| /external/XNNPACK/src/f32-dwconv/gen/ |
| D | up16x9-minmax-fma3-acc2.c | 147 const __m256 vk6x01234567 = _mm256_load_ps(w + 112); in xnn_f32_dwconv_minmax_ukernel_up16x9__fma3_acc2() local
149 vacc01234567p0 = _mm256_fmadd_ps(vi6x01234567, vk6x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up16x9__fma3_acc2()
227 const __m256 vk6x01234567 = _mm256_load_ps(w + 112); in xnn_f32_dwconv_minmax_ukernel_up16x9__fma3_acc2() local
228 vacc01234567p0 = _mm256_fmadd_ps(vi6x01234567, vk6x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up16x9__fma3_acc2()
285 const __m256 vk6x01234567 = _mm256_load_ps(w + 112); in xnn_f32_dwconv_minmax_ukernel_up16x9__fma3_acc2() local
286 vacc01234567p0 = _mm256_fmadd_ps(vi6x01234567, vk6x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up16x9__fma3_acc2()
|
| D | up16x9-minmax-fma3.c | 147 const __m256 vk6x01234567 = _mm256_load_ps(w + 112); in xnn_f32_dwconv_minmax_ukernel_up16x9__fma3() local
149 vacc01234567p0 = _mm256_fmadd_ps(vi6x01234567, vk6x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up16x9__fma3()
224 const __m256 vk6x01234567 = _mm256_load_ps(w + 112); in xnn_f32_dwconv_minmax_ukernel_up16x9__fma3() local
225 vacc01234567p0 = _mm256_fmadd_ps(vi6x01234567, vk6x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up16x9__fma3()
280 const __m256 vk6x01234567 = _mm256_load_ps(w + 112); in xnn_f32_dwconv_minmax_ukernel_up16x9__fma3() local
281 vacc01234567p0 = _mm256_fmadd_ps(vi6x01234567, vk6x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up16x9__fma3()
|
| D | up16x9-minmax-avx.c | 147 const __m256 vk6x01234567 = _mm256_load_ps(w + 112); in xnn_f32_dwconv_minmax_ukernel_up16x9__avx() local
149 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi6x01234567, vk6x01234567)); in xnn_f32_dwconv_minmax_ukernel_up16x9__avx()
224 const __m256 vk6x01234567 = _mm256_load_ps(w + 112); in xnn_f32_dwconv_minmax_ukernel_up16x9__avx() local
225 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi6x01234567, vk6x01234567)); in xnn_f32_dwconv_minmax_ukernel_up16x9__avx()
280 const __m256 vk6x01234567 = _mm256_load_ps(w + 112); in xnn_f32_dwconv_minmax_ukernel_up16x9__avx() local
281 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi6x01234567, vk6x01234567)); in xnn_f32_dwconv_minmax_ukernel_up16x9__avx()
|
| D | up16x9-minmax-avx-acc2.c | 147 const __m256 vk6x01234567 = _mm256_load_ps(w + 112); in xnn_f32_dwconv_minmax_ukernel_up16x9__avx_acc2() local
149 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi6x01234567, vk6x01234567)); in xnn_f32_dwconv_minmax_ukernel_up16x9__avx_acc2()
227 const __m256 vk6x01234567 = _mm256_load_ps(w + 112); in xnn_f32_dwconv_minmax_ukernel_up16x9__avx_acc2() local
228 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi6x01234567, vk6x01234567)); in xnn_f32_dwconv_minmax_ukernel_up16x9__avx_acc2()
285 const __m256 vk6x01234567 = _mm256_load_ps(w + 112); in xnn_f32_dwconv_minmax_ukernel_up16x9__avx_acc2() local
286 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi6x01234567, vk6x01234567)); in xnn_f32_dwconv_minmax_ukernel_up16x9__avx_acc2()
|
| D | up8x9-minmax-fma3.c | 127 const __m256 vk6x01234567 = _mm256_load_ps(w + 56); in xnn_f32_dwconv_minmax_ukernel_up8x9__fma3() local
128 vacc01234567p0 = _mm256_fmadd_ps(vi6x01234567, vk6x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up8x9__fma3()
183 const __m256 vk6x01234567 = _mm256_load_ps(w + 56); in xnn_f32_dwconv_minmax_ukernel_up8x9__fma3() local
184 vacc01234567p0 = _mm256_fmadd_ps(vi6x01234567, vk6x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up8x9__fma3()
|
| D | up8x9-minmax-avx.c | 127 const __m256 vk6x01234567 = _mm256_load_ps(w + 56); in xnn_f32_dwconv_minmax_ukernel_up8x9__avx() local
128 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi6x01234567, vk6x01234567)); in xnn_f32_dwconv_minmax_ukernel_up8x9__avx()
183 const __m256 vk6x01234567 = _mm256_load_ps(w + 56); in xnn_f32_dwconv_minmax_ukernel_up8x9__avx() local
