xref: /aosp_15_r20/external/XNNPACK/scripts/generate-f32-igemm.sh (revision 4bdc94577ba0e567308109d787f7fec7b531ce36)
1#!/bin/sh
2# Copyright 2019 Google LLC
3#
4# This source code is licensed under the BSD-style license found in the
5# LICENSE file in the root directory of this source tree.
6
7#################################### Scalar ###################################
8### Generic C micro-kernels
9tools/xngen src/f32-igemm/scalar.c.in -D MR=1 -D NR=4 -D WASM=0 -D ACTIVATION=LINEAR -o src/f32-igemm/gen/1x4-scalar.c &
10tools/xngen src/f32-igemm/scalar.c.in -D MR=2 -D NR=4 -D WASM=0 -D ACTIVATION=LINEAR -o src/f32-igemm/gen/2x4-scalar.c &
11tools/xngen src/f32-igemm/scalar.c.in -D MR=4 -D NR=2 -D WASM=0 -D ACTIVATION=LINEAR -o src/f32-igemm/gen/4x2-scalar.c &
12tools/xngen src/f32-igemm/scalar.c.in -D MR=4 -D NR=4 -D WASM=0 -D ACTIVATION=LINEAR -o src/f32-igemm/gen/4x4-scalar.c &
13
14tools/xngen src/f32-igemm/scalar.c.in -D MR=1 -D NR=4 -D WASM=0 -D ACTIVATION=RELU   -o src/f32-igemm/gen/1x4-relu-scalar.c &
15tools/xngen src/f32-igemm/scalar.c.in -D MR=2 -D NR=4 -D WASM=0 -D ACTIVATION=RELU   -o src/f32-igemm/gen/2x4-relu-scalar.c &
16tools/xngen src/f32-igemm/scalar.c.in -D MR=4 -D NR=2 -D WASM=0 -D ACTIVATION=RELU   -o src/f32-igemm/gen/4x2-relu-scalar.c &
17tools/xngen src/f32-igemm/scalar.c.in -D MR=4 -D NR=4 -D WASM=0 -D ACTIVATION=RELU   -o src/f32-igemm/gen/4x4-relu-scalar.c &
18
19tools/xngen src/f32-igemm/scalar.c.in -D MR=1 -D NR=4 -D WASM=0 -D ACTIVATION=MINMAX -o src/f32-igemm/gen/1x4-minmax-scalar.c &
20tools/xngen src/f32-igemm/scalar.c.in -D MR=2 -D NR=4 -D WASM=0 -D ACTIVATION=MINMAX -o src/f32-igemm/gen/2x4-minmax-scalar.c &
21tools/xngen src/f32-igemm/scalar.c.in -D MR=4 -D NR=2 -D WASM=0 -D ACTIVATION=MINMAX -o src/f32-igemm/gen/4x2-minmax-scalar.c &
22tools/xngen src/f32-igemm/scalar.c.in -D MR=4 -D NR=4 -D WASM=0 -D ACTIVATION=MINMAX -o src/f32-igemm/gen/4x4-minmax-scalar.c &
23
24### WAsm-specific micro-kernels
25tools/xngen src/f32-igemm/scalar.c.in -D MR=1 -D NR=4 -D WASM=1 -D ACTIVATION=RELU   -o src/f32-igemm/gen/1x4-relu-wasm.c &
26tools/xngen src/f32-igemm/scalar.c.in -D MR=2 -D NR=4 -D WASM=1 -D ACTIVATION=RELU   -o src/f32-igemm/gen/2x4-relu-wasm.c &
27tools/xngen src/f32-igemm/scalar.c.in -D MR=4 -D NR=2 -D WASM=1 -D ACTIVATION=RELU   -o src/f32-igemm/gen/4x2-relu-wasm.c &
28tools/xngen src/f32-igemm/scalar.c.in -D MR=4 -D NR=4 -D WASM=1 -D ACTIVATION=RELU   -o src/f32-igemm/gen/4x4-relu-wasm.c &
29
30tools/xngen src/f32-igemm/scalar.c.in -D MR=1 -D NR=4 -D WASM=1 -D ACTIVATION=MINMAX -o src/f32-igemm/gen/1x4-minmax-wasm.c &
31tools/xngen src/f32-igemm/scalar.c.in -D MR=2 -D NR=4 -D WASM=1 -D ACTIVATION=MINMAX -o src/f32-igemm/gen/2x4-minmax-wasm.c &
32tools/xngen src/f32-igemm/scalar.c.in -D MR=4 -D NR=2 -D WASM=1 -D ACTIVATION=MINMAX -o src/f32-igemm/gen/4x2-minmax-wasm.c &
33tools/xngen src/f32-igemm/scalar.c.in -D MR=4 -D NR=4 -D WASM=1 -D ACTIVATION=MINMAX -o src/f32-igemm/gen/4x4-minmax-wasm.c &
34
35################################## WAsm SIMD ##################################
36### LOAD1+BROADCAST micro-kernels
37tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=1 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=ARM     -o src/f32-igemm/gen/1x8-minmax-wasmsimd-arm-loadsplat.c &
38tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=3 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=ARM     -o src/f32-igemm/gen/3x8-minmax-wasmsimd-arm-loadsplat.c &
39tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=4 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=ARM     -o src/f32-igemm/gen/4x8-minmax-wasmsimd-arm-loadsplat.c &
40tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=5 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=ARM     -o src/f32-igemm/gen/5x8-minmax-wasmsimd-arm-loadsplat.c &
41tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=6 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=ARM     -o src/f32-igemm/gen/6x8-minmax-wasmsimd-arm-loadsplat.c &
42
43tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=1 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=X86     -o src/f32-igemm/gen/1x8-minmax-wasmsimd-x86-loadsplat.c &
44tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=3 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=X86     -o src/f32-igemm/gen/3x8-minmax-wasmsimd-x86-loadsplat.c &
45tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=4 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=X86     -o src/f32-igemm/gen/4x8-minmax-wasmsimd-x86-loadsplat.c &
46tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=5 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=X86     -o src/f32-igemm/gen/5x8-minmax-wasmsimd-x86-loadsplat.c &
47tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=6 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=X86     -o src/f32-igemm/gen/6x8-minmax-wasmsimd-x86-loadsplat.c &
48
49tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=1 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/1x8-minmax-wasmrelaxedsimd-loadsplat.c &
50tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=3 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/3x8-minmax-wasmrelaxedsimd-loadsplat.c &
51tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=4 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/4x8-minmax-wasmrelaxedsimd-loadsplat.c &
52tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=5 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/5x8-minmax-wasmrelaxedsimd-loadsplat.c &
53tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=6 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/6x8-minmax-wasmrelaxedsimd-loadsplat.c &
54
55tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=1 -D NR=8 -D FMA=1 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/1x8-minmax-wasmrelaxedsimd-fma-loadsplat.c &
56tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=3 -D NR=8 -D FMA=1 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/3x8-minmax-wasmrelaxedsimd-fma-loadsplat.c &
57tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=4 -D NR=8 -D FMA=1 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/4x8-minmax-wasmrelaxedsimd-fma-loadsplat.c &
58tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=5 -D NR=8 -D FMA=1 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/5x8-minmax-wasmrelaxedsimd-fma-loadsplat.c &
59tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=6 -D NR=8 -D FMA=1 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/6x8-minmax-wasmrelaxedsimd-fma-loadsplat.c &
60
61tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=1 -D NR=8 -D FMA=0 -D ACTIVATION=RELU                   -o src/f32-igemm/gen/1x8-relu-wasmsimd-loadsplat.c &
62tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=3 -D NR=8 -D FMA=0 -D ACTIVATION=RELU                   -o src/f32-igemm/gen/3x8-relu-wasmsimd-loadsplat.c &
63tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=4 -D NR=8 -D FMA=0 -D ACTIVATION=RELU                   -o src/f32-igemm/gen/4x8-relu-wasmsimd-loadsplat.c &
64tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=5 -D NR=8 -D FMA=0 -D ACTIVATION=RELU                   -o src/f32-igemm/gen/5x8-relu-wasmsimd-loadsplat.c &
65tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=6 -D NR=8 -D FMA=0 -D ACTIVATION=RELU                   -o src/f32-igemm/gen/6x8-relu-wasmsimd-loadsplat.c &
66
67tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=1 -D NR=8 -D FMA=1 -D ACTIVATION=RELU   -D ARCH=RELAXED -o src/f32-igemm/gen/1x8-relu-wasmrelaxedsimd-fma-loadsplat.c &
68tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=3 -D NR=8 -D FMA=1 -D ACTIVATION=RELU   -D ARCH=RELAXED -o src/f32-igemm/gen/3x8-relu-wasmrelaxedsimd-fma-loadsplat.c &
69tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=4 -D NR=8 -D FMA=1 -D ACTIVATION=RELU   -D ARCH=RELAXED -o src/f32-igemm/gen/4x8-relu-wasmrelaxedsimd-fma-loadsplat.c &
70tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=5 -D NR=8 -D FMA=1 -D ACTIVATION=RELU   -D ARCH=RELAXED -o src/f32-igemm/gen/5x8-relu-wasmrelaxedsimd-fma-loadsplat.c &
71tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=6 -D NR=8 -D FMA=1 -D ACTIVATION=RELU   -D ARCH=RELAXED -o src/f32-igemm/gen/6x8-relu-wasmrelaxedsimd-fma-loadsplat.c &
72
73tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=1 -D NR=8 -D FMA=0 -D ACTIVATION=LINEAR                 -o src/f32-igemm/gen/1x8-wasmsimd-loadsplat.c &
74tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=3 -D NR=8 -D FMA=0 -D ACTIVATION=LINEAR                 -o src/f32-igemm/gen/3x8-wasmsimd-loadsplat.c &
75tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=4 -D NR=8 -D FMA=0 -D ACTIVATION=LINEAR                 -o src/f32-igemm/gen/4x8-wasmsimd-loadsplat.c &
76tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=5 -D NR=8 -D FMA=0 -D ACTIVATION=LINEAR                 -o src/f32-igemm/gen/5x8-wasmsimd-loadsplat.c &
77tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=6 -D NR=8 -D FMA=0 -D ACTIVATION=LINEAR                 -o src/f32-igemm/gen/6x8-wasmsimd-loadsplat.c &
78
79tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=1 -D NR=8 -D FMA=1 -D ACTIVATION=LINEAR -D ARCH=RELAXED -o src/f32-igemm/gen/1x8-wasmrelaxedsimd-fma-loadsplat.c &
80tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=3 -D NR=8 -D FMA=1 -D ACTIVATION=LINEAR -D ARCH=RELAXED -o src/f32-igemm/gen/3x8-wasmrelaxedsimd-fma-loadsplat.c &
81tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=4 -D NR=8 -D FMA=1 -D ACTIVATION=LINEAR -D ARCH=RELAXED -o src/f32-igemm/gen/4x8-wasmrelaxedsimd-fma-loadsplat.c &
82tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=5 -D NR=8 -D FMA=1 -D ACTIVATION=LINEAR -D ARCH=RELAXED -o src/f32-igemm/gen/5x8-wasmrelaxedsimd-fma-loadsplat.c &
83tools/xngen src/f32-igemm/wasmsimd-loadsplat.c.in -D MR=6 -D NR=8 -D FMA=1 -D ACTIVATION=LINEAR -D ARCH=RELAXED -o src/f32-igemm/gen/6x8-wasmrelaxedsimd-fma-loadsplat.c &
84
85### LOAD4+DUPLICATE micro-kernels
86tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=1 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=ARM     -o src/f32-igemm/gen/1x8-minmax-wasmsimd-arm-splat.c &
87tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=3 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=ARM     -o src/f32-igemm/gen/3x8-minmax-wasmsimd-arm-splat.c &
88tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=4 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=ARM     -o src/f32-igemm/gen/4x8-minmax-wasmsimd-arm-splat.c &
89tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=5 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=ARM     -o src/f32-igemm/gen/5x8-minmax-wasmsimd-arm-splat.c &
90tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=6 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=ARM     -o src/f32-igemm/gen/6x8-minmax-wasmsimd-arm-splat.c &
91
92tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=1 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=X86     -o src/f32-igemm/gen/1x8-minmax-wasmsimd-x86-splat.c &
93tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=3 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=X86     -o src/f32-igemm/gen/3x8-minmax-wasmsimd-x86-splat.c &
94tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=4 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=X86     -o src/f32-igemm/gen/4x8-minmax-wasmsimd-x86-splat.c &
95tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=5 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=X86     -o src/f32-igemm/gen/5x8-minmax-wasmsimd-x86-splat.c &
96tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=6 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=X86     -o src/f32-igemm/gen/6x8-minmax-wasmsimd-x86-splat.c &
97
98tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=1 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/1x8-minmax-wasmrelaxedsimd-splat.c &
99tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=3 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/3x8-minmax-wasmrelaxedsimd-splat.c &
100tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=4 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/4x8-minmax-wasmrelaxedsimd-splat.c &
101tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=5 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/5x8-minmax-wasmrelaxedsimd-splat.c &
102tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=6 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/6x8-minmax-wasmrelaxedsimd-splat.c &
103
104tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=1 -D NR=8 -D FMA=1 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/1x8-minmax-wasmrelaxedsimd-fma-splat.c &
105tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=3 -D NR=8 -D FMA=1 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/3x8-minmax-wasmrelaxedsimd-fma-splat.c &
106tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=4 -D NR=8 -D FMA=1 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/4x8-minmax-wasmrelaxedsimd-fma-splat.c &
107tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=5 -D NR=8 -D FMA=1 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/5x8-minmax-wasmrelaxedsimd-fma-splat.c &
108tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=6 -D NR=8 -D FMA=1 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/6x8-minmax-wasmrelaxedsimd-fma-splat.c &
109
110tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=1 -D NR=8 -D FMA=0 -D ACTIVATION=RELU                   -o src/f32-igemm/gen/1x8-relu-wasmsimd-splat.c &
111tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=3 -D NR=8 -D FMA=0 -D ACTIVATION=RELU                   -o src/f32-igemm/gen/3x8-relu-wasmsimd-splat.c &
112tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=4 -D NR=8 -D FMA=0 -D ACTIVATION=RELU                   -o src/f32-igemm/gen/4x8-relu-wasmsimd-splat.c &
113tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=5 -D NR=8 -D FMA=0 -D ACTIVATION=RELU                   -o src/f32-igemm/gen/5x8-relu-wasmsimd-splat.c &
114tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=6 -D NR=8 -D FMA=0 -D ACTIVATION=RELU                   -o src/f32-igemm/gen/6x8-relu-wasmsimd-splat.c &
115
116tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=1 -D NR=8 -D FMA=1 -D ACTIVATION=RELU   -D ARCH=RELAXED -o src/f32-igemm/gen/1x8-relu-wasmrelaxedsimd-fma-splat.c &
117tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=3 -D NR=8 -D FMA=1 -D ACTIVATION=RELU   -D ARCH=RELAXED -o src/f32-igemm/gen/3x8-relu-wasmrelaxedsimd-fma-splat.c &
118tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=4 -D NR=8 -D FMA=1 -D ACTIVATION=RELU   -D ARCH=RELAXED -o src/f32-igemm/gen/4x8-relu-wasmrelaxedsimd-fma-splat.c &
119tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=5 -D NR=8 -D FMA=1 -D ACTIVATION=RELU   -D ARCH=RELAXED -o src/f32-igemm/gen/5x8-relu-wasmrelaxedsimd-fma-splat.c &
120tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=6 -D NR=8 -D FMA=1 -D ACTIVATION=RELU   -D ARCH=RELAXED -o src/f32-igemm/gen/6x8-relu-wasmrelaxedsimd-fma-splat.c &
