| /external/XNNPACK/src/f32-dwconv/gen/ |
| D | up8x25-minmax-avx.c | 252 const __m256 vi14x01234567 = _mm256_loadu_ps(i14); in xnn_f32_dwconv_minmax_ukernel_up8x25__avx() local
256 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi14x01234567, vk14x01234567)); in xnn_f32_dwconv_minmax_ukernel_up8x25__avx()
390 const __m256 vi14x01234567 = _mm256_maskload_ps(i14, vmask); in xnn_f32_dwconv_minmax_ukernel_up8x25__avx() local
392 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi14x01234567, vk14x01234567)); in xnn_f32_dwconv_minmax_ukernel_up8x25__avx()
|
| D | up8x25-minmax-fma3-acc2.c | 252 const __m256 vi14x01234567 = _mm256_loadu_ps(i14); in xnn_f32_dwconv_minmax_ukernel_up8x25__fma3_acc2() local
256 vacc01234567p0 = _mm256_fmadd_ps(vi14x01234567, vk14x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up8x25__fma3_acc2()
392 const __m256 vi14x01234567 = _mm256_maskload_ps(i14, vmask); in xnn_f32_dwconv_minmax_ukernel_up8x25__fma3_acc2() local
394 vacc01234567p0 = _mm256_fmadd_ps(vi14x01234567, vk14x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up8x25__fma3_acc2()
|
| D | up8x25-minmax-fma3.c | 252 const __m256 vi14x01234567 = _mm256_loadu_ps(i14); in xnn_f32_dwconv_minmax_ukernel_up8x25__fma3() local
256 vacc01234567p0 = _mm256_fmadd_ps(vi14x01234567, vk14x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up8x25__fma3()
390 const __m256 vi14x01234567 = _mm256_maskload_ps(i14, vmask); in xnn_f32_dwconv_minmax_ukernel_up8x25__fma3() local
392 vacc01234567p0 = _mm256_fmadd_ps(vi14x01234567, vk14x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up8x25__fma3()
|
| D | up16x25-minmax-avx.c | 295 const __m256 vi14x01234567 = _mm256_loadu_ps(i14); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx() local
301 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi14x01234567, vk14x01234567)); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx()
493 const __m256 vi14x01234567 = _mm256_loadu_ps(i14); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx() local
497 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi14x01234567, vk14x01234567)); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx()
631 const __m256 vi14x01234567 = _mm256_maskload_ps(i14, vmask); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx() local
633 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi14x01234567, vk14x01234567)); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx()
|
| D | up16x25-minmax-fma3-acc2.c | 295 const __m256 vi14x01234567 = _mm256_loadu_ps(i14); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3_acc2() local
301 vacc01234567p0 = _mm256_fmadd_ps(vi14x01234567, vk14x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3_acc2()
496 const __m256 vi14x01234567 = _mm256_loadu_ps(i14); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3_acc2() local
500 vacc01234567p0 = _mm256_fmadd_ps(vi14x01234567, vk14x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3_acc2()
636 const __m256 vi14x01234567 = _mm256_maskload_ps(i14, vmask); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3_acc2() local
638 vacc01234567p0 = _mm256_fmadd_ps(vi14x01234567, vk14x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3_acc2()
|
| D | up16x25-minmax-fma3.c | 295 const __m256 vi14x01234567 = _mm256_loadu_ps(i14); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3() local
301 vacc01234567p0 = _mm256_fmadd_ps(vi14x01234567, vk14x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3()
493 const __m256 vi14x01234567 = _mm256_loadu_ps(i14); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3() local
497 vacc01234567p0 = _mm256_fmadd_ps(vi14x01234567, vk14x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3()
631 const __m256 vi14x01234567 = _mm256_maskload_ps(i14, vmask); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3() local
633 vacc01234567p0 = _mm256_fmadd_ps(vi14x01234567, vk14x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3()
|
| D | up8x25-minmax-avx-acc2.c | 252 const __m256 vi14x01234567 = _mm256_loadu_ps(i14); in xnn_f32_dwconv_minmax_ukernel_up8x25__avx_acc2() local
256 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi14x01234567, vk14x01234567)); in xnn_f32_dwconv_minmax_ukernel_up8x25__avx_acc2()
392 const __m256 vi14x01234567 = _mm256_maskload_ps(i14, vmask); in xnn_f32_dwconv_minmax_ukernel_up8x25__avx_acc2() local
394 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi14x01234567, vk14x01234567)); in xnn_f32_dwconv_minmax_ukernel_up8x25__avx_acc2()
|
| D | up16x25-minmax-avx-acc2.c | 295 const __m256 vi14x01234567 = _mm256_loadu_ps(i14); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx_acc2() local
301 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi14x01234567, vk14x01234567)); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx_acc2()
496 const __m256 vi14x01234567 = _mm256_loadu_ps(i14); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx_acc2() local
