From aaa9da1efa83911c7a67d50811ad669a92a7d12f Mon Sep 17 00:00:00 2001 From: David Mansell Date: Fri, 10 Mar 2023 13:48:50 +0000 Subject: arm_gemm: Add SME2 FP16 kernels. Resolves: COMPMID-5966 Change-Id: Ic0d694493178da029a297643855bd0cff01b174f Signed-off-by: David Mansell Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9302 Benchmark: Arm Jenkins Tested-by: Arm Jenkins Reviewed-by: Viet-Hoa Do Comments-Addressed: Arm Jenkins --- .../generic.cpp | 506 +++++++++++++++++++++ 1 file changed, 506 insertions(+) create mode 100644 src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_fp16fp32fp16_mopa_4VLx1VL/generic.cpp (limited to 'src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_fp16fp32fp16_mopa_4VLx1VL/generic.cpp') diff --git a/src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_fp16fp32fp16_mopa_4VLx1VL/generic.cpp b/src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_fp16fp32fp16_mopa_4VLx1VL/generic.cpp new file mode 100644 index 0000000000..8728cff31d --- /dev/null +++ b/src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_fp16fp32fp16_mopa_4VLx1VL/generic.cpp @@ -0,0 +1,506 @@ +/* + * Copyright (c) 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_FEATURE_SVE + +#include "arm_gemm.hpp" + + +#include "../../asmlib.hpp" +#include "../../utils.hpp" + +namespace arm_gemm { + +void sme2_interleaved_nomerge_fp16fp32fp16_mopa_4VLx1VL(const __fp16 *const A, const __fp16 *const B, __fp16 *const C, int ldc, const int M, const int N, const int K, const __fp16 *const bias, const Activation act, bool accumulate, float *const accumulator_buffer) +{ + struct KernelArgs + { + KernelArgs( + const __fp16 *const A, + const __fp16 *const B, + __fp16 *const C, const int ldc, + const int M, const int N, const int K, + const __fp16 *const bias, + const Activation act, + bool accumulate, + float *const accumulator_buffer + ) : A(A), + B(B), kstride_bytes(roundup(K, 2) * sizeof(__fp16)), + C(C), ldcb(ldc * sizeof(__fp16)), + M(M), N(N), K(K), + min(-static_cast<__fp16>(std::numeric_limits::infinity())), + max(static_cast<__fp16>(std::numeric_limits::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 + } + + // Initialise the activation values + switch (act.type) + { + default: + case Activation::Type::None: + break; + case Activation::Type::BoundedReLU: + this->max = static_cast<__fp16>(act.param1); + /* fall through */ + case Activation::Type::ReLU: + this->min = static_cast<__fp16>(0); + break; + } + } + + const __fp16 *const A; + const __fp16 *const B; + const long kstride_bytes; + __fp16 *const C; + const long ldcb; + const long M, N, K; + __fp16 min = -static_cast<__fp16>(std::numeric_limits::infinity()); + __fp16 max = static_cast<__fp16>(std::numeric_limits::infinity()); + + const __fp16 *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 x16, [%x[args], %[offsetof_flags]]\n" + ".inst 0xd503477f // SMSTART ZA\n" + "ptrue p1.b\n" + ".inst 0x25207811 // ptrue pn9.b\n" + "ldr x15, [%x[args], %[offsetof_accumulator_buffer]]\n" + "ldr x14, [%x[args], %[offsetof_accumulator_buffer]]\n" + "tbz