121
122tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=1 -D NR=8 -D FMA=0 -D ACTIVATION=LINEAR                 -o src/f32-igemm/gen/1x8-wasmsimd-splat.c &
123tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=3 -D NR=8 -D FMA=0 -D ACTIVATION=LINEAR                 -o src/f32-igemm/gen/3x8-wasmsimd-splat.c &
124tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=4 -D NR=8 -D FMA=0 -D ACTIVATION=LINEAR                 -o src/f32-igemm/gen/4x8-wasmsimd-splat.c &
125tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=5 -D NR=8 -D FMA=0 -D ACTIVATION=LINEAR                 -o src/f32-igemm/gen/5x8-wasmsimd-splat.c &
126tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=6 -D NR=8 -D FMA=0 -D ACTIVATION=LINEAR                 -o src/f32-igemm/gen/6x8-wasmsimd-splat.c &
127
128tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=1 -D NR=8 -D FMA=1 -D ACTIVATION=LINEAR -D ARCH=RELAXED -o src/f32-igemm/gen/1x8-wasmrelaxedsimd-fma-splat.c &
129tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=3 -D NR=8 -D FMA=1 -D ACTIVATION=LINEAR -D ARCH=RELAXED -o src/f32-igemm/gen/3x8-wasmrelaxedsimd-fma-splat.c &
130tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=4 -D NR=8 -D FMA=1 -D ACTIVATION=LINEAR -D ARCH=RELAXED -o src/f32-igemm/gen/4x8-wasmrelaxedsimd-fma-splat.c &
131tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=5 -D NR=8 -D FMA=1 -D ACTIVATION=LINEAR -D ARCH=RELAXED -o src/f32-igemm/gen/5x8-wasmrelaxedsimd-fma-splat.c &
132tools/xngen src/f32-igemm/wasmsimd-splat.c.in -D MR=6 -D NR=8 -D FMA=1 -D ACTIVATION=LINEAR -D ARCH=RELAXED -o src/f32-igemm/gen/6x8-wasmrelaxedsimd-fma-splat.c &
133
134### LOAD4+PERMUTE micro-kernels
135tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=1 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=ARM     -o src/f32-igemm/gen/1x8s4-minmax-wasmsimd-arm.c &
136tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=3 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=ARM     -o src/f32-igemm/gen/3x8s4-minmax-wasmsimd-arm.c &
137tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=4 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=ARM     -o src/f32-igemm/gen/4x8s4-minmax-wasmsimd-arm.c &
138tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=5 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=ARM     -o src/f32-igemm/gen/5x8s4-minmax-wasmsimd-arm.c &
139tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=6 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=ARM     -o src/f32-igemm/gen/6x8s4-minmax-wasmsimd-arm.c &
140
141tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=1 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=X86     -o src/f32-igemm/gen/1x8s4-minmax-wasmsimd-x86.c &
142tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=3 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=X86     -o src/f32-igemm/gen/3x8s4-minmax-wasmsimd-x86.c &
143tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=4 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=X86     -o src/f32-igemm/gen/4x8s4-minmax-wasmsimd-x86.c &
144tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=5 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=X86     -o src/f32-igemm/gen/5x8s4-minmax-wasmsimd-x86.c &
145tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=6 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=X86     -o src/f32-igemm/gen/6x8s4-minmax-wasmsimd-x86.c &
146
147tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=1 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/1x8s4-minmax-wasmrelaxedsimd.c &
148tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=3 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/3x8s4-minmax-wasmrelaxedsimd.c &
149tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=4 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/4x8s4-minmax-wasmrelaxedsimd.c &
150tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=5 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/5x8s4-minmax-wasmrelaxedsimd.c &
151tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=6 -D NR=8 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/6x8s4-minmax-wasmrelaxedsimd.c &
152
153tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=1 -D NR=8 -D FMA=1 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/1x8s4-minmax-wasmrelaxedsimd-fma.c &
154tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=3 -D NR=8 -D FMA=1 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/3x8s4-minmax-wasmrelaxedsimd-fma.c &
155tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=4 -D NR=8 -D FMA=1 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/4x8s4-minmax-wasmrelaxedsimd-fma.c &
156tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=5 -D NR=8 -D FMA=1 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/5x8s4-minmax-wasmrelaxedsimd-fma.c &
157tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=6 -D NR=8 -D FMA=1 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/6x8s4-minmax-wasmrelaxedsimd-fma.c &
