/external/XNNPACK/src/qc8-dwconv/gen/ |
D | up24x25-minmax-fp32-neon-mul16.c | 418 … const int16x8_t vk17xGHIJKLMN = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qc8_dwconv_minmax_fp32_ukernel_up24x25__neon_mul16() local 424 vaccGHIJ = vmlal_s16(vaccGHIJ, vget_low_s16(vi17xGHIJKLMN), vget_low_s16(vk17xGHIJKLMN)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up24x25__neon_mul16() 425 vaccKLMN = vmlal_s16(vaccKLMN, vget_high_s16(vi17xGHIJKLMN), vget_high_s16(vk17xGHIJKLMN)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up24x25__neon_mul16()
|
D | up24x25-minmax-fp32-neonv8-mul16.c | 418 … const int16x8_t vk17xGHIJKLMN = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qc8_dwconv_minmax_fp32_ukernel_up24x25__neonv8_mul16() local 424 vaccGHIJ = vmlal_s16(vaccGHIJ, vget_low_s16(vi17xGHIJKLMN), vget_low_s16(vk17xGHIJKLMN)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up24x25__neonv8_mul16() 425 vaccKLMN = vmlal_s16(vaccKLMN, vget_high_s16(vi17xGHIJKLMN), vget_high_s16(vk17xGHIJKLMN)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up24x25__neonv8_mul16()
|
D | up24x25-minmax-fp32-avx2-mul32.c | 378 …const __m256i vk17xGHIJKLMN = _mm256_cvtepi8_epi32(_mm_loadl_epi64((const __m128i*) ((uintptr_t) w… in xnn_qc8_dwconv_minmax_fp32_ukernel_up24x25__avx2_mul32() local 383 … vaccGHIJKLMN = _mm256_add_epi32(vaccGHIJKLMN, _mm256_mullo_epi32(vi17xGHIJKLMN, vk17xGHIJKLMN)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up24x25__avx2_mul32()
|
D | up32x25-minmax-fp32-neonv8-mul16.c | 488 … const int16x8_t vk17xGHIJKLMN = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qc8_dwconv_minmax_fp32_ukernel_up32x25__neonv8_mul16() local 496 vaccGHIJ = vmlal_s16(vaccGHIJ, vget_low_s16(vi17xGHIJKLMN), vget_low_s16(vk17xGHIJKLMN)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up32x25__neonv8_mul16() 497 vaccKLMN = vmlal_s16(vaccKLMN, vget_high_s16(vi17xGHIJKLMN), vget_high_s16(vk17xGHIJKLMN)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up32x25__neonv8_mul16()
|
D | up32x25-minmax-fp32-neon-mul16.c | 488 … const int16x8_t vk17xGHIJKLMN = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qc8_dwconv_minmax_fp32_ukernel_up32x25__neon_mul16() local 496 vaccGHIJ = vmlal_s16(vaccGHIJ, vget_low_s16(vi17xGHIJKLMN), vget_low_s16(vk17xGHIJKLMN)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up32x25__neon_mul16() 497 vaccKLMN = vmlal_s16(vaccKLMN, vget_high_s16(vi17xGHIJKLMN), vget_high_s16(vk17xGHIJKLMN)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up32x25__neon_mul16()
|
D | up24x25-minmax-fp32-wasmsimd-mul16-add16.c | 445 …const v128_t vk17xGHIJKLMN = wasm_i16x8_load8x8((const void*) ((uintptr_t) w + 24 * sizeof(int32_t… in xnn_qc8_dwconv_minmax_fp32_ukernel_up24x25__wasmsimd_mul16_add16() local 450 vprodGHIJKLMN = wasm_i16x8_add(vprodGHIJKLMN, wasm_i16x8_mul(vi17xGHIJKLMN, vk17xGHIJKLMN)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up24x25__wasmsimd_mul16_add16()
|
