| /external/XNNPACK/src/f32-dwconv/gen/ |
| D | up8x25-minmax-avx.c | 264 const __m256 vi16x01234567 = _mm256_loadu_ps(i16); in xnn_f32_dwconv_minmax_ukernel_up8x25__avx() local
268 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi16x01234567, vk16x01234567)); in xnn_f32_dwconv_minmax_ukernel_up8x25__avx()
398 const __m256 vi16x01234567 = _mm256_maskload_ps(i16, vmask); in xnn_f32_dwconv_minmax_ukernel_up8x25__avx() local
400 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi16x01234567, vk16x01234567)); in xnn_f32_dwconv_minmax_ukernel_up8x25__avx()
|
| D | up8x25-minmax-fma3-acc2.c | 264 const __m256 vi16x01234567 = _mm256_loadu_ps(i16); in xnn_f32_dwconv_minmax_ukernel_up8x25__fma3_acc2() local
268 vacc01234567p0 = _mm256_fmadd_ps(vi16x01234567, vk16x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up8x25__fma3_acc2()
400 const __m256 vi16x01234567 = _mm256_maskload_ps(i16, vmask); in xnn_f32_dwconv_minmax_ukernel_up8x25__fma3_acc2() local
402 vacc01234567p0 = _mm256_fmadd_ps(vi16x01234567, vk16x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up8x25__fma3_acc2()
|
| D | up8x25-minmax-fma3.c | 264 const __m256 vi16x01234567 = _mm256_loadu_ps(i16); in xnn_f32_dwconv_minmax_ukernel_up8x25__fma3() local
268 vacc01234567p0 = _mm256_fmadd_ps(vi16x01234567, vk16x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up8x25__fma3()
398 const __m256 vi16x01234567 = _mm256_maskload_ps(i16, vmask); in xnn_f32_dwconv_minmax_ukernel_up8x25__fma3() local
400 vacc01234567p0 = _mm256_fmadd_ps(vi16x01234567, vk16x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up8x25__fma3()
|
| D | up16x25-minmax-avx.c | 313 const __m256 vi16x01234567 = _mm256_loadu_ps(i16); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx() local
319 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi16x01234567, vk16x01234567)); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx()
505 const __m256 vi16x01234567 = _mm256_loadu_ps(i16); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx() local
509 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi16x01234567, vk16x01234567)); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx()
639 const __m256 vi16x01234567 = _mm256_maskload_ps(i16, vmask); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx() local
641 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi16x01234567, vk16x01234567)); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx()
|
| D | up16x25-minmax-fma3-acc2.c | 313 const __m256 vi16x01234567 = _mm256_loadu_ps(i16); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3_acc2() local
319 vacc01234567p0 = _mm256_fmadd_ps(vi16x01234567, vk16x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3_acc2()
508 const __m256 vi16x01234567 = _mm256_loadu_ps(i16); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3_acc2() local
512 vacc01234567p0 = _mm256_fmadd_ps(vi16x01234567, vk16x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3_acc2()
644 const __m256 vi16x01234567 = _mm256_maskload_ps(i16, vmask); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3_acc2() local
646 vacc01234567p0 = _mm256_fmadd_ps(vi16x01234567, vk16x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3_acc2()
|
| D | up16x25-minmax-fma3.c | 313 const __m256 vi16x01234567 = _mm256_loadu_ps(i16); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3() local
319 vacc01234567p0 = _mm256_fmadd_ps(vi16x01234567, vk16x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3()
505 const __m256 vi16x01234567 = _mm256_loadu_ps(i16); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3() local
509 vacc01234567p0 = _mm256_fmadd_ps(vi16x01234567, vk16x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3()
639 const __m256 vi16x01234567 = _mm256_maskload_ps(i16, vmask); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3() local
641 vacc01234567p0 = _mm256_fmadd_ps(vi16x01234567, vk16x01234567, vacc01234567p0); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3()
|
| D | up8x25-minmax-avx-acc2.c | 264 const __m256 vi16x01234567 = _mm256_loadu_ps(i16); in xnn_f32_dwconv_minmax_ukernel_up8x25__avx_acc2() local
268 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi16x01234567, vk16x01234567)); in xnn_f32_dwconv_minmax_ukernel_up8x25__avx_acc2()
400 const __m256 vi16x01234567 = _mm256_maskload_ps(i16, vmask); in xnn_f32_dwconv_minmax_ukernel_up8x25__avx_acc2() local
402 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi16x01234567, vk16x01234567)); in xnn_f32_dwconv_minmax_ukernel_up8x25__avx_acc2()
|
| D | up16x25-minmax-avx-acc2.c | 313 const __m256 vi16x01234567 = _mm256_loadu_ps(i16); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx_acc2() local
319 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi16x01234567, vk16x01234567)); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx_acc2()
