diff options
Diffstat (limited to 'src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_s8s32_mopa_1VLx4VL/generic.cpp')
-rw-r--r-- | src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_s8s32_mopa_1VLx4VL/generic.cpp | 341 |
1 files changed, 341 insertions, 0 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_s8s32_mopa_1VLx4VL/generic.cpp b/src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_s8s32_mopa_1VLx4VL/generic.cpp new file mode 100644 index 0000000000..d11faa634d --- /dev/null +++ b/src/core/NEON/kernels/arm_gemm/kernels/sme2_interleaved_nomerge_s8s32_mopa_1VLx4VL/generic.cpp @@ -0,0 +1,341 @@ +/* + * 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 <cstdint> +#include "../../asmlib.hpp" +#include "../../utils.hpp" + +namespace arm_gemm { + +void sme2_interleaved_nomerge_s8s32_mopa_1VLx4VL(const int8_t *const A, const int8_t *const B, int32_t *const C, int ldc, const int M, const int N, const int K, const int32_t *const bias, const Activation, bool accumulate, int32_t *const accumulator_buffer) +{ + struct KernelArgs + { + KernelArgs( + const int8_t *const A, + const int8_t *const B, + int32_t *const C, const int ldc, + const int M, const int N, const int K, + const int32_t *const bias, + + bool accumulate, + int32_t *const accumulator_buffer + ) : A(A), + B(B), kstride_bytes(roundup(K, 4) * sizeof(int8_t)), + C(C), ldcb(ldc * sizeof(int32_t)), + M(M), N(N), K(K), + n_loops(((K / 4) - 1) / 2), n_tail_iters(((K / 4) - 1) % 2), + + 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 + } + } + + const int8_t *const A; + const int8_t *const B; + const long kstride_bytes; + int32_t *const C; + const long ldcb; + const long M, N, K, n_loops, n_tail_iters; + + const int32_t *const bias; + + int32_t *const accumulator_buffer; + uint64_t flags; + }; + + // Construct arguments for this kernel + KernelArgs args(A, B, C, ldc, M, N, K, bias, accumulate, accumulator_buffer); + + __asm__ __volatile__( + "ldr x13, [%x[args], %[offsetof_flags]]\n" + ".inst 0xd503477f // SMSTART ZA\n" + "ptrue p0.b\n" + ".inst 0x25207811 // ptrue pn9.b\n" + "ldr x11, [%x[args], %[offsetof_accumulator_buffer]]\n" + "ldr x10, [%x[args], %[offsetof_accumulator_buffer]]\n" + "tbz x13, #0, 2f\n" + "mov x12, #0x0\n" + "cntw x20\n" + "1:" // Initial accumulator load from buffer: Loop + ".inst 0xa040c564 // ld1w { z4.s-z7.s }, pn9.b/Z, [x11]\n" + ".inst 0xc0840480 // mova za0h.s[x12], { z4.s-z7.s }\n" + ".inst 0xa041c56c // ld1w { z12.s-z15.s }, pn9.b/Z, [x11, #0x4, MUL VL]\n" + ".inst 0xc0840581 // mova za1h.s[x12], { z12.s-z15.s }\n" + ".inst 0xa042c57c // ld1w { z28.s-z31.s }, pn9.b/Z, [x11, #0x8, MUL VL]\n" + ".inst 0xc0840782 // mova za2h.s[x12], { z28.s-z31.s }\n" + ".inst 0xa043c574 // ld1w { z20.s-z23.s }, pn9.b/Z, [x11, #0xc, MUL VL]\n" + ".inst 0xc0840683 // mova za3h.s[x12], { z20.s-z23.s }\n" + "add x12, x12, #0x4\n" + "cmp x12, x20\n" + "addvl x11, x11, #16\n" + "blt 1b\n" + "2:" // Initial accumulator load from buffer: End + "ldr w9, [%x[args], %[offsetof_M]]\n" + "mov x28, #0x0\n" + "mov x27, #0x0\n" + "ldr