shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
llvm-project/llvm/test/CodeGen/AArch64/fp-maximumnum-minimumnum.ll
Jonathan Cohen c6fe567064
[AArch64][MachineCombiner] Combine sequences of gather patterns (#152979) 
Reland of #142941

Squashed with fixes for #150004, #149585

This pattern matches gather-like patterns where
values are loaded per lane into neon registers, and 
replaces it with loads into 2 separate registers, which
will be combined with a zip instruction. This decreases
the critical path length and improves Memory Level
Parallelism.

rdar://151851094
2025-08-18 15:10:59 +03:00

2242 lines
80 KiB
LLVM
Raw Permalink Blame History

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc --mtriple=aarch64 --mattr=+fullfp16 < %s | FileCheck %s --check-prefixes=AARCH64,FULLFP16
; RUN: llc --mtriple=aarch64 < %s | FileCheck %s --check-prefixes=AARCH64,NOFULLFP16
;;;;;;;;;;;;;;;; max_f64
define double @max_nnan_f64(double %a, double %b) {
; AARCH64-LABEL: max_nnan_f64:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fmaxnm d0, d0, d1
; AARCH64-NEXT: ret
entry:
%c = call nnan double @llvm.maximumnum.f64(double %a, double %b)
ret double %c
}
define <2 x double> @max_nnan_v2f64(<2 x double> %a, <2 x double> %b) {
; AARCH64-LABEL: max_nnan_v2f64:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fmaxnm v0.2d, v0.2d, v1.2d
; AARCH64-NEXT: ret
entry:
%c = call nnan <2 x double> @llvm.maximumnum.v2f64(<2 x double> %a, <2 x double> %b)
ret <2 x double> %c
}
define <3 x double> @max_nnan_v3f64(<3 x double> %a, <3 x double> %b) {
; AARCH64-LABEL: max_nnan_v3f64:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: // kill: def $d3 killed $d3 def $q3
; AARCH64-NEXT: // kill: def $d0 killed $d0 def $q0
; AARCH64-NEXT: // kill: def $d4 killed $d4 def $q4
; AARCH64-NEXT: // kill: def $d1 killed $d1 def $q1
; AARCH64-NEXT: // kill: def $d2 killed $d2 def $q2
; AARCH64-NEXT: // kill: def $d5 killed $d5 def $q5
; AARCH64-NEXT: mov v3.d[1], v4.d[0]
; AARCH64-NEXT: mov v0.d[1], v1.d[0]
; AARCH64-NEXT: fmaxnm v2.2d, v2.2d, v5.2d
; AARCH64-NEXT: // kill: def $d2 killed $d2 killed $q2
; AARCH64-NEXT: fmaxnm v0.2d, v0.2d, v3.2d
; AARCH64-NEXT: ext v1.16b, v0.16b, v0.16b, #8
; AARCH64-NEXT: // kill: def $d0 killed $d0 killed $q0
; AARCH64-NEXT: // kill: def $d1 killed $d1 killed $q1
; AARCH64-NEXT: ret
entry:
%c = call nnan <3 x double> @llvm.maximumnum.v3f64(<3 x double> %a, <3 x double> %b)
ret <3 x double> %c
}
define <4 x double> @max_nnan_v4f64(<4 x double> %a, <4 x double> %b) {
; AARCH64-LABEL: max_nnan_v4f64:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fmaxnm v1.2d, v1.2d, v3.2d
; AARCH64-NEXT: fmaxnm v0.2d, v0.2d, v2.2d
; AARCH64-NEXT: ret
entry:
%c = call nnan <4 x double> @llvm.maximumnum.v4f64(<4 x double> %a, <4 x double> %b)
ret <4 x double> %c
}
;;;;;;;;;;;;;;;;;; max_f32
define float @max_nnan_f32(float %a, float %b) {
; AARCH64-LABEL: max_nnan_f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fmaxnm s0, s0, s1
; AARCH64-NEXT: ret
entry:
%c = call nnan float @llvm.maximumnum.f32(float %a, float %b)
ret float %c
}
define <2 x float> @max_nnan_v2f32(<2 x float> %a, <2 x float> %b) {
; AARCH64-LABEL: max_nnan_v2f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fmaxnm v0.2s, v0.2s, v1.2s
; AARCH64-NEXT: ret
entry:
%c = call nnan <2 x float> @llvm.maximumnum.v2f32(<2 x float> %a, <2 x float> %b)
ret <2 x float> %c
}
define <3 x float> @max_nnan_v3f32(<3 x float> %a, <3 x float> %b) {
; AARCH64-LABEL: max_nnan_v3f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fmaxnm v0.4s, v0.4s, v1.4s
; AARCH64-NEXT: ret
entry:
%c = call nnan <3 x float> @llvm.maximumnum.v3f32(<3 x float> %a, <3 x float> %b)
ret <3 x float> %c
}
define <4 x float> @max_nnan_v4f32(<4 x float> %a, <4 x float> %b) {
; AARCH64-LABEL: max_nnan_v4f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fmaxnm v0.4s, v0.4s, v1.4s
; AARCH64-NEXT: ret
entry:
%c = call nnan <4 x float> @llvm.maximumnum.v4f32(<4 x float> %a, <4 x float> %b)
ret <4 x float> %c
}
define <5 x float> @max_nnan_v5f32(<5 x float> %a, <5 x float> %b) {
; AARCH64-LABEL: max_nnan_v5f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: // kill: def $s0 killed $s0 def $q0
; AARCH64-NEXT: // kill: def $s5 killed $s5 def $q5
; AARCH64-NEXT: // kill: def $s1 killed $s1 def $q1
; AARCH64-NEXT: // kill: def $s6 killed $s6 def $q6
; AARCH64-NEXT: // kill: def $s2 killed $s2 def $q2
; AARCH64-NEXT: // kill: def $s7 killed $s7 def $q7
; AARCH64-NEXT: // kill: def $s3 killed $s3 def $q3
; AARCH64-NEXT: mov x8, sp
; AARCH64-NEXT: // kill: def $s4 killed $s4 def $q4
; AARCH64-NEXT: mov v0.s[1], v1.s[0]
; AARCH64-NEXT: mov v5.s[1], v6.s[0]
; AARCH64-NEXT: mov v0.s[2], v2.s[0]
; AARCH64-NEXT: mov v5.s[2], v7.s[0]
; AARCH64-NEXT: ldr s2, [sp, #8]
; AARCH64-NEXT: fmaxnm v4.4s, v4.4s, v2.4s
; AARCH64-NEXT: // kill: def $s4 killed $s4 killed $q4
; AARCH64-NEXT: mov v0.s[3], v3.s[0]
; AARCH64-NEXT: ld1 { v5.s }[3], [x8]
; AARCH64-NEXT: fmaxnm v0.4s, v0.4s, v5.4s
; AARCH64-NEXT: mov s1, v0.s[1]
; AARCH64-NEXT: mov s2, v0.s[2]
; AARCH64-NEXT: mov s3, v0.s[3]
; AARCH64-NEXT: // kill: def $s0 killed $s0 killed $q0
; AARCH64-NEXT: ret
entry:
%c = call nnan <5 x float> @llvm.maximumnum.v5f32(<5 x float> %a, <5 x float> %b)
ret <5 x float> %c
}
define <8 x float> @max_nnan_v8f32(<8 x float> %a, <8 x float> %b) {
; AARCH64-LABEL: max_nnan_v8f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fmaxnm v1.4s, v1.4s, v3.4s
; AARCH64-NEXT: fmaxnm v0.4s, v0.4s, v2.4s
; AARCH64-NEXT: ret
entry:
%c = call nnan <8 x float> @llvm.maximumnum.v8f32(<8 x float> %a, <8 x float> %b)
ret <8 x float> %c
}
;;;;;;;;;;;;;;;;;; max_f16
define half @max_nnan_f16(half %a, half %b) {
; FULLFP16-LABEL: max_nnan_f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fmaxnm h0, h0, h1
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: max_nnan_f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fcvt s0, h0
; NOFULLFP16-NEXT: fmaxnm s0, s0, s1
; NOFULLFP16-NEXT: fcvt h0, s0
; NOFULLFP16-NEXT: ret
entry:
%c = call nnan half @llvm.maximumnum.f16(half %a, half %b)
ret half %c
}
define <2 x half> @max_nnan_v2f16(<2 x half> %a, <2 x half> %b) {
; FULLFP16-LABEL: max_nnan_v2f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fmaxnm v0.4h, v0.4h, v1.4h
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: max_nnan_v2f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: // kill: def $d1 killed $d1 def $q1
; NOFULLFP16-NEXT: // kill: def $d0 killed $d0 def $q0
; NOFULLFP16-NEXT: mov h2, v1.h[1]
; NOFULLFP16-NEXT: mov h3, v0.h[1]
; NOFULLFP16-NEXT: mov h4, v1.h[2]
; NOFULLFP16-NEXT: mov h5, v0.h[2]
; NOFULLFP16-NEXT: fcvt s6, h1
; NOFULLFP16-NEXT: fcvt s7, h0
; NOFULLFP16-NEXT: mov h1, v1.h[3]
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fmaxnm s2, s3, s2
; NOFULLFP16-NEXT: fcvt s3, h4
; NOFULLFP16-NEXT: fcvt s4, h5
; NOFULLFP16-NEXT: fmaxnm s5, s7, s6
; NOFULLFP16-NEXT: mov h6, v0.h[3]
; NOFULLFP16-NEXT: fmaxnm s3, s4, s3
; NOFULLFP16-NEXT: fcvt h2, s2
; NOFULLFP16-NEXT: fcvt h0, s5
; NOFULLFP16-NEXT: fcvt s4, h6
; NOFULLFP16-NEXT: mov v0.h[1], v2.h[0]
; NOFULLFP16-NEXT: fcvt h2, s3
; NOFULLFP16-NEXT: fmaxnm s1, s4, s1
; NOFULLFP16-NEXT: mov v0.h[2], v2.h[0]
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: mov v0.h[3], v1.h[0]
; NOFULLFP16-NEXT: // kill: def $d0 killed $d0 killed $q0
; NOFULLFP16-NEXT: ret
entry:
%c = call nnan <2 x half> @llvm.maximumnum.v2f16(<2 x half> %a, <2 x half> %b)
ret <2 x half> %c
}
define <4 x half> @max_nnan_v4f16(<4 x half> %a, <4 x half> %b) {
; FULLFP16-LABEL: max_nnan_v4f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fmaxnm v0.4h, v0.4h, v1.4h
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: max_nnan_v4f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: // kill: def $d1 killed $d1 def $q1
; NOFULLFP16-NEXT: // kill: def $d0 killed $d0 def $q0
; NOFULLFP16-NEXT: mov h2, v1.h[1]
; NOFULLFP16-NEXT: mov h3, v0.h[1]
; NOFULLFP16-NEXT: mov h4, v1.h[2]
; NOFULLFP16-NEXT: mov h5, v0.h[2]
; NOFULLFP16-NEXT: fcvt s6, h1
; NOFULLFP16-NEXT: fcvt s7, h0
; NOFULLFP16-NEXT: mov h1, v1.h[3]
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fmaxnm s2, s3, s2
; NOFULLFP16-NEXT: fcvt s3, h4
; NOFULLFP16-NEXT: fcvt s4, h5
; NOFULLFP16-NEXT: fmaxnm s5, s7, s6
; NOFULLFP16-NEXT: mov h6, v0.h[3]
; NOFULLFP16-NEXT: fmaxnm s3, s4, s3
; NOFULLFP16-NEXT: fcvt h2, s2
; NOFULLFP16-NEXT: fcvt h0, s5
; NOFULLFP16-NEXT: fcvt s4, h6
; NOFULLFP16-NEXT: mov v0.h[1], v2.h[0]
; NOFULLFP16-NEXT: fcvt h2, s3
; NOFULLFP16-NEXT: fmaxnm s1, s4, s1
; NOFULLFP16-NEXT: mov v0.h[2], v2.h[0]
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: mov v0.h[3], v1.h[0]
; NOFULLFP16-NEXT: // kill: def $d0 killed $d0 killed $q0
; NOFULLFP16-NEXT: ret
entry:
%c = call nnan <4 x half> @llvm.maximumnum.v4f16(<4 x half> %a, <4 x half> %b)
ret <4 x half> %c
}
define <8 x half> @max_nnan_v8f16(<8 x half> %a, <8 x half> %b) {
; FULLFP16-LABEL: max_nnan_v8f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fmaxnm v0.8h, v0.8h, v1.8h
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: max_nnan_v8f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: mov h2, v1.h[1]
; NOFULLFP16-NEXT: mov h3, v0.h[1]
; NOFULLFP16-NEXT: fcvt s4, h1
; NOFULLFP16-NEXT: fcvt s5, h0
; NOFULLFP16-NEXT: mov h6, v1.h[2]
; NOFULLFP16-NEXT: mov h7, v0.h[2]
; NOFULLFP16-NEXT: mov h16, v1.h[3]
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fmaxnm s4, s5, s4
; NOFULLFP16-NEXT: mov h5, v0.h[3]
; NOFULLFP16-NEXT: fcvt s6, h6
; NOFULLFP16-NEXT: fcvt s7, h7
; NOFULLFP16-NEXT: fcvt s16, h16
; NOFULLFP16-NEXT: fmaxnm s3, s3, s2
