File size: 6,130 Bytes
8b7c501 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 |
// Copyright 2019 Google LLC
//
// This source code is licensed under the BSD-style license found in the
// LICENSE file in the root directory of this source tree.
$assert CHANNEL_TILE % 16 == 0
$assert KERNEL_TILE >= 2
$assert ACCUMULATORS >= 1
$ABC = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ"
#include <assert.h>
#include <immintrin.h>
#include <xnnpack/dwconv.h>
#include <xnnpack/intrinsics-polyfill.h>
void xnn_f32_dwconv_minmax_ukernel_${KERNEL_TILE}p${CHANNEL_TILE}c__avx512f${"" if ACCUMULATORS == 1 else "_acc%d" % ACCUMULATORS}(
size_t channels,
size_t output_width,
const float** input,
const float* weights,
float* output,
intptr_t input_stride,
size_t output_increment,
size_t input_offset,
const float* zero,
const union xnn_f32_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
{
assert(channels != 0);
assert(output_width != 0);
const __m512 vmin = _mm512_set1_ps(params->scalar.min);
const __m512 vmax = _mm512_set1_ps(params->scalar.max);
do {
$for K in range(KERNEL_TILE):
const float* i${K} = input[${K}];
assert(i${K} != NULL);
if XNN_UNPREDICTABLE(i${K} != zero) {
i${K} = (const float*) ((uintptr_t) i${K} + input_offset);
}
input = (const float**) ((uintptr_t) input + input_stride);
size_t c = channels;
const float* w = weights;
for (; c >= ${CHANNEL_TILE}; c -= ${CHANNEL_TILE}) {
__m512 vacc${ABC[0:16]}p0 = _mm512_load_ps(w);
$for C in range(16, CHANNEL_TILE, 16):
__m512 vacc${ABC[C:C+16]}p0 = _mm512_load_ps(w + ${C});
$for K in range(KERNEL_TILE):
const __m512 vi${K}x${ABC[0:16]} = _mm512_loadu_ps(i${K});
$for C in range(16, CHANNEL_TILE, 16):
const __m512 vi${K}x${ABC[C:C+16]} = _mm512_loadu_ps(i${K} + ${C});
i${K} += ${CHANNEL_TILE};
$for C in range(0, CHANNEL_TILE, 16):
const __m512 vk${K}x${ABC[C:C+16]} = _mm512_load_ps(w + ${(K + 1) * CHANNEL_TILE + C});
$for C in range(0, CHANNEL_TILE, 16):
$if 1 <= K < ACCUMULATORS:
__m512 vacc${ABC[C:C+16]}p${K} = _mm512_mul_ps(vi${K}x${ABC[C:C+16]}, vk${K}x${ABC[C:C+16]});
$else:
vacc${ABC[C:C+16]}p${K % ACCUMULATORS} = _mm512_fmadd_ps(vi${K}x${ABC[C:C+16]}, vk${K}x${ABC[C:C+16]}, vacc${ABC[C:C+16]}p${K % ACCUMULATORS});
w += ${(KERNEL_TILE + 1) * CHANNEL_TILE};
$if ACCUMULATORS > 1:
// Add up all accumulators to vacc${ABC[0:CHANNEL_TILE]}p0
$ACC_SLICE = 1
$while ACC_SLICE < ACCUMULATORS:
$for A in range(0, ACCUMULATORS, ACC_SLICE * 2):
$if A + ACC_SLICE < ACCUMULATORS:
$for C in range(0, CHANNEL_TILE, 16):
vacc${ABC[C:C+16]}p${A} = _mm512_add_ps(vacc${ABC[C:C+16]}p${A}, vacc${ABC[C:C+16]}p${A + ACC_SLICE});
$ACC_SLICE *= 2
$for C in range(0, CHANNEL_TILE, 16):
