-NEXT: sshll2 v16.8h, v5.16b, #0
; NEON-NOSVE-NEXT: sshll v5.8h, v5.8b, #0
; NEON-NOSVE-NEXT: uzp1 v0.8h, v1.8h, v0.8h
; NEON-NOSVE-NEXT: sshll2 v18.4s, v16.8h, #0
; NEON-NOSVE-NEXT: sshll v16.4s, v16.4h, #0
; NEON-NOSVE-NEXT: sdivr z17.s, p0/m, z17.s, z18.s
; NEON-NOSVE-NEXT: sshll2 v18.4s, v5.8h, #0
; NEON-NOSVE-NEXT: sshll v5.4s, v5.4h, #0
; NEON-NOSVE-NEXT: sdivr z7.s, p0/m, z7.s, z16.s
; NEON-NOSVE-NEXT: sshll2 v16.4s, v6.8h, #0
; NEON-NOSVE-NEXT: sshll v6.4s, v6.4h, #0
; NEON-NOSVE-NEXT: sdivr z16.s, p0/m, z16.s, z18.s
; NEON-NOSVE-NEXT: sdiv z5.s, p0/m, z5.s, z6.s
; NEON-NOSVE-NEXT: sdiv z2.s, p0/m, z2.s, z3.s
; NEON-NOSVE-NEXT: uzp1 v3.8h, v7.8h, v17.8h
; NEON-NOSVE-NEXT: uzp1 v5.8h, v5.8h, v16.8h
; NEON-NOSVE-NEXT: uzp1 v1.8h, v2.8h, v4.8h
; NEON-NOSVE-NEXT: uzp1 v2.16b, v5.16b, v3.16b
; NEON-NOSVE-NEXT: uzp1 v0.16b, v1.16b, v0.16b
; NEON-NOSVE-NEXT: stp q2, q0, [x0]
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: sdiv_v32i8:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #96
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 96
; NONEON-NOSVE-NEXT: ldp q3, q0, [x1]
; NONEON-NOSVE-NEXT: ldp q2, q1, [x0]
; NONEON-NOSVE-NEXT: stp q2, q3, [sp]
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #32]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #63]
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #47]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #46]
; NONEON-NOSVE-NEXT: strb w8, [sp, #95]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #62]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #45]
; NONEON-NOSVE-NEXT: strb w8, [sp, #94]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #61]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #44]
; NONEON-NOSVE-NEXT: strb w8, [sp, #93]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #60]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #43]
; NONEON-NOSVE-NEXT: strb w8, [sp, #92]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #59]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #42]
; NONEON-NOSVE-NEXT: strb w8, [sp, #91]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #58]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #41]
; NONEON-NOSVE-NEXT: strb w8, [sp, #90]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #57]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #40]
; NONEON-NOSVE-NEXT: strb w8, [sp, #89]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #56]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #39]
; NONEON-NOSVE-NEXT: strb w8, [sp, #88]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #55]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #38]
; NONEON-NOSVE-NEXT: strb w8, [sp, #87]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #54]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #37]
; NONEON-NOSVE-NEXT: strb w8, [sp, #86]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #53]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #36]
; NONEON-NOSVE-NEXT: strb w8, [sp, #85]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #52]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #35]
; NONEON-NOSVE-NEXT: strb w8, [sp, #84]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #51]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #34]
; NONEON-NOSVE-NEXT: strb w8, [sp, #83]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #50]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #33]
; NONEON-NOSVE-NEXT: strb w8, [sp, #82]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #49]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #32]
; NONEON-NOSVE-NEXT: strb w8, [sp, #81]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #48]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #15]
; NONEON-NOSVE-NEXT: strb w8, [sp, #80]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #31]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #14]
; NONEON-NOSVE-NEXT: strb w8, [sp, #79]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #30]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #13]
; NONEON-NOSVE-NEXT: strb w8, [sp, #78]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #29]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #12]
; NONEON-NOSVE-NEXT: strb w8, [sp, #77]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #28]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #11]
; NONEON-NOSVE-NEXT: strb w8, [sp, #76]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #27]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #10]
; NONEON-NOSVE-NEXT: strb w8, [sp, #75]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #26]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #9]
; NONEON-NOSVE-NEXT: strb w8, [sp, #74]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #25]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #8]
; NONEON-NOSVE-NEXT: strb w8, [sp, #73]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #24]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #7]
; NONEON-NOSVE-NEXT: strb w8, [sp, #72]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #23]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #6]
; NONEON-NOSVE-NEXT: strb w8, [sp, #71]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #22]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #5]
; NONEON-NOSVE-NEXT: strb w8, [sp, #70]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #21]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #4]
; NONEON-NOSVE-NEXT: strb w8, [sp, #69]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #20]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #3]
; NONEON-NOSVE-NEXT: strb w8, [sp, #68]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #19]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #2]
; NONEON-NOSVE-NEXT: strb w8, [sp, #67]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #18]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp, #1]
; NONEON-NOSVE-NEXT: strb w8, [sp, #66]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #17]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsb w9, [sp]
; NONEON-NOSVE-NEXT: strb w8, [sp, #65]
; NONEON-NOSVE-NEXT: ldrsb w8, [sp, #16]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: strb w8, [sp, #64]
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #64]
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
; NONEON-NOSVE-NEXT: add sp, sp, #96
; NONEON-NOSVE-NEXT: ret
%op1 = load <32 x i8>, ptr %a
%op2 = load <32 x i8>, ptr %b
%res = sdiv <32 x i8> %op1, %op2
store <32 x i8> %res, ptr %a
ret void
}
define <2 x i16> @sdiv_v2i16(<2 x i16> %op1, <2 x i16> %op2) {
; CHECK-LABEL: sdiv_v2i16:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.s, vl2
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT: sxth z1.s, p0/m, z1.s
; CHECK-NEXT: sxth z0.s, p0/m, z0.s
; CHECK-NEXT: sdiv z0.s, p0/m, z0.s, z1.s
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: sdiv_v2i16:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: shl v1.2s, v1.2s, #16
; NEON-NOSVE-NEXT: shl v0.2s, v0.2s, #16
; NEON-NOSVE-NEXT: ptrue p0.s, vl2
; NEON-NOSVE-NEXT: sshr v1.2s, v1.2s, #16
; NEON-NOSVE-NEXT: sshr v0.2s, v0.2s, #16
; NEON-NOSVE-NEXT: sdiv z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: // kill: def $d0 killed $d0 killed $z0
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: sdiv_v2i16:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #32
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #8]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #20]
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #12]
; NONEON-NOSVE-NEXT: ldrsh w10, [sp, #8]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #16]
; NONEON-NOSVE-NEXT: sdiv w9, w10, w9
; NONEON-NOSVE-NEXT: stp w9, w8, [sp, #24]
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT: add sp, sp, #32
; NONEON-NOSVE-NEXT: ret
%res = sdiv <2 x i16> %op1, %op2
ret <2 x i16> %res
}
define <4 x i16> @sdiv_v4i16(<4 x i16> %op1, <4 x i16> %op2) {
; CHECK-LABEL: sdiv_v4i16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT: ptrue p0.s, vl4
; CHECK-NEXT: sunpklo z1.s, z1.h
; CHECK-NEXT: sunpklo z0.s, z0.h
; CHECK-NEXT: sdiv z0.s, p0/m, z0.s, z1.s
; CHECK-NEXT: uzp1 z0.h, z0.h, z0.h
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: sdiv_v4i16:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: sshll v1.4s, v1.4h, #0
; NEON-NOSVE-NEXT: sshll v0.4s, v0.4h, #0
; NEON-NOSVE-NEXT: ptrue p0.s, vl4
; NEON-NOSVE-NEXT: sdiv z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: xtn v0.4h, v0.4s
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: sdiv_v4i16:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #32
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #8]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #22]
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #14]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #12]
; NONEON-NOSVE-NEXT: strh w8, [sp, #30]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #20]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #10]
; NONEON-NOSVE-NEXT: strh w8, [sp, #28]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #18]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #8]
; NONEON-NOSVE-NEXT: strh w8, [sp, #26]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #16]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: strh w8, [sp, #24]
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT: add sp, sp, #32
; NONEON-NOSVE-NEXT: ret
%res = sdiv <4 x i16> %op1, %op2
ret <4 x i16> %res
}
define <8 x i16> @sdiv_v8i16(<8 x i16> %op1, <8 x i16> %op2) {
; CHECK-LABEL: sdiv_v8i16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
; CHECK-NEXT: sunpklo z2.s, z1.h
; CHECK-NEXT: sunpklo z3.s, z0.h
; CHECK-NEXT: ptrue p0.s, vl4
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
; CHECK-NEXT: sunpklo z1.s, z1.h
; CHECK-NEXT: sunpklo z0.s, z0.h
; CHECK-NEXT: sdivr z2.s, p0/m, z2.s, z3.s
; CHECK-NEXT: sdivr z1.s, p0/m, z1.s, z0.s
; CHECK-NEXT: ptrue p0.h, vl4
; CHECK-NEXT: uzp1 z0.h, z2.h, z2.h
; CHECK-NEXT: uzp1 z1.h, z1.h, z1.h
; CHECK-NEXT: splice z0.h, p0, z0.h, z1.h
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: sdiv_v8i16:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: sshll2 v2.4s, v1.8h, #0
; NEON-NOSVE-NEXT: sshll2 v3.4s, v0.8h, #0
; NEON-NOSVE-NEXT: sshll v1.4s, v1.4h, #0
; NEON-NOSVE-NEXT: sshll v0.4s, v0.4h, #0
; NEON-NOSVE-NEXT: ptrue p0.s, vl4
; NEON-NOSVE-NEXT: sdivr z2.s, p0/m, z2.s, z3.s
; NEON-NOSVE-NEXT: sdiv z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: uzp1 v0.8h, v0.8h, v2.8h
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: sdiv_v8i16:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: stp q0, q1, [sp, #-48]!