184 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi6x01234567, vk6x01234567)); in xnn_f32_dwconv_minmax_ukernel_up8x9__avx()
|
| D | up8x9-minmax-fma3-acc2.c | 127 const __m256 vk6x01234567 = _mm256_load_ps(w + 56); in xnn_f32_dwconv_minmax_ukernel_up8x9__fma3_acc2() local
128 vacc01234567p0 = _mm256_fmadd_ps(vi6x01234567, vk6x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up8x9__fma3_acc2()
185 const __m256 vk6x01234567 = _mm256_load_ps(w + 56); in xnn_f32_dwconv_minmax_ukernel_up8x9__fma3_acc2() local
186 vacc01234567p0 = _mm256_fmadd_ps(vi6x01234567, vk6x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up8x9__fma3_acc2()
|
| D | up8x9-minmax-avx-acc2.c | 127 const __m256 vk6x01234567 = _mm256_load_ps(w + 56); in xnn_f32_dwconv_minmax_ukernel_up8x9__avx_acc2() local
128 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi6x01234567, vk6x01234567)); in xnn_f32_dwconv_minmax_ukernel_up8x9__avx_acc2()
185 const __m256 vk6x01234567 = _mm256_load_ps(w + 56); in xnn_f32_dwconv_minmax_ukernel_up8x9__avx_acc2() local
186 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi6x01234567, vk6x01234567)); in xnn_f32_dwconv_minmax_ukernel_up8x9__avx_acc2()
|
| /external/XNNPACK/src/f16-dwconv/gen/ |
| D | up16x9-minmax-neonfp16arith-acc2.c | 135 const float16x8_t vk6x01234567 = vld1q_f16(w); w += 8; in xnn_f16_dwconv_minmax_ukernel_up16x9__neonfp16arith_acc2() local
137 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi6x01234567, vk6x01234567); in xnn_f16_dwconv_minmax_ukernel_up16x9__neonfp16arith_acc2()
195 const float16x8_t vk6x01234567 = vld1q_f16(w + 104); in xnn_f16_dwconv_minmax_ukernel_up16x9__neonfp16arith_acc2() local
196 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi6x01234567, vk6x01234567); in xnn_f16_dwconv_minmax_ukernel_up16x9__neonfp16arith_acc2()
243 const float16x8_t vk6x01234567 = vld1q_f16(w + 112); in xnn_f16_dwconv_minmax_ukernel_up16x9__neonfp16arith_acc2() local
244 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi6x01234567, vk6x01234567); in xnn_f16_dwconv_minmax_ukernel_up16x9__neonfp16arith_acc2()
|
| D | up16x9-minmax-neonfp16arith.c | 135 const float16x8_t vk6x01234567 = vld1q_f16(w); w += 8; in xnn_f16_dwconv_minmax_ukernel_up16x9__neonfp16arith() local
137 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi6x01234567, vk6x01234567); in xnn_f16_dwconv_minmax_ukernel_up16x9__neonfp16arith()
192 const float16x8_t vk6x01234567 = vld1q_f16(w + 104); in xnn_f16_dwconv_minmax_ukernel_up16x9__neonfp16arith() local
193 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi6x01234567, vk6x01234567); in xnn_f16_dwconv_minmax_ukernel_up16x9__neonfp16arith()
238 const float16x8_t vk6x01234567 = vld1q_f16(w + 112); in xnn_f16_dwconv_minmax_ukernel_up16x9__neonfp16arith() local
239 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi6x01234567, vk6x01234567); in xnn_f16_dwconv_minmax_ukernel_up16x9__neonfp16arith()
|
| D | up8x9-minmax-neonfp16arith.c | 115 const float16x8_t vk6x01234567 = vld1q_f16(w); w += 8; in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith() local
116 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi6x01234567, vk6x01234567); in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
161 const float16x8_t vk6x01234567 = vld1q_f16(w); w += 8; in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith() local
162 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi6x01234567, vk6x01234567); in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith()
|
| D | up8x9-minmax-neonfp16arith-acc2.c | 115 const float16x8_t vk6x01234567 = vld1q_f16(w); w += 8; in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith_acc2() local
116 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi6x01234567, vk6x01234567); in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith_acc2()
163 const float16x8_t vk6x01234567 = vld1q_f16(w); w += 8; in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith_acc2() local
164 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi6x01234567, vk6x01234567); in xnn_f16_dwconv_minmax_ukernel_up8x9__neonfp16arith_acc2()
