diff options
author | Viet-Hoa Do <viet-hoa.do@arm.com> | 2022-06-01 11:47:14 +0100 |
---|---|---|
committer | Viet-Hoa Do <viet-hoa.do@arm.com> | 2022-11-28 16:57:42 +0000 |
commit | 03b2971ac69a86f10a1566938d1a25afee15746c (patch) | |
tree | aec7cfc047e1da278b4b71a706cda7b1b0faa158 /src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_s8q_mopa_4VLx1VL | |
parent | 7dc0234331f2150a6b4ac5c2b49de419870f7cf5 (diff) | |
download | ComputeLibrary-03b2971ac69a86f10a1566938d1a25afee15746c.tar.gz |
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 <viet-hoa.do@arm.com>
Change-Id: I2f1ca1d05f8cfeee9309ed1c0a36096a4a6aad5c
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8692
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_s8q_mopa_4VLx1VL')
-rw-r--r-- | src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_s8q_mopa_4VLx1VL/generic.cpp | 507 |
1 files changed, 507 insertions, 0 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_s8q_mopa_4VLx1VL/generic.cpp b/src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_s8q_mopa_4VLx1VL/generic.cpp new file mode 100644 index 0000000000..a738a10418 --- /dev/null +++ b/src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_s8q_mopa_4VLx1VL/generic.cpp @@ -0,0 +1,507 @@ +/* + * 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. + */ +#ifdef __ARM_FEATURE_SVE +#ifdef ARM_COMPUTE_ENABLE_SME2 + +#include "arm_gemm.hpp" + +#include <cstdint> +#include "../../asmlib.hpp" +#include "../../utils.hpp" + +namespace arm_gemm { + +void sme2_interleaved_nomerge_s8q_mopa_4VLx1VL(const int8_t *const A, const int8_t *const B, int8_t *const C, int ldc, const int M, const int N, const int K, const int32_t *const bias, const Requantize32 &rq, const int n_0, bool accumulate, int32_t *const accumulator_buffer) +{ + struct KernelArgs + { + KernelArgs( + const int8_t *const A, + const int8_t *const B, + int8_t *const C, const int ldc, + const int M, const int N, const int K, + const int32_t *const bias, + const Requantize32 &rq, + const int n_0, + bool accumulate, + int32_t *const accumulator_buffer + ) : A(A), + B(B), kstride_bytes(roundup(K, 4) * sizeof(int8_t)), + C(C), ldcb(ldc * sizeof(int8_t)), + M(M), N(N), K(K), + n_loops(((K / 4) - 1) / 2), n_tail_iters(((K / 4) - 1) % 2), + + bias(bias), n_0(n_0), + accumulator_buffer(accumulator_buffer), + flags(0x0) + { + if (accumulate) + { + flags |= 1 << 0; // FILL_ACCUMULATORS_FROM_BUFFER + } + if (C == nullptr) + { + flags |= 1 << 1; // STORE_ACCUMULATORS_TO_BUFFER + } + if (rq.per_channel_requant) + { + flags |= 1 << 2; // PER_CHANNEL_QUANTISATION + } + } + + const int8_t *const A; + const int8_t *const B; + const long kstride_bytes; + int8_t *const C; + const long ldcb; + const long M, N, K, n_loops, n_tail_iters; + int32_t