508 const __m256 vi16x01234567 = _mm256_loadu_ps(i16); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx_acc2() local
512 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi16x01234567, vk16x01234567)); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx_acc2()
644 const __m256 vi16x01234567 = _mm256_maskload_ps(i16, vmask); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx_acc2() local
646 vacc01234567p0 = _mm256_add_ps(vacc01234567p0, _mm256_mul_ps(vi16x01234567, vk16x01234567)); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx_acc2()
|
| /external/XNNPACK/src/qs8-dwconv/gen/ |
| D | up8x25-minmax-fp32-neon-mul16.c | 268 const int16x8_t vi16x01234567 = vmovl_s8(vld1_s8(i16)); i16 += 8; in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16() local
271 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi16x01234567), vget_low_s16(vk16x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
272 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi16x01234567), vget_high_s16(vk16x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
437 const int16x8_t vi16x01234567 = vmovl_s8(vld1_s8(i16)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16() local
440 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi16x01234567), vget_low_s16(vk16x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
441 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi16x01234567), vget_high_s16(vk16x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
|
| D | up8x25-minmax-rndnu-neon-mul16.c | 269 const int16x8_t vi16x01234567 = vmovl_s8(vld1_s8(i16)); i16 += 8; in xnn_qs8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16() local
272 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi16x01234567), vget_low_s16(vk16x01234567)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16()
273 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi16x01234567), vget_high_s16(vk16x01234567)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16()
437 const int16x8_t vi16x01234567 = vmovl_s8(vld1_s8(i16)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16() local
440 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi16x01234567), vget_low_s16(vk16x01234567)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16()
441 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi16x01234567), vget_high_s16(vk16x01234567)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16()
|
| D | up8x25-minmax-fp32-neonv8-mul16.c | 268 const int16x8_t vi16x01234567 = vmovl_s8(vld1_s8(i16)); i16 += 8; in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16() local
271 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi16x01234567), vget_low_s16(vk16x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
272 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi16x01234567), vget_high_s16(vk16x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
436 const int16x8_t vi16x01234567 = vmovl_s8(vld1_s8(i16)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16() local
439 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi16x01234567), vget_low_s16(vk16x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
440 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi16x01234567), vget_high_s16(vk16x01234567)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
|
| D | up8x25-minmax-fp32-sse2-mul16-add16.c | 352 const __m128i vi16x01234567 = _mm_loadl_epi64((const __m128i*) i16); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__sse2_mul16_add16() local
356 … const __m128i vxi16x01234567 = _mm_srai_epi16(_mm_unpacklo_epi8(vi16x01234567, vi16x01234567), 8); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__sse2_mul16_add16()
659 const __m128i vi16x01234567 = _mm_loadl_epi64((const __m128i*) i16); in xnn_qs8_dwconv_minmax_fp32_ukernel_up8x25__sse2_mul16_add16() local
662 … const __m128i vxi16x01234567 = _mm_srai_epi16(_mm_unpacklo_epi8(vi16x01234567, vi16x01234567), 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 | 267 const int16x8_t vi16x01234567 = vmovl_s8(vld1_s8(i16)); i16 += 8; in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16() local
270 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi16x01234567), vget_low_s16(vk16x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
271 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi16x01234567), vget_high_s16(vk16x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
438 const int16x8_t vi16x01234567 = vmovl_s8(vld1_s8(i16)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16() local
441 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi16x01234567), vget_low_s16(vk16x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
442 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi16x01234567), vget_high_s16(vk16x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
|
