/external/XNNPACK/src/f32-velu/gen/ |
D | velu-avx-rr2-lut16-p3-x8.c | 56 const __m128i vidx_hi = _mm_slli_epi32(_mm_castps_si128(_mm256_extractf128_ps(vidx, 1)), 2); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x8() local 60 const uint64_t vidx_hl = (uint64_t) _mm_cvtsi128_si64(vidx_hi); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x8() 61 const uint64_t vidx_hh = (uint64_t) _mm_extract_epi64(vidx_hi, 1); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x8() 73 …(const void*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_cvtsi128_si32(vidx_hi))); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x8() 74 …nst void*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_extract_epi32(vidx_hi, 2))); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x8() 77 … int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_extract_epi32(vidx_hi, 1))), 1); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x8() 78 … int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_extract_epi32(vidx_hi, 3))), 1); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x8() 122 const __m128i vidx_hi = _mm_slli_epi32(_mm_castps_si128(_mm256_extractf128_ps(vidx, 1)), 2); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x8() local 126 const uint64_t vidx_hl = (uint64_t) _mm_cvtsi128_si64(vidx_hi); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x8() 127 const uint64_t vidx_hh = (uint64_t) _mm_extract_epi64(vidx_hi, 1); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x8() [all …]
|
D | velu-neonfma-rr1-lut16-p3-x4.c | 54 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_velu_ukernel__neonfma_rr1_lut16_p3_x4() local 56 … = vld1_dup_s32((const int32_t*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) vidx_hi)); in xnn_f32_velu_ukernel__neonfma_rr1_lut16_p3_x4() 58 …const int32_t*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) (vidx_hi >> 32)), vl_hi, 1… in xnn_f32_velu_ukernel__neonfma_rr1_lut16_p3_x4() 90 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_velu_ukernel__neonfma_rr1_lut16_p3_x4() local 92 … = vld1_dup_s32((const int32_t*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) vidx_hi)); in xnn_f32_velu_ukernel__neonfma_rr1_lut16_p3_x4() 94 …const int32_t*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) (vidx_hi >> 32)), vl_hi, 1… in xnn_f32_velu_ukernel__neonfma_rr1_lut16_p3_x4()
|
D | velu-neon-rr2-lut16-p3-x4.c | 55 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_velu_ukernel__neon_rr2_lut16_p3_x4() local 57 … = vld1_dup_s32((const int32_t*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) vidx_hi)); in xnn_f32_velu_ukernel__neon_rr2_lut16_p3_x4() 59 …const int32_t*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) (vidx_hi >> 32)), vl_hi, 1… in xnn_f32_velu_ukernel__neon_rr2_lut16_p3_x4() 92 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_velu_ukernel__neon_rr2_lut16_p3_x4() local 94 … = vld1_dup_s32((const int32_t*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) vidx_hi)); in xnn_f32_velu_ukernel__neon_rr2_lut16_p3_x4() 96 …const int32_t*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) (vidx_hi >> 32)), vl_hi, 1… in xnn_f32_velu_ukernel__neon_rr2_lut16_p3_x4()
|
D | velu-wasmsimd-arm-rr2-lut16-p3-x4.c | 56 const uint64_t vidx_hi = wasm_i64x2_extract_lane(vidx, 1); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_lut16_p3_x4() local 59 …nst float vl2 = *((const float*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) vidx_hi)); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_lut16_p3_x4() 60 …t vl3 = *((const float*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) (vidx_hi >> 32))); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_lut16_p3_x4() 94 const uint64_t vidx_hi = wasm_i64x2_extract_lane(vidx, 1); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_lut16_p3_x4() local 97 …nst float vl2 = *((const float*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) vidx_hi)); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_lut16_p3_x4() 98 …t vl3 = *((const float*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) (vidx_hi >> 32))); in xnn_f32_velu_ukernel__wasmsimd_arm_rr2_lut16_p3_x4()
|
D | velu-wasmsimd-x86-rr2-lut16-p3-x4.c | 56 const uint64_t vidx_hi = wasm_i64x2_extract_lane(vidx, 1); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x4() local 59 …nst float vl2 = *((const float*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) vidx_hi)); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x4() 60 …t vl3 = *((const float*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) (vidx_hi >> 32))); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x4() 97 const uint64_t vidx_hi = wasm_i64x2_extract_lane(vidx, 1); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x4() local 100 …nst float vl2 = *((const float*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) vidx_hi)); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x4() 101 …t vl3 = *((const float*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) (vidx_hi >> 32))); in xnn_f32_velu_ukernel__wasmsimd_x86_rr2_lut16_p3_x4()
