; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mattr=+sve -force-streaming-compatible < %s | FileCheck %s
; RUN: llc -mattr=+sme -force-streaming < %s | FileCheck %s
; RUN: llc -force-streaming-compatible < %s | FileCheck %s --check-prefix=NONEON-NOSVE
target triple = "aarch64-unknown-linux-gnu"
define <2 x half> @select_v2f16(<2 x half> %op1, <2 x half> %op2, i1 %mask) {
; CHECK-LABEL: select_v2f16:
; CHECK: // %bb.0:
; CHECK-NEXT: mov z2.h, w0
; CHECK-NEXT: ptrue p0.h
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
; CHECK-NEXT: and z2.h, z2.h, #0x1
; CHECK-NEXT: cmpne p0.h, p0/z, z2.h, #0
; CHECK-NEXT: sel z0.h, p0, z0.h, z1.h
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT: ret
;
; NONEON-NOSVE-LABEL: select_v2f16:
; 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: tst w0, #0x1
; NONEON-NOSVE-NEXT: ldr h0, [sp, #22]
; NONEON-NOSVE-NEXT: ldr h1, [sp, #14]
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #12]
; NONEON-NOSVE-NEXT: str h0, [sp, #30]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #20]
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #10]
; NONEON-NOSVE-NEXT: str h0, [sp, #28]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #18]
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #8]
; NONEON-NOSVE-NEXT: str h0, [sp, #26]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #16]
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: str h0, [sp, #24]
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT: add sp, sp, #32
; NONEON-NOSVE-NEXT: ret
%sel = select i1 %mask, <2 x half> %op1, <2 x half> %op2
ret <2 x half> %sel
}
define <4 x half> @select_v4f16(<4 x half> %op1, <4 x half> %op2, i1 %mask) {
; CHECK-LABEL: select_v4f16:
; CHECK: // %bb.0:
; CHECK-NEXT: mov z2.h, w0
; CHECK-NEXT: ptrue p0.h
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
; CHECK-NEXT: and z2.h, z2.h, #0x1
; CHECK-NEXT: cmpne p0.h, p0/z, z2.h, #0
; CHECK-NEXT: sel z0.h, p0, z0.h, z1.h
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT: ret
;
; NONEON-NOSVE-LABEL: select_v4f16:
; 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: tst w0, #0x1
; NONEON-NOSVE-NEXT: ldr h0, [sp, #22]
; NONEON-NOSVE-NEXT: ldr h1, [sp, #14]
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #12]
; NONEON-NOSVE-NEXT: str h0, [sp, #30]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #20]
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #10]
; NONEON-NOSVE-NEXT: str h0, [sp, #28]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #18]
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #8]
; NONEON-NOSVE-NEXT: str h0, [sp, #26]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #16]
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: str h0, [sp, #24]
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT: add sp, sp, #32
; NONEON-NOSVE-NEXT: ret
%sel = select i1 %mask, <4 x half> %op1, <4 x half> %op2
ret <4 x half> %sel
}
define <8 x half> @select_v8f16(<8 x half> %op1, <8 x half> %op2, i1 %mask) {
; CHECK-LABEL: select_v8f16:
; CHECK: // %bb.0:
; CHECK-NEXT: mov z2.h, w0
; CHECK-NEXT: ptrue p0.h
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
; CHECK-NEXT: and z2.h, z2.h, #0x1
; CHECK-NEXT: cmpne p0.h, p0/z, z2.h, #0
; CHECK-NEXT: sel z0.h, p0, z0.h, z1.h
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
; CHECK-NEXT: ret
;
; NONEON-NOSVE-LABEL: select_v8f16:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: stp q0, q1, [sp, #-48]!