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 48
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #30]
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #14]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #12]
; NONEON-NOSVE-NEXT: strh w8, [sp, #46]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #28]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #10]
; NONEON-NOSVE-NEXT: strh w8, [sp, #44]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #26]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #8]
; NONEON-NOSVE-NEXT: strh w8, [sp, #42]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #24]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #6]
; NONEON-NOSVE-NEXT: strh w8, [sp, #40]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #22]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #4]
; NONEON-NOSVE-NEXT: strh w8, [sp, #38]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #20]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #2]
; NONEON-NOSVE-NEXT: strh w8, [sp, #36]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #18]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp]
; NONEON-NOSVE-NEXT: strh w8, [sp, #34]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #16]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: strh w8, [sp, #32]
; NONEON-NOSVE-NEXT: ldr q0, [sp, #32]
; NONEON-NOSVE-NEXT: add sp, sp, #48
; NONEON-NOSVE-NEXT: ret
%res = sdiv <8 x i16> %op1, %op2
ret <8 x i16> %res
}
define void @sdiv_v16i16(ptr %a, ptr %b) {
; CHECK-LABEL: sdiv_v16i16:
; CHECK: // %bb.0:
; CHECK-NEXT: ldp q4, q1, [x1]
; CHECK-NEXT: ptrue p0.s, vl4
; CHECK-NEXT: ldr q0, [x0, #16]
; CHECK-NEXT: sunpklo z2.s, z1.h
; CHECK-NEXT: sunpklo z3.s, z0.h
; CHECK-NEXT: sunpklo z5.s, z4.h
; CHECK-NEXT: ext z4.b, z4.b, z4.b, #8
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
; CHECK-NEXT: sdivr z2.s, p0/m, z2.s, z3.s
; CHECK-NEXT: ldr q3, [x0]
; CHECK-NEXT: sunpklo z4.s, z4.h
; CHECK-NEXT: sunpklo z1.s, z1.h
; CHECK-NEXT: sunpklo z0.s, z0.h
; CHECK-NEXT: sunpklo z6.s, z3.h
; CHECK-NEXT: ext z3.b, z3.b, z3.b, #8
; CHECK-NEXT: sunpklo z3.s, z3.h
; CHECK-NEXT: sdivr z5.s, p0/m, z5.s, z6.s
; CHECK-NEXT: uzp1 z2.h, z2.h, z2.h
; CHECK-NEXT: sdiv z3.s, p0/m, z3.s, z4.s
; CHECK-NEXT: sdiv z0.s, p0/m, z0.s, z1.s
; CHECK-NEXT: uzp1 z1.h, z5.h, z5.h
; CHECK-NEXT: ptrue p0.h, vl4
; CHECK-NEXT: uzp1 z3.h, z3.h, z3.h
; CHECK-NEXT: splice z1.h, p0, z1.h, z3.h
; CHECK-NEXT: uzp1 z0.h, z0.h, z0.h
; CHECK-NEXT: splice z2.h, p0, z2.h, z0.h
; CHECK-NEXT: stp q1, q2, [x0]
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: sdiv_v16i16:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ldp q4, q1, [x1]
; NEON-NOSVE-NEXT: ptrue p0.s, vl4
; NEON-NOSVE-NEXT: ldr q0, [x0, #16]
; NEON-NOSVE-NEXT: sshll2 v2.4s, v1.8h, #0
; NEON-NOSVE-NEXT: sshll2 v3.4s, v0.8h, #0
; NEON-NOSVE-NEXT: sshll2 v5.4s, v4.8h, #0
; NEON-NOSVE-NEXT: sshll v4.4s, v4.4h, #0
; NEON-NOSVE-NEXT: sshll v1.4s, v1.4h, #0
; NEON-NOSVE-NEXT: sshll v0.4s, v0.4h, #0
; NEON-NOSVE-NEXT: sdivr z2.s, p0/m, z2.s, z3.s
; NEON-NOSVE-NEXT: ldr q3, [x0]
; NEON-NOSVE-NEXT: sshll2 v6.4s, v3.8h, #0
; NEON-NOSVE-NEXT: sshll v3.4s, v3.4h, #0
; NEON-NOSVE-NEXT: sdivr z5.s, p0/m, z5.s, z6.s
; NEON-NOSVE-NEXT: sdiv z3.s, p0/m, z3.s, z4.s
; NEON-NOSVE-NEXT: sdiv z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: uzp1 v1.8h, v3.8h, v5.8h
; NEON-NOSVE-NEXT: uzp1 v0.8h, v0.8h, v2.8h
; NEON-NOSVE-NEXT: stp q1, q0, [x0]
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: sdiv_v16i16:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #96
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 96
; NONEON-NOSVE-NEXT: ldp q3, q0, [x1]
; NONEON-NOSVE-NEXT: ldp q2, q1, [x0]
; NONEON-NOSVE-NEXT: stp q2, q3, [sp]
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #32]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #62]
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #46]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #44]
; NONEON-NOSVE-NEXT: strh w8, [sp, #94]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #60]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #42]
; NONEON-NOSVE-NEXT: strh w8, [sp, #92]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #58]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #40]
; NONEON-NOSVE-NEXT: strh w8, [sp, #90]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #56]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #38]
; NONEON-NOSVE-NEXT: strh w8, [sp, #88]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #54]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #36]
; NONEON-NOSVE-NEXT: strh w8, [sp, #86]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #52]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #34]
; NONEON-NOSVE-NEXT: strh w8, [sp, #84]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #50]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #32]
; NONEON-NOSVE-NEXT: strh w8, [sp, #82]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #48]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #14]
; NONEON-NOSVE-NEXT: strh w8, [sp, #80]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #30]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #12]
; NONEON-NOSVE-NEXT: strh w8, [sp, #78]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #28]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #10]
; NONEON-NOSVE-NEXT: strh w8, [sp, #76]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #26]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #8]
; NONEON-NOSVE-NEXT: strh w8, [sp, #74]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #24]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #6]
; NONEON-NOSVE-NEXT: strh w8, [sp, #72]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #22]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #4]
; NONEON-NOSVE-NEXT: strh w8, [sp, #70]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #20]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp, #2]
; NONEON-NOSVE-NEXT: strh w8, [sp, #68]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #18]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrsh w9, [sp]
; NONEON-NOSVE-NEXT: strh w8, [sp, #66]
; NONEON-NOSVE-NEXT: ldrsh w8, [sp, #16]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: strh w8, [sp, #64]
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #64]
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
; NONEON-NOSVE-NEXT: add sp, sp, #96
; NONEON-NOSVE-NEXT: ret
%op1 = load <16 x i16>, ptr %a
%op2 = load <16 x i16>, ptr %b
%res = sdiv <16 x i16> %op1, %op2
store <16 x i16> %res, ptr %a
ret void
}
define <2 x i32> @sdiv_v2i32(<2 x i32> %op1, <2 x i32> %op2) {
; CHECK-LABEL: sdiv_v2i32:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.s, vl2
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
; CHECK-NEXT: sdiv z0.s, p0/m, z0.s, z1.s
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: sdiv_v2i32:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ptrue p0.s, vl2
; NEON-NOSVE-NEXT: // kill: def $d0 killed $d0 def $z0
; NEON-NOSVE-NEXT: // kill: def $d1 killed $d1 def $z1
; NEON-NOSVE-NEXT: sdiv z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: // kill: def $d0 killed $d0 killed $z0
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: sdiv_v2i32:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #32
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #8]
; NONEON-NOSVE-NEXT: ldp w9, w10, [sp, #8]
; NONEON-NOSVE-NEXT: ldr w8, [sp, #20]
; NONEON-NOSVE-NEXT: sdiv w11, w10, w8
; NONEON-NOSVE-NEXT: ldr w8, [sp, #16]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: stp w8, w11, [sp, #24]
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT: add sp, sp, #32
; NONEON-NOSVE-NEXT: ret
%res = sdiv <2 x i32> %op1, %op2
ret <2 x i32> %res
}
define <4 x i32> @sdiv_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
; CHECK-LABEL: sdiv_v4i32:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.s, vl4
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
; CHECK-NEXT: sdiv z0.s, p0/m, z0.s, z1.s
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: sdiv_v4i32:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ptrue p0.s, vl4
; NEON-NOSVE-NEXT: // kill: def $q0 killed $q0 def $z0
; NEON-NOSVE-NEXT: // kill: def $q1 killed $q1 def $z1
; NEON-NOSVE-NEXT: sdiv z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: // kill: def $q0 killed $q0 killed $z0
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: sdiv_v4i32:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: stp q0, q1, [sp, #-48]!