|
D | velu-avx-rr2-lut16-p3-x16.c | 171 const __m128i vidx_hi = _mm_slli_epi32(_mm_castps_si128(_mm256_extractf128_ps(vidx, 1)), 2); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x16() local 175 const uint64_t vidx_hl = (uint64_t) _mm_cvtsi128_si64(vidx_hi); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x16() 176 const uint64_t vidx_hh = (uint64_t) _mm_extract_epi64(vidx_hi, 1); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x16() 188 …(const void*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_cvtsi128_si32(vidx_hi))); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x16() 189 …nst void*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_extract_epi32(vidx_hi, 2))); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x16() 192 … int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_extract_epi32(vidx_hi, 1))), 1); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x16() 193 … int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_extract_epi32(vidx_hi, 3))), 1); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x16() 237 const __m128i vidx_hi = _mm_slli_epi32(_mm_castps_si128(_mm256_extractf128_ps(vidx, 1)), 2); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x16() local 241 const uint64_t vidx_hl = (uint64_t) _mm_cvtsi128_si64(vidx_hi); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x16() 242 const uint64_t vidx_hh = (uint64_t) _mm_extract_epi64(vidx_hi, 1); in xnn_f32_velu_ukernel__avx_rr2_lut16_p3_x16() [all …]
|
D | velu-sse41-rr2-lut16-p3-x4.c | 57 const uint64_t vidx_hi = (uint64_t) _mm_extract_epi64(vidx, 1); in xnn_f32_velu_ukernel__sse41_rr2_lut16_p3_x4() local 59 …mm_cvtsi32_si128(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) vidx_hi))); in xnn_f32_velu_ukernel__sse41_rr2_lut16_p3_x4() 61 …l_hl, *((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) (vidx_hi >> 32))), 1); in xnn_f32_velu_ukernel__sse41_rr2_lut16_p3_x4() 100 const uint64_t vidx_hi = (uint64_t) _mm_extract_epi64(vidx, 1); in xnn_f32_velu_ukernel__sse41_rr2_lut16_p3_x4() local 102 …mm_cvtsi32_si128(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) vidx_hi))); in xnn_f32_velu_ukernel__sse41_rr2_lut16_p3_x4() 104 …l_hl, *((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) (vidx_hi >> 32))), 1); in xnn_f32_velu_ukernel__sse41_rr2_lut16_p3_x4()
|
D | velu-sse2-rr2-lut16-p3-x4.c | 57 const uint64_t vidx_hi = (uint64_t) _mm_cvtsi128_si64(_mm_unpackhi_epi64(vidx, vidx)); in xnn_f32_velu_ukernel__sse2_rr2_lut16_p3_x4() local 59 …mm_cvtsi32_si128(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) vidx_hi))); in xnn_f32_velu_ukernel__sse2_rr2_lut16_p3_x4() 62 …32_si128(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) (vidx_hi >> 32)))); in xnn_f32_velu_ukernel__sse2_rr2_lut16_p3_x4() 105 const uint64_t vidx_hi = (uint64_t) _mm_cvtsi128_si64(_mm_unpackhi_epi64(vidx, vidx)); in xnn_f32_velu_ukernel__sse2_rr2_lut16_p3_x4() local 107 …mm_cvtsi32_si128(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) vidx_hi))); in xnn_f32_velu_ukernel__sse2_rr2_lut16_p3_x4() 110 …32_si128(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) (vidx_hi >> 32)))); in xnn_f32_velu_ukernel__sse2_rr2_lut16_p3_x4()
|
/external/XNNPACK/src/math/ |
D | expm1minus-avx-rr2-lut16-p3.c | 74 const __m128i vidx_hi = _mm_slli_epi32(_mm_castps_si128(_mm256_extractf128_ps(vidx, 1)), 2); in xnn_math_f32_expm1minus__avx_rr2_lut16_p3() local 78 const uint64_t vidx_hl = (uint64_t) _mm_cvtsi128_si64(vidx_hi); in xnn_math_f32_expm1minus__avx_rr2_lut16_p3() 79 const uint64_t vidx_hh = (uint64_t) _mm_extract_epi64(vidx_hi, 1); in xnn_math_f32_expm1minus__avx_rr2_lut16_p3() 91 …(const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_cvtsi128_si32(vidx_hi)))); in xnn_math_f32_expm1minus__avx_rr2_lut16_p3() 92 …nst int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_extract_epi32(vidx_hi, 2)))); in xnn_math_f32_expm1minus__avx_rr2_lut16_p3() 95 … int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_extract_epi32(vidx_hi, 1))), 1); in xnn_math_f32_expm1minus__avx_rr2_lut16_p3() 96 … int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_extract_epi32(vidx_hi, 3))), 1); in xnn_math_f32_expm1minus__avx_rr2_lut16_p3()
|
D | sigmoid-avx-rr2-lut64-p2-div.c | 78 const __m128i vidx_hi = _mm_slli_epi32(_mm_castps_si128(_mm256_extractf128_ps(vidx, 1)), 2); in xnn_math_f32_sigmoid__avx_rr2_lut64_p2_div() local 82 const uint64_t vidx_hl = (uint64_t) _mm_cvtsi128_si64(vidx_hi); in xnn_math_f32_sigmoid__avx_rr2_lut64_p2_div() 83 const uint64_t vidx_hh = (uint64_t) _mm_extract_epi64(vidx_hi, 1); in xnn_math_f32_sigmoid__avx_rr2_lut64_p2_div() 95 …(const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) _mm_cvtsi128_si32(vidx_hi)))); in xnn_math_f32_sigmoid__avx_rr2_lut64_p2_div() 96 …nst int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) _mm_extract_epi32(vidx_hi, 2)))); in xnn_math_f32_sigmoid__avx_rr2_lut64_p2_div() 99 … int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) _mm_extract_epi32(vidx_hi, 1))), 1); in xnn_math_f32_sigmoid__avx_rr2_lut64_p2_div() 100 … int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) _mm_extract_epi32(vidx_hi, 3))), 1); in xnn_math_f32_sigmoid__avx_rr2_lut64_p2_div()