|
| D | up16x9-minmax-fma3.c | 150 const __m256 vk6x01234567 = _mm256_cvtph_ps(_mm_loadu_si128((const __m128i*) (w + 112))); in xnn_f16_dwconv_minmax_ukernel_up16x9__fma3() local
152 …vacc01234567p0 = _mm256_cvtph_ps(_mm256_cvtps_ph(_mm256_fmadd_ps(vi6x01234567, vk6x01234567, vacc0… in xnn_f16_dwconv_minmax_ukernel_up16x9__fma3()
227 const __m256 vk6x01234567 = _mm256_cvtph_ps(_mm_load_si128((const __m128i*) (w + 112))); in xnn_f16_dwconv_minmax_ukernel_up16x9__fma3() local
228 …vacc01234567p0 = _mm256_cvtph_ps(_mm256_cvtps_ph(_mm256_fmadd_ps(vi6x01234567, vk6x01234567, vacc0… in xnn_f16_dwconv_minmax_ukernel_up16x9__fma3()
289 const __m256 vk6x01234567 = _mm256_cvtph_ps(_mm_load_si128((const __m128i*) (w + 112))); in xnn_f16_dwconv_minmax_ukernel_up16x9__fma3() local
290 …vacc01234567p0 = _mm256_cvtph_ps(_mm256_cvtps_ph(_mm256_fmadd_ps(vi6x01234567, vk6x01234567, vacc0… in xnn_f16_dwconv_minmax_ukernel_up16x9__fma3()
|
| D | up8x9-minmax-fma3-acc2.c | 130 const __m256 vk6x01234567 = _mm256_cvtph_ps(_mm_loadu_si128((const __m128i*) (w + 56))); in xnn_f16_dwconv_minmax_ukernel_up8x9__fma3_acc2() local
131 …vacc01234567p0 = _mm256_cvtph_ps(_mm256_cvtps_ph(_mm256_fmadd_ps(vi6x01234567, vk6x01234567, vacc0… in xnn_f16_dwconv_minmax_ukernel_up8x9__fma3_acc2()
194 const __m256 vk6x01234567 = _mm256_cvtph_ps(_mm_load_si128((const __m128i*) (w + 56))); in xnn_f16_dwconv_minmax_ukernel_up8x9__fma3_acc2() local
195 …vacc01234567p0 = _mm256_cvtph_ps(_mm256_cvtps_ph(_mm256_fmadd_ps(vi6x01234567, vk6x01234567, vacc0… in xnn_f16_dwconv_minmax_ukernel_up8x9__fma3_acc2()
|
| D | up8x9-minmax-fma3.c | 130 const __m256 vk6x01234567 = _mm256_cvtph_ps(_mm_loadu_si128((const __m128i*) (w + 56))); in xnn_f16_dwconv_minmax_ukernel_up8x9__fma3() local
131 …vacc01234567p0 = _mm256_cvtph_ps(_mm256_cvtps_ph(_mm256_fmadd_ps(vi6x01234567, vk6x01234567, vacc0… in xnn_f16_dwconv_minmax_ukernel_up8x9__fma3()
192 const __m256 vk6x01234567 = _mm256_cvtph_ps(_mm_load_si128((const __m128i*) (w + 56))); in xnn_f16_dwconv_minmax_ukernel_up8x9__fma3() local
193 …vacc01234567p0 = _mm256_cvtph_ps(_mm256_cvtps_ph(_mm256_fmadd_ps(vi6x01234567, vk6x01234567, vacc0… in xnn_f16_dwconv_minmax_ukernel_up8x9__fma3()
|
| D | up32x9-minmax-neonfp16arith-acc2.c | 175 const float16x8_t vk6x01234567 = vld1q_f16(w); w += 8; in xnn_f16_dwconv_minmax_ukernel_up32x9__neonfp16arith_acc2() local
179 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi6x01234567, vk6x01234567); in xnn_f16_dwconv_minmax_ukernel_up32x9__neonfp16arith_acc2()
259 const float16x8_t vk6x01234567 = vld1q_f16(w + 216); in xnn_f16_dwconv_minmax_ukernel_up32x9__neonfp16arith_acc2() local
260 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi6x01234567, vk6x01234567); in xnn_f16_dwconv_minmax_ukernel_up32x9__neonfp16arith_acc2()
307 const float16x8_t vk6x01234567 = vld1q_f16(w + 224); in xnn_f16_dwconv_minmax_ukernel_up32x9__neonfp16arith_acc2() local
308 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi6x01234567, vk6x01234567); in xnn_f16_dwconv_minmax_ukernel_up32x9__neonfp16arith_acc2()
|
| /external/XNNPACK/src/qs8-dwconv/gen/ |
| D | up8x9-minmax-fp32-neonv8-mul16.c | 129 … const int16x8_t vk6x01234567 = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x9__neonv8_mul16() local
131 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi6x01234567), vget_low_s16(vk6x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x9__neonv8_mul16()
132 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi6x01234567), vget_high_s16(vk6x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x9__neonv8_mul16()
211 const int16x8_t vk6x01234567 = vmovl_s8(vld1_s8((const void*) ((const int8_t*) w + 48))); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x9__neonv8_mul16() local
213 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi6x01234567), vget_low_s16(vk6x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x9__neonv8_mul16()