158
159tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=1 -D NR=8 -D FMA=0 -D ACTIVATION=RELU                   -o src/f32-igemm/gen/1x8s4-relu-wasmsimd.c &
160tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=3 -D NR=8 -D FMA=0 -D ACTIVATION=RELU                   -o src/f32-igemm/gen/3x8s4-relu-wasmsimd.c &
161tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=4 -D NR=8 -D FMA=0 -D ACTIVATION=RELU                   -o src/f32-igemm/gen/4x8s4-relu-wasmsimd.c &
162tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=5 -D NR=8 -D FMA=0 -D ACTIVATION=RELU                   -o src/f32-igemm/gen/5x8s4-relu-wasmsimd.c &
163tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=6 -D NR=8 -D FMA=0 -D ACTIVATION=RELU                   -o src/f32-igemm/gen/6x8s4-relu-wasmsimd.c &
164
165tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=1 -D NR=8 -D FMA=1 -D ACTIVATION=RELU   -D ARCH=RELAXED -o src/f32-igemm/gen/1x8s4-relu-wasmrelaxedsimd-fma.c &
166tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=3 -D NR=8 -D FMA=1 -D ACTIVATION=RELU   -D ARCH=RELAXED -o src/f32-igemm/gen/3x8s4-relu-wasmrelaxedsimd-fma.c &
167tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=4 -D NR=8 -D FMA=1 -D ACTIVATION=RELU   -D ARCH=RELAXED -o src/f32-igemm/gen/4x8s4-relu-wasmrelaxedsimd-fma.c &
168tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=5 -D NR=8 -D FMA=1 -D ACTIVATION=RELU   -D ARCH=RELAXED -o src/f32-igemm/gen/5x8s4-relu-wasmrelaxedsimd-fma.c &
169tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=6 -D NR=8 -D FMA=1 -D ACTIVATION=RELU   -D ARCH=RELAXED -o src/f32-igemm/gen/6x8s4-relu-wasmrelaxedsimd-fma.c &
170
171tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=1 -D NR=8 -D FMA=0 -D ACTIVATION=LINEAR                 -o src/f32-igemm/gen/1x8s4-wasmsimd.c &
172tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=3 -D NR=8 -D FMA=0 -D ACTIVATION=LINEAR                 -o src/f32-igemm/gen/3x8s4-wasmsimd.c &
173tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=4 -D NR=8 -D FMA=0 -D ACTIVATION=LINEAR                 -o src/f32-igemm/gen/4x8s4-wasmsimd.c &
174tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=5 -D NR=8 -D FMA=0 -D ACTIVATION=LINEAR                 -o src/f32-igemm/gen/5x8s4-wasmsimd.c &
175tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=6 -D NR=8 -D FMA=0 -D ACTIVATION=LINEAR                 -o src/f32-igemm/gen/6x8s4-wasmsimd.c &
176
177tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=1 -D NR=8 -D FMA=1 -D ACTIVATION=LINEAR -D ARCH=RELAXED -o src/f32-igemm/gen/1x8s4-wasmrelaxedsimd-fma.c &
178tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=3 -D NR=8 -D FMA=1 -D ACTIVATION=LINEAR -D ARCH=RELAXED -o src/f32-igemm/gen/3x8s4-wasmrelaxedsimd-fma.c &
179tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=4 -D NR=8 -D FMA=1 -D ACTIVATION=LINEAR -D ARCH=RELAXED -o src/f32-igemm/gen/4x8s4-wasmrelaxedsimd-fma.c &
180tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=5 -D NR=8 -D FMA=1 -D ACTIVATION=LINEAR -D ARCH=RELAXED -o src/f32-igemm/gen/5x8s4-wasmrelaxedsimd-fma.c &
181tools/xngen src/f32-igemm/wasmsimd-s4.c.in -D MR=6 -D NR=8 -D FMA=1 -D ACTIVATION=LINEAR -D ARCH=RELAXED -o src/f32-igemm/gen/6x8s4-wasmrelaxedsimd-fma.c &
182
183### MRx2 micro-kernels
184tools/xngen src/f32-igemm/MRx2c4-wasmsimd.c.in -D MR=4 -D NR=2 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=ARM     -o src/f32-igemm/gen/4x2c4-minmax-wasmsimd-arm.c &
185tools/xngen src/f32-igemm/MRx2c4-wasmsimd.c.in -D MR=4 -D NR=2 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=X86     -o src/f32-igemm/gen/4x2c4-minmax-wasmsimd-x86.c &
186tools/xngen src/f32-igemm/MRx2c4-wasmsimd.c.in -D MR=4 -D NR=2 -D FMA=0 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/4x2c4-minmax-wasmrelaxedsimd.c &
187tools/xngen src/f32-igemm/MRx2c4-wasmsimd.c.in -D MR=4 -D NR=2 -D FMA=1 -D ACTIVATION=MINMAX -D ARCH=RELAXED -o src/f32-igemm/gen/4x2c4-minmax-wasmrelaxedsimd-fma.c &
188tools/xngen src/f32-igemm/MRx2c4-wasmsimd.c.in -D MR=4 -D NR=2 -D FMA=0 -D ACTIVATION=RELU                   -o src/f32-igemm/gen/4x2c4-relu-wasmsimd.c &
189tools/xngen src/f32-igemm/MRx2c4-wasmsimd.c.in -D MR=4 -D NR=2 -D FMA=1 -D ACTIVATION=RELU   -D ARCH=RELAXED -o src/f32-igemm/gen/4x2c4-relu-wasmrelaxedsimd-fma.c &
190tools/xngen src/f32-igemm/MRx2c4-wasmsimd.c.in -D MR=4 -D NR=2 -D FMA=0 -D ACTIVATION=LINEAR                 -o src/f32-igemm/gen/4x2c4-wasmsimd.c &
191tools/xngen src/f32-igemm/MRx2c4-wasmsimd.c.in -D MR=4 -D NR=2 -D FMA=1 -D ACTIVATION=LINEAR -D ARCH=RELAXED -o src/f32-igemm/gen/4x2c4-wasmrelaxedsimd-fma.c &
192
193############################### AArch64 assembly ##############################
194### LD64 micro-kernels
195tools/xngen src/f32-igemm/4x8-aarch64-neonfma-ld64.S.in -o src/f32-igemm/gen/4x8-minmax-aarch64-neonfma-ld64.S &
196tools/xngen src/f32-igemm/6x8-aarch64-neonfma-ld64.S.in -o src/f32-igemm/gen/6x8-minmax-aarch64-neonfma-ld64.S &
