/external/XNNPACK/src/qs8-gemm/gen/ |
D | 3x8c8-minmax-neon-mlal-padal.c | 97 const int8x8_t vb6x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal() local 161 int16x8_t vprod0x6 = vmull_s8(vb6x0, va0x0); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal() 162 int16x8_t vprod1x6 = vmull_s8(vb6x0, va1x0); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal() 163 int16x8_t vprod2x6 = vmull_s8(vb6x0, va2x0); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal()
|
D | 4x8c8-minmax-neon-mlal-padal.c | 113 const int8x8_t vb6x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() local 195 int16x8_t vprod0x6 = vmull_s8(vb6x0, va0x0); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() 196 int16x8_t vprod1x6 = vmull_s8(vb6x0, va1x0); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() 197 int16x8_t vprod2x6 = vmull_s8(vb6x0, va2x0); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() 198 int16x8_t vprod3x6 = vmull_s8(vb6x0, va3x0); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal()
|
D | 2x8c8-minmax-neon-mlal-padal.c | 81 const int8x8_t vb6x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mlal_padal() local 127 int16x8_t vprod0x6 = vmull_s8(vb6x0, va0x0); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mlal_padal() 128 int16x8_t vprod1x6 = vmull_s8(vb6x0, va1x0); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mlal_padal()
|
D | 1x8c8-minmax-neon-mlal-padal.c | 65 const int8x8_t vb6x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mlal_padal() local 93 int16x8_t vprod0x6 = vmull_s8(vb6x0, va0x0); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mlal_padal()
|
D | 1x16c8-minmax-neon-mlal-padal.c | 73 const int8x8_t vb6x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mlal_padal() local 109 int16x8_t vprod0x6 = vmull_s8(vb6x0, va0x0); in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mlal_padal()
|
D | 2x16c8-minmax-neon-mlal-padal.c | 97 const int8x8_t vb6x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal() local 151 int16x8_t vprod0x6 = vmull_s8(vb6x0, va0x0); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal() 152 int16x8_t vprod1x6 = vmull_s8(vb6x0, va1x0); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal()
|
D | 3x16c8-minmax-neon-mlal-padal.c | 121 const int8x8_t vb6x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal() local 193 int16x8_t vprod0x6 = vmull_s8(vb6x0, va0x0); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal() 194 int16x8_t vprod1x6 = vmull_s8(vb6x0, va1x0); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal() 195 int16x8_t vprod2x6 = vmull_s8(vb6x0, va2x0); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal()
|
D | 4x16c8-minmax-neon-mlal-padal.c | 145 const int8x8_t vb6x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal() local 235 int16x8_t vprod0x6 = vmull_s8(vb6x0, va0x0); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal() 236 int16x8_t vprod1x6 = vmull_s8(vb6x0, va1x0); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal() 237 int16x8_t vprod2x6 = vmull_s8(vb6x0, va2x0); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal() 238 int16x8_t vprod3x6 = vmull_s8(vb6x0, va3x0); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal()
|
/external/XNNPACK/src/qs8-igemm/gen/ |
D | 4x8c8-minmax-neon-mlal-padal.c | 130 const int8x8_t vb6x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mlal_padal() local 212 int16x8_t vprod0x6 = vmull_s8(vb6x0, va0x0); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mlal_padal() 213 int16x8_t vprod1x6 = vmull_s8(vb6x0, va1x0); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mlal_padal() 214 int16x8_t vprod2x6 = vmull_s8(vb6x0, va2x0); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mlal_padal() 215 int16x8_t vprod3x6 = vmull_s8(vb6x0, va3x0); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mlal_padal()
|
D | 2x8c8-minmax-neon-mlal-padal.c | 94 const int8x8_t vb6x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mlal_padal() local 140 int16x8_t vprod0x6 = vmull_s8(vb6x0, va0x0); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mlal_padal() 141 int16x8_t vprod1x6 = vmull_s8(vb6x0, va1x0); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mlal_padal()
|
D | 3x8c8-minmax-neon-mlal-padal.c | 112 const int8x8_t vb6x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal() local 176 int16x8_t vprod0x6 = vmull_s8(vb6x0, va0x0); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal() 177 int16x8_t vprod1x6 = vmull_s8(vb6x0, va1x0); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal() 178 int16x8_t vprod2x6 = vmull_s8(vb6x0, va2x0); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal()
|
D | 1x8c8-minmax-neon-mlal-padal.c | 76 const int8x8_t vb6x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_igemm_minmax_ukernel_1x8c8__neon_mlal_padal() local 104 int16x8_t vprod0x6 = vmull_s8(vb6x0, va0x0); in xnn_qs8_igemm_minmax_ukernel_1x8c8__neon_mlal_padal()
|
D | 1x16c8-minmax-neon-mlal-padal.c | 84 const int8x8_t vb6x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mlal_padal() local 120 int16x8_t vprod0x6 = vmull_s8(vb6x0, va0x0); in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mlal_padal()
|
D | 2x16c8-minmax-neon-mlal-padal.c | 110 const int8x8_t vb6x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal() local 164 int16x8_t vprod0x6 = vmull_s8(vb6x0, va0x0); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal() 165 int16x8_t vprod1x6 = vmull_s8(vb6x0, va1x0); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal()
|
D | 4x16c8-minmax-neon-mlal-padal.c | 162 const int8x8_t vb6x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal() local 252 int16x8_t vprod0x6 = vmull_s8(vb6x0, va0x0); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal() 253 int16x8_t vprod1x6 = vmull_s8(vb6x0, va1x0); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal() 254 int16x8_t vprod2x6 = vmull_s8(vb6x0, va2x0); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal() 255 int16x8_t vprod3x6 = vmull_s8(vb6x0, va3x0); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal()
|
D | 3x16c8-minmax-neon-mlal-padal.c | 136 const int8x8_t vb6x0 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof( int8_t)); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal() local 208 int16x8_t vprod0x6 = vmull_s8(vb6x0, va0x0); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal() 209 int16x8_t vprod1x6 = vmull_s8(vb6x0, va1x0); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal() 210 int16x8_t vprod2x6 = vmull_s8(vb6x0, va2x0); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
|