; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
; RUN: llc -mtriple=aarch64 -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD,CHECK-SD-NOFP16
; RUN: llc -mtriple=aarch64 -mattr=+fullfp16 -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD,CHECK-SD-FP16
; RUN: llc -mtriple=aarch64 -global-isel -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-GI,CHECK-GI-NOFP16
; RUN: llc -mtriple=aarch64 -mattr=+fullfp16 -global-isel -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-GI,CHECK-GI-FP16
define fp128 @f128_fp128(fp128 %a, fp128 %b, fp128 %d, fp128 %e) {
; CHECK-SD-LABEL: f128_fp128:
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: sub sp, sp, #48
; CHECK-SD-NEXT: str x30, [sp, #32] // 8-byte Folded Spill
; CHECK-SD-NEXT: .cfi_def_cfa_offset 48
; CHECK-SD-NEXT: .cfi_offset w30, -16
; CHECK-SD-NEXT: stp q2, q3, [sp] // 32-byte Folded Spill
; CHECK-SD-NEXT: bl __lttf2
; CHECK-SD-NEXT: ldr q0, [sp, #16] // 16-byte Folded Reload
; CHECK-SD-NEXT: cmp w0, #0
; CHECK-SD-NEXT: b.ge .LBB0_2
; CHECK-SD-NEXT: // %bb.1: // %entry
; CHECK-SD-NEXT: ldr q0, [sp] // 16-byte Folded Reload
; CHECK-SD-NEXT: .LBB0_2: // %entry
; CHECK-SD-NEXT: ldr x30, [sp, #32] // 8-byte Folded Reload
; CHECK-SD-NEXT: add sp, sp, #48
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: f128_fp128:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: sub sp, sp, #48
; CHECK-GI-NEXT: str x30, [sp, #32] // 8-byte Folded Spill
; CHECK-GI-NEXT: .cfi_def_cfa_offset 48
; CHECK-GI-NEXT: .cfi_offset w30, -16
; CHECK-GI-NEXT: stp q3, q2, [sp] // 32-byte Folded Spill
; CHECK-GI-NEXT: bl __lttf2
; CHECK-GI-NEXT: ldp q3, q2, [sp] // 32-byte Folded Reload
; CHECK-GI-NEXT: cmp w0, #0
; CHECK-GI-NEXT: ldr x30, [sp, #32] // 8-byte Folded Reload
; CHECK-GI-NEXT: mov d0, v2.d[1]
; CHECK-GI-NEXT: mov d1, v3.d[1]
; CHECK-GI-NEXT: fcsel d2, d2, d3, lt
; CHECK-GI-NEXT: fmov x8, d2
; CHECK-GI-NEXT: fcsel d1, d0, d1, lt
; CHECK-GI-NEXT: mov v0.d[0], x8
; CHECK-GI-NEXT: fmov x8, d1
; CHECK-GI-NEXT: mov v0.d[1], x8
; CHECK-GI-NEXT: add sp, sp, #48
; CHECK-GI-NEXT: ret
entry:
%c = fcmp olt fp128 %a, %b
%s = select i1 %c, fp128 %d, fp128 %e
ret fp128 %s
}
define i128 @f128_i128(fp128 %a, fp128 %b, i128 %d, i128 %e) {
; CHECK-SD-LABEL: f128_i128:
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: sub sp, sp, #80
; CHECK-SD-NEXT: str x30, [sp, #32] // 8-byte Folded Spill
; CHECK-SD-NEXT: stp x22, x21, [sp, #48] // 16-byte Folded Spill
; CHECK-SD-NEXT: stp x20, x19, [sp, #64] // 16-byte Folded Spill
; CHECK-SD-NEXT: .cfi_def_cfa_offset 80
; CHECK-SD-NEXT: .cfi_offset w19, -8
; CHECK-SD-NEXT: .cfi_offset w20, -16
; CHECK-SD-NEXT: .cfi_offset w21, -24
; CHECK-SD-NEXT: .cfi_offset w22, -32
; CHECK-SD-NEXT: .cfi_offset w30, -48
; CHECK-SD-NEXT: mov x19, x3
; CHECK-SD-NEXT: mov x20, x2
; CHECK-SD-NEXT: mov x21, x1
; CHECK-SD-NEXT: mov x22, x0
; CHECK-SD-NEXT: stp q0, q1, [sp] // 32-byte Folded Spill
; CHECK-SD-NEXT: bl __lttf2
; CHECK-SD-NEXT: ldp q0, q1, [sp] // 32-byte Folded Reload
; CHECK-SD-NEXT: cmp w0, #0
; CHECK-SD-NEXT: csel x20, x22, x20, lt
; CHECK-SD-NEXT: bl __lttf2
; CHECK-SD-NEXT: mov w8, w0
; CHECK-SD-NEXT: mov x0, x20
; CHECK-SD-NEXT: ldr x30, [sp, #32] // 8-byte Folded Reload
; CHECK-SD-NEXT: cmp w8, #0
; CHECK-SD-NEXT: csel x1, x21, x19, lt
; CHECK-SD-NEXT: ldp x20, x19, [sp, #64] // 16-byte Folded Reload
; CHECK-SD-NEXT: ldp x22, x21, [sp, #48] // 16-byte Folded Reload
; CHECK-SD-NEXT: add sp, sp, #80
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: f128_i128:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: str x30, [sp, #-48]! // 8-byte Folded Spill
; CHECK-GI-NEXT: stp x22, x21, [sp, #16] // 16-byte Folded Spill
; CHECK-GI-NEXT: stp x20, x19, [sp, #32] // 16-byte Folded Spill
; CHECK-GI-NEXT: .cfi_def_cfa_offset 48
; CHECK-GI-NEXT: .cfi_offset w19, -8
; CHECK-GI-NEXT: .cfi_offset w20, -16
; CHECK-GI-NEXT: .cfi_offset w21, -24
; CHECK-GI-NEXT: .cfi_offset w22, -32
; CHECK-GI-NEXT: .cfi_offset w30, -48
; CHECK-GI-NEXT: mov x19, x0
; CHECK-GI-NEXT: mov x20, x1
; CHECK-GI-NEXT: mov x21, x2
; CHECK-GI-NEXT: mov x22, x3
; CHECK-GI-NEXT: bl __lttf2
; CHECK-GI-NEXT: cmp w0, #0
; CHECK-GI-NEXT: csel x0, x19, x21, lt
; CHECK-GI-NEXT: csel x1, x20, x22, lt
; CHECK-GI-NEXT: ldp x20, x19, [sp, #32] // 16-byte Folded Reload
; CHECK-GI-NEXT: ldp x22, x21, [sp, #16] // 16-byte Folded Reload
; CHECK-GI-NEXT: ldr x30, [sp], #48 // 8-byte Folded Reload
; CHECK-GI-NEXT: ret
entry:
%c = fcmp olt fp128 %a, %b
%s = select i1 %c, i128 %d, i128 %e
ret i128 %s
}
define double @f128_double(fp128 %a, fp128 %b, double %d, double %e) {
; CHECK-SD-LABEL: f128_double:
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: stp d9, d8, [sp, #-32]! // 16-byte Folded Spill
; CHECK-SD-NEXT: str x30, [sp, #16] // 8-byte Folded Spill
; CHECK-SD-NEXT: .cfi_def_cfa_offset 32
; CHECK-SD-NEXT: .cfi_offset w30, -16
; CHECK-SD-NEXT: .cfi_offset b8, -24
; CHECK-SD-NEXT: .cfi_offset b9, -32
; CHECK-SD-NEXT: fmov d8, d3
; CHECK-SD-NEXT: fmov d9, d2
; CHECK-SD-NEXT: bl __lttf2
; CHECK-SD-NEXT: cmp w0, #0
; CHECK-SD-NEXT: ldr x30, [sp, #16] // 8-byte Folded Reload
; CHECK-SD-NEXT: fcsel d0, d9, d8, lt
; CHECK-SD-NEXT: ldp d9, d8, [sp], #32 // 16-byte Folded Reload
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: f128_double:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: stp d9, d8, [sp, #-32]! // 16-byte Folded Spill
; CHECK-GI-NEXT: str x30, [sp, #16] // 8-byte Folded Spill
; CHECK-GI-NEXT: .cfi_def_cfa_offset 32
; CHECK-GI-NEXT: .cfi_offset w30, -16
; CHECK-GI-NEXT: .cfi_offset b8, -24
; CHECK-GI-NEXT: .cfi_offset b9, -32
; CHECK-GI-NEXT: fmov d8, d2
; CHECK-GI-NEXT: fmov d9, d3
; CHECK-GI-NEXT: bl __lttf2
; CHECK-GI-NEXT: cmp w0, #0
; CHECK-GI-NEXT: ldr x30, [sp, #16] // 8-byte Folded Reload
; CHECK-GI-NEXT: fcsel d0, d8, d9, lt
; CHECK-GI-NEXT: ldp d9, d8, [sp], #32 // 16-byte Folded Reload
; CHECK-GI-NEXT: ret
entry:
%c = fcmp olt fp128 %a, %b
%s = select i1 %c, double %d, double %e
ret double %s
}
define float @f128_float(fp128 %a, fp128 %b, float %d, float %e) {
; CHECK-SD-LABEL: f128_float:
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: stp d9, d8, [sp, #-32]! // 16-byte Folded Spill
; CHECK-SD-NEXT: str x30, [sp, #16] // 8-byte Folded Spill
; CHECK-SD-NEXT: .cfi_def_cfa_offset 32
; CHECK-SD-NEXT: .cfi_offset w30, -16
; CHECK-SD-NEXT: .cfi_offset b8, -24
; CHECK-SD-NEXT: .cfi_offset b9, -32
; CHECK-SD-NEXT: fmov s8, s3
; CHECK-SD-NEXT: fmov s9, s2
; CHECK-SD-NEXT: bl __lttf2
; CHECK-SD-NEXT: cmp w0, #0
; CHECK-SD-NEXT: ldr x30, [sp, #16] // 8-byte Folded Reload
; CHECK-SD-NEXT: fcsel s0, s9, s8, lt
; CHECK-SD-NEXT: ldp d9, d8, [sp], #32 // 16-byte Folded Reload
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: f128_float:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: stp d9, d8, [sp, #-32]! // 16-byte Folded Spill
; CHECK-GI-NEXT: str x30, [sp, #16] // 8-byte Folded Spill
; CHECK-GI-NEXT: .cfi_def_cfa_offset 32
; CHECK-GI-NEXT: .cfi_offset w30, -16
; CHECK-GI-NEXT: .cfi_offset b8, -24
; CHECK-GI-NEXT: .cfi_offset b9, -32
; CHECK-GI-NEXT: fmov s8, s2
; CHECK-GI-NEXT: fmov s9, s3
; CHECK-GI-NEXT: bl __lttf2
; CHECK-GI-NEXT: cmp w0, #0
; CHECK-GI-NEXT: ldr x30, [sp, #16] // 8-byte Folded Reload
; CHECK-GI-NEXT: fcsel s0, s8, s9, lt
; CHECK-GI-NEXT: ldp d9, d8, [sp], #32 // 16-byte Folded Reload
; CHECK-GI-NEXT: ret
entry:
%c = fcmp olt fp128 %a, %b
%s = select i1 %c, float %d, float %e
ret float %s
}
define i32 @f128_i32(fp128 %a, fp128 %b, i32 %d, i32 %e) {
; CHECK-SD-LABEL: f128_i32:
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: str x30, [sp, #-32]! // 8-byte Folded Spill
; CHECK-SD-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-SD-NEXT: .cfi_def_cfa_offset 32
; CHECK-SD-NEXT: .cfi_offset w19, -8
; CHECK-SD-NEXT: .cfi_offset w20, -16
; CHECK-SD-NEXT: .cfi_offset w30, -32
; CHECK-SD-NEXT: mov w19, w1
; CHECK-SD-NEXT: mov w20, w0
; CHECK-SD-NEXT: bl __lttf2
; CHECK-SD-NEXT: cmp w0, #0
; CHECK-SD-NEXT: csel w0, w20, w19, lt
; CHECK-SD-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-SD-NEXT: ldr x30, [sp], #32 // 8-byte Folded Reload
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: f128_i32:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: str x30, [sp, #-32]! // 8-byte Folded Spill
; CHECK-GI-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-GI-NEXT: .cfi_def_cfa_offset 32
; CHECK-GI-NEXT: .cfi_offset w19, -8
; CHECK-GI-NEXT: .cfi_offset w20, -16
; CHECK-GI-NEXT: .cfi_offset w30, -32
; CHECK-GI-NEXT: mov w19, w0
; CHECK-GI-NEXT: mov w20, w1
; CHECK-GI-NEXT: bl __lttf2
; CHECK-GI-NEXT: cmp w0, #0
; CHECK-GI-NEXT: csel w0, w19, w20, lt
; CHECK-GI-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-GI-NEXT: ldr x30, [sp], #32 // 8-byte Folded Reload
; CHECK-GI-NEXT: ret
entry:
%c = fcmp olt fp128 %a, %b
%s = select i1 %c, i32 %d, i32 %e
ret i32 %s
}
define half @f128_half(fp128 %a, fp128 %b, half %d, half %e) {
; CHECK-SD-NOFP16-LABEL: f128_half:
; CHECK-SD-NOFP16: // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT: stp d9, d8, [sp, #-32]! // 16-byte Folded Spill
; CHECK-SD-NOFP16-NEXT: str x30, [sp, #16] // 8-byte Folded Spill
; CHECK-SD-NOFP16-NEXT: .cfi_def_cfa_offset 32
; CHECK-SD-NOFP16-NEXT: .cfi_offset w30, -16
; CHECK-SD-NOFP16-NEXT: .cfi_offset b8, -24
; CHECK-SD-NOFP16-NEXT: .cfi_offset b9, -32
; CHECK-SD-NOFP16-NEXT: fmov s8, s3
; CHECK-SD-NOFP16-NEXT: fmov s9, s2
; CHECK-SD-NOFP16-NEXT: bl __lttf2
; CHECK-SD-NOFP16-NEXT: cmp w0, #0
; CHECK-SD-NOFP16-NEXT: ldr x30, [sp, #16] // 8-byte Folded Reload
; CHECK-SD-NOFP16-NEXT: fcsel s0, s9, s8, lt
; CHECK-SD-NOFP16-NEXT: // kill: def $h0 killed $h0 killed $s0
; CHECK-SD-NOFP16-NEXT: ldp d9, d8, [sp], #32 // 16-byte Folded Reload
; CHECK-SD-NOFP16-NEXT: ret
;
; CHECK-SD-FP16-LABEL: f128_half:
; CHECK-SD-FP16: // %bb.0: // %entry
; CHECK-SD-FP16-NEXT: stp d9, d8, [sp, #-32]! // 16-byte Folded Spill
; CHECK-SD-FP16-NEXT: str x30, [sp, #16] // 8-byte Folded Spill
; CHECK-SD-FP16-NEXT: .cfi_def_cfa_offset 32
; CHECK-SD-FP16-NEXT: .cfi_offset w30, -16
; CHECK-SD-FP16-NEXT: .cfi_offset b8, -24
; CHECK-SD-FP16-NEXT: .cfi_offset b9, -32
; CHECK-SD-FP16-NEXT: fmov s8, s3
; CHECK-SD-FP16-NEXT: fmov s9, s2
; CHECK-SD-FP16-NEXT: bl __lttf2