; NOFULLFP16-NEXT: fcvt s5, h5
; NOFULLFP16-NEXT: fcvt h2, s4
; NOFULLFP16-NEXT: fmaxnm s4, s7, s6
; NOFULLFP16-NEXT: mov h6, v1.h[4]
; NOFULLFP16-NEXT: mov h7, v0.h[4]
; NOFULLFP16-NEXT: fcvt h3, s3
; NOFULLFP16-NEXT: fmaxnm s5, s5, s16
; NOFULLFP16-NEXT: mov h16, v0.h[5]
; NOFULLFP16-NEXT: fcvt h4, s4
; NOFULLFP16-NEXT: mov v2.h[1], v3.h[0]
; NOFULLFP16-NEXT: fcvt s3, h6
; NOFULLFP16-NEXT: fcvt s6, h7
; NOFULLFP16-NEXT: mov h7, v1.h[5]
; NOFULLFP16-NEXT: fcvt h5, s5
; NOFULLFP16-NEXT: fcvt s16, h16
; NOFULLFP16-NEXT: mov v2.h[2], v4.h[0]
; NOFULLFP16-NEXT: mov h4, v1.h[6]
; NOFULLFP16-NEXT: fmaxnm s3, s6, s3
; NOFULLFP16-NEXT: mov h6, v0.h[6]
; NOFULLFP16-NEXT: fcvt s7, h7
; NOFULLFP16-NEXT: mov h1, v1.h[7]
; NOFULLFP16-NEXT: mov h0, v0.h[7]
; NOFULLFP16-NEXT: mov v2.h[3], v5.h[0]
; NOFULLFP16-NEXT: fcvt s4, h4
; NOFULLFP16-NEXT: fcvt h3, s3
; NOFULLFP16-NEXT: fcvt s5, h6
; NOFULLFP16-NEXT: fmaxnm s6, s16, s7
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fcvt s0, h0
; NOFULLFP16-NEXT: mov v2.h[4], v3.h[0]
; NOFULLFP16-NEXT: fmaxnm s4, s5, s4
; NOFULLFP16-NEXT: fcvt h3, s6
; NOFULLFP16-NEXT: fmaxnm s0, s0, s1
; NOFULLFP16-NEXT: mov v2.h[5], v3.h[0]
; NOFULLFP16-NEXT: fcvt h3, s4
; NOFULLFP16-NEXT: fcvt h0, s0
; NOFULLFP16-NEXT: mov v2.h[6], v3.h[0]
; NOFULLFP16-NEXT: mov v2.h[7], v0.h[0]
; NOFULLFP16-NEXT: mov v0.16b, v2.16b
; NOFULLFP16-NEXT: ret
entry:
%c = call nnan <8 x half> @llvm.maximumnum.v8f16(<8 x half> %a, <8 x half> %b)
ret <8 x half> %c
}
define <9 x half> @max_nnan_v9f16(<9 x half> %a, <9 x half> %b) {
; FULLFP16-LABEL: max_nnan_v9f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: // kill: def $h0 killed $h0 def $q0
; FULLFP16-NEXT: // kill: def $h1 killed $h1 def $q1
; FULLFP16-NEXT: // kill: def $h2 killed $h2 def $q2
; FULLFP16-NEXT: add x9, sp, #16
; FULLFP16-NEXT: // kill: def $h3 killed $h3 def $q3
; FULLFP16-NEXT: // kill: def $h4 killed $h4 def $q4
; FULLFP16-NEXT: // kill: def $h5 killed $h5 def $q5
; FULLFP16-NEXT: // kill: def $h6 killed $h6 def $q6
; FULLFP16-NEXT: // kill: def $h7 killed $h7 def $q7
; FULLFP16-NEXT: mov v0.h[1], v1.h[0]
; FULLFP16-NEXT: ldr h1, [sp, #8]
; FULLFP16-NEXT: ld1 { v1.h }[1], [x9]
; FULLFP16-NEXT: add x9, sp, #24
; FULLFP16-NEXT: mov v0.h[2], v2.h[0]
; FULLFP16-NEXT: ldr h2, [sp, #72]
; FULLFP16-NEXT: ld1 { v1.h }[2], [x9]
; FULLFP16-NEXT: add x9, sp, #32
; FULLFP16-NEXT: mov v0.h[3], v3.h[0]
; FULLFP16-NEXT: ld1 { v1.h }[3], [x9]
; FULLFP16-NEXT: add x9, sp, #40
; FULLFP16-NEXT: ldr h3, [sp]
; FULLFP16-NEXT: ld1 { v1.h }[4], [x9]
; FULLFP16-NEXT: add x9, sp, #48
; FULLFP16-NEXT: fmaxnm v2.8h, v3.8h, v2.8h
; FULLFP16-NEXT: mov v0.h[4], v4.h[0]
; FULLFP16-NEXT: ld1 { v1.h }[5], [x9]
; FULLFP16-NEXT: add x9, sp, #56
; FULLFP16-NEXT: str h2, [x8, #16]
; FULLFP16-NEXT: mov v0.h[5], v5.h[0]
; FULLFP16-NEXT: ld1 { v1.h }[6], [x9]
; FULLFP16-NEXT: add x9, sp, #64
; FULLFP16-NEXT: mov v0.h[6], v6.h[0]
; FULLFP16-NEXT: ld1 { v1.h }[7], [x9]
; FULLFP16-NEXT: mov v0.h[7], v7.h[0]
; FULLFP16-NEXT: fmaxnm v0.8h, v0.8h, v1.8h
; FULLFP16-NEXT: str q0, [x8]
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: max_nnan_v9f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: ldr h16, [sp, #16]
; NOFULLFP16-NEXT: ldr h17, [sp, #8]
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fcvt s0, h0
; NOFULLFP16-NEXT: ldr h18, [sp, #24]
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s16, h16
; NOFULLFP16-NEXT: fcvt s17, h17
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fcvt s18, h18
; NOFULLFP16-NEXT: fcvt s4, h4
; NOFULLFP16-NEXT: fcvt s5, h5
; NOFULLFP16-NEXT: fmaxnm s1, s1, s16
; NOFULLFP16-NEXT: fmaxnm s0, s0, s17
; NOFULLFP16-NEXT: ldr h16, [sp, #32]
; NOFULLFP16-NEXT: fmaxnm s2, s2, s18
; NOFULLFP16-NEXT: ldr h17, [sp, #40]
; NOFULLFP16-NEXT: fcvt s16, h16
; NOFULLFP16-NEXT: fcvt s17, h17
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: fcvt h0, s0
; NOFULLFP16-NEXT: fcvt h2, s2
; NOFULLFP16-NEXT: fmaxnm s3, s3, s16
; NOFULLFP16-NEXT: fmaxnm s4, s4, s17
; NOFULLFP16-NEXT: mov v0.h[1], v1.h[0]
; NOFULLFP16-NEXT: ldr h1, [sp, #48]
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fcvt h3, s3
; NOFULLFP16-NEXT: fcvt h4, s4
; NOFULLFP16-NEXT: mov v0.h[2], v2.h[0]
; NOFULLFP16-NEXT: ldr h2, [sp, #56]
; NOFULLFP16-NEXT: fmaxnm s1, s5, s1
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: ldr h5, [sp, #64]
; NOFULLFP16-NEXT: mov v0.h[3], v3.h[0]
; NOFULLFP16-NEXT: fcvt s3, h6
; NOFULLFP16-NEXT: ldr h6, [sp, #72]
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: fcvt s6, h6
; NOFULLFP16-NEXT: mov v0.h[4], v4.h[0]
; NOFULLFP16-NEXT: fmaxnm s2, s3, s2
; NOFULLFP16-NEXT: fcvt s3, h5
; NOFULLFP16-NEXT: fcvt s4, h7
; NOFULLFP16-NEXT: ldr h5, [sp]
; NOFULLFP16-NEXT: fcvt s5, h5
; NOFULLFP16-NEXT: mov v0.h[5], v1.h[0]
; NOFULLFP16-NEXT: fcvt h1, s2
; NOFULLFP16-NEXT: fmaxnm s2, s4, s3
; NOFULLFP16-NEXT: fmaxnm s3, s5, s6
; NOFULLFP16-NEXT: mov v0.h[6], v1.h[0]
; NOFULLFP16-NEXT: fcvt h1, s2
; NOFULLFP16-NEXT: fcvt h2, s3
; NOFULLFP16-NEXT: mov v0.h[7], v1.h[0]
; NOFULLFP16-NEXT: str h2, [x8, #16]
; NOFULLFP16-NEXT: str q0, [x8]
; NOFULLFP16-NEXT: ret
entry:
%c = call nnan <9 x half> @llvm.maximumnum.v9f16(<9 x half> %a, <9 x half> %b)
ret <9 x half> %c
}
define <16 x half> @max_nnan_v16f16(<16 x half> %a, <16 x half> %b) {
; FULLFP16-LABEL: max_nnan_v16f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fmaxnm v1.8h, v1.8h, v3.8h
; FULLFP16-NEXT: fmaxnm v0.8h, v0.8h, v2.8h
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: max_nnan_v16f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: mov h6, v2.h[1]
; NOFULLFP16-NEXT: mov h7, v0.h[1]
; NOFULLFP16-NEXT: fcvt s4, h2
; NOFULLFP16-NEXT: fcvt s5, h0
; NOFULLFP16-NEXT: mov h16, v3.h[1]
; NOFULLFP16-NEXT: mov h17, v1.h[1]
; NOFULLFP16-NEXT: mov h18, v2.h[2]
; NOFULLFP16-NEXT: mov h19, v0.h[2]
; NOFULLFP16-NEXT: fcvt s20, h3
; NOFULLFP16-NEXT: fcvt s21, h1
; NOFULLFP16-NEXT: mov h22, v3.h[2]
; NOFULLFP16-NEXT: mov h23, v1.h[2]
; NOFULLFP16-NEXT: fcvt s6, h6
; NOFULLFP16-NEXT: fcvt s7, h7
; NOFULLFP16-NEXT: mov h24, v0.h[6]
; NOFULLFP16-NEXT: fmaxnm s4, s5, s4
; NOFULLFP16-NEXT: fcvt s5, h16
; NOFULLFP16-NEXT: fcvt s16, h17
; NOFULLFP16-NEXT: fcvt s17, h18
; NOFULLFP16-NEXT: fcvt s18, h19
; NOFULLFP16-NEXT: mov h19, v0.h[3]
; NOFULLFP16-NEXT: fmaxnm s20, s21, s20
; NOFULLFP16-NEXT: fcvt s21, h22
; NOFULLFP16-NEXT: mov h22, v3.h[3]
; NOFULLFP16-NEXT: fmaxnm s6, s7, s6
; NOFULLFP16-NEXT: mov h7, v2.h[3]
; NOFULLFP16-NEXT: mov h25, v1.h[6]
; NOFULLFP16-NEXT: fcvt h4, s4
; NOFULLFP16-NEXT: fmaxnm s5, s16, s5
; NOFULLFP16-NEXT: fcvt s16, h23
; NOFULLFP16-NEXT: mov h23, v1.h[3]
; NOFULLFP16-NEXT: fmaxnm s17, s18, s17
; NOFULLFP16-NEXT: fcvt s18, h19
; NOFULLFP16-NEXT: fcvt h6, s6
; NOFULLFP16-NEXT: fcvt s7, h7
; NOFULLFP16-NEXT: fcvt h19, s5
; NOFULLFP16-NEXT: fcvt h5, s20
; NOFULLFP16-NEXT: fmaxnm s16, s16, s21
; NOFULLFP16-NEXT: fcvt s20, h23
; NOFULLFP16-NEXT: fcvt h17, s17
; NOFULLFP16-NEXT: mov h21, v2.h[4]
; NOFULLFP16-NEXT: mov h23, v1.h[4]
; NOFULLFP16-NEXT: mov v4.h[1], v6.h[0]
; NOFULLFP16-NEXT: fcvt s6, h22
; NOFULLFP16-NEXT: mov h22, v0.h[4]
; NOFULLFP16-NEXT: fmaxnm s7, s18, s7
; NOFULLFP16-NEXT: mov h18, v3.h[4]
; NOFULLFP16-NEXT: mov v5.h[1], v19.h[0]
; NOFULLFP16-NEXT: fcvt h16, s16
; NOFULLFP16-NEXT: fmaxnm s6, s20, s6
; NOFULLFP16-NEXT: mov v4.h[2], v17.h[0]
; NOFULLFP16-NEXT: fcvt s17, h21
; NOFULLFP16-NEXT: fcvt s19, h22
; NOFULLFP16-NEXT: fcvt h7, s7
; NOFULLFP16-NEXT: fcvt s18, h18
; NOFULLFP16-NEXT: fcvt s20, h23
; NOFULLFP16-NEXT: mov h21, v2.h[5]
; NOFULLFP16-NEXT: mov h22, v0.h[5]
; NOFULLFP16-NEXT: mov v5.h[2], v16.h[0]
; NOFULLFP16-NEXT: mov h16, v3.h[5]
; NOFULLFP16-NEXT: mov h23, v1.h[5]
; NOFULLFP16-NEXT: fcvt h6, s6
; NOFULLFP16-NEXT: mov h0, v0.h[7]
; NOFULLFP16-NEXT: mov h1, v1.h[7]
; NOFULLFP16-NEXT: fmaxnm s17, s19, s17
; NOFULLFP16-NEXT: mov h19, v2.h[6]
; NOFULLFP16-NEXT: mov v4.h[3], v7.h[0]
; NOFULLFP16-NEXT: fmaxnm s18, s20, s18
; NOFULLFP16-NEXT: mov h20, v3.h[6]
; NOFULLFP16-NEXT: fcvt s7, h21
; NOFULLFP16-NEXT: fcvt s21, h22
; NOFULLFP16-NEXT: fcvt s22, h24
; NOFULLFP16-NEXT: mov h2, v2.h[7]
; NOFULLFP16-NEXT: mov v5.h[3], v6.h[0]
; NOFULLFP16-NEXT: fcvt s6, h16
; NOFULLFP16-NEXT: fcvt s16, h23
; NOFULLFP16-NEXT: fcvt h17, s17
; NOFULLFP16-NEXT: fcvt s19, h19
; NOFULLFP16-NEXT: fcvt s23, h25
; NOFULLFP16-NEXT: fcvt h18, s18
; NOFULLFP16-NEXT: fcvt s20, h20
; NOFULLFP16-NEXT: mov h3, v3.h[7]
; NOFULLFP16-NEXT: fmaxnm s7, s21, s7
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s0, h0
; NOFULLFP16-NEXT: fmaxnm s6, s16, s6
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: mov v4.h[4], v17.h[0]
; NOFULLFP16-NEXT: fmaxnm s16, s22, s19
; NOFULLFP16-NEXT: mov v5.h[4], v18.h[0]
; NOFULLFP16-NEXT: fmaxnm s17, s23, s20
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fcvt h7, s7
; NOFULLFP16-NEXT: fmaxnm s0, s0, s2
; NOFULLFP16-NEXT: fcvt h6, s6
; NOFULLFP16-NEXT: fcvt h2, s16
; NOFULLFP16-NEXT: fmaxnm s1, s1, s3
; NOFULLFP16-NEXT: mov v4.h[5], v7.h[0]
; NOFULLFP16-NEXT: fcvt h0, s0
; NOFULLFP16-NEXT: mov v5.h[5], v6.h[0]
; NOFULLFP16-NEXT: fcvt h6, s17
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: mov v4.h[6], v2.h[0]
; NOFULLFP16-NEXT: mov v5.h[6], v6.h[0]
; NOFULLFP16-NEXT: mov v4.h[7], v0.h[0]
; NOFULLFP16-NEXT: mov v5.h[7], v1.h[0]
; NOFULLFP16-NEXT: mov v0.16b, v4.16b
; NOFULLFP16-NEXT: mov v1.16b, v5.16b
; NOFULLFP16-NEXT: ret
entry:
%c = call nnan <16 x half> @llvm.maximumnum.v16f16(<16 x half> %a, <16 x half> %b)
ret <16 x half> %c
}
;;;;;;;;;;;;;;;; min_f64
define double @min_nnan_f64(double %a, double %b) {
