| /external/XNNPACK/src/f32-argmaxpool/ |
| D | 9p8x-wasmsimd-c4.c | 76 v128_t vidx = wasm_i32x4_splat(0); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4() local
80 vidx = wasm_v128_bitselect(wasm_i32x4_splat(1), vidx, vm1); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4()
84 vidx = wasm_v128_bitselect(wasm_i32x4_splat(2), vidx, vm2); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4()
88 vidx = wasm_v128_bitselect(wasm_i32x4_splat(3), vidx, vm3); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4()
92 vidx = wasm_v128_bitselect(wasm_i32x4_splat(4), vidx, vm4); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4()
96 vidx = wasm_v128_bitselect(wasm_i32x4_splat(5), vidx, vm5); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4()
100 vidx = wasm_v128_bitselect(wasm_i32x4_splat(6), vidx, vm6); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4()
104 vidx = wasm_v128_bitselect(wasm_i32x4_splat(7), vidx, vm7); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4()
108 vidx = wasm_v128_bitselect(wasm_i32x4_splat(8), vidx, vm8); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4()
112 wasm_v128_store(ib, vidx); in xnn_f32_argmaxpool_ukernel_9p8x__wasmsimd_c4()
[all …]
|
| D | 9p8x-neon-c4.c | 67 uint32x4_t vidx = vmovq_n_u32(0); in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4() local
71 vidx = vbslq_u32(vm1, vmovq_n_u32(1), vidx); in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4()
75 vidx = vbslq_u32(vm2, vmovq_n_u32(2), vidx); in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4()
79 vidx = vbslq_u32(vm3, vmovq_n_u32(3), vidx); in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4()
83 vidx = vbslq_u32(vm4, vmovq_n_u32(4), vidx); in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4()
87 vidx = vbslq_u32(vm5, vmovq_n_u32(5), vidx); in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4()
91 vidx = vbslq_u32(vm6, vmovq_n_u32(6), vidx); in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4()
95 vidx = vbslq_u32(vm7, vmovq_n_u32(7), vidx); in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4()
99 vidx = vbslq_u32(vm8, vmovq_n_u32(8), vidx); in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4()
102 vst1q_u32(ib, vidx); ib += 4; in xnn_f32_argmaxpool_ukernel_9p8x__neon_c4()
[all …]
|
| D | 9x-wasmsimd-c4.c | 96 v128_t vidx = wasm_i32x4_splat(0); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4() local
100 vidx = wasm_v128_bitselect(wasm_i32x4_splat(1), vidx, vm1); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4()
104 vidx = wasm_v128_bitselect(wasm_i32x4_splat(2), vidx, vm2); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4()
108 vidx = wasm_v128_bitselect(wasm_i32x4_splat(3), vidx, vm3); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4()
112 vidx = wasm_v128_bitselect(wasm_i32x4_splat(4), vidx, vm4); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4()
116 vidx = wasm_v128_bitselect(wasm_i32x4_splat(5), vidx, vm5); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4()
120 vidx = wasm_v128_bitselect(wasm_i32x4_splat(6), vidx, vm6); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4()
124 vidx = wasm_v128_bitselect(wasm_i32x4_splat(7), vidx, vm7); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4()
128 vidx = wasm_v128_bitselect(wasm_i32x4_splat(8), vidx, vm8); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4()