; CHECK-SD-FP16-NEXT: cmp w0, #0
; CHECK-SD-FP16-NEXT: ldr x30, [sp, #16] // 8-byte Folded Reload
; CHECK-SD-FP16-NEXT: fcsel h0, h9, h8, lt
; CHECK-SD-FP16-NEXT: ldp d9, d8, [sp], #32 // 16-byte Folded Reload
; CHECK-SD-FP16-NEXT: ret
;
; CHECK-GI-LABEL: f128_half:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: stp d9, d8, [sp, #-32]! // 16-byte Folded Spill
; CHECK-GI-NEXT: str x30, [sp, #16] // 8-byte Folded Spill
; CHECK-GI-NEXT: .cfi_def_cfa_offset 32
; CHECK-GI-NEXT: .cfi_offset w30, -16
; CHECK-GI-NEXT: .cfi_offset b8, -24
; CHECK-GI-NEXT: .cfi_offset b9, -32
; CHECK-GI-NEXT: fmov s8, s2
; CHECK-GI-NEXT: fmov s9, s3
; CHECK-GI-NEXT: bl __lttf2
; CHECK-GI-NEXT: fmov w8, s8
; CHECK-GI-NEXT: fmov w9, s9
; CHECK-GI-NEXT: cmp w0, #0
; CHECK-GI-NEXT: ldr x30, [sp, #16] // 8-byte Folded Reload
; CHECK-GI-NEXT: csel w8, w8, w9, lt
; CHECK-GI-NEXT: fmov s0, w8
; CHECK-GI-NEXT: // kill: def $h0 killed $h0 killed $s0
; CHECK-GI-NEXT: ldp d9, d8, [sp], #32 // 16-byte Folded Reload
; CHECK-GI-NEXT: ret
entry:
%c = fcmp olt fp128 %a, %b
%s = select i1 %c, half %d, half %e
ret half %s
}
define double @f64_double(double %a, double %b, double %d, double %e) {
; CHECK-LABEL: f64_double:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: fcmp d0, d1
; CHECK-NEXT: fcsel d0, d2, d3, mi
; CHECK-NEXT: ret
entry:
%c = fcmp olt double %a, %b
%s = select i1 %c, double %d, double %e
ret double %s
}
define i32 @f64_i32(double %a, double %b, i32 %d, i32 %e) {
; CHECK-LABEL: f64_i32:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: fcmp d0, d1
; CHECK-NEXT: csel w0, w0, w1, mi
; CHECK-NEXT: ret
entry:
%c = fcmp olt double %a, %b
%s = select i1 %c, i32 %d, i32 %e
ret i32 %s
}
define float @f32_float(float %a, float %b, float %d, float %e) {
; CHECK-LABEL: f32_float:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: fcmp s0, s1
; CHECK-NEXT: fcsel s0, s2, s3, mi
; CHECK-NEXT: ret
entry:
%c = fcmp olt float %a, %b
%s = select i1 %c, float %d, float %e
ret float %s
}
define i32 @f32_i32(float %a, float %b, i32 %d, i32 %e) {
; CHECK-LABEL: f32_i32:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: fcmp s0, s1
; CHECK-NEXT: csel w0, w0, w1, mi
; CHECK-NEXT: ret
entry:
%c = fcmp olt float %a, %b
%s = select i1 %c, i32 %d, i32 %e
ret i32 %s
}
define half @f16_half(half %a, half %b, half %d, half %e) {
; CHECK-SD-NOFP16-LABEL: f16_half:
; CHECK-SD-NOFP16: // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT: fcvt s1, h1
; CHECK-SD-NOFP16-NEXT: fcvt s0, h0
; CHECK-SD-NOFP16-NEXT: // kill: def $h3 killed $h3 def $s3
; CHECK-SD-NOFP16-NEXT: // kill: def $h2 killed $h2 def $s2
; CHECK-SD-NOFP16-NEXT: fcmp s0, s1
; CHECK-SD-NOFP16-NEXT: fcsel s0, s2, s3, mi
; CHECK-SD-NOFP16-NEXT: // kill: def $h0 killed $h0 killed $s0
; CHECK-SD-NOFP16-NEXT: ret
;
; CHECK-SD-FP16-LABEL: f16_half:
; CHECK-SD-FP16: // %bb.0: // %entry
; CHECK-SD-FP16-NEXT: fcmp h0, h1
; CHECK-SD-FP16-NEXT: fcsel h0, h2, h3, mi
; CHECK-SD-FP16-NEXT: ret
;
; CHECK-GI-NOFP16-LABEL: f16_half:
; CHECK-GI-NOFP16: // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT: fcvt s0, h0
; CHECK-GI-NOFP16-NEXT: fcvt s1, h1
; CHECK-GI-NOFP16-NEXT: // kill: def $h2 killed $h2 def $s2
; CHECK-GI-NOFP16-NEXT: // kill: def $h3 killed $h3 def $s3
; CHECK-GI-NOFP16-NEXT: fmov w8, s2
; CHECK-GI-NOFP16-NEXT: fmov w9, s3
; CHECK-GI-NOFP16-NEXT: fcmp s0, s1
; CHECK-GI-NOFP16-NEXT: csel w8, w8, w9, mi
; CHECK-GI-NOFP16-NEXT: fmov s0, w8
; CHECK-GI-NOFP16-NEXT: // kill: def $h0 killed $h0 killed $s0
; CHECK-GI-NOFP16-NEXT: ret
;
; CHECK-GI-FP16-LABEL: f16_half:
; CHECK-GI-FP16: // %bb.0: // %entry
; CHECK-GI-FP16-NEXT: // kill: def $h2 killed $h2 def $s2
; CHECK-GI-FP16-NEXT: // kill: def $h3 killed $h3 def $s3
; CHECK-GI-FP16-NEXT: fcmp h0, h1
; CHECK-GI-FP16-NEXT: fmov w8, s2
; CHECK-GI-FP16-NEXT: fmov w9, s3
; CHECK-GI-FP16-NEXT: csel w8, w8, w9, mi
; CHECK-GI-FP16-NEXT: fmov s0, w8
; CHECK-GI-FP16-NEXT: // kill: def $h0 killed $h0 killed $s0
; CHECK-GI-FP16-NEXT: ret
entry:
%c = fcmp olt half %a, %b
%s = select i1 %c, half %d, half %e
ret half %s
}
define i32 @f16_i32(half %a, half %b, i32 %d, i32 %e) {
; CHECK-SD-NOFP16-LABEL: f16_i32:
; CHECK-SD-NOFP16: // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT: fcvt s1, h1
; CHECK-SD-NOFP16-NEXT: fcvt s0, h0
; CHECK-SD-NOFP16-NEXT: fcmp s0, s1
; CHECK-SD-NOFP16-NEXT: csel w0, w0, w1, mi
; CHECK-SD-NOFP16-NEXT: ret
;
; CHECK-SD-FP16-LABEL: f16_i32:
; CHECK-SD-FP16: // %bb.0: // %entry
; CHECK-SD-FP16-NEXT: fcmp h0, h1
; CHECK-SD-FP16-NEXT: csel w0, w0, w1, mi
; CHECK-SD-FP16-NEXT: ret
;
; CHECK-GI-NOFP16-LABEL: f16_i32:
; CHECK-GI-NOFP16: // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT: fcvt s0, h0
; CHECK-GI-NOFP16-NEXT: fcvt s1, h1
; CHECK-GI-NOFP16-NEXT: fcmp s0, s1
; CHECK-GI-NOFP16-NEXT: csel w0, w0, w1, mi
; CHECK-GI-NOFP16-NEXT: ret
;
; CHECK-GI-FP16-LABEL: f16_i32:
; CHECK-GI-FP16: // %bb.0: // %entry
; CHECK-GI-FP16-NEXT: fcmp h0, h1
; CHECK-GI-FP16-NEXT: csel w0, w0, w1, mi
; CHECK-GI-FP16-NEXT: ret
entry:
%c = fcmp olt half %a, %b
%s = select i1 %c, i32 %d, i32 %e
ret i32 %s
}
define <2 x fp128> @v2f128_fp128(<2 x fp128> %a, <2 x fp128> %b, <2 x fp128> %d, <2 x fp128> %e) {
; CHECK-SD-LABEL: v2f128_fp128:
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: sub sp, sp, #112
; CHECK-SD-NEXT: str x30, [sp, #96] // 8-byte Folded Spill
; CHECK-SD-NEXT: .cfi_def_cfa_offset 112
; CHECK-SD-NEXT: .cfi_offset w30, -16
; CHECK-SD-NEXT: stp q4, q5, [sp] // 32-byte Folded Spill
; CHECK-SD-NEXT: stp q1, q3, [sp, #32] // 32-byte Folded Spill
; CHECK-SD-NEXT: mov v1.16b, v2.16b
; CHECK-SD-NEXT: stp q7, q6, [sp, #64] // 32-byte Folded Spill
; CHECK-SD-NEXT: bl __lttf2
; CHECK-SD-NEXT: cmp w0, #0
; CHECK-SD-NEXT: b.ge .LBB12_2
; CHECK-SD-NEXT: // %bb.1: // %entry
; CHECK-SD-NEXT: ldr q0, [sp] // 16-byte Folded Reload
; CHECK-SD-NEXT: str q0, [sp, #80] // 16-byte Folded Spill
; CHECK-SD-NEXT: .LBB12_2: // %entry
; CHECK-SD-NEXT: ldp q0, q1, [sp, #32] // 32-byte Folded Reload
; CHECK-SD-NEXT: bl __lttf2
; CHECK-SD-NEXT: ldr q1, [sp, #64] // 16-byte Folded Reload
; CHECK-SD-NEXT: cmp w0, #0
; CHECK-SD-NEXT: b.ge .LBB12_4
; CHECK-SD-NEXT: // %bb.3: // %entry
; CHECK-SD-NEXT: ldr q1, [sp, #16] // 16-byte Folded Reload
; CHECK-SD-NEXT: .LBB12_4: // %entry
; CHECK-SD-NEXT: ldr q0, [sp, #80] // 16-byte Folded Reload
; CHECK-SD-NEXT: ldr x30, [sp, #96] // 8-byte Folded Reload
; CHECK-SD-NEXT: add sp, sp, #112
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: v2f128_fp128:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: sub sp, sp, #112
; CHECK-GI-NEXT: stp x30, x19, [sp, #96] // 16-byte Folded Spill
; CHECK-GI-NEXT: .cfi_def_cfa_offset 112
; CHECK-GI-NEXT: .cfi_offset w19, -8
; CHECK-GI-NEXT: .cfi_offset w30, -16
; CHECK-GI-NEXT: stp q3, q1, [sp] // 32-byte Folded Spill
; CHECK-GI-NEXT: mov v1.16b, v2.16b
; CHECK-GI-NEXT: stp q6, q4, [sp, #32] // 32-byte Folded Spill
; CHECK-GI-NEXT: stp q7, q5, [sp, #64] // 32-byte Folded Spill
; CHECK-GI-NEXT: bl __lttf2
; CHECK-GI-NEXT: ldp q1, q0, [sp] // 32-byte Folded Reload
; CHECK-GI-NEXT: mov w19, w0
; CHECK-GI-NEXT: bl __lttf2
; CHECK-GI-NEXT: ldp q3, q2, [sp, #32] // 32-byte Folded Reload
; CHECK-GI-NEXT: cmp w19, #0
; CHECK-GI-NEXT: ldp x30, x19, [sp, #96] // 16-byte Folded Reload
; CHECK-GI-NEXT: mov d0, v2.d[1]
; CHECK-GI-NEXT: mov d1, v3.d[1]
; CHECK-GI-NEXT: fcsel d2, d2, d3, lt
; CHECK-GI-NEXT: fmov x8, d2
; CHECK-GI-NEXT: fcsel d3, d0, d1, lt
; CHECK-GI-NEXT: ldp q5, q0, [sp, #64] // 32-byte Folded Reload
; CHECK-GI-NEXT: cmp w0, #0
; CHECK-GI-NEXT: mov d1, v0.d[1]
; CHECK-GI-NEXT: mov d4, v5.d[1]
; CHECK-GI-NEXT: fcsel d0, d0, d5, lt
; CHECK-GI-NEXT: fmov x9, d0
; CHECK-GI-NEXT: mov v0.d[0], x8
; CHECK-GI-NEXT: fmov x8, d3
; CHECK-GI-NEXT: fcsel d2, d1, d4, lt
; CHECK-GI-NEXT: mov v1.d[0], x9
; CHECK-GI-NEXT: fmov x9, d2
; CHECK-GI-NEXT: mov v0.d[1], x8
; CHECK-GI-NEXT: mov v1.d[1], x9
; CHECK-GI-NEXT: add sp, sp, #112
; CHECK-GI-NEXT: ret
entry:
%c = fcmp olt <2 x fp128> %a, %b
%s = select <2 x i1> %c, <2 x fp128> %d, <2 x fp128> %e
ret <2 x fp128> %s
}
define <3 x fp128> @v3f128_fp128(<3 x fp128> %a, <3 x fp128> %b, <3 x fp128> %d, <3 x fp128> %e) {
; CHECK-SD-LABEL: v3f128_fp128:
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: sub sp, sp, #112
; CHECK-SD-NEXT: str x30, [sp, #96] // 8-byte Folded Spill
; CHECK-SD-NEXT: .cfi_def_cfa_offset 112
; CHECK-SD-NEXT: .cfi_offset w30, -16
; CHECK-SD-NEXT: stp q1, q4, [sp] // 32-byte Folded Spill
; CHECK-SD-NEXT: mov v1.16b, v3.16b
; CHECK-SD-NEXT: stp q2, q5, [sp, #32] // 32-byte Folded Spill
; CHECK-SD-NEXT: stp q6, q7, [sp, #64] // 32-byte Folded Spill
; CHECK-SD-NEXT: bl __lttf2
; CHECK-SD-NEXT: cmp w0, #0
; CHECK-SD-NEXT: b.lt .LBB13_2
; CHECK-SD-NEXT: // %bb.1:
; CHECK-SD-NEXT: ldr q0, [sp, #128]
; CHECK-SD-NEXT: str q0, [sp, #64] // 16-byte Folded Spill
; CHECK-SD-NEXT: .LBB13_2: // %entry
; CHECK-SD-NEXT: ldp q0, q1, [sp] // 32-byte Folded Reload
; CHECK-SD-NEXT: bl __lttf2
; CHECK-SD-NEXT: cmp w0, #0
; CHECK-SD-NEXT: b.lt .LBB13_4
; CHECK-SD-NEXT: // %bb.3:
; CHECK-SD-NEXT: ldr q0, [sp, #144]
; CHECK-SD-NEXT: str q0, [sp, #80] // 16-byte Folded Spill
; CHECK-SD-NEXT: .LBB13_4: // %entry
; CHECK-SD-NEXT: ldp q0, q1, [sp, #32] // 32-byte Folded Reload
; CHECK-SD-NEXT: bl __lttf2
; CHECK-SD-NEXT: add x8, sp, #160
; CHECK-SD-NEXT: cmp w0, #0
; CHECK-SD-NEXT: add x9, sp, #112
; CHECK-SD-NEXT: csel x8, x9, x8, lt
; CHECK-SD-NEXT: ldp q0, q1, [sp, #64] // 32-byte Folded Reload
; CHECK-SD-NEXT: ldr q2, [x8]
; CHECK-SD-NEXT: ldr x30, [sp, #96] // 8-byte Folded Reload
; CHECK-SD-NEXT: add sp, sp, #112
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: v3f128_fp128:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: sub sp, sp, #192
; CHECK-GI-NEXT: str x30, [sp, #160] // 8-byte Folded Spill
; CHECK-GI-NEXT: stp x20, x19, [sp, #176] // 16-byte Folded Spill
; CHECK-GI-NEXT: .cfi_def_cfa_offset 192
; CHECK-GI-NEXT: .cfi_offset w19, -8
; CHECK-GI-NEXT: .cfi_offset w20, -16
; CHECK-GI-NEXT: .cfi_offset w30, -32
; CHECK-GI-NEXT: stp q4, q1, [sp] // 32-byte Folded Spill
; CHECK-GI-NEXT: mov v1.16b, v3.16b
; CHECK-GI-NEXT: stp q5, q2, [sp, #32] // 32-byte Folded Spill
; CHECK-GI-NEXT: ldr q2, [sp, #192]
