blob: a83aed3b994a405ef6ca9a20bc5497eab69be7a6 [file] [log] [blame]
XNNPACK Teamb455b122019-09-27 18:10:33 -07001#!/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### Microkernels without unrolling
Marat Dukhan355ab432020-04-09 19:01:52 -07009tools/xngen src/f32-spmm/scalar.c.in -D MR=1 -D NR=1 -D UNROLL=1 -o src/f32-spmm/gen/1x1-minmax-scalar.c
10tools/xngen src/f32-spmm/scalar.c.in -D MR=2 -D NR=1 -D UNROLL=1 -o src/f32-spmm/gen/2x1-minmax-scalar.c
11tools/xngen src/f32-spmm/scalar.c.in -D MR=4 -D NR=1 -D UNROLL=1 -o src/f32-spmm/gen/4x1-minmax-scalar.c
12tools/xngen src/f32-spmm/scalar.c.in -D MR=8 -D NR=1 -D UNROLL=1 -o src/f32-spmm/gen/8x1-minmax-scalar.c
13tools/xngen src/f32-spmm/scalar.c.in -D MR=8 -D NR=2 -D UNROLL=1 -o src/f32-spmm/gen/8x2-minmax-scalar.c
14tools/xngen src/f32-spmm/scalar.c.in -D MR=8 -D NR=4 -D UNROLL=1 -o src/f32-spmm/gen/8x4-minmax-scalar.c
XNNPACK Teamb455b122019-09-27 18:10:33 -070015### Microkernels with software pipelining
Marat Dukhan355ab432020-04-09 19:01:52 -070016tools/xngen src/f32-spmm/scalar-pipelined.c.in -D MR=1 -D NR=1 -o src/f32-spmm/gen/1x1-minmax-scalar-pipelined.c
17tools/xngen src/f32-spmm/scalar-pipelined.c.in -D MR=2 -D NR=1 -o src/f32-spmm/gen/2x1-minmax-scalar-pipelined.c
18tools/xngen src/f32-spmm/scalar-pipelined.c.in -D MR=4 -D NR=1 -o src/f32-spmm/gen/4x1-minmax-scalar-pipelined.c
19tools/xngen src/f32-spmm/scalar-pipelined.c.in -D MR=8 -D NR=1 -o src/f32-spmm/gen/8x1-minmax-scalar-pipelined.c
XNNPACK Teamb455b122019-09-27 18:10:33 -070020
21################################### ARM NEON ##################################
22### Microkernels without unrolling
Marat Dukhan5b86c432020-12-06 19:15:03 -080023tools/xngen src/f32-spmm/neon.c.in -D MR=4 -D NR=1 -D UNROLL=1 -D FMA=0 -o src/f32-spmm/gen/4x1-minmax-neon.c
24tools/xngen src/f32-spmm/neon.c.in -D MR=8 -D NR=1 -D UNROLL=1 -D FMA=0 -o src/f32-spmm/gen/8x1-minmax-neon.c
25tools/xngen src/f32-spmm/neon.c.in -D MR=12 -D NR=1 -D UNROLL=1 -D FMA=0 -o src/f32-spmm/gen/12x1-minmax-neon.c
26tools/xngen src/f32-spmm/neon.c.in -D MR=16 -D NR=1 -D UNROLL=1 -D FMA=0 -o src/f32-spmm/gen/16x1-minmax-neon.c
27tools/xngen src/f32-spmm/neon.c.in -D MR=32 -D NR=1 -D UNROLL=1 -D FMA=0 -o src/f32-spmm/gen/32x1-minmax-neon.c
28
Marat Dukhan355ab432020-04-09 19:01:52 -070029tools/xngen src/f32-spmm/neon.c.in -D MR=4 -D NR=1 -D UNROLL=1 -D FMA=1 -o src/f32-spmm/gen/4x1-minmax-neonfma.c
30tools/xngen src/f32-spmm/neon.c.in -D MR=8 -D NR=1 -D UNROLL=1 -D FMA=1 -o src/f32-spmm/gen/8x1-minmax-neonfma.c
31tools/xngen src/f32-spmm/neon.c.in -D MR=12 -D NR=1 -D UNROLL=1 -D FMA=1 -o src/f32-spmm/gen/12x1-minmax-neonfma.c