min = std::numeric_limits<int8_t>::min(); + int32_t max = std::numeric_limits<int8_t>::max(); + + const int32_t *const bias; + const int n_0; + + int32_t *const accumulator_buffer; + uint64_t flags; + }; + + // Construct arguments for this kernel + KernelArgs args(A, B, C, ldc, M, N, K, bias, rq, n_0, accumulate, accumulator_buffer); + + __asm__ __volatile__( + "ldr x15, [%x[args], %[offsetof_flags]]\n" + ".inst 0xd503477f // SMSTART ZA\n" + "ptrue p1.b\n" + ".inst 0x25207810 // ptrue pn8.b\n" + "ldr x14, [%x[args], %[offsetof_accumulator_buffer]]\n" + "ldr x13, [%x[args], %[offsetof_accumulator_buffer]]\n" + "tbz x15, #0, 2f\n" + "mov x12, #0x0\n" + "cntw x19\n" + "1:" // Initial accumulator load from buffer: Loop + ".inst 0xa040c1dc // ld1w { z28.s-z31.s }, pn8.b/Z, [x14]\n" + ".inst 0xc0840780 // mova za0h.s[x12], { z28.s-z31.s }\n" + ".inst 0xa041c1cc // ld1w { z12.s-z15.s }, pn8.b/Z, [x14, #0x4, MUL VL]\n" + ".inst 0xc0840581 // mova za1h.s[x12], { z12.s-z15.s }\n" + ".inst 0xa042c1d4 // ld1w { z20.s-z23.s }, pn8.b/Z, [x14, #0x8, MUL VL]\n" + ".inst 0xc0840682 // mova za2h.s[x12], { z20.s-z23.s }\n" + ".inst 0xa043c1d8 // ld1w { z24.s-z27.s }, pn8.b/Z, [x14, #0xc, MUL VL]\n" + ".inst 0xc0840703 // mova za3h.s[x12], { z24.s-z27.s }\n" + "add x12, x12, #0x4\n" + "cmp x12, x19\n" + "addvl x14, x14, #16\n" + "blt 1b\n" + "2:" // Initial accumulator load from buffer: End + "ldr w11, [%x[args], %[offsetof_M]]\n" + "mov x10, #0x0\n" + "mov x9, #0x0\n" + "ldr w28, [%x[args], %[offsetof_N]]\n" + "ldr x27, [%x[args], %[offsetof_A]]\n" + "3:" // M and N loop + "mov x26, x27\n" + "whilelt p0.s, x9, x28\n" + "tbnz x15, #0, 4f\n" + "ldr x19, [%x[args], %[offsetof_bias]]\n" + ".inst 0xc00800ff // zero { zad0, zad1, zad2, zad3, zad4, zad5, zad6, zad7 }\n" + "cbz x19, 5f\n" + "ldnt1w { z15.s }, p0/Z, [x19, x9, LSL #2]\n" + ".inst 0xc09025e0 // addha za0.s, p1/M, p1/M, z15.s\n" + ".inst 0xc09025e1 // addha za1.s, p1/M, p1/M, z15.s\n" + ".inst 0xc09025e2 // addha za2.s, p1/M, p1/M, z15.s\n" + ".inst 0xc09025e3 // addha za3.s, p1/M, p1/M, z15.s\n" + "4:" // Prepare accumulators: Test for last block + "mov x19, x9\n" + "mov x20, x10\n" + "incw x19\n" + "incw x20, ALL, MUL #4\n" + "cmp x19, x28\n" + "csel x20, x10, x20, LT\n" + "mov x19, x15\n" + "bfm x15, XZR, #0x0, #0x0 // bfc x15, #0x0, #0x1\n" + "cmp x20, x11\n" + "csel x15, x19, x15, LT\n" + "5:" // Prepare accumulators: End + "ldr x19, [%x[args], %[offsetof_K]]\n" + "add x19, x19, #0x3\n" + "lsr x19, x19, #0x2\n" + "ldr x22, [%x[args], %[offsetof_B]]\n" + "lsr x21, x19, #0x2\n" + "and x20, x19, #0x3\n" + "ldr x19, [%x[args], %[offsetof_kstride_bytes]]\n" + "madd x22, x9, x19, x22\n" // bptr = B + n * kstride_bytes + "cbz x21, 8f\n" + "subs