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 48
; NONEON-NOSVE-NEXT: ldp w9, w10, [sp, #8]
; NONEON-NOSVE-NEXT: ldr w8, [sp, #28]
; NONEON-NOSVE-NEXT: sdiv w11, w10, w8
; NONEON-NOSVE-NEXT: ldr w8, [sp, #24]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldp w9, w10, [sp]
; NONEON-NOSVE-NEXT: stp w8, w11, [sp, #40]
; NONEON-NOSVE-NEXT: ldr w8, [sp, #20]
; NONEON-NOSVE-NEXT: sdiv w11, w10, w8
; NONEON-NOSVE-NEXT: ldr w8, [sp, #16]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: stp w8, w11, [sp, #32]
; NONEON-NOSVE-NEXT: ldr q0, [sp, #32]
; NONEON-NOSVE-NEXT: add sp, sp, #48
; NONEON-NOSVE-NEXT: ret
%res = sdiv <4 x i32> %op1, %op2
ret <4 x i32> %res
}
define void @sdiv_v8i32(ptr %a, ptr %b) {
; CHECK-LABEL: sdiv_v8i32:
; CHECK: // %bb.0:
; CHECK-NEXT: ldp q0, q3, [x1]
; CHECK-NEXT: ptrue p0.s, vl4
; CHECK-NEXT: ldp q1, q2, [x0]
; CHECK-NEXT: sdivr z0.s, p0/m, z0.s, z1.s
; CHECK-NEXT: movprfx z1, z2
; CHECK-NEXT: sdiv z1.s, p0/m, z1.s, z3.s
; CHECK-NEXT: stp q0, q1, [x0]
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: sdiv_v8i32:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ldp q0, q3, [x1]
; NEON-NOSVE-NEXT: ptrue p0.s, vl4
; NEON-NOSVE-NEXT: ldp q1, q2, [x0]
; NEON-NOSVE-NEXT: sdivr z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: movprfx z1, z2
; NEON-NOSVE-NEXT: sdiv z1.s, p0/m, z1.s, z3.s
; NEON-NOSVE-NEXT: stp q0, q1, [x0]
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: sdiv_v8i32:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #96
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 96
; NONEON-NOSVE-NEXT: ldp q3, q0, [x1]
; NONEON-NOSVE-NEXT: ldp q2, q1, [x0]
; NONEON-NOSVE-NEXT: stp q2, q3, [sp]
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #32]
; NONEON-NOSVE-NEXT: ldp w9, w10, [sp, #40]
; NONEON-NOSVE-NEXT: ldr w8, [sp, #60]
; NONEON-NOSVE-NEXT: sdiv w11, w10, w8
; NONEON-NOSVE-NEXT: ldr w8, [sp, #56]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldp w9, w10, [sp, #32]
; NONEON-NOSVE-NEXT: stp w8, w11, [sp, #88]
; NONEON-NOSVE-NEXT: ldr w8, [sp, #52]
; NONEON-NOSVE-NEXT: sdiv w11, w10, w8
; NONEON-NOSVE-NEXT: ldr w8, [sp, #48]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldp w9, w10, [sp, #8]
; NONEON-NOSVE-NEXT: stp w8, w11, [sp, #80]
; NONEON-NOSVE-NEXT: ldr w8, [sp, #28]
; NONEON-NOSVE-NEXT: sdiv w11, w10, w8
; NONEON-NOSVE-NEXT: ldr w8, [sp, #24]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldp w9, w10, [sp]
; NONEON-NOSVE-NEXT: stp w8, w11, [sp, #72]
; NONEON-NOSVE-NEXT: ldr w8, [sp, #20]
; NONEON-NOSVE-NEXT: sdiv w11, w10, w8
; NONEON-NOSVE-NEXT: ldr w8, [sp, #16]
; NONEON-NOSVE-NEXT: sdiv w8, w9, w8
; NONEON-NOSVE-NEXT: stp w8, w11, [sp, #64]
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #64]
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
; NONEON-NOSVE-NEXT: add sp, sp, #96
; NONEON-NOSVE-NEXT: ret
%op1 = load <8 x i32>, ptr %a
%op2 = load <8 x i32>, ptr %b
%res = sdiv <8 x i32> %op1, %op2
store <8 x i32> %res, ptr %a
ret void
}
define <1 x i64> @sdiv_v1i64(<1 x i64> %op1, <1 x i64> %op2) {
; CHECK-LABEL: sdiv_v1i64:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.d, vl1
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
; CHECK-NEXT: sdiv z0.d, p0/m, z0.d, z1.d
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: sdiv_v1i64:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ptrue p0.d, vl1
; NEON-NOSVE-NEXT: // kill: def $d0 killed $d0 def $z0
; NEON-NOSVE-NEXT: // kill: def $d1 killed $d1 def $z1
; NEON-NOSVE-NEXT: sdiv z0.d, p0/m, z0.d, z1.d
; NEON-NOSVE-NEXT: // kill: def $d0 killed $d0 killed $z0
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: sdiv_v1i64:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #16
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
; NONEON-NOSVE-NEXT: fmov x8, d1
; NONEON-NOSVE-NEXT: fmov x9, d0
; NONEON-NOSVE-NEXT: sdiv x8, x9, x8
; NONEON-NOSVE-NEXT: str x8, [sp, #8]
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
; NONEON-NOSVE-NEXT: add sp, sp, #16
; NONEON-NOSVE-NEXT: ret
%res = sdiv <1 x i64> %op1, %op2
ret <1 x i64> %res
}
define <2 x i64> @sdiv_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
; CHECK-LABEL: sdiv_v2i64:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.d, vl2
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
; CHECK-NEXT: sdiv z0.d, p0/m, z0.d, z1.d
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: sdiv_v2i64:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ptrue p0.d, vl2
; NEON-NOSVE-NEXT: // kill: def $q0 killed $q0 def $z0
; NEON-NOSVE-NEXT: // kill: def $q1 killed $q1 def $z1
; NEON-NOSVE-NEXT: sdiv z0.d, p0/m, z0.d, z1.d
; NEON-NOSVE-NEXT: // kill: def $q0 killed $q0 killed $z0
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: sdiv_v2i64:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: stp q0, q1, [sp, #-48]!
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 48
; NONEON-NOSVE-NEXT: ldp x9, x10, [sp]
; NONEON-NOSVE-NEXT: ldr x8, [sp, #24]
; NONEON-NOSVE-NEXT: sdiv x11, x10, x8
; NONEON-NOSVE-NEXT: ldr x8, [sp, #16]
; NONEON-NOSVE-NEXT: sdiv x8, x9, x8
; NONEON-NOSVE-NEXT: stp x8, x11, [sp, #32]
; NONEON-NOSVE-NEXT: ldr q0, [sp, #32]
; NONEON-NOSVE-NEXT: add sp, sp, #48
; NONEON-NOSVE-NEXT: ret
%res = sdiv <2 x i64> %op1, %op2
ret <2 x i64> %res
}
define void @sdiv_v4i64(ptr %a, ptr %b) {
; CHECK-LABEL: sdiv_v4i64:
; CHECK: // %bb.0:
; CHECK-NEXT: ldp q0, q3, [x1]
; CHECK-NEXT: ptrue p0.d, vl2
; CHECK-NEXT: ldp q1, q2, [x0]
; CHECK-NEXT: sdivr z0.d, p0/m, z0.d, z1.d
; CHECK-NEXT: movprfx z1, z2
; CHECK-NEXT: sdiv z1.d, p0/m, z1.d, z3.d
; CHECK-NEXT: stp q0, q1, [x0]
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: sdiv_v4i64:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ldp q0, q3, [x1]
; NEON-NOSVE-NEXT: ptrue p0.d, vl2
; NEON-NOSVE-NEXT: ldp q1, q2, [x0]
; NEON-NOSVE-NEXT: sdivr z0.d, p0/m, z0.d, z1.d
; NEON-NOSVE-NEXT: movprfx z1, z2
; NEON-NOSVE-NEXT: sdiv z1.d, p0/m, z1.d, z3.d
; NEON-NOSVE-NEXT: stp q0, q1, [x0]
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: sdiv_v4i64:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #96
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 96
; NONEON-NOSVE-NEXT: ldp q3, q0, [x1]
; NONEON-NOSVE-NEXT: ldp q2, q1, [x0]
; NONEON-NOSVE-NEXT: stp q2, q3, [sp]
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #32]
; NONEON-NOSVE-NEXT: ldp x9, x10, [sp, #32]
; NONEON-NOSVE-NEXT: ldr x8, [sp, #56]
; NONEON-NOSVE-NEXT: sdiv x11, x10, x8
; NONEON-NOSVE-NEXT: ldr x8, [sp, #48]
; NONEON-NOSVE-NEXT: sdiv x8, x9, x8
; NONEON-NOSVE-NEXT: ldp x9, x10, [sp]
; NONEON-NOSVE-NEXT: stp x8, x11, [sp, #80]
; NONEON-NOSVE-NEXT: ldr x8, [sp, #24]
; NONEON-NOSVE-NEXT: sdiv x11, x10, x8
; NONEON-NOSVE-NEXT: ldr x8, [sp, #16]
; NONEON-NOSVE-NEXT: sdiv x8, x9, x8
; NONEON-NOSVE-NEXT: stp x8, x11, [sp, #64]
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #64]
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
; NONEON-NOSVE-NEXT: add sp, sp, #96
; NONEON-NOSVE-NEXT: ret
%op1 = load <4 x i64>, ptr %a
%op2 = load <4 x i64>, ptr %b
%res = sdiv <4 x i64> %op1, %op2
store <4 x i64> %res, ptr %a
ret void
}
;
; UDIV
;
define <4 x i8> @udiv_v4i8(<4 x i8> %op1, <4 x i8> %op2) {
; CHECK-LABEL: udiv_v4i8:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT: ptrue p0.s, vl4
; CHECK-NEXT: and z0.h, z0.h, #0xff
; CHECK-NEXT: and z1.h, z1.h, #0xff
; CHECK-NEXT: uunpklo z1.s, z1.h
; CHECK-NEXT: uunpklo z0.s, z0.h
; CHECK-NEXT: udiv z0.s, p0/m, z0.s, z1.s
; CHECK-NEXT: uzp1 z0.h, z0.h, z0.h
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: udiv_v4i8:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: bic v0.4h, #255, lsl #8
; NEON-NOSVE-NEXT: bic v1.4h, #255, lsl #8
; NEON-NOSVE-NEXT: ptrue p0.s, vl4
; NEON-NOSVE-NEXT: ushll v1.4s, v1.4h, #0
; NEON-NOSVE-NEXT: ushll v0.4s, v0.4h, #0
; NEON-NOSVE-NEXT: udiv z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: xtn v0.4h, v0.4s
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: udiv_v4i8:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #32
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #8]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #22]
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #14]
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #12]
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #10]
; NONEON-NOSVE-NEXT: ldrb w12, [sp, #8]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #20]
; NONEON-NOSVE-NEXT: udiv w9, w10, w9
; NONEON-NOSVE-NEXT: ldrb w10, [sp, #18]
; NONEON-NOSVE-NEXT: strh w8, [sp, #30]
; NONEON-NOSVE-NEXT: udiv w10, w11, w10
; NONEON-NOSVE-NEXT: ldrb w11, [sp, #16]
; NONEON-NOSVE-NEXT: strh w9, [sp, #28]
; NONEON-NOSVE-NEXT: udiv w11, w12, w11
; NONEON-NOSVE-NEXT: strh w10, [sp, #26]
; NONEON-NOSVE-NEXT: strh w11, [sp, #24]
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT: add sp, sp, #32
; NONEON-NOSVE-NEXT: ret
%res = udiv <4 x i8> %op1, %op2
ret <4 x i8> %res
}
define <8 x i8> @udiv_v8i8(<8 x i8> %op1, <8 x i8> %op2) {
; CHECK-LABEL: udiv_v8i8:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT: ptrue p0.s, vl4
; CHECK-NEXT: uunpklo z1.h, z1.b
; CHECK-NEXT: uunpklo z0.h, z0.b
; CHECK-NEXT: uunpklo z2.s, z1.h
; CHECK-NEXT: uunpklo z3.s, z0.h
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
; CHECK-NEXT: uunpklo z1.s, z1.h
; CHECK-NEXT: uunpklo z0.s, z0.h
; CHECK-NEXT: udivr z2.s, p0/m, z2.s, z3.s
; CHECK-NEXT: udiv z0.s, p0/m, z0.s, z1.s
; CHECK-NEXT: ptrue p0.h, vl4
; CHECK-NEXT: uzp1 z1.h, z2.h, z2.h
; CHECK-NEXT: uzp1 z0.h, z0.h, z0.h
; CHECK-NEXT: splice z1.h, p0, z1.h, z0.h
; CHECK-NEXT: uzp1 z0.b, z1.b, z1.b
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: udiv_v8i8:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ushll v1.8h, v1.8b, #0
; NEON-NOSVE-NEXT: ushll v0.8h, v0.8b, #0
; NEON-NOSVE-NEXT: ptrue p0.s, vl4
; NEON-NOSVE-NEXT: ushll2 v2.4s, v1.8h, #0
; NEON-NOSVE-NEXT: ushll2 v3.4s, v0.8h, #0
; NEON-NOSVE-NEXT: ushll v1.4s, v1.4h, #0
; NEON-NOSVE-NEXT: ushll v0.4s, v0.4h, #0