; AARCH64-LABEL: min_nnan_f64:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm d0, d0, d1
; AARCH64-NEXT: ret
entry:
%c = call nnan double @llvm.minimumnum.f64(double %a, double %b)
ret double %c
}
define <2 x double> @min_nnan_v2f64(<2 x double> %a, <2 x double> %b) {
; AARCH64-LABEL: min_nnan_v2f64:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm v0.2d, v0.2d, v1.2d
; AARCH64-NEXT: ret
entry:
%c = call nnan <2 x double> @llvm.minimumnum.v2f64(<2 x double> %a, <2 x double> %b)
ret <2 x double> %c
}
define <3 x double> @min_nnan_v3f64(<3 x double> %a, <3 x double> %b) {
; AARCH64-LABEL: min_nnan_v3f64:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: // kill: def $d3 killed $d3 def $q3
; AARCH64-NEXT: // kill: def $d0 killed $d0 def $q0
; AARCH64-NEXT: // kill: def $d4 killed $d4 def $q4
; AARCH64-NEXT: // kill: def $d1 killed $d1 def $q1
; AARCH64-NEXT: // kill: def $d2 killed $d2 def $q2
; AARCH64-NEXT: // kill: def $d5 killed $d5 def $q5
; AARCH64-NEXT: mov v3.d[1], v4.d[0]
; AARCH64-NEXT: mov v0.d[1], v1.d[0]
; AARCH64-NEXT: fminnm v2.2d, v2.2d, v5.2d
; AARCH64-NEXT: // kill: def $d2 killed $d2 killed $q2
; AARCH64-NEXT: fminnm v0.2d, v0.2d, v3.2d
; AARCH64-NEXT: ext v1.16b, v0.16b, v0.16b, #8
; AARCH64-NEXT: // kill: def $d0 killed $d0 killed $q0
; AARCH64-NEXT: // kill: def $d1 killed $d1 killed $q1
; AARCH64-NEXT: ret
entry:
%c = call nnan <3 x double> @llvm.minimumnum.v3f64(<3 x double> %a, <3 x double> %b)
ret <3 x double> %c
}
define <4 x double> @min_nnan_v4f64(<4 x double> %a, <4 x double> %b) {
; AARCH64-LABEL: min_nnan_v4f64:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm v1.2d, v1.2d, v3.2d
; AARCH64-NEXT: fminnm v0.2d, v0.2d, v2.2d
; AARCH64-NEXT: ret
entry:
%c = call nnan <4 x double> @llvm.minimumnum.v4f64(<4 x double> %a, <4 x double> %b)
ret <4 x double> %c
}
;;;;;;;;;;;;;;;;;; min_f32
define float @min_nnan_f32(float %a, float %b) {
; AARCH64-LABEL: min_nnan_f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm s0, s0, s1
; AARCH64-NEXT: ret
entry:
%c = call nnan float @llvm.minimumnum.f32(float %a, float %b)
ret float %c
}
define <2 x float> @min_nnan_v2f32(<2 x float> %a, <2 x float> %b) {
; AARCH64-LABEL: min_nnan_v2f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm v0.2s, v0.2s, v1.2s
; AARCH64-NEXT: ret
entry:
%c = call nnan <2 x float> @llvm.minimumnum.v2f32(<2 x float> %a, <2 x float> %b)
ret <2 x float> %c
}
define <3 x float> @min_nnan_v3f32(<3 x float> %a, <3 x float> %b) {
; AARCH64-LABEL: min_nnan_v3f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm v0.4s, v0.4s, v1.4s
; AARCH64-NEXT: ret
entry:
%c = call nnan <3 x float> @llvm.minimumnum.v3f32(<3 x float> %a, <3 x float> %b)
ret <3 x float> %c
}
define <4 x float> @min_nnan_v4f32(<4 x float> %a, <4 x float> %b) {
; AARCH64-LABEL: min_nnan_v4f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm v0.4s, v0.4s, v1.4s
; AARCH64-NEXT: ret
entry:
%c = call nnan <4 x float> @llvm.minimumnum.v4f32(<4 x float> %a, <4 x float> %b)
ret <4 x float> %c
}
define <5 x float> @min_nnan_v5f32(<5 x float> %a, <5 x float> %b) {
; AARCH64-LABEL: min_nnan_v5f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: // kill: def $s0 killed $s0 def $q0
; AARCH64-NEXT: // kill: def $s5 killed $s5 def $q5
; AARCH64-NEXT: // kill: def $s1 killed $s1 def $q1
; AARCH64-NEXT: // kill: def $s6 killed $s6 def $q6
; AARCH64-NEXT: // kill: def $s2 killed $s2 def $q2
; AARCH64-NEXT: // kill: def $s7 killed $s7 def $q7
; AARCH64-NEXT: // kill: def $s3 killed $s3 def $q3
; AARCH64-NEXT: mov x8, sp
; AARCH64-NEXT: // kill: def $s4 killed $s4 def $q4
; AARCH64-NEXT: mov v0.s[1], v1.s[0]
; AARCH64-NEXT: mov v5.s[1], v6.s[0]
; AARCH64-NEXT: mov v0.s[2], v2.s[0]
; AARCH64-NEXT: mov v5.s[2], v7.s[0]
; AARCH64-NEXT: ldr s2, [sp, #8]
; AARCH64-NEXT: fminnm v4.4s, v4.4s, v2.4s
; AARCH64-NEXT: // kill: def $s4 killed $s4 killed $q4
; AARCH64-NEXT: mov v0.s[3], v3.s[0]
; AARCH64-NEXT: ld1 { v5.s }[3], [x8]
; AARCH64-NEXT: fminnm v0.4s, v0.4s, v5.4s
; AARCH64-NEXT: mov s1, v0.s[1]
; AARCH64-NEXT: mov s2, v0.s[2]
; AARCH64-NEXT: mov s3, v0.s[3]
; AARCH64-NEXT: // kill: def $s0 killed $s0 killed $q0
; AARCH64-NEXT: ret
entry:
%c = call nnan <5 x float> @llvm.minimumnum.v5f32(<5 x float> %a, <5 x float> %b)
ret <5 x float> %c
}
define <8 x float> @min_nnan_v8f32(<8 x float> %a, <8 x float> %b) {
; AARCH64-LABEL: min_nnan_v8f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm v1.4s, v1.4s, v3.4s
; AARCH64-NEXT: fminnm v0.4s, v0.4s, v2.4s
; AARCH64-NEXT: ret
entry:
%c = call nnan <8 x float> @llvm.minimumnum.v8f32(<8 x float> %a, <8 x float> %b)
ret <8 x float> %c
}
;;;;;;;;;;;;;;;;;; min_f16
define half @min_nnan_f16(half %a, half %b) {
; FULLFP16-LABEL: min_nnan_f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fminnm h0, h0, h1
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: min_nnan_f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fcvt s0, h0
; NOFULLFP16-NEXT: fminnm s0, s0, s1
; NOFULLFP16-NEXT: fcvt h0, s0
; NOFULLFP16-NEXT: ret
entry:
%c = call nnan half @llvm.minimumnum.f16(half %a, half %b)
ret half %c
}
define <2 x half> @min_nnan_v2f16(<2 x half> %a, <2 x half> %b) {
; FULLFP16-LABEL: min_nnan_v2f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fminnm v0.4h, v0.4h, v1.4h
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: min_nnan_v2f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: // kill: def $d1 killed $d1 def $q1
; NOFULLFP16-NEXT: // kill: def $d0 killed $d0 def $q0
; NOFULLFP16-NEXT: mov h2, v1.h[1]
; NOFULLFP16-NEXT: mov h3, v0.h[1]
; NOFULLFP16-NEXT: mov h4, v1.h[2]
; NOFULLFP16-NEXT: mov h5, v0.h[2]
; NOFULLFP16-NEXT: fcvt s6, h1
; NOFULLFP16-NEXT: fcvt s7, h0
; NOFULLFP16-NEXT: mov h1, v1.h[3]
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fminnm s2, s3, s2
; NOFULLFP16-NEXT: fcvt s3, h4
; NOFULLFP16-NEXT: fcvt s4, h5
; NOFULLFP16-NEXT: fminnm s5, s7, s6
; NOFULLFP16-NEXT: mov h6, v0.h[3]
; NOFULLFP16-NEXT: fminnm s3, s4, s3
; NOFULLFP16-NEXT: fcvt h2, s2
; NOFULLFP16-NEXT: fcvt h0, s5
; NOFULLFP16-NEXT: fcvt s4, h6
; NOFULLFP16-NEXT: mov v0.h[1], v2.h[0]
; NOFULLFP16-NEXT: fcvt h2, s3
; NOFULLFP16-NEXT: fminnm s1, s4, s1
; NOFULLFP16-NEXT: mov v0.h[2], v2.h[0]
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: mov v0.h[3], v1.h[0]
; NOFULLFP16-NEXT: // kill: def $d0 killed $d0 killed $q0
; NOFULLFP16-NEXT: ret
entry:
%c = call nnan <2 x half> @llvm.minimumnum.v2f16(<2 x half> %a, <2 x half> %b)
ret <2 x half> %c
}
define <4 x half> @min_nnan_v4f16(<4 x half> %a, <4 x half> %b) {
; FULLFP16-LABEL: min_nnan_v4f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fminnm v0.4h, v0.4h, v1.4h
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: min_nnan_v4f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: // kill: def $d1 killed $d1 def $q1
; NOFULLFP16-NEXT: // kill: def $d0 killed $d0 def $q0
; NOFULLFP16-NEXT: mov h2, v1.h[1]
; NOFULLFP16-NEXT: mov h3, v0.h[1]
; NOFULLFP16-NEXT: mov h4, v1.h[2]
; NOFULLFP16-NEXT: mov h5, v0.h[2]
; NOFULLFP16-NEXT: fcvt s6, h1
; NOFULLFP16-NEXT: fcvt s7, h0
; NOFULLFP16-NEXT: mov h1, v1.h[3]
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fminnm s2, s3, s2
; NOFULLFP16-NEXT: fcvt s3, h4
; NOFULLFP16-NEXT: fcvt s4, h5
; NOFULLFP16-NEXT: fminnm s5, s7, s6
; NOFULLFP16-NEXT: mov h6, v0.h[3]
; NOFULLFP16-NEXT: fminnm s3, s4, s3
; NOFULLFP16-NEXT: fcvt h2, s2
; NOFULLFP16-NEXT: fcvt h0, s5
; NOFULLFP16-NEXT: fcvt s4, h6
; NOFULLFP16-NEXT: mov v0.h[1], v2.h[0]
; NOFULLFP16-NEXT: fcvt h2, s3
; NOFULLFP16-NEXT: fminnm s1, s4, s1
; NOFULLFP16-NEXT: mov v0.h[2], v2.h[0]
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: mov v0.h[3], v1.h[0]
; NOFULLFP16-NEXT: // kill: def $d0 killed $d0 killed $q0
; NOFULLFP16-NEXT: ret
entry:
%c = call nnan <4 x half> @llvm.minimumnum.v4f16(<4 x half> %a, <4 x half> %b)
ret <4 x half> %c
}
define <8 x half> @min_nnan_v8f16(<8 x half> %a, <8 x half> %b) {
; FULLFP16-LABEL: min_nnan_v8f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fminnm v0.8h, v0.8h, v1.8h
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: min_nnan_v8f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: mov h2, v1.h[1]
; NOFULLFP16-NEXT: mov h3, v0.h[1]
; NOFULLFP16-NEXT: fcvt s4, h1
; NOFULLFP16-NEXT: fcvt s5, h0
; NOFULLFP16-NEXT: mov h6, v1.h[2]
; NOFULLFP16-NEXT: mov h7, v0.h[2]
; NOFULLFP16-NEXT: mov h16, v1.h[3]
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fminnm s4, s5, s4
; NOFULLFP16-NEXT: mov h5, v0.h[3]
; NOFULLFP16-NEXT: fcvt s6, h6
; NOFULLFP16-NEXT: fcvt s7, h7
; NOFULLFP16-NEXT: fcvt s16, h16
; NOFULLFP16-NEXT: fminnm s3, s3, s2
; NOFULLFP16-NEXT: fcvt s5, h5
; NOFULLFP16-NEXT: fcvt h2, s4
; NOFULLFP16-NEXT: fminnm s4, s7, s6
; NOFULLFP16-NEXT: mov h6, v1.h[4]
; NOFULLFP16-NEXT: mov h7, v0.h[4]
; NOFULLFP16-NEXT: fcvt h3, s3
; NOFULLFP16-NEXT: fminnm s5, s5, s16
; NOFULLFP16-NEXT: mov h16, v0.h[5]
; NOFULLFP16-NEXT: fcvt h4, s4
; NOFULLFP16-NEXT: mov v2.h[1], v3.h[0]
; NOFULLFP16-NEXT: fcvt s3, h6
; NOFULLFP16-NEXT: fcvt s6, h7
; NOFULLFP16-NEXT: mov h7, v1.h[5]
; NOFULLFP16-NEXT: fcvt h5, s5
; NOFULLFP16-NEXT: fcvt s16, h16
; NOFULLFP16-NEXT: mov v2.h[2], v4.h[0]
; NOFULLFP16-NEXT: mov h4, v1.h[6]
; NOFULLFP16-NEXT: fminnm s3, s6, s3
; NOFULLFP16-NEXT: mov h6, v0.h[6]
; NOFULLFP16-NEXT: fcvt s7, h7
; NOFULLFP16-NEXT: mov h1, v1.h[7]
; NOFULLFP16-NEXT: mov h0, v0.h[7]
; NOFULLFP16-NEXT: mov v2.h[3], v5.h[0]
; NOFULLFP16-NEXT: fcvt s4, h4
; NOFULLFP16-NEXT: fcvt h3, s3
; NOFULLFP16-NEXT: fcvt s5, h6
; NOFULLFP16-NEXT: fminnm s6, s16, s7
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fcvt s0, h0
; NOFULLFP16-NEXT: mov v2.h[4], v3.h[0]
; NOFULLFP16-NEXT: fminnm s4, s5, s4