197
198### LD128 micro-kernels
199tools/xngen src/f32-igemm/4x8-aarch64-neonfma-ld128.S.in -o src/f32-igemm/gen/4x8-minmax-aarch64-neonfma-ld128.S &
200tools/xngen src/f32-igemm/6x8-aarch64-neonfma-ld128.S.in -o src/f32-igemm/gen/6x8-minmax-aarch64-neonfma-ld128.S &
201
202### MRx2 micro-kernels
203tools/xngen src/f32-igemm/4x2-aarch64-neonfma-ld64.S.in  -o src/f32-igemm/gen/4x2-minmax-aarch64-neonfma-ld64.S &
204tools/xngen src/f32-igemm/4x2-aarch64-neonfma-cortex-a75.S.in -D PREFETCH=0 -o src/f32-igemm/gen/4x2-minmax-aarch64-neonfma-cortex-a75.S &
205tools/xngen src/f32-igemm/4x2-aarch64-neonfma-cortex-a75.S.in -D PREFETCH=1 -o src/f32-igemm/gen/4x2-minmax-aarch64-neonfma-prfm-cortex-a75.S &
206
207### Cortex A53 micro-kernels
208tools/xngen src/f32-igemm/1x8-aarch64-neonfma-cortex-a53.S.in -D PREFETCH=0 -o src/f32-igemm/gen/1x8-minmax-aarch64-neonfma-cortex-a53.S &
209tools/xngen src/f32-igemm/1x8-aarch64-neonfma-cortex-a53.S.in -D PREFETCH=1 -o src/f32-igemm/gen/1x8-minmax-aarch64-neonfma-prfm-cortex-a53.S &
210tools/xngen src/f32-igemm/4x8-aarch64-neonfma-cortex-a53.S.in -D PREFETCH=0 -o src/f32-igemm/gen/4x8-minmax-aarch64-neonfma-cortex-a53.S &
211tools/xngen src/f32-igemm/4x8-aarch64-neonfma-cortex-a53.S.in -D PREFETCH=1 -o src/f32-igemm/gen/4x8-minmax-aarch64-neonfma-prfm-cortex-a53.S &
212tools/xngen src/f32-igemm/6x8-aarch64-neonfma-cortex-a53.S.in -D PREFETCH=0 -o src/f32-igemm/gen/6x8-minmax-aarch64-neonfma-cortex-a53.S &
213tools/xngen src/f32-igemm/6x8-aarch64-neonfma-cortex-a53.S.in -D PREFETCH=1 -o src/f32-igemm/gen/6x8-minmax-aarch64-neonfma-prfm-cortex-a53.S &
214
215### Cortex A75 micro-kernels
216tools/xngen src/f32-igemm/1x8-aarch64-neonfma-cortex-a75.S.in -D PREFETCH=0 -o src/f32-igemm/gen/1x8-minmax-aarch64-neonfma-cortex-a75.S &
217tools/xngen src/f32-igemm/1x8-aarch64-neonfma-cortex-a75.S.in -D PREFETCH=1 -o src/f32-igemm/gen/1x8-minmax-aarch64-neonfma-prfm-cortex-a75.S &
218tools/xngen src/f32-igemm/4x8-aarch64-neonfma-cortex-a75.S.in -D PREFETCH=0 -o src/f32-igemm/gen/4x8-minmax-aarch64-neonfma-cortex-a75.S &
219tools/xngen src/f32-igemm/4x8-aarch64-neonfma-cortex-a75.S.in -D PREFETCH=1 -o src/f32-igemm/gen/4x8-minmax-aarch64-neonfma-prfm-cortex-a75.S &
220tools/xngen src/f32-igemm/5x8-aarch64-neonfma-cortex-a75.S.in -D PREFETCH=0 -o src/f32-igemm/gen/5x8-minmax-aarch64-neonfma-cortex-a75.S &
221tools/xngen src/f32-igemm/5x8-aarch64-neonfma-cortex-a75.S.in -D PREFETCH=1 -o src/f32-igemm/gen/5x8-minmax-aarch64-neonfma-prfm-cortex-a75.S &
222tools/xngen src/f32-igemm/6x8-aarch64-neonfma-cortex-a75.S.in -D PREFETCH=0 -o src/f32-igemm/gen/6x8-minmax-aarch64-neonfma-cortex-a75.S &
223tools/xngen src/f32-igemm/6x8-aarch64-neonfma-cortex-a75.S.in -D PREFETCH=1 -o src/f32-igemm/gen/6x8-minmax-aarch64-neonfma-prfm-cortex-a75.S &
224
225############################### AArch32 assembly ##############################
226tools/xngen src/f32-igemm/4x8-aarch32-neon-cortex-a53.S.in       -D PREFETCH=0 -o src/f32-igemm/gen/4x8-minmax-aarch32-neon-cortex-a53.S &
227tools/xngen src/f32-igemm/4x8-aarch32-neon-cortex-a53.S.in       -D PREFETCH=1 -o src/f32-igemm/gen/4x8-minmax-aarch32-neon-prfm-cortex-a53.S &
228tools/xngen src/f32-igemm/4x8-aarch32-neon-cortex-a75.S.in       -D PREFETCH=0 -o src/f32-igemm/gen/4x8-minmax-aarch32-neon-cortex-a75.S &
229tools/xngen src/f32-igemm/4x8-aarch32-neon-cortex-a75.S.in       -D PREFETCH=1 -o src/f32-igemm/gen/4x8-minmax-aarch32-neon-prfm-cortex-a75.S &
230tools/xngen src/f32-igemm/4x8-minmax-aarch32-neon-cortex-a7.S.in -D PREFETCH=1 -o src/f32-igemm/gen/4x8-minmax-aarch32-neon-cortex-a7.S &
231tools/xngen src/f32-igemm/4x8-minmax-aarch32-neon-ld64.S.in      -D PREFETCH=0 -o src/f32-igemm/gen/4x8-minmax-aarch32-neon-ld64.S &
232
233################################### ARM NEON ##################################
234### LD64 micro-kernels
235tools/xngen src/f32-igemm/neon-ld64.c.in      -D MR=1 -D NR=8 -D FMA=0 -D DUP=0 -o src/f32-igemm/gen/1x8-minmax-neon-lane-ld64.c &
236tools/xngen src/f32-igemm/neon-ld64.c.in      -D MR=1 -D NR=8 -D FMA=1 -D DUP=0 -o src/f32-igemm/gen/1x8-minmax-neonfma-lane-ld64.c &
237tools/xngen src/f32-igemm/neon-ld64.c.in      -D MR=4 -D NR=4 -D FMA=0 -D DUP=0 -o src/f32-igemm/gen/4x4-minmax-neon-lane-ld64.c &
238tools/xngen src/f32-igemm/neon-ld64.c.in      -D MR=4 -D NR=4 -D FMA=1 -D DUP=0 -o src/f32-igemm/gen/4x4-minmax-neonfma-lane-ld64.c &
239tools/xngen src/f32-igemm/neon-ld64.c.in      -D MR=4 -D NR=8 -D FMA=0 -D DUP=0 -o src/f32-igemm/gen/4x8-minmax-neon-lane-ld64.c &
240tools/xngen src/f32-igemm/neon-ld64.c.in      -D MR=4 -D NR=8 -D FMA=1 -D DUP=0 -o src/f32-igemm/gen/4x8-minmax-neonfma-lane-ld64.c &
241tools/xngen src/f32-igemm/neon-ld64.c.in      -D MR=6 -D NR=8 -D FMA=0 -D DUP=0 -o src/f32-igemm/gen/6x8-minmax-neon-lane-ld64.c &
242tools/xngen src/f32-igemm/neon-ld64.c.in      -D MR=6 -D NR=8 -D FMA=1 -D DUP=0 -o src/f32-igemm/gen/6x8-minmax-neonfma-lane-ld64.c &
243### LD128 micro-kernels