w26, [%x[args], %[offsetof_N]]\n" + "ldr x25, [%x[args], %[offsetof_A]]\n" + "3:" // M and N loop + "mov x24, x25\n" + ".inst 0x25ba6770 // whilelt pn8.s, x27, x26, VLx4\n" + "tbnz x13, #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" + ".inst 0xa11bc29b // ldnt1w { z19.s, z23.s, z27.s, z31.s }, p8/Z, [x20, x27, LSL #2]\n" + ".inst 0xc0900260 // addha za0.s, p0/M, p0/M, z19.s\n" + ".inst 0xc09002e1 // addha za1.s, p0/M, p0/M, z23.s\n" + ".inst 0xc0900362 // addha za2.s, p0/M, p0/M, z27.s\n" + ".inst 0xc09003e3 // addha za3.s, p0/M, p0/M, z31.s\n" + "4:" // Prepare accumulators: Test for last block + "mov x20, x27\n" + "mov x21, x28\n" + "incw x20, ALL, MUL #4\n" + "incw x21\n" + "cmp x20, x26\n" + "csel x21, x28, x21, LT\n" + "mov x20, x13\n" + "bfm x13, XZR, #0x0, #0x0 // bfc x13, #0x0, #0x1\n" + "cmp x21, x9\n" + "csel x13, x20, x13, LT\n" + "5:" // Prepare accumulators: End + "ldr x20, [%x[args], %[offsetof_K]]\n" + "add x20, x20, #0x3\n" + "lsr x20, x20, #0x2\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, x27, x20, x23\n" // bptr = B + n * kstride_bytes + "cbz x22, 8f\n" + "subs x22, x22, #0x1\n" + "ld1b { z30.b }, p0/Z, [x24]\n" + ".inst 0xa04086e1 // ldnt1b { z0.b-z3.b }, pn9.b/Z, [x23]\n" + "ld1b { z21.b }, p0/Z, [x24, #1, MUL VL]\n" + ".inst 0xa04186f9 // ldnt1b { z24.b-z27.b }, pn9.b/Z, [x23, #0x4, MUL VL]\n" + "ld1b { z28.b }, p0/Z, [x24, #2, MUL VL]\n" + ".inst 0xa04286e5 // ldnt1b { z4.b-z7.b }, pn9.b/Z, [x23, #0x8, MUL VL]\n" + "ld1b { z11.b }, p0/Z, [x24, #3, MUL VL]\n" + "addvl x24, x24, #4\n" + ".inst 0xa04386f1 // ldnt1b { z16.b-z19.b }, pn9.b/Z, [x23, #0xc, MUL VL]\n" + "addvl x23, x23, #16\n" + "ble 7f\n" + "6:" // K loop + ".inst 0xa08003c0 // smopa za0.s, p0/M, p0/M, z30.b, z0.b\n" + "subs x22, x22, #0x1\n" + ".inst 0xa08103c1 // smopa za1.s, p0/M, p0/M, z30.b, z1.b\n" + ".inst 0xa08203c2 // smopa za2.s, p0/M, p0/M, z30.b, z2.b\n" + ".inst 0xa08303c3 // smopa za3.s, p0/M, p0/M, z30.b, z3.b\n" + "ld1b { z30.b }, p0/Z, [x24]\n" + ".inst 0xa09802a0 // smopa za0.s, p0/M, p0/M, z21.b, z24.b\n" + ".inst 0xa04086e1 // ldnt1b { z0.b-z3.b }, pn9.b/Z, [x23]\n" + ".inst 0xa09902a1 // smopa za1.s, p0/M, p0/M, z21.b, z25.b\n" + ".inst 0xa09a02a2 // smopa za2.s, p0/M, p0/M, z21.b, z26.b\n" + ".inst 0xa09b02a3 // smopa za3.s, p0/M, p0/M, z21.b, z27.b\n" + "ld1b { z21.b }, p0/Z, [x24, #1, MUL VL]\n" + ".inst 0xa0840380 // smopa za0.s, p0/M, p0/M, z28.b, z4.b\n" + ".inst 0xa04186f9 // ldnt1b { z24.b-z27.b }, pn9.b/Z, [x23, #0x4, MUL VL]\n" + ".inst 0xa0850381 // smopa za1.s, p0/M, p0/M, z28.b, z5.b\n" + ".inst 0xa0860382 // smopa za2.s, p0/M, p0/M, z28.b, z6.b\n" + ".inst 0xa0870383 // smopa za3.s, p0/M, p0/M, z28.b, z7.b\n" + "ld1b { z28.b }, p0/Z, [x24, #2, MUL VL]\n" + ".inst 0xa04286e5 // ldnt1b { z4.b-z7.b }, pn9.b/Z, [x23, #0x8, MUL VL]\n" + ".inst 0xa0900160 // smopa za0.s, p0/M, p0/M, z11.b, z16.b\n" + ".inst 0xa0910161 // smopa za1.s, p0/M, p0/M, z11.b, z17.b\n" + ".inst 0xa0920162 // smopa za2.s, p0/M, p0/M, z11.b, z18.b\n" + ".inst 0xa0930163 // smopa za3.s, p0/M, p0/M, z11.b, z19.b\n" + "ld1b { z11.b }, p0/Z, [x24, #3, MUL