; NOFULLFP16-NEXT: fcvt h3, s6
; NOFULLFP16-NEXT: fminnm s0, s0, s1
; NOFULLFP16-NEXT: mov v2.h[5], v3.h[0]
; NOFULLFP16-NEXT: fcvt h3, s4
; NOFULLFP16-NEXT: fcvt h0, s0
; NOFULLFP16-NEXT: mov v2.h[6], v3.h[0]
; NOFULLFP16-NEXT: mov v2.h[7], v0.h[0]
; NOFULLFP16-NEXT: mov v0.16b, v2.16b
; NOFULLFP16-NEXT: ret
entry:
%c = call nnan <8 x half> @llvm.minimumnum.v8f16(<8 x half> %a, <8 x half> %b)
ret <8 x half> %c
}
define <9 x half> @min_nnan_v9f16(<9 x half> %a, <9 x half> %b) {
; FULLFP16-LABEL: min_nnan_v9f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: // kill: def $h0 killed $h0 def $q0
; FULLFP16-NEXT: // kill: def $h1 killed $h1 def $q1
; FULLFP16-NEXT: // kill: def $h2 killed $h2 def $q2
; FULLFP16-NEXT: add x9, sp, #16
; FULLFP16-NEXT: // kill: def $h3 killed $h3 def $q3
; FULLFP16-NEXT: // kill: def $h4 killed $h4 def $q4
; FULLFP16-NEXT: // kill: def $h5 killed $h5 def $q5
; FULLFP16-NEXT: // kill: def $h6 killed $h6 def $q6
; FULLFP16-NEXT: // kill: def $h7 killed $h7 def $q7
; FULLFP16-NEXT: mov v0.h[1], v1.h[0]
; FULLFP16-NEXT: ldr h1, [sp, #8]
; FULLFP16-NEXT: ld1 { v1.h }[1], [x9]
; FULLFP16-NEXT: add x9, sp, #24
; FULLFP16-NEXT: mov v0.h[2], v2.h[0]
; FULLFP16-NEXT: ldr h2, [sp, #72]
; FULLFP16-NEXT: ld1 { v1.h }[2], [x9]
; FULLFP16-NEXT: add x9, sp, #32
; FULLFP16-NEXT: mov v0.h[3], v3.h[0]
; FULLFP16-NEXT: ld1 { v1.h }[3], [x9]
; FULLFP16-NEXT: add x9, sp, #40
; FULLFP16-NEXT: ldr h3, [sp]
; FULLFP16-NEXT: ld1 { v1.h }[4], [x9]
; FULLFP16-NEXT: add x9, sp, #48
; FULLFP16-NEXT: fminnm v2.8h, v3.8h, v2.8h
; FULLFP16-NEXT: mov v0.h[4], v4.h[0]
; FULLFP16-NEXT: ld1 { v1.h }[5], [x9]
; FULLFP16-NEXT: add x9, sp, #56
; FULLFP16-NEXT: str h2, [x8, #16]
; FULLFP16-NEXT: mov v0.h[5], v5.h[0]
; FULLFP16-NEXT: ld1 { v1.h }[6], [x9]
; FULLFP16-NEXT: add x9, sp, #64
; FULLFP16-NEXT: mov v0.h[6], v6.h[0]
; FULLFP16-NEXT: ld1 { v1.h }[7], [x9]
; FULLFP16-NEXT: mov v0.h[7], v7.h[0]
; FULLFP16-NEXT: fminnm v0.8h, v0.8h, v1.8h
; FULLFP16-NEXT: str q0, [x8]
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: min_nnan_v9f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: ldr h16, [sp, #16]
; NOFULLFP16-NEXT: ldr h17, [sp, #8]
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fcvt s0, h0
; NOFULLFP16-NEXT: ldr h18, [sp, #24]
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s16, h16
; NOFULLFP16-NEXT: fcvt s17, h17
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fcvt s18, h18
; NOFULLFP16-NEXT: fcvt s4, h4
; NOFULLFP16-NEXT: fcvt s5, h5
; NOFULLFP16-NEXT: fminnm s1, s1, s16
; NOFULLFP16-NEXT: fminnm s0, s0, s17
; NOFULLFP16-NEXT: ldr h16, [sp, #32]
; NOFULLFP16-NEXT: fminnm s2, s2, s18
; NOFULLFP16-NEXT: ldr h17, [sp, #40]
; NOFULLFP16-NEXT: fcvt s16, h16
; NOFULLFP16-NEXT: fcvt s17, h17
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: fcvt h0, s0
; NOFULLFP16-NEXT: fcvt h2, s2
; NOFULLFP16-NEXT: fminnm s3, s3, s16
; NOFULLFP16-NEXT: fminnm s4, s4, s17
; NOFULLFP16-NEXT: mov v0.h[1], v1.h[0]
; NOFULLFP16-NEXT: ldr h1, [sp, #48]
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fcvt h3, s3
; NOFULLFP16-NEXT: fcvt h4, s4
; NOFULLFP16-NEXT: mov v0.h[2], v2.h[0]
; NOFULLFP16-NEXT: ldr h2, [sp, #56]
; NOFULLFP16-NEXT: fminnm s1, s5, s1
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: ldr h5, [sp, #64]
; NOFULLFP16-NEXT: mov v0.h[3], v3.h[0]
; NOFULLFP16-NEXT: fcvt s3, h6
; NOFULLFP16-NEXT: ldr h6, [sp, #72]
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: fcvt s6, h6
; NOFULLFP16-NEXT: mov v0.h[4], v4.h[0]
; NOFULLFP16-NEXT: fminnm s2, s3, s2
; NOFULLFP16-NEXT: fcvt s3, h5
; NOFULLFP16-NEXT: fcvt s4, h7
; NOFULLFP16-NEXT: ldr h5, [sp]
; NOFULLFP16-NEXT: fcvt s5, h5
; NOFULLFP16-NEXT: mov v0.h[5], v1.h[0]
; NOFULLFP16-NEXT: fcvt h1, s2
; NOFULLFP16-NEXT: fminnm s2, s4, s3
; NOFULLFP16-NEXT: fminnm s3, s5, s6
; NOFULLFP16-NEXT: mov v0.h[6], v1.h[0]
; NOFULLFP16-NEXT: fcvt h1, s2
; NOFULLFP16-NEXT: fcvt h2, s3
; NOFULLFP16-NEXT: mov v0.h[7], v1.h[0]
; NOFULLFP16-NEXT: str h2, [x8, #16]
; NOFULLFP16-NEXT: str q0, [x8]
; NOFULLFP16-NEXT: ret
entry:
%c = call nnan <9 x half> @llvm.minimumnum.v9f16(<9 x half> %a, <9 x half> %b)
ret <9 x half> %c
}
define <16 x half> @min_nnan_v16f16(<16 x half> %a, <16 x half> %b) {
; FULLFP16-LABEL: min_nnan_v16f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fminnm v1.8h, v1.8h, v3.8h
; FULLFP16-NEXT: fminnm v0.8h, v0.8h, v2.8h
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: min_nnan_v16f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: mov h6, v2.h[1]
; NOFULLFP16-NEXT: mov h7, v0.h[1]
; NOFULLFP16-NEXT: fcvt s4, h2
; NOFULLFP16-NEXT: fcvt s5, h0
; NOFULLFP16-NEXT: mov h16, v3.h[1]
; NOFULLFP16-NEXT: mov h17, v1.h[1]
; NOFULLFP16-NEXT: mov h18, v2.h[2]
; NOFULLFP16-NEXT: mov h19, v0.h[2]
; NOFULLFP16-NEXT: fcvt s20, h3
; NOFULLFP16-NEXT: fcvt s21, h1
; NOFULLFP16-NEXT: mov h22, v3.h[2]
; NOFULLFP16-NEXT: mov h23, v1.h[2]
; NOFULLFP16-NEXT: fcvt s6, h6
; NOFULLFP16-NEXT: fcvt s7, h7
; NOFULLFP16-NEXT: mov h24, v0.h[6]
; NOFULLFP16-NEXT: fminnm s4, s5, s4
; NOFULLFP16-NEXT: fcvt s5, h16
; NOFULLFP16-NEXT: fcvt s16, h17
; NOFULLFP16-NEXT: fcvt s17, h18
; NOFULLFP16-NEXT: fcvt s18, h19
; NOFULLFP16-NEXT: mov h19, v0.h[3]
; NOFULLFP16-NEXT: fminnm s20, s21, s20
; NOFULLFP16-NEXT: fcvt s21, h22
; NOFULLFP16-NEXT: mov h22, v3.h[3]
; NOFULLFP16-NEXT: fminnm s6, s7, s6
; NOFULLFP16-NEXT: mov h7, v2.h[3]
; NOFULLFP16-NEXT: mov h25, v1.h[6]
; NOFULLFP16-NEXT: fcvt h4, s4
; NOFULLFP16-NEXT: fminnm s5, s16, s5
; NOFULLFP16-NEXT: fcvt s16, h23
; NOFULLFP16-NEXT: mov h23, v1.h[3]
; NOFULLFP16-NEXT: fminnm s17, s18, s17
; NOFULLFP16-NEXT: fcvt s18, h19
; NOFULLFP16-NEXT: fcvt h6, s6
; NOFULLFP16-NEXT: fcvt s7, h7
; NOFULLFP16-NEXT: fcvt h19, s5
; NOFULLFP16-NEXT: fcvt h5, s20
; NOFULLFP16-NEXT: fminnm s16, s16, s21
; NOFULLFP16-NEXT: fcvt s20, h23
; NOFULLFP16-NEXT: fcvt h17, s17
; NOFULLFP16-NEXT: mov h21, v2.h[4]
; NOFULLFP16-NEXT: mov h23, v1.h[4]
; NOFULLFP16-NEXT: mov v4.h[1], v6.h[0]
; NOFULLFP16-NEXT: fcvt s6, h22
; NOFULLFP16-NEXT: mov h22, v0.h[4]
; NOFULLFP16-NEXT: fminnm s7, s18, s7
; NOFULLFP16-NEXT: mov h18, v3.h[4]
; NOFULLFP16-NEXT: mov v5.h[1], v19.h[0]
; NOFULLFP16-NEXT: fcvt h16, s16
; NOFULLFP16-NEXT: fminnm s6, s20, s6
; NOFULLFP16-NEXT: mov v4.h[2], v17.h[0]
; NOFULLFP16-NEXT: fcvt s17, h21
; NOFULLFP16-NEXT: fcvt s19, h22
; NOFULLFP16-NEXT: fcvt h7, s7
; NOFULLFP16-NEXT: fcvt s18, h18
; NOFULLFP16-NEXT: fcvt s20, h23
; NOFULLFP16-NEXT: mov h21, v2.h[5]
; NOFULLFP16-NEXT: mov h22, v0.h[5]
; NOFULLFP16-NEXT: mov v5.h[2], v16.h[0]
; NOFULLFP16-NEXT: mov h16, v3.h[5]
; NOFULLFP16-NEXT: mov h23, v1.h[5]
; NOFULLFP16-NEXT: fcvt h6, s6
; NOFULLFP16-NEXT: mov h0, v0.h[7]
; NOFULLFP16-NEXT: mov h1, v1.h[7]
; NOFULLFP16-NEXT: fminnm s17, s19, s17
; NOFULLFP16-NEXT: mov h19, v2.h[6]
; NOFULLFP16-NEXT: mov v4.h[3], v7.h[0]
; NOFULLFP16-NEXT: fminnm s18, s20, s18
; NOFULLFP16-NEXT: mov h20, v3.h[6]
; NOFULLFP16-NEXT: fcvt s7, h21
; NOFULLFP16-NEXT: fcvt s21, h22
; NOFULLFP16-NEXT: fcvt s22, h24
; NOFULLFP16-NEXT: mov h2, v2.h[7]
; NOFULLFP16-NEXT: mov v5.h[3], v6.h[0]
; NOFULLFP16-NEXT: fcvt s6, h16
; NOFULLFP16-NEXT: fcvt s16, h23
; NOFULLFP16-NEXT: fcvt h17, s17
; NOFULLFP16-NEXT: fcvt s19, h19
; NOFULLFP16-NEXT: fcvt s23, h25
; NOFULLFP16-NEXT: fcvt h18, s18
; NOFULLFP16-NEXT: fcvt s20, h20
; NOFULLFP16-NEXT: mov h3, v3.h[7]
; NOFULLFP16-NEXT: fminnm s7, s21, s7
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s0, h0
; NOFULLFP16-NEXT: fminnm s6, s16, s6
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: mov v4.h[4], v17.h[0]
; NOFULLFP16-NEXT: fminnm s16, s22, s19
; NOFULLFP16-NEXT: mov v5.h[4], v18.h[0]
; NOFULLFP16-NEXT: fminnm s17, s23, s20
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fcvt h7, s7
; NOFULLFP16-NEXT: fminnm s0, s0, s2
; NOFULLFP16-NEXT: fcvt h6, s6
; NOFULLFP16-NEXT: fcvt h2, s16
; NOFULLFP16-NEXT: fminnm s1, s1, s3
; NOFULLFP16-NEXT: mov v4.h[5], v7.h[0]
; NOFULLFP16-NEXT: fcvt h0, s0
; NOFULLFP16-NEXT: mov v5.h[5], v6.h[0]
; NOFULLFP16-NEXT: fcvt h6, s17
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: mov v4.h[6], v2.h[0]
; NOFULLFP16-NEXT: mov v5.h[6], v6.h[0]
; NOFULLFP16-NEXT: mov v4.h[7], v0.h[0]
; NOFULLFP16-NEXT: mov v5.h[7], v1.h[0]
; NOFULLFP16-NEXT: mov v0.16b, v4.16b
; NOFULLFP16-NEXT: mov v1.16b, v5.16b
; NOFULLFP16-NEXT: ret
entry:
%c = call nnan <16 x half> @llvm.minimumnum.v16f16(<16 x half> %a, <16 x half> %b)
ret <16 x half> %c
}
;;;;;;;;;;;;;;;; max_f64
define double @max_f64(double %a, double %b) {
; AARCH64-LABEL: max_f64:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm d1, d1, d1
; AARCH64-NEXT: fminnm d0, d0, d0
; AARCH64-NEXT: fmaxnm d0, d0, d1
; AARCH64-NEXT: ret
entry:
%c = call double @llvm.maximumnum.f64(double %a, double %b)
ret double %c
}
define <2 x double> @max_v2f64(<2 x double> %a, <2 x double> %b) {
; AARCH64-LABEL: max_v2f64:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm v1.2d, v1.2d, v1.2d
; AARCH64-NEXT: fminnm v0.2d, v0.2d, v0.2d
; AARCH64-NEXT: fmaxnm v0.2d, v0.2d, v1.2d
; AARCH64-NEXT: ret
entry:
%c = call <2 x double> @llvm.maximumnum.v2f64(<2 x double> %a, <2 x double> %b)
ret <2 x double> %c
}
define <3 x double> @max_v3f64(<3 x double> %a, <3 x double> %b) {
; AARCH64-LABEL: max_v3f64:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: // kill: def $d3 killed $d3 def $q3
; AARCH64-NEXT: // kill: def $d0 killed $d0 def $q0
; AARCH64-NEXT: // kill: def $d4 killed $d4 def $q4
; AARCH64-NEXT: // kill: def $d1 killed $d1 def $q1
; AARCH64-NEXT: // kill: def $d2 killed $d2 def $q2
; AARCH64-NEXT: // kill: def $d5 killed $d5 def $q5