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 48
; NONEON-NOSVE-NEXT: ldr h0, [sp, #30]
; NONEON-NOSVE-NEXT: ldr h1, [sp, #14]
; NONEON-NOSVE-NEXT: tst w0, #0x1
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #12]
; NONEON-NOSVE-NEXT: str h0, [sp, #46]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #28]
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #10]
; NONEON-NOSVE-NEXT: str h0, [sp, #44]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #26]
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #8]
; NONEON-NOSVE-NEXT: str h0, [sp, #42]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #24]
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #6]
; NONEON-NOSVE-NEXT: str h0, [sp, #40]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #22]
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #4]
; NONEON-NOSVE-NEXT: str h0, [sp, #38]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #20]
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #2]
; NONEON-NOSVE-NEXT: str h0, [sp, #36]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #18]
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp]
; NONEON-NOSVE-NEXT: str h0, [sp, #34]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #16]
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: str h0, [sp, #32]
; NONEON-NOSVE-NEXT: ldr q0, [sp, #32]
; NONEON-NOSVE-NEXT: add sp, sp, #48
; NONEON-NOSVE-NEXT: ret
%sel = select i1 %mask, <8 x half> %op1, <8 x half> %op2
ret <8 x half> %sel
}
define void @select_v16f16(ptr %a, ptr %b, i1 %mask) {
; CHECK-LABEL: select_v16f16:
; CHECK: // %bb.0:
; CHECK-NEXT: mov z0.h, w2
; CHECK-NEXT: ptrue p0.h
; CHECK-NEXT: and z0.h, z0.h, #0x1
; CHECK-NEXT: cmpne p0.h, p0/z, z0.h, #0
; CHECK-NEXT: ldr q0, [x0]
; CHECK-NEXT: ldr q1, [x0, #16]
; CHECK-NEXT: ldr q2, [x1]
; CHECK-NEXT: ldr q3, [x1, #16]
; CHECK-NEXT: sel z0.h, p0, z0.h, z2.h
; CHECK-NEXT: sel z1.h, p0, z1.h, z3.h
; CHECK-NEXT: stp q0, q1, [x0]
; CHECK-NEXT: ret
;
; NONEON-NOSVE-LABEL: select_v16f16:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #96
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 96
; NONEON-NOSVE-NEXT: ldr q0, [x0]
; NONEON-NOSVE-NEXT: ldr q1, [x0, #16]
; NONEON-NOSVE-NEXT: tst w2, #0x1
; NONEON-NOSVE-NEXT: ldr q2, [x1]
; NONEON-NOSVE-NEXT: ldr q3, [x1, #16]
; NONEON-NOSVE-NEXT: str q0, [sp]
; NONEON-NOSVE-NEXT: stp q1, q3, [sp, #16]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #30]
; NONEON-NOSVE-NEXT: ldr h1, [sp, #46]
; NONEON-NOSVE-NEXT: str q2, [sp, #48]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #44]
; NONEON-NOSVE-NEXT: str h0, [sp, #78]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #28]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #42]
; NONEON-NOSVE-NEXT: str h0, [sp, #76]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #26]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #40]
; NONEON-NOSVE-NEXT: str h0, [sp, #74]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #24]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #38]
; NONEON-NOSVE-NEXT: str h0, [sp, #72]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #22]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #36]
; NONEON-NOSVE-NEXT: str h0, [sp, #70]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #20]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #34]
; NONEON-NOSVE-NEXT: str h0, [sp, #68]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #18]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #32]
; NONEON-NOSVE-NEXT: str h0, [sp, #66]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #16]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #62]
