From 03b2971ac69a86f10a1566938d1a25afee15746c Mon Sep 17 00:00:00 2001 From: Viet-Hoa Do Date: Wed, 1 Jun 2022 11:47:14 +0100 Subject: Integrate SME2 kernels * Add SME/SME2 detection. * Integrate SME2 implementation for: - Normal convolution - Winograd - Depthwise convolution - Pooling Resolves: COMPMID-5700 Signed-off-by: Viet-Hoa Do Change-Id: I2f1ca1d05f8cfeee9309ed1c0a36096a4a6aad5c Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8692 Reviewed-by: Gunes Bayir Tested-by: Arm Jenkins Comments-Addressed: Arm Jenkins --- .../sme_transpose_interleave_4VL_2x2_fp32bf16.hpp | 189 +++++++++++++++++++++ 1 file changed, 189 insertions(+) create mode 100644 src/core/NEON/kernels/arm_gemm/transforms/sme_transpose_interleave_4VL_2x2_fp32bf16.hpp (limited to 'src/core/NEON/kernels/arm_gemm/transforms/sme_transpose_interleave_4VL_2x2_fp32bf16.hpp') diff --git a/src/core/NEON/kernels/arm_gemm/transforms/sme_transpose_interleave_4VL_2x2_fp32bf16.hpp b/src/core/NEON/kernels/arm_gemm/transforms/sme_transpose_interleave_4VL_2x2_fp32bf16.hpp new file mode 100644 index 0000000000..59ac4bf50b --- /dev/null +++ b/src/core/NEON/kernels/arm_gemm/transforms/sme_transpose_interleave_4VL_2x2_fp32bf16.hpp @@ -0,0 +1,189 @@ +/* + * Copyright (c) 2022 Arm Limited. + * + * SPDX-License-Identifier: MIT + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to + * deal in the Software without restriction, including without limitation the + * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or + * sell copies of the Software, and to permit persons to whom the Software is + * furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + */ + +#pragma once + +#ifdef __ARM_FEATURE_SVE + + +namespace { + +void sme_transpose_interleave_4VL_2x2_fp32bf16(bfloat16 *out, const float *in, size_t width, size_t in_stride, size_t height) +{ + float *pad_row = reinterpret_cast(alloca(width * sizeof(float))); + + if (height % 2) { + memset(pad_row, 0, width * sizeof(float)); + } + + size_t out_stride = 4 * roundup(height, 2) * sme::get_vector_length(); + + __asm__ __volatile__( + ".inst 0xd503477f // SMSTART ZA\n" + "cmp %x[height], #0x4\n" + "ptrue p4.b\n" + "blt 4f\n" + "1:" // Main row loop: Head + "mov x25, %x[in]\n" + "add x24, x25, %x[in_stride]\n" + "add x23, x24, %x[in_stride]\n" + "add x22, x23, %x[in_stride]\n" + "add %x[in], x22, %x[in_stride]\n" + "mov x21, %x[out]\n" + "sub %x[height], %x[height], #0x4\n" + "mov x20, %x[width]\n" + "2:" // Main row loop: Column loop + "mov x19, x20\n" + "whilelt p3.s, XZR, x19\n" + "ld1w { z16.s }, p3/Z, [x25]\n" + ".inst 0x658ab218 // bfcvt z24.h, p4/M, z16.s\n" + "decw x19\n" + "whilelt p2.s, XZR, x19\n" + "ld1w { z16.s }, p2/Z, [x25, #1, MUL VL]\n" + ".inst 0x658ab217 // bfcvt z23.h, p4/M, z16.s\n" + "decw x19\n" + "whilelt p1.s, XZR, x19\n" + "ld1w { z16.s }, p1/Z, [x25, #2, MUL VL]\n" + ".inst 0x658ab216 // bfcvt z22.h, p4/M, z16.s\n" + "decw x19\n" + "whilelt p0.s, XZR, x19\n" + "ld1w { z16.s }, p0/Z, [x25, #3, MUL VL]\n" + ".inst 0x658ab215 // bfcvt z21.h, p4/M, z16.s\n" + "ld1w { z16.s }, p3/Z, [x23]\n" + ".inst 0x658ab214 // bfcvt z20.h, p4/M, z16.s\n" + "decw x20, ALL, MUL #4\n" + "cmp x20, #0x0\n" + "ld1w { z16.s }, p2/Z, [x23, #1, MUL VL]\n" + ".inst 0x658ab213 // bfcvt z19.h, p4/M, z16.s\n" + "addvl x25, x25, #4\n" + "ld1w { z16.s }, p1/Z, [x23, #2, MUL VL]\n" + ".inst 0x658ab212 // bfcvt z18.h, p4/M, z16.s\n" + "ld1w { z16.s }, p0/Z, [x23, #3, MUL VL]\n" + ".inst 0x658ab211 // bfcvt z17.h, p4/M, z16.s\n" + "addvl x23, x23, #4\n" + "ld1w { z16.s }, p3/Z, [x24]\n" + ".inst 0x648ab218 // bfcvtnt z24.h, p4/M, z16.s\n" + "ld1w { z16.s }, p2/Z, [x24, #1, MUL VL]\n" + ".inst 0x648ab217 // bfcvtnt z23.h, p4/M, z16.s\n" + "ld1w { z16.s }, p1/Z, [x24, #2, MUL VL]\n" + ".inst 