| D | up8x25-minmax-fp32-neon-mul16.c | 267 const int16x8_t vi16x01234567 = vmovl_s8(vld1_s8(i16)); i16 += 8; in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16() local
270 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi16x01234567), vget_low_s16(vk16x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
271 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi16x01234567), vget_high_s16(vk16x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
439 const int16x8_t vi16x01234567 = vmovl_s8(vld1_s8(i16)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16() local
442 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi16x01234567), vget_low_s16(vk16x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
443 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi16x01234567), vget_high_s16(vk16x01234567)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
|
| /external/XNNPACK/src/f16-dwconv/gen/ |
| D | up8x25-minmax-neonfp16arith.c | 234 const float16x8_t vi16x01234567 = vld1q_f16(i16); i16 += 8; in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith() local
236 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi16x01234567, vk16x01234567); in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
344 const float16x8_t vi16x01234567 = vld1q_f16(i16); in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith() local
346 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi16x01234567, vk16x01234567); in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith()
|
| D | up8x25-minmax-neonfp16arith-acc2.c | 234 const float16x8_t vi16x01234567 = vld1q_f16(i16); i16 += 8; in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith_acc2() local
236 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi16x01234567, vk16x01234567); in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith_acc2()
346 const float16x8_t vi16x01234567 = vld1q_f16(i16); in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith_acc2() local
348 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi16x01234567, vk16x01234567); in xnn_f16_dwconv_minmax_ukernel_up8x25__neonfp16arith_acc2()
|
| D | up16x25-minmax-neonfp16arith-acc2.c | 283 const float16x8_t vi16x01234567 = vld1q_f16(i16); i16 += 8; in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith_acc2() local
287 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi16x01234567, vk16x01234567); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith_acc2()
426 const float16x8_t vi16x01234567 = vld1q_f16(i16); i16 += 8; in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith_acc2() local
428 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi16x01234567, vk16x01234567); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith_acc2()
538 const float16x8_t vi16x01234567 = vld1q_f16(i16); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith_acc2() local
540 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi16x01234567, vk16x01234567); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith_acc2()
|
| D | up16x25-minmax-neonfp16arith.c | 283 const float16x8_t vi16x01234567 = vld1q_f16(i16); i16 += 8; in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith() local
287 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi16x01234567, vk16x01234567); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith()
423 const float16x8_t vi16x01234567 = vld1q_f16(i16); i16 += 8; in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith() local
425 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi16x01234567, vk16x01234567); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith()
533 const float16x8_t vi16x01234567 = vld1q_f16(i16); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith() local
535 vacc01234567p0 = vfmaq_f16(vacc01234567p0, vi16x01234567, vk16x01234567); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith()
|
| D | up8x25-minmax-fma3-acc2.c | 267 const __m256 vi16x01234567 = _mm256_cvtph_ps(_mm_loadu_si128((const __m128i*) i16)); in xnn_f16_dwconv_minmax_ukernel_up8x25__fma3_acc2() local
271 …vacc01234567p0 = _mm256_cvtph_ps(_mm256_cvtps_ph(_mm256_fmadd_ps(vi16x01234567, vk16x01234567, vac… in xnn_f16_dwconv_minmax_ukernel_up8x25__fma3_acc2()
418 const __m256 vi16x01234567 = _mm256_cvtph_ps(_mm_loadu_si128((const __m128i*) i16)); in xnn_f16_dwconv_minmax_ukernel_up8x25__fma3_acc2() local
421 …vacc01234567p0 = _mm256_cvtph_ps(_mm256_cvtps_ph(_mm256_fmadd_ps(vi16x01234567, vk16x01234567, vac… in xnn_f16_dwconv_minmax_ukernel_up8x25__fma3_acc2()
|
| D | up8x25-minmax-fma3.c | 267 const __m256 vi16x01234567 = _mm256_cvtph_ps(_mm_loadu_si128((const __m128i*) i16)); in xnn_f16_dwconv_minmax_ukernel_up8x25__fma3() local
271 …vacc01234567p0 = _mm256_cvtph_ps(_mm256_cvtps_ph(_mm256_fmadd_ps(vi16x01234567, vk16x01234567, vac… in xnn_f16_dwconv_minmax_ukernel_up8x25__fma3()
416 const __m256 vi16x01234567 = _mm256_cvtph_ps(_mm_loadu_si128((const __m128i*) i16)); in xnn_f16_dwconv_minmax_ukernel_up8x25__fma3() local