32tools/xngen src/f32-spmm/neon.c.in -D MR=16 -D NR=1 -D UNROLL=1 -D FMA=1 -o src/f32-spmm/gen/16x1-minmax-neonfma.c
Frank Barchard846c0c62020-10-26 15:01:39 -070033tools/xngen src/f32-spmm/neon.c.in -D MR=32 -D NR=1 -D UNROLL=1 -D FMA=1 -o src/f32-spmm/gen/32x1-minmax-neonfma.c
Marat Dukhan5b86c432020-12-06 19:15:03 -080034
XNNPACK Teamb455b122019-09-27 18:10:33 -070035### Microkernels with 2X unrolling
Marat Dukhan5b86c432020-12-06 19:15:03 -080036tools/xngen src/f32-spmm/neon.c.in -D MR=4 -D NR=1 -D UNROLL=2 -D FMA=0 -o src/f32-spmm/gen/4x1-minmax-neon-x2.c
37tools/xngen src/f32-spmm/neon.c.in -D MR=8 -D NR=1 -D UNROLL=2 -D FMA=0 -o src/f32-spmm/gen/8x1-minmax-neon-x2.c
38tools/xngen src/f32-spmm/neon.c.in -D MR=16 -D NR=1 -D UNROLL=2 -D FMA=0 -o src/f32-spmm/gen/16x1-minmax-neon-x2.c
39tools/xngen src/f32-spmm/neon.c.in -D MR=32 -D NR=1 -D UNROLL=2 -D FMA=0 -o src/f32-spmm/gen/32x1-minmax-neon-x2.c
40
Frank Barchardbeca6522020-10-30 22:34:35 -070041tools/xngen src/f32-spmm/neon.c.in -D MR=4 -D NR=1 -D UNROLL=2 -D FMA=1 -o src/f32-spmm/gen/4x1-minmax-neonfma-x2.c
42tools/xngen src/f32-spmm/neon.c.in -D MR=8 -D NR=1 -D UNROLL=2 -D FMA=1 -o src/f32-spmm/gen/8x1-minmax-neonfma-x2.c
43tools/xngen src/f32-spmm/neon.c.in -D MR=16 -D NR=1 -D UNROLL=2 -D FMA=1 -o src/f32-spmm/gen/16x1-minmax-neonfma-x2.c
44tools/xngen src/f32-spmm/neon.c.in -D MR=32 -D NR=1 -D UNROLL=2 -D FMA=1 -o src/f32-spmm/gen/32x1-minmax-neonfma-x2.c
Marat Dukhan5b86c432020-12-06 19:15:03 -080045
46### Microkernels for blocks of output channels
Marat Dukhan355ab432020-04-09 19:01:52 -070047tools/xngen src/f32-spmm/neon-blocked.c.in -D MR=4 -D NR=2 -D UNROLL=1 -D FMA=1 -o src/f32-spmm/gen/4x2-minmax-neonfma.c
48tools/xngen src/f32-spmm/neon-blocked.c.in -D MR=8 -D NR=2 -D UNROLL=1 -D FMA=1 -o src/f32-spmm/gen/8x2-minmax-neonfma.c
49tools/xngen src/f32-spmm/neon-blocked.c.in -D MR=12 -D NR=2 -D UNROLL=1 -D FMA=1 -o src/f32-spmm/gen/12x2-minmax-neonfma.c
50tools/xngen src/f32-spmm/neon-blocked.c.in -D MR=16 -D NR=2 -D UNROLL=1 -D FMA=1 -o src/f32-spmm/gen/16x2-minmax-neonfma.c
Frank Barchard846c0c62020-10-26 15:01:39 -070051tools/xngen src/f32-spmm/neon-blocked.c.in -D MR=32 -D NR=2 -D UNROLL=1 -D FMA=1 -o src/f32-spmm/gen/32x2-minmax-neonfma.c
Marat Dukhan355ab432020-04-09 19:01:52 -070052tools/xngen src/f32-spmm/neon-blocked.c.in -D MR=4 -D NR=4 -D UNROLL=1 -D FMA=1 -o src/f32-spmm/gen/4x4-minmax-neonfma.c
53tools/xngen src/f32-spmm/neon-blocked.c.in -D MR=8 -D NR=4 -D UNROLL=1 -D FMA=1 -o src/f32-spmm/gen/8x4-minmax-neonfma.c