0x648ab216 // bfcvtnt z22.h, p4/M, z16.s\n" + "ld1w { z16.s }, p0/Z, [x24, #3, MUL VL]\n" + "addvl x24, x24, #4\n" + ".inst 0x648ab215 // bfcvtnt z21.h, p4/M, z16.s\n" + "ld1w { z16.s }, p3/Z, [x22]\n" + ".inst 0x648ab214 // bfcvtnt z20.h, p4/M, z16.s\n" + "ld1w { z16.s }, p2/Z, [x22, #1, MUL VL]\n" + ".inst 0x648ab213 // bfcvtnt z19.h, p4/M, z16.s\n" + "ld1w { z16.s }, p1/Z, [x22, #2, MUL VL]\n" + ".inst 0x648ab212 // bfcvtnt z18.h, p4/M, z16.s\n" + "ld1w { z16.s }, p0/Z, [x22, #3, MUL VL]\n" + "addvl x22, x22, #4\n" + ".inst 0x648ab211 // bfcvtnt z17.h, p4/M, z16.s\n" + "st1h { z24.h }, p4, [x21]\n" + "st1h { z23.h }, p4, [x21, #1, MUL VL]\n" + "st1h { z22.h }, p4, [x21, #2, MUL VL]\n" + "st1h { z21.h }, p4, [x21, #3, MUL VL]\n" + "st1h { z20.h }, p4, [x21, #4, MUL VL]\n" + "st1h { z19.h }, p4, [x21, #5, MUL VL]\n" + "st1h { z18.h }, p4, [x21, #6, MUL VL]\n" + "st1h { z17.h }, p4, [x21, #7, MUL VL]\n" + "add x21, x21, %x[out_stride]\n" + "bgt 2b\n" + "3:" // Main row loop: Column loop skip + "cmp %x[height], #0x4\n" + "addvl %x[out], %x[out], #8\n" + "bge 1b\n" + "cbz %x[height], 8f\n" + "4:" // Main loop skip + "5:" // Tail row loop: Head + "mov x25, %x[in]\n" + "add x24, x25, %x[in_stride]\n" + "cmp %x[height], #0x1\n" + "add %x[in], x24, %x[in_stride]\n" + "mov x21, %x[out]\n" + "csel x24, x24, %x[pad_row], GT\n" + "sub %x[height], %x[height], #0x2\n" + "mov x20, %x[width]\n" + "6:" // Tail row loop: Column loop + "mov x19, x20\n" + "whilelt p3.s, XZR, x19\n" + "ld1w { z16.s }, p3/Z, [x25]\n" + ".inst 0x658ab214 // bfcvt z20.h, p4/M, z16.s\n" + "decw x19\n" + "whilelt p2.s, XZR, x19\n" + "ld1w { z16.s }, p2/Z, [x25, #1, MUL VL]\n" + ".inst 0x658ab213 // bfcvt z19.h, p4/M, z16.s\n" + "decw x19\n" + "whilelt p1.s, XZR, x19\n" + "ld1w { z16.s }, p1/Z, [x25, #2, MUL VL]\n" + ".inst 0x658ab212 // bfcvt z18.h, p4/M, z16.s\n" + "decw x19\n" + "whilelt p0.s, XZR, x19\n" + "ld1w { z16.s }, p0/Z, [x25, #3, MUL VL]\n" + ".inst 0x658ab211 // bfcvt z17.h, p4/M, z16.s\n" + "ld1w { z16.s }, p3/Z, [x24]\n" + "decw x20, ALL, MUL #4\n" + "cmp x20, #0x0\n" + ".inst 0x648ab214 // bfcvtnt z20.h, p4/M, z16.s\n" + "ld1w { z16.s }, p2/Z, [x24, #1, MUL VL]\n" + "addvl x25, x25, #4\n" + ".inst 0x648ab213 // bfcvtnt z19.h, p4/M, z16.s\n" + "ld1w { z16.s }, p1/Z, [x24, #2, MUL VL]\n" + ".inst 0x648ab212 // bfcvtnt z18.h, p4/M, z16.s\n" + "ld1w { z16.s }, p0/Z, [x24, #3, MUL VL]\n" + "addvl x24, x24, #4\n" + ".inst 0x648ab211 // bfcvtnt z17.h, p4/M, z16.s\n" + "st1h { z20.h }, p4, [x21]\n" + "st1h { z19.h }, p4, [x21, #1, MUL VL]\n" + "st1h { z18.h }, p4, [x21, #2, MUL VL]\n" + "st1h { z17.h }, p4, [x21, #3, MUL VL]\n" + "add x21, x21, %x[out_stride]\n" + "bgt 6b\n" + "7:" // Tail row loop: Column loop skip + "cmp %x[height], #0x1\n" + "addvl %x[out], %x[out], #4\n" + "bge 5b\n" + "8:" // Done + ".inst 0xd503467f // SMSTOP\n" + : [height] "+&r" (height), [in] "+&r" (in), [out] "+&r" (out) + : [in_stride] "r" (in_stride), [out_stride] "r" (out_stride), [pad_row] "r" (pad_row), [width] "r" (width) + : "cc", "memory", "p0", "p1", "p2", "p3", "p4", "p5", "p6", "p7", "p8", "p9", "p10", "p11", "p12", "p13", "p14", "p15", "x19", "x20", "x21", "x22", "x23", "x24", "x25", "z0", "z1", "z2", "z3", "z4", "z5", "z6", "z7", "z8", "z9", "z10", "z11", "z12", "z13", "z14", "z15", "z16", "z17", "z18", "z19", "z20", "z21", "z22", "z23", "z24", "z25", "z26", "z27", "z28", "z29", "z30", "z31" + ); +} + +} // anonymous namespace +template<> +void Transform<4, 2, true, VLType::SME>( + bfloat16 *out, const float *in, int stride, int x0, int xmax, int k0, int kmax) +{ + sme_transpose_interleave_4VL_2x2_fp32bf16( + out, + in + k0 * stride + x0, + (xmax-x0), + stride * sizeof(float), + (kmax-k0) + ); +} + +#endif -- cgit v1.2.1