; NONEON-NOSVE-NEXT: str h0, [sp, #64]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #14]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #60]
; NONEON-NOSVE-NEXT: str h0, [sp, #94]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #12]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #58]
; NONEON-NOSVE-NEXT: str h0, [sp, #92]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #10]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #56]
; NONEON-NOSVE-NEXT: str h0, [sp, #90]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #8]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #54]
; NONEON-NOSVE-NEXT: str h0, [sp, #88]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #6]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #52]
; NONEON-NOSVE-NEXT: str h0, [sp, #86]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #4]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #50]
; NONEON-NOSVE-NEXT: str h0, [sp, #84]
; NONEON-NOSVE-NEXT: ldr h0, [sp, #2]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: ldr h1, [sp, #48]
; NONEON-NOSVE-NEXT: str h0, [sp, #82]
; NONEON-NOSVE-NEXT: ldr h0, [sp]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: str h0, [sp, #80]
; NONEON-NOSVE-NEXT: ldp q1, q0, [sp, #64]
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
; NONEON-NOSVE-NEXT: add sp, sp, #96
; NONEON-NOSVE-NEXT: ret
%op1 = load volatile <16 x half>, ptr %a
%op2 = load volatile <16 x half>, ptr %b
%sel = select i1 %mask, <16 x half> %op1, <16 x half> %op2
store <16 x half> %sel, ptr %a
ret void
}
define <2 x float> @select_v2f32(<2 x float> %op1, <2 x float> %op2, i1 %mask) {
; CHECK-LABEL: select_v2f32:
; CHECK: // %bb.0:
; CHECK-NEXT: and w8, w0, #0x1
; CHECK-NEXT: ptrue p0.s
; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
; CHECK-NEXT: // kill: def $d1 killed $d1 def $z1
; CHECK-NEXT: mov z2.s, w8
; CHECK-NEXT: cmpne p0.s, p0/z, z2.s, #0
; CHECK-NEXT: sel z0.s, p0, z0.s, z1.s
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
; CHECK-NEXT: ret
;
; NONEON-NOSVE-LABEL: select_v2f32:
; 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: tst w0, #0x1
; NONEON-NOSVE-NEXT: ldp s1, s2, [sp, #8]
; NONEON-NOSVE-NEXT: ldr s0, [sp, #20]
; NONEON-NOSVE-NEXT: fcsel s3, s2, s0, ne
; NONEON-NOSVE-NEXT: ldr s0, [sp, #16]
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: stp s0, s3, [sp, #24]
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT: add sp, sp, #32
; NONEON-NOSVE-NEXT: ret
%sel = select i1 %mask, <2 x float> %op1, <2 x float> %op2
ret <2 x float> %sel
}
define <4 x float> @select_v4f32(<4 x float> %op1, <4 x float> %op2, i1 %mask) {
; CHECK-LABEL: select_v4f32:
; CHECK: // %bb.0:
; CHECK-NEXT: and w8, w0, #0x1
; CHECK-NEXT: ptrue p0.s
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
; CHECK-NEXT: mov z2.s, w8
; CHECK-NEXT: cmpne p0.s, p0/z, z2.s, #0
; CHECK-NEXT: sel z0.s, p0, z0.s, z1.s
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
; CHECK-NEXT: ret
;
; NONEON-NOSVE-LABEL: select_v4f32:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: stp q0, q1, [sp, #-48]!
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 48
; NONEON-NOSVE-NEXT: ldp s1, s2, [sp, #8]
; NONEON-NOSVE-NEXT: tst w0, #0x1
; NONEON-NOSVE-NEXT: ldr s0, [sp, #28]
; NONEON-NOSVE-NEXT: fcsel s3, s2, s0, ne
; NONEON-NOSVE-NEXT: ldr s0, [sp, #24]
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: ldp s1, s2, [sp]
; NONEON-NOSVE-NEXT: stp s0, s3, [sp, #40]
; NONEON-NOSVE-NEXT: ldr s0, [sp, #20]
; NONEON-NOSVE-NEXT: fcsel s3, s2, s0, ne
; NONEON-NOSVE-NEXT: ldr s0, [sp, #16]
; NONEON-NOSVE-NEXT: fcsel s0, s1, s0, ne
; NONEON-NOSVE-NEXT: stp s0, s3, [sp, #32]
; NONEON-NOSVE-NEXT: ldr q0, [sp, #32]