; AARCH64-NEXT: mov v0.d[1], v1.d[0]
; AARCH64-NEXT: mov v3.d[1], v4.d[0]
; AARCH64-NEXT: fminnm v2.2d, v2.2d, v2.2d
; AARCH64-NEXT: fminnm v1.2d, v3.2d, v3.2d
; AARCH64-NEXT: fminnm v0.2d, v0.2d, v0.2d
; AARCH64-NEXT: fmaxnm v0.2d, v0.2d, v1.2d
; AARCH64-NEXT: fminnm v1.2d, v5.2d, v5.2d
; AARCH64-NEXT: fmaxnm v2.2d, v2.2d, v1.2d
; AARCH64-NEXT: ext v1.16b, v0.16b, v0.16b, #8
; AARCH64-NEXT: // kill: def $d0 killed $d0 killed $q0
; AARCH64-NEXT: // kill: def $d1 killed $d1 killed $q1
; AARCH64-NEXT: // kill: def $d2 killed $d2 killed $q2
; AARCH64-NEXT: ret
entry:
%c = call <3 x double> @llvm.maximumnum.v3f64(<3 x double> %a, <3 x double> %b)
ret <3 x double> %c
}
define <4 x double> @max_v4f64(<4 x double> %a, <4 x double> %b) {
; AARCH64-LABEL: max_v4f64:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm v2.2d, v2.2d, v2.2d
; AARCH64-NEXT: fminnm v0.2d, v0.2d, v0.2d
; AARCH64-NEXT: fminnm v3.2d, v3.2d, v3.2d
; AARCH64-NEXT: fminnm v1.2d, v1.2d, v1.2d
; AARCH64-NEXT: fmaxnm v0.2d, v0.2d, v2.2d
; AARCH64-NEXT: fmaxnm v1.2d, v1.2d, v3.2d
; AARCH64-NEXT: ret
entry:
%c = call <4 x double> @llvm.maximumnum.v4f64(<4 x double> %a, <4 x double> %b)
ret <4 x double> %c
}
;;;;;;;;;;;;;;;;;; max_f32
define float @max_f32(float %a, float %b) {
; AARCH64-LABEL: max_f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm s1, s1, s1
; AARCH64-NEXT: fminnm s0, s0, s0
; AARCH64-NEXT: fmaxnm s0, s0, s1
; AARCH64-NEXT: ret
entry:
%c = call float @llvm.maximumnum.f32(float %a, float %b)
ret float %c
}
define <2 x float> @max_v2f32(<2 x float> %a, <2 x float> %b) {
; AARCH64-LABEL: max_v2f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm v1.2s, v1.2s, v1.2s
; AARCH64-NEXT: fminnm v0.2s, v0.2s, v0.2s
; AARCH64-NEXT: fmaxnm v0.2s, v0.2s, v1.2s
; AARCH64-NEXT: ret
entry:
%c = call <2 x float> @llvm.maximumnum.v2f32(<2 x float> %a, <2 x float> %b)
ret <2 x float> %c
}
define <3 x float> @max_v3f32(<3 x float> %a, <3 x float> %b) {
; AARCH64-LABEL: max_v3f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm v1.4s, v1.4s, v1.4s
; AARCH64-NEXT: fminnm v0.4s, v0.4s, v0.4s
; AARCH64-NEXT: fmaxnm v0.4s, v0.4s, v1.4s
; AARCH64-NEXT: ret
entry:
%c = call <3 x float> @llvm.maximumnum.v3f32(<3 x float> %a, <3 x float> %b)
ret <3 x float> %c
}
define <4 x float> @max_v4f32(<4 x float> %a, <4 x float> %b) {
; AARCH64-LABEL: max_v4f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm v1.4s, v1.4s, v1.4s
; AARCH64-NEXT: fminnm v0.4s, v0.4s, v0.4s
; AARCH64-NEXT: fmaxnm v0.4s, v0.4s, v1.4s
; AARCH64-NEXT: ret
entry:
%c = call <4 x float> @llvm.maximumnum.v4f32(<4 x float> %a, <4 x float> %b)
ret <4 x float> %c
}
define <5 x float> @max_v5f32(<5 x float> %a, <5 x float> %b) {
; AARCH64-LABEL: max_v5f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: // kill: def $s0 killed $s0 def $q0
; AARCH64-NEXT: // kill: def $s5 killed $s5 def $q5
; AARCH64-NEXT: // kill: def $s1 killed $s1 def $q1
; AARCH64-NEXT: // kill: def $s6 killed $s6 def $q6
; AARCH64-NEXT: // kill: def $s2 killed $s2 def $q2
; AARCH64-NEXT: // kill: def $s7 killed $s7 def $q7
; AARCH64-NEXT: // kill: def $s3 killed $s3 def $q3
; AARCH64-NEXT: mov x8, sp
; AARCH64-NEXT: // kill: def $s4 killed $s4 def $q4
; AARCH64-NEXT: mov v0.s[1], v1.s[0]
; AARCH64-NEXT: mov v5.s[1], v6.s[0]
; AARCH64-NEXT: mov v0.s[2], v2.s[0]
; AARCH64-NEXT: mov v5.s[2], v7.s[0]
; AARCH64-NEXT: ldr s2, [sp, #8]
; AARCH64-NEXT: fminnm v2.4s, v2.4s, v2.4s
; AARCH64-NEXT: mov v0.s[3], v3.s[0]
; AARCH64-NEXT: ld1 { v5.s }[3], [x8]
; AARCH64-NEXT: fminnm v3.4s, v4.4s, v4.4s
; AARCH64-NEXT: fminnm v1.4s, v5.4s, v5.4s
; AARCH64-NEXT: fminnm v0.4s, v0.4s, v0.4s
; AARCH64-NEXT: fmaxnm v4.4s, v3.4s, v2.4s
; AARCH64-NEXT: // kill: def $s4 killed $s4 killed $q4
; AARCH64-NEXT: fmaxnm v0.4s, v0.4s, v1.4s
; AARCH64-NEXT: mov s1, v0.s[1]
; AARCH64-NEXT: mov s2, v0.s[2]
; AARCH64-NEXT: mov s3, v0.s[3]
; AARCH64-NEXT: // kill: def $s0 killed $s0 killed $q0
; AARCH64-NEXT: ret
entry:
%c = call <5 x float> @llvm.maximumnum.v5f32(<5 x float> %a, <5 x float> %b)
ret <5 x float> %c
}
define <8 x float> @max_v8f32(<8 x float> %a, <8 x float> %b) {
; AARCH64-LABEL: max_v8f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm v2.4s, v2.4s, v2.4s
; AARCH64-NEXT: fminnm v0.4s, v0.4s, v0.4s
; AARCH64-NEXT: fminnm v3.4s, v3.4s, v3.4s
; AARCH64-NEXT: fminnm v1.4s, v1.4s, v1.4s
; AARCH64-NEXT: fmaxnm v0.4s, v0.4s, v2.4s
; AARCH64-NEXT: fmaxnm v1.4s, v1.4s, v3.4s
; AARCH64-NEXT: ret
entry:
%c = call <8 x float> @llvm.maximumnum.v8f32(<8 x float> %a, <8 x float> %b)
ret <8 x float> %c
}
;;;;;;;;;;;;;;;;;; max_f16
define half @max_f16(half %a, half %b) {
; FULLFP16-LABEL: max_f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fminnm h1, h1, h1
; FULLFP16-NEXT: fminnm h0, h0, h0
; FULLFP16-NEXT: fmaxnm h0, h0, h1
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: max_f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fcvt s0, h0
; NOFULLFP16-NEXT: fmaxnm s0, s0, s1
; NOFULLFP16-NEXT: fcvt h0, s0
; NOFULLFP16-NEXT: ret
entry:
%c = call half @llvm.maximumnum.f16(half %a, half %b)
ret half %c
}
define <2 x half> @max_v2f16(<2 x half> %a, <2 x half> %b) {
; FULLFP16-LABEL: max_v2f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fminnm v1.4h, v1.4h, v1.4h
; FULLFP16-NEXT: fminnm v0.4h, v0.4h, v0.4h
; FULLFP16-NEXT: fmaxnm v0.4h, v0.4h, v1.4h
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: max_v2f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: // kill: def $d1 killed $d1 def $q1
; NOFULLFP16-NEXT: // kill: def $d0 killed $d0 def $q0
; NOFULLFP16-NEXT: mov h2, v1.h[1]
; NOFULLFP16-NEXT: mov h3, v0.h[1]
; NOFULLFP16-NEXT: mov h4, v1.h[2]
; NOFULLFP16-NEXT: mov h5, v0.h[2]
; NOFULLFP16-NEXT: fcvt s6, h1
; NOFULLFP16-NEXT: fcvt s7, h0
; NOFULLFP16-NEXT: mov h1, v1.h[3]
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fmaxnm s2, s3, s2
; NOFULLFP16-NEXT: fcvt s3, h4
; NOFULLFP16-NEXT: fcvt s4, h5
; NOFULLFP16-NEXT: fmaxnm s5, s7, s6
; NOFULLFP16-NEXT: mov h6, v0.h[3]
; NOFULLFP16-NEXT: fmaxnm s3, s4, s3
; NOFULLFP16-NEXT: fcvt h2, s2
; NOFULLFP16-NEXT: fcvt h0, s5
; NOFULLFP16-NEXT: fcvt s4, h6
; NOFULLFP16-NEXT: mov v0.h[1], v2.h[0]
; NOFULLFP16-NEXT: fcvt h2, s3
; NOFULLFP16-NEXT: fmaxnm s1, s4, s1
; NOFULLFP16-NEXT: mov v0.h[2], v2.h[0]
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: mov v0.h[3], v1.h[0]
; NOFULLFP16-NEXT: // kill: def $d0 killed $d0 killed $q0
; NOFULLFP16-NEXT: ret
entry:
%c = call <2 x half> @llvm.maximumnum.v2f16(<2 x half> %a, <2 x half> %b)
ret <2 x half> %c
}
define <4 x half> @max_v4f16(<4 x half> %a, <4 x half> %b) {
; FULLFP16-LABEL: max_v4f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fminnm v1.4h, v1.4h, v1.4h
; FULLFP16-NEXT: fminnm v0.4h, v0.4h, v0.4h
; FULLFP16-NEXT: fmaxnm v0.4h, v0.4h, v1.4h
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: max_v4f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: // kill: def $d1 killed $d1 def $q1
; NOFULLFP16-NEXT: // kill: def $d0 killed $d0 def $q0
; NOFULLFP16-NEXT: mov h2, v1.h[1]
; NOFULLFP16-NEXT: mov h3, v0.h[1]
; NOFULLFP16-NEXT: mov h4, v1.h[2]
; NOFULLFP16-NEXT: mov h5, v0.h[2]
; NOFULLFP16-NEXT: fcvt s6, h1
; NOFULLFP16-NEXT: fcvt s7, h0
; NOFULLFP16-NEXT: mov h1, v1.h[3]
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fmaxnm s2, s3, s2
; NOFULLFP16-NEXT: fcvt s3, h4
; NOFULLFP16-NEXT: fcvt s4, h5
; NOFULLFP16-NEXT: fmaxnm s5, s7, s6
; NOFULLFP16-NEXT: mov h6, v0.h[3]
; NOFULLFP16-NEXT: fmaxnm s3, s4, s3
; NOFULLFP16-NEXT: fcvt h2, s2
; NOFULLFP16-NEXT: fcvt h0, s5
; NOFULLFP16-NEXT: fcvt s4, h6
; NOFULLFP16-NEXT: mov v0.h[1], v2.h[0]
; NOFULLFP16-NEXT: fcvt h2, s3
; NOFULLFP16-NEXT: fmaxnm s1, s4, s1
; NOFULLFP16-NEXT: mov v0.h[2], v2.h[0]
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: mov v0.h[3], v1.h[0]
; NOFULLFP16-NEXT: // kill: def $d0 killed $d0 killed $q0
; NOFULLFP16-NEXT: ret
entry:
%c = call <4 x half> @llvm.maximumnum.v4f16(<4 x half> %a, <4 x half> %b)
ret <4 x half> %c
}
define <8 x half> @max_v8f16(<8 x half> %a, <8 x half> %b) {
; FULLFP16-LABEL: max_v8f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fminnm v1.8h, v1.8h, v1.8h
; FULLFP16-NEXT: fminnm v0.8h, v0.8h, v0.8h
; FULLFP16-NEXT: fmaxnm v0.8h, v0.8h, v1.8h
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: max_v8f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: mov h2, v1.h[1]
; NOFULLFP16-NEXT: mov h3, v0.h[1]
; NOFULLFP16-NEXT: fcvt s4, h1
; NOFULLFP16-NEXT: fcvt s5, h0
; NOFULLFP16-NEXT: mov h6, v1.h[2]
; NOFULLFP16-NEXT: mov h7, v0.h[2]
; NOFULLFP16-NEXT: mov h16, v1.h[3]
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fmaxnm s4, s5, s4
; NOFULLFP16-NEXT: mov h5, v0.h[3]
; NOFULLFP16-NEXT: fcvt s6, h6
; NOFULLFP16-NEXT: fcvt s7, h7
; NOFULLFP16-NEXT: fcvt s16, h16
; NOFULLFP16-NEXT: fmaxnm s3, s3, s2
; NOFULLFP16-NEXT: fcvt s5, h5
; NOFULLFP16-NEXT: fcvt h2, s4
; NOFULLFP16-NEXT: fmaxnm s4, s7, s6
; NOFULLFP16-NEXT: mov h6, v1.h[4]
; NOFULLFP16-NEXT: mov h7, v0.h[4]
; NOFULLFP16-NEXT: fcvt h3, s3
; NOFULLFP16-NEXT: fmaxnm s5, s5, s16
; NOFULLFP16-NEXT: mov h16, v0.h[5]
; NOFULLFP16-NEXT: fcvt h4, s4
; NOFULLFP16-NEXT: mov v2.h[1], v3.h[0]
; NOFULLFP16-NEXT: fcvt s3, h6
; NOFULLFP16-NEXT: fcvt s6, h7
; NOFULLFP16-NEXT: mov h7, v1.h[5]
; NOFULLFP16-NEXT: fcvt h5, s5
; NOFULLFP16-NEXT: fcvt s16, h16
; NOFULLFP16-NEXT: mov v2.h[2], v4.h[0]
; NOFULLFP16-NEXT: mov h4, v1.h[6]
; NOFULLFP16-NEXT: fmaxnm s3, s6, s3
; NOFULLFP16-NEXT: mov h6, v0.h[6]
; NOFULLFP16-NEXT: fcvt s7, h7
; NOFULLFP16-NEXT: mov h1, v1.h[7]
; NOFULLFP16-NEXT: mov h0, v0.h[7]
; NOFULLFP16-NEXT: mov v2.h[3], v5.h[0]
; NOFULLFP16-NEXT: fcvt s4, h4
