/external/XNNPACK/src/qs8-dwconv/gen/ |
D | up16x25-minmax-rndnu-neon-mul16.c | 348 … const int16x8_t vk17x89ABCDEF = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up16x25__neon_mul16() local 352 vacc89AB = vmlal_s16(vacc89AB, vget_low_s16(vi17x89ABCDEF), vget_low_s16(vk17x89ABCDEF)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up16x25__neon_mul16() 353 vaccCDEF = vmlal_s16(vaccCDEF, vget_high_s16(vi17x89ABCDEF), vget_high_s16(vk17x89ABCDEF)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up16x25__neon_mul16()
|
D | up16x25-minmax-fp32-neonv8-mul16.c | 347 … const int16x8_t vk17x89ABCDEF = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qs8_dwconv_minmax_fp32_ukernel_up16x25__neonv8_mul16() local 351 vacc89AB = vmlal_s16(vacc89AB, vget_low_s16(vi17x89ABCDEF), vget_low_s16(vk17x89ABCDEF)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up16x25__neonv8_mul16() 352 vaccCDEF = vmlal_s16(vaccCDEF, vget_high_s16(vi17x89ABCDEF), vget_high_s16(vk17x89ABCDEF)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up16x25__neonv8_mul16()
|
D | up16x25-minmax-fp32-neon-mul16.c | 347 … const int16x8_t vk17x89ABCDEF = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qs8_dwconv_minmax_fp32_ukernel_up16x25__neon_mul16() local 351 vacc89AB = vmlal_s16(vacc89AB, vget_low_s16(vi17x89ABCDEF), vget_low_s16(vk17x89ABCDEF)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up16x25__neon_mul16() 352 vaccCDEF = vmlal_s16(vaccCDEF, vget_high_s16(vi17x89ABCDEF), vget_high_s16(vk17x89ABCDEF)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up16x25__neon_mul16()
|
D | up16x25-minmax-fp32-wasmsimd-mul16-add16.c | 374 …const v128_t vk17x89ABCDEF = wasm_i16x8_load8x8((const void*) ((uintptr_t) w + 16 * sizeof(int32_t… in xnn_qs8_dwconv_minmax_fp32_ukernel_up16x25__wasmsimd_mul16_add16() local 378 vprod89ABCDEF = wasm_i16x8_add(vprod89ABCDEF, wasm_i16x8_mul(vi17x89ABCDEF, vk17x89ABCDEF)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up16x25__wasmsimd_mul16_add16()
|
D | up16x25-minmax-rndnu-neon-mla8-ld64.c | 345 const int8x8_t vk17x89ABCDEF = vld1_s8(w); w = (const void*) ((const int8_t*) w + 8); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up16x25__neon_mla8_ld64() local 348 vprod89ABCDEF = vmlal_s8(vprod89ABCDEF, vi17x89ABCDEF, vk17x89ABCDEF); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up16x25__neon_mla8_ld64()
|
D | up16x25-minmax-fp32-avx2-mul32.c | 324 …const __m256i vk17x89ABCDEF = _mm256_cvtepi8_epi32(_mm_loadl_epi64((const __m128i*) ((uintptr_t) w… in xnn_qs8_dwconv_minmax_fp32_ukernel_up16x25__avx2_mul32() local 328 … vacc89ABCDEF = _mm256_add_epi32(vacc89ABCDEF, _mm256_mullo_epi32(vi17x89ABCDEF, vk17x89ABCDEF)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up16x25__avx2_mul32()
|
D | up24x25-minmax-rndnu-neon-mul16.c | 418 … const int16x8_t vk17x89ABCDEF = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up24x25__neon_mul16() local 424 vacc89AB = vmlal_s16(vacc89AB, vget_low_s16(vi17x89ABCDEF), vget_low_s16(vk17x89ABCDEF)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up24x25__neon_mul16() 425 vaccCDEF = vmlal_s16(vaccCDEF, vget_high_s16(vi17x89ABCDEF), vget_high_s16(vk17x89ABCDEF)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up24x25__neon_mul16()
|
/external/XNNPACK/src/qc8-dwconv/gen/ |
D | up16x25-minmax-fp32-neon-mul16.c | 346 … const int16x8_t vk17x89ABCDEF = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qc8_dwconv_minmax_fp32_ukernel_up16x25__neon_mul16() local 350 vacc89AB = vmlal_s16(vacc89AB, vget_low_s16(vi17x89ABCDEF), vget_low_s16(vk17x89ABCDEF)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up16x25__neon_mul16() 351 vaccCDEF = vmlal_s16(vaccCDEF, vget_high_s16(vi17x89ABCDEF), vget_high_s16(vk17x89ABCDEF)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up16x25__neon_mul16()
