/external/XNNPACK/src/qs8-igemm/gen/ |
D | 4x8c2s4-minmax-rndnu-neon-mlal.c | 180 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c2s4__neon_mlal() local 181 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c2s4__neon_mlal() 182 vprod1x0123c2 = vmlal_s8(vprod1x0123c2, vb0123c2x1, va1x1); in xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c2s4__neon_mlal() 183 vprod2x0123c2 = vmlal_s8(vprod2x0123c2, vb0123c2x1, va2x1); in xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c2s4__neon_mlal() 184 vprod3x0123c2 = vmlal_s8(vprod3x0123c2, vb0123c2x1, va3x1); in xnn_qs8_igemm_minmax_rndnu_ukernel_4x8c2s4__neon_mlal()
|
D | 3x8c2s4-minmax-rndnu-neon-mlal.c | 151 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c2s4__neon_mlal() local 152 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c2s4__neon_mlal() 153 vprod1x0123c2 = vmlal_s8(vprod1x0123c2, vb0123c2x1, va1x1); in xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c2s4__neon_mlal() 154 vprod2x0123c2 = vmlal_s8(vprod2x0123c2, vb0123c2x1, va2x1); in xnn_qs8_igemm_minmax_rndnu_ukernel_3x8c2s4__neon_mlal()
|
D | 2x8c2s4-minmax-fp32-neonv8-mlal.c | 123 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qs8_igemm_minmax_fp32_ukernel_2x8c2s4__neonv8_mlal() local 124 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qs8_igemm_minmax_fp32_ukernel_2x8c2s4__neonv8_mlal() 125 vprod1x0123c2 = vmlal_s8(vprod1x0123c2, vb0123c2x1, va1x1); in xnn_qs8_igemm_minmax_fp32_ukernel_2x8c2s4__neonv8_mlal()
|
D | 2x8c2s4-minmax-rndnu-neon-mlal.c | 122 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c2s4__neon_mlal() local 123 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c2s4__neon_mlal() 124 vprod1x0123c2 = vmlal_s8(vprod1x0123c2, vb0123c2x1, va1x1); in xnn_qs8_igemm_minmax_rndnu_ukernel_2x8c2s4__neon_mlal()
|
D | 2x8c2s4-minmax-fp32-neon-mlal.c | 122 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qs8_igemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal() local 123 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qs8_igemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal() 124 vprod1x0123c2 = vmlal_s8(vprod1x0123c2, vb0123c2x1, va1x1); in xnn_qs8_igemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
|
D | 1x8c2s4-minmax-fp32-neonv8-mlal.c | 94 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qs8_igemm_minmax_fp32_ukernel_1x8c2s4__neonv8_mlal() local 95 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qs8_igemm_minmax_fp32_ukernel_1x8c2s4__neonv8_mlal()
|
D | 1x8c2s4-minmax-rndnu-neon-mlal.c | 93 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c2s4__neon_mlal() local 94 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qs8_igemm_minmax_rndnu_ukernel_1x8c2s4__neon_mlal()
|
D | 1x8c2s4-minmax-fp32-neon-mlal.c | 93 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qs8_igemm_minmax_fp32_ukernel_1x8c2s4__neon_mlal() local 94 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qs8_igemm_minmax_fp32_ukernel_1x8c2s4__neon_mlal()
|
/external/XNNPACK/src/qs8-gemm/gen/ |
D | 4x8c2s4-minmax-rndnu-neon-mlal.c | 163 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qs8_gemm_minmax_rndnu_ukernel_4x8c2s4__neon_mlal() local 164 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qs8_gemm_minmax_rndnu_ukernel_4x8c2s4__neon_mlal() 165 vprod1x0123c2 = vmlal_s8(vprod1x0123c2, vb0123c2x1, va1x1); in xnn_qs8_gemm_minmax_rndnu_ukernel_4x8c2s4__neon_mlal() 166 vprod2x0123c2 = vmlal_s8(vprod2x0123c2, vb0123c2x1, va2x1); in xnn_qs8_gemm_minmax_rndnu_ukernel_4x8c2s4__neon_mlal() 167 vprod3x0123c2 = vmlal_s8(vprod3x0123c2, vb0123c2x1, va3x1); in xnn_qs8_gemm_minmax_rndnu_ukernel_4x8c2s4__neon_mlal()
|
D | 3x8c2s4-minmax-rndnu-neon-mlal.c | 136 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qs8_gemm_minmax_rndnu_ukernel_3x8c2s4__neon_mlal() local 137 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qs8_gemm_minmax_rndnu_ukernel_3x8c2s4__neon_mlal() 138 vprod1x0123c2 = vmlal_s8(vprod1x0123c2, vb0123c2x1, va1x1); in xnn_qs8_gemm_minmax_rndnu_ukernel_3x8c2s4__neon_mlal() 139 vprod2x0123c2 = vmlal_s8(vprod2x0123c2, vb0123c2x1, va2x1); in xnn_qs8_gemm_minmax_rndnu_ukernel_3x8c2s4__neon_mlal()
|
D | 2x8c2s4-minmax-rndnu-neon-mlal.c | 109 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qs8_gemm_minmax_rndnu_ukernel_2x8c2s4__neon_mlal() local 110 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qs8_gemm_minmax_rndnu_ukernel_2x8c2s4__neon_mlal() 111 vprod1x0123c2 = vmlal_s8(vprod1x0123c2, vb0123c2x1, va1x1); in xnn_qs8_gemm_minmax_rndnu_ukernel_2x8c2s4__neon_mlal()
|