D | up32x25-minmax-fp32-avx2-mul32.c | 430 …const __m256i vk17xGHIJKLMN = _mm256_cvtepi8_epi32(_mm_loadl_epi64((const __m128i*) ((uintptr_t) w… in xnn_qc8_dwconv_minmax_fp32_ukernel_up32x25__avx2_mul32() local 437 … vaccGHIJKLMN = _mm256_add_epi32(vaccGHIJKLMN, _mm256_mullo_epi32(vi17xGHIJKLMN, vk17xGHIJKLMN)); in xnn_qc8_dwconv_minmax_fp32_ukernel_up32x25__avx2_mul32()
|
/external/XNNPACK/src/qs8-dwconv/gen/ |
D | up24x25-minmax-rndnu-neon-mul16.c | 420 … const int16x8_t vk17xGHIJKLMN = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up24x25__neon_mul16() local 426 vaccGHIJ = vmlal_s16(vaccGHIJ, vget_low_s16(vi17xGHIJKLMN), vget_low_s16(vk17xGHIJKLMN)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up24x25__neon_mul16() 427 vaccKLMN = vmlal_s16(vaccKLMN, vget_high_s16(vi17xGHIJKLMN), vget_high_s16(vk17xGHIJKLMN)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up24x25__neon_mul16()
|
D | up24x25-minmax-fp32-neon-mul16.c | 419 … const int16x8_t vk17xGHIJKLMN = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qs8_dwconv_minmax_fp32_ukernel_up24x25__neon_mul16() local 425 vaccGHIJ = vmlal_s16(vaccGHIJ, vget_low_s16(vi17xGHIJKLMN), vget_low_s16(vk17xGHIJKLMN)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up24x25__neon_mul16() 426 vaccKLMN = vmlal_s16(vaccKLMN, vget_high_s16(vi17xGHIJKLMN), vget_high_s16(vk17xGHIJKLMN)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up24x25__neon_mul16()
|
D | up24x25-minmax-fp32-neonv8-mul16.c | 419 … const int16x8_t vk17xGHIJKLMN = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qs8_dwconv_minmax_fp32_ukernel_up24x25__neonv8_mul16() local 425 vaccGHIJ = vmlal_s16(vaccGHIJ, vget_low_s16(vi17xGHIJKLMN), vget_low_s16(vk17xGHIJKLMN)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up24x25__neonv8_mul16() 426 vaccKLMN = vmlal_s16(vaccKLMN, vget_high_s16(vi17xGHIJKLMN), vget_high_s16(vk17xGHIJKLMN)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up24x25__neonv8_mul16()
|
D | up24x25-minmax-fp32-avx2-mul32.c | 378 …const __m256i vk17xGHIJKLMN = _mm256_cvtepi8_epi32(_mm_loadl_epi64((const __m128i*) ((uintptr_t) w… in xnn_qs8_dwconv_minmax_fp32_ukernel_up24x25__avx2_mul32() local 383 … vaccGHIJKLMN = _mm256_add_epi32(vaccGHIJKLMN, _mm256_mullo_epi32(vi17xGHIJKLMN, vk17xGHIJKLMN)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up24x25__avx2_mul32()
|
D | up32x25-minmax-rndnu-neon-mul16.c | 490 … const int16x8_t vk17xGHIJKLMN = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up32x25__neon_mul16() local 498 vaccGHIJ = vmlal_s16(vaccGHIJ, vget_low_s16(vi17xGHIJKLMN), vget_low_s16(vk17xGHIJKLMN)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up32x25__neon_mul16() 499 vaccKLMN = vmlal_s16(vaccKLMN, vget_high_s16(vi17xGHIJKLMN), vget_high_s16(vk17xGHIJKLMN)); in xnn_qs8_dwconv_minmax_rndnu_ukernel_up32x25__neon_mul16()
|
