diff options
Diffstat (limited to 'src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_bf16fp32_mopa_1VLx4VL/generic.cpp')
-rw-r--r-- | src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_bf16fp32_mopa_1VLx4VL/generic.cpp | 418 |
1 files changed, 418 insertions, 0 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_bf16fp32_mopa_1VLx4VL/generic.cpp b/src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_bf16fp32_mopa_1VLx4VL/generic.cpp new file mode 100644 index 0000000000..8105300cb7 --- /dev/null +++ b/src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_bf16fp32_mopa_1VLx4VL/generic.cpp @@ -0,0 +1,418 @@ +/* + * Copyright (c) 2022-2023 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_COMPUTE_ENABLE_SME2 + +#include "arm_gemm.hpp" + +#include "../../bfloat.hpp" +#include "../../asmlib.hpp" +#include "../../utils.hpp" + +namespace arm_gemm { + +void sme2_interleaved_nomerge_bf16fp32_mopa_1VLx4VL(const bfloat16 *const A, const bfloat16 *const B, float *const C, int ldc, const int M, const int N, const int K, const float *const bias, const Activation act, bool accumulate, float *const accumulator_buffer) +{ + struct KernelArgs + { + KernelArgs( + const bfloat16 *const A, + const bfloat16 *const B, + float *const C, const int ldc, + const int M, const int N, const int K, + const float *const bias, + const Activation act, + bool accumulate, + float *const accumulator_buffer + ) : A(A), + B(B), kstride_bytes(roundup(K, 2) * sizeof(bfloat16)), + C(C), ldcb(ldc * sizeof(float)), + M(M), N(N), K(K), + n_loops(((K / 2) - 1) / 2), n_tail_iters(((K / 2) - 1) % 2), + min(-std::numeric_limits<float>::infinity()), + max(std::numeric_limits<float>::infinity()), + bias(bias), + 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 (act.type == Activation::Type::None) + { + flags |= 1 << 2; // SKIP_ACTIVATION + } + + // Initialise the activation values + switch (act.type) + { + default: + case Activation::Type::None: + break; + case Activation::Type::BoundedReLU: + this->max = static_cast<float>(act.param1); + /* fall through */ + case Activation::Type::ReLU: + this->min = static_cast<float>(0); + break; + } + } + + const bfloat16 *const A; + const bfloat16 *const B; + const long kstride_bytes; + float *const C; + const long ldcb; + const long M, N, K, n_loops, n_tail_iters; + float min = -std::numeric_limits<float>::infinity(); + float max = std::numeric_limits<float>::infinity(); + + const float *const bias; + + float *const accumulator_buffer; + uint64_t flags; + }; + + // Construct arguments for this kernel + KernelArgs args(A, B, C, ldc, M, N, K, bias, act, accumulate, accumulator_buffer); + + __asm__ __volatile__( + "ldr x15, [%x[args], %[offsetof_flags]]\n" + ".inst 0xd503477f // SMSTART ZA\n" + "ptrue p0.b\n" + ".inst 0x25207811 // ptrue pn9.