x16, #0, 2f\n" + "mov x12, #0x0\n" + "cntw x20\n" + "1:" // Initial accumulator load from buffer: Loop + ".inst 0xa040c5f8 // ld1w { z24.s-z27.s }, pn9.b/Z, [x15]\n" + ".inst 0xa041c5e4 // ld1w { z4.s-z7.s }, pn9.b/Z, [x15, #0x4, MUL VL]\n" + ".inst 0xa042c5e8 // ld1w { z8.s-z11.s }, pn9.b/Z, [x15, #0x8, MUL VL]\n" + ".inst 0xa043c5f4 // ld1w { z20.s-z23.s }, pn9.b/Z, [x15, #0xc, MUL VL]\n" + ".inst 0xc0840700 // mova za0h.s[x12], { z24.s-z27.s }\n" + "addvl x15, x15, #16\n" + ".inst 0xc0840481 // mova za1h.s[x12], { z4.s-z7.s }\n" + ".inst 0xc0840502 // mova za2h.s[x12], { z8.s-z11.s }\n" + ".inst 0xc0840683 // mova za3h.s[x12], { z20.s-z23.s }\n" + "add x12, x12, #0x4\n" + "cmp x12, x20\n" + "blt 1b\n" + "2:" // Initial accumulator load from buffer: End + "ldr w13, [%x[args], %[offsetof_M]]\n" + "mov x11, #0x0\n" + "mov x10, #0x0\n" + "ldr w9, [%x[args], %[offsetof_N]]\n" + "ldr x28, [%x[args], %[offsetof_A]]\n" + "3:" // M and N loop + "mov x27, x28\n" + "whilelt p8.s, x10, x9\n" + "tbnz x16, #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" + "whilelt p0.h, x10, x9\n" + "fmov z5.h, #0.0\n" + "fmov z18.h, #1.0\n" + "ld1h { z31.h }, p0/Z, [x20, x10, LSL #1]\n" + "zip1 z15.h, z31.h, z5.h\n" + ".inst 0x81af2640 // fmopa za0.s, p1/M, p1/M, z18.h, z15.h\n" + ".inst 0x81af2641 // fmopa za1.s, p1/M, p1/M, z18.h, z15.h\n" + ".inst 0x81af2642 // fmopa za2.s, p1/M, p1/M, z18.h, z15.h\n" + ".inst 0x81af2643 // fmopa za3.s, p1/M, p1/M, z18.h, z15.h\n" + "4:" // Prepare accumulators: Test for last block + "mov x20, x10\n" + "mov x21, x11\n" + "incw x20\n" + "incw x21, ALL, MUL #4\n" + "cmp x20, x9\n" + "mov x20, x16\n" + "csel x21, x11, x21, LT\n" + "bfm x16, XZR, #0x0, #0x0 // bfc x16, #0x0, #0x1\n" + "cmp x21, x13\n" + "csel x16, x20, x16, LT\n" + "5:" // Prepare accumulators: End + "ldr x20, [%x[args], %[offsetof_K]]\n" + "ldr x23, [%x[args], %[offsetof_B]]\n" + "ldr x22, [%x[args], %[offsetof_kstride_bytes]]\n" + "add x20, x20, #0x1\n" + "lsr x20, x20, #0x1\n" + "lsr x21, x20, #0x2\n" + "and x20, x20, #0x3\n" + "madd x23, x10, x22, x23\n" // bptr = B + n * kstride_bytes + "cbz x21, 8f\n" + "subs x21, x21, #0x1\n" + ".inst 0xa140a773 // ld1h { z19.h, z23.h, z27.h, z31.h }, pn9.b/Z, [x27]\n" + "ldnt1h { z17.h }, p1/Z, [x23]\n" + ".inst 0xa041a76c // ld1h { z12.h-z15.h }, pn9.b/Z, [x27, #0x4, MUL VL]\n" + "ldnt1h { z26.h }, p1/Z, [x23, #1, MUL VL]\n" + ".inst 0xa042a760 // ld1h { z0.h-z3.h }, pn9.b/Z, [x27, #0x8, MUL VL]\n" + "ldnt1h { z30.h }, p1/Z, [x23, #2, MUL VL]\n" + ".inst 0xa143a770 // ld1h { z16.h, z20.h, z24.h, z28.h }, pn9.b/Z, [x27, #0xc, MUL VL]\n" + "addvl x27, x27, #16\n" + "ldnt1h { z18.h }, p1/Z, [x23, #3, MUL VL]\n" + "addvl x23, x23, #4\n" + "ble 7f\n" + "6:" // K loop + ".inst 0x81b12660 // fmopa za0.s, p1/M, p1/M, z19.h, z17.h\n" + "subs x21, x21, #0x1\n" + ".inst 0x81b126e1 // fmopa za1.s, p1/M, p1/M, z23.h, z17.h\n" + ".inst 0x81b12762 // fmopa za2.s, p1/M, p1/M, z27.h, z17.h\n" + ".inst 0x81b127e3 // fmopa za3.s, p1/M, p1/M, z31.h, z17.h\n" + ".inst 