D | up32x25-minmax-fp32-neon-mul16.c | 489 … const int16x8_t vk17xGHIJKLMN = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qs8_dwconv_minmax_fp32_ukernel_up32x25__neon_mul16() local 497 vaccGHIJ = vmlal_s16(vaccGHIJ, vget_low_s16(vi17xGHIJKLMN), vget_low_s16(vk17xGHIJKLMN)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up32x25__neon_mul16() 498 vaccKLMN = vmlal_s16(vaccKLMN, vget_high_s16(vi17xGHIJKLMN), vget_high_s16(vk17xGHIJKLMN)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up32x25__neon_mul16()
|
D | up32x25-minmax-fp32-neonv8-mul16.c | 489 … const int16x8_t vk17xGHIJKLMN = vmovl_s8(vld1_s8(w)); w = (const void*) ((const int8_t*) w + 8); in xnn_qs8_dwconv_minmax_fp32_ukernel_up32x25__neonv8_mul16() local 497 vaccGHIJ = vmlal_s16(vaccGHIJ, vget_low_s16(vi17xGHIJKLMN), vget_low_s16(vk17xGHIJKLMN)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up32x25__neonv8_mul16() 498 vaccKLMN = vmlal_s16(vaccKLMN, vget_high_s16(vi17xGHIJKLMN), vget_high_s16(vk17xGHIJKLMN)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up32x25__neonv8_mul16()
|
D | up24x25-minmax-fp32-wasmsimd-mul16-add16.c | 445 …const v128_t vk17xGHIJKLMN = wasm_i16x8_load8x8((const void*) ((uintptr_t) w + 24 * sizeof(int32_t… in xnn_qs8_dwconv_minmax_fp32_ukernel_up24x25__wasmsimd_mul16_add16() local 450 vprodGHIJKLMN = wasm_i16x8_add(vprodGHIJKLMN, wasm_i16x8_mul(vi17xGHIJKLMN, vk17xGHIJKLMN)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up24x25__wasmsimd_mul16_add16()
|
D | up32x25-minmax-fp32-avx2-mul32.c | 430 …const __m256i vk17xGHIJKLMN = _mm256_cvtepi8_epi32(_mm_loadl_epi64((const __m128i*) ((uintptr_t) w… in xnn_qs8_dwconv_minmax_fp32_ukernel_up32x25__avx2_mul32() local 437 … vaccGHIJKLMN = _mm256_add_epi32(vaccGHIJKLMN, _mm256_mullo_epi32(vi17xGHIJKLMN, vk17xGHIJKLMN)); in xnn_qs8_dwconv_minmax_fp32_ukernel_up32x25__avx2_mul32()
|
D | up24x25-minmax-fp32-wasmsimd-mul16.c | 499 …const v128_t vk17xGHIJKLMN = wasm_i16x8_load8x8((const void*) ((uintptr_t) w + 24 * sizeof(int32_t… in xnn_qs8_dwconv_minmax_fp32_ukernel_up24x25__wasmsimd_mul16() local 504 vprodGHIJKLMN = wasm_i16x8_mul(vi17xGHIJKLMN, vk17xGHIJKLMN); in xnn_qs8_dwconv_minmax_fp32_ukernel_up24x25__wasmsimd_mul16()
|
/external/XNNPACK/src/qu8-dwconv/gen/ |
D | up24x25-minmax-fp32-neonv8-mul16.c | 420 …const int16x8_t vk17xGHIJKLMN = vreinterpretq_s16_u16(vsubl_u8(vld1_u8(w), vkernel_zero_point)); w… in xnn_qu8_dwconv_minmax_fp32_ukernel_up24x25__neonv8_mul16() local 426 vaccGHIJ = vmlal_s16(vaccGHIJ, vget_low_s16(vi17xGHIJKLMN), vget_low_s16(vk17xGHIJKLMN)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up24x25__neonv8_mul16() 427 vaccKLMN = vmlal_s16(vaccKLMN, vget_high_s16(vi17xGHIJKLMN), vget_high_s16(vk17xGHIJKLMN)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up24x25__neonv8_mul16()
|
D | up24x25-minmax-rndnu-neon-mul16.c | 421 …const int16x8_t vk17xGHIJKLMN = vreinterpretq_s16_u16(vsubl_u8(vld1_u8(w), vkernel_zero_point)); w… in xnn_qu8_dwconv_minmax_rndnu_ukernel_up24x25__neon_mul16() local 427 vaccGHIJ = vmlal_s16(vaccGHIJ, vget_low_s16(vi17xGHIJKLMN), vget_low_s16(vk17xGHIJKLMN)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up24x25__neon_mul16() 428 vaccKLMN = vmlal_s16(vaccKLMN, vget_high_s16(vi17xGHIJKLMN), vget_high_s16(vk17xGHIJKLMN)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up24x25__neon_mul16()