VL]\n" + "addvl x24, x24, #4\n" + ".inst 0xa04386f1 // ldnt1b { z16.b-z19.b }, pn9.b/Z, [x23, #0xc, MUL VL]\n" + "addvl x23, x23, #16\n" + "bgt 6b\n" + "7:" // K loop tail + ".inst 0xa08003c0 // smopa za0.s, p0/M, p0/M, z30.b, z0.b\n" + ".inst 0xa08103c1 // smopa za1.s, p0/M, p0/M, z30.b, z1.b\n" + ".inst 0xa08203c2 // smopa za2.s, p0/M, p0/M, z30.b, z2.b\n" + ".inst 0xa08303c3 // smopa za3.s, p0/M, p0/M, z30.b, z3.b\n" + ".inst 0xa09802a0 // smopa za0.s, p0/M, p0/M, z21.b, z24.b\n" + ".inst 0xa09902a1 // smopa za1.s, p0/M, p0/M, z21.b, z25.b\n" + ".inst 0xa09a02a2 // smopa za2.s, p0/M, p0/M, z21.b, z26.b\n" + ".inst 0xa09b02a3 // smopa za3.s, p0/M, p0/M, z21.b, z27.b\n" + ".inst 0xa0840380 // smopa za0.s, p0/M, p0/M, z28.b, z4.b\n" + ".inst 0xa0850381 // smopa za1.s, p0/M, p0/M, z28.b, z5.b\n" + ".inst 0xa0860382 // smopa za2.s, p0/M, p0/M, z28.b, z6.b\n" + ".inst 0xa0870383 // smopa za3.s, p0/M, p0/M, z28.b, z7.b\n" + ".inst 0xa0900160 // smopa za0.s, p0/M, p0/M, z11.b, z16.b\n" + ".inst 0xa0910161 // smopa za1.s, p0/M, p0/M, z11.b, z17.b\n" + ".inst 0xa0920162 // smopa za2.s, p0/M, p0/M, z11.b, z18.b\n" + ".inst 0xa0930163 // smopa za3.s, p0/M, p0/M, z11.b, z19.b\n" + "8:" // K oddments + "cbz x21, 10f\n" + "9:" // K oddments: Loop + "ld1b { z22.b }, p0/Z, [x24]\n" + "subs x21, x21, #0x1\n" + "addvl x24, x24, #1\n" + ".inst 0xa14086f1 // ld1b { z17.b, z21.b, z25.b, z29.b }, pn9.b/Z, [x23]\n" + "addvl x23, x23, #4\n" + ".inst 0xa09102c0 // smopa za0.s, p0/M, p0/M, z22.b, z17.b\n" + ".inst 0xa09502c1 // smopa za1.s, p0/M, p0/M, z22.b, z21.b\n" + ".inst 0xa09902c2 // smopa za2.s, p0/M, p0/M, z22.b, z25.b\n" + ".inst 0xa09d02c3 // smopa za3.s, p0/M, p0/M, z22.b, z29.b\n" + "bgt 9b\n" + "10:" // K oddments: End + "tbz x13, #1, 14f\n" + "tbz x13, #0, 12f\n" + "mov x12, #0x0\n" + "cntw x20\n" + "11:" // Store to partial result buffer: Store and refill: Loop + ".inst 0xa040c570 // ld1w { z16.s-z19.s }, pn9.b/Z, [x11]\n" + ".inst 0xc0860404 // mova { z4.s-z7.s }, za0h.s[x12]\n" + ".inst 0xc0840600 // mova za0h.s[x12], { z16.s-z19.s }\n" + ".inst 0xc086042c // mova { z12.s-z15.s }, za1h.s[x12]\n" + ".inst 0xa041c570 // ld1w { z16.s-z19.s }, pn9.b/Z, [x11, #0x4, MUL VL]\n" + ".inst 0xc0840601 // mova za1h.s[x12], { z16.s-z19.s }\n" + ".inst 0xc086045c // mova { z28.s-z31.s }, za2h.s[x12]\n" + ".inst 0xc0860470 // mova { z16.s-z19.s }, za3h.s[x12]\n" + ".inst 0xa042c560 // ld1w { z0.s-z3.s }, pn9.b/Z, [x11, #0x8, MUL VL]\n" + ".inst 0xc0840402 // mova za2h.s[x12], { z0.s-z3.s }\n" + ".inst 0xa043c574 // ld1w { z20.s-z23.s }, pn9.b/Z, [x11, #0xc, MUL VL]\n" + ".inst 0xc0840683 // mova za3h.s[x12], { z20.s-z23.s }\n" + "add x12, x12, #0x4\n" + "cmp x12, x20\n" + ".inst 0xa060c544 // st1w { z4.s-z7.s }, pn9.b, [x10]\n" + "addvl x11, x11, #16\n" + ".inst 0xa061c54c // st1w { z12.s-z15.s }, pn9.b, [x10, #0x4, MUL VL]\n" + ".inst 0xa062c55c // st1w { z28.s-z31.s }, pn9.b, [x10, #0x8, MUL VL]\n" + ".inst 0xa063c550 // st1w { z16.s-z19.s }, pn9.b, [x10, #0xc, MUL VL]\n" + "addvl x10, x10, #16\n" + "blt 11b\n" + "b 20f\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 