54tools/xngen src/f32-spmm/neon-blocked.c.in -D MR=12 -D NR=4 -D UNROLL=1 -D FMA=1 -o src/f32-spmm/gen/12x4-minmax-neonfma.c
55tools/xngen src/f32-spmm/neon-blocked.c.in -D MR=16 -D NR=4 -D UNROLL=1 -D FMA=1 -o src/f32-spmm/gen/16x4-minmax-neonfma.c
Frank Barchard846c0c62020-10-26 15:01:39 -070056tools/xngen src/f32-spmm/neon-blocked.c.in -D MR=32 -D NR=4 -D UNROLL=1 -D FMA=1 -o src/f32-spmm/gen/32x4-minmax-neonfma.c
Marat Dukhan5b86c432020-12-06 19:15:03 -080057
XNNPACK Teamb455b122019-09-27 18:10:33 -070058### Microkernels with software pipelining
Marat Dukhan5b86c432020-12-06 19:15:03 -080059tools/xngen src/f32-spmm/neon-pipelined.c.in -D MR=4 -D NR=1 -D FMA=0 -o src/f32-spmm/gen/4x1-minmax-neon-pipelined.c
60tools/xngen src/f32-spmm/neon-pipelined.c.in -D MR=8 -D NR=1 -D FMA=0 -o src/f32-spmm/gen/8x1-minmax-neon-pipelined.c
61tools/xngen src/f32-spmm/neon-pipelined.c.in -D MR=16 -D NR=1 -D FMA=0 -o src/f32-spmm/gen/16x1-minmax-neon-pipelined.c
62tools/xngen src/f32-spmm/neon-pipelined.c.in -D MR=32 -D NR=1 -D FMA=0 -o src/f32-spmm/gen/32x1-minmax-neon-pipelined.c
63
Marat Dukhan355ab432020-04-09 19:01:52 -070064tools/xngen src/f32-spmm/neon-pipelined.c.in -D MR=4 -D NR=1 -D FMA=1 -o src/f32-spmm/gen/4x1-minmax-neonfma-pipelined.c
65tools/xngen src/f32-spmm/neon-pipelined.c.in -D MR=8 -D NR=1 -D FMA=1 -o src/f32-spmm/gen/8x1-minmax-neonfma-pipelined.c
66tools/xngen src/f32-spmm/neon-pipelined.c.in -D MR=16 -D NR=1 -D FMA=1 -o src/f32-spmm/gen/16x1-minmax-neonfma-pipelined.c
Frank Barchard846c0c62020-10-26 15:01:39 -070067tools/xngen src/f32-spmm/neon-pipelined.c.in -D MR=32 -D NR=1 -D FMA=1 -o src/f32-spmm/gen/32x1-minmax-neonfma-pipelined.c
XNNPACK Teamb455b122019-09-27 18:10:33 -070068
69################################### x86 SSE ###################################
70### Microkernels without unrolling
Marat Dukhan355ab432020-04-09 19:01:52 -070071tools/xngen src/f32-spmm/sse.c.in -D MR=4 -D NR=1 -D UNROLL=1 -o src/f32-spmm/gen/4x1-minmax-sse.c
Frank Barchard846c0c62020-10-26 15:01:39 -070072tools/xngen src/f32-spmm/sse.c.in -D MR=8 -D NR=1 -D UNROLL=1 -o src/f32-spmm/gen/8x1-minmax-sse.c
Erich Elsen6e80fdc2020-06-09 15:35:37 -070073tools/xngen src/f32-spmm/sse.c.in -D MR=16 -D NR=1 -D UNROLL=1 -o src/f32-spmm/gen/16x1-minmax-sse.c
Frank Barchard846c0c62020-10-26 15:01:39 -070074tools/xngen src/f32-spmm/sse.c.in -D MR=32 -D NR=1 -D UNROLL=1 -o src/f32-spmm/gen/32x1-minmax-sse.c
Erich Elsen6e80fdc2020-06-09 15:35:37 -070075
Frank Barchardc451e8a2020-10-21 17:13:12 -070076################################### WASM SIMD ###################################
Frank Barchard9e053402020-10-19 15:29:08 -070077### Microkernels without unrolling.
78tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=4 -D NR=1 -D X86=0 -D UNROLL=1 -o src/f32-spmm/gen/4x1-minmax-wasmsimd-arm.c
79tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=8 -D NR=1 -D X86=0 -D UNROLL=1 -o src/f32-spmm/gen/8x1-minmax-wasmsimd-arm.c
80tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=16 -D NR=1 -D X86=0 -D UNROLL=1 -o src/f32-spmm/gen/16x1-minmax-wasmsimd-arm.c
Frank Barchard846c0c62020-10-26 15:01:39 -070081tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=32 -D NR=1 -D X86=0 -D UNROLL=1 -o src/f32-spmm/gen/32x1-minmax-wasmsimd-arm.c
Frank Barchard9e053402020-10-19 15:29:08 -070082tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=4 -D NR=1 -D X86=1 -D UNROLL=1 -o src/f32-spmm/gen/4x1-minmax-wasmsimd-x86.c
83tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=8 -D NR=1 -D X86=1 -D UNROLL=1 -o src/f32-spmm/gen/8x1-minmax-wasmsimd-x86.c
84tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=16 -D NR=1 -D X86=1 -D UNROLL=1 -o src/f32-spmm/gen/16x1-minmax-wasmsimd-x86.c
Frank Barchard846c0c62020-10-26 15:01:39 -070085tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=32 -D NR=1 -D X86=1 -D UNROLL=1 -o src/f32-spmm/gen/32x1-minmax-wasmsimd-x86.c
XNNPACK Teamb455b122019-09-27 18:10:33 -070086
Frank Barchardc451e8a2020-10-21 17:13:12 -070087### Microkernels with 2X unrolling
Frank Barchardbeca6522020-10-30 22:34:35 -070088tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=4 -D NR=1 -D X86=0 -D UNROLL=2 -o src/f32-spmm/gen/4x1-minmax-wasmsimd-arm-x2.c
89tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=8 -D NR=1 -D X86=0 -D UNROLL=2 -o src/f32-spmm/gen/8x1-minmax-wasmsimd-arm-x2.c
90tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=16 -D NR=1 -D X86=0 -D UNROLL=2 -o src/f32-spmm/gen/16x1-minmax-wasmsimd-arm-x2.c
91tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=32 -D NR=1 -D X86=0 -D UNROLL=2 -o src/f32-spmm/gen/32x1-minmax-wasmsimd-arm-x2.c
92tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=4 -D NR=1 -D X86=1 -D UNROLL=2 -o src/f32-spmm/gen/4x1-minmax-wasmsimd-x86-x2.c
93tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=8 -D NR=1 -D X86=1 -D UNROLL=2 -o src/f32-spmm/gen/8x1-minmax-wasmsimd-x86-x2.c
94tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=16 -D NR=1 -D X86=1 -D UNROLL=2 -o src/f32-spmm/gen/16x1-minmax-wasmsimd-x86-x2.c
95tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=32 -D NR=1 -D X86=1 -D UNROLL=2 -o src/f32-spmm/gen/32x1-minmax-wasmsimd-x86-x2.c
Frank Barchardc451e8a2020-10-21 17:13:12 -070096
97### Microkernels with 4X unrolling
Frank Barchardbeca6522020-10-30 22:34:35 -070098tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=4 -D NR=1 -D X86=0 -D UNROLL=4 -o src/f32-spmm/gen/4x1-minmax-wasmsimd-arm-x4.c
99tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=8 -D NR=1 -D X86=0 -D UNROLL=4 -o src/f32-spmm/gen/8x1-minmax-wasmsimd-arm-x4.c
100tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=16 -D NR=1 -D X86=0 -D UNROLL=4 -o src/f32-spmm/gen/16x1-minmax-wasmsimd-arm-x4.c
101tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=32 -D NR=1 -D X86=0 -D UNROLL=4 -o src/f32-spmm/gen/32x1-minmax-wasmsimd-arm-x4.c
102tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=4 -D NR=1 -D X86=1 -D UNROLL=4 -o src/f32-spmm/gen/4x1-minmax-wasmsimd-x86-x4.c
103tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=8 -D NR=1 -D X86=1 -D UNROLL=4 -o src/f32-spmm/gen/8x1-minmax-wasmsimd-x86-x4.c
104tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=16 -D NR=1 -D X86=1 -D UNROLL=4 -o src/f32-spmm/gen/16x1-minmax-wasmsimd-x86-x4.c
105tools/xngen src/f32-spmm/wasmsimd.c.in -D MR=32 -D NR=1 -D X86=1 -D UNROLL=4 -o src/f32-spmm/gen/32x1-minmax-wasmsimd-x86-x4.c
Frank Barchardc451e8a2020-10-21 17:13:12 -0700106
Frank Barchard8ef44cd2020-11-03 12:30:23 -0800107### Microkernels with software pipelining
108tools/xngen src/f32-spmm/wasmsimd-pipelined.c.in -D MR=4 -D NR=1 -D X86=0 -D UNROLL=1 -o src/f32-spmm/gen/4x1-minmax-wasmsimd-arm-pipelined.c
109tools/xngen src/f32-spmm/wasmsimd-pipelined.c.in -D MR=8 -D NR=1 -D X86=0 -D UNROLL=1 -o src/f32-spmm/gen/8x1-minmax-wasmsimd-arm-pipelined.c
110tools/xngen src/f32-spmm/wasmsimd-pipelined.c.in -D MR=16 -D NR=1 -D X86=0 -D UNROLL=1 -o src/f32-spmm/gen/16x1-minmax-wasmsimd-arm-pipelined.c
111tools/xngen src/f32-spmm/wasmsimd-pipelined.c.in -D MR=32 -D NR=1 -D X86=0 -D UNROLL=1 -o src/f32-spmm/gen/32x1-minmax-wasmsimd-arm-pipelined.c
112tools/xngen src/f32-spmm/wasmsimd-pipelined.c.in -D MR=4 -D NR=1 -D X86=1 -D UNROLL=1 -o src/f32-spmm/gen/4x1-minmax-wasmsimd-x86-pipelined.c
113tools/xngen src/f32-spmm/wasmsimd-pipelined.c.in -D MR=8 -D NR=1 -D X86=1 -D UNROLL=1 -o src/f32-spmm/gen/8x1-minmax-wasmsimd-x86-pipelined.c
114tools/xngen src/f32-spmm/wasmsimd-pipelined.c.in -D MR=16 -D NR=1 -D X86=1 -D UNROLL=1 -o src/f32-spmm/gen/16x1-minmax-wasmsimd-x86-pipelined.c
115tools/xngen src/f32-spmm/wasmsimd-pipelined.c.in -D MR=32 -D NR=1 -D X86=1 -D UNROLL=1 -o src/f32-spmm/gen/32x1-minmax-wasmsimd-x86-pipelined.c
116
117### Microkernels with software pipelining and 2X unrolling
118tools/xngen src/f32-spmm/wasmsimd-pipelined.c.in -D MR=4 -D NR=1 -D X86=0 -D UNROLL=2 -o src/f32-spmm/gen/4x1-minmax-wasmsimd-arm-pipelined-x2.c
119tools/xngen src/f32-spmm/wasmsimd-pipelined.c.in -D MR=8 -D NR=1 -D X86=0 -D UNROLL=2 -o src/f32-spmm/gen/8x1-minmax-wasmsimd-arm-pipelined-x2.c
120tools/xngen src/f32-spmm/wasmsimd-pipelined.c.in -D MR=16 -D NR=1 -D X86=0 -D UNROLL=2 -o src/f32-spmm/gen/16x1-minmax-wasmsimd-arm-pipelined-x2.c
121tools/xngen src/f32-spmm/wasmsimd-pipelined.c.in -D MR=32 -D NR=1 -D X86=0 -D UNROLL=2 -o src/f32-spmm/gen/32x1-minmax-wasmsimd-arm-pipelined-x2.c
122tools/xngen src/f32-spmm/wasmsimd-pipelined.c.in -D MR=4 -D NR=1 -D X86=1 -D UNROLL=2 -o src/f32-spmm/gen/4x1-minmax-wasmsimd-x86-pipelined-x2.c
123tools/xngen src/f32-spmm/wasmsimd-pipelined.c.in -D MR=8 -D NR=1 -D X86=1 -D UNROLL=2 -o src/f32-spmm/gen/8x1-minmax-wasmsimd-x86-pipelined-x2.c
124tools/xngen src/f32-spmm/wasmsimd-pipelined.c.in -D MR=16 -D NR=1 -D X86=1 -D UNROLL=2 -o src/f32-spmm/gen/16x1-minmax-wasmsimd-x86-pipelined-x2.c
125tools/xngen src/f32-spmm/wasmsimd-pipelined.c.in -D MR=32 -D NR=1 -D X86=1 -D UNROLL=2 -o src/f32-spmm/gen/32x1-minmax-wasmsimd-x86-pipelined-x2.c
126
XNNPACK Teamb455b122019-09-27 18:10:33 -0700127################################## Unit tests #################################
Marat Dukhan355ab432020-04-09 19:01:52 -0700128tools/generate-spmm-test.py --spec test/f32-spmm-minmax.yaml --output test/f32-spmm-minmax.cc