; NEON-NOSVE-NEXT: udivr z2.s, p0/m, z2.s, z3.s
; NEON-NOSVE-NEXT: udiv z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: uzp1 v0.8h, v0.8h, v2.8h
; NEON-NOSVE-NEXT: xtn v0.8b, v0.8h
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: udiv_v8i8:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #32
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #8]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #23]
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #15]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #14]
; NONEON-NOSVE-NEXT: strb w8, [sp, #31]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #22]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #13]
; NONEON-NOSVE-NEXT: strb w8, [sp, #30]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #21]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #12]
; NONEON-NOSVE-NEXT: strb w8, [sp, #29]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #20]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #11]
; NONEON-NOSVE-NEXT: strb w8, [sp, #28]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #19]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #10]
; NONEON-NOSVE-NEXT: strb w8, [sp, #27]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #18]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #9]
; NONEON-NOSVE-NEXT: strb w8, [sp, #26]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #17]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #8]
; NONEON-NOSVE-NEXT: strb w8, [sp, #25]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #16]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: strb w8, [sp, #24]
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT: add sp, sp, #32
; NONEON-NOSVE-NEXT: ret
%res = udiv <8 x i8> %op1, %op2
ret <8 x i8> %res
}
define <16 x i8> @udiv_v16i8(<16 x i8> %op1, <16 x i8> %op2) {
; CHECK-LABEL: udiv_v16i8:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
; CHECK-NEXT: uunpklo z2.h, z1.b
; CHECK-NEXT: uunpklo z3.h, z0.b
; CHECK-NEXT: ptrue p0.s, vl4
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
; CHECK-NEXT: uunpklo z4.s, z2.h
; CHECK-NEXT: uunpklo z5.s, z3.h
; CHECK-NEXT: ext z2.b, z2.b, z2.b, #8
; CHECK-NEXT: ext z3.b, z3.b, z3.b, #8
; CHECK-NEXT: uunpklo z1.h, z1.b
; CHECK-NEXT: uunpklo z0.h, z0.b
; CHECK-NEXT: uunpklo z2.s, z2.h
; CHECK-NEXT: uunpklo z3.s, z3.h
; CHECK-NEXT: udivr z4.s, p0/m, z4.s, z5.s
; CHECK-NEXT: uunpklo z5.s, z0.h
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
; CHECK-NEXT: uunpklo z0.s, z0.h
; CHECK-NEXT: udivr z2.s, p0/m, z2.s, z3.s
; CHECK-NEXT: uunpklo z3.s, z1.h
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
; CHECK-NEXT: uunpklo z1.s, z1.h
; CHECK-NEXT: udivr z3.s, p0/m, z3.s, z5.s
; CHECK-NEXT: uzp1 z2.h, z2.h, z2.h
; CHECK-NEXT: udiv z0.s, p0/m, z0.s, z1.s
; CHECK-NEXT: uzp1 z1.h, z4.h, z4.h
; CHECK-NEXT: ptrue p0.h, vl4
; CHECK-NEXT: uzp1 z3.h, z3.h, z3.h
; CHECK-NEXT: splice z1.h, p0, z1.h, z2.h
; CHECK-NEXT: uzp1 z0.h, z0.h, z0.h
; CHECK-NEXT: splice z3.h, p0, z3.h, z0.h
; CHECK-NEXT: uzp1 z0.b, z1.b, z1.b
; CHECK-NEXT: ptrue p0.b, vl8
; CHECK-NEXT: uzp1 z1.b, z3.b, z3.b
; CHECK-NEXT: splice z0.b, p0, z0.b, z1.b
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: udiv_v16i8:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ushll2 v2.8h, v1.16b, #0
; NEON-NOSVE-NEXT: ushll2 v3.8h, v0.16b, #0
; NEON-NOSVE-NEXT: ushll v1.8h, v1.8b, #0
; NEON-NOSVE-NEXT: ushll v0.8h, v0.8b, #0
; NEON-NOSVE-NEXT: ptrue p0.s, vl4
; NEON-NOSVE-NEXT: ushll2 v4.4s, v2.8h, #0
; NEON-NOSVE-NEXT: ushll2 v5.4s, v3.8h, #0
; NEON-NOSVE-NEXT: ushll v2.4s, v2.4h, #0
; NEON-NOSVE-NEXT: ushll v3.4s, v3.4h, #0
; NEON-NOSVE-NEXT: udivr z4.s, p0/m, z4.s, z5.s
; NEON-NOSVE-NEXT: ushll2 v5.4s, v0.8h, #0
; NEON-NOSVE-NEXT: ushll v0.4s, v0.4h, #0
; NEON-NOSVE-NEXT: udivr z2.s, p0/m, z2.s, z3.s
; NEON-NOSVE-NEXT: ushll2 v3.4s, v1.8h, #0
; NEON-NOSVE-NEXT: ushll v1.4s, v1.4h, #0
; NEON-NOSVE-NEXT: udivr z3.s, p0/m, z3.s, z5.s
; NEON-NOSVE-NEXT: udiv z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: uzp1 v1.8h, v2.8h, v4.8h
; NEON-NOSVE-NEXT: uzp1 v0.8h, v0.8h, v3.8h
; NEON-NOSVE-NEXT: uzp1 v0.16b, v0.16b, v1.16b
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: udiv_v16i8:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: stp q0, q1, [sp, #-48]!
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 48
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #31]
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #15]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #14]
; NONEON-NOSVE-NEXT: strb w8, [sp, #47]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #30]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #13]
; NONEON-NOSVE-NEXT: strb w8, [sp, #46]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #29]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #12]
; NONEON-NOSVE-NEXT: strb w8, [sp, #45]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #28]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #11]
; NONEON-NOSVE-NEXT: strb w8, [sp, #44]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #27]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #10]
; NONEON-NOSVE-NEXT: strb w8, [sp, #43]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #26]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #9]
; NONEON-NOSVE-NEXT: strb w8, [sp, #42]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #25]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #8]
; NONEON-NOSVE-NEXT: strb w8, [sp, #41]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #24]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #7]
; NONEON-NOSVE-NEXT: strb w8, [sp, #40]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #23]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #6]
; NONEON-NOSVE-NEXT: strb w8, [sp, #39]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #22]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #5]
; NONEON-NOSVE-NEXT: strb w8, [sp, #38]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #21]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #4]
; NONEON-NOSVE-NEXT: strb w8, [sp, #37]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #20]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #3]
; NONEON-NOSVE-NEXT: strb w8, [sp, #36]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #19]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #2]
; NONEON-NOSVE-NEXT: strb w8, [sp, #35]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #18]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #1]
; NONEON-NOSVE-NEXT: strb w8, [sp, #34]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #17]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp]
; NONEON-NOSVE-NEXT: strb w8, [sp, #33]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #16]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: strb w8, [sp, #32]
; NONEON-NOSVE-NEXT: ldr q0, [sp, #32]
; NONEON-NOSVE-NEXT: add sp, sp, #48
; NONEON-NOSVE-NEXT: ret
%res = udiv <16 x i8> %op1, %op2
ret <16 x i8> %res
}
define void @udiv_v32i8(ptr %a, ptr %b) {
; CHECK-LABEL: udiv_v32i8:
; CHECK: // %bb.0:
; CHECK-NEXT: ldp q6, q3, [x1]
; CHECK-NEXT: ptrue p0.s, vl4
; CHECK-NEXT: ldr q2, [x0, #16]
; CHECK-NEXT: uunpklo z1.h, z3.b
; CHECK-NEXT: uunpklo z4.h, z2.b
; CHECK-NEXT: ext z3.b, z3.b, z3.b, #8
; CHECK-NEXT: ext z2.b, z2.b, z2.b, #8
; CHECK-NEXT: uunpklo z7.h, z6.b
; CHECK-NEXT: ext z6.b, z6.b, z6.b, #8
; CHECK-NEXT: uunpklo z3.h, z3.b
; CHECK-NEXT: uunpklo z0.s, z1.h
; CHECK-NEXT: uunpklo z5.s, z4.h
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
; CHECK-NEXT: ext z4.b, z4.b, z4.b, #8
; CHECK-NEXT: uunpklo z17.s, z7.h
; CHECK-NEXT: ext z7.b, z7.b, z7.b, #8
; CHECK-NEXT: uunpklo z6.h, z6.b
; CHECK-NEXT: uunpklo z1.s, z1.h
; CHECK-NEXT: uunpklo z4.s, z4.h
; CHECK-NEXT: uunpklo z7.s, z7.h
; CHECK-NEXT: udivr z0.s, p0/m, z0.s, z5.s
; CHECK-NEXT: udivr z1.s, p0/m, z1.s, z4.s
; CHECK-NEXT: uunpklo z4.h, z2.b
; CHECK-NEXT: uunpklo z2.s, z3.h
; CHECK-NEXT: ext z3.b, z3.b, z3.b, #8
; CHECK-NEXT: uzp1 z0.h, z0.h, z0.h
; CHECK-NEXT: uunpklo z5.s, z4.h
; CHECK-NEXT: ext z4.b, z4.b, z4.b, #8
; CHECK-NEXT: uunpklo z3.s, z3.h
; CHECK-NEXT: uunpklo z4.s, z4.h
; CHECK-NEXT: udivr z2.s, p0/m, z2.s, z5.s
; CHECK-NEXT: ldr q5, [x0]
; CHECK-NEXT: uunpklo z16.h, z5.b
; CHECK-NEXT: ext z5.b, z5.b, z5.b, #8
; CHECK-NEXT: uzp1 z1.h, z1.h, z1.h
; CHECK-NEXT: uunpklo z5.h, z5.b
; CHECK-NEXT: uunpklo z18.s, z16.h
; CHECK-NEXT: ext z16.b, z16.b, z16.b, #8
; CHECK-NEXT: uunpklo z16.s, z16.h
; CHECK-NEXT: udivr z17.s, p0/m, z17.s, z18.s
; CHECK-NEXT: uunpklo z18.s, z5.h
; CHECK-NEXT: ext z5.b, z5.b, z5.b, #8
; CHECK-NEXT: uunpklo z5.s, z5.h
; CHECK-NEXT: uzp1 z2.h, z2.h, z2.h
; CHECK-NEXT: udivr z7.s, p0/m, z7.s, z16.s
; CHECK-NEXT: uunpklo z16.s, z6.h
; CHECK-NEXT: ext z6.b, z6.b, z6.b, #8
; CHECK-NEXT: uunpklo z6.s, z6.h
; CHECK-NEXT: udivr z16.s, p0/m, z16.s, z18.s
; CHECK-NEXT: udiv z5.s, p0/m, z5.s, z6.s
; CHECK-NEXT: uzp1 z6.h, z7.h, z7.h
; CHECK-NEXT: uzp1 z7.h, z16.h, z16.h
; CHECK-NEXT: udivr z3.s, p0/m, z3.s, z4.s
; CHECK-NEXT: uzp1 z4.h, z17.h, z17.h
; CHECK-NEXT: ptrue p0.h, vl4
; CHECK-NEXT: splice z0.h, p0, z0.h, z1.h
; CHECK-NEXT: uzp1 z5.h, z5.h, z5.h
; CHECK-NEXT: splice z4.h, p0, z4.h, z6.h
; CHECK-NEXT: splice z7.h, p0, z7.h, z5.h
; CHECK-NEXT: uzp1 z0.b, z0.b, z0.b
; CHECK-NEXT: uzp1 z1.b, z4.b, z4.b
; CHECK-NEXT: uzp1 z3.h, z3.h, z3.h
; CHECK-NEXT: splice z2.h, p0, z2.h, z3.h
; CHECK-NEXT: uzp1 z3.b, z7.b, z7.b
; CHECK-NEXT: ptrue p0.b, vl8
; CHECK-NEXT: splice z1.b, p0, z1.b, z3.b
; CHECK-NEXT: uzp1 z2.b, z2.b, z2.b
; CHECK-NEXT: splice z0.b, p0, z0.b, z2.b
; CHECK-NEXT: stp q1, q0, [x0]
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: udiv_v32i8:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ldp q6, q3, [x1]
; NEON-NOSVE-NEXT: ptrue p0.s, vl4
; NEON-NOSVE-NEXT: ldr q2, [x0, #16]
; NEON-NOSVE-NEXT: ushll2 v1.8h, v3.16b, #0
; NEON-NOSVE-NEXT: ushll2 v4.8h, v2.16b, #0
; NEON-NOSVE-NEXT: ushll v3.8h, v3.8b, #0