; CHECK-GI-NEXT: str q2, [sp, #144] // 16-byte Folded Spill
; CHECK-GI-NEXT: ldr q2, [sp, #208]
; CHECK-GI-NEXT: stp q2, q6, [sp, #64] // 32-byte Folded Spill
; CHECK-GI-NEXT: ldr q2, [sp, #224]
; CHECK-GI-NEXT: stp q7, q2, [sp, #96] // 32-byte Folded Spill
; CHECK-GI-NEXT: ldr q2, [sp, #240]
; CHECK-GI-NEXT: str q2, [sp, #128] // 16-byte Folded Spill
; CHECK-GI-NEXT: bl __lttf2
; CHECK-GI-NEXT: ldp q1, q0, [sp] // 32-byte Folded Reload
; CHECK-GI-NEXT: mov w19, w0
; CHECK-GI-NEXT: bl __lttf2
; CHECK-GI-NEXT: ldp q1, q0, [sp, #32] // 32-byte Folded Reload
; CHECK-GI-NEXT: mov w20, w0
; CHECK-GI-NEXT: bl __lttf2
; CHECK-GI-NEXT: ldp q5, q4, [sp, #64] // 32-byte Folded Reload
; CHECK-GI-NEXT: cmp w19, #0
; CHECK-GI-NEXT: ldp q7, q6, [sp, #96] // 32-byte Folded Reload
; CHECK-GI-NEXT: ldr x30, [sp, #160] // 8-byte Folded Reload
; CHECK-GI-NEXT: mov d0, v4.d[1]
; CHECK-GI-NEXT: mov d1, v5.d[1]
; CHECK-GI-NEXT: fcsel d4, d4, d5, lt
; CHECK-GI-NEXT: mov d2, v7.d[1]
; CHECK-GI-NEXT: mov d3, v6.d[1]
; CHECK-GI-NEXT: fmov x8, d4
; CHECK-GI-NEXT: fcsel d5, d0, d1, lt
; CHECK-GI-NEXT: cmp w20, #0
; CHECK-GI-NEXT: fcsel d1, d7, d6, lt
; CHECK-GI-NEXT: ldp q7, q0, [sp, #128] // 32-byte Folded Reload
; CHECK-GI-NEXT: fcsel d3, d2, d3, lt
; CHECK-GI-NEXT: cmp w0, #0
; CHECK-GI-NEXT: ldp x20, x19, [sp, #176] // 16-byte Folded Reload
; CHECK-GI-NEXT: mov d2, v0.d[1]
; CHECK-GI-NEXT: mov d6, v7.d[1]
; CHECK-GI-NEXT: fcsel d7, d0, d7, lt
; CHECK-GI-NEXT: mov v0.d[0], x8
; CHECK-GI-NEXT: fmov x8, d1
; CHECK-GI-NEXT: fmov x9, d7
; CHECK-GI-NEXT: fcsel d4, d2, d6, lt
; CHECK-GI-NEXT: mov v1.d[0], x8
; CHECK-GI-NEXT: fmov x8, d5
; CHECK-GI-NEXT: mov v2.d[0], x9
; CHECK-GI-NEXT: fmov x9, d3
; CHECK-GI-NEXT: fmov x10, d4
; CHECK-GI-NEXT: mov v0.d[1], x8
; CHECK-GI-NEXT: mov v1.d[1], x9
; CHECK-GI-NEXT: mov v2.d[1], x10
; CHECK-GI-NEXT: add sp, sp, #192
; CHECK-GI-NEXT: ret
entry:
%c = fcmp olt <3 x fp128> %a, %b
%s = select <3 x i1> %c, <3 x fp128> %d, <3 x fp128> %e
ret <3 x fp128> %s
}
define <2 x double> @v2f128_double(<2 x fp128> %a, <2 x fp128> %b, <2 x double> %d, <2 x double> %e) {
; CHECK-SD-LABEL: v2f128_double:
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: sub sp, sp, #96
; CHECK-SD-NEXT: str x30, [sp, #80] // 8-byte Folded Spill
; CHECK-SD-NEXT: .cfi_def_cfa_offset 96
; CHECK-SD-NEXT: .cfi_offset w30, -16
; CHECK-SD-NEXT: str q0, [sp] // 16-byte Folded Spill
; CHECK-SD-NEXT: mov v0.16b, v1.16b
; CHECK-SD-NEXT: mov v1.16b, v3.16b
; CHECK-SD-NEXT: stp q4, q5, [sp, #48] // 32-byte Folded Spill
; CHECK-SD-NEXT: str q2, [sp, #32] // 16-byte Folded Spill
; CHECK-SD-NEXT: bl __lttf2
; CHECK-SD-NEXT: cmp w0, #0
; CHECK-SD-NEXT: ldr q1, [sp, #32] // 16-byte Folded Reload
; CHECK-SD-NEXT: cset w8, lt
; CHECK-SD-NEXT: sbfx x8, x8, #0, #1
; CHECK-SD-NEXT: fmov d0, x8
; CHECK-SD-NEXT: str q0, [sp, #16] // 16-byte Folded Spill
; CHECK-SD-NEXT: ldr q0, [sp] // 16-byte Folded Reload
; CHECK-SD-NEXT: bl __lttf2
; CHECK-SD-NEXT: cmp w0, #0
; CHECK-SD-NEXT: ldr q1, [sp, #16] // 16-byte Folded Reload
; CHECK-SD-NEXT: ldr x30, [sp, #80] // 8-byte Folded Reload
; CHECK-SD-NEXT: cset w8, lt
; CHECK-SD-NEXT: sbfx x8, x8, #0, #1
; CHECK-SD-NEXT: fmov d0, x8
; CHECK-SD-NEXT: mov v0.d[1], v1.d[0]
; CHECK-SD-NEXT: ldp q2, q1, [sp, #48] // 32-byte Folded Reload
; CHECK-SD-NEXT: bsl v0.16b, v2.16b, v1.16b
; CHECK-SD-NEXT: add sp, sp, #96
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: v2f128_double:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: sub sp, sp, #80
; CHECK-GI-NEXT: stp x30, x19, [sp, #64] // 16-byte Folded Spill
; CHECK-GI-NEXT: .cfi_def_cfa_offset 80
; CHECK-GI-NEXT: .cfi_offset w19, -8
; CHECK-GI-NEXT: .cfi_offset w30, -16
; CHECK-GI-NEXT: stp q3, q1, [sp] // 32-byte Folded Spill
; CHECK-GI-NEXT: mov v1.16b, v2.16b
; CHECK-GI-NEXT: stp q5, q4, [sp, #32] // 32-byte Folded Spill
; CHECK-GI-NEXT: bl __lttf2
; CHECK-GI-NEXT: ldp q1, q0, [sp] // 32-byte Folded Reload
; CHECK-GI-NEXT: cmp w0, #0
; CHECK-GI-NEXT: cset w19, lt
; CHECK-GI-NEXT: bl __lttf2
; CHECK-GI-NEXT: mov v0.d[0], x19
; CHECK-GI-NEXT: cmp w0, #0
; CHECK-GI-NEXT: cset w8, lt
; CHECK-GI-NEXT: ldp q2, q1, [sp, #32] // 32-byte Folded Reload
; CHECK-GI-NEXT: ldp x30, x19, [sp, #64] // 16-byte Folded Reload
; CHECK-GI-NEXT: mov v0.d[1], x8
; CHECK-GI-NEXT: shl v0.2d, v0.2d, #63
; CHECK-GI-NEXT: sshr v0.2d, v0.2d, #63
; CHECK-GI-NEXT: bsl v0.16b, v1.16b, v2.16b
; CHECK-GI-NEXT: add sp, sp, #80
; CHECK-GI-NEXT: ret
entry:
%c = fcmp olt <2 x fp128> %a, %b
%s = select <2 x i1> %c, <2 x double> %d, <2 x double> %e
ret <2 x double> %s
}
define <3 x double> @v3f128_double(<3 x fp128> %a, <3 x fp128> %b, <3 x double> %d, <3 x double> %e) {
; CHECK-SD-LABEL: v3f128_double:
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: sub sp, sp, #160
; CHECK-SD-NEXT: str x30, [sp, #144] // 8-byte Folded Spill
; CHECK-SD-NEXT: .cfi_def_cfa_offset 160
; CHECK-SD-NEXT: .cfi_offset w30, -16
; CHECK-SD-NEXT: stp q2, q5, [sp, #112] // 32-byte Folded Spill
; CHECK-SD-NEXT: // kill: def $d6 killed $d6 def $q6
; CHECK-SD-NEXT: // kill: def $d7 killed $d7 def $q7
; CHECK-SD-NEXT: ldr d5, [sp, #184]
; CHECK-SD-NEXT: str q3, [sp, #64] // 16-byte Folded Spill
; CHECK-SD-NEXT: ldp d3, d2, [sp, #168]
; CHECK-SD-NEXT: mov v6.d[1], v7.d[0]
; CHECK-SD-NEXT: str q0, [sp, #16] // 16-byte Folded Spill
; CHECK-SD-NEXT: mov v0.16b, v1.16b
; CHECK-SD-NEXT: mov v1.16b, v4.16b
; CHECK-SD-NEXT: str q5, [sp, #96] // 16-byte Folded Spill
; CHECK-SD-NEXT: ldr d5, [sp, #160]
; CHECK-SD-NEXT: mov v3.d[1], v2.d[0]
; CHECK-SD-NEXT: str q5, [sp, #80] // 16-byte Folded Spill
; CHECK-SD-NEXT: stp q6, q3, [sp, #32] // 32-byte Folded Spill
; CHECK-SD-NEXT: bl __lttf2
; CHECK-SD-NEXT: cmp w0, #0
; CHECK-SD-NEXT: ldr q1, [sp, #64] // 16-byte Folded Reload
; CHECK-SD-NEXT: cset w8, lt
; CHECK-SD-NEXT: sbfx x8, x8, #0, #1
; CHECK-SD-NEXT: fmov d0, x8
; CHECK-SD-NEXT: str q0, [sp] // 16-byte Folded Spill
; CHECK-SD-NEXT: ldr q0, [sp, #16] // 16-byte Folded Reload
; CHECK-SD-NEXT: bl __lttf2
; CHECK-SD-NEXT: cmp w0, #0
; CHECK-SD-NEXT: ldr q0, [sp] // 16-byte Folded Reload
; CHECK-SD-NEXT: cset w8, lt
; CHECK-SD-NEXT: sbfx x8, x8, #0, #1
; CHECK-SD-NEXT: fmov d1, x8
; CHECK-SD-NEXT: mov v1.d[1], v0.d[0]
; CHECK-SD-NEXT: str q1, [sp, #64] // 16-byte Folded Spill
; CHECK-SD-NEXT: ldp q0, q1, [sp, #112] // 32-byte Folded Reload
; CHECK-SD-NEXT: bl __lttf2
; CHECK-SD-NEXT: ldp q1, q0, [sp, #32] // 32-byte Folded Reload
; CHECK-SD-NEXT: cmp w0, #0
; CHECK-SD-NEXT: ldp q2, q4, [sp, #64] // 32-byte Folded Reload
; CHECK-SD-NEXT: cset w8, lt
; CHECK-SD-NEXT: sbfx x8, x8, #0, #1
; CHECK-SD-NEXT: ldr q3, [sp, #96] // 16-byte Folded Reload
; CHECK-SD-NEXT: ldr x30, [sp, #144] // 8-byte Folded Reload
; CHECK-SD-NEXT: bit v0.16b, v1.16b, v2.16b
; CHECK-SD-NEXT: fmov d2, x8
; CHECK-SD-NEXT: bsl v2.16b, v4.16b, v3.16b
; CHECK-SD-NEXT: ext v1.16b, v0.16b, v0.16b, #8
; CHECK-SD-NEXT: // kill: def $d0 killed $d0 killed $q0
; CHECK-SD-NEXT: // kill: def $d2 killed $d2 killed $q2
; CHECK-SD-NEXT: // kill: def $d1 killed $d1 killed $q1
; CHECK-SD-NEXT: add sp, sp, #160
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: v3f128_double:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: sub sp, sp, #176
; CHECK-GI-NEXT: str x30, [sp, #128] // 8-byte Folded Spill
; CHECK-GI-NEXT: stp x22, x21, [sp, #144] // 16-byte Folded Spill
; CHECK-GI-NEXT: stp x20, x19, [sp, #160] // 16-byte Folded Spill
; CHECK-GI-NEXT: .cfi_def_cfa_offset 176
; CHECK-GI-NEXT: .cfi_offset w19, -8
; CHECK-GI-NEXT: .cfi_offset w20, -16
; CHECK-GI-NEXT: .cfi_offset w21, -24
; CHECK-GI-NEXT: .cfi_offset w22, -32
; CHECK-GI-NEXT: .cfi_offset w30, -48
; CHECK-GI-NEXT: stp q4, q1, [sp] // 32-byte Folded Spill
; CHECK-GI-NEXT: mov v1.16b, v3.16b
; CHECK-GI-NEXT: ldr x19, [sp, #176]
; CHECK-GI-NEXT: stp q5, q2, [sp, #32] // 32-byte Folded Spill
; CHECK-GI-NEXT: ldr d2, [sp, #184]
; CHECK-GI-NEXT: ldr x20, [sp, #200]
; CHECK-GI-NEXT: // kill: def $d6 killed $d6 def $q6
; CHECK-GI-NEXT: // kill: def $d7 killed $d7 def $q7
; CHECK-GI-NEXT: str q7, [sp, #64] // 16-byte Folded Spill
; CHECK-GI-NEXT: str q2, [sp, #112] // 16-byte Folded Spill
; CHECK-GI-NEXT: ldr d2, [sp, #192]
; CHECK-GI-NEXT: stp q6, q2, [sp, #80] // 32-byte Folded Spill
; CHECK-GI-NEXT: bl __lttf2
; CHECK-GI-NEXT: ldp q1, q0, [sp] // 32-byte Folded Reload
; CHECK-GI-NEXT: cmp w0, #0
; CHECK-GI-NEXT: cset w21, lt
; CHECK-GI-NEXT: bl __lttf2
; CHECK-GI-NEXT: ldp q1, q0, [sp, #32] // 32-byte Folded Reload
; CHECK-GI-NEXT: cmp w0, #0
; CHECK-GI-NEXT: cset w22, lt
; CHECK-GI-NEXT: bl __lttf2
; CHECK-GI-NEXT: sbfx x8, x21, #0, #1
; CHECK-GI-NEXT: ldp q3, q2, [sp, #64] // 32-byte Folded Reload
; CHECK-GI-NEXT: cmp w0, #0
; CHECK-GI-NEXT: ldr x30, [sp, #128] // 8-byte Folded Reload
; CHECK-GI-NEXT: mov v0.d[0], x8
; CHECK-GI-NEXT: mov v1.d[0], x8
; CHECK-GI-NEXT: sbfx x8, x22, #0, #1
; CHECK-GI-NEXT: mov v2.d[1], v3.d[0]
; CHECK-GI-NEXT: ldp q4, q3, [sp, #96] // 32-byte Folded Reload
; CHECK-GI-NEXT: ldp x22, x21, [sp, #144] // 16-byte Folded Reload
; CHECK-GI-NEXT: mov v0.d[1], x8
; CHECK-GI-NEXT: mov v1.d[1], x8
; CHECK-GI-NEXT: mov v3.d[1], v4.d[0]
; CHECK-GI-NEXT: cset w8, lt
; CHECK-GI-NEXT: sbfx x8, x8, #0, #1
; CHECK-GI-NEXT: and v1.16b, v2.16b, v1.16b
; CHECK-GI-NEXT: bic v0.16b, v3.16b, v0.16b
; CHECK-GI-NEXT: and x9, x19, x8
; CHECK-GI-NEXT: bic x8, x20, x8
; CHECK-GI-NEXT: ldp x20, x19, [sp, #160] // 16-byte Folded Reload
; CHECK-GI-NEXT: orr x8, x9, x8
; CHECK-GI-NEXT: orr v0.16b, v1.16b, v0.16b
; CHECK-GI-NEXT: fmov d2, x8
; CHECK-GI-NEXT: mov d1, v0.d[1]
; CHECK-GI-NEXT: // kill: def $d0 killed $d0 killed $q0
; CHECK-GI-NEXT: add sp, sp, #176
; CHECK-GI-NEXT: ret
entry:
%c = fcmp olt <3 x fp128> %a, %b
%s = select <3 x i1> %c, <3 x double> %d, <3 x double> %e
ret <3 x double> %s
}
define <2 x double> @v2f64_double(<2 x double> %a, <2 x double> %b, <2 x double> %d, <2 x double> %e) {
; CHECK-LABEL: v2f64_double:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: fcmgt v0.2d, v1.2d, v0.2d
; CHECK-NEXT: bsl v0.16b, v2.16b, v3.16b
; CHECK-NEXT: ret
entry:
%c = fcmp olt <2 x double> %a, %b
%s = select <2 x i1> %c, <2 x double> %d, <2 x double> %e
ret <2 x double> %s
}
define <3 x double> @v3f64_double(<3 x double> %a, <3 x double> %b, <3 x double> %d, <3 x double> %e) {