214 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi6x01234567), vget_high_s16(vk6x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x9__neonv8_mul16()
|
| D | up8x9-minmax-rndnu-neon-mul16.c | 130 … const int16x8_t vk6x01234567 = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up8x9__neon_mul16() local
132 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi6x01234567), vget_low_s16(vk6x01234567)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up8x9__neon_mul16()
133 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi6x01234567), vget_high_s16(vk6x01234567)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up8x9__neon_mul16()
212 const int16x8_t vk6x01234567 = vmovl_s8(vld1_s8((const void*) ((const int8_t*) w + 48))); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up8x9__neon_mul16() local
214 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi6x01234567), vget_low_s16(vk6x01234567)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up8x9__neon_mul16()
215 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi6x01234567), vget_high_s16(vk6x01234567)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up8x9__neon_mul16()
|
| D | up8x9-minmax-fp32-neon-mul16.c | 129 … const int16x8_t vk6x01234567 = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x9__neon_mul16() local
131 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi6x01234567), vget_low_s16(vk6x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x9__neon_mul16()
132 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi6x01234567), vget_high_s16(vk6x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x9__neon_mul16()
212 const int16x8_t vk6x01234567 = vmovl_s8(vld1_s8((const void*) ((const int8_t*) w + 48))); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x9__neon_mul16() local
214 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi6x01234567), vget_low_s16(vk6x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x9__neon_mul16()
215 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi6x01234567), vget_high_s16(vk6x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x9__neon_mul16()
|
| /external/XNNPACK/src/qc8-dwconv/gen/ |
| D | up8x9-minmax-fp32-neonv8-mul16.c | 128 … const int16x8_t vk6x01234567 = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x9__neonv8_mul16() local
130 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi6x01234567), vget_low_s16(vk6x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x9__neonv8_mul16()
131 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi6x01234567), vget_high_s16(vk6x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x9__neonv8_mul16()
213 const int16x8_t vk6x01234567 = vmovl_s8(vld1_s8((const void*) ((const int8_t*) w + 48))); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x9__neonv8_mul16() local
215 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi6x01234567), vget_low_s16(vk6x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x9__neonv8_mul16()
216 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi6x01234567), vget_high_s16(vk6x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x9__neonv8_mul16()
|
| D | up8x9-minmax-fp32-neon-mul16.c | 128 … const int16x8_t vk6x01234567 = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x9__neon_mul16() local
130 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi6x01234567), vget_low_s16(vk6x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x9__neon_mul16()
131 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi6x01234567), vget_high_s16(vk6x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x9__neon_mul16()
214 const int16x8_t vk6x01234567 = vmovl_s8(vld1_s8((const void*) ((const int8_t*) w + 48))); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x9__neon_mul16() local
216 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi6x01234567), vget_low_s16(vk6x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x9__neon_mul16()
217 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi6x01234567), vget_high_s16(vk6x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x9__neon_mul16()