; NEON-NOSVE-NEXT: ushll v2.8h, v2.8b, #0
; NEON-NOSVE-NEXT: ushll2 v7.8h, v6.16b, #0
; NEON-NOSVE-NEXT: ushll v6.8h, v6.8b, #0
; NEON-NOSVE-NEXT: ushll2 v0.4s, v1.8h, #0
; NEON-NOSVE-NEXT: ushll2 v5.4s, v4.8h, #0
; NEON-NOSVE-NEXT: ushll v1.4s, v1.4h, #0
; NEON-NOSVE-NEXT: ushll v4.4s, v4.4h, #0
; NEON-NOSVE-NEXT: ushll2 v17.4s, v7.8h, #0
; NEON-NOSVE-NEXT: ushll v7.4s, v7.4h, #0
; NEON-NOSVE-NEXT: udivr z0.s, p0/m, z0.s, z5.s
; NEON-NOSVE-NEXT: ushll2 v5.4s, v2.8h, #0
; NEON-NOSVE-NEXT: ushll v2.4s, v2.4h, #0
; NEON-NOSVE-NEXT: udivr z1.s, p0/m, z1.s, z4.s
; NEON-NOSVE-NEXT: ushll2 v4.4s, v3.8h, #0
; NEON-NOSVE-NEXT: ushll v3.4s, v3.4h, #0
; NEON-NOSVE-NEXT: udivr z4.s, p0/m, z4.s, z5.s
; NEON-NOSVE-NEXT: ldr q5, [x0]
; NEON-NOSVE-NEXT: ushll2 v16.8h, v5.16b, #0
; NEON-NOSVE-NEXT: ushll v5.8h, v5.8b, #0
; NEON-NOSVE-NEXT: uzp1 v0.8h, v1.8h, v0.8h
; NEON-NOSVE-NEXT: ushll2 v18.4s, v16.8h, #0
; NEON-NOSVE-NEXT: ushll v16.4s, v16.4h, #0
; NEON-NOSVE-NEXT: udivr z17.s, p0/m, z17.s, z18.s
; NEON-NOSVE-NEXT: ushll2 v18.4s, v5.8h, #0
; NEON-NOSVE-NEXT: ushll v5.4s, v5.4h, #0
; NEON-NOSVE-NEXT: udivr z7.s, p0/m, z7.s, z16.s
; NEON-NOSVE-NEXT: ushll2 v16.4s, v6.8h, #0
; NEON-NOSVE-NEXT: ushll v6.4s, v6.4h, #0
; NEON-NOSVE-NEXT: udivr z16.s, p0/m, z16.s, z18.s
; NEON-NOSVE-NEXT: udiv z5.s, p0/m, z5.s, z6.s
; NEON-NOSVE-NEXT: udiv z2.s, p0/m, z2.s, z3.s
; NEON-NOSVE-NEXT: uzp1 v3.8h, v7.8h, v17.8h
; NEON-NOSVE-NEXT: uzp1 v5.8h, v5.8h, v16.8h
; NEON-NOSVE-NEXT: uzp1 v1.8h, v2.8h, v4.8h
; NEON-NOSVE-NEXT: uzp1 v2.16b, v5.16b, v3.16b
; NEON-NOSVE-NEXT: uzp1 v0.16b, v1.16b, v0.16b
; NEON-NOSVE-NEXT: stp q2, q0, [x0]
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: udiv_v32i8:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #96
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 96
; NONEON-NOSVE-NEXT: ldp q3, q0, [x1]
; NONEON-NOSVE-NEXT: ldp q2, q1, [x0]
; NONEON-NOSVE-NEXT: stp q2, q3, [sp]
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #32]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #63]
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #47]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #46]
; NONEON-NOSVE-NEXT: strb w8, [sp, #95]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #62]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #45]
; NONEON-NOSVE-NEXT: strb w8, [sp, #94]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #61]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #44]
; NONEON-NOSVE-NEXT: strb w8, [sp, #93]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #60]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #43]
; NONEON-NOSVE-NEXT: strb w8, [sp, #92]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #59]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #42]
; NONEON-NOSVE-NEXT: strb w8, [sp, #91]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #58]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #41]
; NONEON-NOSVE-NEXT: strb w8, [sp, #90]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #57]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #40]
; NONEON-NOSVE-NEXT: strb w8, [sp, #89]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #56]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #39]
; NONEON-NOSVE-NEXT: strb w8, [sp, #88]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #55]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #38]
; NONEON-NOSVE-NEXT: strb w8, [sp, #87]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #54]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #37]
; NONEON-NOSVE-NEXT: strb w8, [sp, #86]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #53]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #36]
; NONEON-NOSVE-NEXT: strb w8, [sp, #85]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #52]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #35]
; NONEON-NOSVE-NEXT: strb w8, [sp, #84]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #51]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #34]
; NONEON-NOSVE-NEXT: strb w8, [sp, #83]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #50]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #33]
; NONEON-NOSVE-NEXT: strb w8, [sp, #82]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #49]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #32]
; NONEON-NOSVE-NEXT: strb w8, [sp, #81]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #48]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #15]
; NONEON-NOSVE-NEXT: strb w8, [sp, #80]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #31]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #14]
; NONEON-NOSVE-NEXT: strb w8, [sp, #79]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #30]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #13]
; NONEON-NOSVE-NEXT: strb w8, [sp, #78]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #29]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #12]
; NONEON-NOSVE-NEXT: strb w8, [sp, #77]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #28]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #11]
; NONEON-NOSVE-NEXT: strb w8, [sp, #76]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #27]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #10]
; NONEON-NOSVE-NEXT: strb w8, [sp, #75]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #26]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #9]
; NONEON-NOSVE-NEXT: strb w8, [sp, #74]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #25]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #8]
; NONEON-NOSVE-NEXT: strb w8, [sp, #73]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #24]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #7]
; NONEON-NOSVE-NEXT: strb w8, [sp, #72]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #23]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #6]
; NONEON-NOSVE-NEXT: strb w8, [sp, #71]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #22]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #5]
; NONEON-NOSVE-NEXT: strb w8, [sp, #70]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #21]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #4]
; NONEON-NOSVE-NEXT: strb w8, [sp, #69]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #20]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #3]
; NONEON-NOSVE-NEXT: strb w8, [sp, #68]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #19]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #2]
; NONEON-NOSVE-NEXT: strb w8, [sp, #67]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #18]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp, #1]
; NONEON-NOSVE-NEXT: strb w8, [sp, #66]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #17]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrb w9, [sp]
; NONEON-NOSVE-NEXT: strb w8, [sp, #65]
; NONEON-NOSVE-NEXT: ldrb w8, [sp, #16]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: strb w8, [sp, #64]
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #64]
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
; NONEON-NOSVE-NEXT: add sp, sp, #96
; NONEON-NOSVE-NEXT: ret
%op1 = load <32 x i8>, ptr %a
%op2 = load <32 x i8>, ptr %b
%res = udiv <32 x i8> %op1, %op2
store <32 x i8> %res, ptr %a
ret void
}
define <2 x i16> @udiv_v2i16(<2 x i16> %op1, <2 x i16> %op2) {
; CHECK-LABEL: udiv_v2i16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT: ptrue p0.s, vl2
; CHECK-NEXT: and z1.s, z1.s, #0xffff
; CHECK-NEXT: and z0.s, z0.s, #0xffff
; CHECK-NEXT: udiv z0.s, p0/m, z0.s, z1.s
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: udiv_v2i16:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: movi d2, #0x00ffff0000ffff
; NEON-NOSVE-NEXT: ptrue p0.s, vl2
; NEON-NOSVE-NEXT: and v1.8b, v1.8b, v2.8b
; NEON-NOSVE-NEXT: and v0.8b, v0.8b, v2.8b
; NEON-NOSVE-NEXT: udiv z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: // kill: def $d0 killed $d0 killed $z0
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: udiv_v2i16:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #32
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #8]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #20]
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #12]
; NONEON-NOSVE-NEXT: ldrh w10, [sp, #8]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #16]
; NONEON-NOSVE-NEXT: udiv w9, w10, w9
; NONEON-NOSVE-NEXT: stp w9, w8, [sp, #24]
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT: add sp, sp, #32
; NONEON-NOSVE-NEXT: ret
%res = udiv <2 x i16> %op1, %op2
ret <2 x i16> %res
}
define <4 x i16> @udiv_v4i16(<4 x i16> %op1, <4 x i16> %op2) {
; CHECK-LABEL: udiv_v4i16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT: ptrue p0.s, vl4
; CHECK-NEXT: uunpklo z1.s, z1.h
; CHECK-NEXT: uunpklo z0.s, z0.h
; CHECK-NEXT: udiv z0.s, p0/m, z0.s, z1.s
; CHECK-NEXT: uzp1 z0.h, z0.h, z0.h
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: udiv_v4i16:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ushll v1.4s, v1.4h, #0
; NEON-NOSVE-NEXT: ushll v0.4s, v0.4h, #0
; NEON-NOSVE-NEXT: ptrue p0.s, vl4
; NEON-NOSVE-NEXT: udiv z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: xtn v0.4h, v0.4s
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: udiv_v4i16:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #32
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #8]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #22]
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #14]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #12]
; NONEON-NOSVE-NEXT: strh w8, [sp, #30]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #20]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #10]
; NONEON-NOSVE-NEXT: strh w8, [sp, #28]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #18]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #8]
; NONEON-NOSVE-NEXT: strh w8, [sp, #26]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #16]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: strh w8, [sp, #24]