132 wasm_v128_store(index, vidx); in xnn_f32_argmaxpool_ukernel_9x__wasmsimd_c4()
[all …]
|
| D | 9p8x-sse2-c4.c | 76 __m128i vidx = _mm_setzero_si128(); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4() local
80 vidx = _mm_or_si128(_mm_andnot_si128(vm1, vidx), _mm_and_si128(vm1, _mm_set1_epi32(1))); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4()
84 vidx = _mm_or_si128(_mm_andnot_si128(vm2, vidx), _mm_and_si128(vm2, _mm_set1_epi32(2))); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4()
88 vidx = _mm_or_si128(_mm_andnot_si128(vm3, vidx), _mm_and_si128(vm3, _mm_set1_epi32(3))); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4()
92 vidx = _mm_or_si128(_mm_andnot_si128(vm4, vidx), _mm_and_si128(vm4, _mm_set1_epi32(4))); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4()
96 vidx = _mm_or_si128(_mm_andnot_si128(vm5, vidx), _mm_and_si128(vm5, _mm_set1_epi32(5))); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4()
100 vidx = _mm_or_si128(_mm_andnot_si128(vm6, vidx), _mm_and_si128(vm6, _mm_set1_epi32(6))); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4()
104 vidx = _mm_or_si128(_mm_andnot_si128(vm7, vidx), _mm_and_si128(vm7, _mm_set1_epi32(7))); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4()
108 vidx = _mm_or_si128(_mm_andnot_si128(vm8, vidx), _mm_and_si128(vm8, _mm_set1_epi32(8))); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4()
112 _mm_store_si128((__m128i*) ib, vidx); in xnn_f32_argmaxpool_ukernel_9p8x__sse2_c4()
[all …]
|
| D | 9x-neon-c4.c | 86 uint32x4_t vidx = vmovq_n_u32(0); in xnn_f32_argmaxpool_ukernel_9x__neon_c4() local
90 vidx = vbslq_u32(vm1, vmovq_n_u32(1), vidx); in xnn_f32_argmaxpool_ukernel_9x__neon_c4()
94 vidx = vbslq_u32(vm2, vmovq_n_u32(2), vidx); in xnn_f32_argmaxpool_ukernel_9x__neon_c4()
98 vidx = vbslq_u32(vm3, vmovq_n_u32(3), vidx); in xnn_f32_argmaxpool_ukernel_9x__neon_c4()
102 vidx = vbslq_u32(vm4, vmovq_n_u32(4), vidx); in xnn_f32_argmaxpool_ukernel_9x__neon_c4()
106 vidx = vbslq_u32(vm5, vmovq_n_u32(5), vidx); in xnn_f32_argmaxpool_ukernel_9x__neon_c4()
110 vidx = vbslq_u32(vm6, vmovq_n_u32(6), vidx); in xnn_f32_argmaxpool_ukernel_9x__neon_c4()
114 vidx = vbslq_u32(vm7, vmovq_n_u32(7), vidx); in xnn_f32_argmaxpool_ukernel_9x__neon_c4()
118 vidx = vbslq_u32(vm8, vmovq_n_u32(8), vidx); in xnn_f32_argmaxpool_ukernel_9x__neon_c4()
121 vst1q_u32(index, vidx); index += 4; in xnn_f32_argmaxpool_ukernel_9x__neon_c4()
[all …]
|
| D | 9x-sse2-c4.c | 95 __m128i vidx = _mm_setzero_si128(); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4() local
99 vidx = _mm_or_si128(_mm_andnot_si128(vm1, vidx), _mm_and_si128(vm1, _mm_set1_epi32(1))); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4()
103 vidx = _mm_or_si128(_mm_andnot_si128(vm2, vidx), _mm_and_si128(vm2, _mm_set1_epi32(2))); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4()
107 vidx = _mm_or_si128(_mm_andnot_si128(vm3, vidx), _mm_and_si128(vm3, _mm_set1_epi32(3))); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4()
111 vidx = _mm_or_si128(_mm_andnot_si128(vm4, vidx), _mm_and_si128(vm4, _mm_set1_epi32(4))); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4()
115 vidx = _mm_or_si128(_mm_andnot_si128(vm5, vidx), _mm_and_si128(vm5, _mm_set1_epi32(5))); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4()