__m512 vacc${ABC[C:C+16]} = _mm512_max_ps(vmin, vacc${ABC[C:C+16]}p0);
$for C in range(0, CHANNEL_TILE, 16):
vacc${ABC[C:C+16]} = _mm512_min_ps(vmax, vacc${ABC[C:C+16]});
_mm512_storeu_ps(output, vacc${ABC[0:16]});
$for C in range(16, CHANNEL_TILE, 16):
_mm512_storeu_ps(output + ${C}, vacc${ABC[C:C+16]});
output += ${CHANNEL_TILE};
}
$if CHANNEL_TILE > 16:
for (; c >= 16; c -= 16) {
__m512 vacc${ABC[0:16]}p0 = _mm512_load_ps(w);
$for K in range(KERNEL_TILE):
const __m512 vi${K}x${ABC[0:16]} = _mm512_loadu_ps(i${K});
i${K} += 16;
const __m512 vk${K}x${ABC[0:16]} = _mm512_load_ps(w + ${(K + 1) * CHANNEL_TILE});
$if 1 <= K < ACCUMULATORS:
__m512 vacc${ABC[0:16]}p${K} = _mm512_mul_ps(vi${K}x${ABC[0:16]}, vk${K}x${ABC[0:16]});
$else:
vacc${ABC[0:16]}p${K % ACCUMULATORS} = _mm512_fmadd_ps(vi${K}x${ABC[0:16]}, vk${K}x${ABC[0:16]}, vacc${ABC[0:16]}p${K % ACCUMULATORS});
w += 16;
$if ACCUMULATORS > 1:
// Add up all accumulators to vacc${ABC[0:16]}p0
$ACC_SLICE = 1
$while ACC_SLICE < ACCUMULATORS:
$for A in range(0, ACCUMULATORS, ACC_SLICE * 2):
$if A + ACC_SLICE < ACCUMULATORS:
vacc${ABC[0:16]}p${A} = _mm512_add_ps(vacc${ABC[0:16]}p${A}, vacc${ABC[0:16]}p${A + ACC_SLICE});
$ACC_SLICE *= 2
__m512 vacc${ABC[0:16]} = _mm512_max_ps(vmin, vacc${ABC[0:16]}p0);
vacc${ABC[0:16]} = _mm512_min_ps(vmax, vacc${ABC[0:16]});
_mm512_storeu_ps(output, vacc${ABC[0:16]});
output += 16;
}
if XNN_UNLIKELY(c != 0) {
assert(c >= 1);
assert(c <= 16);
// Prepare mask for valid 32-bit elements (depends on nc).
const __mmask16 vmask = _cvtu32_mask16((uint16_t) ((uint32_t) (UINT32_C(1) << c) - UINT32_C(1)));
__m512 vacc${ABC[0:16]}p0 = _mm512_maskz_loadu_ps(vmask, w);
$for K in range(KERNEL_TILE):
const __m512 vi${K}x${ABC[0:16]} = _mm512_maskz_loadu_ps(vmask, i${K});
const __m512 vk${K}x${ABC[0:16]} = _mm512_maskz_loadu_ps(vmask, w + ${(K + 1) * CHANNEL_TILE});
$if 1 <= K < ACCUMULATORS:
__m512 vacc${ABC[0:16]}p${K} = _mm512_mul_ps(vi${K}x${ABC[0:16]}, vk${K}x${ABC[0:16]});
$else:
vacc${ABC[0:16]}p${K % ACCUMULATORS} = _mm512_fmadd_ps(vi${K}x${ABC[0:16]}, vk${K}x${ABC[0:16]}, vacc${ABC[0:16]}p${K % ACCUMULATORS});
$if ACCUMULATORS > 1:
// Add up all accumulators to vacc${ABC[0:16]}p0
$ACC_SLICE = 1
$while ACC_SLICE < ACCUMULATORS:
$for A in range(0, ACCUMULATORS, ACC_SLICE * 2):
$if A + ACC_SLICE < ACCUMULATORS:
vacc${ABC[0:16]}p${A} = _mm512_add_ps(vacc${ABC[0:16]}p${A}, vacc${ABC[0:16]}p${A + ACC_SLICE});
$ACC_SLICE *= 2
__m512 vacc${ABC[0:16]} = _mm512_max_ps(vmin, vacc${ABC[0:16]}p0);
vacc${ABC[0:16]} = _mm512_min_ps(vmax, vacc${ABC[0:16]});
_mm512_mask_storeu_ps(output, vmask, vacc${ABC[0:16]});
output += c;
}
output = (float*) ((uintptr_t) output + output_increment);
} while (--output_width != 0);
}
|