File: fp8-sve-fmla.ll

package info (click to toggle)
llvm-toolchain-20 1%3A20.1.6-1~exp1
  • links: PTS, VCS
  • area: main
  • in suites: experimental
  • size: 2,111,304 kB
  • sloc: cpp: 7,438,677; ansic: 1,393,822; asm: 1,012,926; python: 241,650; f90: 86,635; objc: 75,479; lisp: 42,144; pascal: 17,286; sh: 10,027; ml: 5,082; perl: 4,730; awk: 3,523; makefile: 3,349; javascript: 2,251; xml: 892; fortran: 672
file content (114 lines) | stat: -rw-r--r-- 5,397 bytes parent folder | download | duplicates (6)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: llc -mattr=+sve2,+fp8,+fp8fma  < %s | FileCheck %s
; RUN: llc -mattr=+sme,+fp8,+ssve-fp8fma --force-streaming < %s | FileCheck %s

target triple = "aarch64-linux"

define <vscale x 8 x half> @fmla_2way_bot(<vscale x 8 x half> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2) {
; CHECK-LABEL: fmla_2way_bot:
; CHECK:       // %bb.0:
; CHECK-NEXT:    fmlalb z0.h, z1.b, z2.b
; CHECK-NEXT:    ret
    %r = call <vscale x 8 x half> @llvm.aarch64.sve.fp8.fmlalb.nxv8f16(<vscale x 8 x half> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2)
    ret <vscale x 8 x half> %r
}

define <vscale x 8 x half> @fmla_2way_top(<vscale x 8 x half> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2) {
; CHECK-LABEL: fmla_2way_top:
; CHECK:       // %bb.0:
; CHECK-NEXT:    fmlalt z0.h, z1.b, z2.b
; CHECK-NEXT:    ret
    %r = call <vscale x 8 x half> @llvm.aarch64.sve.fp8.fmlalt.nxv8f16(<vscale x 8 x half> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2)
    ret <vscale x 8 x half> %r
}

define <vscale x 8 x half> @fmla_2way_bot_lane(<vscale x 8 x half> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2) {
; CHECK-LABEL: fmla_2way_bot_lane:
; CHECK:       // %bb.0:
; CHECK-NEXT:    fmlalb z0.h, z1.b, z2.b[3]
; CHECK-NEXT:    ret
    %r = call <vscale x 8 x half> @llvm.aarch64.sve.fp8.fmlalb.lane.nxv8f16(<vscale x 8 x half> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2, i32 3)
    ret <vscale x 8 x half> %r
}

define <vscale x 8 x half> @fmla_2way_top_lane(<vscale x 8 x half> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2) {
; CHECK-LABEL: fmla_2way_top_lane:
; CHECK:       // %bb.0:
; CHECK-NEXT:    fmlalt z0.h, z1.b, z2.b[3]
; CHECK-NEXT:    ret
    %r = call <vscale x 8 x half> @llvm.aarch64.sve.fp8.fmlalt.lane.nxv8f16(<vscale x 8 x half> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2, i32 3)
    ret <vscale x 8 x half> %r
}

define <vscale x 4 x float> @fmla_4way_bb(<vscale x 4 x float> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2) {
; CHECK-LABEL: fmla_4way_bb:
; CHECK:       // %bb.0:
; CHECK-NEXT:    fmlallbb z0.s, z1.b, z2.b
; CHECK-NEXT:    ret
    %r = call <vscale x 4 x float> @llvm.aarch64.sve.fp8.fmlallbb.nxv4f32(<vscale x 4 x float> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2)
    ret <vscale x 4 x float> %r
}

define <vscale x 4 x float> @fmla_4way_bt(<vscale x 4 x float> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2) {
; CHECK-LABEL: fmla_4way_bt:
; CHECK:       // %bb.0:
; CHECK-NEXT:    fmlallbt z0.s, z1.b, z2.b
; CHECK-NEXT:    ret
    %r = call <vscale x 4 x float> @llvm.aarch64.sve.fp8.fmlallbt.nxv4f32(<vscale x 4 x float> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2)
    ret <vscale x 4 x float> %r
}

define <vscale x 4 x float> @fmla_4way_tb(<vscale x 4 x float> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2) {
; CHECK-LABEL: fmla_4way_tb:
; CHECK:       // %bb.0:
; CHECK-NEXT:    fmlalltb z0.s, z1.b, z2.b
; CHECK-NEXT:    ret
    %r = call <vscale x 4 x float> @llvm.aarch64.sve.fp8.fmlalltb.nxv4f32(<vscale x 4 x float> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2)
    ret <vscale x 4 x float> %r
}

define <vscale x 4 x float> @fmla_4way_tt(<vscale x 4 x float> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2) {
; CHECK-LABEL: fmla_4way_tt:
; CHECK:       // %bb.0:
; CHECK-NEXT:    fmlalltt z0.s, z1.b, z2.b
; CHECK-NEXT:    ret
    %r = call <vscale x 4 x float> @llvm.aarch64.sve.fp8.fmlalltt.nxv4f32(<vscale x 4 x float> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2)
    ret <vscale x 4 x float> %r
}

define <vscale x 4 x float> @fmla_4way_bb_lane(<vscale x 4 x float> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2) {
; CHECK-LABEL: fmla_4way_bb_lane:
; CHECK:       // %bb.0:
; CHECK-NEXT:    fmlallbb z0.s, z1.b, z2.b[3]
; CHECK-NEXT:    ret
    %r = call <vscale x 4 x float> @llvm.aarch64.sve.fp8.fmlallbb.lane.nxv4f32(<vscale x 4 x float> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2, i32 3)
    ret <vscale x 4 x float> %r
}

define <vscale x 4 x float> @fmla_4way_bt_lane(<vscale x 4 x float> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2) {
; CHECK-LABEL: fmla_4way_bt_lane:
; CHECK:       // %bb.0:
; CHECK-NEXT:    fmlallbt z0.s, z1.b, z2.b[3]
; CHECK-NEXT:    ret
    %r = call <vscale x 4 x float> @llvm.aarch64.sve.fp8.fmlallbt.lane.nxv4f32(<vscale x 4 x float> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2, i32 3)
    ret <vscale x 4 x float> %r
}

define <vscale x 4 x float> @fmla_4way_tb_lane(<vscale x 4 x float> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2) {
; CHECK-LABEL: fmla_4way_tb_lane:
; CHECK:       // %bb.0:
; CHECK-NEXT:    fmlalltb z0.s, z1.b, z2.b[3]
; CHECK-NEXT:    ret
    %r = call <vscale x 4 x float> @llvm.aarch64.sve.fp8.fmlalltb.lane.nxv4f32(<vscale x 4 x float> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2, i32 3)
    ret <vscale x 4 x float> %r
}

define <vscale x 4 x float> @fmla_4way_tt_lane(<vscale x 4 x float> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2) {
; CHECK-LABEL: fmla_4way_tt_lane:
; CHECK:       // %bb.0:
; CHECK-NEXT:    fmlalltt z0.s, z1.b, z2.b[3]
; CHECK-NEXT:    ret
    %r = call <vscale x 4 x float> @llvm.aarch64.sve.fp8.fmlalltt.lane.nxv4f32(<vscale x 4 x float> %a, <vscale x 16 x i8> %s1, <vscale x 16 x i8> %s2, i32 3)
    ret <vscale x 4 x float> %r
}