119 vidx = _mm_or_si128(_mm_andnot_si128(vm6, vidx), _mm_and_si128(vm6, _mm_set1_epi32(6))); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4()
123 vidx = _mm_or_si128(_mm_andnot_si128(vm7, vidx), _mm_and_si128(vm7, _mm_set1_epi32(7))); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4()
127 vidx = _mm_or_si128(_mm_andnot_si128(vm8, vidx), _mm_and_si128(vm8, _mm_set1_epi32(8))); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4()
131 _mm_storeu_si128((__m128i*) index, vidx); in xnn_f32_argmaxpool_ukernel_9x__sse2_c4()
[all …]
|
| D | 9p8x-scalar-c1.c | 67 uint32_t vidx = 0; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1() local
71 vidx = 1; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
76 vidx = 2; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
81 vidx = 3; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
86 vidx = 4; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
91 vidx = 5; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
96 vidx = 6; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
101 vidx = 7; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
106 vidx = 8; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
110 *ib++ = vidx; in xnn_f32_argmaxpool_ukernel_9p8x__scalar_c1()
[all …]
|
| D | 4x-wasmsimd-c4.c | 61 v128_t vidx = wasm_i32x4_splat(0); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4() local
65 vidx = wasm_v128_bitselect(wasm_i32x4_splat(1), vidx, vm1); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4()
69 vidx = wasm_v128_bitselect(wasm_i32x4_splat(2), vidx, vm2); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4()
73 vidx = wasm_v128_bitselect(wasm_i32x4_splat(3), vidx, vm3); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4()
77 wasm_v128_store(index, vidx); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4()
87 v128_t vidx = wasm_i32x4_splat(0); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4() local
91 vidx = wasm_v128_bitselect(wasm_i32x4_splat(1), vidx, vm1); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4()
95 vidx = wasm_v128_bitselect(wasm_i32x4_splat(2), vidx, vm2); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4()
99 vidx = wasm_v128_bitselect(wasm_i32x4_splat(3), vidx, vm3); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4()
103 *((double*) index) = wasm_f64x2_extract_lane(vidx, 0); in xnn_f32_argmaxpool_ukernel_4x__wasmsimd_c4()
[all …]
|
| D | 4x-sse2-c4.c | 60 __m128i vidx = _mm_setzero_si128(); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4() local
64 vidx = _mm_or_si128(_mm_andnot_si128(vm1, vidx), _mm_and_si128(vm1, _mm_set1_epi32(1))); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4()
68 vidx = _mm_or_si128(_mm_andnot_si128(vm2, vidx), _mm_and_si128(vm2, _mm_set1_epi32(2))); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4()
72 vidx = _mm_or_si128(_mm_andnot_si128(vm3, vidx), _mm_and_si128(vm3, _mm_set1_epi32(3))); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4()
76 _mm_storeu_si128((__m128i*) index, vidx); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4()
86 __m128i vidx = _mm_setzero_si128(); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4() local
90 vidx = _mm_or_si128(_mm_andnot_si128(vm1, vidx), _mm_and_si128(vm1, _mm_set1_epi32(1))); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4()