|
/external/XNNPACK/src/f32-sigmoid/gen/ |
D | neonfma-rr1-lut64-p2-div-x4.c | 46 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_div_x4() local 48 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_div_x4() 50 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_div_x4() 80 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_div_x4() local 82 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_div_x4() 84 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_div_x4()
|
D | wasmsimd-lut64-p2-div-x4.c | 48 const uint64_t vidx_hi = wasm_i64x2_extract_lane(vidx, 1); in xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x4() local 51 …nst float vl2 = *((const float*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) vidx_hi)); in xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x4() 52 …t vl3 = *((const float*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) (vidx_hi >> 32))); in xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x4() 84 const uint64_t vidx_hi = wasm_i64x2_extract_lane(vidx, 1); in xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x4() local 87 …nst float vl2 = *((const float*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) vidx_hi)); in xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x4() 88 …t vl3 = *((const float*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) (vidx_hi >> 32))); in xnn_f32_sigmoid_ukernel__wasmsimd_lut64_p2_div_x4()
|
D | neonfma-rr1-lut2048-p1-div-x4.c | 46 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_div_x4() local 48 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_div_x4() 50 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_div_x4() 79 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_div_x4() local 81 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_div_x4() 83 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_div_x4()
|
D | neonfma-rr1-lut2048-p1-nr1recps1fma-x4.c | 46 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr1recps1fma_x4() local 48 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr1recps1fma_x4() 50 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr1recps1fma_x4() 83 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr1recps1fma_x4() local 85 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr1recps1fma_x4() 87 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr1recps1fma_x4()
|
D | neonfma-rr1-lut2048-p1-nr2recps-x4.c | 46 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr2recps_x4() local 48 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr2recps_x4() 50 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr2recps_x4() 83 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr2recps_x4() local 85 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr2recps_x4() 87 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr2recps_x4()
|
D | neonfma-rr1-lut2048-p1-nr2fma-x4.c | 46 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr2fma_x4() local 48 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr2fma_x4() 50 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr2fma_x4() 83 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr2fma_x4() local 85 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr2fma_x4() 87 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut2048_p1_nr2fma_x4()
|
D | neonfma-rr1-lut64-p2-nr2recps-x4.c | 46 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr2recps_x4() local 48 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr2recps_x4() 50 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr2recps_x4() 84 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr2recps_x4() local 86 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr2recps_x4() 88 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr2recps_x4()
|
D | neon-rr2-lut2048-p1-nr2recps-x4.c | 47 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neon_rr2_lut2048_p1_nr2recps_x4() local 49 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neon_rr2_lut2048_p1_nr2recps_x4() 51 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neon_rr2_lut2048_p1_nr2recps_x4() 85 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neon_rr2_lut2048_p1_nr2recps_x4() local 87 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neon_rr2_lut2048_p1_nr2recps_x4() 89 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_2048[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neon_rr2_lut2048_p1_nr2recps_x4()
|