|
D | up24x25-minmax-fp32-neon-mul16.c | 420 …const int16x8_t vk17xGHIJKLMN = vreinterpretq_s16_u16(vsubl_u8(vld1_u8(w), vkernel_zero_point)); w… in xnn_qu8_dwconv_minmax_fp32_ukernel_up24x25__neon_mul16() local 426 vaccGHIJ = vmlal_s16(vaccGHIJ, vget_low_s16(vi17xGHIJKLMN), vget_low_s16(vk17xGHIJKLMN)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up24x25__neon_mul16() 427 vaccKLMN = vmlal_s16(vaccKLMN, vget_high_s16(vi17xGHIJKLMN), vget_high_s16(vk17xGHIJKLMN)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up24x25__neon_mul16()
|
D | up32x25-minmax-fp32-neonv8-mul16.c | 490 …const int16x8_t vk17xGHIJKLMN = vreinterpretq_s16_u16(vsubl_u8(vld1_u8(w), vkernel_zero_point)); w… in xnn_qu8_dwconv_minmax_fp32_ukernel_up32x25__neonv8_mul16() local 498 vaccGHIJ = vmlal_s16(vaccGHIJ, vget_low_s16(vi17xGHIJKLMN), vget_low_s16(vk17xGHIJKLMN)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up32x25__neonv8_mul16() 499 vaccKLMN = vmlal_s16(vaccKLMN, vget_high_s16(vi17xGHIJKLMN), vget_high_s16(vk17xGHIJKLMN)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up32x25__neonv8_mul16()
|
D | up32x25-minmax-fp32-neon-mul16.c | 490 …const int16x8_t vk17xGHIJKLMN = vreinterpretq_s16_u16(vsubl_u8(vld1_u8(w), vkernel_zero_point)); w… in xnn_qu8_dwconv_minmax_fp32_ukernel_up32x25__neon_mul16() local 498 vaccGHIJ = vmlal_s16(vaccGHIJ, vget_low_s16(vi17xGHIJKLMN), vget_low_s16(vk17xGHIJKLMN)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up32x25__neon_mul16() 499 vaccKLMN = vmlal_s16(vaccKLMN, vget_high_s16(vi17xGHIJKLMN), vget_high_s16(vk17xGHIJKLMN)); in xnn_qu8_dwconv_minmax_fp32_ukernel_up32x25__neon_mul16()
|
D | up32x25-minmax-rndnu-neon-mul16.c | 491 …const int16x8_t vk17xGHIJKLMN = vreinterpretq_s16_u16(vsubl_u8(vld1_u8(w), vkernel_zero_point)); w… in xnn_qu8_dwconv_minmax_rndnu_ukernel_up32x25__neon_mul16() local 499 vaccGHIJ = vmlal_s16(vaccGHIJ, vget_low_s16(vi17xGHIJKLMN), vget_low_s16(vk17xGHIJKLMN)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up32x25__neon_mul16() 500 vaccKLMN = vmlal_s16(vaccKLMN, vget_high_s16(vi17xGHIJKLMN), vget_high_s16(vk17xGHIJKLMN)); in xnn_qu8_dwconv_minmax_rndnu_ukernel_up32x25__neon_mul16()
|
/external/XNNPACK/src/f16-dwconv/gen/ |
D | up32x25-minmax-neonfp16arith-acc2.c | 400 const float16x8_t vk17xGHIJKLMN = vld1q_f16(w); w += 8; in xnn_f16_dwconv_minmax_ukernel_up32x25__neonfp16arith_acc2() local 404 vaccGHIJKLMNp1 = vfmaq_f16(vaccGHIJKLMNp1, vi17xGHIJKLMN, vk17xGHIJKLMN); in xnn_f16_dwconv_minmax_ukernel_up32x25__neonfp16arith_acc2()
|
D | up32x25-minmax-neonfp16arith.c | 400 const float16x8_t vk17xGHIJKLMN = vld1q_f16(w); w += 8; in xnn_f16_dwconv_minmax_ukernel_up32x25__neonfp16arith() local 404 vaccGHIJKLMNp0 = vfmaq_f16(vaccGHIJKLMNp0, vi17xGHIJKLMN, vk17xGHIJKLMN); in xnn_f16_dwconv_minmax_ukernel_up32x25__neonfp16arith()
|