500 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi14x01234567, vk14x01234567)); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx_acc2()
636 const __m256 vi14x01234567 = _mm256_maskload_ps(i14, vmask); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx_acc2() local
638 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi14x01234567, vk14x01234567)); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx_acc2()
|
| /external/XNNPACK/src/qs8-dwconv/gen/ |
| D | up8x25-minmax-fp32-neon-mul16.c | 256 const int16x8_t vi14x01234567 = vmovl_s8(vld1_s8(i14)); i14 += 8; in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16() local
259 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi14x01234567), vget_low_s16(vk14x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
260 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi14x01234567), vget_high_s16(vk14x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
427 const int16x8_t vi14x01234567 = vmovl_s8(vld1_s8(i14)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16() local
430 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi14x01234567), vget_low_s16(vk14x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
431 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi14x01234567), vget_high_s16(vk14x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
|
| D | up8x25-minmax-rndnu-neon-mul16.c | 257 const int16x8_t vi14x01234567 = vmovl_s8(vld1_s8(i14)); i14 += 8; in xnn_qs8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16() local
260 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi14x01234567), vget_low_s16(vk14x01234567)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16()
261 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi14x01234567), vget_high_s16(vk14x01234567)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16()
427 const int16x8_t vi14x01234567 = vmovl_s8(vld1_s8(i14)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16() local
430 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi14x01234567), vget_low_s16(vk14x01234567)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16()
431 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi14x01234567), vget_high_s16(vk14x01234567)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16()
|
| D | up8x25-minmax-fp32-neonv8-mul16.c | 256 const int16x8_t vi14x01234567 = vmovl_s8(vld1_s8(i14)); i14 += 8; in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16() local
259 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi14x01234567), vget_low_s16(vk14x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
260 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi14x01234567), vget_high_s16(vk14x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
426 const int16x8_t vi14x01234567 = vmovl_s8(vld1_s8(i14)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16() local
429 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi14x01234567), vget_low_s16(vk14x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
430 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi14x01234567), vget_high_s16(vk14x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
|
| D | up8x25-minmax-fp32-sse2-mul16-add16.c | 329 const __m128i vi14x01234567 = _mm_loadl_epi64((const __m128i*) i14); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__sse2_mul16_add16() local
333 … const __m128i vxi14x01234567 = _mm_srai_epi16(_mm_unpacklo_epi8(vi14x01234567, vi14x01234567), 8); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__sse2_mul16_add16()
638 const __m128i vi14x01234567 = _mm_loadl_epi64((const __m128i*) i14); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__sse2_mul16_add16() local
641 … const __m128i vxi14x01234567 = _mm_srai_epi16(_mm_unpacklo_epi8(vi14x01234567, vi14x01234567), 8); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__sse2_mul16_add16()
|
| /external/XNNPACK/src/qc8-dwconv/gen/ |
| D | up8x25-minmax-fp32-neonv8-mul16.c | 255 const int16x8_t vi14x01234567 = vmovl_s8(vld1_s8(i14)); i14 += 8; in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16() local
258 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi14x01234567), vget_low_s16(vk14x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
259 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi14x01234567), vget_high_s16(vk14x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
428 const int16x8_t vi14x01234567 = vmovl_s8(vld1_s8(i14)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16() local
431 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi14x01234567), vget_low_s16(vk14x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
432 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi14x01234567), vget_high_s16(vk14x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
|