0xa140a773 // ld1h { z19.h, z23.h, z27.h, z31.h }, pn9.b/Z, [x27]\n" + ".inst 0x81ba2580 // fmopa za0.s, p1/M, p1/M, z12.h, z26.h\n" + "ld1h { z17.h }, p1/Z, [x23]\n" + ".inst 0x81ba25a1 // fmopa za1.s, p1/M, p1/M, z13.h, z26.h\n" + ".inst 0x81ba25c2 // fmopa za2.s, p1/M, p1/M, z14.h, z26.h\n" + ".inst 0x81ba25e3 // fmopa za3.s, p1/M, p1/M, z15.h, z26.h\n" + ".inst 0xa041a76c // ld1h { z12.h-z15.h }, pn9.b/Z, [x27, #0x4, MUL VL]\n" + ".inst 0x81be2400 // fmopa za0.s, p1/M, p1/M, z0.h, z30.h\n" + "ld1h { z26.h }, p1/Z, [x23, #1, MUL VL]\n" + ".inst 0x81be2421 // fmopa za1.s, p1/M, p1/M, z1.h, z30.h\n" + ".inst 0x81be2442 // fmopa za2.s, p1/M, p1/M, z2.h, z30.h\n" + ".inst 0x81be2463 // fmopa za3.s, p1/M, p1/M, z3.h, z30.h\n" + ".inst 0xa042a760 // ld1h { z0.h-z3.h }, pn9.b/Z, [x27, #0x8, MUL VL]\n" + "ld1h { z30.h }, p1/Z, [x23, #2, MUL VL]\n" + ".inst 0x81b22600 // fmopa za0.s, p1/M, p1/M, z16.h, z18.h\n" + ".inst 0x81b22681 // fmopa za1.s, p1/M, p1/M, z20.h, z18.h\n" + ".inst 0x81b22702 // fmopa za2.s, p1/M, p1/M, z24.h, z18.h\n" + ".inst 0x81b22783 // fmopa za3.s, p1/M, p1/M, z28.h, z18.h\n" + ".inst 0xa143a770 // ld1h { z16.h, z20.h, z24.h, z28.h }, pn9.b/Z, [x27, #0xc, MUL VL]\n" + "addvl x27, x27, #16\n" + "ld1h { z18.h }, p1/Z, [x23, #3, MUL VL]\n" + "addvl x23, x23, #4\n" + "bgt 6b\n" + "7:" // K loop tail + ".inst 0x81b12660 // fmopa za0.s, p1/M, p1/M, z19.h, z17.h\n" + ".inst 0x81b126e1 // fmopa za1.s, p1/M, p1/M, z23.h, z17.h\n" + ".inst 0x81b12762 // fmopa za2.s, p1/M, p1/M, z27.h, z17.h\n" + ".inst 0x81b127e3 // fmopa za3.s, p1/M, p1/M, z31.h, z17.h\n" + ".inst 0x81ba2580 // fmopa za0.s, p1/M, p1/M, z12.h, z26.h\n" + ".inst 0x81ba25a1 // fmopa za1.s, p1/M, p1/M, z13.h, z26.h\n" + ".inst 0x81ba25c2 // fmopa za2.s, p1/M, p1/M, z14.h, z26.h\n" + ".inst 0x81ba25e3 // fmopa za3.s, p1/M, p1/M, z15.h, z26.h\n" + ".inst 0x81be2400 // fmopa za0.s, p1/M, p1/M, z0.h, z30.h\n" + ".inst 0x81be2421 // fmopa za1.s, p1/M, p1/M, z1.h, z30.h\n" + ".inst 0x81be2442 // fmopa za2.s, p1/M, p1/M, z2.h, z30.h\n" + ".inst 0x81be2463 // fmopa za3.s, p1/M, p1/M, z3.h, z30.h\n" + ".inst 0x81b22600 // fmopa za0.s, p1/M, p1/M, z16.h, z18.h\n" + ".inst 0x81b22681 // fmopa za1.s, p1/M, p1/M, z20.h, z18.h\n" + ".inst 0x81b22702 // fmopa za2.s, p1/M, p1/M, z24.h, z18.h\n" + ".inst 0x81b22783 // fmopa za3.s, p1/M, p1/M, z28.h, z18.h\n" + "8:" // K oddments + "cbz x20, 10f\n" + "9:" // K oddments: Loop + ".inst 0xa140a773 // ld1h { z19.h, z23.h, z27.h, z31.h }, pn9.b/Z, [x27]\n" + "subs x20, x20, #0x1\n" + "addvl x27, x27, #4\n" + "ld1h { z17.h }, p1/Z, [x23]\n" + "addvl x23, x23, #1\n" + ".inst 0x81b12660 // fmopa za0.s, p1/M, p1/M, z19.h, z17.h\n" + ".inst 0x81b126e1 // fmopa za1.s, p1/M, p1/M, z23.h, z17.h\n" + ".inst 0x81b12762 // fmopa za2.s, p1/M, p1/M, z27.h, z17.h\n" + ".inst 0x81b127e3 // fmopa za3.s, p1/M, p1/M, z31.h, z17.h\n" + "bgt 9b\n" + "10:" // K oddments: End + "tbz x16, #1, 14f\n" + "tbz x16, #0, 12f\n" + "mov x12, #0x0\n" + "cntw x20\n" + "11:" // Store to partial result buffer: Store