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT: add sp, sp, #32
; NONEON-NOSVE-NEXT: ret
%res = udiv <4 x i16> %op1, %op2
ret <4 x i16> %res
}
define <8 x i16> @udiv_v8i16(<8 x i16> %op1, <8 x i16> %op2) {
; CHECK-LABEL: udiv_v8i16:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
; CHECK-NEXT: uunpklo z2.s, z1.h
; CHECK-NEXT: uunpklo z3.s, z0.h
; CHECK-NEXT: ptrue p0.s, vl4
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
; CHECK-NEXT: uunpklo z1.s, z1.h
; CHECK-NEXT: uunpklo z0.s, z0.h
; CHECK-NEXT: udivr z2.s, p0/m, z2.s, z3.s
; CHECK-NEXT: udivr z1.s, p0/m, z1.s, z0.s
; CHECK-NEXT: ptrue p0.h, vl4
; CHECK-NEXT: uzp1 z0.h, z2.h, z2.h
; CHECK-NEXT: uzp1 z1.h, z1.h, z1.h
; CHECK-NEXT: splice z0.h, p0, z0.h, z1.h
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: udiv_v8i16:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ushll2 v2.4s, v1.8h, #0
; NEON-NOSVE-NEXT: ushll2 v3.4s, v0.8h, #0
; NEON-NOSVE-NEXT: ushll v1.4s, v1.4h, #0
; NEON-NOSVE-NEXT: ushll v0.4s, v0.4h, #0
; NEON-NOSVE-NEXT: ptrue p0.s, vl4
; NEON-NOSVE-NEXT: udivr z2.s, p0/m, z2.s, z3.s
; NEON-NOSVE-NEXT: udiv z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: uzp1 v0.8h, v0.8h, v2.8h
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: udiv_v8i16:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: stp q0, q1, [sp, #-48]!
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 48
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #30]
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #14]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #12]
; NONEON-NOSVE-NEXT: strh w8, [sp, #46]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #28]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #10]
; NONEON-NOSVE-NEXT: strh w8, [sp, #44]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #26]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #8]
; NONEON-NOSVE-NEXT: strh w8, [sp, #42]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #24]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #6]
; NONEON-NOSVE-NEXT: strh w8, [sp, #40]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #22]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #4]
; NONEON-NOSVE-NEXT: strh w8, [sp, #38]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #20]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #2]
; NONEON-NOSVE-NEXT: strh w8, [sp, #36]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #18]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp]
; NONEON-NOSVE-NEXT: strh w8, [sp, #34]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #16]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: strh w8, [sp, #32]
; NONEON-NOSVE-NEXT: ldr q0, [sp, #32]
; NONEON-NOSVE-NEXT: add sp, sp, #48
; NONEON-NOSVE-NEXT: ret
%res = udiv <8 x i16> %op1, %op2
ret <8 x i16> %res
}
define void @udiv_v16i16(ptr %a, ptr %b) {
; CHECK-LABEL: udiv_v16i16:
; CHECK: // %bb.0:
; CHECK-NEXT: ldp q4, q1, [x1]
; CHECK-NEXT: ptrue p0.s, vl4
; CHECK-NEXT: ldr q0, [x0, #16]
; CHECK-NEXT: uunpklo z2.s, z1.h
; CHECK-NEXT: uunpklo z3.s, z0.h
; CHECK-NEXT: uunpklo z5.s, z4.h
; CHECK-NEXT: ext z4.b, z4.b, z4.b, #8
; CHECK-NEXT: ext z1.b, z1.b, z1.b, #8
; CHECK-NEXT: ext z0.b, z0.b, z0.b, #8
; CHECK-NEXT: udivr z2.s, p0/m, z2.s, z3.s
; CHECK-NEXT: ldr q3, [x0]
; CHECK-NEXT: uunpklo z4.s, z4.h
; CHECK-NEXT: uunpklo z1.s, z1.h
; CHECK-NEXT: uunpklo z0.s, z0.h
; CHECK-NEXT: uunpklo z6.s, z3.h
; CHECK-NEXT: ext z3.b, z3.b, z3.b, #8
; CHECK-NEXT: uunpklo z3.s, z3.h
; CHECK-NEXT: udivr z5.s, p0/m, z5.s, z6.s
; CHECK-NEXT: uzp1 z2.h, z2.h, z2.h
; CHECK-NEXT: udiv z3.s, p0/m, z3.s, z4.s
; CHECK-NEXT: udiv z0.s, p0/m, z0.s, z1.s
; CHECK-NEXT: uzp1 z1.h, z5.h, z5.h
; CHECK-NEXT: ptrue p0.h, vl4
; CHECK-NEXT: uzp1 z3.h, z3.h, z3.h
; CHECK-NEXT: splice z1.h, p0, z1.h, z3.h
; CHECK-NEXT: uzp1 z0.h, z0.h, z0.h
; CHECK-NEXT: splice z2.h, p0, z2.h, z0.h
; CHECK-NEXT: stp q1, q2, [x0]
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: udiv_v16i16:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ldp q4, q1, [x1]
; NEON-NOSVE-NEXT: ptrue p0.s, vl4
; NEON-NOSVE-NEXT: ldr q0, [x0, #16]
; NEON-NOSVE-NEXT: ushll2 v2.4s, v1.8h, #0
; NEON-NOSVE-NEXT: ushll2 v3.4s, v0.8h, #0
; NEON-NOSVE-NEXT: ushll2 v5.4s, v4.8h, #0
; NEON-NOSVE-NEXT: ushll v4.4s, v4.4h, #0
; NEON-NOSVE-NEXT: ushll v1.4s, v1.4h, #0
; NEON-NOSVE-NEXT: ushll v0.4s, v0.4h, #0
; NEON-NOSVE-NEXT: udivr z2.s, p0/m, z2.s, z3.s
; NEON-NOSVE-NEXT: ldr q3, [x0]
; NEON-NOSVE-NEXT: ushll2 v6.4s, v3.8h, #0
; NEON-NOSVE-NEXT: ushll v3.4s, v3.4h, #0
; NEON-NOSVE-NEXT: udivr z5.s, p0/m, z5.s, z6.s
; NEON-NOSVE-NEXT: udiv z3.s, p0/m, z3.s, z4.s
; NEON-NOSVE-NEXT: udiv z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: uzp1 v1.8h, v3.8h, v5.8h
; NEON-NOSVE-NEXT: uzp1 v0.8h, v0.8h, v2.8h
; NEON-NOSVE-NEXT: stp q1, q0, [x0]
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: udiv_v16i16:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #96
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 96
; NONEON-NOSVE-NEXT: ldp q3, q0, [x1]
; NONEON-NOSVE-NEXT: ldp q2, q1, [x0]
; NONEON-NOSVE-NEXT: stp q2, q3, [sp]
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #32]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #62]
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #46]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #44]
; NONEON-NOSVE-NEXT: strh w8, [sp, #94]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #60]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #42]
; NONEON-NOSVE-NEXT: strh w8, [sp, #92]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #58]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #40]
; NONEON-NOSVE-NEXT: strh w8, [sp, #90]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #56]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #38]
; NONEON-NOSVE-NEXT: strh w8, [sp, #88]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #54]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #36]
; NONEON-NOSVE-NEXT: strh w8, [sp, #86]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #52]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #34]
; NONEON-NOSVE-NEXT: strh w8, [sp, #84]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #50]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #32]
; NONEON-NOSVE-NEXT: strh w8, [sp, #82]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #48]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #14]
; NONEON-NOSVE-NEXT: strh w8, [sp, #80]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #30]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #12]
; NONEON-NOSVE-NEXT: strh w8, [sp, #78]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #28]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #10]
; NONEON-NOSVE-NEXT: strh w8, [sp, #76]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #26]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #8]
; NONEON-NOSVE-NEXT: strh w8, [sp, #74]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #24]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #6]
; NONEON-NOSVE-NEXT: strh w8, [sp, #72]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #22]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #4]
; NONEON-NOSVE-NEXT: strh w8, [sp, #70]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #20]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp, #2]
; NONEON-NOSVE-NEXT: strh w8, [sp, #68]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #18]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldrh w9, [sp]
; NONEON-NOSVE-NEXT: strh w8, [sp, #66]
; NONEON-NOSVE-NEXT: ldrh w8, [sp, #16]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: strh w8, [sp, #64]
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #64]
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
; NONEON-NOSVE-NEXT: add sp, sp, #96
; NONEON-NOSVE-NEXT: ret
%op1 = load <16 x i16>, ptr %a
%op2 = load <16 x i16>, ptr %b
%res = udiv <16 x i16> %op1, %op2
store <16 x i16> %res, ptr %a
ret void
}
define <2 x i32> @udiv_v2i32(<2 x i32> %op1, <2 x i32> %op2) {
; CHECK-LABEL: udiv_v2i32:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.s, vl2
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
; CHECK-NEXT: udiv z0.s, p0/m, z0.s, z1.s
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: udiv_v2i32:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ptrue p0.s, vl2
; NEON-NOSVE-NEXT: // kill: def $d0 killed $d0 def $z0
; NEON-NOSVE-NEXT: // kill: def $d1 killed $d1 def $z1
; NEON-NOSVE-NEXT: udiv z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: // kill: def $d0 killed $d0 killed $z0
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: udiv_v2i32:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #32
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
; NONEON-NOSVE-NEXT: stp d0, d1, [sp, #8]
; NONEON-NOSVE-NEXT: ldp w9, w10, [sp, #8]
; NONEON-NOSVE-NEXT: ldr w8, [sp, #20]
; NONEON-NOSVE-NEXT: udiv w11, w10, w8
; NONEON-NOSVE-NEXT: ldr w8, [sp, #16]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: stp w8, w11, [sp, #24]
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT: add sp, sp, #32
; NONEON-NOSVE-NEXT: ret
%res = udiv <2 x i32> %op1, %op2
ret <2 x i32> %res
}
define <4 x i32> @udiv_v4i32(<4 x i32> %op1, <4 x i32> %op2) {
; CHECK-LABEL: udiv_v4i32:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.s, vl4
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
; CHECK-NEXT: udiv z0.s, p0/m, z0.s, z1.s
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: udiv_v4i32:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ptrue p0.s, vl4
; NEON-NOSVE-NEXT: // kill: def $q0 killed $q0 def $z0
; NEON-NOSVE-NEXT: // kill: def $q1 killed $q1 def $z1
; NEON-NOSVE-NEXT: udiv z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: // kill: def $q0 killed $q0 killed $z0
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: udiv_v4i32:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: stp q0, q1, [sp, #-48]!