|
| D | up8x9-minmax-fp32-sse2-mul16-add16.c | 158 …const __m128i vk6x01234567 = _mm_loadl_epi64((const __m128i*) ((uintptr_t) w + 8 * sizeof(int32_t)… in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x9__sse2_mul16_add16() local
162 … const __m128i vxk6x01234567 = _mm_srai_epi16(_mm_unpacklo_epi8(vk6x01234567, vk6x01234567), 8); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x9__sse2_mul16_add16()
293 …const __m128i vk6x01234567 = _mm_loadl_epi64((const __m128i*) ((uintptr_t) w + 8 * sizeof(int32_t)… in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x9__sse2_mul16_add16() local
296 … const __m128i vxk6x01234567 = _mm_srai_epi16(_mm_unpacklo_epi8(vk6x01234567, vk6x01234567), 8); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x9__sse2_mul16_add16()
|
| /external/XNNPACK/src/qu8-dwconv/gen/ |
| D | up8x9-minmax-fp32-neonv8-mul16.c | 130 …const int16x8_t vk6x01234567 = vreinterpretq_s16_u16(vsubl_u8(vld1_u8(w), vkernel_zero_point)); w … in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x9__neonv8_mul16() local
132 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi6x01234567), vget_low_s16(vk6x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x9__neonv8_mul16()
133 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi6x01234567), vget_high_s16(vk6x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x9__neonv8_mul16()
212 …const int16x8_t vk6x01234567 = vreinterpretq_s16_u16(vsubl_u8(vld1_u8((const void*) ((const uint8_… in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x9__neonv8_mul16() local
214 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi6x01234567), vget_low_s16(vk6x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x9__neonv8_mul16()
215 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi6x01234567), vget_high_s16(vk6x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x9__neonv8_mul16()
|
| D | up8x9-minmax-rndnu-neon-mul16.c | 131 …const int16x8_t vk6x01234567 = vreinterpretq_s16_u16(vsubl_u8(vld1_u8(w), vkernel_zero_point)); w … in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x9__neon_mul16() local
133 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi6x01234567), vget_low_s16(vk6x01234567)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x9__neon_mul16()
134 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi6x01234567), vget_high_s16(vk6x01234567)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x9__neon_mul16()
213 …const int16x8_t vk6x01234567 = vreinterpretq_s16_u16(vsubl_u8(vld1_u8((const void*) ((const uint8_… in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x9__neon_mul16() local
215 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi6x01234567), vget_low_s16(vk6x01234567)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x9__neon_mul16()
216 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi6x01234567), vget_high_s16(vk6x01234567)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x9__neon_mul16()
|
| D | up8x9-minmax-fp32-neon-mul16.c | 130 …const int16x8_t vk6x01234567 = vreinterpretq_s16_u16(vsubl_u8(vld1_u8(w), vkernel_zero_point)); w … in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x9__neon_mul16() local
132 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi6x01234567), vget_low_s16(vk6x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x9__neon_mul16()
133 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi6x01234567), vget_high_s16(vk6x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x9__neon_mul16()
213 …const int16x8_t vk6x01234567 = vreinterpretq_s16_u16(vsubl_u8(vld1_u8((const void*) ((const uint8_… in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x9__neon_mul16() local
215 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi6x01234567), vget_low_s16(vk6x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x9__neon_mul16()
216 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi6x01234567), vget_high_s16(vk6x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x9__neon_mul16()
|