| D | up8x25-minmax-fp32-neon-mul16.c | 255 const int16x8_t vi14x01234567 = vmovl_s8(vld1_s8(i14)); i14 += 8; in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16() local
258 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi14x01234567), vget_low_s16(vk14x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
259 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi14x01234567), vget_high_s16(vk14x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
429 const int16x8_t vi14x01234567 = vmovl_s8(vld1_s8(i14)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16() local
432 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi14x01234567), vget_low_s16(vk14x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
433 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi14x01234567), vget_high_s16(vk14x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
|
| /external/XNNPACK/src/f16-dwconv/gen/ |
| D | up8x25-minmax-neonfp16arith.c | 226 const float16x8_t vi14x01234567 = vld1q_f16(i14); i14 += 8; in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith() local
228 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi14x01234567, vk14x01234567); in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
336 const float16x8_t vi14x01234567 = vld1q_f16(i14); in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith() local
338 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi14x01234567, vk14x01234567); in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
|
| D | up8x25-minmax-neonfp16arith-acc2.c | 226 const float16x8_t vi14x01234567 = vld1q_f16(i14); i14 += 8; in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith_acc2() local
228 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi14x01234567, vk14x01234567); in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith_acc2()
338 const float16x8_t vi14x01234567 = vld1q_f16(i14); in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith_acc2() local
340 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi14x01234567, vk14x01234567); in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith_acc2()
|
| D | up16x25-minmax-neonfp16arith-acc2.c | 269 const float16x8_t vi14x01234567 = vld1q_f16(i14); i14 += 8; in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith_acc2() local
273 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi14x01234567, vk14x01234567); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith_acc2()
418 const float16x8_t vi14x01234567 = vld1q_f16(i14); i14 += 8; in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith_acc2() local
420 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi14x01234567, vk14x01234567); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith_acc2()
530 const float16x8_t vi14x01234567 = vld1q_f16(i14); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith_acc2() local
532 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi14x01234567, vk14x01234567); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith_acc2()
|
| D | up16x25-minmax-neonfp16arith.c | 269 const float16x8_t vi14x01234567 = vld1q_f16(i14); i14 += 8; in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith() local
273 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi14x01234567, vk14x01234567); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith()
415 const float16x8_t vi14x01234567 = vld1q_f16(i14); i14 += 8; in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith() local
417 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi14x01234567, vk14x01234567); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith()
525 const float16x8_t vi14x01234567 = vld1q_f16(i14); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith() local
527 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi14x01234567, vk14x01234567); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith()
|
| D | up8x25-minmax-fma3-acc2.c | 255 const __m256 vi14x01234567 = _mm256_cvtph_ps(_mm_loadu_si128((const __m128i*) i14)); in xnn_f16_dwconv_minmax_ukernel_up8x25__fma3_acc2() local
259 …vacc01234567p0 = _mm256_cvtph_ps(_mm256_cvtps_ph(_mm256_fmadd_ps(vi14x01234567, vk14x01234567, vac… in xnn_f16_dwconv_minmax_ukernel_up8x25__fma3_acc2()
408 const __m256 vi14x01234567 = _mm256_cvtph_ps(_mm_loadu_si128((const __m128i*) i14)); in xnn_f16_dwconv_minmax_ukernel_up8x25__fma3_acc2() local
411 …vacc01234567p0 = _mm256_cvtph_ps(_mm256_cvtps_ph(_mm256_fmadd_ps(vi14x01234567, vk14x01234567, vac… in xnn_f16_dwconv_minmax_ukernel_up8x25__fma3_acc2()
|
| D | up8x25-minmax-fma3.c | 255 const __m256 vi14x01234567 = _mm256_cvtph_ps(_mm_loadu_si128((const __m128i*) i14)); in xnn_f16_dwconv_minmax_ukernel_up8x25__fma3() local
259 …vacc01234567p0 = _mm256_cvtph_ps(_mm256_cvtps_ph(_mm256_fmadd_ps(vi14x01234567, vk14x01234567, vac… in xnn_f16_dwconv_minmax_ukernel_up8x25__fma3()
