1*4bdc9457SAndroid Build Coastguard Worker#!/bin/sh 2*4bdc9457SAndroid Build Coastguard Worker# Copyright 2019 Google LLC 3*4bdc9457SAndroid Build Coastguard Worker# 4*4bdc9457SAndroid Build Coastguard Worker# This source code is licensed under the BSD-style license found in the 5*4bdc9457SAndroid Build Coastguard Worker# LICENSE file in the root directory of this source tree. 6*4bdc9457SAndroid Build Coastguard Worker 7*4bdc9457SAndroid Build Coastguard Worker############################### AArch64 assembly ############################## 8*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/1x16-aarch64-neonfp16arith-ld32.S.in -D INC=0 -o src/f16-gemm/gen/1x16-minmax-aarch64-neonfp16arith-ld32.S & 9*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/4x16-aarch64-neonfp16arith-ld32.S.in -D INC=0 -o src/f16-gemm/gen/4x16-minmax-aarch64-neonfp16arith-ld32.S & 10*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/4x16-aarch64-neonfp16arith-ld64.S.in -D INC=0 -o src/f16-gemm/gen/4x16-minmax-aarch64-neonfp16arith-ld64.S & 11*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/6x16-aarch64-neonfp16arith-ld32.S.in -D INC=0 -o src/f16-gemm/gen/6x16-minmax-aarch64-neonfp16arith-ld32.S & 12*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/6x16-aarch64-neonfp16arith-ld64.S.in -D INC=0 -o src/f16-gemm/gen/6x16-minmax-aarch64-neonfp16arith-ld64.S & 13*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/1x16-aarch64-neonfp16arith-ld32.S.in -D INC=1 -o src/f16-gemm/gen-inc/1x16inc-minmax-aarch64-neonfp16arith-ld32.S & 14*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/4x16-aarch64-neonfp16arith-ld32.S.in -D INC=1 -o src/f16-gemm/gen-inc/4x16inc-minmax-aarch64-neonfp16arith-ld32.S & 15*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/6x16-aarch64-neonfp16arith-ld32.S.in -D INC=1 -o src/f16-gemm/gen-inc/6x16inc-minmax-aarch64-neonfp16arith-ld32.S & 16*4bdc9457SAndroid Build Coastguard Worker 17*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/1x8-aarch64-neonfp16arith-ld64.S.in -D INC=0 -o src/f16-gemm/gen/1x8-minmax-aarch64-neonfp16arith-ld64.S & 18*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/4x8-aarch64-neonfp16arith-ld64.S.in -D INC=0 -o src/f16-gemm/gen/4x8-minmax-aarch64-neonfp16arith-ld64.S & 19*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/6x8-aarch64-neonfp16arith-ld64.S.in -D INC=0 -o src/f16-gemm/gen/6x8-minmax-aarch64-neonfp16arith-ld64.S & 20*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/8x8-aarch64-neonfp16arith-ld64.S.in -D INC=0 -o src/f16-gemm/gen/8x8-minmax-aarch64-neonfp16arith-ld64.S & 21*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/1x8-aarch64-neonfp16arith-ld64.S.in -D INC=1 -o src/f16-gemm/gen-inc/1x8inc-minmax-aarch64-neonfp16arith-ld64.S & 22*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/4x8-aarch64-neonfp16arith-ld64.S.in -D INC=1 -o src/f16-gemm/gen-inc/4x8inc-minmax-aarch64-neonfp16arith-ld64.S & 23*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/6x8-aarch64-neonfp16arith-ld64.S.in -D INC=1 -o src/f16-gemm/gen-inc/6x8inc-minmax-aarch64-neonfp16arith-ld64.S & 24*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/8x8-aarch64-neonfp16arith-ld64.S.in -D INC=1 -o src/f16-gemm/gen-inc/8x8inc-minmax-aarch64-neonfp16arith-ld64.S & 25*4bdc9457SAndroid Build Coastguard Worker 26*4bdc9457SAndroid Build Coastguard Worker### Cortex A55r0 micro-kernel 27*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/6x16-aarch64-neonfp16arith-cortex-a55r0.S.in -D INC=0 -o src/f16-gemm/gen/6x16-minmax-aarch64-neonfp16arith-cortex-a55r0.S & 28*4bdc9457SAndroid Build Coastguard Worker 29*4bdc9457SAndroid Build Coastguard Worker### Cortex A55 micro-kernels 30*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/6x16-aarch64-neonfp16arith-cortex-a55.S.in -D INC=0 -o src/f16-gemm/gen/6x16-minmax-aarch64-neonfp16arith-cortex-a55.S & 31*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/6x16-aarch64-neonfp16arith-cortex-a55.S.in -D INC=1 -o src/f16-gemm/gen-inc/6x16inc-minmax-aarch64-neonfp16arith-cortex-a55.S & 32*4bdc9457SAndroid Build Coastguard Worker 33*4bdc9457SAndroid Build Coastguard Worker### Cortex A75 micro-kernels 34*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/6x16-aarch64-neonfp16arith-cortex-a75.S.in -D INC=0 -o src/f16-gemm/gen/6x16-minmax-aarch64-neonfp16arith-cortex-a75.S & 35*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/6x16-aarch64-neonfp16arith-cortex-a75.S.in -D INC=1 -o src/f16-gemm/gen-inc/6x16inc-minmax-aarch64-neonfp16arith-cortex-a75.S & 36*4bdc9457SAndroid Build Coastguard Worker 37*4bdc9457SAndroid Build Coastguard Worker########################## ARM NEON with FP16 compute ######################### 38*4bdc9457SAndroid Build Coastguard Worker### LD64 micro-kernels 39*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/neonfp16arith-ld64.c.in -D MR=1 -D NR=8 -D INC=0 -o src/f16-gemm/gen/1x8-minmax-neonfp16arith-ld64.c & 40*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/neonfp16arith-ld64.c.in -D MR=4 -D NR=8 -D INC=0 -o src/f16-gemm/gen/4x8-minmax-neonfp16arith-ld64.c & 41*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/neonfp16arith-ld64.c.in -D MR=6 -D NR=8 -D INC=0 -o src/f16-gemm/gen/6x8-minmax-neonfp16arith-ld64.c & 42*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/neonfp16arith-ld64.c.in -D MR=8 -D NR=8 -D INC=0 -o src/f16-gemm/gen/8x8-minmax-neonfp16arith-ld64.c & 43*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/neonfp16arith-ld64.c.in -D MR=1 -D NR=8 -D INC=1 -o src/f16-gemm/gen-inc/1x8inc-minmax-neonfp16arith-ld64.c & 44*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/neonfp16arith-ld64.c.in -D MR=4 -D NR=8 -D INC=1 -o src/f16-gemm/gen-inc/4x8inc-minmax-neonfp16arith-ld64.c & 45*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/neonfp16arith-ld64.c.in -D MR=6 -D NR=8 -D INC=1 -o src/f16-gemm/gen-inc/6x8inc-minmax-neonfp16arith-ld64.c & 46*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/neonfp16arith-ld64.c.in -D MR=8 -D NR=8 -D INC=1 -o src/f16-gemm/gen-inc/8x8inc-minmax-neonfp16arith-ld64.c & 47*4bdc9457SAndroid Build Coastguard Worker 48*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/neonfp16arith-ld64.c.in -D MR=1 -D NR=16 -D INC=0 -o src/f16-gemm/gen/1x16-minmax-neonfp16arith-ld64.c & 49*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/neonfp16arith-ld64.c.in -D MR=4 -D NR=16 -D INC=0 -o src/f16-gemm/gen/4x16-minmax-neonfp16arith-ld64.c & 50*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/neonfp16arith-ld64.c.in -D MR=6 -D NR=16 -D INC=0 -o src/f16-gemm/gen/6x16-minmax-neonfp16arith-ld64.c & 51*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/neonfp16arith-ld64.c.in -D MR=8 -D NR=16 -D INC=0 -o src/f16-gemm/gen/8x16-minmax-neonfp16arith-ld64.c & 52*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/neonfp16arith-ld64.c.in -D MR=1 -D NR=16 -D INC=1 -o src/f16-gemm/gen-inc/1x16inc-minmax-neonfp16arith-ld64.c & 53*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/neonfp16arith-ld64.c.in -D MR=4 -D NR=16 -D INC=1 -o src/f16-gemm/gen-inc/4x16inc-minmax-neonfp16arith-ld64.c & 54*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/neonfp16arith-ld64.c.in -D MR=6 -D NR=16 -D INC=1 -o src/f16-gemm/gen-inc/6x16inc-minmax-neonfp16arith-ld64.c & 55*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/neonfp16arith-ld64.c.in -D MR=8 -D NR=16 -D INC=1 -o src/f16-gemm/gen-inc/8x16inc-minmax-neonfp16arith-ld64.c & 56*4bdc9457SAndroid Build Coastguard Worker 57*4bdc9457SAndroid Build Coastguard Worker################################### x86 AVX2 ################################### 58*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/avx2-broadcast.c.in -D MR=1 -D NR=8 -o src/f16-gemm/gen/1x8-minmax-avx2-broadcast.c & 59*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/avx2-broadcast.c.in -D MR=4 -D NR=8 -o src/f16-gemm/gen/4x8-minmax-avx2-broadcast.c & 60*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/avx2-broadcast.c.in -D MR=5 -D NR=8 -o src/f16-gemm/gen/5x8-minmax-avx2-broadcast.c & 61*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/avx2-broadcast.c.in -D MR=6 -D NR=8 -o src/f16-gemm/gen/6x8-minmax-avx2-broadcast.c & 62*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/avx2-broadcast.c.in -D MR=7 -D NR=8 -o src/f16-gemm/gen/7x8-minmax-avx2-broadcast.c & 63*4bdc9457SAndroid Build Coastguard Worker 64*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/avx2-broadcast.c.in -D MR=1 -D NR=16 -o src/f16-gemm/gen/1x16-minmax-avx2-broadcast.c & 65*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/avx2-broadcast.c.in -D MR=3 -D NR=16 -o src/f16-gemm/gen/3x16-minmax-avx2-broadcast.c & 66*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/avx2-broadcast.c.in -D MR=4 -D NR=16 -o src/f16-gemm/gen/4x16-minmax-avx2-broadcast.c & 67*4bdc9457SAndroid Build Coastguard Workertools/xngen src/f16-gemm/avx2-broadcast.c.in -D MR=5 -D NR=16 -o src/f16-gemm/gen/5x16-minmax-avx2-broadcast.c & 68*4bdc9457SAndroid Build Coastguard Worker 69*4bdc9457SAndroid Build Coastguard Worker################################## Unit tests ################################# 70*4bdc9457SAndroid Build Coastguard Workertools/generate-gemm-test.py --spec test/f16-gemm-minmax.yaml --output test/f16-gemm-minmax.cc & 71*4bdc9457SAndroid Build Coastguard Worker 72*4bdc9457SAndroid Build Coastguard Workerwait 73