; CHECK-SD-LABEL: v3f64_double:
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: // kill: def $d3 killed $d3 def $q3
; CHECK-SD-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-SD-NEXT: // kill: def $d4 killed $d4 def $q4
; CHECK-SD-NEXT: // kill: def $d1 killed $d1 def $q1
; CHECK-SD-NEXT: // kill: def $d6 killed $d6 def $q6
; CHECK-SD-NEXT: // kill: def $d7 killed $d7 def $q7
; CHECK-SD-NEXT: // kill: def $d2 killed $d2 def $q2
; CHECK-SD-NEXT: // kill: def $d5 killed $d5 def $q5
; CHECK-SD-NEXT: ldr d16, [sp, #24]
; CHECK-SD-NEXT: ldr d17, [sp]
; CHECK-SD-NEXT: mov v3.d[1], v4.d[0]
; CHECK-SD-NEXT: mov v0.d[1], v1.d[0]
; CHECK-SD-NEXT: mov v6.d[1], v7.d[0]
; CHECK-SD-NEXT: ldp d1, d4, [sp, #8]
; CHECK-SD-NEXT: fcmgt v2.2d, v5.2d, v2.2d
; CHECK-SD-NEXT: mov v1.d[1], v4.d[0]
; CHECK-SD-NEXT: fcmgt v0.2d, v3.2d, v0.2d
; CHECK-SD-NEXT: bsl v2.16b, v17.16b, v16.16b
; CHECK-SD-NEXT: // kill: def $d2 killed $d2 killed $q2
; CHECK-SD-NEXT: bsl v0.16b, v6.16b, v1.16b
; CHECK-SD-NEXT: ext v1.16b, v0.16b, v0.16b, #8
; CHECK-SD-NEXT: // kill: def $d0 killed $d0 killed $q0
; CHECK-SD-NEXT: // kill: def $d1 killed $d1 killed $q1
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: v3f64_double:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-GI-NEXT: // kill: def $d3 killed $d3 def $q3
; CHECK-GI-NEXT: // kill: def $d1 killed $d1 def $q1
; CHECK-GI-NEXT: // kill: def $d4 killed $d4 def $q4
; CHECK-GI-NEXT: // kill: def $d6 killed $d6 def $q6
; CHECK-GI-NEXT: // kill: def $d7 killed $d7 def $q7
; CHECK-GI-NEXT: fcmp d2, d5
; CHECK-GI-NEXT: ldr x8, [sp]
; CHECK-GI-NEXT: ldr x10, [sp, #24]
; CHECK-GI-NEXT: mov v0.d[1], v1.d[0]
; CHECK-GI-NEXT: mov v3.d[1], v4.d[0]
; CHECK-GI-NEXT: mov v6.d[1], v7.d[0]
; CHECK-GI-NEXT: ldp d1, d4, [sp, #8]
; CHECK-GI-NEXT: cset w9, mi
; CHECK-GI-NEXT: sbfx x9, x9, #0, #1
; CHECK-GI-NEXT: fcmgt v0.2d, v3.2d, v0.2d
; CHECK-GI-NEXT: mov v1.d[1], v4.d[0]
; CHECK-GI-NEXT: and x8, x8, x9
; CHECK-GI-NEXT: bic x9, x10, x9
; CHECK-GI-NEXT: orr x8, x8, x9
; CHECK-GI-NEXT: fmov d2, x8
; CHECK-GI-NEXT: bsl v0.16b, v6.16b, v1.16b
; CHECK-GI-NEXT: mov d1, v0.d[1]
; CHECK-GI-NEXT: // kill: def $d0 killed $d0 killed $q0
; CHECK-GI-NEXT: ret
entry:
%c = fcmp olt <3 x double> %a, %b
%s = select <3 x i1> %c, <3 x double> %d, <3 x double> %e
ret <3 x double> %s
}
define <4 x double> @v4f64_double(<4 x double> %a, <4 x double> %b, <4 x double> %d, <4 x double> %e) {
; CHECK-SD-LABEL: v4f64_double:
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: fcmgt v1.2d, v3.2d, v1.2d
; CHECK-SD-NEXT: fcmgt v0.2d, v2.2d, v0.2d
; CHECK-SD-NEXT: bsl v1.16b, v5.16b, v7.16b
; CHECK-SD-NEXT: bsl v0.16b, v4.16b, v6.16b
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: v4f64_double:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: fcmgt v0.2d, v2.2d, v0.2d
; CHECK-GI-NEXT: fcmgt v1.2d, v3.2d, v1.2d
; CHECK-GI-NEXT: bsl v0.16b, v4.16b, v6.16b
; CHECK-GI-NEXT: bsl v1.16b, v5.16b, v7.16b
; CHECK-GI-NEXT: ret
entry:
%c = fcmp olt <4 x double> %a, %b
%s = select <4 x i1> %c, <4 x double> %d, <4 x double> %e
ret <4 x double> %s
}
define <2 x i32> @v2f64_i32(<2 x double> %a, <2 x double> %b, <2 x i32> %d, <2 x i32> %e) {
; CHECK-LABEL: v2f64_i32:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: fcmgt v0.2d, v1.2d, v0.2d
; CHECK-NEXT: xtn v0.2s, v0.2d
; CHECK-NEXT: bsl v0.8b, v2.8b, v3.8b
; CHECK-NEXT: ret
entry:
%c = fcmp olt <2 x double> %a, %b
%s = select <2 x i1> %c, <2 x i32> %d, <2 x i32> %e
ret <2 x i32> %s
}
define <3 x i32> @v3f64_i32(<3 x double> %a, <3 x double> %b, <3 x i32> %d, <3 x i32> %e) {
; CHECK-SD-LABEL: v3f64_i32:
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: // kill: def $d3 killed $d3 def $q3
; CHECK-SD-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-SD-NEXT: // kill: def $d4 killed $d4 def $q4
; CHECK-SD-NEXT: // kill: def $d1 killed $d1 def $q1
; CHECK-SD-NEXT: // kill: def $d5 killed $d5 def $q5
; CHECK-SD-NEXT: // kill: def $d2 killed $d2 def $q2
; CHECK-SD-NEXT: mov v0.d[1], v1.d[0]
; CHECK-SD-NEXT: mov v3.d[1], v4.d[0]
; CHECK-SD-NEXT: fcmgt v1.2d, v5.2d, v2.2d
; CHECK-SD-NEXT: fcmgt v0.2d, v3.2d, v0.2d
; CHECK-SD-NEXT: uzp1 v0.4s, v0.4s, v1.4s
; CHECK-SD-NEXT: bsl v0.16b, v6.16b, v7.16b
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: v3f64_i32:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-GI-NEXT: // kill: def $d1 killed $d1 def $q1
; CHECK-GI-NEXT: // kill: def $d3 killed $d3 def $q3
; CHECK-GI-NEXT: mov w8, #31 // =0x1f
; CHECK-GI-NEXT: // kill: def $d4 killed $d4 def $q4
; CHECK-GI-NEXT: fcmp d2, d5
; CHECK-GI-NEXT: mov v0.d[1], v1.d[0]
; CHECK-GI-NEXT: mov v3.d[1], v4.d[0]
; CHECK-GI-NEXT: mov v1.s[0], w8
; CHECK-GI-NEXT: cset w9, mi
; CHECK-GI-NEXT: mov v2.d[0], x9
; CHECK-GI-NEXT: mov w9, #-1 // =0xffffffff
; CHECK-GI-NEXT: fcmgt v0.2d, v3.2d, v0.2d
; CHECK-GI-NEXT: mov v1.s[1], w8
; CHECK-GI-NEXT: mov v3.s[0], w9
; CHECK-GI-NEXT: mov v1.s[2], w8
; CHECK-GI-NEXT: uzp1 v0.4s, v0.4s, v2.4s
; CHECK-GI-NEXT: mov v3.s[1], w9
; CHECK-GI-NEXT: ushl v0.4s, v0.4s, v1.4s
; CHECK-GI-NEXT: neg v1.4s, v1.4s
; CHECK-GI-NEXT: mov v3.s[2], w9
; CHECK-GI-NEXT: sshl v0.4s, v0.4s, v1.4s
; CHECK-GI-NEXT: eor v1.16b, v0.16b, v3.16b
; CHECK-GI-NEXT: and v0.16b, v6.16b, v0.16b
; CHECK-GI-NEXT: and v1.16b, v7.16b, v1.16b
; CHECK-GI-NEXT: orr v0.16b, v0.16b, v1.16b
; CHECK-GI-NEXT: ret
entry:
%c = fcmp olt <3 x double> %a, %b
%s = select <3 x i1> %c, <3 x i32> %d, <3 x i32> %e
ret <3 x i32> %s
}
define <4 x i32> @v4f64_i32(<4 x double> %a, <4 x double> %b, <4 x i32> %d, <4 x i32> %e) {
; CHECK-SD-LABEL: v4f64_i32:
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: fcmgt v1.2d, v3.2d, v1.2d
; CHECK-SD-NEXT: fcmgt v0.2d, v2.2d, v0.2d
; CHECK-SD-NEXT: uzp1 v0.4s, v0.4s, v1.4s
; CHECK-SD-NEXT: bsl v0.16b, v4.16b, v5.16b
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: v4f64_i32:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: fcmgt v0.2d, v2.2d, v0.2d
; CHECK-GI-NEXT: fcmgt v1.2d, v3.2d, v1.2d
; CHECK-GI-NEXT: uzp1 v0.4s, v0.4s, v1.4s
; CHECK-GI-NEXT: shl v0.4s, v0.4s, #31
; CHECK-GI-NEXT: sshr v0.4s, v0.4s, #31
; CHECK-GI-NEXT: bsl v0.16b, v4.16b, v5.16b
; CHECK-GI-NEXT: ret
entry:
%c = fcmp olt <4 x double> %a, %b
%s = select <4 x i1> %c, <4 x i32> %d, <4 x i32> %e
ret <4 x i32> %s
}
define <2 x float> @v2f32_float(<2 x float> %a, <2 x float> %b, <2 x float> %d, <2 x float> %e) {
; CHECK-LABEL: v2f32_float:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: fcmgt v0.2s, v1.2s, v0.2s
; CHECK-NEXT: bsl v0.8b, v2.8b, v3.8b
; CHECK-NEXT: ret
entry:
%c = fcmp olt <2 x float> %a, %b
%s = select <2 x i1> %c, <2 x float> %d, <2 x float> %e
ret <2 x float> %s
}
define <3 x float> @v3f32_float(<3 x float> %a, <3 x float> %b, <3 x float> %d, <3 x float> %e) {
; CHECK-SD-LABEL: v3f32_float:
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: fcmgt v0.4s, v1.4s, v0.4s
; CHECK-SD-NEXT: bsl v0.16b, v2.16b, v3.16b
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: v3f32_float:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: mov w8, #31 // =0x1f
; CHECK-GI-NEXT: mov w9, #-1 // =0xffffffff
; CHECK-GI-NEXT: fcmgt v0.4s, v1.4s, v0.4s
; CHECK-GI-NEXT: mov v4.s[0], w8
; CHECK-GI-NEXT: mov v5.s[0], w9
; CHECK-GI-NEXT: mov v4.s[1], w8
; CHECK-GI-NEXT: mov v5.s[1], w9
; CHECK-GI-NEXT: mov v4.s[2], w8
; CHECK-GI-NEXT: mov v5.s[2], w9
; CHECK-GI-NEXT: ushl v0.4s, v0.4s, v4.4s
; CHECK-GI-NEXT: neg v1.4s, v4.4s
; CHECK-GI-NEXT: sshl v0.4s, v0.4s, v1.4s
; CHECK-GI-NEXT: eor v1.16b, v0.16b, v5.16b
; CHECK-GI-NEXT: and v0.16b, v2.16b, v0.16b
; CHECK-GI-NEXT: and v1.16b, v3.16b, v1.16b
; CHECK-GI-NEXT: orr v0.16b, v0.16b, v1.16b
; CHECK-GI-NEXT: ret
entry:
%c = fcmp olt <3 x float> %a, %b
%s = select <3 x i1> %c, <3 x float> %d, <3 x float> %e
ret <3 x float> %s
}
define <4 x float> @v4f32_float(<4 x float> %a, <4 x float> %b, <4 x float> %d, <4 x float> %e) {
; CHECK-LABEL: v4f32_float:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: fcmgt v0.4s, v1.4s, v0.4s
; CHECK-NEXT: bsl v0.16b, v2.16b, v3.16b
; CHECK-NEXT: ret
entry:
%c = fcmp olt <4 x float> %a, %b
%s = select <4 x i1> %c, <4 x float> %d, <4 x float> %e
ret <4 x float> %s
}
define <8 x float> @v8f32_float(<8 x float> %a, <8 x float> %b, <8 x float> %d, <8 x float> %e) {
; CHECK-SD-LABEL: v8f32_float:
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: fcmgt v1.4s, v3.4s, v1.4s
; CHECK-SD-NEXT: fcmgt v0.4s, v2.4s, v0.4s
; CHECK-SD-NEXT: bsl v1.16b, v5.16b, v7.16b
; CHECK-SD-NEXT: bsl v0.16b, v4.16b, v6.16b
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: v8f32_float:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: fcmgt v0.4s, v2.4s, v0.4s
; CHECK-GI-NEXT: fcmgt v1.4s, v3.4s, v1.4s
; CHECK-GI-NEXT: bsl v0.16b, v4.16b, v6.16b
; CHECK-GI-NEXT: bsl v1.16b, v5.16b, v7.16b
; CHECK-GI-NEXT: ret
entry:
%c = fcmp olt <8 x float> %a, %b
%s = select <8 x i1> %c, <8 x float> %d, <8 x float> %e
ret <8 x float> %s
}
define <2 x i32> @v2f32_i32(<2 x float> %a, <2 x float> %b, <2 x i32> %d, <2 x i32> %e) {
; CHECK-LABEL: v2f32_i32:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: fcmgt v0.2s, v1.2s, v0.2s
; CHECK-NEXT: bsl v0.8b, v2.8b, v3.8b
; CHECK-NEXT: ret
entry:
%c = fcmp olt <2 x float> %a, %b
%s = select <2 x i1> %c, <2 x i32> %d, <2 x i32> %e
ret <2 x i32> %s
}
define <3 x i32> @v3f32_i32(<3 x float> %a, <3 x float> %b, <3 x i32> %d, <3 x i32> %e) {
; CHECK-SD-LABEL: v3f32_i32:
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: fcmgt v0.4s, v1.4s, v0.4s
; CHECK-SD-NEXT: bsl v0.16b, v2.16b, v3.16b
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: v3f32_i32:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: mov w8, #31 // =0x1f
; CHECK-GI-NEXT: mov w9, #-1 // =0xffffffff
; CHECK-GI-NEXT: fcmgt v0.4s, v1.4s, v0.4s
; CHECK-GI-NEXT: mov v4.s[0], w8
; CHECK-GI-NEXT: mov v5.s[0], w9
; CHECK-GI-NEXT: mov v4.s[1], w8
; CHECK-GI-NEXT: mov v5.s[1], w9
; CHECK-GI-NEXT: mov v4.s[2], w8
; CHECK-GI-NEXT: mov v5.s[2], w9
; CHECK-GI-NEXT: ushl v0.4s, v0.4s, v4.4s
; CHECK-GI-NEXT: neg v1.4s, v4.4s
; CHECK-GI-NEXT: sshl v0.4s, v0.4s, v1.4s
; CHECK-GI-NEXT: eor v1.16b, v0.16b, v5.16b
; CHECK-GI-NEXT: and v0.16b, v2.16b, v0.16b
; CHECK-GI-NEXT: and v1.16b, v3.16b, v1.16b