0xc086040c // mova { z12.s-z15.s }, za0h.s[x12]\n" + ".inst 0xc0860424 // mova { z4.s-z7.s }, za1h.s[x12]\n" + ".inst 0xa060c54c // st1w { z12.s-z15.s }, pn9.b, [x10]\n" + ".inst 0xc0860440 // mova { z0.s-z3.s }, za2h.s[x12]\n" + ".inst 0xc0860478 // mova { z24.s-z27.s }, za3h.s[x12]\n" + ".inst 0xa061c544 // st1w { z4.s-z7.s }, pn9.b, [x10, #0x4, MUL VL]\n" + "add x12, x12, #0x4\n" + "cmp x12, x20\n" + ".inst 0xa062c540 // st1w { z0.s-z3.s }, pn9.b, [x10, #0x8, MUL VL]\n" + ".inst 0xa063c558 // st1w { z24.s-z27.s }, pn9.b, [x10, #0xc, MUL VL]\n" + "addvl x10, x10, #16\n" + "blt 13b\n" + "b 20f\n" + "14:" // Store to output array + "ldr x23, [%x[args], %[offsetof_C]]\n" + "sub x21, x9, x28\n" + "cntw x20\n" + "ldr x22, [%x[args], %[offsetof_ldcb]]\n" + "cmp x21, x20\n" + "csel x20, x21, x20, LT\n" + "add x23, x23, x27, LSL #2\n" // C += n + "lsr x21, x20, #0x2\n" + "madd x23, x28, x22, x23\n" // C += m * ldc + "mov x12, #0x0\n" + "and x20, x20, #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" + ".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 0xa160c2e0 // st1w { z0.s, z4.s, z8.s, z12.s }, p8, [x23]\n" + "add x23, x23, x22\n" + ".inst 0xa160c2e1 // st1w { z1.s, z5.s, z9.s, z13.s }, p8, [x23]\n" + "add x23, x23, x22\n" + "add x12, x12, #0x4\n" + ".inst 0xa160c2e2 // st1w { z2.s, z6.s, z10.s, z14.s }, p8, [x23]\n" + "add x23, x23, x22\n" + "cmp x12, x21, LSL #2\n" + ".inst 0xa160c2e3 // st1w { z3.s, z7.s, z11.s, z15.s }, p8, [x23]\n" + "add x23, x23, x22\n" + "blt 15b\n" + "16:" // Store to output array: 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 0xa160c2f0 // st1w { z16.s, z20.s, z24.s, z28.s }, p8, [x23]\n" + "add x23, x23, x22\n" + "beq 17f\n" + "subs x20, x20, #0x1\n" + ".inst 0xa160c2f1 // st1w { z17.s, z21.s, z25.s, z29.s }, p8, [x23]\n" + "add x23, x23, x22\n" + "beq 17f\n" + ".inst 0xa160c2f2 // st1w { z18.s, z22.s, z26.s, z30.s }, p8, [x23]\n" + "17:" // Store to output array: Accumulator row 0 oddments: End + "18:" // Store to output array: End + "tbz x13, #0, 20f\n" + "mov x12, #0x0\n" + "cntw x20\n" + "19:" // Store to output array: Refill accumulators: Loop + ".inst 0xa040c568 // ld1w { z8.s-z11.s }, pn9.b/Z, [x11]\n" + ".inst 0xc0840500 // mova za0h.s[x12], { z8.s-z11.s }\n" + ".inst 0xa041c570 // ld1w { z16.s-z19.s }, pn9.b/Z, [x11, #0x4, MUL VL]\n" + ".inst 0xc0840601 // mova za1h.s[x12], { z16.s-z19.s }\n" + ".inst 0xa042c570 // ld1w { z16.s-z19.s }, pn9.b/Z, [x11, #0x8, MUL VL]\n" + ".inst 0xc0840602 // mova za2h.s[x12], { z16.s-z19.s }\n" + ".inst 0xa043c568 // ld1w { z8.s-z11.s }, pn9.b/Z, [x11, #0xc, MUL VL]\n" + ".inst 0xc0840503 // mova za3h.s[x12], { z8.s-z11.s }\n" + "add x12, x12, #0x4\n" + "cmp x12, x20\n" + "addvl x11, x11, #16\n" + "blt 19b\n" + "20:" // End block + "incw x27, ALL, MUL #4\n" + "cmp x27, x26\n" + "blt 3b\n" + "incw x28\n" + "cmp x28, x9\n" + "mov x27, #0x0\n" + "mov x25, x24\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_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", "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 |