x21, x21, #0x1\n" + ".inst 0xa1408352 // ld1b { z18.b, z22.b, z26.b, z30.b }, pn8.b/Z, [x26]\n" + "ldnt1b { z0.b }, p1/Z, [x22]\n" + ".inst 0xa1418353 // ld1b { z19.b, z23.b, z27.b, z31.b }, pn8.b/Z, [x26, #0x4, MUL VL]\n" + "ldnt1b { z9.b }, p1/Z, [x22, #1, MUL VL]\n" + ".inst 0xa1428350 // ld1b { z16.b, z20.b, z24.b, z28.b }, pn8.b/Z, [x26, #0x8, MUL VL]\n" + "ldnt1b { z21.b }, p1/Z, [x22, #2, MUL VL]\n" + ".inst 0xa1438342 // ld1b { z2.b, z6.b, z10.b, z14.b }, pn8.b/Z, [x26, #0xc, MUL VL]\n" + "addvl x26, x26, #16\n" + "ldnt1b { z12.b }, p1/Z, [x22, #3, MUL VL]\n" + "addvl x22, x22, #4\n" + "ble 7f\n" + "6:" // K loop + ".inst 0xa0802640 // smopa za0.s, p1/M, p1/M, z18.b, z0.b\n" + "subs x21, x21, #0x1\n" + ".inst 0xa08026c1 // smopa za1.s, p1/M, p1/M, z22.b, z0.b\n" + ".inst 0xa0802742 // smopa za2.s, p1/M, p1/M, z26.b, z0.b\n" + ".inst 0xa08027c3 // smopa za3.s, p1/M, p1/M, z30.b, z0.b\n" + ".inst 0xa1408352 // ld1b { z18.b, z22.b, z26.b, z30.b }, pn8.b/Z, [x26]\n" + ".inst 0xa0892660 // smopa za0.s, p1/M, p1/M, z19.b, z9.b\n" + "ldnt1b { z0.b }, p1/Z, [x22]\n" + ".inst 0xa08926e1 // smopa za1.s, p1/M, p1/M, z23.b, z9.b\n" + ".inst 0xa0892762 // smopa za2.s, p1/M, p1/M, z27.b, z9.b\n" + ".inst 0xa08927e3 // smopa za3.s, p1/M, p1/M, z31.b, z9.b\n" + ".inst 0xa1418353 // ld1b { z19.b, z23.b, z27.b, z31.b }, pn8.b/Z, [x26, #0x4, MUL VL]\n" + ".inst 0xa0952600 // smopa za0.s, p1/M, p1/M, z16.b, z21.b\n" + "ldnt1b { z9.b }, p1/Z, [x22, #1, MUL VL]\n" + ".inst 0xa0952681 // smopa za1.s, p1/M, p1/M, z20.b, z21.b\n" + ".inst 0xa0952702 // smopa za2.s, p1/M, p1/M, z24.b, z21.b\n" + ".inst 0xa0952783 // smopa za3.s, p1/M, p1/M, z28.b, z21.b\n" + ".inst 0xa1428350 // ld1b { z16.b, z20.b, z24.b, z28.b }, pn8.b/Z, [x26, #0x8, MUL VL]\n" + "ldnt1b { z21.b }, p1/Z, [x22, #2, MUL VL]\n" + ".inst 0xa08c2440 // smopa za0.s, p1/M, p1/M, z2.b, z12.b\n" + ".inst 0xa08c24c1 // smopa za1.s, p1/M, p1/M, z6.b, z12.b\n" + ".inst 0xa08c2542 // smopa za2.s, p1/M, p1/M, z10.b, z12.b\n" + ".inst 0xa08c25c3 // smopa za3.s, p1/M, p1/M, z14.b, z12.b\n" + ".inst 0xa1438342 // ld1b { z2.b, z6.b, z10.b, z14.b }, pn8.b/Z, [x26, #0xc, MUL VL]\n" + "addvl x26, x26, #16\n" + "ldnt1b { z12.b }, p1/Z, [x22, #3, MUL VL]\n" + "addvl x22, x22, #4\n" + "bgt 6b\n" + "7:" // K loop tail + ".inst 0xa0802640 // smopa za0.s, p1/M, p1/M, z18.b, z0.b\n" + ".inst 0xa08026c1 // smopa za1.s, p1/M, p1/M, z22.b, z0.b\n" + ".inst 0xa0802742 // smopa za2.s, p1/M, p1/M, z26.b, z0.b\n" + ".inst 0xa08027c3 // smopa za3.s, p1/M, p1/M, z30.b, z0.b\n" + ".inst 0xa0892660 // smopa za0.s, p1/M, p1/M, z19.b, z9.b\n" + ".inst 0xa08926e1 // smopa za1.s, p1/M, p1/M, z23.b, z9.b\n" + ".inst 0xa0892762 // smopa za2.s, p1/M, p1/M, z27.b, z9.b\n" + ".inst 0xa08927e3 // smopa za3.s, p1/M, p1/M, z31.b, z9.b\n" + ".inst 0xa0952600 // smopa za0.s, p1/M, p1/M, z16.b, z21.b\n" + ".inst 0xa0952681 // smopa za1.s, p1/M, p1/M, z20.b, z21.b\n" + ".inst 0xa0952702 // smopa za2.s, p1/M, p1/M, z24.b, z21.b\n" + ".inst 0xa0952783 // smopa za3.s, p1/M, p1/M, z28.b, z21.b\n" + ".inst 0xa08c2440 // smopa za0.s, p1/M, p1/M, z2.b, z12.b\n" + ".inst 0xa08c24c1 // smopa za1.s, p1/M, p1/M, z6.b, z12.b\n" + ".inst 0xa08c2542 // smopa za2.s, p1/M, p1/M, z10.b, z12.b\n" + ".inst 0xa08c25c3 // smopa za3.s, p1/M, p1/M, z14.b, z12.b\n" + "8:" // K oddments + "cbz x20, 10f\n" + "9:" // K oddments: Loop + ".inst 0xa1408352 // ld1b { z18.b, z22.b, z26.b, z30.b }, pn8.b/Z, [x26]\n" + "subs x20, x20, #0x1\n" + "addvl x26, x26, #4\n" + "ld1b { z0.b }, p1/Z, [x22]\n" + "addvl x22, x22, #1\n" + ".inst 0xa0802640 // smopa za0.s, p1/M, p1/M, z18.b, z0.b\n" + ".inst 0xa08026c1 // smopa za1.s, p1/M, p1/M, z22.b, z0.b\n" + ".inst 0xa0802742 // smopa za2.s, p1/M, p1/M, z26.b, z0.b\n" + ".inst 0xa08027c3 // smopa za3.s, p1/M, p1/M, z30.b, z0.b\n" + "bgt 9b\n" + "10:" // K oddments: End + ".inst 0xa040c340 // ld1w { z0.s-z3.s }, pn8.b/Z, [x26]\n" + "addvl x26, x26, #4\n" + ".inst 0xc0912400 // addva za0.s, p1/M, p1/M, z0.s\n" + ".inst 0xc0912421 // addva za1.s, p1/M, p1/M, z1.s\n" + ".inst 0xc0912442 // addva za2.s, p1/M, p1/M, z2.s\n" + ".inst 0xc0912463 // addva za3.s, p1/M, p1/M, z3.s\n" + "tbz x15, #1, 14f\n" + "tbz x15, #0, 12f\n" + "mov x12, #0x0\n" + "cntw x19\n" + "11:" // Store to partial result buffer: Store and refill: Loop + ".inst 0xa040c1d4 // ld1w { z20.s-z23.s }, pn8.b/Z, [x14]\n" + ".inst 0xc0860410 // mova { z16.s-z19.s }, za0h.s[x12]\n" + ".inst 0xc0840680 // mova za0h.s[x12], { z20.s-z23.s }\n" + ".inst 0xc0860428 // mova { z8.s-z11.s }, za1h.s[x12]\n" + ".inst 0xa041c1c4 // ld1w { z4.s-z7.s }, pn8.b/Z, [x14, #0x4, MUL VL]\n" + ".inst 0xc0840481 // mova za1h.s[x12], { z4.s-z7.s }\n" + ".inst 0xc086044c // mova { z12.s-z15.s }, za2h.s[x12]\n" + ".inst 0xc086047c // mova { z28.s-z31.s }, za3h.s[x12]\n" + ".inst 0xa042c1c4 // ld1w { z4.s-z7.s }, pn8.b/Z, [x14, #0x8, MUL VL]\n" + ".inst 0xc0840482 // mova za2h.s[x12], { z4.s-z7.s }\n" + ".inst 0xa043c1d4 // ld1w { z20.s-z23.s }, pn8.b/Z, [x14, #0xc, MUL VL]\n" + ".inst 0xc0840683 // mova za3h.s[x12], { z20.s-z23.s }\n" + "add x12, x12, #0x4\n" + "cmp x12, x19\n" + ".inst 0xa060c1b0 // st1w { z16.s-z19.s }, pn8.b, [x13]\n" + "addvl x14, x14, #16\n" + ".inst 0xa061c1a8 // st1w { z8.s-z11.s }, pn8.b, [x13, #0x4, MUL VL]\n" + ".inst 0xa062c1ac // st1w { z12.s-z15.s }, pn8.b, [x13, #0x8, MUL VL]\n" + ".inst 0xa063c1bc // st1w { z28.s-z31.s }, pn8.b, [x13, #0xc, MUL VL]\n" + "addvl x13, x13, #16\n" + "blt 11b\n" + "b 30f\n" + "12:" // Store to