; CHECK-GI-NEXT: orr v0.16b, v0.16b, v1.16b
; CHECK-GI-NEXT: ret
entry:
%c = fcmp olt <3 x float> %a, %b
%s = select <3 x i1> %c, <3 x i32> %d, <3 x i32> %e
ret <3 x i32> %s
}
define <4 x i32> @v4f32_i32(<4 x float> %a, <4 x float> %b, <4 x i32> %d, <4 x i32> %e) {
; CHECK-LABEL: v4f32_i32:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: fcmgt v0.4s, v1.4s, v0.4s
; CHECK-NEXT: bsl v0.16b, v2.16b, v3.16b
; CHECK-NEXT: ret
entry:
%c = fcmp olt <4 x float> %a, %b
%s = select <4 x i1> %c, <4 x i32> %d, <4 x i32> %e
ret <4 x i32> %s
}
define <8 x i32> @v8f32_i32(<8 x float> %a, <8 x float> %b, <8 x i32> %d, <8 x i32> %e) {
; CHECK-SD-LABEL: v8f32_i32:
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: fcmgt v1.4s, v3.4s, v1.4s
; CHECK-SD-NEXT: fcmgt v0.4s, v2.4s, v0.4s
; CHECK-SD-NEXT: bsl v1.16b, v5.16b, v7.16b
; CHECK-SD-NEXT: bsl v0.16b, v4.16b, v6.16b
; CHECK-SD-NEXT: ret
;
; CHECK-GI-LABEL: v8f32_i32:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: fcmgt v0.4s, v2.4s, v0.4s
; CHECK-GI-NEXT: fcmgt v1.4s, v3.4s, v1.4s
; CHECK-GI-NEXT: bsl v0.16b, v4.16b, v6.16b
; CHECK-GI-NEXT: bsl v1.16b, v5.16b, v7.16b
; CHECK-GI-NEXT: ret
entry:
%c = fcmp olt <8 x float> %a, %b
%s = select <8 x i1> %c, <8 x i32> %d, <8 x i32> %e
ret <8 x i32> %s
}
define <7 x half> @v7f16_half(<7 x half> %a, <7 x half> %b, <7 x half> %d, <7 x half> %e) {
; CHECK-SD-NOFP16-LABEL: v7f16_half:
; CHECK-SD-NOFP16: // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT: mov h4, v1.h[1]
; CHECK-SD-NOFP16-NEXT: mov h5, v0.h[1]
; CHECK-SD-NOFP16-NEXT: fcvt s6, h1
; CHECK-SD-NOFP16-NEXT: fcvt s7, h0
; CHECK-SD-NOFP16-NEXT: mov h16, v1.h[2]
; CHECK-SD-NOFP16-NEXT: fcvt s4, h4
; CHECK-SD-NOFP16-NEXT: fcvt s5, h5
; CHECK-SD-NOFP16-NEXT: fcmp s5, s4
; CHECK-SD-NOFP16-NEXT: mov h4, v0.h[2]
; CHECK-SD-NOFP16-NEXT: mov h5, v1.h[3]
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s7, s6
; CHECK-SD-NOFP16-NEXT: fcvt s7, h16
; CHECK-SD-NOFP16-NEXT: fcvt s4, h4
; CHECK-SD-NOFP16-NEXT: mov h6, v0.h[3]
; CHECK-SD-NOFP16-NEXT: fcvt s5, h5
; CHECK-SD-NOFP16-NEXT: mov h16, v0.h[4]
; CHECK-SD-NOFP16-NEXT: csetm w9, mi
; CHECK-SD-NOFP16-NEXT: fcmp s4, s7
; CHECK-SD-NOFP16-NEXT: fmov s4, w9
; CHECK-SD-NOFP16-NEXT: fcvt s6, h6
; CHECK-SD-NOFP16-NEXT: mov h7, v1.h[4]
; CHECK-SD-NOFP16-NEXT: fcvt s16, h16
; CHECK-SD-NOFP16-NEXT: mov v4.h[1], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s6, s5
; CHECK-SD-NOFP16-NEXT: mov h5, v1.h[5]
; CHECK-SD-NOFP16-NEXT: mov h6, v0.h[5]
; CHECK-SD-NOFP16-NEXT: fcvt s7, h7
; CHECK-SD-NOFP16-NEXT: mov v4.h[2], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcvt s5, h5
; CHECK-SD-NOFP16-NEXT: fcvt s6, h6
; CHECK-SD-NOFP16-NEXT: fcmp s16, s7
; CHECK-SD-NOFP16-NEXT: mov h7, v1.h[6]
; CHECK-SD-NOFP16-NEXT: mov h16, v0.h[6]
; CHECK-SD-NOFP16-NEXT: mov h1, v1.h[7]
; CHECK-SD-NOFP16-NEXT: mov h0, v0.h[7]
; CHECK-SD-NOFP16-NEXT: mov v4.h[3], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s6, s5
; CHECK-SD-NOFP16-NEXT: fcvt s5, h7
; CHECK-SD-NOFP16-NEXT: fcvt s6, h16
; CHECK-SD-NOFP16-NEXT: fcvt s1, h1
; CHECK-SD-NOFP16-NEXT: fcvt s0, h0
; CHECK-SD-NOFP16-NEXT: mov v4.h[4], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s6, s5
; CHECK-SD-NOFP16-NEXT: mov v4.h[5], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s0, s1
; CHECK-SD-NOFP16-NEXT: mov v4.h[6], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: mov v4.h[7], w8
; CHECK-SD-NOFP16-NEXT: mov v0.16b, v4.16b
; CHECK-SD-NOFP16-NEXT: bsl v0.16b, v2.16b, v3.16b
; CHECK-SD-NOFP16-NEXT: ret
;
; CHECK-SD-FP16-LABEL: v7f16_half:
; CHECK-SD-FP16: // %bb.0: // %entry
; CHECK-SD-FP16-NEXT: fcmgt v0.8h, v1.8h, v0.8h
; CHECK-SD-FP16-NEXT: bsl v0.16b, v2.16b, v3.16b
; CHECK-SD-FP16-NEXT: ret
;
; CHECK-GI-NOFP16-LABEL: v7f16_half:
; CHECK-GI-NOFP16: // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT: mov w8, #15 // =0xf
; CHECK-GI-NOFP16-NEXT: mov v4.h[0], v0.h[4]
; CHECK-GI-NOFP16-NEXT: mov v6.h[0], v1.h[4]
; CHECK-GI-NOFP16-NEXT: fmov s5, w8
; CHECK-GI-NOFP16-NEXT: mov w9, #65535 // =0xffff
; CHECK-GI-NOFP16-NEXT: fmov s7, w9
; CHECK-GI-NOFP16-NEXT: mov v5.h[1], w8
; CHECK-GI-NOFP16-NEXT: mov v4.h[1], v0.h[5]
; CHECK-GI-NOFP16-NEXT: mov v6.h[1], v1.h[5]
; CHECK-GI-NOFP16-NEXT: mov v7.h[1], w9
; CHECK-GI-NOFP16-NEXT: mov v5.h[2], w8
; CHECK-GI-NOFP16-NEXT: mov v4.h[2], v0.h[6]
; CHECK-GI-NOFP16-NEXT: mov v6.h[2], v1.h[6]
; CHECK-GI-NOFP16-NEXT: mov v7.h[2], w9
; CHECK-GI-NOFP16-NEXT: fcvtl v0.4s, v0.4h
; CHECK-GI-NOFP16-NEXT: fcvtl v1.4s, v1.4h
; CHECK-GI-NOFP16-NEXT: mov v5.h[3], w8
; CHECK-GI-NOFP16-NEXT: fcvtl v4.4s, v4.4h
; CHECK-GI-NOFP16-NEXT: fcvtl v6.4s, v6.4h
; CHECK-GI-NOFP16-NEXT: mov v7.h[3], w9
; CHECK-GI-NOFP16-NEXT: fcmgt v0.4s, v1.4s, v0.4s
; CHECK-GI-NOFP16-NEXT: mov v5.h[4], w8
; CHECK-GI-NOFP16-NEXT: fcmgt v1.4s, v6.4s, v4.4s
; CHECK-GI-NOFP16-NEXT: mov v7.h[4], w9
; CHECK-GI-NOFP16-NEXT: mov v5.h[5], w8
; CHECK-GI-NOFP16-NEXT: uzp1 v0.8h, v0.8h, v1.8h
; CHECK-GI-NOFP16-NEXT: mov v7.h[5], w9
; CHECK-GI-NOFP16-NEXT: mov v5.h[6], w8
; CHECK-GI-NOFP16-NEXT: mov v7.h[6], w9
; CHECK-GI-NOFP16-NEXT: ushl v0.8h, v0.8h, v5.8h
; CHECK-GI-NOFP16-NEXT: neg v1.8h, v5.8h
; CHECK-GI-NOFP16-NEXT: sshl v0.8h, v0.8h, v1.8h
; CHECK-GI-NOFP16-NEXT: eor v1.16b, v0.16b, v7.16b
; CHECK-GI-NOFP16-NEXT: and v0.16b, v2.16b, v0.16b
; CHECK-GI-NOFP16-NEXT: and v1.16b, v3.16b, v1.16b
; CHECK-GI-NOFP16-NEXT: orr v0.16b, v0.16b, v1.16b
; CHECK-GI-NOFP16-NEXT: ret
;
; CHECK-GI-FP16-LABEL: v7f16_half:
; CHECK-GI-FP16: // %bb.0: // %entry
; CHECK-GI-FP16-NEXT: mov w8, #15 // =0xf
; CHECK-GI-FP16-NEXT: mov w9, #65535 // =0xffff
; CHECK-GI-FP16-NEXT: fcmgt v0.8h, v1.8h, v0.8h
; CHECK-GI-FP16-NEXT: fmov s4, w8
; CHECK-GI-FP16-NEXT: fmov s5, w9
; CHECK-GI-FP16-NEXT: mov v4.h[1], w8
; CHECK-GI-FP16-NEXT: mov v5.h[1], w9
; CHECK-GI-FP16-NEXT: mov v4.h[2], w8
; CHECK-GI-FP16-NEXT: mov v5.h[2], w9
; CHECK-GI-FP16-NEXT: mov v4.h[3], w8
; CHECK-GI-FP16-NEXT: mov v5.h[3], w9
; CHECK-GI-FP16-NEXT: mov v4.h[4], w8
; CHECK-GI-FP16-NEXT: mov v5.h[4], w9
; CHECK-GI-FP16-NEXT: mov v4.h[5], w8
; CHECK-GI-FP16-NEXT: mov v5.h[5], w9
; CHECK-GI-FP16-NEXT: mov v4.h[6], w8
; CHECK-GI-FP16-NEXT: mov v5.h[6], w9
; CHECK-GI-FP16-NEXT: ushl v0.8h, v0.8h, v4.8h
; CHECK-GI-FP16-NEXT: neg v1.8h, v4.8h
; CHECK-GI-FP16-NEXT: sshl v0.8h, v0.8h, v1.8h
; CHECK-GI-FP16-NEXT: eor v1.16b, v0.16b, v5.16b
; CHECK-GI-FP16-NEXT: and v0.16b, v2.16b, v0.16b
; CHECK-GI-FP16-NEXT: and v1.16b, v3.16b, v1.16b
; CHECK-GI-FP16-NEXT: orr v0.16b, v0.16b, v1.16b
; CHECK-GI-FP16-NEXT: ret
entry:
%c = fcmp olt <7 x half> %a, %b
%s = select <7 x i1> %c, <7 x half> %d, <7 x half> %e
ret <7 x half> %s
}
define <4 x half> @v4f16_half(<4 x half> %a, <4 x half> %b, <4 x half> %d, <4 x half> %e) {
; CHECK-SD-NOFP16-LABEL: v4f16_half:
; CHECK-SD-NOFP16: // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT: fcvtl v0.4s, v0.4h
; CHECK-SD-NOFP16-NEXT: fcvtl v1.4s, v1.4h
; CHECK-SD-NOFP16-NEXT: fcmgt v0.4s, v1.4s, v0.4s
; CHECK-SD-NOFP16-NEXT: xtn v0.4h, v0.4s
; CHECK-SD-NOFP16-NEXT: bsl v0.8b, v2.8b, v3.8b
; CHECK-SD-NOFP16-NEXT: ret
;
; CHECK-SD-FP16-LABEL: v4f16_half:
; CHECK-SD-FP16: // %bb.0: // %entry
; CHECK-SD-FP16-NEXT: fcmgt v0.4h, v1.4h, v0.4h
; CHECK-SD-FP16-NEXT: bsl v0.8b, v2.8b, v3.8b
; CHECK-SD-FP16-NEXT: ret
;
; CHECK-GI-NOFP16-LABEL: v4f16_half:
; CHECK-GI-NOFP16: // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT: fcvtl v0.4s, v0.4h
; CHECK-GI-NOFP16-NEXT: fcvtl v1.4s, v1.4h
; CHECK-GI-NOFP16-NEXT: fcmgt v0.4s, v1.4s, v0.4s
; CHECK-GI-NOFP16-NEXT: xtn v0.4h, v0.4s
; CHECK-GI-NOFP16-NEXT: bsl v0.8b, v2.8b, v3.8b
; CHECK-GI-NOFP16-NEXT: ret
;
; CHECK-GI-FP16-LABEL: v4f16_half:
; CHECK-GI-FP16: // %bb.0: // %entry
; CHECK-GI-FP16-NEXT: fcmgt v0.4h, v1.4h, v0.4h
; CHECK-GI-FP16-NEXT: bsl v0.8b, v2.8b, v3.8b
; CHECK-GI-FP16-NEXT: ret
entry:
%c = fcmp olt <4 x half> %a, %b
%s = select <4 x i1> %c, <4 x half> %d, <4 x half> %e
ret <4 x half> %s
}
define <8 x half> @v8f16_half(<8 x half> %a, <8 x half> %b, <8 x half> %d, <8 x half> %e) {
; CHECK-SD-NOFP16-LABEL: v8f16_half:
; CHECK-SD-NOFP16: // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT: mov h4, v1.h[1]
; CHECK-SD-NOFP16-NEXT: mov h5, v0.h[1]
; CHECK-SD-NOFP16-NEXT: fcvt s6, h1
; CHECK-SD-NOFP16-NEXT: fcvt s7, h0
; CHECK-SD-NOFP16-NEXT: mov h16, v1.h[2]
; CHECK-SD-NOFP16-NEXT: fcvt s4, h4
; CHECK-SD-NOFP16-NEXT: fcvt s5, h5
; CHECK-SD-NOFP16-NEXT: fcmp s5, s4
; CHECK-SD-NOFP16-NEXT: mov h4, v0.h[2]
; CHECK-SD-NOFP16-NEXT: mov h5, v1.h[3]
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s7, s6
; CHECK-SD-NOFP16-NEXT: fcvt s7, h16
; CHECK-SD-NOFP16-NEXT: fcvt s4, h4
; CHECK-SD-NOFP16-NEXT: mov h6, v0.h[3]
; CHECK-SD-NOFP16-NEXT: fcvt s5, h5
; CHECK-SD-NOFP16-NEXT: mov h16, v0.h[4]
; CHECK-SD-NOFP16-NEXT: csetm w9, mi
; CHECK-SD-NOFP16-NEXT: fcmp s4, s7
; CHECK-SD-NOFP16-NEXT: fmov s4, w9
; CHECK-SD-NOFP16-NEXT: fcvt s6, h6
; CHECK-SD-NOFP16-NEXT: mov h7, v1.h[4]
; CHECK-SD-NOFP16-NEXT: fcvt s16, h16
; CHECK-SD-NOFP16-NEXT: mov v4.h[1], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s6, s5
; CHECK-SD-NOFP16-NEXT: mov h5, v1.h[5]
; CHECK-SD-NOFP16-NEXT: mov h6, v0.h[5]
; CHECK-SD-NOFP16-NEXT: fcvt s7, h7
; CHECK-SD-NOFP16-NEXT: mov v4.h[2], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcvt s5, h5
; CHECK-SD-NOFP16-NEXT: fcvt s6, h6
; CHECK-SD-NOFP16-NEXT: fcmp s16, s7
; CHECK-SD-NOFP16-NEXT: mov h7, v1.h[6]
; CHECK-SD-NOFP16-NEXT: mov h16, v0.h[6]
; CHECK-SD-NOFP16-NEXT: mov h1, v1.h[7]
; CHECK-SD-NOFP16-NEXT: mov h0, v0.h[7]
; CHECK-SD-NOFP16-NEXT: mov v4.h[3], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s6, s5
; CHECK-SD-NOFP16-NEXT: fcvt s5, h7
; CHECK-SD-NOFP16-NEXT: fcvt s6, h16
; CHECK-SD-NOFP16-NEXT: fcvt s1, h1
; CHECK-SD-NOFP16-NEXT: fcvt s0, h0
; CHECK-SD-NOFP16-NEXT: mov v4.h[4], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s6, s5
; CHECK-SD-NOFP16-NEXT: mov v4.h[5], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s0, s1
; CHECK-SD-NOFP16-NEXT: mov v4.h[6], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: mov v4.h[7], w8
; CHECK-SD-NOFP16-NEXT: mov v0.16b, v4.16b
; CHECK-SD-NOFP16-NEXT: bsl v0.16b, v2.16b, v3.16b
; CHECK-SD-NOFP16-NEXT: ret
;
; CHECK-SD-FP16-LABEL: v8f16_half:
; CHECK-SD-FP16: // %bb.0: // %entry