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 48
; NONEON-NOSVE-NEXT: ldp w9, w10, [sp, #8]
; NONEON-NOSVE-NEXT: ldr w8, [sp, #28]
; NONEON-NOSVE-NEXT: udiv w11, w10, w8
; NONEON-NOSVE-NEXT: ldr w8, [sp, #24]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldp w9, w10, [sp]
; NONEON-NOSVE-NEXT: stp w8, w11, [sp, #40]
; NONEON-NOSVE-NEXT: ldr w8, [sp, #20]
; NONEON-NOSVE-NEXT: udiv w11, w10, w8
; NONEON-NOSVE-NEXT: ldr w8, [sp, #16]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: stp w8, w11, [sp, #32]
; NONEON-NOSVE-NEXT: ldr q0, [sp, #32]
; NONEON-NOSVE-NEXT: add sp, sp, #48
; NONEON-NOSVE-NEXT: ret
%res = udiv <4 x i32> %op1, %op2
ret <4 x i32> %res
}
define void @udiv_v8i32(ptr %a, ptr %b) {
; CHECK-LABEL: udiv_v8i32:
; CHECK: // %bb.0:
; CHECK-NEXT: ldp q0, q3, [x1]
; CHECK-NEXT: ptrue p0.s, vl4
; CHECK-NEXT: ldp q1, q2, [x0]
; CHECK-NEXT: udivr z0.s, p0/m, z0.s, z1.s
; CHECK-NEXT: movprfx z1, z2
; CHECK-NEXT: udiv z1.s, p0/m, z1.s, z3.s
; CHECK-NEXT: stp q0, q1, [x0]
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: udiv_v8i32:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ldp q0, q3, [x1]
; NEON-NOSVE-NEXT: ptrue p0.s, vl4
; NEON-NOSVE-NEXT: ldp q1, q2, [x0]
; NEON-NOSVE-NEXT: udivr z0.s, p0/m, z0.s, z1.s
; NEON-NOSVE-NEXT: movprfx z1, z2
; NEON-NOSVE-NEXT: udiv z1.s, p0/m, z1.s, z3.s
; NEON-NOSVE-NEXT: stp q0, q1, [x0]
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: udiv_v8i32:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #96
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 96
; NONEON-NOSVE-NEXT: ldp q3, q0, [x1]
; NONEON-NOSVE-NEXT: ldp q2, q1, [x0]
; NONEON-NOSVE-NEXT: stp q2, q3, [sp]
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #32]
; NONEON-NOSVE-NEXT: ldp w9, w10, [sp, #40]
; NONEON-NOSVE-NEXT: ldr w8, [sp, #60]
; NONEON-NOSVE-NEXT: udiv w11, w10, w8
; NONEON-NOSVE-NEXT: ldr w8, [sp, #56]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldp w9, w10, [sp, #32]
; NONEON-NOSVE-NEXT: stp w8, w11, [sp, #88]
; NONEON-NOSVE-NEXT: ldr w8, [sp, #52]
; NONEON-NOSVE-NEXT: udiv w11, w10, w8
; NONEON-NOSVE-NEXT: ldr w8, [sp, #48]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldp w9, w10, [sp, #8]
; NONEON-NOSVE-NEXT: stp w8, w11, [sp, #80]
; NONEON-NOSVE-NEXT: ldr w8, [sp, #28]
; NONEON-NOSVE-NEXT: udiv w11, w10, w8
; NONEON-NOSVE-NEXT: ldr w8, [sp, #24]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: ldp w9, w10, [sp]
; NONEON-NOSVE-NEXT: stp w8, w11, [sp, #72]
; NONEON-NOSVE-NEXT: ldr w8, [sp, #20]
; NONEON-NOSVE-NEXT: udiv w11, w10, w8
; NONEON-NOSVE-NEXT: ldr w8, [sp, #16]
; NONEON-NOSVE-NEXT: udiv w8, w9, w8
; NONEON-NOSVE-NEXT: stp w8, w11, [sp, #64]
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #64]
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
; NONEON-NOSVE-NEXT: add sp, sp, #96
; NONEON-NOSVE-NEXT: ret
%op1 = load <8 x i32>, ptr %a
%op2 = load <8 x i32>, ptr %b
%res = udiv <8 x i32> %op1, %op2
store <8 x i32> %res, ptr %a
ret void
}
define <1 x i64> @udiv_v1i64(<1 x i64> %op1, <1 x i64> %op2) {
; CHECK-LABEL: udiv_v1i64:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.d, vl1
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
; CHECK-NEXT: udiv z0.d, p0/m, z0.d, z1.d
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: udiv_v1i64:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ptrue p0.d, vl1
; NEON-NOSVE-NEXT: // kill: def $d0 killed $d0 def $z0
; NEON-NOSVE-NEXT: // kill: def $d1 killed $d1 def $z1
; NEON-NOSVE-NEXT: udiv z0.d, p0/m, z0.d, z1.d
; NEON-NOSVE-NEXT: // kill: def $d0 killed $d0 killed $z0
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: udiv_v1i64:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #16
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
; NONEON-NOSVE-NEXT: fmov x8, d1
; NONEON-NOSVE-NEXT: fmov x9, d0
; NONEON-NOSVE-NEXT: udiv x8, x9, x8
; NONEON-NOSVE-NEXT: str x8, [sp, #8]
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
; NONEON-NOSVE-NEXT: add sp, sp, #16
; NONEON-NOSVE-NEXT: ret
%res = udiv <1 x i64> %op1, %op2
ret <1 x i64> %res
}
define <2 x i64> @udiv_v2i64(<2 x i64> %op1, <2 x i64> %op2) {
; CHECK-LABEL: udiv_v2i64:
; CHECK: // %bb.0:
; CHECK-NEXT: ptrue p0.d, vl2
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
; CHECK-NEXT: udiv z0.d, p0/m, z0.d, z1.d
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: udiv_v2i64:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ptrue p0.d, vl2
; NEON-NOSVE-NEXT: // kill: def $q0 killed $q0 def $z0
; NEON-NOSVE-NEXT: // kill: def $q1 killed $q1 def $z1
; NEON-NOSVE-NEXT: udiv z0.d, p0/m, z0.d, z1.d
; NEON-NOSVE-NEXT: // kill: def $q0 killed $q0 killed $z0
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: udiv_v2i64:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: stp q0, q1, [sp, #-48]!