partial result buffer: Store only + "mov x12, #0x0\n" + "cntw x19\n" + "13:" // Store to partial result buffer: Store only: Loop + ".inst 0xc0860410 // mova { z16.s-z19.s }, za0h.s[x12]\n" + ".inst 0xc086042c // mova { z12.s-z15.s }, za1h.s[x12]\n" + ".inst 0xa060c1b0 // st1w { z16.s-z19.s }, pn8.b, [x13]\n" + ".inst 0xc0860454 // mova { z20.s-z23.s }, za2h.s[x12]\n" + ".inst 0xc0860478 // mova { z24.s-z27.s }, za3h.s[x12]\n" + ".inst 0xa061c1ac // st1w { z12.s-z15.s }, pn8.b, [x13, #0x4, MUL VL]\n" + "add x12, x12, #0x4\n" + "cmp x12, x19\n" + ".inst 0xa062c1b4 // st1w { z20.s-z23.s }, pn8.b, [x13, #0x8, MUL VL]\n" + ".inst 0xa063c1b8 // st1w { z24.s-z27.s }, pn8.b, [x13, #0xc, MUL VL]\n" + "addvl x13, x13, #16\n" + "blt 13b\n" + "b 30f\n" + "14:" // Store to output array + "ldr x25, [%x[args], %[offsetof_C]]\n" + "add x25, x25, x9\n" // C += n + "sub x24, x11, x10\n" + "ld1rw { z8.s }, p1/Z, [%x[rq], %[offsetof_Requantize32_per_layer_mul]]\n" + "ldr x23, [%x[args], %[offsetof_ldcb]]\n" + "madd x25, x10, x23, x25\n" // C += m * ldc + "ld1rw { z7.s }, p1/Z, [%x[rq], %[offsetof_Requantize32_per_layer_right_shift]]\n" + "ld1rw { z6.s }, p1/Z, [%x[rq], %[offsetof_Requantize32_c_offset]]\n" + "ld1rw { z5.s }, p1/Z, [%x[rq], %[offsetof_Requantize32_minval]]\n" + "ld1rw { z4.s }, p1/Z, [%x[rq], %[offsetof_Requantize32_maxval]]\n" + "tbz x15, #2, 15f\n" + "ldr w20, [%x[args], %[offsetof_n_0]]\n" + "add x20, x20, x9\n" + "ldr x19, [%x[rq], %[offsetof_Requantize32_per_channel_muls]]\n" + "add x19, x19, x20, LSL #2\n" + "ld1w { z8.s }, p0/Z, [x19]\n" + "ldr x19, [%x[rq], %[offsetof_Requantize32_per_channel_right_shifts]]\n" + "add x19, x19, x20, LSL #2\n" + "ld1w { z7.s }, p0/Z, [x19]\n" + "15:" // Store to output array: Load per-channel parameters: End + "cntw x22\n" + "whilelt p0.s, x9, x28\n" + "cmp x24, x22\n" + "csel x21, x24, x22, LT\n" + "lsr x20, x21, #0x2\n" + "mov x12, #0x0\n" + "and x19, x21, #0x3\n" + "cbz x20, 17f\n" + "16:" // Store to output array: Accumulator row 0 loop + ".inst 0xc086040c // mova { z12.s-z15.s }, za0h.s[x12]\n" + ".inst 0xc1a8ac0c // sqdmulh { z12.s-z15.s }, { z12.s-z15.s }, z8.s\n" + "add x12, x12, #0x4\n" + ".inst 0xc1a7aa2c // srshl { z12.s-z15.s }, { z12.s-z15.s }, z7.s\n" + "cmp x12, x20, LSL #2\n" + ".inst 0xc1a6ab0c // add { z12.s-z15.s }, { z12.s-z15.s }, z6.s\n" + ".inst 0xc1a4ccac // sclamp { z12.s-z15.s }, z5.s, z4.s\n" + "st1b { z12.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "st1b { z13.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "st1b { z14.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "st1b { z15.