; CHECK-SD-FP16-NEXT: fcmgt v0.8h, v1.8h, v0.8h
; CHECK-SD-FP16-NEXT: bsl v0.16b, v2.16b, v3.16b
; CHECK-SD-FP16-NEXT: ret
;
; CHECK-GI-NOFP16-LABEL: v8f16_half:
; CHECK-GI-NOFP16: // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT: fcvtl v4.4s, v0.4h
; CHECK-GI-NOFP16-NEXT: fcvtl2 v0.4s, v0.8h
; CHECK-GI-NOFP16-NEXT: fcvtl v5.4s, v1.4h
; CHECK-GI-NOFP16-NEXT: fcvtl2 v1.4s, v1.8h
; CHECK-GI-NOFP16-NEXT: fcmgt v4.4s, v5.4s, v4.4s
; CHECK-GI-NOFP16-NEXT: fcmgt v0.4s, v1.4s, v0.4s
; CHECK-GI-NOFP16-NEXT: uzp1 v0.8h, v4.8h, v0.8h
; CHECK-GI-NOFP16-NEXT: shl v0.8h, v0.8h, #15
; CHECK-GI-NOFP16-NEXT: sshr v0.8h, v0.8h, #15
; CHECK-GI-NOFP16-NEXT: bsl v0.16b, v2.16b, v3.16b
; CHECK-GI-NOFP16-NEXT: ret
;
; CHECK-GI-FP16-LABEL: v8f16_half:
; CHECK-GI-FP16: // %bb.0: // %entry
; CHECK-GI-FP16-NEXT: fcmgt v0.8h, v1.8h, v0.8h
; CHECK-GI-FP16-NEXT: bsl v0.16b, v2.16b, v3.16b
; CHECK-GI-FP16-NEXT: ret
entry:
%c = fcmp olt <8 x half> %a, %b
%s = select <8 x i1> %c, <8 x half> %d, <8 x half> %e
ret <8 x half> %s
}
define <16 x half> @v16f16_half(<16 x half> %a, <16 x half> %b, <16 x half> %d, <16 x half> %e) {
; CHECK-SD-NOFP16-LABEL: v16f16_half:
; CHECK-SD-NOFP16: // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT: mov h16, v3.h[1]
; CHECK-SD-NOFP16-NEXT: mov h17, v1.h[1]
; CHECK-SD-NOFP16-NEXT: mov h18, v3.h[2]
; CHECK-SD-NOFP16-NEXT: mov h19, v1.h[2]
; CHECK-SD-NOFP16-NEXT: fcvt s20, h3
; CHECK-SD-NOFP16-NEXT: fcvt s21, h1
; CHECK-SD-NOFP16-NEXT: fcvt s16, h16
; CHECK-SD-NOFP16-NEXT: fcvt s17, h17
; CHECK-SD-NOFP16-NEXT: fcvt s18, h18
; CHECK-SD-NOFP16-NEXT: fcvt s19, h19
; CHECK-SD-NOFP16-NEXT: fcmp s17, s16
; CHECK-SD-NOFP16-NEXT: mov h16, v3.h[3]
; CHECK-SD-NOFP16-NEXT: mov h17, v1.h[3]
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s21, s20
; CHECK-SD-NOFP16-NEXT: mov h20, v3.h[4]
; CHECK-SD-NOFP16-NEXT: mov h21, v1.h[4]
; CHECK-SD-NOFP16-NEXT: fcvt s16, h16
; CHECK-SD-NOFP16-NEXT: fcvt s17, h17
; CHECK-SD-NOFP16-NEXT: csetm w14, mi
; CHECK-SD-NOFP16-NEXT: fcmp s19, s18
; CHECK-SD-NOFP16-NEXT: mov h18, v3.h[5]
; CHECK-SD-NOFP16-NEXT: mov h19, v1.h[5]
; CHECK-SD-NOFP16-NEXT: fcvt s20, h20
; CHECK-SD-NOFP16-NEXT: fcvt s21, h21
; CHECK-SD-NOFP16-NEXT: csetm w13, mi
; CHECK-SD-NOFP16-NEXT: fcmp s17, s16
; CHECK-SD-NOFP16-NEXT: mov h16, v3.h[6]
; CHECK-SD-NOFP16-NEXT: mov h17, v1.h[6]
; CHECK-SD-NOFP16-NEXT: fcvt s18, h18
; CHECK-SD-NOFP16-NEXT: mov h3, v3.h[7]
; CHECK-SD-NOFP16-NEXT: fcvt s19, h19
; CHECK-SD-NOFP16-NEXT: mov h1, v1.h[7]
; CHECK-SD-NOFP16-NEXT: csetm w11, mi
; CHECK-SD-NOFP16-NEXT: fcmp s21, s20
; CHECK-SD-NOFP16-NEXT: fcvt s16, h16
; CHECK-SD-NOFP16-NEXT: fcvt s17, h17
; CHECK-SD-NOFP16-NEXT: fcvt s3, h3
; CHECK-SD-NOFP16-NEXT: csetm w12, mi
; CHECK-SD-NOFP16-NEXT: fcmp s19, s18
; CHECK-SD-NOFP16-NEXT: mov h18, v2.h[1]
; CHECK-SD-NOFP16-NEXT: mov h19, v0.h[1]
; CHECK-SD-NOFP16-NEXT: fcvt s1, h1
; CHECK-SD-NOFP16-NEXT: csetm w10, mi
; CHECK-SD-NOFP16-NEXT: fcmp s17, s16
; CHECK-SD-NOFP16-NEXT: fcvt s16, h18
; CHECK-SD-NOFP16-NEXT: mov h18, v2.h[2]
; CHECK-SD-NOFP16-NEXT: fcvt s17, h19
; CHECK-SD-NOFP16-NEXT: mov h19, v0.h[2]
; CHECK-SD-NOFP16-NEXT: csetm w9, mi
; CHECK-SD-NOFP16-NEXT: fcmp s1, s3
; CHECK-SD-NOFP16-NEXT: fcvt s1, h2
; CHECK-SD-NOFP16-NEXT: fcvt s3, h0
; CHECK-SD-NOFP16-NEXT: fcvt s18, h18
; CHECK-SD-NOFP16-NEXT: csetm w15, mi
; CHECK-SD-NOFP16-NEXT: fcmp s17, s16
; CHECK-SD-NOFP16-NEXT: mov h16, v2.h[3]
; CHECK-SD-NOFP16-NEXT: mov h17, v0.h[3]
; CHECK-SD-NOFP16-NEXT: fcvt s19, h19
; CHECK-SD-NOFP16-NEXT: csetm w16, mi
; CHECK-SD-NOFP16-NEXT: fcmp s3, s1
; CHECK-SD-NOFP16-NEXT: fmov s1, w14
; CHECK-SD-NOFP16-NEXT: fcvt s16, h16
; CHECK-SD-NOFP16-NEXT: fcvt s17, h17
; CHECK-SD-NOFP16-NEXT: csetm w14, mi
; CHECK-SD-NOFP16-NEXT: fcmp s19, s18
; CHECK-SD-NOFP16-NEXT: mov h18, v2.h[4]
; CHECK-SD-NOFP16-NEXT: fmov s3, w14
; CHECK-SD-NOFP16-NEXT: mov h19, v0.h[4]
; CHECK-SD-NOFP16-NEXT: mov v1.h[1], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s17, s16
; CHECK-SD-NOFP16-NEXT: mov h16, v2.h[5]
; CHECK-SD-NOFP16-NEXT: mov v3.h[1], w16
; CHECK-SD-NOFP16-NEXT: mov h17, v0.h[5]
; CHECK-SD-NOFP16-NEXT: fcvt s18, h18
; CHECK-SD-NOFP16-NEXT: fcvt s19, h19
; CHECK-SD-NOFP16-NEXT: mov v1.h[2], w13
; CHECK-SD-NOFP16-NEXT: fcvt s16, h16
; CHECK-SD-NOFP16-NEXT: mov v3.h[2], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcvt s17, h17
; CHECK-SD-NOFP16-NEXT: fcmp s19, s18
; CHECK-SD-NOFP16-NEXT: mov h18, v2.h[6]
; CHECK-SD-NOFP16-NEXT: mov h19, v0.h[6]
; CHECK-SD-NOFP16-NEXT: mov v1.h[3], w11
; CHECK-SD-NOFP16-NEXT: mov h2, v2.h[7]
; CHECK-SD-NOFP16-NEXT: mov h0, v0.h[7]
; CHECK-SD-NOFP16-NEXT: mov v3.h[3], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s17, s16
; CHECK-SD-NOFP16-NEXT: fcvt s16, h18
; CHECK-SD-NOFP16-NEXT: fcvt s17, h19
; CHECK-SD-NOFP16-NEXT: mov v1.h[4], w12
; CHECK-SD-NOFP16-NEXT: fcvt s2, h2
; CHECK-SD-NOFP16-NEXT: fcvt s0, h0
; CHECK-SD-NOFP16-NEXT: mov v3.h[4], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s17, s16
; CHECK-SD-NOFP16-NEXT: mov v1.h[5], w10
; CHECK-SD-NOFP16-NEXT: mov v3.h[5], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s0, s2
; CHECK-SD-NOFP16-NEXT: mov v1.h[6], w9
; CHECK-SD-NOFP16-NEXT: mov v3.h[6], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: mov v1.h[7], w15
; CHECK-SD-NOFP16-NEXT: mov v3.h[7], w8
; CHECK-SD-NOFP16-NEXT: bsl v1.16b, v5.16b, v7.16b
; CHECK-SD-NOFP16-NEXT: mov v0.16b, v3.16b
; CHECK-SD-NOFP16-NEXT: bsl v0.16b, v4.16b, v6.16b
; CHECK-SD-NOFP16-NEXT: ret
;
; CHECK-SD-FP16-LABEL: v16f16_half:
; CHECK-SD-FP16: // %bb.0: // %entry
; CHECK-SD-FP16-NEXT: fcmgt v1.8h, v3.8h, v1.8h
; CHECK-SD-FP16-NEXT: fcmgt v0.8h, v2.8h, v0.8h
; CHECK-SD-FP16-NEXT: bsl v1.16b, v5.16b, v7.16b
; CHECK-SD-FP16-NEXT: bsl v0.16b, v4.16b, v6.16b
; CHECK-SD-FP16-NEXT: ret
;
; CHECK-GI-NOFP16-LABEL: v16f16_half:
; CHECK-GI-NOFP16: // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT: fcvtl v16.4s, v0.4h
; CHECK-GI-NOFP16-NEXT: fcvtl2 v0.4s, v0.8h
; CHECK-GI-NOFP16-NEXT: fcvtl v17.4s, v1.4h
; CHECK-GI-NOFP16-NEXT: fcvtl2 v1.4s, v1.8h
; CHECK-GI-NOFP16-NEXT: fcvtl v18.4s, v2.4h
; CHECK-GI-NOFP16-NEXT: fcvtl2 v2.4s, v2.8h
; CHECK-GI-NOFP16-NEXT: fcvtl v19.4s, v3.4h
; CHECK-GI-NOFP16-NEXT: fcvtl2 v3.4s, v3.8h
; CHECK-GI-NOFP16-NEXT: fcmgt v16.4s, v18.4s, v16.4s
; CHECK-GI-NOFP16-NEXT: fcmgt v0.4s, v2.4s, v0.4s
; CHECK-GI-NOFP16-NEXT: fcmgt v2.4s, v19.4s, v17.4s
; CHECK-GI-NOFP16-NEXT: fcmgt v1.4s, v3.4s, v1.4s
; CHECK-GI-NOFP16-NEXT: uzp1 v0.8h, v16.8h, v0.8h
; CHECK-GI-NOFP16-NEXT: uzp1 v1.8h, v2.8h, v1.8h
; CHECK-GI-NOFP16-NEXT: shl v0.8h, v0.8h, #15
; CHECK-GI-NOFP16-NEXT: shl v1.8h, v1.8h, #15
; CHECK-GI-NOFP16-NEXT: sshr v0.8h, v0.8h, #15
; CHECK-GI-NOFP16-NEXT: sshr v1.8h, v1.8h, #15
; CHECK-GI-NOFP16-NEXT: bsl v0.16b, v4.16b, v6.16b
; CHECK-GI-NOFP16-NEXT: bsl v1.16b, v5.16b, v7.16b
; CHECK-GI-NOFP16-NEXT: ret
;
; CHECK-GI-FP16-LABEL: v16f16_half:
; CHECK-GI-FP16: // %bb.0: // %entry
; CHECK-GI-FP16-NEXT: fcmgt v0.8h, v2.8h, v0.8h
; CHECK-GI-FP16-NEXT: fcmgt v1.8h, v3.8h, v1.8h
; CHECK-GI-FP16-NEXT: bsl v0.16b, v4.16b, v6.16b
; CHECK-GI-FP16-NEXT: bsl v1.16b, v5.16b, v7.16b
; CHECK-GI-FP16-NEXT: ret
entry:
%c = fcmp olt <16 x half> %a, %b
%s = select <16 x i1> %c, <16 x half> %d, <16 x half> %e
ret <16 x half> %s
}
define <7 x i32> @v7f16_i32(<7 x half> %a, <7 x half> %b, <7 x i32> %d, <7 x i32> %e) {
; CHECK-SD-NOFP16-LABEL: v7f16_i32:
; CHECK-SD-NOFP16: // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT: mov h2, v1.h[1]
; CHECK-SD-NOFP16-NEXT: mov h3, v0.h[1]
; CHECK-SD-NOFP16-NEXT: mov h4, v1.h[2]
; CHECK-SD-NOFP16-NEXT: mov h5, v0.h[2]
; CHECK-SD-NOFP16-NEXT: fcvt s6, h1
; CHECK-SD-NOFP16-NEXT: fcvt s7, h0
; CHECK-SD-NOFP16-NEXT: fcvt s2, h2
; CHECK-SD-NOFP16-NEXT: fcvt s3, h3
; CHECK-SD-NOFP16-NEXT: fcvt s4, h4
; CHECK-SD-NOFP16-NEXT: fcvt s5, h5
; CHECK-SD-NOFP16-NEXT: fcmp s3, s2
; CHECK-SD-NOFP16-NEXT: mov h2, v1.h[3]
; CHECK-SD-NOFP16-NEXT: mov h3, v0.h[3]
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s7, s6
; CHECK-SD-NOFP16-NEXT: mov h6, v1.h[5]
; CHECK-SD-NOFP16-NEXT: mov h7, v0.h[5]
; CHECK-SD-NOFP16-NEXT: fcvt s2, h2
; CHECK-SD-NOFP16-NEXT: fcvt s3, h3
; CHECK-SD-NOFP16-NEXT: csetm w9, mi
; CHECK-SD-NOFP16-NEXT: fcmp s5, s4
; CHECK-SD-NOFP16-NEXT: mov h4, v1.h[4]
; CHECK-SD-NOFP16-NEXT: mov h5, v0.h[4]
; CHECK-SD-NOFP16-NEXT: fcvt s6, h6
; CHECK-SD-NOFP16-NEXT: fcvt s7, h7
; CHECK-SD-NOFP16-NEXT: csetm w10, mi
; CHECK-SD-NOFP16-NEXT: fcmp s3, s2
; CHECK-SD-NOFP16-NEXT: fcvt s2, h4
; CHECK-SD-NOFP16-NEXT: mov h4, v1.h[6]
; CHECK-SD-NOFP16-NEXT: mov h1, v1.h[7]
; CHECK-SD-NOFP16-NEXT: fcvt s3, h5
; CHECK-SD-NOFP16-NEXT: mov h5, v0.h[6]
; CHECK-SD-NOFP16-NEXT: mov h0, v0.h[7]
; CHECK-SD-NOFP16-NEXT: csetm w11, mi
; CHECK-SD-NOFP16-NEXT: fcmp s7, s6
; CHECK-SD-NOFP16-NEXT: fcvt s1, h1
; CHECK-SD-NOFP16-NEXT: csetm w12, mi
; CHECK-SD-NOFP16-NEXT: fcmp s3, s2
; CHECK-SD-NOFP16-NEXT: fcvt s2, h4
; CHECK-SD-NOFP16-NEXT: fcvt s3, h5
; CHECK-SD-NOFP16-NEXT: fmov s4, w9
; CHECK-SD-NOFP16-NEXT: fcvt s0, h0
; CHECK-SD-NOFP16-NEXT: add x9, sp, #8
; CHECK-SD-NOFP16-NEXT: csetm w13, mi
; CHECK-SD-NOFP16-NEXT: fmov s5, w13
; CHECK-SD-NOFP16-NEXT: mov v4.h[1], w8
; CHECK-SD-NOFP16-NEXT: mov x8, sp
; CHECK-SD-NOFP16-NEXT: fcmp s3, s2
; CHECK-SD-NOFP16-NEXT: fmov s2, w7
; CHECK-SD-NOFP16-NEXT: fmov s3, w0
; CHECK-SD-NOFP16-NEXT: mov v5.h[1], w12
; CHECK-SD-NOFP16-NEXT: ld1 { v2.s }[1], [x8]
; CHECK-SD-NOFP16-NEXT: mov v3.s[1], w1
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: mov v4.h[2], w10
; CHECK-SD-NOFP16-NEXT: fcmp s0, s1
; CHECK-SD-NOFP16-NEXT: fmov s1, w4
; CHECK-SD-NOFP16-NEXT: ldr s0, [sp, #24]
; CHECK-SD-NOFP16-NEXT: mov v5.h[2], w8
; CHECK-SD-NOFP16-NEXT: ld1 { v2.s }[2], [x9]
; CHECK-SD-NOFP16-NEXT: add x9, sp, #32
; CHECK-SD-NOFP16-NEXT: mov v3.s[2], w2
; CHECK-SD-NOFP16-NEXT: mov v1.s[1], w5
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: mov v4.h[3], w11
; CHECK-SD-NOFP16-NEXT: ld1 { v0.s }[1], [x9]
; CHECK-SD-NOFP16-NEXT: mov v5.h[3], w8
; CHECK-SD-NOFP16-NEXT: add x8, sp, #16
; CHECK-SD-NOFP16-NEXT: ld1 { v2.s }[3], [x8]
; CHECK-SD-NOFP16-NEXT: mov v3.s[3], w3
; CHECK-SD-NOFP16-NEXT: add x8, sp, #40
; CHECK-SD-NOFP16-NEXT: mov v1.s[2], w6
; CHECK-SD-NOFP16-NEXT: sshll v4.4s, v4.4h, #0