; NOFULLFP16-NEXT: fcvt h3, s3
; NOFULLFP16-NEXT: fcvt s5, h6
; NOFULLFP16-NEXT: fmaxnm s6, s16, s7
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fcvt s0, h0
; NOFULLFP16-NEXT: mov v2.h[4], v3.h[0]
; NOFULLFP16-NEXT: fmaxnm s4, s5, s4
; NOFULLFP16-NEXT: fcvt h3, s6
; NOFULLFP16-NEXT: fmaxnm s0, s0, s1
; NOFULLFP16-NEXT: mov v2.h[5], v3.h[0]
; NOFULLFP16-NEXT: fcvt h3, s4
; NOFULLFP16-NEXT: fcvt h0, s0
; NOFULLFP16-NEXT: mov v2.h[6], v3.h[0]
; NOFULLFP16-NEXT: mov v2.h[7], v0.h[0]
; NOFULLFP16-NEXT: mov v0.16b, v2.16b
; NOFULLFP16-NEXT: ret
entry:
%c = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> %a, <8 x half> %b)
ret <8 x half> %c
}
define <9 x half> @max_v9f16(<9 x half> %a, <9 x half> %b) {
; FULLFP16-LABEL: max_v9f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: // kill: def $h0 killed $h0 def $q0
; FULLFP16-NEXT: // kill: def $h1 killed $h1 def $q1
; FULLFP16-NEXT: // kill: def $h2 killed $h2 def $q2
; FULLFP16-NEXT: add x9, sp, #16
; FULLFP16-NEXT: // kill: def $h3 killed $h3 def $q3
; FULLFP16-NEXT: // kill: def $h4 killed $h4 def $q4
; FULLFP16-NEXT: add x10, sp, #40
; FULLFP16-NEXT: // kill: def $h5 killed $h5 def $q5
; FULLFP16-NEXT: // kill: def $h6 killed $h6 def $q6
; FULLFP16-NEXT: // kill: def $h7 killed $h7 def $q7
; FULLFP16-NEXT: mov v0.h[1], v1.h[0]
; FULLFP16-NEXT: ldr h1, [sp, #8]
; FULLFP16-NEXT: ld1 { v1.h }[1], [x9]
; FULLFP16-NEXT: add x9, sp, #24
; FULLFP16-NEXT: mov v0.h[2], v2.h[0]
; FULLFP16-NEXT: ld1 { v1.h }[2], [x9]
; FULLFP16-NEXT: add x9, sp, #32
; FULLFP16-NEXT: mov v0.h[3], v3.h[0]
; FULLFP16-NEXT: ld1 { v1.h }[3], [x9]
; FULLFP16-NEXT: ldr h2, [x10]
; FULLFP16-NEXT: add x9, sp, #48
; FULLFP16-NEXT: ldr h3, [sp, #72]
; FULLFP16-NEXT: ld1 { v2.h }[1], [x9]
; FULLFP16-NEXT: add x9, sp, #56
; FULLFP16-NEXT: fminnm v3.8h, v3.8h, v3.8h
; FULLFP16-NEXT: mov v0.h[4], v4.h[0]
; FULLFP16-NEXT: ld1 { v2.h }[2], [x9]
; FULLFP16-NEXT: add x9, sp, #64
; FULLFP16-NEXT: mov v0.h[5], v5.h[0]
; FULLFP16-NEXT: ld1 { v2.h }[3], [x9]
; FULLFP16-NEXT: zip1 v1.2d, v1.2d, v2.2d
; FULLFP16-NEXT: ldr h2, [sp]
; FULLFP16-NEXT: mov v0.h[6], v6.h[0]
; FULLFP16-NEXT: fminnm v2.8h, v2.8h, v2.8h
; FULLFP16-NEXT: fminnm v1.8h, v1.8h, v1.8h
; FULLFP16-NEXT: mov v0.h[7], v7.h[0]
; FULLFP16-NEXT: fmaxnm v2.8h, v2.8h, v3.8h
; FULLFP16-NEXT: fminnm v0.8h, v0.8h, v0.8h
; FULLFP16-NEXT: str h2, [x8, #16]
; FULLFP16-NEXT: fmaxnm v0.8h, v0.8h, v1.8h
; FULLFP16-NEXT: str q0, [x8]
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: max_v9f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: ldr h16, [sp, #16]
; NOFULLFP16-NEXT: ldr h17, [sp, #8]
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fcvt s0, h0
; NOFULLFP16-NEXT: ldr h18, [sp, #24]
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s16, h16
; NOFULLFP16-NEXT: fcvt s17, h17
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fcvt s18, h18
; NOFULLFP16-NEXT: fcvt s4, h4
; NOFULLFP16-NEXT: fcvt s5, h5
; NOFULLFP16-NEXT: fmaxnm s1, s1, s16
; NOFULLFP16-NEXT: fmaxnm s0, s0, s17
; NOFULLFP16-NEXT: ldr h16, [sp, #32]
; NOFULLFP16-NEXT: fmaxnm s2, s2, s18
; NOFULLFP16-NEXT: ldr h17, [sp, #40]
; NOFULLFP16-NEXT: fcvt s16, h16
; NOFULLFP16-NEXT: fcvt s17, h17
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: fcvt h0, s0
; NOFULLFP16-NEXT: fcvt h2, s2
; NOFULLFP16-NEXT: fmaxnm s3, s3, s16
; NOFULLFP16-NEXT: fmaxnm s4, s4, s17
; NOFULLFP16-NEXT: mov v0.h[1], v1.h[0]
; NOFULLFP16-NEXT: ldr h1, [sp, #48]
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fcvt h3, s3
; NOFULLFP16-NEXT: fcvt h4, s4
; NOFULLFP16-NEXT: mov v0.h[2], v2.h[0]
; NOFULLFP16-NEXT: ldr h2, [sp, #56]
; NOFULLFP16-NEXT: fmaxnm s1, s5, s1
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: ldr h5, [sp, #64]
; NOFULLFP16-NEXT: mov v0.h[3], v3.h[0]
; NOFULLFP16-NEXT: fcvt s3, h6
; NOFULLFP16-NEXT: ldr h6, [sp, #72]
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: fcvt s6, h6
; NOFULLFP16-NEXT: mov v0.h[4], v4.h[0]
; NOFULLFP16-NEXT: fmaxnm s2, s3, s2
; NOFULLFP16-NEXT: fcvt s3, h5
; NOFULLFP16-NEXT: fcvt s4, h7
; NOFULLFP16-NEXT: ldr h5, [sp]
; NOFULLFP16-NEXT: fcvt s5, h5
; NOFULLFP16-NEXT: mov v0.h[5], v1.h[0]
; NOFULLFP16-NEXT: fcvt h1, s2
; NOFULLFP16-NEXT: fmaxnm s2, s4, s3
; NOFULLFP16-NEXT: fmaxnm s3, s5, s6
; NOFULLFP16-NEXT: mov v0.h[6], v1.h[0]
; NOFULLFP16-NEXT: fcvt h1, s2
; NOFULLFP16-NEXT: fcvt h2, s3
; NOFULLFP16-NEXT: mov v0.h[7], v1.h[0]
; NOFULLFP16-NEXT: str h2, [x8, #16]
; NOFULLFP16-NEXT: str q0, [x8]
; NOFULLFP16-NEXT: ret
entry:
%c = call <9 x half> @llvm.maximumnum.v9f16(<9 x half> %a, <9 x half> %b)
ret <9 x half> %c
}
define <16 x half> @max_v16f16(<16 x half> %a, <16 x half> %b) {
; FULLFP16-LABEL: max_v16f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fminnm v2.8h, v2.8h, v2.8h
; FULLFP16-NEXT: fminnm v0.8h, v0.8h, v0.8h
; FULLFP16-NEXT: fminnm v3.8h, v3.8h, v3.8h
; FULLFP16-NEXT: fminnm v1.8h, v1.8h, v1.8h
; FULLFP16-NEXT: fmaxnm v0.8h, v0.8h, v2.8h
; FULLFP16-NEXT: fmaxnm v1.8h, v1.8h, v3.8h
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: max_v16f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: mov h6, v2.h[1]
; NOFULLFP16-NEXT: mov h7, v0.h[1]
; NOFULLFP16-NEXT: fcvt s4, h2
; NOFULLFP16-NEXT: fcvt s5, h0
; NOFULLFP16-NEXT: mov h16, v3.h[1]
; NOFULLFP16-NEXT: mov h17, v1.h[1]
; NOFULLFP16-NEXT: mov h18, v2.h[2]
; NOFULLFP16-NEXT: mov h19, v0.h[2]
; NOFULLFP16-NEXT: fcvt s20, h3
; NOFULLFP16-NEXT: fcvt s21, h1
; NOFULLFP16-NEXT: mov h22, v3.h[2]
; NOFULLFP16-NEXT: mov h23, v1.h[2]
; NOFULLFP16-NEXT: fcvt s6, h6
; NOFULLFP16-NEXT: fcvt s7, h7
; NOFULLFP16-NEXT: mov h24, v0.h[6]
; NOFULLFP16-NEXT: fmaxnm s4, s5, s4
; NOFULLFP16-NEXT: fcvt s5, h16
; NOFULLFP16-NEXT: fcvt s16, h17
; NOFULLFP16-NEXT: fcvt s17, h18
; NOFULLFP16-NEXT: fcvt s18, h19
; NOFULLFP16-NEXT: mov h19, v0.h[3]
; NOFULLFP16-NEXT: fmaxnm s20, s21, s20
; NOFULLFP16-NEXT: fcvt s21, h22
; NOFULLFP16-NEXT: mov h22, v3.h[3]
; NOFULLFP16-NEXT: fmaxnm s6, s7, s6
; NOFULLFP16-NEXT: mov h7, v2.h[3]
; NOFULLFP16-NEXT: mov h25, v1.h[6]
; NOFULLFP16-NEXT: fcvt h4, s4
; NOFULLFP16-NEXT: fmaxnm s5, s16, s5
; NOFULLFP16-NEXT: fcvt s16, h23
; NOFULLFP16-NEXT: mov h23, v1.h[3]
; NOFULLFP16-NEXT: fmaxnm s17, s18, s17
; NOFULLFP16-NEXT: fcvt s18, h19
; NOFULLFP16-NEXT: fcvt h6, s6
; NOFULLFP16-NEXT: fcvt s7, h7
; NOFULLFP16-NEXT: fcvt h19, s5
; NOFULLFP16-NEXT: fcvt h5, s20
; NOFULLFP16-NEXT: fmaxnm s16, s16, s21
; NOFULLFP16-NEXT: fcvt s20, h23
; NOFULLFP16-NEXT: fcvt h17, s17
; NOFULLFP16-NEXT: mov h21, v2.h[4]
; NOFULLFP16-NEXT: mov h23, v1.h[4]
; NOFULLFP16-NEXT: mov v4.h[1], v6.h[0]
; NOFULLFP16-NEXT: fcvt s6, h22
; NOFULLFP16-NEXT: mov h22, v0.h[4]
; NOFULLFP16-NEXT: fmaxnm s7, s18, s7
; NOFULLFP16-NEXT: mov h18, v3.h[4]
; NOFULLFP16-NEXT: mov v5.h[1], v19.h[0]
; NOFULLFP16-NEXT: fcvt h16, s16
; NOFULLFP16-NEXT: fmaxnm s6, s20, s6
; NOFULLFP16-NEXT: mov v4.h[2], v17.h[0]
; NOFULLFP16-NEXT: fcvt s17, h21
; NOFULLFP16-NEXT: fcvt s19, h22
; NOFULLFP16-NEXT: fcvt h7, s7
; NOFULLFP16-NEXT: fcvt s18, h18
; NOFULLFP16-NEXT: fcvt s20, h23
; NOFULLFP16-NEXT: mov h21, v2.h[5]
; NOFULLFP16-NEXT: mov h22, v0.h[5]
; NOFULLFP16-NEXT: mov v5.h[2], v16.h[0]
; NOFULLFP16-NEXT: mov h16, v3.h[5]
; NOFULLFP16-NEXT: mov h23, v1.h[5]
; NOFULLFP16-NEXT: fcvt h6, s6
; NOFULLFP16-NEXT: mov h0, v0.h[7]
; NOFULLFP16-NEXT: mov h1, v1.h[7]
; NOFULLFP16-NEXT: fmaxnm s17, s19, s17
; NOFULLFP16-NEXT: mov h19, v2.h[6]
; NOFULLFP16-NEXT: mov v4.h[3], v7.h[0]
; NOFULLFP16-NEXT: fmaxnm s18, s20, s18
; NOFULLFP16-NEXT: mov h20, v3.h[6]
; NOFULLFP16-NEXT: fcvt s7, h21
; NOFULLFP16-NEXT: fcvt s21, h22
; NOFULLFP16-NEXT: fcvt s22, h24
; NOFULLFP16-NEXT: mov h2, v2.h[7]
; NOFULLFP16-NEXT: mov v5.h[3], v6.h[0]
; NOFULLFP16-NEXT: fcvt s6, h16
; NOFULLFP16-NEXT: fcvt s16, h23
; NOFULLFP16-NEXT: fcvt h17, s17
; NOFULLFP16-NEXT: fcvt s19, h19
; NOFULLFP16-NEXT: fcvt s23, h25
; NOFULLFP16-NEXT: fcvt h18, s18
; NOFULLFP16-NEXT: fcvt s20, h20
; NOFULLFP16-NEXT: mov h3, v3.h[7]
; NOFULLFP16-NEXT: fmaxnm s7, s21, s7
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s0, h0
; NOFULLFP16-NEXT: fmaxnm s6, s16, s6
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: mov v4.h[4], v17.h[0]
; NOFULLFP16-NEXT: fmaxnm s16, s22, s19
; NOFULLFP16-NEXT: mov v5.h[4], v18.h[0]
; NOFULLFP16-NEXT: fmaxnm s17, s23, s20
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fcvt h7, s7
; NOFULLFP16-NEXT: fmaxnm s0, s0, s2
; NOFULLFP16-NEXT: fcvt h6, s6
; NOFULLFP16-NEXT: fcvt h2, s16
; NOFULLFP16-NEXT: fmaxnm s1, s1, s3
; NOFULLFP16-NEXT: mov v4.h[5], v7.h[0]
; NOFULLFP16-NEXT: fcvt h0, s0
; NOFULLFP16-NEXT: mov v5.h[5], v6.h[0]
; NOFULLFP16-NEXT: fcvt h6, s17
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: mov v4.h[6], v2.h[0]
; NOFULLFP16-NEXT: mov v5.h[6], v6.h[0]
; NOFULLFP16-NEXT: mov v4.h[7], v0.h[0]
; NOFULLFP16-NEXT: mov v5.h[7], v1.h[0]
; NOFULLFP16-NEXT: mov v0.16b, v4.16b
; NOFULLFP16-NEXT: mov v1.16b, v5.16b
; NOFULLFP16-NEXT: ret
entry:
%c = call <16 x half> @llvm.maximumnum.v16f16(<16 x half> %a, <16 x half> %b)
ret <16 x half> %c
}
;;;;;;;;;;;;;;;; min_f64
define double @min_f64(double %a, double %b) {
; AARCH64-LABEL: min_f64:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm d1, d1, d1
; AARCH64-NEXT: fminnm d0, d0, d0
; AARCH64-NEXT: fminnm d0, d0, d1
; AARCH64-NEXT: ret
entry:
%c = call double @llvm.minimumnum.f64(double %a, double %b)
ret double %c
}
define <2 x double> @min_v2f64(<2 x double> %a, <2 x double> %b) {
; AARCH64-LABEL: min_v2f64:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm v1.2d, v1.2d, v1.2d
; AARCH64-NEXT: fminnm v0.2d, v0.2d, v0.2d
; AARCH64-NEXT: fminnm v0.2d, v0.2d, v1.2d
; AARCH64-NEXT: ret