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "blt 16b\n" + "17:" // Store to output array: Accumulator row 0 oddments + "cbz x19, 18f\n" + ".inst 0xc0860410 // mova { z16.s-z19.s }, za0h.s[x12]\n" + ".inst 0xc1a8ac10 // sqdmulh { z16.s-z19.s }, { z16.s-z19.s }, z8.s\n" + "subs x19, x19, #0x1\n" + ".inst 0xc1a7aa30 // srshl { z16.s-z19.s }, { z16.s-z19.s }, z7.s\n" + ".inst 0xc1a6ab10 // add { z16.s-z19.s }, { z16.s-z19.s }, z6.s\n" + ".inst 0xc1a4ccb0 // sclamp { z16.s-z19.s }, z5.s, z4.s\n" + "st1b { z16.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "beq 18f\n" + "subs x19, x19, #0x1\n" + "st1b { z17.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "beq 18f\n" + "st1b { z18.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "18:" // Store to output array: Accumulator row 0 oddments: End + "subs x24, x24, x21\n" + "beq 28f\n" + "whilelt p0.s, x9, x28\n" + "cmp x24, x22\n" + "csel x21, x24, x22, LT\n" + "lsr x20, x21, #0x2\n" + "mov x12, #0x0\n" + "and x19, x21, #0x3\n" + "cbz x20, 20f\n" + "19:" // Store to output array: Accumulator row 1 loop + ".inst 0xc0860430 // mova { z16.s-z19.s }, za1h.s[x12]\n" + ".inst 0xc1a8ac10 // sqdmulh { z16.s-z19.s }, { z16.s-z19.s }, z8.s\n" + "add x12, x12, #0x4\n" + ".inst 0xc1a7aa30 // srshl { z16.s-z19.s }, { z16.s-z19.s }, z7.s\n" + "cmp x12, x20, LSL #2\n" + ".inst 0xc1a6ab10 // add { z16.s-z19.s }, { z16.s-z19.s }, z6.s\n" + ".inst 0xc1a4ccb0 // sclamp { z16.s-z19.s }, z5.s, z4.s\n" + "st1b { z16.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "st1b { z17.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "st1b { z18.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "st1b { z19.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "blt 19b\n" + "20:" // Store to output array: Accumulator row 1 oddments + "cbz x19, 21f\n" + ".inst 0xc086043c // mova { z28.s-z31.s }, za1h.s[x12]\n" + ".inst 0xc1a8ac1c // sqdmulh { z28.s-z31.s }, { z28.s-z31.s }, z8.s\n" + "subs x19, x19, #0x1\n" + ".inst 0xc1a7aa3c // srshl { z28.s-z31.s }, { z28.s-z31.s }, z7.s\n" + ".inst 0xc1a6ab1c // add { z28.s-z31.s }, { z28.s-z31.s }, z6.s\n" + ".inst 0xc1a4ccbc // sclamp { z28.s-z31.s }, z5.s, z4.s\n" + "st1b { z28.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "beq 21f\n" + "subs x19, x19, #0x1\n" + "st1b { z29.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "beq 21f\n" + "st1b { z30.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "21:" // Store to output array: Accumulator row 1 oddments: End + "subs x24, x24, x21\n" + "beq 28f\n" + "whilelt p0.s, x9, x28\n" + "cmp x24, x22\n" + "csel x21, x24, x22, LT\n" + "lsr x20, x21, #0x2\n" + "mov x12, #0x0\n" + "and x19, x21, #0x3\n" + "cbz x20, 23f\n" + "22:" // Store to output array: Accumulator row 2 loop + ".inst 0xc0860458 // mova { z24.s-z27.s }, za2h.s[x12]\n" + ".inst 0xc1a8ac18 // sqdmulh { z24.s-z27.s }, { z24.s-z27.s }, z8.s\n" + "add x12, x12, #0x4\n" + ".inst 0xc1a7aa38 // srshl { z24.s-z27.s }, { z24.s-z27.s }, z7.s\n" + "cmp x12, x20, LSL #2\n" + ".inst 0xc1a6ab18 // add { z24.s-z27.s }, { z24.s-z27.s }, z6.s\n" + ".inst 0xc1a4ccb8 // sclamp { z24.s-z27.s }, z5.s, z4.s\n" + "st1b { z24.