94 vidx = _mm_or_si128(_mm_andnot_si128(vm2, vidx), _mm_and_si128(vm2, _mm_set1_epi32(2))); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4()
98 vidx = _mm_or_si128(_mm_andnot_si128(vm3, vidx), _mm_and_si128(vm3, _mm_set1_epi32(3))); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4()
102 _mm_storel_epi64((__m128i*) index, vidx); in xnn_f32_argmaxpool_ukernel_4x__sse2_c4()
[all …]
|
| D | 4x-neon-c4.c | 56 uint32x4_t vidx = vmovq_n_u32(0); in xnn_f32_argmaxpool_ukernel_4x__neon_c4() local
60 vidx = vbslq_u32(vm1, vmovq_n_u32(1), vidx); in xnn_f32_argmaxpool_ukernel_4x__neon_c4()
64 vidx = vbslq_u32(vm2, vmovq_n_u32(2), vidx); in xnn_f32_argmaxpool_ukernel_4x__neon_c4()
68 vidx = vbslq_u32(vm3, vmovq_n_u32(3), vidx); in xnn_f32_argmaxpool_ukernel_4x__neon_c4()
71 vst1q_u32(index, vidx); index += 4; in xnn_f32_argmaxpool_ukernel_4x__neon_c4()
80 uint32x4_t vidx = vmovq_n_u32(0); in xnn_f32_argmaxpool_ukernel_4x__neon_c4() local
84 vidx = vbslq_u32(vm1, vmovq_n_u32(1), vidx); in xnn_f32_argmaxpool_ukernel_4x__neon_c4()
88 vidx = vbslq_u32(vm2, vmovq_n_u32(2), vidx); in xnn_f32_argmaxpool_ukernel_4x__neon_c4()
92 vidx = vbslq_u32(vm3, vmovq_n_u32(3), vidx); in xnn_f32_argmaxpool_ukernel_4x__neon_c4()
95 uint32x2_t vidx_lo = vget_low_u32(vidx); in xnn_f32_argmaxpool_ukernel_4x__neon_c4()
[all …]
|
| D | 9x-scalar-c1.c | 85 uint32_t vidx = 0; in xnn_f32_argmaxpool_ukernel_9x__scalar_c1() local
89 vidx = 1; in xnn_f32_argmaxpool_ukernel_9x__scalar_c1()
94 vidx = 2; in xnn_f32_argmaxpool_ukernel_9x__scalar_c1()
99 vidx = 3; in xnn_f32_argmaxpool_ukernel_9x__scalar_c1()
104 vidx = 4; in xnn_f32_argmaxpool_ukernel_9x__scalar_c1()
109 vidx = 5; in xnn_f32_argmaxpool_ukernel_9x__scalar_c1()
114 vidx = 6; in xnn_f32_argmaxpool_ukernel_9x__scalar_c1()
119 vidx = 7; in xnn_f32_argmaxpool_ukernel_9x__scalar_c1()
124 vidx = 8; in xnn_f32_argmaxpool_ukernel_9x__scalar_c1()
128 *index++ = vidx; in xnn_f32_argmaxpool_ukernel_9x__scalar_c1()
|
| /external/XNNPACK/src/f32-velu/ |
| D | sse-rr2-lut16-p3.c.in | 60 …const __m128i vidx${ABC[N:N+4]} = _mm_slli_epi32(_mm_and_si128(_mm_castps_si128(vn${ABC[N:N+4]}), …
65 const uint64_t vidx${ABC[N:N+2]} = (uint64_t) _mm_cvtsi128_si64(vidx${ABC[N:N+4]});
67 const uint64_t vidx${ABC[N+2:N+4]} = (uint64_t) _mm_extract_epi64(vidx${ABC[N:N+4]}, 1);
69 …const uint64_t vidx${ABC[N+2:N+4]} = (uint64_t) _mm_cvtsi128_si64(_mm_unpackhi_epi64(vidx${ABC[N:N…
70 …_si128(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) vidx${ABC[N:N+2]})));
71 …i128(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) vidx${ABC[N+2:N+4]})));
73 …, *((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) (vidx${ABC[N:N+2]} >> 32)…
75 …8(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) (vidx${ABC[N:N+2]} >> 32)…
78 …, *((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) (vidx${ABC[N+2:N+4]} >> 3…
80 …8(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) (vidx${ABC[N+2:N+4]} >> 3…
[all …]
|
| D | avx-rr2-lut16-p3.c.in | 58 const __m256 vidx${N} = _mm256_and_ps(vn${N}, vindex_mask);
60 … const __m128i vidx${N}_lo = _mm_slli_epi32(_mm_castps_si128(_mm256_castps256_ps128(vidx${N})), 2);
61 …const __m128i vidx${N}_hi = _mm_slli_epi32(_mm_castps_si128(_mm256_extractf128_ps(vidx${N}, 1)), 2…
63 const uint64_t vidx${N}_ll = (uint64_t) _mm_cvtsi128_si64(vidx${N}_lo);