406 const __m256 vi14x01234567 = _mm256_cvtph_ps(_mm_loadu_si128((const __m128i*) i14)); in xnn_f16_dwconv_minmax_ukernel_up8x25__fma3() local
409 …vacc01234567p0 = _mm256_cvtph_ps(_mm256_cvtps_ph(_mm256_fmadd_ps(vi14x01234567, vk14x01234567, vac… in xnn_f16_dwconv_minmax_ukernel_up8x25__fma3()
|
| /external/XNNPACK/src/qu8-dwconv/gen/ |
| D | up8x25-minmax-fp32-neonv8-mul16.c | 257 const int16x8_t vi14x01234567 = vreinterpretq_s16_u16(vmovl_u8(vld1_u8(i14))); i14 += 8; in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16() local
260 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi14x01234567), vget_low_s16(vk14x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
261 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi14x01234567), vget_high_s16(vk14x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
427 const int16x8_t vi14x01234567 = vreinterpretq_s16_u16(vmovl_u8(vld1_u8(i14))); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16() local
430 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi14x01234567), vget_low_s16(vk14x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
431 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi14x01234567), vget_high_s16(vk14x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
|
| D | up8x25-minmax-fp32-neon-mul16.c | 257 const int16x8_t vi14x01234567 = vreinterpretq_s16_u16(vmovl_u8(vld1_u8(i14))); i14 += 8; in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16() local
260 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi14x01234567), vget_low_s16(vk14x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
261 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi14x01234567), vget_high_s16(vk14x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
428 const int16x8_t vi14x01234567 = vreinterpretq_s16_u16(vmovl_u8(vld1_u8(i14))); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16() local
431 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi14x01234567), vget_low_s16(vk14x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
432 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi14x01234567), vget_high_s16(vk14x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
|
| D | up8x25-minmax-rndnu-neon-mul8.c | 287 const uint8x8_t vi14x01234567 = vld1_u8(i14); i14 += 8; in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul8() local
290 vprod01234567 = vmull_u8(vi14x01234567, vk14x01234567); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul8()
291 vsum01234567 = vaddw_u8(vsum01234567, vi14x01234567); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul8()
479 const int16x8_t vi14x01234567 = vreinterpretq_s16_u16(vmovl_u8(vld1_u8(i14))); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul8() local
482 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi14x01234567), vget_low_s16(vk14x01234567)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul8()
483 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi14x01234567), vget_high_s16(vk14x01234567)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul8()
|
| D | up8x25-minmax-fp32-wasmsimd-mul16.c | 307 const v128_t vi14x01234567 = wasm_u16x8_load8x8(i14); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__wasmsimd_mul16() local
309 vsumx01234567 = wasm_i16x8_add(vsumx01234567, vi14x01234567); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__wasmsimd_mul16()
312 vprod01234567 = wasm_i16x8_mul(vi14x01234567, vk14x01234567); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__wasmsimd_mul16()
582 const v128_t vi14x01234567 = wasm_u16x8_load8x8(i14); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__wasmsimd_mul16() local
584 vsumx01234567 = wasm_i16x8_add(vsumx01234567, vi14x01234567); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__wasmsimd_mul16()
586 vprod01234567 = wasm_i16x8_mul(vi14x01234567, vk14x01234567); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__wasmsimd_mul16()
|
| D | up8x25-minmax-rndnu-neon-mul16.c | 258 const int16x8_t vi14x01234567 = vreinterpretq_s16_u16(vmovl_u8(vld1_u8(i14))); i14 += 8; in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16() local
261 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi14x01234567), vget_low_s16(vk14x01234567)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16()
262 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi14x01234567), vget_high_s16(vk14x01234567)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16()
428 const int16x8_t vi14x01234567 = vreinterpretq_s16_u16(vmovl_u8(vld1_u8(i14))); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16() local
431 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi14x01234567), vget_low_s16(vk14x01234567)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16()
432 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi14x01234567), vget_high_s16(vk14x01234567)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16()
|