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 x20\n" + "1:" // Initial accumulator load from buffer: Loop + ".inst 0xa040c5c4 // ld1w { z4.s-z7.s }, pn9.b/Z, [x14]\n" + ".inst 0xc0840480 // mova za0h.s[x12], { z4.s-z7.s }\n" + ".inst 0xa041c5d8 // ld1w { z24.s-z27.s }, pn9.b/Z, [x14, #0x4, MUL VL]\n" + ".inst 0xc0840701 // mova za1h.s[x12], { z24.s-z27.s }\n" + ".inst 0xa042c5d4 // ld1w { z20.s-z23.s }, pn9.b/Z, [x14, #0x8, MUL VL]\n" + ".inst 0xc0840682 // mova za2h.s[x12], { z20.s-z23.s }\n" + ".inst 0xa043c5c4 // ld1w { z4.s-z7.s }, pn9.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, x20\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" + ".inst 0x25bc6530 // whilelt pn8.s, x9, x28, VLx4\n" + "tbnz x15, #0, 4f\n" + "ldr x20, [%x[args], %[offsetof_bias]]\n" + ".inst 0xc00800ff // zero { zad0, zad1, zad2, zad3, zad4, zad5, zad6, zad7 }\n" + "cbz x20, 5f\n" + "fmov z6.s, #1.0\n" + ".inst 0xa009c29d // ldnt1w { z28.s-z31.s }, p8/Z, [x20, x9, LSL #2]\n" + ".inst 0x809c00c0 // fmopa za0.s, p0/M, p0/M, z6.s, z28.s\n" + ".inst 0x809d00c1 // fmopa za1.s, p0/M, p0/M, z6.s, z29.s\n" + ".inst 0x809e00c2 // fmopa za2.s, p0/M, p0/M, z6.s, z30.s\n" + ".inst 0x809f00c3 // fmopa za3.s, p0/M, p0/M, z6.s, z31.s\n" + "4:" // Prepare accumulators: Test for last block + "mov x20, x9\n" + "mov x21, x10\n" + "incw x20, ALL, MUL #4\n" + "incw x21\n" + "cmp x20, x28\n" + "csel x21, x10, x21, LT\n" + "mov x20, x15\n" + "bfm x15, XZR, #0x0, #0x0 // bfc x15, #0x0, #0x1\n" + "cmp x21, x11\n" + "csel x15, x20, x15, LT\n" + "5:" // Prepare accumulators: End + "ldr x20, [%x[args], %[offsetof_K]]\n" + "add x20, x20, #0x1\n" + "lsr x20, x20, #0x1\n" + "ldr x23, [%x[args], %[offsetof_B]]\n" + "lsr x22, x20, #0x2\n" + "and x21, x20, #0x3\n" + "ldr x20, [%x[args], %[offsetof_kstride_bytes]]\n" + "madd x23, x9, x20, x23\n" // bptr = B + n * kstride_bytes + "cbz x22, 8f\n" + "subs x22, x22, #0x1\n" + "ld1h { z28.h }, p0/Z, [x26]\n" + ".inst 0xa040a6e9 // ldnt1h { z8.h-z11.h }, pn9.b/Z, [x23]\n" + "ld1h { z22.h }, p0/Z, [x26, #1, MUL VL]\n" + ".inst 0xa041a6ed // ldnt1h { z12.h-z15.h }, pn9.b/Z, [x23, #0x4, MUL VL]\n" + "ld1h { z30.h }, p0/Z, [x26, #2, MUL VL]\n" + ".inst 0xa042a6e5 // ldnt1h { z4.h-z7.h }, pn9.b/Z, [x23, #0x8, MUL VL]\n" + "ld1h { z20.h }, p0/Z, [x26, #3, MUL VL]\n" + "addvl x26, x26, #4\n" + ".inst 0xa143a6fb // ldnt1h { z19.h, z23.h, z27.h, z31.h }, pn9.b/Z, [x23, #0xc, MUL VL]\n" + "addvl x23, x23, #16\n" + "ble 7f\n" + "6:" // K loop + ".inst 0x81880380 // bfmopa za0.s, p0/M, p0/M, z28.h, z8.h\n" + "subs