|
D | up16x25-minmax-fp32-neonv8-mul16.c | 346 … const int16x8_t vk17x89ABCDEF = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qc8_dwconv_minmax_fp32_ukernel_up16x25__neonv8_mul16() local 350 vacc89AB = vmlal_s16(vacc89AB, vget_low_s16(vi17x89ABCDEF), vget_low_s16(vk17x89ABCDEF)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up16x25__neonv8_mul16() 351 vaccCDEF = vmlal_s16(vaccCDEF, vget_high_s16(vi17x89ABCDEF), vget_high_s16(vk17x89ABCDEF)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up16x25__neonv8_mul16()
|
D | up16x25-minmax-fp32-avx2-mul32.c | 324 …const __m256i vk17x89ABCDEF = _mm256_cvtepi8_epi32(_mm_loadl_epi64((const __m128i*) ((uintptr_t) w… in xnn_qc8_dwconv_minmax_fp32_ukernel_up16x25__avx2_mul32() local 328 … vacc89ABCDEF = _mm256_add_epi32(vacc89ABCDEF, _mm256_mullo_epi32(vi17x89ABCDEF, vk17x89ABCDEF)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up16x25__avx2_mul32()
|
D | up16x25-minmax-fp32-neon-mla8-ld64.c | 343 const int8x8_t vk17x89ABCDEF = vld1_s8(w); w = (const void*) ((const int8_t*) w + 8); in xnn_qc8_dwconv_minmax_fp32_ukernel_up16x25__neon_mla8_ld64() local 346 vprod89ABCDEF = vmlal_s8(vprod89ABCDEF, vi17x89ABCDEF, vk17x89ABCDEF); in xnn_qc8_dwconv_minmax_fp32_ukernel_up16x25__neon_mla8_ld64()
|
D | up16x25-minmax-fp32-neonv8-mla8-ld64.c | 343 const int8x8_t vk17x89ABCDEF = vld1_s8(w); w = (const void*) ((const int8_t*) w + 8); in xnn_qc8_dwconv_minmax_fp32_ukernel_up16x25__neonv8_mla8_ld64() local 346 vprod89ABCDEF = vmlal_s8(vprod89ABCDEF, vi17x89ABCDEF, vk17x89ABCDEF); in xnn_qc8_dwconv_minmax_fp32_ukernel_up16x25__neonv8_mla8_ld64()
|
D | up16x25-minmax-fp32-wasmsimd-mul16-add16.c | 374 …const v128_t vk17x89ABCDEF = wasm_i16x8_load8x8((const void*) ((uintptr_t) w + 16 * sizeof(int32_t… in xnn_qc8_dwconv_minmax_fp32_ukernel_up16x25__wasmsimd_mul16_add16() local 378 vprod89ABCDEF = wasm_i16x8_add(vprod89ABCDEF, wasm_i16x8_mul(vi17x89ABCDEF, vk17x89ABCDEF)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up16x25__wasmsimd_mul16_add16()
|
D | up16x25-minmax-fp32-sse2-mul16-add16.c | 476 …const __m128i vk17x89ABCDEF = _mm_loadl_epi64((const __m128i*) ((uintptr_t) w + 16 * sizeof(int32_… in xnn_qc8_dwconv_minmax_fp32_ukernel_up16x25__sse2_mul16_add16() local 482 … const __m128i vxk17x89ABCDEF = _mm_srai_epi16(_mm_unpacklo_epi8(vk17x89ABCDEF, vk17x89ABCDEF), 8); in xnn_qc8_dwconv_minmax_fp32_ukernel_up16x25__sse2_mul16_add16()
|
D | up24x25-minmax-fp32-neon-mul16.c | 416 … const int16x8_t vk17x89ABCDEF = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qc8_dwconv_minmax_fp32_ukernel_up24x25__neon_mul16() local 422 vacc89AB = vmlal_s16(vacc89AB, vget_low_s16(vi17x89ABCDEF), vget_low_s16(vk17x89ABCDEF)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up24x25__neon_mul16() 423 vaccCDEF = vmlal_s16(vaccCDEF, vget_high_s16(vi17x89ABCDEF), vget_high_s16(vk17x89ABCDEF)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up24x25__neon_mul16()
|
/external/XNNPACK/src/qu8-dwconv/gen/ |
D | up16x25-minmax-fp32-neon-mul16.c | 348 …const int16x8_t vk17x89ABCDEF = vreinterpretq_s16_u16(vsubl_u8(vld1_u8(w), vkernel_zero_point)); w… in xnn_qu8_dwconv_minmax_fp32_ukernel_up16x25__neon_mul16() local 352 vacc89AB = vmlal_s16(vacc89AB, vget_low_s16(vi17x89ABCDEF), vget_low_s16(vk17x89ABCDEF)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up16x25__neon_mul16() 353 vaccCDEF = vmlal_s16(vaccCDEF, vget_high_s16(vi17x89ABCDEF), vget_high_s16(vk17x89ABCDEF)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up16x25__neon_mul16()