; NONEON-NOSVE-NEXT: add sp, sp, #48
; NONEON-NOSVE-NEXT: ret
%sel = select i1 %mask, <4 x float> %op1, <4 x float> %op2
ret <4 x float> %sel
}
define void @select_v8f32(ptr %a, ptr %b, i1 %mask) {
; CHECK-LABEL: select_v8f32:
; CHECK: // %bb.0:
; CHECK-NEXT: and w8, w2, #0x1
; CHECK-NEXT: ptrue p0.s
; CHECK-NEXT: mov z0.s, w8
; CHECK-NEXT: cmpne p0.s, p0/z, z0.s, #0
; CHECK-NEXT: ldr q0, [x0]
; CHECK-NEXT: ldr q1, [x0, #16]
; CHECK-NEXT: ldr q2, [x1]
; CHECK-NEXT: ldr q3, [x1, #16]
; CHECK-NEXT: sel z0.s, p0, z0.s, z2.s
; CHECK-NEXT: sel z1.s, p0, z1.s, z3.s
; CHECK-NEXT: stp q0, q1, [x0]
; CHECK-NEXT: ret
;
; NONEON-NOSVE-LABEL: select_v8f32:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #96
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 96
; NONEON-NOSVE-NEXT: ldr q0, [x0]
; NONEON-NOSVE-NEXT: ldr q1, [x0, #16]
; NONEON-NOSVE-NEXT: tst w2, #0x1
; NONEON-NOSVE-NEXT: ldr q2, [x1]
; NONEON-NOSVE-NEXT: ldr q3, [x1, #16]
; NONEON-NOSVE-NEXT: str q0, [sp]
; NONEON-NOSVE-NEXT: stp q1, q3, [sp, #16]
; NONEON-NOSVE-NEXT: str q2, [sp, #48]
; NONEON-NOSVE-NEXT: ldp s1, s2, [sp, #40]
; NONEON-NOSVE-NEXT: ldr s0, [sp, #28]
; NONEON-NOSVE-NEXT: fcsel s3, s0, s2, ne
; NONEON-NOSVE-NEXT: ldr s0, [sp, #24]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: ldp s1, s2, [sp, #32]
; NONEON-NOSVE-NEXT: stp s0, s3, [sp, #72]
; NONEON-NOSVE-NEXT: ldr s0, [sp, #20]
; NONEON-NOSVE-NEXT: fcsel s3, s0, s2, ne
; NONEON-NOSVE-NEXT: ldr s0, [sp, #16]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: ldp s1, s2, [sp, #56]
; NONEON-NOSVE-NEXT: stp s0, s3, [sp, #64]
; NONEON-NOSVE-NEXT: ldr s0, [sp, #12]
; NONEON-NOSVE-NEXT: fcsel s3, s0, s2, ne
; NONEON-NOSVE-NEXT: ldr s0, [sp, #8]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: ldp s1, s2, [sp, #48]
; NONEON-NOSVE-NEXT: stp s0, s3, [sp, #88]
; NONEON-NOSVE-NEXT: ldr s0, [sp, #4]
; NONEON-NOSVE-NEXT: fcsel s3, s0, s2, ne
; NONEON-NOSVE-NEXT: ldr s0, [sp]
; NONEON-NOSVE-NEXT: fcsel s0, s0, s1, ne
; NONEON-NOSVE-NEXT: stp s0, s3, [sp, #80]
; NONEON-NOSVE-NEXT: ldp q1, q0, [sp, #64]
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
; NONEON-NOSVE-NEXT: add sp, sp, #96
; NONEON-NOSVE-NEXT: ret
%op1 = load volatile <8 x float>, ptr %a
%op2 = load volatile <8 x float>, ptr %b
%sel = select i1 %mask, <8 x float> %op1, <8 x float> %op2
store <8 x float> %sel, ptr %a
ret void
}
define <1 x double> @select_v1f64(<1 x double> %op1, <1 x double> %op2, i1 %mask) {
; CHECK-LABEL: select_v1f64:
; CHECK: // %bb.0:
; CHECK-NEXT: tst w0, #0x1
; CHECK-NEXT: fcsel d0, d0, d1, ne
; CHECK-NEXT: ret
;
; NONEON-NOSVE-LABEL: select_v1f64:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #16
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 16
; NONEON-NOSVE-NEXT: tst w0, #0x1
; NONEON-NOSVE-NEXT: fcsel d0, d0, d1, ne
; NONEON-NOSVE-NEXT: str d0, [sp, #8]
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
; NONEON-NOSVE-NEXT: add sp, sp, #16
; NONEON-NOSVE-NEXT: ret
%sel = select i1 %mask, <1 x double> %op1, <1 x double> %op2
ret <1 x double> %sel
}
define <2 x double> @select_v2f64(<2 x double> %op1, <2 x double> %op2, i1 %mask) {
; CHECK-LABEL: select_v2f64:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $w0 killed $w0 def $x0
; CHECK-NEXT: and x8, x0, #0x1
; CHECK-NEXT: ptrue p0.d
; CHECK-NEXT: // kill: def $q0 killed $q0 def $z0
; CHECK-NEXT: // kill: def $q1 killed $q1 def $z1
; CHECK-NEXT: mov z2.d, x8
; CHECK-NEXT: cmpne p0.d, p0/z, z2.d, #0
; CHECK-NEXT: sel z0.d, p0, z0.d, z1.d
; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0
; CHECK-NEXT: ret
;
; NONEON-NOSVE-LABEL: select_v2f64:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: stp q0, q1, [sp, #-48]!
