/external/XNNPACK/src/qs8-dwconv/gen/ |
D | up16x9-minmax-neon-mul16.c | 177 const int16x8_t vi8x89ABCDEF = vmovl_s8(vld1_s8(i8)); i8 += 8; in xnn_qs8_dwconv_minmax_ukernel_up16x9__neon_mul16() local 182 vacc89AB = vmlal_s16(vacc89AB, vget_low_s16(vi8x89ABCDEF), vget_low_s16(vk8x89ABCDEF)); in xnn_qs8_dwconv_minmax_ukernel_up16x9__neon_mul16() 183 vaccCDEF = vmlal_s16(vaccCDEF, vget_high_s16(vi8x89ABCDEF), vget_high_s16(vk8x89ABCDEF)); in xnn_qs8_dwconv_minmax_ukernel_up16x9__neon_mul16()
|
D | up24x9-minmax-neon-mul16.c | 211 const int16x8_t vi8x89ABCDEF = vmovl_s8(vld1_s8(i8)); i8 += 8; in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16() local 218 vacc89AB = vmlal_s16(vacc89AB, vget_low_s16(vi8x89ABCDEF), vget_low_s16(vk8x89ABCDEF)); in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16() 219 vaccCDEF = vmlal_s16(vaccCDEF, vget_high_s16(vi8x89ABCDEF), vget_high_s16(vk8x89ABCDEF)); in xnn_qs8_dwconv_minmax_ukernel_up24x9__neon_mul16()
|
D | up16x9-minmax-avx2-mul32.c | 161 const __m256i vi8x89ABCDEF = _mm256_cvtepi8_epi32(_mm_loadl_epi64((const __m128i*) (i8 + 8))); in xnn_qs8_dwconv_minmax_ukernel_up16x9__avx2_mul32() local 166 vacc89ABCDEF = _mm256_add_epi32(vacc89ABCDEF, _mm256_mullo_epi32(vi8x89ABCDEF, vk8x89ABCDEF)); in xnn_qs8_dwconv_minmax_ukernel_up16x9__avx2_mul32()
|
D | up32x9-minmax-neon-mul16.c | 245 const int16x8_t vi8x89ABCDEF = vmovl_s8(vld1_s8(i8)); i8 += 8; in xnn_qs8_dwconv_minmax_ukernel_up32x9__neon_mul16() local 254 vacc89AB = vmlal_s16(vacc89AB, vget_low_s16(vi8x89ABCDEF), vget_low_s16(vk8x89ABCDEF)); in xnn_qs8_dwconv_minmax_ukernel_up32x9__neon_mul16() 255 vaccCDEF = vmlal_s16(vaccCDEF, vget_high_s16(vi8x89ABCDEF), vget_high_s16(vk8x89ABCDEF)); in xnn_qs8_dwconv_minmax_ukernel_up32x9__neon_mul16()
|
D | up24x9-minmax-avx2-mul32.c | 186 const __m256i vi8x89ABCDEF = _mm256_cvtepi8_epi32(_mm_loadl_epi64((const __m128i*) (i8 + 8))); in xnn_qs8_dwconv_minmax_ukernel_up24x9__avx2_mul32() local 193 vacc89ABCDEF = _mm256_add_epi32(vacc89ABCDEF, _mm256_mullo_epi32(vi8x89ABCDEF, vk8x89ABCDEF)); in xnn_qs8_dwconv_minmax_ukernel_up24x9__avx2_mul32()
|
D | up16x9-minmax-wasmsimd-mul16.c | 203 const v128_t vi8x89ABCDEF = wasm_i16x8_load_8x8(i8 + 8); in xnn_qs8_dwconv_minmax_ukernel_up16x9__wasmsimd_mul16() local 208 const v128_t vprod8x89ABCDEF = wasm_i16x8_mul(vi8x89ABCDEF, vk8x89ABCDEF); in xnn_qs8_dwconv_minmax_ukernel_up16x9__wasmsimd_mul16()
|
D | up16x9-minmax-ssse3-mul16.c | 259 const __m128i vi8x89ABCDEF = _mm_loadl_epi64((const __m128i*) (i8 + 8)); in xnn_qs8_dwconv_minmax_ukernel_up16x9__ssse3_mul16() local 265 …st __m128i vxi8x89ABCDEF = _mm_unpacklo_epi8(vi8x89ABCDEF, _mm_cmpgt_epi8(_mm_setzero_si128(), vi8… in xnn_qs8_dwconv_minmax_ukernel_up16x9__ssse3_mul16()
|