D | neonfma-rr1-lut64-p2-nr2fma-x4.c | 46 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr2fma_x4() local 48 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr2fma_x4() 50 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr2fma_x4() 84 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr2fma_x4() local 86 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr2fma_x4() 88 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr2fma_x4()
|
D | neon-rr2-lut64-p2-nr2recps-x4.c | 47 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neon_rr2_lut64_p2_nr2recps_x4() local 49 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neon_rr2_lut64_p2_nr2recps_x4() 51 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neon_rr2_lut64_p2_nr2recps_x4() 86 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neon_rr2_lut64_p2_nr2recps_x4() local 88 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neon_rr2_lut64_p2_nr2recps_x4() 90 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neon_rr2_lut64_p2_nr2recps_x4()
|
D | neonfma-rr1-lut64-p2-nr1recps1fma-x4.c | 46 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr1recps1fma_x4() local 48 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr1recps1fma_x4() 50 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr1recps1fma_x4() 84 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr1recps1fma_x4() local 86 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) vidx_hi]); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr1recps1fma_x4() 88 vl_hi = vld1_lane_f32(&xnn_table_exp2minus_k_over_64[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_sigmoid_ukernel__neonfma_rr1_lut64_p2_nr1recps1fma_x4()
|
D | sse41-lut64-p2-div-x4.c | 50 const uint64_t vidx_hi = (uint64_t) _mm_extract_epi64(vidx, 1); in xnn_f32_sigmoid_ukernel__sse41_lut64_p2_div_x4() local 52 …mm_cvtsi32_si128(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) vidx_hi))); in xnn_f32_sigmoid_ukernel__sse41_lut64_p2_div_x4() 54 …l_hl, *((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) (vidx_hi >> 32))), 1); in xnn_f32_sigmoid_ukernel__sse41_lut64_p2_div_x4() 92 const uint64_t vidx_hi = (uint64_t) _mm_extract_epi64(vidx, 1); in xnn_f32_sigmoid_ukernel__sse41_lut64_p2_div_x4() local 94 …mm_cvtsi32_si128(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) vidx_hi))); in xnn_f32_sigmoid_ukernel__sse41_lut64_p2_div_x4() 96 …l_hl, *((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) (vidx_hi >> 32))), 1); in xnn_f32_sigmoid_ukernel__sse41_lut64_p2_div_x4()
|
D | sse2-lut64-p2-div-x4.c | 50 const uint64_t vidx_hi = (uint64_t) _mm_cvtsi128_si64(_mm_unpackhi_epi64(vidx, vidx)); in xnn_f32_sigmoid_ukernel__sse2_lut64_p2_div_x4() local 52 …mm_cvtsi32_si128(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) vidx_hi))); in xnn_f32_sigmoid_ukernel__sse2_lut64_p2_div_x4() 55 …32_si128(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) (vidx_hi >> 32)))); in xnn_f32_sigmoid_ukernel__sse2_lut64_p2_div_x4() 97 const uint64_t vidx_hi = (uint64_t) _mm_cvtsi128_si64(_mm_unpackhi_epi64(vidx, vidx)); in xnn_f32_sigmoid_ukernel__sse2_lut64_p2_div_x4() local 99 …mm_cvtsi32_si128(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) vidx_hi))); in xnn_f32_sigmoid_ukernel__sse2_lut64_p2_div_x4() 102 …32_si128(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) (vidx_hi >> 32)))); in xnn_f32_sigmoid_ukernel__sse2_lut64_p2_div_x4()
|
/external/XNNPACK/src/f32-raddstoreexpminusmax/gen/ |
D | neonfma-lut64-p2-x4.c | 74 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x4() local 76 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2_k_over_64[(uint32_t) vidx_hi]); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x4() 78 vl_hi = vld1_lane_f32(&xnn_table_exp2_k_over_64[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x4() 150 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x4() local 152 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2_k_over_64[(uint32_t) vidx_hi]); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x4() 154 vl_hi = vld1_lane_f32(&xnn_table_exp2_k_over_64[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_raddstoreexpminusmax_ukernel__neonfma_lut64_p2_x4()
|
D | neon-lut64-p2-x4.c | 75 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_raddstoreexpminusmax_ukernel__neon_lut64_p2_x4() local 77 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2_k_over_64[(uint32_t) vidx_hi]); in xnn_f32_raddstoreexpminusmax_ukernel__neon_lut64_p2_x4() 79 vl_hi = vld1_lane_f32(&xnn_table_exp2_k_over_64[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_raddstoreexpminusmax_ukernel__neon_lut64_p2_x4() 151 const uint64_t vidx_hi = vgetq_lane_u64(vidx, 1); in xnn_f32_raddstoreexpminusmax_ukernel__neon_lut64_p2_x4() local 153 float32x2_t vl_hi = vld1_dup_f32(&xnn_table_exp2_k_over_64[(uint32_t) vidx_hi]); in xnn_f32_raddstoreexpminusmax_ukernel__neon_lut64_p2_x4() 155 vl_hi = vld1_lane_f32(&xnn_table_exp2_k_over_64[(uint32_t) (vidx_hi >> 32)], vl_hi, 1); in xnn_f32_raddstoreexpminusmax_ukernel__neon_lut64_p2_x4()
|