64 const uint64_t vidx${N}_lh = (uint64_t) _mm_extract_epi64(vidx${N}_lo, 1);
65 const uint64_t vidx${N}_hl = (uint64_t) _mm_cvtsi128_si64(vidx${N}_hi);
66 const uint64_t vidx${N}_hh = (uint64_t) _mm_extract_epi64(vidx${N}_hi, 1);
67 …_mm_loadu_si32((const void*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) vidx${N}_ll));
68 …_mm_loadu_si32((const void*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) vidx${N}_lh));
69 …_mm_loadu_si32((const void*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) vidx${N}_hl));
[all …]
|
| /external/XNNPACK/src/f32-sigmoid/ |
| D | sse-lut64-p2-div.c.in | 56 …const __m128i vidx${ABC[N:N+4]} = _mm_slli_epi32(_mm_and_si128(_mm_castps_si128(vn${ABC[N:N+4]}), …
60 const uint64_t vidx${ABC[N:N+2]} = (uint64_t) _mm_cvtsi128_si64(vidx${ABC[N:N+4]});
62 const uint64_t vidx${ABC[N+2:N+4]} = (uint64_t) _mm_extract_epi64(vidx${ABC[N:N+4]}, 1);
64 …const uint64_t vidx${ABC[N+2:N+4]} = (uint64_t) _mm_cvtsi128_si64(_mm_unpackhi_epi64(vidx${ABC[N:N…
65 …_si128(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) vidx${ABC[N:N+2]})));
66 …i128(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) vidx${ABC[N+2:N+4]})));
68 …, *((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) (vidx${ABC[N:N+2]} >> 32)…
70 …8(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) (vidx${ABC[N:N+2]} >> 32)…
73 …, *((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) (vidx${ABC[N+2:N+4]} >> 3…
75 …8(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) (vidx${ABC[N+2:N+4]} >> 3…
[all …]
|
| /external/XNNPACK/src/f32-sigmoid/gen/ |
| D | sse2-lut64-p2-div-x4.c | 47 const __m128i vidx = _mm_slli_epi32(_mm_and_si128(_mm_castps_si128(vn), vindex_mask), 2); in xnn_f32_sigmoid_ukernel__sse2_lut64_p2_div_x4() local
49 const uint64_t vidx_lo = (uint64_t) _mm_cvtsi128_si64(vidx); in xnn_f32_sigmoid_ukernel__sse2_lut64_p2_div_x4()
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()
58 …(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) _mm_cvtsi128_si32(vidx)))); in xnn_f32_sigmoid_ukernel__sse2_lut64_p2_div_x4()
59 …(const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) _mm_extract_epi16(vidx, 4)))); in xnn_f32_sigmoid_ukernel__sse2_lut64_p2_div_x4()
60 …(const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) _mm_extract_epi16(vidx, 2)))); in xnn_f32_sigmoid_ukernel__sse2_lut64_p2_div_x4()
62 …(const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) _mm_extract_epi16(vidx, 6)))); in xnn_f32_sigmoid_ukernel__sse2_lut64_p2_div_x4()
94 const __m128i vidx = _mm_slli_epi32(_mm_and_si128(_mm_castps_si128(vn), vindex_mask), 2); in xnn_f32_sigmoid_ukernel__sse2_lut64_p2_div_x4() local
96 const uint64_t vidx_lo = (uint64_t) _mm_cvtsi128_si64(vidx); 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()
[all …]
|
| D | sse41-lut64-p2-div-x4.c | 47 const __m128i vidx = _mm_slli_epi32(_mm_and_si128(_mm_castps_si128(vn), vindex_mask), 2); in xnn_f32_sigmoid_ukernel__sse41_lut64_p2_div_x4() local
49 const uint64_t vidx_lo = (uint64_t) _mm_cvtsi128_si64(vidx); in xnn_f32_sigmoid_ukernel__sse41_lut64_p2_div_x4()
50 const uint64_t vidx_hi = (uint64_t) _mm_extract_epi64(vidx, 1); in xnn_f32_sigmoid_ukernel__sse41_lut64_p2_div_x4()
56 …(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) _mm_cvtsi128_si32(vidx)))); in xnn_f32_sigmoid_ukernel__sse41_lut64_p2_div_x4()