244tools/xngen src/f32-igemm/neon-ld128.c.in     -D MR=4 -D NR=8 -D FMA=0 -D DUP=0 -o src/f32-igemm/gen/4x8-minmax-neon-lane-ld128.c &
245tools/xngen src/f32-igemm/neon-ld128.c.in     -D MR=4 -D NR=8 -D FMA=1 -D DUP=0 -o src/f32-igemm/gen/4x8-minmax-neonfma-lane-ld128.c &
246tools/xngen src/f32-igemm/neon-ld128.c.in     -D MR=6 -D NR=8 -D FMA=0 -D DUP=0 -o src/f32-igemm/gen/6x8-minmax-neon-lane-ld128.c &
247tools/xngen src/f32-igemm/neon-ld128.c.in     -D MR=6 -D NR=8 -D FMA=1 -D DUP=0 -o src/f32-igemm/gen/6x8-minmax-neonfma-lane-ld128.c &
248### MRx2 micro-kernels
249tools/xngen src/f32-igemm/MRx2-neon-ld64.c.in -D MR=4 -D NR=2 -D FMA=0 -D DUP=0 -o src/f32-igemm/gen/4x2-minmax-neon-lane-ld64.c &
250tools/xngen src/f32-igemm/MRx2-neon-ld64.c.in -D MR=4 -D NR=2 -D FMA=1 -D DUP=0 -o src/f32-igemm/gen/4x2-minmax-neonfma-lane-ld64.c &
251tools/xngen src/f32-igemm/MRx2-neon-ld64.c.in -D MR=6 -D NR=2 -D FMA=0 -D DUP=0 -o src/f32-igemm/gen/6x2-minmax-neon-lane-ld64.c &
252tools/xngen src/f32-igemm/MRx2-neon-ld64.c.in -D MR=6 -D NR=2 -D FMA=1 -D DUP=0 -o src/f32-igemm/gen/6x2-minmax-neonfma-lane-ld64.c &
253### DUP LD64 micro-kernels
254tools/xngen src/f32-igemm/neon-ld64.c.in      -D MR=1 -D NR=8 -D FMA=0 -D DUP=1 -o src/f32-igemm/gen/1x8-minmax-neon-dup-ld64.c &
255tools/xngen src/f32-igemm/neon-ld64.c.in      -D MR=1 -D NR=8 -D FMA=1 -D DUP=1 -o src/f32-igemm/gen/1x8-minmax-neonfma-dup-ld64.c &
256tools/xngen src/f32-igemm/neon-ld64.c.in      -D MR=4 -D NR=8 -D FMA=0 -D DUP=1 -o src/f32-igemm/gen/4x8-minmax-neon-dup-ld64.c &
257tools/xngen src/f32-igemm/neon-ld64.c.in      -D MR=4 -D NR=8 -D FMA=1 -D DUP=1 -o src/f32-igemm/gen/4x8-minmax-neonfma-dup-ld64.c &
258tools/xngen src/f32-igemm/neon-ld64.c.in      -D MR=6 -D NR=8 -D FMA=0 -D DUP=1 -o src/f32-igemm/gen/6x8-minmax-neon-dup-ld64.c &
259tools/xngen src/f32-igemm/neon-ld64.c.in      -D MR=6 -D NR=8 -D FMA=1 -D DUP=1 -o src/f32-igemm/gen/6x8-minmax-neonfma-dup-ld64.c &
260### DUP LD128 micro-kernels
261tools/xngen src/f32-igemm/neon-ld128.c.in     -D MR=4 -D NR=8 -D FMA=0 -D DUP=1 -o src/f32-igemm/gen/4x8-minmax-neon-dup-ld128.c &
262tools/xngen src/f32-igemm/neon-ld128.c.in     -D MR=4 -D NR=8 -D FMA=1 -D DUP=1 -o src/f32-igemm/gen/4x8-minmax-neonfma-dup-ld128.c &
263tools/xngen src/f32-igemm/neon-ld128.c.in     -D MR=6 -D NR=8 -D FMA=0 -D DUP=1 -o src/f32-igemm/gen/6x8-minmax-neon-dup-ld128.c &
264tools/xngen src/f32-igemm/neon-ld128.c.in     -D MR=6 -D NR=8 -D FMA=1 -D DUP=1 -o src/f32-igemm/gen/6x8-minmax-neonfma-dup-ld128.c &
265### LOAD4+PERMUTE micro-kernels
266tools/xngen src/f32-igemm/neon-shuffle.c.in   -D MR=1 -D NR=8 -D FMA=0 -o src/f32-igemm/gen/1x8s4-minmax-neon.c &
267tools/xngen src/f32-igemm/neon-shuffle.c.in   -D MR=1 -D NR=8 -D FMA=1 -o src/f32-igemm/gen/1x8s4-minmax-neonfma.c &
268tools/xngen src/f32-igemm/neon-shuffle.c.in   -D MR=4 -D NR=8 -D FMA=0 -o src/f32-igemm/gen/4x8s4-minmax-neon.c &
269tools/xngen src/f32-igemm/neon-shuffle.c.in   -D MR=4 -D NR=8 -D FMA=1 -o src/f32-igemm/gen/4x8s4-minmax-neonfma.c &
270tools/xngen src/f32-igemm/neon-shuffle.c.in   -D MR=6 -D NR=8 -D FMA=0 -o src/f32-igemm/gen/6x8s4-minmax-neon.c &
271tools/xngen src/f32-igemm/neon-shuffle.c.in   -D MR=6 -D NR=8 -D FMA=1 -o src/f32-igemm/gen/6x8s4-minmax-neonfma.c &
272tools/xngen src/f32-igemm/neon-shuffle.c.in   -D MR=8 -D NR=8 -D FMA=0 -o src/f32-igemm/gen/8x8s4-minmax-neon.c &
273tools/xngen src/f32-igemm/neon-shuffle.c.in   -D MR=8 -D NR=8 -D FMA=1 -o src/f32-igemm/gen/8x8s4-minmax-neonfma.c &
274
275################################### x86 SSE ###################################
276### LOAD1+BROADCAST micro-kernels
277tools/xngen src/f32-igemm/sse-load1.c.in -D MR=1 -D NR=8 -o src/f32-igemm/gen/1x8-minmax-sse-load1.c &
278tools/xngen src/f32-igemm/sse-load1.c.in -D MR=3 -D NR=8 -o src/f32-igemm/gen/3x8-minmax-sse-load1.c &
279tools/xngen src/f32-igemm/sse-load1.c.in -D MR=4 -D NR=8 -o src/f32-igemm/gen/4x8-minmax-sse-load1.c &
280tools/xngen src/f32-igemm/sse-load1.c.in -D MR=5 -D NR=8 -o src/f32-igemm/gen/5x8-minmax-sse-load1.c &
281
282### LOAD4+DUPLICATE micro-kernels
283tools/xngen src/f32-igemm/sse-dup.c.in -D MR=1 -D NR=8 -D SSE=1 -o src/f32-igemm/gen/1x8-minmax-sse-dup.c &
284tools/xngen src/f32-igemm/sse-dup.c.in -D MR=3 -D NR=8 -D SSE=1 -o src/f32-igemm/gen/3x8-minmax-sse-dup.c &
285tools/xngen src/f32-igemm/sse-dup.c.in -D MR=4 -D NR=8 -D SSE=1 -o src/f32-igemm/gen/4x8-minmax-sse-dup.c &
286tools/xngen src/f32-igemm/sse-dup.c.in -D MR=5 -D NR=8 -D SSE=1 -o src/f32-igemm/gen/5x8-minmax-sse-dup.c &
287
288tools/xngen src/f32-igemm/sse-dup.c.in -D MR=1 -D NR=8 -D SSE=2 -o src/f32-igemm/gen/1x8-minmax-sse2-dup.c &
289tools/xngen src/f32-igemm/sse-dup.c.in -D MR=3 -D NR=8 -D SSE=2 -o src/f32-igemm/gen/3x8-minmax-sse2-dup.c &
290tools/xngen src/f32-igemm/sse-dup.c.in -D MR=4 -D NR=8 -D SSE=2 -o src/f32-igemm/gen/4x8-minmax-sse2-dup.c &
291tools/xngen src/f32-igemm/sse-dup.c.in -D MR=5 -D NR=8 -D SSE=2 -o src/f32-igemm/gen/5x8-minmax-sse2-dup.c &
292
293### LOAD4+PERMUTE micro-kernels