x22, x22, #0x1\n" + ".inst 0x81890381 // bfmopa za1.s, p0/M, p0/M, z28.h, z9.h\n" + ".inst 0x818a0382 // bfmopa za2.s, p0/M, p0/M, z28.h, z10.h\n" + ".inst 0x818b0383 // bfmopa za3.s, p0/M, p0/M, z28.h, z11.h\n" + "ld1h { z28.h }, p0/Z, [x26]\n" + ".inst 0x818c02c0 // bfmopa za0.s, p0/M, p0/M, z22.h, z12.h\n" + ".inst 0xa040a6e9 // ldnt1h { z8.h-z11.h }, pn9.b/Z, [x23]\n" + ".inst 0x818d02c1 // bfmopa za1.s, p0/M, p0/M, z22.h, z13.h\n" + ".inst 0x818e02c2 // bfmopa za2.s, p0/M, p0/M, z22.h, z14.h\n" + ".inst 0x818f02c3 // bfmopa za3.s, p0/M, p0/M, z22.h, z15.h\n" + "ld1h { z22.h }, p0/Z, [x26, #1, MUL VL]\n" + ".inst 0x818403c0 // bfmopa za0.s, p0/M, p0/M, z30.h, z4.h\n" + ".inst 0xa041a6ed // ldnt1h { z12.h-z15.h }, pn9.b/Z, [x23, #0x4, MUL VL]\n" + ".inst 0x818503c1 // bfmopa za1.s, p0/M, p0/M, z30.h, z5.h\n" + ".inst 0x818603c2 // bfmopa za2.s, p0/M, p0/M, z30.h, z6.h\n" + ".inst 0x818703c3 // bfmopa za3.s, p0/M, p0/M, z30.h, z7.h\n" + "ld1h { z30.h }, p0/Z, [x26, #2, MUL VL]\n" + ".inst 0xa042a6e5 // ldnt1h { z4.h-z7.h }, pn9.b/Z, [x23, #0x8, MUL VL]\n" + ".inst 0x81930280 // bfmopa za0.s, p0/M, p0/M, z20.h, z19.h\n" + ".inst 0x81970281 // bfmopa za1.s, p0/M, p0/M, z20.h, z23.h\n" + ".inst 0x819b0282 // bfmopa za2.s, p0/M, p0/M, z20.h, z27.h\n" + ".inst 0x819f0283 // bfmopa za3.s, p0/M, p0/M, z20.h, z31.h\n" + "ld1h { z20.h }, p0/Z, [x26, #3, MUL VL]\n" + "addvl x26, x26, #4\n" + ".inst 0xa143a6fb // ldnt1h { z19.h, z23.h, z27.h, z31.h }, pn9.b/Z, [x23, #0xc, MUL VL]\n" + "addvl x23, x23, #16\n" + "bgt 6b\n" + "7:" // K loop tail + ".inst 0x81880380 // bfmopa za0.s, p0/M, p0/M, z28.h, z8.h\n" + ".inst 0x81890381 // bfmopa za1.s, p0/M, p0/M, z28.h, z9.h\n" + ".inst 0x818a0382 // bfmopa za2.s, p0/M, p0/M, z28.h, z10.h\n" + ".inst 0x818b0383 // bfmopa za3.s, p0/M, p0/M, z28.h, z11.h\n" + ".inst 0x818c02c0 // bfmopa za0.s, p0/M, p0/M, z22.h, z12.h\n" + ".inst 0x818d02c1 // bfmopa za1.s, p0/M, p0/M, z22.h, z13.h\n" + ".inst 0x818e02c2 // bfmopa za2.s, p0/M, p0/M, z22.h, z14.h\n" + ".inst 0x818f02c3 // bfmopa za3.s, p0/M, p0/M, z22.h, z15.h\n" + ".inst 0x818403c0 // bfmopa za0.s, p0/M, p0/M, z30.h, z4.h\n" + ".inst 0x818503c1 // bfmopa za1.s, p0/M, p0/M, z30.h, z5.h\n" + ".inst 0x818603c2 // bfmopa za2.s, p0/M, p0/M, z30.h, z6.h\n" + ".inst 0x818703c3 // bfmopa za3.s, p0/M, p0/M, z30.h, z7.h\n" + ".inst 0x81930280 // bfmopa za0.s, p0/M, p0/M, z20.h, z19.h\n" + ".inst 0x81970281 // bfmopa za1.s, p0/M, p0/M, z20.h, z23.h\n" + ".inst 0x819b0282 // bfmopa za2.s, p0/M, p0/M, z20.h, z27.h\n" + ".inst 0x819f0283 // bfmopa za3.s, p0/M, p0/M, z20.h, z31.h\n" + "8:" // K oddments + "cbz x21, 10f\n" + "9:" // K oddments: Loop + "ld1h { z8.h }, p0/Z, [x26]\n" + "subs