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 48
; NONEON-NOSVE-NEXT: ldp x9, x10, [sp]
; NONEON-NOSVE-NEXT: ldr x8, [sp, #24]
; NONEON-NOSVE-NEXT: udiv x11, x10, x8
; NONEON-NOSVE-NEXT: ldr x8, [sp, #16]
; NONEON-NOSVE-NEXT: udiv x8, x9, x8
; NONEON-NOSVE-NEXT: stp x8, x11, [sp, #32]
; NONEON-NOSVE-NEXT: ldr q0, [sp, #32]
; NONEON-NOSVE-NEXT: add sp, sp, #48
; NONEON-NOSVE-NEXT: ret
%res = udiv <2 x i64> %op1, %op2
ret <2 x i64> %res
}
define void @udiv_v4i64(ptr %a, ptr %b) {
; CHECK-LABEL: udiv_v4i64:
; CHECK: // %bb.0:
; CHECK-NEXT: ldp q0, q3, [x1]
; CHECK-NEXT: ptrue p0.d, vl2
; CHECK-NEXT: ldp q1, q2, [x0]
; CHECK-NEXT: udivr z0.d, p0/m, z0.d, z1.d
; CHECK-NEXT: movprfx z1, z2
; CHECK-NEXT: udiv z1.d, p0/m, z1.d, z3.d
; CHECK-NEXT: stp q0, q1, [x0]
; CHECK-NEXT: ret
;
; NEON-NOSVE-LABEL: udiv_v4i64:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: ldp q0, q3, [x1]
; NEON-NOSVE-NEXT: ptrue p0.d, vl2
; NEON-NOSVE-NEXT: ldp q1, q2, [x0]
; NEON-NOSVE-NEXT: udivr z0.d, p0/m, z0.d, z1.d
; NEON-NOSVE-NEXT: movprfx z1, z2
; NEON-NOSVE-NEXT: udiv z1.d, p0/m, z1.d, z3.d
; NEON-NOSVE-NEXT: stp q0, q1, [x0]
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: udiv_v4i64:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #96
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 96
; NONEON-NOSVE-NEXT: ldp q3, q0, [x1]
; NONEON-NOSVE-NEXT: ldp q2, q1, [x0]
; NONEON-NOSVE-NEXT: stp q2, q3, [sp]
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #32]
; NONEON-NOSVE-NEXT: ldp x9, x10, [sp, #32]
; NONEON-NOSVE-NEXT: ldr x8, [sp, #56]
; NONEON-NOSVE-NEXT: udiv x11, x10, x8
; NONEON-NOSVE-NEXT: ldr x8, [sp, #48]
; NONEON-NOSVE-NEXT: udiv x8, x9, x8
; NONEON-NOSVE-NEXT: ldp x9, x10, [sp]
; NONEON-NOSVE-NEXT: stp x8, x11, [sp, #80]
; NONEON-NOSVE-NEXT: ldr x8, [sp, #24]
; NONEON-NOSVE-NEXT: udiv x11, x10, x8
; NONEON-NOSVE-NEXT: ldr x8, [sp, #16]
; NONEON-NOSVE-NEXT: udiv x8, x9, x8
; NONEON-NOSVE-NEXT: stp x8, x11, [sp, #64]
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #64]
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
; NONEON-NOSVE-NEXT: add sp, sp, #96
; NONEON-NOSVE-NEXT: ret
%op1 = load <4 x i64>, ptr %a
%op2 = load <4 x i64>, ptr %b
%res = udiv <4 x i64> %op1, %op2
store <4 x i64> %res, ptr %a
ret void
}
define void @udiv_constantsplat_v8i32(ptr %a) {
; SVE-LABEL: udiv_constantsplat_v8i32:
; SVE: // %bb.0:
; SVE-NEXT: mov w8, #8969 // =0x2309
; SVE-NEXT: ldp q1, q2, [x0]
; SVE-NEXT: movk w8, #22765, lsl #16
; SVE-NEXT: ptrue p0.s, vl4
; SVE-NEXT: mov z0.s, w8
; SVE-NEXT: movprfx z3, z1
; SVE-NEXT: umulh z3.s, p0/m, z3.s, z0.s
; SVE-NEXT: umulh z0.s, p0/m, z0.s, z2.s
; SVE-NEXT: sub z1.s, z1.s, z3.s
; SVE-NEXT: sub z2.s, z2.s, z0.s
; SVE-NEXT: lsr z1.s, z1.s, #1
; SVE-NEXT: lsr z2.s, z2.s, #1
; SVE-NEXT: add z1.s, z1.s, z3.s
; SVE-NEXT: add z0.s, z2.s, z0.s
; SVE-NEXT: lsr z1.s, z1.s, #6
; SVE-NEXT: lsr z0.s, z0.s, #6
; SVE-NEXT: stp q1, q0, [x0]
; SVE-NEXT: ret
;
; SVE2-LABEL: udiv_constantsplat_v8i32:
; SVE2: // %bb.0:
; SVE2-NEXT: mov w8, #8969 // =0x2309
; SVE2-NEXT: ldp q1, q2, [x0]
; SVE2-NEXT: movk w8, #22765, lsl #16
; SVE2-NEXT: mov z0.s, w8
; SVE2-NEXT: umulh z3.s, z1.s, z0.s
; SVE2-NEXT: umulh z0.s, z2.s, z0.s
; SVE2-NEXT: sub z1.s, z1.s, z3.s
; SVE2-NEXT: sub z2.s, z2.s, z0.s
; SVE2-NEXT: usra z3.s, z1.s, #1
; SVE2-NEXT: usra z0.s, z2.s, #1
; SVE2-NEXT: lsr z1.s, z3.s, #6
; SVE2-NEXT: lsr z0.s, z0.s, #6
; SVE2-NEXT: stp q1, q0, [x0]
; SVE2-NEXT: ret
;
; NEON-NOSVE-LABEL: udiv_constantsplat_v8i32:
; NEON-NOSVE: // %bb.0:
; NEON-NOSVE-NEXT: mov w8, #8969 // =0x2309
; NEON-NOSVE-NEXT: ldp q1, q2, [x0]
; NEON-NOSVE-NEXT: movk w8, #22765, lsl #16
; NEON-NOSVE-NEXT: dup v0.4s, w8
; NEON-NOSVE-NEXT: umull2 v3.2d, v1.4s, v0.4s
; NEON-NOSVE-NEXT: umull v4.2d, v1.2s, v0.2s
; NEON-NOSVE-NEXT: umull2 v5.2d, v2.4s, v0.4s
; NEON-NOSVE-NEXT: umull v0.2d, v2.2s, v0.2s
; NEON-NOSVE-NEXT: uzp2 v3.4s, v4.4s, v3.4s
; NEON-NOSVE-NEXT: uzp2 v0.4s, v0.4s, v5.4s
; NEON-NOSVE-NEXT: sub v1.4s, v1.4s, v3.4s
; NEON-NOSVE-NEXT: sub v2.4s, v2.4s, v0.4s
; NEON-NOSVE-NEXT: usra v3.4s, v1.4s, #1
; NEON-NOSVE-NEXT: usra v0.4s, v2.4s, #1
; NEON-NOSVE-NEXT: ushr v1.4s, v3.4s, #6
; NEON-NOSVE-NEXT: ushr v0.4s, v0.4s, #6
; NEON-NOSVE-NEXT: stp q1, q0, [x0]
; NEON-NOSVE-NEXT: ret
;
; NONEON-NOSVE-LABEL: udiv_constantsplat_v8i32:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: ldp q1, q0, [x0]
; NONEON-NOSVE-NEXT: mov w8, #8969 // =0x2309
; NONEON-NOSVE-NEXT: movk w8, #22765, lsl #16
; NONEON-NOSVE-NEXT: stp q1, q0, [sp, #-64]!
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
; NONEON-NOSVE-NEXT: ldr w9, [sp, #28]
; NONEON-NOSVE-NEXT: umull x10, w9, w8
; NONEON-NOSVE-NEXT: lsr x10, x10, #32
; NONEON-NOSVE-NEXT: sub w9, w9, w10
; NONEON-NOSVE-NEXT: add w9, w10, w9, lsr #1
; NONEON-NOSVE-NEXT: lsr w11, w9, #6
; NONEON-NOSVE-NEXT: ldr w9, [sp, #24]
; NONEON-NOSVE-NEXT: umull x10, w9, w8
; NONEON-NOSVE-NEXT: lsr x10, x10, #32
; NONEON-NOSVE-NEXT: sub w9, w9, w10
; NONEON-NOSVE-NEXT: add w9, w10, w9, lsr #1
; NONEON-NOSVE-NEXT: lsr w9, w9, #6
; NONEON-NOSVE-NEXT: stp w9, w11, [sp, #56]
; NONEON-NOSVE-NEXT: ldr w9, [sp, #20]
; NONEON-NOSVE-NEXT: umull x10, w9, w8
; NONEON-NOSVE-NEXT: lsr x10, x10, #32
; NONEON-NOSVE-NEXT: sub w9, w9, w10
; NONEON-NOSVE-NEXT: add w9, w10, w9, lsr #1
; NONEON-NOSVE-NEXT: lsr w11, w9, #6
; NONEON-NOSVE-NEXT: ldr w9, [sp, #16]
; NONEON-NOSVE-NEXT: umull x10, w9, w8
; NONEON-NOSVE-NEXT: lsr x10, x10, #32
; NONEON-NOSVE-NEXT: sub w9, w9, w10
; NONEON-NOSVE-NEXT: add w9, w10, w9, lsr #1
; NONEON-NOSVE-NEXT: lsr w9, w9, #6
; NONEON-NOSVE-NEXT: stp w9, w11, [sp, #48]
; NONEON-NOSVE-NEXT: ldr w9, [sp, #12]
; NONEON-NOSVE-NEXT: umull x10, w9, w8
; NONEON-NOSVE-NEXT: lsr x10, x10, #32
; NONEON-NOSVE-NEXT: sub w9, w9, w10
; NONEON-NOSVE-NEXT: add w9, w10, w9, lsr #1
; NONEON-NOSVE-NEXT: lsr w11, w9, #6
; NONEON-NOSVE-NEXT: ldr w9, [sp, #8]
; NONEON-NOSVE-NEXT: umull x10, w9, w8
; NONEON-NOSVE-NEXT: lsr x10, x10, #32
; NONEON-NOSVE-NEXT: sub w9, w9, w10
; NONEON-NOSVE-NEXT: add w9, w10, w9, lsr #1
; NONEON-NOSVE-NEXT: lsr w9, w9, #6
; NONEON-NOSVE-NEXT: stp w9, w11, [sp, #40]
; NONEON-NOSVE-NEXT: ldr w9, [sp, #4]
; NONEON-NOSVE-NEXT: umull x10, w9, w8
; NONEON-NOSVE-NEXT: lsr x10, x10, #32
; NONEON-NOSVE-NEXT: sub w9, w9, w10
; NONEON-NOSVE-NEXT: add w9, w10, w9, lsr #1
; NONEON-NOSVE-NEXT: lsr w11, w9, #6
; NONEON-NOSVE-NEXT: ldr w9, [sp]
; NONEON-NOSVE-NEXT: umull x8, w9, w8
; NONEON-NOSVE-NEXT: lsr x8, x8, #32
; NONEON-NOSVE-NEXT: sub w9, w9, w8
; NONEON-NOSVE-NEXT: add w8, w8, w9, lsr #1
; NONEON-NOSVE-NEXT: lsr w8, w8, #6
; NONEON-NOSVE-NEXT: stp w8, w11, [sp, #32]
; NONEON-NOSVE-NEXT: ldp q0, q1, [sp, #32]
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
; NONEON-NOSVE-NEXT: add sp, sp, #64
; NONEON-NOSVE-NEXT: ret
%op1 = load <8 x i32>, ptr %a
%res = udiv <8 x i32> %op1, <i32 95, i32 95, i32 95, i32 95, i32 95, i32 95, i32 95, i32 95>
store <8 x i32> %res, ptr %a
ret void
}
Codebase Browser
llvm