and refill: Loop + ".inst 0xa040c5e0 // ld1w { z0.s-z3.s }, pn9.b/Z, [x15]\n" + ".inst 0xc0860404 // mova { z4.s-z7.s }, za0h.s[x12]\n" + ".inst 0xc0860434 // mova { z20.s-z23.s }, za1h.s[x12]\n" + ".inst 0xa041c5ec // ld1w { z12.s-z15.s }, pn9.b/Z, [x15, #0x4, MUL VL]\n" + ".inst 0xc086045c // mova { z28.s-z31.s }, za2h.s[x12]\n" + ".inst 0xc0860478 // mova { z24.s-z27.s }, za3h.s[x12]\n" + ".inst 0xa042c5e8 // ld1w { z8.s-z11.s }, pn9.b/Z, [x15, #0x8, MUL VL]\n" + ".inst 0xa043c5f0 // ld1w { z16.s-z19.s }, pn9.b/Z, [x15, #0xc, MUL VL]\n" + ".inst 0xc0840400 // mova za0h.s[x12], { z0.s-z3.s }\n" + "addvl x15, x15, #16\n" + ".inst 0xc0840581 // mova za1h.s[x12], { z12.s-z15.s }\n" + ".inst 0xa060c5c4 // st1w { z4.s-z7.s }, pn9.b, [x14]\n" + ".inst 0xc0840502 // mova za2h.s[x12], { z8.s-z11.s }\n" + ".inst 0xa061c5d4 // st1w { z20.s-z23.s }, pn9.b, [x14, #0x4, MUL VL]\n" + ".inst 0xc0840603 // mova za3h.s[x12], { z16.s-z19.s }\n" + "add x12, x12, #0x4\n" + ".inst 0xa062c5dc // st1w { z28.s-z31.s }, pn9.b, [x14, #0x8, MUL VL]\n" + "cmp x12, x20\n" + ".inst 0xa063c5d8 // st1w { z24.s-z27.s }, pn9.b, [x14, #0xc, MUL VL]\n" + "addvl x14, x14, #16\n" + "blt 11b\n" + "b 29f\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 0xc086041c // mova { z28.s-z31.s }, za0h.s[x12]\n" + ".inst 0xc0860438 // mova { z24.s-z27.s }, za1h.s[x12]\n" + ".inst 0xc086044c // mova { z12.s-z15.s }, za2h.s[x12]\n" + ".inst 0xc0860470 // mova { z16.s-z19.s }, za3h.s[x12]\n" + ".inst 0xa060c5dc // st1w { z28.s-z31.s }, pn9.b, [x14]\n" + "add x12, x12, #0x4\n" + ".inst 0xa061c5d8 // st1w { z24.s-z27.s }, pn9.b, [x14, #0x4, MUL VL]\n" + "cmp x12, x20\n" + ".inst 0xa062c5cc // st1w { z12.s-z15.s }, pn9.b, [x14, #0x8, MUL VL]\n" + ".inst 0xa063c5d0 // st1w { z16.s-z19.s }, pn9.b, [x14, #0xc, MUL VL]\n" + "addvl x14, x14, #16\n" + "blt 13b\n" + "b 29f\n" + "14:" // Store to output array + "ldr x26, [%x[args], %[offsetof_C]]\n" + "sub x25, x13, x11\n" + "cntw x24\n" + "ld1rh { z29.h }, p1/Z, [%x[args], %[offsetof_KernelArgs_min]]\n" + "ldr x23, [%x[args], %[offsetof_ldcb]]\n" + "whilelt p0.s, x10, x9\n" + "cmp x25, x24\n" + "ld1rh { z28.h }, p1/Z, [%x[args], %[offsetof_KernelArgs_max]]\n" + "csel x22, x25, x24, LT\n" + "mov x12, #0x0\n" + "add x26, x26, x10, LSL #1\n" // C += n + "lsr x21, x22, #0x2\n" + "madd x26, x11, x23, x26\n" // C += m * ldc + "and x20, x22, #0x3\n" + "cbz x21, 16f\n" + "15:" // Store to output array: Accumulator row 0 loop + ".inst 0xc0860400 // mova { z0.s-z3.s }, za0h.s[x12]\n" + "add x12, x12, #0x4\n" + "fcvt z0.h, p1/m, z0.s\n" + "fcvt z1.h, p1/m, z1.s\n" + "fcvt z2.h, p1/m, z2.s\n" + "fcvt z3.h, p1/m, z3.s\n" + "cmp x12, x21, LSL #2\n" + ".inst 0xc17ccba0 // fclamp { z0.h-z3.h }, z29.h, z28.h\n" + "st1h { z0.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "st1h { z1.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "st1h { z2.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "st1h { z3.