419 …vacc01234567p0 = _mm256_cvtph_ps(_mm256_cvtps_ph(_mm256_fmadd_ps(vi16x01234567, vk16x01234567, vac… in xnn_f16_dwconv_minmax_ukernel_up8x25__fma3()
|
| /external/XNNPACK/src/qu8-dwconv/gen/ |
| D | up8x25-minmax-fp32-neonv8-mul16.c | 269 const int16x8_t vi16x01234567 = vreinterpretq_s16_u16(vmovl_u8(vld1_u8(i16))); i16 += 8; in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16() local
272 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi16x01234567), vget_low_s16(vk16x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
273 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi16x01234567), vget_high_s16(vk16x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
437 const int16x8_t vi16x01234567 = vreinterpretq_s16_u16(vmovl_u8(vld1_u8(i16))); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16() local
440 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi16x01234567), vget_low_s16(vk16x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
441 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi16x01234567), vget_high_s16(vk16x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neonv8_mul16()
|
| D | up8x25-minmax-fp32-neon-mul16.c | 269 const int16x8_t vi16x01234567 = vreinterpretq_s16_u16(vmovl_u8(vld1_u8(i16))); i16 += 8; in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16() local
272 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi16x01234567), vget_low_s16(vk16x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
273 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi16x01234567), vget_high_s16(vk16x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
438 const int16x8_t vi16x01234567 = vreinterpretq_s16_u16(vmovl_u8(vld1_u8(i16))); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16() local
441 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi16x01234567), vget_low_s16(vk16x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
442 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi16x01234567), vget_high_s16(vk16x01234567)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__neon_mul16()
|
| D | up8x25-minmax-rndnu-neon-mul8.c | 303 const uint8x8_t vi16x01234567 = vld1_u8(i16); i16 += 8; in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul8() local
306 vprod01234567 = vmull_u8(vi16x01234567, vk16x01234567); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul8()
307 vsum01234567 = vaddw_u8(vsum01234567, vi16x01234567); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul8()
489 const int16x8_t vi16x01234567 = vreinterpretq_s16_u16(vmovl_u8(vld1_u8(i16))); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul8() local
492 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi16x01234567), vget_low_s16(vk16x01234567)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul8()
493 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi16x01234567), vget_high_s16(vk16x01234567)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul8()
|
| D | up8x25-minmax-fp32-wasmsimd-mul16.c | 327 const v128_t vi16x01234567 = wasm_u16x8_load8x8(i16); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__wasmsimd_mul16() local
329 vsumx01234567 = wasm_i16x8_add(vsumx01234567, vi16x01234567); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__wasmsimd_mul16()
332 vprod01234567 = wasm_i16x8_mul(vi16x01234567, vk16x01234567); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__wasmsimd_mul16()
600 const v128_t vi16x01234567 = wasm_u16x8_load8x8(i16); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__wasmsimd_mul16() local
602 vsumx01234567 = wasm_i16x8_add(vsumx01234567, vi16x01234567); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__wasmsimd_mul16()
604 vprod01234567 = wasm_i16x8_mul(vi16x01234567, vk16x01234567); in xnn_qu8_dwconv_minmax_fp32_ukernel_up8x25__wasmsimd_mul16()
|
| D | up8x25-minmax-rndnu-neon-mul16.c | 270 const int16x8_t vi16x01234567 = vreinterpretq_s16_u16(vmovl_u8(vld1_u8(i16))); i16 += 8; in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16() local
273 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi16x01234567), vget_low_s16(vk16x01234567)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16()
274 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi16x01234567), vget_high_s16(vk16x01234567)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16()
438 const int16x8_t vi16x01234567 = vreinterpretq_s16_u16(vmovl_u8(vld1_u8(i16))); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16() local
441 vacc0123 = vmlal_s16(vacc0123, vget_low_s16(vi16x01234567), vget_low_s16(vk16x01234567)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16()
442 vacc4567 = vmlal_s16(vacc4567, vget_high_s16(vi16x01234567), vget_high_s16(vk16x01234567)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up8x25__neon_mul16()
|