x21, x21, #0x1\n" + "addvl x26, x26, #1\n" + ".inst 0xa140a6e3 // ld1h { z3.h, z7.h, z11.h, z15.h }, pn9.b/Z, [x23]\n" + "addvl x23, x23, #4\n" + ".inst 0x81830100 // bfmopa za0.s, p0/M, p0/M, z8.h, z3.h\n" + ".inst 0x81870101 // bfmopa za1.s, p0/M, p0/M, z8.h, z7.h\n" + ".inst 0x818b0102 // bfmopa za2.s, p0/M, p0/M, z8.h, z11.h\n" + ".inst 0x818f0103 // bfmopa za3.s, p0/M, p0/M, z8.h, z15.h\n" + "bgt 9b\n" + "10:" // K oddments: End + "tbz x15, #1, 14f\n" + "tbz x15, #0, 12f\n" + "mov x12, #0x0\n" + "cntw x20\n" + "11:" // Store to partial result buffer: Store and refill: Loop + ".inst 0xa040c5d4 // ld1w { z20.s-z23.s }, pn9.b/Z, [x14]\n" + ".inst 0xc0860408 // mova { z8.s-z11.s }, za0h.s[x12]\n" + ".inst 0xc0840680 // mova za0h.s[x12], { z20.s-z23.s }\n" + ".inst 0xc0860424 // mova { z4.s-z7.s }, za1h.s[x12]\n" + ".inst 0xa041c5cc // ld1w { z12.s-z15.s }, pn9.b/Z, [x14, #0x4, MUL VL]\n" + ".inst 0xc0840581 // mova za1h.s[x12], { z12.s-z15.s }\n" + ".inst 0xc086044c // mova { z12.s-z15.s }, za2h.s[x12]\n" + ".inst 0xc0860460 // mova { z0.s-z3.s }, za3h.s[x12]\n" + ".inst 0xa042c5d0 // ld1w { z16.s-z19.s }, pn9.b/Z, [x14, #0x8, MUL VL]\n" + ".inst 0xc0840602 // mova za2h.s[x12], { z16.s-z19.s }\n" + ".inst 0xa043c5d0 // ld1w { z16.s-z19.s }, pn9.b/Z, [x14, #0xc, MUL VL]\n" + ".inst 0xc0840603 // mova za3h.s[x12], { z16.s-z19.s }\n" + "add x12, x12, #0x4\n" + "cmp x12, x20\n" + ".inst 0xa060c5a8 // st1w { z8.s-z11.s }, pn9.b, [x13]\n" + "addvl x14, x14, #16\n" + ".inst 0xa061c5a4 // st1w { z4.s-z7.s }, pn9.b, [x13, #0x4, MUL VL]\n" + ".inst 0xa062c5ac // st1w { z12.s-z15.s }, pn9.b, [x13, #0x8, MUL VL]\n" + ".inst 0xa063c5a0 // st1w { z0.s-z3.s }, pn9.b, [x13, #0xc, MUL VL]\n" + "addvl x13, x13, #16\n" + "blt 11b\n" + "b 24f\n" + "12:" // Store to partial result buffer: Store only + "mov x12, #0x0\n" + "cntw x20\n" + "13:" // Store to partial result buffer: Store only: Loop + ".inst 0xc0860410 // mova { z16.s-z19.s }, za0h.s[x12]\n" + ".inst 0xc0860438 // mova { z24.s-z27.s }, za1h.s[x12]\n" + ".inst 0xa060c5b0 // st1w { z16.s-z19.s }, pn9.b, [x13]\n" + ".inst 0xc0860448 // mova { z8.s-z11.s }, za2h.s[x12]\n" + ".inst 0xc0860464 // mova { z4.s-z7.s }, za3h.s[x12]\n" + ".inst 0xa061c5b8 // st1w { z24.s-z27.s }, pn9.b, [x13, #0x4, MUL VL]\n" + "add x12, x12, #0x4\n" + "cmp x12, x20\n" + ".inst 0xa062c5a8 // st1w { z8.s-z11.s }, pn9.b, [x13, #0x8, MUL VL]\n" + ".inst 0xa063c5a4 // st1w { z4.s-z7.s }, pn9.b, [x13, #0xc, MUL VL]\n" + "addvl x13, x13, #16\n" + "blt 13b\n" + "b 24f\n" + "14:" // Store to output array + "ldr x25, [%x[args], %[offsetof_C]]\n" + "add x25, x25, x9, LSL #2\n" // C += n + "sub x24, x11, x10\n" + "ldr