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 48
; NONEON-NOSVE-NEXT: ldp d1, d2, [sp]
; NONEON-NOSVE-NEXT: tst w0, #0x1
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT: fcsel d3, d2, d0, ne
; NONEON-NOSVE-NEXT: ldr d0, [sp, #16]
; NONEON-NOSVE-NEXT: fcsel d0, d1, d0, ne
; NONEON-NOSVE-NEXT: stp d0, d3, [sp, #32]
; NONEON-NOSVE-NEXT: ldr q0, [sp, #32]
; NONEON-NOSVE-NEXT: add sp, sp, #48
; NONEON-NOSVE-NEXT: ret
%sel = select i1 %mask, <2 x double> %op1, <2 x double> %op2
ret <2 x double> %sel
}
define void @select_v4f64(ptr %a, ptr %b, i1 %mask) {
; CHECK-LABEL: select_v4f64:
; CHECK: // %bb.0:
; CHECK-NEXT: // kill: def $w2 killed $w2 def $x2
; CHECK-NEXT: and x8, x2, #0x1
; CHECK-NEXT: ptrue p0.d
; CHECK-NEXT: mov z0.d, x8
; CHECK-NEXT: cmpne p0.d, p0/z, z0.d, #0
; CHECK-NEXT: ldr q0, [x0]
; CHECK-NEXT: ldr q1, [x0, #16]
; CHECK-NEXT: ldr q2, [x1]
; CHECK-NEXT: ldr q3, [x1, #16]
; CHECK-NEXT: sel z0.d, p0, z0.d, z2.d
; CHECK-NEXT: sel z1.d, p0, z1.d, z3.d
; CHECK-NEXT: stp q0, q1, [x0]
; CHECK-NEXT: ret
;
; NONEON-NOSVE-LABEL: select_v4f64:
; NONEON-NOSVE: // %bb.0:
; NONEON-NOSVE-NEXT: sub sp, sp, #96
; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 96
; NONEON-NOSVE-NEXT: ldr q0, [x0]
; NONEON-NOSVE-NEXT: ldr q1, [x0, #16]
; NONEON-NOSVE-NEXT: tst w2, #0x1
; NONEON-NOSVE-NEXT: ldr q2, [x1]
; NONEON-NOSVE-NEXT: ldr q3, [x1, #16]
; NONEON-NOSVE-NEXT: str q0, [sp]
; NONEON-NOSVE-NEXT: stp q1, q3, [sp, #16]
; NONEON-NOSVE-NEXT: str q2, [sp, #48]
; NONEON-NOSVE-NEXT: ldp d1, d2, [sp, #32]
; NONEON-NOSVE-NEXT: ldr d0, [sp, #24]
; NONEON-NOSVE-NEXT: fcsel d3, d0, d2, ne
; NONEON-NOSVE-NEXT: ldr d0, [sp, #16]
; NONEON-NOSVE-NEXT: fcsel d0, d0, d1, ne
; NONEON-NOSVE-NEXT: ldp d1, d2, [sp, #48]
; NONEON-NOSVE-NEXT: stp d0, d3, [sp, #64]
; NONEON-NOSVE-NEXT: ldr d0, [sp, #8]
; NONEON-NOSVE-NEXT: fcsel d3, d0, d2, ne
; NONEON-NOSVE-NEXT: ldr d0, [sp]
; NONEON-NOSVE-NEXT: fcsel d0, d0, d1, ne
; NONEON-NOSVE-NEXT: stp d0, d3, [sp, #80]
; NONEON-NOSVE-NEXT: ldp q1, q0, [sp, #64]
; NONEON-NOSVE-NEXT: stp q0, q1, [x0]
; NONEON-NOSVE-NEXT: add sp, sp, #96
; NONEON-NOSVE-NEXT: ret
%op1 = load volatile <4 x double>, ptr %a
%op2 = load volatile <4 x double>, ptr %b
%sel = select i1 %mask, <4 x double> %op1, <4 x double> %op2
store <4 x double> %sel, ptr %a
ret void
}
Codebase Browser
llvm