entry:
%c = call <2 x double> @llvm.minimumnum.v2f64(<2 x double> %a, <2 x double> %b)
ret <2 x double> %c
}
define <3 x double> @min_v3f64(<3 x double> %a, <3 x double> %b) {
; AARCH64-LABEL: min_v3f64:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: // kill: def $d3 killed $d3 def $q3
; AARCH64-NEXT: // kill: def $d0 killed $d0 def $q0
; AARCH64-NEXT: // kill: def $d4 killed $d4 def $q4
; AARCH64-NEXT: // kill: def $d1 killed $d1 def $q1
; AARCH64-NEXT: // kill: def $d2 killed $d2 def $q2
; AARCH64-NEXT: // kill: def $d5 killed $d5 def $q5
; AARCH64-NEXT: mov v0.d[1], v1.d[0]
; AARCH64-NEXT: mov v3.d[1], v4.d[0]
; AARCH64-NEXT: fminnm v2.2d, v2.2d, v2.2d
; AARCH64-NEXT: fminnm v1.2d, v3.2d, v3.2d
; AARCH64-NEXT: fminnm v0.2d, v0.2d, v0.2d
; AARCH64-NEXT: fminnm v0.2d, v0.2d, v1.2d
; AARCH64-NEXT: fminnm v1.2d, v5.2d, v5.2d
; AARCH64-NEXT: fminnm v2.2d, v2.2d, v1.2d
; AARCH64-NEXT: ext v1.16b, v0.16b, v0.16b, #8
; AARCH64-NEXT: // kill: def $d0 killed $d0 killed $q0
; AARCH64-NEXT: // kill: def $d1 killed $d1 killed $q1
; AARCH64-NEXT: // kill: def $d2 killed $d2 killed $q2
; AARCH64-NEXT: ret
entry:
%c = call <3 x double> @llvm.minimumnum.v3f64(<3 x double> %a, <3 x double> %b)
ret <3 x double> %c
}
define <4 x double> @min_v4f64(<4 x double> %a, <4 x double> %b) {
; AARCH64-LABEL: min_v4f64:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm v2.2d, v2.2d, v2.2d
; AARCH64-NEXT: fminnm v0.2d, v0.2d, v0.2d
; AARCH64-NEXT: fminnm v3.2d, v3.2d, v3.2d
; AARCH64-NEXT: fminnm v1.2d, v1.2d, v1.2d
; AARCH64-NEXT: fminnm v0.2d, v0.2d, v2.2d
; AARCH64-NEXT: fminnm v1.2d, v1.2d, v3.2d
; AARCH64-NEXT: ret
entry:
%c = call <4 x double> @llvm.minimumnum.v4f64(<4 x double> %a, <4 x double> %b)
ret <4 x double> %c
}
;;;;;;;;;;;;;;;;;; min_f32
define float @min_f32(float %a, float %b) {
; AARCH64-LABEL: min_f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm s1, s1, s1
; AARCH64-NEXT: fminnm s0, s0, s0
; AARCH64-NEXT: fminnm s0, s0, s1
; AARCH64-NEXT: ret
entry:
%c = call float @llvm.minimumnum.f32(float %a, float %b)
ret float %c
}
define <2 x float> @min_v2f32(<2 x float> %a, <2 x float> %b) {
; AARCH64-LABEL: min_v2f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm v1.2s, v1.2s, v1.2s
; AARCH64-NEXT: fminnm v0.2s, v0.2s, v0.2s
; AARCH64-NEXT: fminnm v0.2s, v0.2s, v1.2s
; AARCH64-NEXT: ret
entry:
%c = call <2 x float> @llvm.minimumnum.v2f32(<2 x float> %a, <2 x float> %b)
ret <2 x float> %c
}
define <3 x float> @min_v3f32(<3 x float> %a, <3 x float> %b) {
; AARCH64-LABEL: min_v3f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm v1.4s, v1.4s, v1.4s
; AARCH64-NEXT: fminnm v0.4s, v0.4s, v0.4s
; AARCH64-NEXT: fminnm v0.4s, v0.4s, v1.4s
; AARCH64-NEXT: ret
entry:
%c = call <3 x float> @llvm.minimumnum.v3f32(<3 x float> %a, <3 x float> %b)
ret <3 x float> %c
}
define <4 x float> @min_v4f32(<4 x float> %a, <4 x float> %b) {
; AARCH64-LABEL: min_v4f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm v1.4s, v1.4s, v1.4s
; AARCH64-NEXT: fminnm v0.4s, v0.4s, v0.4s
; AARCH64-NEXT: fminnm v0.4s, v0.4s, v1.4s
; AARCH64-NEXT: ret
entry:
%c = call <4 x float> @llvm.minimumnum.v4f32(<4 x float> %a, <4 x float> %b)
ret <4 x float> %c
}
define <5 x float> @min_v5f32(<5 x float> %a, <5 x float> %b) {
; AARCH64-LABEL: min_v5f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: // kill: def $s0 killed $s0 def $q0
; AARCH64-NEXT: // kill: def $s5 killed $s5 def $q5
; AARCH64-NEXT: // kill: def $s1 killed $s1 def $q1
; AARCH64-NEXT: // kill: def $s6 killed $s6 def $q6
; AARCH64-NEXT: // kill: def $s2 killed $s2 def $q2
; AARCH64-NEXT: // kill: def $s7 killed $s7 def $q7
; AARCH64-NEXT: // kill: def $s3 killed $s3 def $q3
; AARCH64-NEXT: mov x8, sp
; AARCH64-NEXT: // kill: def $s4 killed $s4 def $q4
; AARCH64-NEXT: mov v0.s[1], v1.s[0]
; AARCH64-NEXT: mov v5.s[1], v6.s[0]
; AARCH64-NEXT: mov v0.s[2], v2.s[0]
; AARCH64-NEXT: mov v5.s[2], v7.s[0]
; AARCH64-NEXT: ldr s2, [sp, #8]
; AARCH64-NEXT: fminnm v2.4s, v2.4s, v2.4s
; AARCH64-NEXT: mov v0.s[3], v3.s[0]
; AARCH64-NEXT: ld1 { v5.s }[3], [x8]
; AARCH64-NEXT: fminnm v3.4s, v4.4s, v4.4s
; AARCH64-NEXT: fminnm v1.4s, v5.4s, v5.4s
; AARCH64-NEXT: fminnm v0.4s, v0.4s, v0.4s
; AARCH64-NEXT: fminnm v4.4s, v3.4s, v2.4s
; AARCH64-NEXT: // kill: def $s4 killed $s4 killed $q4
; AARCH64-NEXT: fminnm v0.4s, v0.4s, v1.4s
; AARCH64-NEXT: mov s1, v0.s[1]
; AARCH64-NEXT: mov s2, v0.s[2]
; AARCH64-NEXT: mov s3, v0.s[3]
; AARCH64-NEXT: // kill: def $s0 killed $s0 killed $q0
; AARCH64-NEXT: ret
entry:
%c = call <5 x float> @llvm.minimumnum.v5f32(<5 x float> %a, <5 x float> %b)
ret <5 x float> %c
}
define <8 x float> @min_v8f32(<8 x float> %a, <8 x float> %b) {
; AARCH64-LABEL: min_v8f32:
; AARCH64: // %bb.0: // %entry
; AARCH64-NEXT: fminnm v2.4s, v2.4s, v2.4s
; AARCH64-NEXT: fminnm v0.4s, v0.4s, v0.4s
; AARCH64-NEXT: fminnm v3.4s, v3.4s, v3.4s
; AARCH64-NEXT: fminnm v1.4s, v1.4s, v1.4s
; AARCH64-NEXT: fminnm v0.4s, v0.4s, v2.4s
; AARCH64-NEXT: fminnm v1.4s, v1.4s, v3.4s
; AARCH64-NEXT: ret
entry:
%c = call <8 x float> @llvm.minimumnum.v8f32(<8 x float> %a, <8 x float> %b)
ret <8 x float> %c
}
;;;;;;;;;;;;;;;;;; min_f16
define half @min_f16(half %a, half %b) {
; FULLFP16-LABEL: min_f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fminnm h1, h1, h1
; FULLFP16-NEXT: fminnm h0, h0, h0
; FULLFP16-NEXT: fminnm h0, h0, h1
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: min_f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fcvt s0, h0
; NOFULLFP16-NEXT: fminnm s0, s0, s1
; NOFULLFP16-NEXT: fcvt h0, s0
; NOFULLFP16-NEXT: ret
entry:
%c = call half @llvm.minimumnum.f16(half %a, half %b)
ret half %c
}
define <2 x half> @min_v2f16(<2 x half> %a, <2 x half> %b) {
; FULLFP16-LABEL: min_v2f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fminnm v1.4h, v1.4h, v1.4h
; FULLFP16-NEXT: fminnm v0.4h, v0.4h, v0.4h
; FULLFP16-NEXT: fminnm v0.4h, v0.4h, v1.4h
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: min_v2f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: // kill: def $d1 killed $d1 def $q1
; NOFULLFP16-NEXT: // kill: def $d0 killed $d0 def $q0
; NOFULLFP16-NEXT: mov h2, v1.h[1]
; NOFULLFP16-NEXT: mov h3, v0.h[1]
; NOFULLFP16-NEXT: mov h4, v1.h[2]
; NOFULLFP16-NEXT: mov h5, v0.h[2]
; NOFULLFP16-NEXT: fcvt s6, h1
; NOFULLFP16-NEXT: fcvt s7, h0
; NOFULLFP16-NEXT: mov h1, v1.h[3]
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fminnm s2, s3, s2
; NOFULLFP16-NEXT: fcvt s3, h4
; NOFULLFP16-NEXT: fcvt s4, h5
; NOFULLFP16-NEXT: fminnm s5, s7, s6
; NOFULLFP16-NEXT: mov h6, v0.h[3]
; NOFULLFP16-NEXT: fminnm s3, s4, s3
; NOFULLFP16-NEXT: fcvt h2, s2
; NOFULLFP16-NEXT: fcvt h0, s5
; NOFULLFP16-NEXT: fcvt s4, h6
; NOFULLFP16-NEXT: mov v0.h[1], v2.h[0]
; NOFULLFP16-NEXT: fcvt h2, s3
; NOFULLFP16-NEXT: fminnm s1, s4, s1
; NOFULLFP16-NEXT: mov v0.h[2], v2.h[0]
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: mov v0.h[3], v1.h[0]
; NOFULLFP16-NEXT: // kill: def $d0 killed $d0 killed $q0
; NOFULLFP16-NEXT: ret
entry:
%c = call <2 x half> @llvm.minimumnum.v2f16(<2 x half> %a, <2 x half> %b)
ret <2 x half> %c
}
define <4 x half> @min_v4f16(<4 x half> %a, <4 x half> %b) {
; FULLFP16-LABEL: min_v4f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fminnm v1.4h, v1.4h, v1.4h
; FULLFP16-NEXT: fminnm v0.4h, v0.4h, v0.4h
; FULLFP16-NEXT: fminnm v0.4h, v0.4h, v1.4h
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: min_v4f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: // kill: def $d1 killed $d1 def $q1
; NOFULLFP16-NEXT: // kill: def $d0 killed $d0 def $q0
; NOFULLFP16-NEXT: mov h2, v1.h[1]
; NOFULLFP16-NEXT: mov h3, v0.h[1]
; NOFULLFP16-NEXT: mov h4, v1.h[2]
; NOFULLFP16-NEXT: mov h5, v0.h[2]
; NOFULLFP16-NEXT: fcvt s6, h1
; NOFULLFP16-NEXT: fcvt s7, h0
; NOFULLFP16-NEXT: mov h1, v1.h[3]
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fminnm s2, s3, s2
; NOFULLFP16-NEXT: fcvt s3, h4
; NOFULLFP16-NEXT: fcvt s4, h5
; NOFULLFP16-NEXT: fminnm s5, s7, s6
; NOFULLFP16-NEXT: mov h6, v0.h[3]
; NOFULLFP16-NEXT: fminnm s3, s4, s3
; NOFULLFP16-NEXT: fcvt h2, s2
; NOFULLFP16-NEXT: fcvt h0, s5
; NOFULLFP16-NEXT: fcvt s4, h6
; NOFULLFP16-NEXT: mov v0.h[1], v2.h[0]
; NOFULLFP16-NEXT: fcvt h2, s3
; NOFULLFP16-NEXT: fminnm s1, s4, s1
; NOFULLFP16-NEXT: mov v0.h[2], v2.h[0]
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: mov v0.h[3], v1.h[0]
; NOFULLFP16-NEXT: // kill: def $d0 killed $d0 killed $q0
; NOFULLFP16-NEXT: ret
entry:
%c = call <4 x half> @llvm.minimumnum.v4f16(<4 x half> %a, <4 x half> %b)
ret <4 x half> %c
}
define <8 x half> @min_v8f16(<8 x half> %a, <8 x half> %b) {
; FULLFP16-LABEL: min_v8f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fminnm v1.8h, v1.8h, v1.8h
; FULLFP16-NEXT: fminnm v0.8h, v0.8h, v0.8h
; FULLFP16-NEXT: fminnm v0.8h, v0.8h, v1.8h
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: min_v8f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: mov h2, v1.h[1]
; NOFULLFP16-NEXT: mov h3, v0.h[1]
; NOFULLFP16-NEXT: fcvt s4, h1
; NOFULLFP16-NEXT: fcvt s5, h0
; NOFULLFP16-NEXT: mov h6, v1.h[2]
; NOFULLFP16-NEXT: mov h7, v0.h[2]
; NOFULLFP16-NEXT: mov h16, v1.h[3]
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fminnm s4, s5, s4
; NOFULLFP16-NEXT: mov h5, v0.h[3]
; NOFULLFP16-NEXT: fcvt s6, h6
; NOFULLFP16-NEXT: fcvt s7, h7
; NOFULLFP16-NEXT: fcvt s16, h16
; NOFULLFP16-NEXT: fminnm s3, s3, s2
; NOFULLFP16-NEXT: fcvt s5, h5
; NOFULLFP16-NEXT: fcvt h2, s4
; NOFULLFP16-NEXT: fminnm s4, s7, s6
; NOFULLFP16-NEXT: mov h6, v1.h[4]
; NOFULLFP16-NEXT: mov h7, v0.h[4]
; NOFULLFP16-NEXT: fcvt h3, s3
; NOFULLFP16-NEXT: fminnm s5, s5, s16
; NOFULLFP16-NEXT: mov h16, v0.h[5]