; CHECK-SD-NOFP16-NEXT: ld1 { v0.s }[2], [x8]
; CHECK-SD-NOFP16-NEXT: sshll v5.4s, v5.4h, #0
; CHECK-SD-NOFP16-NEXT: bit v2.16b, v3.16b, v4.16b
; CHECK-SD-NOFP16-NEXT: bit v0.16b, v1.16b, v5.16b
; CHECK-SD-NOFP16-NEXT: mov w1, v2.s[1]
; CHECK-SD-NOFP16-NEXT: mov w2, v2.s[2]
; CHECK-SD-NOFP16-NEXT: mov w3, v2.s[3]
; CHECK-SD-NOFP16-NEXT: fmov w0, s2
; CHECK-SD-NOFP16-NEXT: mov w5, v0.s[1]
; CHECK-SD-NOFP16-NEXT: mov w6, v0.s[2]
; CHECK-SD-NOFP16-NEXT: fmov w4, s0
; CHECK-SD-NOFP16-NEXT: ret
;
; CHECK-SD-FP16-LABEL: v7f16_i32:
; CHECK-SD-FP16: // %bb.0: // %entry
; CHECK-SD-FP16-NEXT: fmov s2, w0
; CHECK-SD-FP16-NEXT: fmov s3, w7
; CHECK-SD-FP16-NEXT: mov x8, sp
; CHECK-SD-FP16-NEXT: fmov s5, w4
; CHECK-SD-FP16-NEXT: ldr s4, [sp, #24]
; CHECK-SD-FP16-NEXT: fcmgt v0.8h, v1.8h, v0.8h
; CHECK-SD-FP16-NEXT: add x9, sp, #32
; CHECK-SD-FP16-NEXT: mov v2.s[1], w1
; CHECK-SD-FP16-NEXT: ld1 { v3.s }[1], [x8]
; CHECK-SD-FP16-NEXT: add x8, sp, #8
; CHECK-SD-FP16-NEXT: mov v5.s[1], w5
; CHECK-SD-FP16-NEXT: ld1 { v4.s }[1], [x9]
; CHECK-SD-FP16-NEXT: add x9, sp, #16
; CHECK-SD-FP16-NEXT: sshll v1.4s, v0.4h, #0
; CHECK-SD-FP16-NEXT: sshll2 v0.4s, v0.8h, #0
; CHECK-SD-FP16-NEXT: ld1 { v3.s }[2], [x8]
; CHECK-SD-FP16-NEXT: add x8, sp, #40
; CHECK-SD-FP16-NEXT: mov v2.s[2], w2
; CHECK-SD-FP16-NEXT: ld1 { v4.s }[2], [x8]
; CHECK-SD-FP16-NEXT: mov v5.s[2], w6
; CHECK-SD-FP16-NEXT: ld1 { v3.s }[3], [x9]
; CHECK-SD-FP16-NEXT: mov v2.s[3], w3
; CHECK-SD-FP16-NEXT: bsl v0.16b, v5.16b, v4.16b
; CHECK-SD-FP16-NEXT: bsl v1.16b, v2.16b, v3.16b
; CHECK-SD-FP16-NEXT: mov w5, v0.s[1]
; CHECK-SD-FP16-NEXT: mov w6, v0.s[2]
; CHECK-SD-FP16-NEXT: fmov w4, s0
; CHECK-SD-FP16-NEXT: mov w1, v1.s[1]
; CHECK-SD-FP16-NEXT: mov w2, v1.s[2]
; CHECK-SD-FP16-NEXT: mov w3, v1.s[3]
; CHECK-SD-FP16-NEXT: fmov w0, s1
; CHECK-SD-FP16-NEXT: ret
;
; CHECK-GI-NOFP16-LABEL: v7f16_i32:
; CHECK-GI-NOFP16: // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT: mov v2.h[0], v0.h[4]
; CHECK-GI-NOFP16-NEXT: mov v3.h[0], v1.h[4]
; CHECK-GI-NOFP16-NEXT: mov w8, #31 // =0x1f
; CHECK-GI-NOFP16-NEXT: mov v4.s[0], w8
; CHECK-GI-NOFP16-NEXT: mov w9, #-1 // =0xffffffff
; CHECK-GI-NOFP16-NEXT: mov v5.s[0], w0
; CHECK-GI-NOFP16-NEXT: mov v6.s[0], w9
; CHECK-GI-NOFP16-NEXT: mov v7.s[0], w7
; CHECK-GI-NOFP16-NEXT: ldr s16, [sp]
; CHECK-GI-NOFP16-NEXT: ldr s17, [sp, #24]
; CHECK-GI-NOFP16-NEXT: ldr s18, [sp, #32]
; CHECK-GI-NOFP16-NEXT: mov v2.h[1], v0.h[5]
; CHECK-GI-NOFP16-NEXT: mov v3.h[1], v1.h[5]
; CHECK-GI-NOFP16-NEXT: mov v4.s[1], w8
; CHECK-GI-NOFP16-NEXT: mov v5.s[1], w1
; CHECK-GI-NOFP16-NEXT: mov v17.s[1], v18.s[0]
; CHECK-GI-NOFP16-NEXT: mov v6.s[1], w9
; CHECK-GI-NOFP16-NEXT: mov v7.s[1], v16.s[0]
; CHECK-GI-NOFP16-NEXT: ldr s16, [sp, #8]
; CHECK-GI-NOFP16-NEXT: mov v2.h[2], v0.h[6]
; CHECK-GI-NOFP16-NEXT: mov v3.h[2], v1.h[6]
; CHECK-GI-NOFP16-NEXT: fcvtl v0.4s, v0.4h
; CHECK-GI-NOFP16-NEXT: mov v4.s[2], w8
; CHECK-GI-NOFP16-NEXT: fcvtl v1.4s, v1.4h
; CHECK-GI-NOFP16-NEXT: mov v5.s[2], w2
; CHECK-GI-NOFP16-NEXT: mov v6.s[2], w9
; CHECK-GI-NOFP16-NEXT: mov v7.s[2], v16.s[0]
; CHECK-GI-NOFP16-NEXT: ldr s16, [sp, #40]
; CHECK-GI-NOFP16-NEXT: fcvtl v2.4s, v2.4h
; CHECK-GI-NOFP16-NEXT: fcvtl v3.4s, v3.4h
; CHECK-GI-NOFP16-NEXT: mov v17.s[2], v16.s[0]
; CHECK-GI-NOFP16-NEXT: fcmgt v0.4s, v1.4s, v0.4s
; CHECK-GI-NOFP16-NEXT: mov v5.s[3], w3
; CHECK-GI-NOFP16-NEXT: fcmgt v2.4s, v3.4s, v2.4s
; CHECK-GI-NOFP16-NEXT: mov v3.s[0], w4
; CHECK-GI-NOFP16-NEXT: ushl v2.4s, v2.4s, v4.4s
; CHECK-GI-NOFP16-NEXT: neg v4.4s, v4.4s
; CHECK-GI-NOFP16-NEXT: mov v3.s[1], w5
; CHECK-GI-NOFP16-NEXT: sshl v2.4s, v2.4s, v4.4s
; CHECK-GI-NOFP16-NEXT: ldr s4, [sp, #16]
; CHECK-GI-NOFP16-NEXT: mov v3.s[2], w6
; CHECK-GI-NOFP16-NEXT: mov v7.s[3], v4.s[0]
; CHECK-GI-NOFP16-NEXT: eor v1.16b, v2.16b, v6.16b
; CHECK-GI-NOFP16-NEXT: and v2.16b, v3.16b, v2.16b
; CHECK-GI-NOFP16-NEXT: and v1.16b, v17.16b, v1.16b
; CHECK-GI-NOFP16-NEXT: bsl v0.16b, v5.16b, v7.16b
; CHECK-GI-NOFP16-NEXT: orr v1.16b, v2.16b, v1.16b
; CHECK-GI-NOFP16-NEXT: mov s2, v0.s[1]
; CHECK-GI-NOFP16-NEXT: mov s3, v0.s[2]
; CHECK-GI-NOFP16-NEXT: mov s4, v0.s[3]
; CHECK-GI-NOFP16-NEXT: fmov w0, s0
; CHECK-GI-NOFP16-NEXT: mov s5, v1.s[1]
; CHECK-GI-NOFP16-NEXT: mov s6, v1.s[2]
; CHECK-GI-NOFP16-NEXT: fmov w4, s1
; CHECK-GI-NOFP16-NEXT: fmov w1, s2
; CHECK-GI-NOFP16-NEXT: fmov w2, s3
; CHECK-GI-NOFP16-NEXT: fmov w3, s4
; CHECK-GI-NOFP16-NEXT: fmov w5, s5
; CHECK-GI-NOFP16-NEXT: fmov w6, s6
; CHECK-GI-NOFP16-NEXT: ret
;
; CHECK-GI-FP16-LABEL: v7f16_i32:
; CHECK-GI-FP16: // %bb.0: // %entry
; CHECK-GI-FP16-NEXT: fcmgt v0.8h, v1.8h, v0.8h
; CHECK-GI-FP16-NEXT: mov w9, #31 // =0x1f
; CHECK-GI-FP16-NEXT: mov v4.s[0], w0
; CHECK-GI-FP16-NEXT: mov v2.s[0], w9
; CHECK-GI-FP16-NEXT: mov v5.s[0], w7
; CHECK-GI-FP16-NEXT: ldr s6, [sp]
; CHECK-GI-FP16-NEXT: mov v7.s[0], w4
; CHECK-GI-FP16-NEXT: ldr s16, [sp, #32]
; CHECK-GI-FP16-NEXT: ldr s17, [sp, #8]
; CHECK-GI-FP16-NEXT: umov w8, v0.h[4]
; CHECK-GI-FP16-NEXT: umov w10, v0.h[5]
; CHECK-GI-FP16-NEXT: mov v4.s[1], w1
; CHECK-GI-FP16-NEXT: mov v2.s[1], w9
; CHECK-GI-FP16-NEXT: mov v5.s[1], v6.s[0]
; CHECK-GI-FP16-NEXT: ldr s6, [sp, #24]
; CHECK-GI-FP16-NEXT: mov v7.s[1], w5
; CHECK-GI-FP16-NEXT: mov v6.s[1], v16.s[0]
; CHECK-GI-FP16-NEXT: ldr s16, [sp, #40]
; CHECK-GI-FP16-NEXT: mov v1.s[0], w8
; CHECK-GI-FP16-NEXT: umov w8, v0.h[6]
; CHECK-GI-FP16-NEXT: ushll v0.4s, v0.4h, #0
; CHECK-GI-FP16-NEXT: mov v2.s[2], w9
; CHECK-GI-FP16-NEXT: mov v4.s[2], w2
; CHECK-GI-FP16-NEXT: mov v5.s[2], v17.s[0]
; CHECK-GI-FP16-NEXT: mov v7.s[2], w6
; CHECK-GI-FP16-NEXT: shl v0.4s, v0.4s, #31
; CHECK-GI-FP16-NEXT: mov v6.s[2], v16.s[0]
; CHECK-GI-FP16-NEXT: mov v1.s[1], w10
; CHECK-GI-FP16-NEXT: mov w10, #-1 // =0xffffffff
; CHECK-GI-FP16-NEXT: mov v3.s[0], w10
; CHECK-GI-FP16-NEXT: mov v4.s[3], w3
; CHECK-GI-FP16-NEXT: sshr v0.4s, v0.4s, #31
; CHECK-GI-FP16-NEXT: mov v1.s[2], w8
; CHECK-GI-FP16-NEXT: mov v3.s[1], w10
; CHECK-GI-FP16-NEXT: ushl v1.4s, v1.4s, v2.4s
; CHECK-GI-FP16-NEXT: neg v2.4s, v2.4s
; CHECK-GI-FP16-NEXT: mov v3.s[2], w10
; CHECK-GI-FP16-NEXT: sshl v1.4s, v1.4s, v2.4s
; CHECK-GI-FP16-NEXT: ldr s2, [sp, #16]
; CHECK-GI-FP16-NEXT: mov v5.s[3], v2.s[0]
; CHECK-GI-FP16-NEXT: eor v3.16b, v1.16b, v3.16b
; CHECK-GI-FP16-NEXT: and v1.16b, v7.16b, v1.16b
; CHECK-GI-FP16-NEXT: and v2.16b, v6.16b, v3.16b
; CHECK-GI-FP16-NEXT: bsl v0.16b, v4.16b, v5.16b
; CHECK-GI-FP16-NEXT: orr v1.16b, v1.16b, v2.16b
; CHECK-GI-FP16-NEXT: mov s2, v0.s[1]
; CHECK-GI-FP16-NEXT: mov s3, v0.s[2]
; CHECK-GI-FP16-NEXT: mov s4, v0.s[3]
; CHECK-GI-FP16-NEXT: fmov w0, s0
; CHECK-GI-FP16-NEXT: mov s5, v1.s[1]
; CHECK-GI-FP16-NEXT: mov s6, v1.s[2]
; CHECK-GI-FP16-NEXT: fmov w4, s1
; CHECK-GI-FP16-NEXT: fmov w1, s2
; CHECK-GI-FP16-NEXT: fmov w2, s3
; CHECK-GI-FP16-NEXT: fmov w3, s4
; CHECK-GI-FP16-NEXT: fmov w5, s5
; CHECK-GI-FP16-NEXT: fmov w6, s6
; CHECK-GI-FP16-NEXT: ret
entry:
%c = fcmp olt <7 x half> %a, %b
%s = select <7 x i1> %c, <7 x i32> %d, <7 x i32> %e
ret <7 x i32> %s
}
define <4 x i32> @v4f16_i32(<4 x half> %a, <4 x half> %b, <4 x i32> %d, <4 x i32> %e) {
; CHECK-SD-NOFP16-LABEL: v4f16_i32:
; CHECK-SD-NOFP16: // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT: fcvtl v0.4s, v0.4h
; CHECK-SD-NOFP16-NEXT: fcvtl v1.4s, v1.4h
; CHECK-SD-NOFP16-NEXT: fcmgt v0.4s, v1.4s, v0.4s
; CHECK-SD-NOFP16-NEXT: bsl v0.16b, v2.16b, v3.16b
; CHECK-SD-NOFP16-NEXT: ret
;
; CHECK-SD-FP16-LABEL: v4f16_i32:
; CHECK-SD-FP16: // %bb.0: // %entry
; CHECK-SD-FP16-NEXT: fcmgt v0.4h, v1.4h, v0.4h
; CHECK-SD-FP16-NEXT: sshll v0.4s, v0.4h, #0
; CHECK-SD-FP16-NEXT: bsl v0.16b, v2.16b, v3.16b
; CHECK-SD-FP16-NEXT: ret
;
; CHECK-GI-NOFP16-LABEL: v4f16_i32:
; CHECK-GI-NOFP16: // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT: fcvtl v0.4s, v0.4h
; CHECK-GI-NOFP16-NEXT: fcvtl v1.4s, v1.4h
; CHECK-GI-NOFP16-NEXT: fcmgt v0.4s, v1.4s, v0.4s
; CHECK-GI-NOFP16-NEXT: bsl v0.16b, v2.16b, v3.16b
; CHECK-GI-NOFP16-NEXT: ret
;
; CHECK-GI-FP16-LABEL: v4f16_i32:
; CHECK-GI-FP16: // %bb.0: // %entry
; CHECK-GI-FP16-NEXT: fcmgt v0.4h, v1.4h, v0.4h
; CHECK-GI-FP16-NEXT: ushll v0.4s, v0.4h, #0
; CHECK-GI-FP16-NEXT: shl v0.4s, v0.4s, #31
; CHECK-GI-FP16-NEXT: sshr v0.4s, v0.4s, #31
; CHECK-GI-FP16-NEXT: bsl v0.16b, v2.16b, v3.16b
; CHECK-GI-FP16-NEXT: ret
entry:
%c = fcmp olt <4 x half> %a, %b
%s = select <4 x i1> %c, <4 x i32> %d, <4 x i32> %e
ret <4 x i32> %s
}
define <8 x i32> @v8f16_i32(<8 x half> %a, <8 x half> %b, <8 x i32> %d, <8 x i32> %e) {
; CHECK-SD-NOFP16-LABEL: v8f16_i32:
; CHECK-SD-NOFP16: // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT: mov h6, v1.h[5]
; CHECK-SD-NOFP16-NEXT: mov h7, v0.h[5]
; CHECK-SD-NOFP16-NEXT: mov h16, v1.h[4]
; CHECK-SD-NOFP16-NEXT: mov h17, v0.h[4]
; CHECK-SD-NOFP16-NEXT: mov h18, v1.h[6]
; CHECK-SD-NOFP16-NEXT: mov h19, v0.h[6]
; CHECK-SD-NOFP16-NEXT: fcvt s6, h6
; CHECK-SD-NOFP16-NEXT: fcvt s7, h7
; CHECK-SD-NOFP16-NEXT: fcvt s16, h16
; CHECK-SD-NOFP16-NEXT: fcvt s17, h17
; CHECK-SD-NOFP16-NEXT: fcvt s18, h18
; CHECK-SD-NOFP16-NEXT: fcvt s19, h19
; CHECK-SD-NOFP16-NEXT: fcmp s7, s6
; CHECK-SD-NOFP16-NEXT: mov h6, v1.h[7]
; CHECK-SD-NOFP16-NEXT: mov h7, v0.h[7]
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s17, s16
; CHECK-SD-NOFP16-NEXT: mov h16, v1.h[1]
; CHECK-SD-NOFP16-NEXT: mov h17, v0.h[1]
; CHECK-SD-NOFP16-NEXT: fcvt s6, h6
; CHECK-SD-NOFP16-NEXT: fcvt s7, h7
; CHECK-SD-NOFP16-NEXT: csetm w9, mi
; CHECK-SD-NOFP16-NEXT: fcmp s19, s18
; CHECK-SD-NOFP16-NEXT: fcvt s16, h16
; CHECK-SD-NOFP16-NEXT: fcvt s17, h17
; CHECK-SD-NOFP16-NEXT: csetm w10, mi
; CHECK-SD-NOFP16-NEXT: fcmp s7, s6
; CHECK-SD-NOFP16-NEXT: fcvt s6, h1
; CHECK-SD-NOFP16-NEXT: fcvt s7, h0
; CHECK-SD-NOFP16-NEXT: csetm w11, mi
; CHECK-SD-NOFP16-NEXT: fcmp s17, s16
; CHECK-SD-NOFP16-NEXT: mov h16, v1.h[2]
; CHECK-SD-NOFP16-NEXT: mov h17, v0.h[2]
; CHECK-SD-NOFP16-NEXT: mov h1, v1.h[3]
; CHECK-SD-NOFP16-NEXT: mov h0, v0.h[3]
; CHECK-SD-NOFP16-NEXT: csetm w12, mi
; CHECK-SD-NOFP16-NEXT: fcmp s7, s6
; CHECK-SD-NOFP16-NEXT: fcvt s6, h16
; CHECK-SD-NOFP16-NEXT: fmov s16, w9
; CHECK-SD-NOFP16-NEXT: fcvt s7, h17
; CHECK-SD-NOFP16-NEXT: fcvt s1, h1
; CHECK-SD-NOFP16-NEXT: fcvt s0, h0
; CHECK-SD-NOFP16-NEXT: csetm w13, mi
; CHECK-SD-NOFP16-NEXT: fmov s17, w13