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "st1b { z25.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "st1b { z26.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "st1b { z27.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "blt 22b\n" + "23:" // Store to output array: Accumulator row 2 oddments + "cbz x19, 24f\n" + ".inst 0xc086044c // mova { z12.s-z15.s }, za2h.s[x12]\n" + ".inst 0xc1a8ac0c // sqdmulh { z12.s-z15.s }, { z12.s-z15.s }, z8.s\n" + "subs x19, x19, #0x1\n" + ".inst 0xc1a7aa2c // srshl { z12.s-z15.s }, { z12.s-z15.s }, z7.s\n" + ".inst 0xc1a6ab0c // add { z12.s-z15.s }, { z12.s-z15.s }, z6.s\n" + ".inst 0xc1a4ccac // sclamp { z12.s-z15.s }, z5.s, z4.s\n" + "st1b { z12.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "beq 24f\n" + "subs x19, x19, #0x1\n" + "st1b { z13.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "beq 24f\n" + "st1b { z14.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "24:" // Store to output array: Accumulator row 2 oddments: End + "subs x24, x24, x21\n" + "beq 28f\n" + "whilelt p0.s, x9, x28\n" + "cmp x24, x22\n" + "csel x19, x24, x22, LT\n" + "lsr x20, x19, #0x2\n" + "mov x12, #0x0\n" + "and x19, x19, #0x3\n" + "cbz x20, 26f\n" + "25:" // Store to output array: Accumulator row 3 loop + ".inst 0xc0860474 // mova { z20.s-z23.s }, za3h.s[x12]\n" + ".inst 0xc1a8ac14 // sqdmulh { z20.s-z23.s }, { z20.s-z23.s }, z8.s\n" + "add x12, x12, #0x4\n" + ".inst 0xc1a7aa34 // srshl { z20.s-z23.s }, { z20.s-z23.s }, z7.s\n" + "cmp x12, x20, LSL #2\n" + ".inst 0xc1a6ab14 // add { z20.s-z23.s }, { z20.s-z23.s }, z6.s\n" + ".inst 0xc1a4ccb4 // sclamp { z20.s-z23.s }, z5.s, z4.s\n" + "st1b { z20.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "st1b { z21.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "st1b { z22.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "st1b { z23.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "blt 25b\n" + "26:" // Store to output array: Accumulator row 3 oddments + "cbz x19, 27f\n" + ".inst 0xc0860460 // mova { z0.s-z3.s }, za3h.s[x12]\n" + ".inst 0xc1a8ac00 // sqdmulh { z0.s-z3.s }, { z0.s-z3.s }, z8.s\n" + "subs x19, x19, #0x1\n" + ".inst 0xc1a7aa20 // srshl { z0.s-z3.s }, { z0.s-z3.s }, z7.s\n" + ".inst 0xc1a6ab00 // add { z0.s-z3.s }, { z0.s-z3.s }, z6.s\n" + ".inst 0xc1a4cca0 // sclamp { z0.s-z3.s }, z5.s, z4.s\n" + "st1b { z0.