D | up16x9-minmax-sse2-mul16.c | 259 const __m128i vi8x89ABCDEF = _mm_loadl_epi64((const __m128i*) (i8 + 8)); in xnn_qs8_dwconv_minmax_ukernel_up16x9__sse2_mul16() local 265 …st __m128i vxi8x89ABCDEF = _mm_unpacklo_epi8(vi8x89ABCDEF, _mm_cmpgt_epi8(_mm_setzero_si128(), vi8… in xnn_qs8_dwconv_minmax_ukernel_up16x9__sse2_mul16()
|
D | up32x9-minmax-avx2-mul32.c | 211 const __m256i vi8x89ABCDEF = _mm256_cvtepi8_epi32(_mm_loadl_epi64((const __m128i*) (i8 + 8))); in xnn_qs8_dwconv_minmax_ukernel_up32x9__avx2_mul32() local 220 vacc89ABCDEF = _mm256_add_epi32(vacc89ABCDEF, _mm256_mullo_epi32(vi8x89ABCDEF, vk8x89ABCDEF)); in xnn_qs8_dwconv_minmax_ukernel_up32x9__avx2_mul32()
|
D | up24x9-minmax-wasmsimd-mul16.c | 245 const v128_t vi8x89ABCDEF = wasm_i16x8_load_8x8(i8 + 8); in xnn_qs8_dwconv_minmax_ukernel_up24x9__wasmsimd_mul16() local 252 const v128_t vprod8x89ABCDEF = wasm_i16x8_mul(vi8x89ABCDEF, vk8x89ABCDEF); in xnn_qs8_dwconv_minmax_ukernel_up24x9__wasmsimd_mul16()
|
D | up16x9-minmax-sse41-mul16.c | 261 const __m128i vi8x89ABCDEF = _mm_loadl_epi64((const __m128i*) (i8 + 8)); in xnn_qs8_dwconv_minmax_ukernel_up16x9__sse41_mul16() local 262 const __m128i vxi8x89ABCDEF = _mm_cvtepi8_epi16(vi8x89ABCDEF); in xnn_qs8_dwconv_minmax_ukernel_up16x9__sse41_mul16()
|
D | up24x9-minmax-ssse3-mul16.c | 325 const __m128i vi8x89ABCDEF = _mm_loadl_epi64((const __m128i*) (i8 + 8)); in xnn_qs8_dwconv_minmax_ukernel_up24x9__ssse3_mul16() local 333 …st __m128i vxi8x89ABCDEF = _mm_unpacklo_epi8(vi8x89ABCDEF, _mm_cmpgt_epi8(_mm_setzero_si128(), vi8… in xnn_qs8_dwconv_minmax_ukernel_up24x9__ssse3_mul16()
|
D | up24x9-minmax-sse2-mul16.c | 325 const __m128i vi8x89ABCDEF = _mm_loadl_epi64((const __m128i*) (i8 + 8)); in xnn_qs8_dwconv_minmax_ukernel_up24x9__sse2_mul16() local 333 …st __m128i vxi8x89ABCDEF = _mm_unpacklo_epi8(vi8x89ABCDEF, _mm_cmpgt_epi8(_mm_setzero_si128(), vi8… in xnn_qs8_dwconv_minmax_ukernel_up24x9__sse2_mul16()
|
D | up24x9-minmax-sse41-mul16.c | 327 const __m128i vi8x89ABCDEF = _mm_loadl_epi64((const __m128i*) (i8 + 8)); in xnn_qs8_dwconv_minmax_ukernel_up24x9__sse41_mul16() local 328 const __m128i vxi8x89ABCDEF = _mm_cvtepi8_epi16(vi8x89ABCDEF); in xnn_qs8_dwconv_minmax_ukernel_up24x9__sse41_mul16()
|
/external/XNNPACK/src/f32-dwconv/gen/ |
D | up16x9-minmax-fma3.c | 164 const __m256 vi8x89ABCDEF = _mm256_loadu_ps(i8 + 8); in xnn_f32_dwconv_minmax_ukernel_up16x9__fma3() local 170 vacc89ABCDEFp0 = _mm256_fmadd_ps(vi8x89ABCDEF, vk8x89ABCDEF, vacc89ABCDEFp0); in xnn_f32_dwconv_minmax_ukernel_up16x9__fma3()
|
D | up16x9-minmax-fma3-acc2.c | 164 const __m256 vi8x89ABCDEF = _mm256_loadu_ps(i8 + 8); in xnn_f32_dwconv_minmax_ukernel_up16x9__fma3_acc2() local 170 vacc89ABCDEFp0 = _mm256_fmadd_ps(vi8x89ABCDEF, vk8x89ABCDEF, vacc89ABCDEFp0); in xnn_f32_dwconv_minmax_ukernel_up16x9__fma3_acc2()