D | 2x8c2s4-minmax-fp32-neonv8-mlal.c | 110 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qs8_gemm_minmax_fp32_ukernel_2x8c2s4__neonv8_mlal() local 111 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qs8_gemm_minmax_fp32_ukernel_2x8c2s4__neonv8_mlal() 112 vprod1x0123c2 = vmlal_s8(vprod1x0123c2, vb0123c2x1, va1x1); in xnn_qs8_gemm_minmax_fp32_ukernel_2x8c2s4__neonv8_mlal()
|
D | 2x8c2s4-minmax-fp32-neon-mlal.c | 109 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qs8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal() local 110 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qs8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal() 111 vprod1x0123c2 = vmlal_s8(vprod1x0123c2, vb0123c2x1, va1x1); in xnn_qs8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
|
D | 1x8c2s4-minmax-fp32-neon-mlal.c | 82 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qs8_gemm_minmax_fp32_ukernel_1x8c2s4__neon_mlal() local 83 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qs8_gemm_minmax_fp32_ukernel_1x8c2s4__neon_mlal()
|
D | 1x8c2s4-minmax-fp32-neonv8-mlal.c | 83 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qs8_gemm_minmax_fp32_ukernel_1x8c2s4__neonv8_mlal() local 84 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qs8_gemm_minmax_fp32_ukernel_1x8c2s4__neonv8_mlal()
|
D | 1x8c2s4-minmax-rndnu-neon-mlal.c | 82 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qs8_gemm_minmax_rndnu_ukernel_1x8c2s4__neon_mlal() local 83 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qs8_gemm_minmax_rndnu_ukernel_1x8c2s4__neon_mlal()
|
D | 4x8c2-minmax-rndnu-neon-mlal-ld4r.c | 173 … const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const void*) ((uintptr_t) w + 8 * sizeof(int8_t)); in xnn_qs8_gemm_minmax_rndnu_ukernel_4x8c2__neon_mlal_ld4r() local 174 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0c2x1); in xnn_qs8_gemm_minmax_rndnu_ukernel_4x8c2__neon_mlal_ld4r() 175 vprod1x0123c2 = vmlal_s8(vprod1x0123c2, vb0123c2x1, va1c2x1); in xnn_qs8_gemm_minmax_rndnu_ukernel_4x8c2__neon_mlal_ld4r() 176 vprod2x0123c2 = vmlal_s8(vprod2x0123c2, vb0123c2x1, va2c2x1); in xnn_qs8_gemm_minmax_rndnu_ukernel_4x8c2__neon_mlal_ld4r() 177 vprod3x0123c2 = vmlal_s8(vprod3x0123c2, vb0123c2x1, va3c2x1); in xnn_qs8_gemm_minmax_rndnu_ukernel_4x8c2__neon_mlal_ld4r()
|
/external/XNNPACK/src/qc8-gemm/gen/ |
D | 2x8c2s4-minmax-fp32-neon-mlal.c | 109 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal() local 110 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal() 111 vprod1x0123c2 = vmlal_s8(vprod1x0123c2, vb0123c2x1, va1x1); in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
|
D | 2x8c2s4-minmax-fp32-neonv8-mlal.c | 110 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neonv8_mlal() local 111 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neonv8_mlal() 112 vprod1x0123c2 = vmlal_s8(vprod1x0123c2, vb0123c2x1, va1x1); in xnn_qc8_gemm_minmax_fp32_ukernel_2x8c2s4__neonv8_mlal()
|
D | 1x8c2s4-minmax-fp32-neonv8-mlal.c | 83 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qc8_gemm_minmax_fp32_ukernel_1x8c2s4__neonv8_mlal() local 84 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qc8_gemm_minmax_fp32_ukernel_1x8c2s4__neonv8_mlal()
|
D | 1x8c2s4-minmax-fp32-neon-mlal.c | 82 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qc8_gemm_minmax_fp32_ukernel_1x8c2s4__neon_mlal() local 83 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qc8_gemm_minmax_fp32_ukernel_1x8c2s4__neon_mlal()
|
/external/XNNPACK/src/qc8-igemm/gen/ |
D | 2x8c2s4-minmax-fp32-neon-mlal.c | 122 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qc8_igemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal() local 123 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qc8_igemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal() 124 vprod1x0123c2 = vmlal_s8(vprod1x0123c2, vb0123c2x1, va1x1); in xnn_qc8_igemm_minmax_fp32_ukernel_2x8c2s4__neon_mlal()
|
D | 2x8c2s4-minmax-fp32-neonv8-mlal.c | 123 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qc8_igemm_minmax_fp32_ukernel_2x8c2s4__neonv8_mlal() local 124 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qc8_igemm_minmax_fp32_ukernel_2x8c2s4__neonv8_mlal() 125 vprod1x0123c2 = vmlal_s8(vprod1x0123c2, vb0123c2x1, va1x1); in xnn_qc8_igemm_minmax_fp32_ukernel_2x8c2s4__neonv8_mlal()
|
D | 1x8c2s4-minmax-fp32-neonv8-mlal.c | 94 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qc8_igemm_minmax_fp32_ukernel_1x8c2s4__neonv8_mlal() local 95 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qc8_igemm_minmax_fp32_ukernel_1x8c2s4__neonv8_mlal()
|
D | 1x8c2s4-minmax-fp32-neon-mlal.c | 93 const int8x8_t vb0123c2x1 = vld1_s8(w); w = (const int8_t*) w + 8; in xnn_qc8_igemm_minmax_fp32_ukernel_1x8c2s4__neon_mlal() local 94 vprod0x0123c2 = vmlal_s8(vprod0x0123c2, vb0123c2x1, va0x1); in xnn_qc8_igemm_minmax_fp32_ukernel_1x8c2s4__neon_mlal()
|