/external/XNNPACK/src/qs8-gemm/gen/ |
D | 3x8c8-minmax-neon-mlal-padal.c | 93 const int8x8_t vb2x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal() local 121 int16x8_t vprod0x2 = vmull_s8(vb2x0, va0x0); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal() 122 int16x8_t vprod1x2 = vmull_s8(vb2x0, va1x0); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal() 123 int16x8_t vprod2x2 = vmull_s8(vb2x0, va2x0); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal()
|
D | 4x8c8-minmax-neon-mlal-padal.c | 109 const int8x8_t vb2x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() local 143 int16x8_t vprod0x2 = vmull_s8(vb2x0, va0x0); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() 144 int16x8_t vprod1x2 = vmull_s8(vb2x0, va1x0); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() 145 int16x8_t vprod2x2 = vmull_s8(vb2x0, va2x0); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() 146 int16x8_t vprod3x2 = vmull_s8(vb2x0, va3x0); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal()
|
D | 2x8c8-minmax-neon-mlal-padal.c | 77 const int8x8_t vb2x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mlal_padal() local 99 int16x8_t vprod0x2 = vmull_s8(vb2x0, va0x0); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mlal_padal() 100 int16x8_t vprod1x2 = vmull_s8(vb2x0, va1x0); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mlal_padal()
|
D | 1x8c8-minmax-neon-mlal-padal.c | 61 const int8x8_t vb2x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mlal_padal() local 77 int16x8_t vprod0x2 = vmull_s8(vb2x0, va0x0); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mlal_padal()
|
D | 1x16c8-minmax-neon-mlal-padal.c | 69 const int8x8_t vb2x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mlal_padal() local 93 int16x8_t vprod0x2 = vmull_s8(vb2x0, va0x0); in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mlal_padal()
|
D | 2x16c8-minmax-neon-mlal-padal.c | 93 const int8x8_t vb2x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal() local 123 int16x8_t vprod0x2 = vmull_s8(vb2x0, va0x0); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal() 124 int16x8_t vprod1x2 = vmull_s8(vb2x0, va1x0); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal()
|
D | 3x16c8-minmax-neon-mlal-padal.c | 117 const int8x8_t vb2x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal() local 153 int16x8_t vprod0x2 = vmull_s8(vb2x0, va0x0); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal() 154 int16x8_t vprod1x2 = vmull_s8(vb2x0, va1x0); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal() 155 int16x8_t vprod2x2 = vmull_s8(vb2x0, va2x0); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal()
|
D | 4x16c8-minmax-neon-mlal-padal.c | 141 const int8x8_t vb2x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal() local 183 int16x8_t vprod0x2 = vmull_s8(vb2x0, va0x0); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal() 184 int16x8_t vprod1x2 = vmull_s8(vb2x0, va1x0); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal() 185 int16x8_t vprod2x2 = vmull_s8(vb2x0, va2x0); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal() 186 int16x8_t vprod3x2 = vmull_s8(vb2x0, va3x0); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal()
|
/external/XNNPACK/src/qs8-igemm/gen/ |
D | 4x8c8-minmax-neon-mlal-padal.c | 126 const int8x8_t vb2x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mlal_padal() local 160 int16x8_t vprod0x2 = vmull_s8(vb2x0, va0x0); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mlal_padal() 161 int16x8_t vprod1x2 = vmull_s8(vb2x0, va1x0); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mlal_padal() 162 int16x8_t vprod2x2 = vmull_s8(vb2x0, va2x0); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mlal_padal() 163 int16x8_t vprod3x2 = vmull_s8(vb2x0, va3x0); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mlal_padal()
|
D | 2x8c8-minmax-neon-mlal-padal.c | 90 const int8x8_t vb2x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mlal_padal() local 112 int16x8_t vprod0x2 = vmull_s8(vb2x0, va0x0); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mlal_padal() 113 int16x8_t vprod1x2 = vmull_s8(vb2x0, va1x0); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mlal_padal()
|
D | 3x8c8-minmax-neon-mlal-padal.c | 108 const int8x8_t vb2x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal() local 136 int16x8_t vprod0x2 = vmull_s8(vb2x0, va0x0); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal() 137 int16x8_t vprod1x2 = vmull_s8(vb2x0, va1x0); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal() 138 int16x8_t vprod2x2 = vmull_s8(vb2x0, va2x0); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal()
|
D | 1x8c8-minmax-neon-mlal-padal.c | 72 const int8x8_t vb2x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_igemm_minmax_ukernel_1x8c8__neon_mlal_padal() local 88 int16x8_t vprod0x2 = vmull_s8(vb2x0, va0x0); in xnn_qs8_igemm_minmax_ukernel_1x8c8__neon_mlal_padal()
|
D | 1x16c8-minmax-neon-mlal-padal.c | 80 const int8x8_t vb2x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mlal_padal() local 104 int16x8_t vprod0x2 = vmull_s8(vb2x0, va0x0); in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mlal_padal()
|
D | 2x16c8-minmax-neon-mlal-padal.c | 106 const int8x8_t vb2x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal() local 136 int16x8_t vprod0x2 = vmull_s8(vb2x0, va0x0); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal() 137 int16x8_t vprod1x2 = vmull_s8(vb2x0, va1x0); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal()
|
D | 4x16c8-minmax-neon-mlal-padal.c | 158 const int8x8_t vb2x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal() local 200 int16x8_t vprod0x2 = vmull_s8(vb2x0, va0x0); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal() 201 int16x8_t vprod1x2 = vmull_s8(vb2x0, va1x0); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal() 202 int16x8_t vprod2x2 = vmull_s8(vb2x0, va2x0); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal() 203 int16x8_t vprod3x2 = vmull_s8(vb2x0, va3x0); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal()
|
D | 3x16c8-minmax-neon-mlal-padal.c | 132 const int8x8_t vb2x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal() local 168 int16x8_t vprod0x2 = vmull_s8(vb2x0, va0x0); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal() 169 int16x8_t vprod1x2 = vmull_s8(vb2x0, va1x0); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal() 170 int16x8_t vprod2x2 = vmull_s8(vb2x0, va2x0); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
|