|
D | up16x9-minmax-avx.c | 164 const __m256 vi8x89ABCDEF = _mm256_loadu_ps(i8 + 8); in xnn_f32_dwconv_minmax_ukernel_up16x9__avx() local 170 vacc89ABCDEFp0 = _mm256_add_ps(vacc89ABCDEFp0, _mm256_mul_ps(vi8x89ABCDEF, vk8x89ABCDEF)); in xnn_f32_dwconv_minmax_ukernel_up16x9__avx()
|
D | up16x25-minmax-fma3.c | 244 const __m256 vi8x89ABCDEF = _mm256_loadu_ps(i8 + 8); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3() local 250 vacc89ABCDEFp0 = _mm256_fmadd_ps(vi8x89ABCDEF, vk8x89ABCDEF, vacc89ABCDEFp0); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3()
|
D | up16x25-minmax-fma3-acc2.c | 244 const __m256 vi8x89ABCDEF = _mm256_loadu_ps(i8 + 8); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3_acc2() local 250 vacc89ABCDEFp0 = _mm256_fmadd_ps(vi8x89ABCDEF, vk8x89ABCDEF, vacc89ABCDEFp0); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3_acc2()
|
D | up16x25-minmax-avx.c | 244 const __m256 vi8x89ABCDEF = _mm256_loadu_ps(i8 + 8); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx() local 250 vacc89ABCDEFp0 = _mm256_add_ps(vacc89ABCDEFp0, _mm256_mul_ps(vi8x89ABCDEF, vk8x89ABCDEF)); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx()
|
D | up16x25-minmax-avx-acc2.c | 244 const __m256 vi8x89ABCDEF = _mm256_loadu_ps(i8 + 8); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx_acc2() local 250 vacc89ABCDEFp0 = _mm256_add_ps(vacc89ABCDEFp0, _mm256_mul_ps(vi8x89ABCDEF, vk8x89ABCDEF)); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx_acc2()
|
/external/XNNPACK/src/f16-dwconv/gen/ |
D | up16x9-minmax-neonfp16arith.c | 148 const float16x8_t vi8x89ABCDEF = vld1q_f16(i8); i8 += 8; in xnn_f16_dwconv_minmax_ukernel_up16x9__neonfp16arith() local 152 vacc89ABCDEFp0 = vfmaq_f16(vacc89ABCDEFp0, vi8x89ABCDEF, vk8x89ABCDEF); in xnn_f16_dwconv_minmax_ukernel_up16x9__neonfp16arith()
|
D | up16x9-minmax-neonfp16arith-acc2.c | 148 const float16x8_t vi8x89ABCDEF = vld1q_f16(i8); i8 += 8; in xnn_f16_dwconv_minmax_ukernel_up16x9__neonfp16arith_acc2() local 152 vacc89ABCDEFp0 = vfmaq_f16(vacc89ABCDEFp0, vi8x89ABCDEF, vk8x89ABCDEF); in xnn_f16_dwconv_minmax_ukernel_up16x9__neonfp16arith_acc2()
|
D | up16x25-minmax-neonfp16arith-acc2.c | 228 const float16x8_t vi8x89ABCDEF = vld1q_f16(i8); i8 += 8; in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith_acc2() local 232 vacc89ABCDEFp0 = vfmaq_f16(vacc89ABCDEFp0, vi8x89ABCDEF, vk8x89ABCDEF); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith_acc2()
|
D | up16x25-minmax-neonfp16arith.c | 228 const float16x8_t vi8x89ABCDEF = vld1q_f16(i8); i8 += 8; in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith() local 232 vacc89ABCDEFp0 = vfmaq_f16(vacc89ABCDEFp0, vi8x89ABCDEF, vk8x89ABCDEF); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith()
|