294tools/xngen src/f32-igemm/sse-shuffle.c.in -D MR=1 -D NR=8 -o src/f32-igemm/gen/1x8s4-minmax-sse.c &
295tools/xngen src/f32-igemm/sse-shuffle.c.in -D MR=3 -D NR=8 -o src/f32-igemm/gen/3x8s4-minmax-sse.c &
296tools/xngen src/f32-igemm/sse-shuffle.c.in -D MR=4 -D NR=8 -o src/f32-igemm/gen/4x8s4-minmax-sse.c &
297tools/xngen src/f32-igemm/sse-shuffle.c.in -D MR=5 -D NR=8 -o src/f32-igemm/gen/5x8s4-minmax-sse.c &
298
299### MRx2 micro-kernels
300tools/xngen src/f32-igemm/MRx2c4-sse.c.in -D MR=4 -D NR=2 -o src/f32-igemm/gen/4x2c4-minmax-sse.c &
301
302################################### x86 AVX ###################################
303### AVX+BROADCAST micro-kernels
304tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=1 -D NR=8 -D FMA=0 -o src/f32-igemm/gen/1x8-minmax-avx-broadcast.c &
305tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=4 -D NR=8 -D FMA=0 -o src/f32-igemm/gen/4x8-minmax-avx-broadcast.c &
306tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=5 -D NR=8 -D FMA=0 -o src/f32-igemm/gen/5x8-minmax-avx-broadcast.c &
307tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=6 -D NR=8 -D FMA=0 -o src/f32-igemm/gen/6x8-minmax-avx-broadcast.c &
308tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=7 -D NR=8 -D FMA=0 -o src/f32-igemm/gen/7x8-minmax-avx-broadcast.c &
309tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=1 -D NR=16 -D FMA=0 -o src/f32-igemm/gen/1x16-minmax-avx-broadcast.c &
310tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=3 -D NR=16 -D FMA=0 -o src/f32-igemm/gen/3x16-minmax-avx-broadcast.c &
311tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=4 -D NR=16 -D FMA=0 -o src/f32-igemm/gen/4x16-minmax-avx-broadcast.c &
312tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=5 -D NR=16 -D FMA=0 -o src/f32-igemm/gen/5x16-minmax-avx-broadcast.c &
313### FMA3+BROADCAST micro-kernels
314tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=1 -D NR=8 -D FMA=3 -o src/f32-igemm/gen/1x8-minmax-fma3-broadcast.c &
315tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=4 -D NR=8 -D FMA=3 -o src/f32-igemm/gen/4x8-minmax-fma3-broadcast.c &
316tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=5 -D NR=8 -D FMA=3 -o src/f32-igemm/gen/5x8-minmax-fma3-broadcast.c &
317tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=6 -D NR=8 -D FMA=3 -o src/f32-igemm/gen/6x8-minmax-fma3-broadcast.c &
318tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=7 -D NR=8 -D FMA=3 -o src/f32-igemm/gen/7x8-minmax-fma3-broadcast.c &
319tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=8 -D NR=8 -D FMA=3 -o src/f32-igemm/gen/8x8-minmax-fma3-broadcast.c &
320tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=1 -D NR=16 -D FMA=3 -o src/f32-igemm/gen/1x16-minmax-fma3-broadcast.c &
321tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=3 -D NR=16 -D FMA=3 -o src/f32-igemm/gen/3x16-minmax-fma3-broadcast.c &
322tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=4 -D NR=16 -D FMA=3 -o src/f32-igemm/gen/4x16-minmax-fma3-broadcast.c &
323tools/xngen src/f32-igemm/avx-broadcast.c.in -D MR=5 -D NR=16 -D FMA=3 -o src/f32-igemm/gen/5x16-minmax-fma3-broadcast.c &
324
325tools/xngen src/f32-igemm/avx-shuffle4.c.in -D MR=1 -D NR=16 -D FMA=3 -o src/f32-igemm/gen/1x16s4-minmax-fma3-broadcast.c &
326tools/xngen src/f32-igemm/avx-shuffle4.c.in -D MR=3 -D NR=16 -D FMA=3 -o src/f32-igemm/gen/3x16s4-minmax-fma3-broadcast.c &
327tools/xngen src/f32-igemm/avx-shuffle4.c.in -D MR=4 -D NR=16 -D FMA=3 -o src/f32-igemm/gen/4x16s4-minmax-fma3-broadcast.c &
328tools/xngen src/f32-igemm/avx-shuffle4.c.in -D MR=5 -D NR=16 -D FMA=3 -o src/f32-igemm/gen/5x16s4-minmax-fma3-broadcast.c &
329
330################################# x86 AVX-512 #################################
331### AVX512F+BROADCAST micro-kernels
332tools/xngen src/f32-igemm/avx512-broadcast.c.in -D MR=1 -D NR=16 -o src/f32-igemm/gen/1x16-minmax-avx512f-broadcast.c &
333tools/xngen src/f32-igemm/avx512-broadcast.c.in -D MR=4 -D NR=16 -o src/f32-igemm/gen/4x16-minmax-avx512f-broadcast.c &
334tools/xngen src/f32-igemm/avx512-broadcast.c.in -D MR=5 -D NR=16 -o src/f32-igemm/gen/5x16-minmax-avx512f-broadcast.c &
335tools/xngen src/f32-igemm/avx512-broadcast.c.in -D MR=6 -D NR=16 -o src/f32-igemm/gen/6x16-minmax-avx512f-broadcast.c &
336tools/xngen src/f32-igemm/avx512-broadcast.c.in -D MR=7 -D NR=16 -o src/f32-igemm/gen/7x16-minmax-avx512f-broadcast.c &
337tools/xngen src/f32-igemm/avx512-broadcast.c.in -D MR=8 -D NR=16 -o src/f32-igemm/gen/8x16-minmax-avx512f-broadcast.c &
338
339################################## Unit tests #################################
340tools/generate-gemm-test.py --spec test/f32-igemm.yaml --output test/f32-igemm.cc --output test/f32-igemm-2.cc &
341tools/generate-gemm-test.py --spec test/f32-igemm-relu.yaml --output test/f32-igemm-relu.cc --output test/f32-igemm-relu-2.cc &
342tools/generate-gemm-test.py --spec test/f32-igemm-minmax.yaml --output test/f32-igemm-minmax.cc --output test/f32-igemm-minmax-2.cc &
343
344wait
345