|
D | up16x25-minmax-fp32-neonv8-mul16.c | 348 …const int16x8_t vk17x89ABCDEF = vreinterpretq_s16_u16(vsubl_u8(vld1_u8(w), vkernel_zero_point)); w… in xnn_qu8_dwconv_minmax_fp32_ukernel_up16x25__neonv8_mul16() local 352 vacc89AB = vmlal_s16(vacc89AB, vget_low_s16(vi17x89ABCDEF), vget_low_s16(vk17x89ABCDEF)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up16x25__neonv8_mul16() 353 vaccCDEF = vmlal_s16(vaccCDEF, vget_high_s16(vi17x89ABCDEF), vget_high_s16(vk17x89ABCDEF)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up16x25__neonv8_mul16()
|
D | up16x25-minmax-rndnu-neon-mul16.c | 349 …const int16x8_t vk17x89ABCDEF = vreinterpretq_s16_u16(vsubl_u8(vld1_u8(w), vkernel_zero_point)); w… in xnn_qu8_dwconv_minmax_rndnu_ukernel_up16x25__neon_mul16() local 353 vacc89AB = vmlal_s16(vacc89AB, vget_low_s16(vi17x89ABCDEF), vget_low_s16(vk17x89ABCDEF)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up16x25__neon_mul16() 354 vaccCDEF = vmlal_s16(vaccCDEF, vget_high_s16(vi17x89ABCDEF), vget_high_s16(vk17x89ABCDEF)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up16x25__neon_mul16()
|
D | up16x25-minmax-fp32-avx2-mul32.c | 325 …const __m256i vk17x89ABCDEF = _mm256_sub_epi32(_mm256_cvtepu8_epi32(_mm_loadl_epi64((const __m128i… in xnn_qu8_dwconv_minmax_fp32_ukernel_up16x25__avx2_mul32() local 329 … vacc89ABCDEF = _mm256_add_epi32(vacc89ABCDEF, _mm256_mullo_epi32(vi17x89ABCDEF, vk17x89ABCDEF)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up16x25__avx2_mul32()
|
/external/XNNPACK/src/f32-dwconv/gen/ |
D | up16x25-minmax-avx.c | 327 const __m256 vk17x89ABCDEF = _mm256_load_ps(w + 296); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx() local 329 vacc89ABCDEFp0 = _mm256_add_ps(vacc89ABCDEFp0, _mm256_mul_ps(vi17x89ABCDEF, vk17x89ABCDEF)); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx()
|
D | up16x25-minmax-fma3-acc2.c | 327 const __m256 vk17x89ABCDEF = _mm256_load_ps(w + 296); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3_acc2() local 329 vacc89ABCDEFp1 = _mm256_fmadd_ps(vi17x89ABCDEF, vk17x89ABCDEF, vacc89ABCDEFp1); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3_acc2()
|
D | up16x25-minmax-fma3.c | 327 const __m256 vk17x89ABCDEF = _mm256_load_ps(w + 296); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3() local 329 vacc89ABCDEFp0 = _mm256_fmadd_ps(vi17x89ABCDEF, vk17x89ABCDEF, vacc89ABCDEFp0); in xnn_f32_dwconv_minmax_ukernel_up16x25__fma3()
|
D | up16x25-minmax-avx-acc2.c | 327 const __m256 vk17x89ABCDEF = _mm256_load_ps(w + 296); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx_acc2() local 329 vacc89ABCDEFp1 = _mm256_add_ps(vacc89ABCDEFp1, _mm256_mul_ps(vi17x89ABCDEF, vk17x89ABCDEF)); in xnn_f32_dwconv_minmax_ukernel_up16x25__avx_acc2()
|
/external/XNNPACK/src/f16-dwconv/gen/ |
D | up16x25-minmax-neonfp16arith-acc2.c | 293 const float16x8_t vk17x89ABCDEF = vld1q_f16(w); w += 8; in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith_acc2() local 295 vacc89ABCDEFp1 = vfmaq_f16(vacc89ABCDEFp1, vi17x89ABCDEF, vk17x89ABCDEF); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith_acc2()
|
D | up16x25-minmax-neonfp16arith.c | 293 const float16x8_t vk17x89ABCDEF = vld1q_f16(w); w += 8; in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith() local 295 vacc89ABCDEFp0 = vfmaq_f16(vacc89ABCDEFp0, vi17x89ABCDEF, vk17x89ABCDEF); in xnn_f16_dwconv_minmax_ukernel_up16x25__neonfp16arith()
|