; CHECK-SD-NOFP16-NEXT: mov v16.h[1], w8
; CHECK-SD-NOFP16-NEXT: fcmp s7, s6
; CHECK-SD-NOFP16-NEXT: mov v17.h[1], w12
; CHECK-SD-NOFP16-NEXT: mov v16.h[2], w10
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s0, s1
; CHECK-SD-NOFP16-NEXT: mov v17.h[2], w8
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: mov v16.h[3], w11
; CHECK-SD-NOFP16-NEXT: mov v17.h[3], w8
; CHECK-SD-NOFP16-NEXT: sshll v1.4s, v16.4h, #0
; CHECK-SD-NOFP16-NEXT: sshll v0.4s, v17.4h, #0
; CHECK-SD-NOFP16-NEXT: bsl v1.16b, v3.16b, v5.16b
; CHECK-SD-NOFP16-NEXT: bsl v0.16b, v2.16b, v4.16b
; CHECK-SD-NOFP16-NEXT: ret
;
; CHECK-SD-FP16-LABEL: v8f16_i32:
; CHECK-SD-FP16: // %bb.0: // %entry
; CHECK-SD-FP16-NEXT: fcmgt v0.8h, v1.8h, v0.8h
; CHECK-SD-FP16-NEXT: sshll v6.4s, v0.4h, #0
; CHECK-SD-FP16-NEXT: sshll2 v0.4s, v0.8h, #0
; CHECK-SD-FP16-NEXT: mov v1.16b, v0.16b
; CHECK-SD-FP16-NEXT: mov v0.16b, v6.16b
; CHECK-SD-FP16-NEXT: bsl v1.16b, v3.16b, v5.16b
; CHECK-SD-FP16-NEXT: bsl v0.16b, v2.16b, v4.16b
; CHECK-SD-FP16-NEXT: ret
;
; CHECK-GI-NOFP16-LABEL: v8f16_i32:
; CHECK-GI-NOFP16: // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT: fcvtl v6.4s, v0.4h
; CHECK-GI-NOFP16-NEXT: fcvtl v7.4s, v1.4h
; CHECK-GI-NOFP16-NEXT: fcvtl2 v0.4s, v0.8h
; CHECK-GI-NOFP16-NEXT: fcvtl2 v1.4s, v1.8h
; CHECK-GI-NOFP16-NEXT: fcmgt v6.4s, v7.4s, v6.4s
; CHECK-GI-NOFP16-NEXT: fcmgt v1.4s, v1.4s, v0.4s
; CHECK-GI-NOFP16-NEXT: mov v0.16b, v6.16b
; CHECK-GI-NOFP16-NEXT: bsl v1.16b, v3.16b, v5.16b
; CHECK-GI-NOFP16-NEXT: bsl v0.16b, v2.16b, v4.16b
; CHECK-GI-NOFP16-NEXT: ret
;
; CHECK-GI-FP16-LABEL: v8f16_i32:
; CHECK-GI-FP16: // %bb.0: // %entry
; CHECK-GI-FP16-NEXT: fcmgt v0.8h, v1.8h, v0.8h
; CHECK-GI-FP16-NEXT: ushll v1.4s, v0.4h, #0
; CHECK-GI-FP16-NEXT: ushll2 v0.4s, v0.8h, #0
; CHECK-GI-FP16-NEXT: shl v1.4s, v1.4s, #31
; CHECK-GI-FP16-NEXT: shl v0.4s, v0.4s, #31
; CHECK-GI-FP16-NEXT: sshr v1.4s, v1.4s, #31
; CHECK-GI-FP16-NEXT: sshr v6.4s, v0.4s, #31
; CHECK-GI-FP16-NEXT: mov v0.16b, v1.16b
; CHECK-GI-FP16-NEXT: mov v1.16b, v6.16b
; CHECK-GI-FP16-NEXT: bsl v0.16b, v2.16b, v4.16b
; CHECK-GI-FP16-NEXT: bsl v1.16b, v3.16b, v5.16b
; CHECK-GI-FP16-NEXT: ret
entry:
%c = fcmp olt <8 x half> %a, %b
%s = select <8 x i1> %c, <8 x i32> %d, <8 x i32> %e
ret <8 x i32> %s
}
define <16 x i32> @v16f16_i32(<16 x half> %a, <16 x half> %b, <16 x i32> %d, <16 x i32> %e) {
; CHECK-SD-NOFP16-LABEL: v16f16_i32:
; CHECK-SD-NOFP16: // %bb.0: // %entry
; CHECK-SD-NOFP16-NEXT: mov h16, v3.h[5]
; CHECK-SD-NOFP16-NEXT: mov h17, v1.h[5]
; CHECK-SD-NOFP16-NEXT: mov h18, v3.h[4]
; CHECK-SD-NOFP16-NEXT: mov h19, v1.h[4]
; CHECK-SD-NOFP16-NEXT: mov h20, v3.h[6]
; CHECK-SD-NOFP16-NEXT: mov h21, v1.h[6]
; CHECK-SD-NOFP16-NEXT: fcvt s16, h16
; CHECK-SD-NOFP16-NEXT: fcvt s17, h17
; CHECK-SD-NOFP16-NEXT: fcvt s18, h18
; CHECK-SD-NOFP16-NEXT: fcvt s19, h19
; CHECK-SD-NOFP16-NEXT: fcvt s20, h20
; CHECK-SD-NOFP16-NEXT: fcvt s21, h21
; CHECK-SD-NOFP16-NEXT: fcmp s17, s16
; CHECK-SD-NOFP16-NEXT: mov h16, v3.h[7]
; CHECK-SD-NOFP16-NEXT: mov h17, v1.h[7]
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s19, s18
; CHECK-SD-NOFP16-NEXT: mov h18, v3.h[1]
; CHECK-SD-NOFP16-NEXT: mov h19, v1.h[1]
; CHECK-SD-NOFP16-NEXT: fcvt s16, h16
; CHECK-SD-NOFP16-NEXT: fcvt s17, h17
; CHECK-SD-NOFP16-NEXT: csetm w10, mi
; CHECK-SD-NOFP16-NEXT: fcmp s21, s20
; CHECK-SD-NOFP16-NEXT: fcvt s20, h3
; CHECK-SD-NOFP16-NEXT: fcvt s18, h18
; CHECK-SD-NOFP16-NEXT: fcvt s21, h1
; CHECK-SD-NOFP16-NEXT: fcvt s19, h19
; CHECK-SD-NOFP16-NEXT: csetm w9, mi
; CHECK-SD-NOFP16-NEXT: fcmp s17, s16
; CHECK-SD-NOFP16-NEXT: mov h16, v3.h[2]
; CHECK-SD-NOFP16-NEXT: mov h17, v1.h[2]
; CHECK-SD-NOFP16-NEXT: mov h3, v3.h[3]
; CHECK-SD-NOFP16-NEXT: mov h1, v1.h[3]
; CHECK-SD-NOFP16-NEXT: csetm w11, mi
; CHECK-SD-NOFP16-NEXT: fcmp s19, s18
; CHECK-SD-NOFP16-NEXT: mov h18, v2.h[5]
; CHECK-SD-NOFP16-NEXT: fcvt s16, h16
; CHECK-SD-NOFP16-NEXT: mov h19, v0.h[5]
; CHECK-SD-NOFP16-NEXT: fcvt s17, h17
; CHECK-SD-NOFP16-NEXT: fcvt s3, h3
; CHECK-SD-NOFP16-NEXT: fcvt s1, h1
; CHECK-SD-NOFP16-NEXT: csetm w12, mi
; CHECK-SD-NOFP16-NEXT: fcmp s21, s20
; CHECK-SD-NOFP16-NEXT: fcvt s18, h18
; CHECK-SD-NOFP16-NEXT: fcvt s19, h19
; CHECK-SD-NOFP16-NEXT: csetm w14, mi
; CHECK-SD-NOFP16-NEXT: fcmp s17, s16
; CHECK-SD-NOFP16-NEXT: mov h16, v2.h[4]
; CHECK-SD-NOFP16-NEXT: mov h17, v0.h[4]
; CHECK-SD-NOFP16-NEXT: csetm w13, mi
; CHECK-SD-NOFP16-NEXT: fcmp s1, s3
; CHECK-SD-NOFP16-NEXT: mov h1, v2.h[6]
; CHECK-SD-NOFP16-NEXT: mov h3, v0.h[6]
; CHECK-SD-NOFP16-NEXT: fcvt s16, h16
; CHECK-SD-NOFP16-NEXT: fcvt s17, h17
; CHECK-SD-NOFP16-NEXT: csetm w15, mi
; CHECK-SD-NOFP16-NEXT: fcmp s19, s18
; CHECK-SD-NOFP16-NEXT: mov h18, v2.h[7]
; CHECK-SD-NOFP16-NEXT: mov h19, v0.h[7]
; CHECK-SD-NOFP16-NEXT: fcvt s1, h1
; CHECK-SD-NOFP16-NEXT: fcvt s3, h3
; CHECK-SD-NOFP16-NEXT: csetm w16, mi
; CHECK-SD-NOFP16-NEXT: fcmp s17, s16
; CHECK-SD-NOFP16-NEXT: mov h16, v2.h[1]
; CHECK-SD-NOFP16-NEXT: mov h17, v0.h[1]
; CHECK-SD-NOFP16-NEXT: fcvt s18, h18
; CHECK-SD-NOFP16-NEXT: fcvt s19, h19
; CHECK-SD-NOFP16-NEXT: csetm w17, mi
; CHECK-SD-NOFP16-NEXT: fcmp s3, s1
; CHECK-SD-NOFP16-NEXT: fcvt s1, h16
; CHECK-SD-NOFP16-NEXT: fcvt s16, h2
; CHECK-SD-NOFP16-NEXT: fcvt s3, h17
; CHECK-SD-NOFP16-NEXT: fcvt s17, h0
; CHECK-SD-NOFP16-NEXT: csetm w18, mi
; CHECK-SD-NOFP16-NEXT: fcmp s19, s18
; CHECK-SD-NOFP16-NEXT: fmov s18, w14
; CHECK-SD-NOFP16-NEXT: fmov s19, w17
; CHECK-SD-NOFP16-NEXT: csetm w0, mi
; CHECK-SD-NOFP16-NEXT: fcmp s3, s1
; CHECK-SD-NOFP16-NEXT: mov h1, v2.h[2]
; CHECK-SD-NOFP16-NEXT: mov h3, v0.h[2]
; CHECK-SD-NOFP16-NEXT: mov h2, v2.h[3]
; CHECK-SD-NOFP16-NEXT: mov h0, v0.h[3]
; CHECK-SD-NOFP16-NEXT: mov v18.h[1], w12
; CHECK-SD-NOFP16-NEXT: mov v19.h[1], w16
; CHECK-SD-NOFP16-NEXT: csetm w1, mi
; CHECK-SD-NOFP16-NEXT: fcmp s17, s16
; CHECK-SD-NOFP16-NEXT: fmov s16, w10
; CHECK-SD-NOFP16-NEXT: fcvt s1, h1
; CHECK-SD-NOFP16-NEXT: fcvt s3, h3
; CHECK-SD-NOFP16-NEXT: fcvt s2, h2
; CHECK-SD-NOFP16-NEXT: fcvt s0, h0
; CHECK-SD-NOFP16-NEXT: csetm w2, mi
; CHECK-SD-NOFP16-NEXT: mov v16.h[1], w8
; CHECK-SD-NOFP16-NEXT: mov v18.h[2], w13
; CHECK-SD-NOFP16-NEXT: fmov s17, w2
; CHECK-SD-NOFP16-NEXT: mov v19.h[2], w18
; CHECK-SD-NOFP16-NEXT: fcmp s3, s1
; CHECK-SD-NOFP16-NEXT: mov v17.h[1], w1
; CHECK-SD-NOFP16-NEXT: mov v16.h[2], w9
; CHECK-SD-NOFP16-NEXT: mov v18.h[3], w15
; CHECK-SD-NOFP16-NEXT: mov v19.h[3], w0
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: fcmp s0, s2
; CHECK-SD-NOFP16-NEXT: mov v17.h[2], w8
; CHECK-SD-NOFP16-NEXT: mov v16.h[3], w11
; CHECK-SD-NOFP16-NEXT: csetm w8, mi
; CHECK-SD-NOFP16-NEXT: mov v17.h[3], w8
; CHECK-SD-NOFP16-NEXT: sshll v2.4s, v16.4h, #0
; CHECK-SD-NOFP16-NEXT: sshll v16.4s, v18.4h, #0
; CHECK-SD-NOFP16-NEXT: ldp q0, q18, [sp]
; CHECK-SD-NOFP16-NEXT: sshll v1.4s, v17.4h, #0
; CHECK-SD-NOFP16-NEXT: sshll v17.4s, v19.4h, #0
; CHECK-SD-NOFP16-NEXT: ldp q19, q3, [sp, #32]
; CHECK-SD-NOFP16-NEXT: bit v0.16b, v4.16b, v1.16b
; CHECK-SD-NOFP16-NEXT: mov v1.16b, v17.16b
; CHECK-SD-NOFP16-NEXT: bit v3.16b, v7.16b, v2.16b
; CHECK-SD-NOFP16-NEXT: mov v2.16b, v16.16b
; CHECK-SD-NOFP16-NEXT: bsl v1.16b, v5.16b, v18.16b
; CHECK-SD-NOFP16-NEXT: bsl v2.16b, v6.16b, v19.16b
; CHECK-SD-NOFP16-NEXT: ret
;
; CHECK-SD-FP16-LABEL: v16f16_i32:
; CHECK-SD-FP16: // %bb.0: // %entry
; CHECK-SD-FP16-NEXT: fcmgt v0.8h, v2.8h, v0.8h
; CHECK-SD-FP16-NEXT: fcmgt v1.8h, v3.8h, v1.8h
; CHECK-SD-FP16-NEXT: ldp q2, q20, [sp]
; CHECK-SD-FP16-NEXT: ldp q18, q19, [sp, #32]
; CHECK-SD-FP16-NEXT: sshll v3.4s, v0.4h, #0
; CHECK-SD-FP16-NEXT: sshll v16.4s, v1.4h, #0
; CHECK-SD-FP16-NEXT: sshll2 v17.4s, v1.8h, #0
; CHECK-SD-FP16-NEXT: sshll2 v1.4s, v0.8h, #0
; CHECK-SD-FP16-NEXT: mov v0.16b, v3.16b
; CHECK-SD-FP16-NEXT: mov v3.16b, v17.16b
; CHECK-SD-FP16-NEXT: bsl v1.16b, v5.16b, v20.16b
; CHECK-SD-FP16-NEXT: bsl v0.16b, v4.16b, v2.16b
; CHECK-SD-FP16-NEXT: mov v2.16b, v16.16b
; CHECK-SD-FP16-NEXT: bsl v3.16b, v7.16b, v19.16b
; CHECK-SD-FP16-NEXT: bsl v2.16b, v6.16b, v18.16b
; CHECK-SD-FP16-NEXT: ret
;
; CHECK-GI-NOFP16-LABEL: v16f16_i32:
; CHECK-GI-NOFP16: // %bb.0: // %entry
; CHECK-GI-NOFP16-NEXT: fcvtl v16.4s, v0.4h
; CHECK-GI-NOFP16-NEXT: fcvtl2 v0.4s, v0.8h
; CHECK-GI-NOFP16-NEXT: fcvtl v17.4s, v1.4h
; CHECK-GI-NOFP16-NEXT: fcvtl v18.4s, v2.4h
; CHECK-GI-NOFP16-NEXT: fcvtl2 v2.4s, v2.8h
; CHECK-GI-NOFP16-NEXT: fcvtl v19.4s, v3.4h
; CHECK-GI-NOFP16-NEXT: fcvtl2 v1.4s, v1.8h
; CHECK-GI-NOFP16-NEXT: fcvtl2 v3.4s, v3.8h
; CHECK-GI-NOFP16-NEXT: fcmgt v2.4s, v2.4s, v0.4s
; CHECK-GI-NOFP16-NEXT: fcmgt v17.4s, v19.4s, v17.4s
; CHECK-GI-NOFP16-NEXT: fcmgt v16.4s, v18.4s, v16.4s
; CHECK-GI-NOFP16-NEXT: fcmgt v3.4s, v3.4s, v1.4s
; CHECK-GI-NOFP16-NEXT: ldp q0, q1, [sp]
; CHECK-GI-NOFP16-NEXT: ldp q18, q19, [sp, #32]
; CHECK-GI-NOFP16-NEXT: bit v1.16b, v5.16b, v2.16b
; CHECK-GI-NOFP16-NEXT: mov v2.16b, v17.16b
; CHECK-GI-NOFP16-NEXT: bit v0.16b, v4.16b, v16.16b
; CHECK-GI-NOFP16-NEXT: bsl v3.16b, v7.16b, v19.16b
; CHECK-GI-NOFP16-NEXT: bsl v2.16b, v6.16b, v18.16b
; CHECK-GI-NOFP16-NEXT: ret
;
; CHECK-GI-FP16-LABEL: v16f16_i32:
; CHECK-GI-FP16: // %bb.0: // %entry
; CHECK-GI-FP16-NEXT: fcmgt v0.8h, v2.8h, v0.8h
; CHECK-GI-FP16-NEXT: fcmgt v1.8h, v3.8h, v1.8h
; CHECK-GI-FP16-NEXT: ldp q18, q19, [sp, #32]
; CHECK-GI-FP16-NEXT: ushll v2.4s, v0.4h, #0
; CHECK-GI-FP16-NEXT: ushll2 v0.4s, v0.8h, #0
; CHECK-GI-FP16-NEXT: ushll v3.4s, v1.4h, #0
; CHECK-GI-FP16-NEXT: ushll2 v1.4s, v1.8h, #0
; CHECK-GI-FP16-NEXT: shl v2.4s, v2.4s, #31
; CHECK-GI-FP16-NEXT: shl v0.4s, v0.4s, #31
; CHECK-GI-FP16-NEXT: shl v3.4s, v3.4s, #31
; CHECK-GI-FP16-NEXT: shl v1.4s, v1.4s, #31
; CHECK-GI-FP16-NEXT: sshr v2.4s, v2.4s, #31
; CHECK-GI-FP16-NEXT: sshr v16.4s, v0.4s, #31
; CHECK-GI-FP16-NEXT: sshr v3.4s, v3.4s, #31
; CHECK-GI-FP16-NEXT: sshr v17.4s, v1.4s, #31
; CHECK-GI-FP16-NEXT: ldp q0, q1, [sp]
; CHECK-GI-FP16-NEXT: bit v0.16b, v4.16b, v2.16b
; CHECK-GI-FP16-NEXT: mov v2.16b, v3.16b
; CHECK-GI-FP16-NEXT: mov v3.16b, v17.16b
; CHECK-GI-FP16-NEXT: bit v1.16b, v5.16b, v16.16b
; CHECK-GI-FP16-NEXT: bsl v2.16b, v6.16b, v18.16b
; CHECK-GI-FP16-NEXT: bsl v3.16b, v7.16b, v19.16b
; CHECK-GI-FP16-NEXT: ret
entry:
%c = fcmp olt <16 x half> %a, %b
%s = select <16 x i1> %c, <16 x i32> %d, <16 x i32> %e
ret <16 x i32> %s
}