x23, [%x[args], %[offsetof_ldcb]]\n" + "madd x25, x10, x23, x25\n" // C += m * ldc + "tbz x15, #2, 18f\n" + "cntw x20\n" + "cmp x24, x20\n" + "csel x22, x24, x20, LT\n" + "lsr x21, x22, #0x2\n" + "mov x12, #0x0\n" + "and x20, x22, #0x3\n" + "cbz x21, 16f\n" + "15:" // Store to output array: Skip activation: Accumulator row 0 loop + ".inst 0xc0860400 // mova { z0.s-z3.s }, za0h.s[x12]\n" + ".inst 0xc0860424 // mova { z4.s-z7.s }, za1h.s[x12]\n" + ".inst 0xc0860448 // mova { z8.s-z11.s }, za2h.s[x12]\n" + ".inst 0xc086046c // mova { z12.s-z15.s }, za3h.s[x12]\n" + ".inst 0xa160c320 // st1w { z0.s, z4.s, z8.s, z12.s }, p8, [x25]\n" + "add x25, x25, x23\n" + ".inst 0xa160c321 // st1w { z1.s, z5.s, z9.s, z13.s }, p8, [x25]\n" + "add x25, x25, x23\n" + "add x12, x12, #0x4\n" + ".inst 0xa160c322 // st1w { z2.s, z6.s, z10.s, z14.s }, p8, [x25]\n" + "add x25, x25, x23\n" + "cmp x12, x21, LSL #2\n" + ".inst 0xa160c323 // st1w { z3.s, z7.s, z11.s, z15.s }, p8, [x25]\n" + "add x25, x25, x23\n" + "blt 15b\n" + "16:" // Store to output array: Skip activation: Accumulator row 0 oddments + "cbz x20, 17f\n" + "subs x20, x20, #0x1\n" + ".inst 0xc0860410 // mova { z16.s-z19.s }, za0h.s[x12]\n" + ".inst 0xc0860434 // mova { z20.s-z23.s }, za1h.s[x12]\n" + ".inst 0xc0860458 // mova { z24.s-z27.s }, za2h.s[x12]\n" + ".inst 0xc086047c // mova { z28.s-z31.s }, za3h.s[x12]\n" + ".inst 0xa160c330 // st1w { z16.s, z20.s, z24.s, z28.s }, p8, [x25]\n" + "add x25, x25, x23\n" + "beq 17f\n" + "subs x20, x20, #0x1\n" + ".inst 0xa160c331 // st1w { z17.s, z21.s, z25.s, z29.s }, p8, [x25]\n" + "add x25, x25, x23\n" + "beq 17f\n" + ".inst 0xa160c332 // st1w { z18.s, z22.s, z26.s, z30.s }, p8, [x25]\n" + "add x25, x25, x23\n" + "17:" // Store to output array: Skip activation: Accumulator row 0 oddments: End + "subs x24, x24, x22\n" + "beq 18f\n" + "b 22f\n" + "18:" // Store to output array: Skip activation: End + "cntw x20\n" + "cmp x24, x20\n" + "ld1rw { z1.s }, p0/Z, [%x[args], %[offsetof_KernelArgs_min]]\n" + "csel x20, x24, x20, LT\n" + "lsr x21, x20, #0x2\n" + "ld1rw { z0.s }, p0/Z, [%x[args], %[offsetof_KernelArgs_max]]\n" + "mov x12, #0x0\n" + "and x20, x20, #0x3\n" + "cbz x21, 20f\n" + "19:" // Store to output array: Accumulator row 0 loop + ".inst 0xc0860410 // mova { z16.s-z19.s }, za0h.s[x12]\n" + ".inst 0xc0860434 // mova { z20.s-z23.s }, za1h.s[x12]\n" + ".inst 0xc1a0c830 // fclamp { z16.s-z19.s }, z1.s, z0.s\n" + ".inst 0xc1a0c834 // fclamp { z20.s-z23.s }, z1.s, z0.s\n" + ".inst 0xc0860458 // mova { z24.s-z27.s }, za2h.s[x12]\n" + ".inst 0xc086047c // mova { z28.s-z31.s }, za3h.s[x12]\n" + ".inst 0xc1a0c838 // fclamp { z24.s-z27.s }, z1.s, z0.s\n" + ".inst 0xc1a0c83c // fclamp { z28.s-z31.s }, z1.s, z0.s\n" + ".inst 0xa160c330 // st1w { z16.s, z20.s, z24.s, z28.s }, p8, [x25]\n" + "add x25, x25, x23\n" + "add x12, x12, #0x4\n" + ".inst 0xa160c331 // st1w { z17.s, z21.s, z25.s, z29.s }, p8, [x25]\n" + "add x25, x25, x23\n" + "cmp x12, x21, LSL #2\n" + ".inst 0xa160c332 // st1w { z18.s, z22.s, z26.s, z30.s }, p8, [x25]\n" + "add x25, x25, x23\n" + ".inst 0xa160c333 // st1w { z19.s, z23.s, z27.s, z31.s }, p8, [x25]\n" + "add x25, x25, x23\n" + "blt 19b\n" + "20:" // Store to output array: Accumulator row 0 oddments + "cbz x20, 21f\n" + ".inst 0xc0860410 // mova { z16.s-z19.s }, za0h.s[x12]\n" + ".inst 0xc0860434 // mova { z20.s-z23.s }, za1h.s[x12]\n" + ".inst 0xc1a0c830 // fclamp { z16.s-z19.s }, z1.s, z0.s\n" + ".inst 0xc1a0c834 // fclamp { z20.s-z23.s }, z1.s, z0.s\n" + ".inst 0xc0860458 // mova { z24.s-z27.s }, za2h.s[x12]\n" + ".inst 0xc086047c // mova { z28.s-z31.s }, za3h.s[x12]\n" + ".inst 0xc1a0c838 // fclamp { z24.s-z27.s }, z1.s, z0.s\n" + ".inst 0xc1a0c83c // fclamp { z28.s-z31.s }, z1.s, z0.s\n" + "subs x20, x20, #0x1\n" + ".inst 0xa160c330 // st1w { z16.s, z20.s, z24.s, z28.s }, p8, [x25]\n" + "add x25, x25, x23\n" + "beq 21f\n" + "subs x20, x20, #0x1\n" + ".inst 0xa160c331 // st1w { z17.s, z21.s, z25.s, z29.s }, p8, [x25]\n" + "add x25, x25, x23\n" + "beq 21f\n" + ".inst 0xa160c332 // st1w { z18.s, z22.s, z26.s, z30.s }, p8, [x25]\n" + "21:" // Store to output array: Accumulator row 0 oddments: End + "22:" // Store to output array: End + "tbz x15, #0, 24f\n" + "mov x12, #0x0\n" + "cntw x20\n" + "23:" // Store to output array: Refill accumulators: Loop + ".inst 0xa040c5cc // ld1w { z12.s-z15.s }, pn9.b/Z, [x14]\n" + ".inst 0xc0840580 // mova za0h.s[x12], { z12.s-z15.s }\n" + ".inst 0xa041c5d0 // ld1w { z16.s-z19.s }, pn9.b/Z, [x14, #0x4, MUL VL]\n" + ".inst 0xc0840601 // mova za1h.s[x12], { z16.s-z19.s }\n" + ".inst 0xa042c5d0 // ld1w { z16.s-z19.s }, pn9.b/Z, [x14, #0x8, MUL VL]\n" + ".inst 0xc0840602 // mova za2h.s[x12], { z16.s-z19.s }\n" + ".inst 0xa043c5c4 // ld1w { z4.s-z7.s }, pn9.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, x20\n" + "addvl x14, x14, #16\n" + "blt 23b\n" + "24:" // End block + "incw x9, ALL, MUL #4\n" + "cmp x9, x28\n" + "blt 3b\n" + "incw x10\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_KernelArgs_max] "I" (offsetof(KernelArgs, max)), [offsetof_KernelArgs_min] "I" (offsetof(KernelArgs, min)), [offsetof_M] "I" (offsetof(KernelArgs, M)), [offsetof_N] "I" (offsetof(KernelArgs, N)), [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)) + : "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", "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 |