; NOFULLFP16-NEXT: fcvt h4, s4
; NOFULLFP16-NEXT: mov v2.h[1], v3.h[0]
; NOFULLFP16-NEXT: fcvt s3, h6
; NOFULLFP16-NEXT: fcvt s6, h7
; NOFULLFP16-NEXT: mov h7, v1.h[5]
; NOFULLFP16-NEXT: fcvt h5, s5
; NOFULLFP16-NEXT: fcvt s16, h16
; NOFULLFP16-NEXT: mov v2.h[2], v4.h[0]
; NOFULLFP16-NEXT: mov h4, v1.h[6]
; NOFULLFP16-NEXT: fminnm s3, s6, s3
; NOFULLFP16-NEXT: mov h6, v0.h[6]
; NOFULLFP16-NEXT: fcvt s7, h7
; NOFULLFP16-NEXT: mov h1, v1.h[7]
; NOFULLFP16-NEXT: mov h0, v0.h[7]
; NOFULLFP16-NEXT: mov v2.h[3], v5.h[0]
; NOFULLFP16-NEXT: fcvt s4, h4
; NOFULLFP16-NEXT: fcvt h3, s3
; NOFULLFP16-NEXT: fcvt s5, h6
; NOFULLFP16-NEXT: fminnm s6, s16, s7
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fcvt s0, h0
; NOFULLFP16-NEXT: mov v2.h[4], v3.h[0]
; NOFULLFP16-NEXT: fminnm s4, s5, s4
; NOFULLFP16-NEXT: fcvt h3, s6
; NOFULLFP16-NEXT: fminnm s0, s0, s1
; NOFULLFP16-NEXT: mov v2.h[5], v3.h[0]
; NOFULLFP16-NEXT: fcvt h3, s4
; NOFULLFP16-NEXT: fcvt h0, s0
; NOFULLFP16-NEXT: mov v2.h[6], v3.h[0]
; NOFULLFP16-NEXT: mov v2.h[7], v0.h[0]
; NOFULLFP16-NEXT: mov v0.16b, v2.16b
; NOFULLFP16-NEXT: ret
entry:
%c = call <8 x half> @llvm.minimumnum.v8f16(<8 x half> %a, <8 x half> %b)
ret <8 x half> %c
}
define <9 x half> @min_v9f16(<9 x half> %a, <9 x half> %b) {
; FULLFP16-LABEL: min_v9f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: // kill: def $h0 killed $h0 def $q0
; FULLFP16-NEXT: // kill: def $h1 killed $h1 def $q1
; FULLFP16-NEXT: // kill: def $h2 killed $h2 def $q2
; FULLFP16-NEXT: add x9, sp, #16
; FULLFP16-NEXT: // kill: def $h3 killed $h3 def $q3
; FULLFP16-NEXT: // kill: def $h4 killed $h4 def $q4
; FULLFP16-NEXT: add x10, sp, #40
; FULLFP16-NEXT: // kill: def $h5 killed $h5 def $q5
; FULLFP16-NEXT: // kill: def $h6 killed $h6 def $q6
; FULLFP16-NEXT: // kill: def $h7 killed $h7 def $q7
; FULLFP16-NEXT: mov v0.h[1], v1.h[0]
; FULLFP16-NEXT: ldr h1, [sp, #8]
; FULLFP16-NEXT: ld1 { v1.h }[1], [x9]
; FULLFP16-NEXT: add x9, sp, #24
; FULLFP16-NEXT: mov v0.h[2], v2.h[0]
; FULLFP16-NEXT: ld1 { v1.h }[2], [x9]
; FULLFP16-NEXT: add x9, sp, #32
; FULLFP16-NEXT: mov v0.h[3], v3.h[0]
; FULLFP16-NEXT: ld1 { v1.h }[3], [x9]
; FULLFP16-NEXT: ldr h2, [x10]
; FULLFP16-NEXT: add x9, sp, #48
; FULLFP16-NEXT: ldr h3, [sp, #72]
; FULLFP16-NEXT: ld1 { v2.h }[1], [x9]
; FULLFP16-NEXT: add x9, sp, #56
; FULLFP16-NEXT: fminnm v3.8h, v3.8h, v3.8h
; FULLFP16-NEXT: mov v0.h[4], v4.h[0]
; FULLFP16-NEXT: ld1 { v2.h }[2], [x9]
; FULLFP16-NEXT: add x9, sp, #64
; FULLFP16-NEXT: mov v0.h[5], v5.h[0]
; FULLFP16-NEXT: ld1 { v2.h }[3], [x9]
; FULLFP16-NEXT: zip1 v1.2d, v1.2d, v2.2d
; FULLFP16-NEXT: ldr h2, [sp]
; FULLFP16-NEXT: mov v0.h[6], v6.h[0]
; FULLFP16-NEXT: fminnm v2.8h, v2.8h, v2.8h
; FULLFP16-NEXT: fminnm v1.8h, v1.8h, v1.8h
; FULLFP16-NEXT: mov v0.h[7], v7.h[0]
; FULLFP16-NEXT: fminnm v2.8h, v2.8h, v3.8h
; FULLFP16-NEXT: fminnm v0.8h, v0.8h, v0.8h
; FULLFP16-NEXT: str h2, [x8, #16]
; FULLFP16-NEXT: fminnm v0.8h, v0.8h, v1.8h
; FULLFP16-NEXT: str q0, [x8]
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: min_v9f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: ldr h16, [sp, #16]
; NOFULLFP16-NEXT: ldr h17, [sp, #8]
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fcvt s0, h0
; NOFULLFP16-NEXT: ldr h18, [sp, #24]
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s16, h16
; NOFULLFP16-NEXT: fcvt s17, h17
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fcvt s18, h18
; NOFULLFP16-NEXT: fcvt s4, h4
; NOFULLFP16-NEXT: fcvt s5, h5
; NOFULLFP16-NEXT: fminnm s1, s1, s16
; NOFULLFP16-NEXT: fminnm s0, s0, s17
; NOFULLFP16-NEXT: ldr h16, [sp, #32]
; NOFULLFP16-NEXT: fminnm s2, s2, s18
; NOFULLFP16-NEXT: ldr h17, [sp, #40]
; NOFULLFP16-NEXT: fcvt s16, h16
; NOFULLFP16-NEXT: fcvt s17, h17
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: fcvt h0, s0
; NOFULLFP16-NEXT: fcvt h2, s2
; NOFULLFP16-NEXT: fminnm s3, s3, s16
; NOFULLFP16-NEXT: fminnm s4, s4, s17
; NOFULLFP16-NEXT: mov v0.h[1], v1.h[0]
; NOFULLFP16-NEXT: ldr h1, [sp, #48]
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: fcvt h3, s3
; NOFULLFP16-NEXT: fcvt h4, s4
; NOFULLFP16-NEXT: mov v0.h[2], v2.h[0]
; NOFULLFP16-NEXT: ldr h2, [sp, #56]
; NOFULLFP16-NEXT: fminnm s1, s5, s1
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: ldr h5, [sp, #64]
; NOFULLFP16-NEXT: mov v0.h[3], v3.h[0]
; NOFULLFP16-NEXT: fcvt s3, h6
; NOFULLFP16-NEXT: ldr h6, [sp, #72]
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: fcvt s6, h6
; NOFULLFP16-NEXT: mov v0.h[4], v4.h[0]
; NOFULLFP16-NEXT: fminnm s2, s3, s2
; NOFULLFP16-NEXT: fcvt s3, h5
; NOFULLFP16-NEXT: fcvt s4, h7
; NOFULLFP16-NEXT: ldr h5, [sp]
; NOFULLFP16-NEXT: fcvt s5, h5
; NOFULLFP16-NEXT: mov v0.h[5], v1.h[0]
; NOFULLFP16-NEXT: fcvt h1, s2
; NOFULLFP16-NEXT: fminnm s2, s4, s3
; NOFULLFP16-NEXT: fminnm s3, s5, s6
; NOFULLFP16-NEXT: mov v0.h[6], v1.h[0]
; NOFULLFP16-NEXT: fcvt h1, s2
; NOFULLFP16-NEXT: fcvt h2, s3
; NOFULLFP16-NEXT: mov v0.h[7], v1.h[0]
; NOFULLFP16-NEXT: str h2, [x8, #16]
; NOFULLFP16-NEXT: str q0, [x8]
; NOFULLFP16-NEXT: ret
entry:
%c = call <9 x half> @llvm.minimumnum.v9f16(<9 x half> %a, <9 x half> %b)
ret <9 x half> %c
}
define <16 x half> @min_v16f16(<16 x half> %a, <16 x half> %b) {
; FULLFP16-LABEL: min_v16f16:
; FULLFP16: // %bb.0: // %entry
; FULLFP16-NEXT: fminnm v2.8h, v2.8h, v2.8h
; FULLFP16-NEXT: fminnm v0.8h, v0.8h, v0.8h
; FULLFP16-NEXT: fminnm v3.8h, v3.8h, v3.8h
; FULLFP16-NEXT: fminnm v1.8h, v1.8h, v1.8h
; FULLFP16-NEXT: fminnm v0.8h, v0.8h, v2.8h
; FULLFP16-NEXT: fminnm v1.8h, v1.8h, v3.8h
; FULLFP16-NEXT: ret
;
; NOFULLFP16-LABEL: min_v16f16:
; NOFULLFP16: // %bb.0: // %entry
; NOFULLFP16-NEXT: mov h6, v2.h[1]
; NOFULLFP16-NEXT: mov h7, v0.h[1]
; NOFULLFP16-NEXT: fcvt s4, h2
; NOFULLFP16-NEXT: fcvt s5, h0
; NOFULLFP16-NEXT: mov h16, v3.h[1]
; NOFULLFP16-NEXT: mov h17, v1.h[1]
; NOFULLFP16-NEXT: mov h18, v2.h[2]
; NOFULLFP16-NEXT: mov h19, v0.h[2]
; NOFULLFP16-NEXT: fcvt s20, h3
; NOFULLFP16-NEXT: fcvt s21, h1
; NOFULLFP16-NEXT: mov h22, v3.h[2]
; NOFULLFP16-NEXT: mov h23, v1.h[2]
; NOFULLFP16-NEXT: fcvt s6, h6
; NOFULLFP16-NEXT: fcvt s7, h7
; NOFULLFP16-NEXT: mov h24, v0.h[6]
; NOFULLFP16-NEXT: fminnm s4, s5, s4
; NOFULLFP16-NEXT: fcvt s5, h16
; NOFULLFP16-NEXT: fcvt s16, h17
; NOFULLFP16-NEXT: fcvt s17, h18
; NOFULLFP16-NEXT: fcvt s18, h19
; NOFULLFP16-NEXT: mov h19, v0.h[3]
; NOFULLFP16-NEXT: fminnm s20, s21, s20
; NOFULLFP16-NEXT: fcvt s21, h22
; NOFULLFP16-NEXT: mov h22, v3.h[3]
; NOFULLFP16-NEXT: fminnm s6, s7, s6
; NOFULLFP16-NEXT: mov h7, v2.h[3]
; NOFULLFP16-NEXT: mov h25, v1.h[6]
; NOFULLFP16-NEXT: fcvt h4, s4
; NOFULLFP16-NEXT: fminnm s5, s16, s5
; NOFULLFP16-NEXT: fcvt s16, h23
; NOFULLFP16-NEXT: mov h23, v1.h[3]
; NOFULLFP16-NEXT: fminnm s17, s18, s17
; NOFULLFP16-NEXT: fcvt s18, h19
; NOFULLFP16-NEXT: fcvt h6, s6
; NOFULLFP16-NEXT: fcvt s7, h7
; NOFULLFP16-NEXT: fcvt h19, s5
; NOFULLFP16-NEXT: fcvt h5, s20
; NOFULLFP16-NEXT: fminnm s16, s16, s21
; NOFULLFP16-NEXT: fcvt s20, h23
; NOFULLFP16-NEXT: fcvt h17, s17
; NOFULLFP16-NEXT: mov h21, v2.h[4]
; NOFULLFP16-NEXT: mov h23, v1.h[4]
; NOFULLFP16-NEXT: mov v4.h[1], v6.h[0]
; NOFULLFP16-NEXT: fcvt s6, h22
; NOFULLFP16-NEXT: mov h22, v0.h[4]
; NOFULLFP16-NEXT: fminnm s7, s18, s7
; NOFULLFP16-NEXT: mov h18, v3.h[4]
; NOFULLFP16-NEXT: mov v5.h[1], v19.h[0]
; NOFULLFP16-NEXT: fcvt h16, s16
; NOFULLFP16-NEXT: fminnm s6, s20, s6
; NOFULLFP16-NEXT: mov v4.h[2], v17.h[0]
; NOFULLFP16-NEXT: fcvt s17, h21
; NOFULLFP16-NEXT: fcvt s19, h22
; NOFULLFP16-NEXT: fcvt h7, s7
; NOFULLFP16-NEXT: fcvt s18, h18
; NOFULLFP16-NEXT: fcvt s20, h23
; NOFULLFP16-NEXT: mov h21, v2.h[5]
; NOFULLFP16-NEXT: mov h22, v0.h[5]
; NOFULLFP16-NEXT: mov v5.h[2], v16.h[0]
; NOFULLFP16-NEXT: mov h16, v3.h[5]
; NOFULLFP16-NEXT: mov h23, v1.h[5]
; NOFULLFP16-NEXT: fcvt h6, s6
; NOFULLFP16-NEXT: mov h0, v0.h[7]
; NOFULLFP16-NEXT: mov h1, v1.h[7]
; NOFULLFP16-NEXT: fminnm s17, s19, s17
; NOFULLFP16-NEXT: mov h19, v2.h[6]
; NOFULLFP16-NEXT: mov v4.h[3], v7.h[0]
; NOFULLFP16-NEXT: fminnm s18, s20, s18
; NOFULLFP16-NEXT: mov h20, v3.h[6]
; NOFULLFP16-NEXT: fcvt s7, h21
; NOFULLFP16-NEXT: fcvt s21, h22
; NOFULLFP16-NEXT: fcvt s22, h24
; NOFULLFP16-NEXT: mov h2, v2.h[7]
; NOFULLFP16-NEXT: mov v5.h[3], v6.h[0]
; NOFULLFP16-NEXT: fcvt s6, h16
; NOFULLFP16-NEXT: fcvt s16, h23
; NOFULLFP16-NEXT: fcvt h17, s17
; NOFULLFP16-NEXT: fcvt s19, h19
; NOFULLFP16-NEXT: fcvt s23, h25
; NOFULLFP16-NEXT: fcvt h18, s18
; NOFULLFP16-NEXT: fcvt s20, h20
; NOFULLFP16-NEXT: mov h3, v3.h[7]
; NOFULLFP16-NEXT: fminnm s7, s21, s7
; NOFULLFP16-NEXT: fcvt s2, h2
; NOFULLFP16-NEXT: fcvt s0, h0
; NOFULLFP16-NEXT: fminnm s6, s16, s6
; NOFULLFP16-NEXT: fcvt s1, h1
; NOFULLFP16-NEXT: mov v4.h[4], v17.h[0]
; NOFULLFP16-NEXT: fminnm s16, s22, s19
; NOFULLFP16-NEXT: mov v5.h[4], v18.h[0]
; NOFULLFP16-NEXT: fminnm s17, s23, s20
; NOFULLFP16-NEXT: fcvt s3, h3
; NOFULLFP16-NEXT: fcvt h7, s7
; NOFULLFP16-NEXT: fminnm s0, s0, s2
; NOFULLFP16-NEXT: fcvt h6, s6
; NOFULLFP16-NEXT: fcvt h2, s16
; NOFULLFP16-NEXT: fminnm s1, s1, s3
; NOFULLFP16-NEXT: mov v4.h[5], v7.h[0]
; NOFULLFP16-NEXT: fcvt h0, s0
; NOFULLFP16-NEXT: mov v5.h[5], v6.h[0]
; NOFULLFP16-NEXT: fcvt h6, s17
; NOFULLFP16-NEXT: fcvt h1, s1
; NOFULLFP16-NEXT: mov v4.h[6], v2.h[0]
; NOFULLFP16-NEXT: mov v5.h[6], v6.h[0]
; NOFULLFP16-NEXT: mov v4.h[7], v0.h[0]
; NOFULLFP16-NEXT: mov v5.h[7], v1.h[0]
; NOFULLFP16-NEXT: mov v0.16b, v4.16b
; NOFULLFP16-NEXT: mov v1.16b, v5.16b
; NOFULLFP16-NEXT: ret
entry:
%c = call <16 x half> @llvm.minimumnum.v16f16(<16 x half> %a, <16 x half> %b)
ret <16 x half> %c
}
Reference in New Issue View Git Blame Copy Permalink