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "beq 27f\n" + "subs x19, x19, #0x1\n" + "st1b { z1.s }, p0, [x25]\n" + "add x25, x25, x23\n" + "beq 27f\n" + "st1b { z2.s }, p0, [x25]\n" + "27:" // Store to output array: Accumulator row 3 oddments: End + "28:" // Store to output array: End + "tbz x15, #0, 30f\n" + "mov x12, #0x0\n" + "cntw x19\n" + "29:" // Store to output array: Refill accumulators: Loop + ".inst 0xa040c1c4 // ld1w { z4.s-z7.s }, pn8.b/Z, [x14]\n" + ".inst 0xc0840480 // mova za0h.s[x12], { z4.s-z7.s }\n" + ".inst 0xa041c1d0 // ld1w { z16.s-z19.s }, pn8.b/Z, [x14, #0x4, MUL VL]\n" + ".inst 0xc0840601 // mova za1h.s[x12], { z16.s-z19.s }\n" + ".inst 0xa042c1d0 // ld1w { z16.s-z19.s }, pn8.b/Z, [x14, #0x8, MUL VL]\n" + ".inst 0xc0840602 // mova za2h.s[x12], { z16.s-z19.s }\n" + ".inst 0xa043c1c4 // ld1w { z4.s-z7.s }, pn8.b/Z, [x14, #0xc, MUL VL]\n" + ".inst 0xc0840483 // mova za3h.s[x12], { z4.s-z7.s }\n" + "add x12, x12, #0x4\n" + "cmp x12, x19\n" + "addvl x14, x14, #16\n" + "blt 29b\n" + "30:" // End block + "incw x9\n" + "cmp x9, x28\n" + "blt 3b\n" + "incw x10, ALL, MUL #4\n" + "cmp x10, x11\n" + "mov x9, #0x0\n" + "mov x27, x26\n" + "blt 3b\n" + ".inst 0xd503467f // SMSTOP\n" + : + : [args] "r" (&args), [offsetof_A] "I" (offsetof(KernelArgs, A)), [offsetof_B] "I" (offsetof(KernelArgs, B)), [offsetof_C] "I" (offsetof(KernelArgs, C)), [offsetof_K] "I" (offsetof(KernelArgs, K)), [offsetof_M] "I" (offsetof(KernelArgs, M)), [offsetof_N] "I" (offsetof(KernelArgs, N)), [offsetof_Requantize32_c_offset] "I" (offsetof(Requantize32, c_offset)), [offsetof_Requantize32_maxval] "I" (offsetof(Requantize32, maxval)), [offsetof_Requantize32_minval] "I" (offsetof(Requantize32, minval)), [offsetof_Requantize32_per_channel_muls] "I" (offsetof(Requantize32, per_channel_muls)), [offsetof_Requantize32_per_channel_right_shifts] "I" (offsetof(Requantize32, per_channel_right_shifts)), [offsetof_Requantize32_per_layer_mul] "I" (offsetof(Requantize32, per_layer_mul)), [offsetof_Requantize32_per_layer_right_shift] "I" (offsetof(Requantize32, per_layer_right_shift)), [offsetof_accumulator_buffer] "I" (offsetof(KernelArgs, accumulator_buffer)), [offsetof_bias] "I" (offsetof(KernelArgs, bias)), [offsetof_flags] "I" (offsetof(KernelArgs, flags)), [offsetof_kstride_bytes] "I" (offsetof(KernelArgs, kstride_bytes)), [offsetof_ldcb] "I" (offsetof(KernelArgs, ldcb)), [offsetof_n_0] "I" (offsetof(KernelArgs, n_0)), [rq] "r" (&rq) + : "cc", "memory", "p0", "p1", "p2", "p3", "p4", "p5", "p6", "p7", "p8", "p9", "p10", "p11", "p12", "p13", "p14", "p15", "x9", "x10", "x11", "x12", "x13", "x14", "x15", "x19", "x20", "x21", "x22", "x23", "x24", "x25", "x26", "x27", "x28", "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" + ); +} + +} // namespace arm_gemm + +#endif // ARM_COMPUTE_ENABLE_SME2 +#endif // __ARM_FEATURE_SVE |