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "blt 15b\n" + "16:" // Store to output array: Accumulator row 0 oddments + "cbz x20, 17f\n" + ".inst 0xc0860410 // mova { z16.s-z19.s }, za0h.s[x12]\n" + "subs x20, x20, #0x1\n" + "fcvt z16.h, p1/m, z16.s\n" + "fcvt z17.h, p1/m, z17.s\n" + "fcvt z18.h, p1/m, z18.s\n" + "fcvt z19.h, p1/m, z19.s\n" + ".inst 0xc17ccbb0 // fclamp { z16.h-z19.h }, z29.h, z28.h\n" + "st1h { z16.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "beq 17f\n" + "subs x20, x20, #0x1\n" + "st1h { z17.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "beq 17f\n" + "st1h { z18.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "17:" // Store to output array: Accumulator row 0 oddments: End + "subs x25, x25, x22\n" + "beq 27f\n" + "whilelt p0.s, x10, x9\n" + "cmp x25, x24\n" + "csel x22, x25, x24, LT\n" + "mov x12, #0x0\n" + "lsr x21, x22, #0x2\n" + "and x20, x22, #0x3\n" + "cbz x21, 19f\n" + "18:" // Store to output array: Accumulator row 1 loop + ".inst 0xc0860438 // mova { z24.s-z27.s }, za1h.s[x12]\n" + "add x12, x12, #0x4\n" + "fcvt z24.h, p1/m, z24.s\n" + "fcvt z25.h, p1/m, z25.s\n" + "fcvt z26.h, p1/m, z26.s\n" + "fcvt z27.h, p1/m, z27.s\n" + "cmp x12, x21, LSL #2\n" + ".inst 0xc17ccbb8 // fclamp { z24.h-z27.h }, z29.h, z28.h\n" + "st1h { z24.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "st1h { z25.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "st1h { z26.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "st1h { z27.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "blt 18b\n" + "19:" // Store to output array: Accumulator row 1 oddments + "cbz x20, 20f\n" + ".inst 0xc0860420 // mova { z0.s-z3.s }, za1h.s[x12]\n" + "subs x20, x20, #0x1\n" + "fcvt z0.h, p1/m, z0.s\n" + "fcvt z1.h, p1/m, z1.s\n" + "fcvt z2.h, p1/m, z2.s\n" + "fcvt z3.h, p1/m, z3.s\n" + ".inst 0xc17ccba0 // fclamp { z0.h-z3.h }, z29.h, z28.h\n" + "st1h { z0.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "beq 20f\n" + "subs x20, x20, #0x1\n" + "st1h { z1.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "beq 20f\n" + "st1h { z2.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "20:" // Store to output array: Accumulator row 1 oddments: End + "subs x25, x25, x22\n" + "beq 27f\n" + "whilelt p0.s, x10, x9\n" + "cmp x25, x24\n" + "csel x22, x25, x24, LT\n" + "mov x12, #0x0\n" + "lsr x21, x22, #0x2\n" + "and x20, x22, #0x3\n" + "cbz x21, 22f\n" + "21:" // Store to output array: Accumulator row 2 loop + ".inst 0xc0860454 // mova { z20.s-z23.s }, za2h.s[x12]\n" + "add x12, x12, #0x4\n" + "fcvt z20.h, p1/m, z20.s\n" + "fcvt z21.h, p1/m, z21.s\n" + "fcvt z22.h, p1/m, z22.s\n" + "fcvt z23.h, p1/m, z23.s\n" + "cmp x12, x21, LSL #2\n" + ".inst 0xc17ccbb4 // fclamp { z20.h-z23.h }, z29.h, z28.h\n" + "st1h { z20.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "st1h { z21.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "st1h { z22.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "st1h { z23.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "blt 21b\n" + "22:" // Store to output array: Accumulator row 2 oddments + "cbz x20, 23f\n" + ".inst 0xc086044c // mova { z12.s-z15.s }, za2h.s[x12]\n" + "subs x20, x20, #0x1\n" + "fcvt z12.h, p1/m, z12.s\n" + "fcvt z13.h, p1/m, z13.s\n" + "fcvt z14.h, p1/m, z14.s\n" + "fcvt z15.h, p1/m, z15.s\n" + ".inst 0xc17ccbac // fclamp { z12.h-z15.h }, z29.h, z28.h\n" + "st1h { z12.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "beq 23f\n" + "subs x20, x20, #0x1\n" + "st1h { z13.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "beq 23f\n" + "st1h { z14.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "23:" // Store to output array: Accumulator row 2 oddments: End + "subs x25, x25, x22\n" + "beq 27f\n" + "whilelt p0.s, x10, x9\n" + "cmp x25, x24\n" + "csel x20, x25, x24, LT\n" + "mov x12, #0x0\n" + "lsr x21, x20, #0x2\n" + "and x20, x20, #0x3\n" + "cbz x21, 25f\n" + "24:" // Store to output array: Accumulator row 3 loop + ".inst 0xc0860464 // mova { z4.s-z7.s }, za3h.s[x12]\n" + "add x12, x12, #0x4\n" + "fcvt z4.h, p1/m, z4.s\n" + "fcvt z5.h, p1/m, z5.s\n" + "fcvt z6.h, p1/m, z6.s\n" + "fcvt z7.h, p1/m, z7.s\n" + "cmp x12, x21, LSL #2\n" + ".inst 0xc17ccba4 // fclamp { z4.h-z7.h }, z29.h, z28.h\n" + "st1h { z4.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "st1h { z5.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "st1h { z6.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "st1h { z7.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "blt 24b\n" + "25:" // Store to output array: Accumulator row 3 oddments + "cbz x20, 26f\n" + ".inst 0xc0860464 // mova { z4.s-z7.s }, za3h.s[x12]\n" + "subs x20, x20, #0x1\n" + "fcvt z4.h, p1/m, z4.s\n" + "fcvt z5.h, p1/m, z5.s\n" + "fcvt z6.h, p1/m, z6.s\n" + "fcvt z7.h, p1/m, z7.s\n" + ".inst 0xc17ccba4 // fclamp { z4.h-z7.h }, z29.h, z28.h\n" + "st1h { z4.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "beq 26f\n" + "subs x20, x20, #0x1\n" + "st1h { z5.s }, p0, [x26]\n" + "add x26, x26, x23\n" + "beq 26f\n" + "st1h { z6.s }, p0, [x26]\n" + "26:" // Store to output array: Accumulator row 3 oddments: End + "27:" // Store to output array: End + "tbz x16, #0, 29f\n" + "mov x12, #0x0\n" + "cntw x20\n" + "28:" // Store to output array: Refill accumulators: Loop + ".inst 0xa040c5e0 // ld1w { z0.s-z3.s }, pn9.b/Z, [x15]\n" + ".inst 0xa041c5ec // ld1w { z12.s-z15.s }, pn9.b/Z, [x15, #0x4, MUL VL]\n" + ".inst 0xa042c5e4 // ld1w { z4.s-z7.s }, pn9.b/Z, [x15, #0x8, MUL VL]\n" + ".inst 0xa043c5f4 // ld1w { z20.s-z23.s }, pn9.b/Z, [x15, #0xc, MUL VL]\n" + ".inst 0xc0840400 // mova za0h.s[x12], { z0.s-z3.s }\n" + "addvl x15, x15, #16\n" + ".inst 0xc0840581 // mova za1h.s[x12], { z12.s-z15.s }\n" + ".inst 0xc0840482 // mova za2h.s[x12], { z4.s-z7.s }\n" + ".inst 0xc0840683 // mova za3h.s[x12], { z20.s-z23.s }\n" + "add x12, x12, #0x4\n" + "cmp x12, x20\n" + "blt 28b\n" + "29:" // End block + "incw x10\n" + "cmp x10, x9\n" + "blt 3b\n" + "incw x11, ALL, MUL #4\n" + "mov x10, #0x0\n" + "cmp x11, x13\n" + "mov x28, x27\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", "x16", "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_FEATURE_SVE -- cgit v1.2.1