57 …(const int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) _mm_extract_epi16(vidx, 4)))); in xnn_f32_sigmoid_ukernel__sse41_lut64_p2_div_x4()
58 …nst int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) _mm_extract_epi16(vidx, 2))), 1); in xnn_f32_sigmoid_ukernel__sse41_lut64_p2_div_x4()
59 …nst int*) ((uintptr_t) xnn_table_exp2minus_k_over_64 + (uint32_t) _mm_extract_epi16(vidx, 6))), 1); in xnn_f32_sigmoid_ukernel__sse41_lut64_p2_div_x4()
89 const __m128i vidx = _mm_slli_epi32(_mm_and_si128(_mm_castps_si128(vn), vindex_mask), 2); in xnn_f32_sigmoid_ukernel__sse41_lut64_p2_div_x4() local
91 const uint64_t vidx_lo = (uint64_t) _mm_cvtsi128_si64(vidx); 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()
[all …]
|
| /external/XNNPACK/src/f32-velu/gen/ |
| D | velu-sse2-rr2-lut16-p3-x4.c | 54 const __m128i vidx = _mm_slli_epi32(_mm_and_si128(_mm_castps_si128(vn), vindex_mask), 2); in xnn_f32_velu_ukernel__sse2_rr2_lut16_p3_x4() local
56 const uint64_t vidx_lo = (uint64_t) _mm_cvtsi128_si64(vidx); in xnn_f32_velu_ukernel__sse2_rr2_lut16_p3_x4()
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()
65 …(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_cvtsi128_si32(vidx)))); in xnn_f32_velu_ukernel__sse2_rr2_lut16_p3_x4()
66 …(const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_extract_epi16(vidx, 4)))); in xnn_f32_velu_ukernel__sse2_rr2_lut16_p3_x4()
67 …(const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_extract_epi16(vidx, 2)))); in xnn_f32_velu_ukernel__sse2_rr2_lut16_p3_x4()
69 …(const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_extract_epi16(vidx, 6)))); in xnn_f32_velu_ukernel__sse2_rr2_lut16_p3_x4()
102 const __m128i vidx = _mm_slli_epi32(_mm_and_si128(_mm_castps_si128(vn), vindex_mask), 2); in xnn_f32_velu_ukernel__sse2_rr2_lut16_p3_x4() local
104 const uint64_t vidx_lo = (uint64_t) _mm_cvtsi128_si64(vidx); 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()
[all …]
|
| D | velu-sse41-rr2-lut16-p3-x4.c | 54 const __m128i vidx = _mm_slli_epi32(_mm_and_si128(_mm_castps_si128(vn), vindex_mask), 2); in xnn_f32_velu_ukernel__sse41_rr2_lut16_p3_x4() local
56 const uint64_t vidx_lo = (uint64_t) _mm_cvtsi128_si64(vidx); in xnn_f32_velu_ukernel__sse41_rr2_lut16_p3_x4()
57 const uint64_t vidx_hi = (uint64_t) _mm_extract_epi64(vidx, 1); in xnn_f32_velu_ukernel__sse41_rr2_lut16_p3_x4()
63 …(*((const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_cvtsi128_si32(vidx)))); in xnn_f32_velu_ukernel__sse41_rr2_lut16_p3_x4()
64 …(const int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_extract_epi16(vidx, 4)))); in xnn_f32_velu_ukernel__sse41_rr2_lut16_p3_x4()
65 …nst int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_extract_epi16(vidx, 2))), 1); in xnn_f32_velu_ukernel__sse41_rr2_lut16_p3_x4()
66 …nst int*) ((uintptr_t) xnn_table_exp2minus_k_over_16 + (uint32_t) _mm_extract_epi16(vidx, 6))), 1); in xnn_f32_velu_ukernel__sse41_rr2_lut16_p3_x4()
97 const __m128i vidx = _mm_slli_epi32(_mm_and_si128(_mm_castps_si128(vn), vindex_mask), 2); in xnn_f32_velu_ukernel__sse41_rr2_lut16_p3_x4() local
99 const uint64_t vidx_lo = (uint64_t) _mm_cvtsi128_si64(vidx); 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()
[all …]
|
| /external/icu/icu4j/main/classes/core/src/com/ibm/icu/impl/ |
| D | ICUDebug.java | 40 int i = 0, vidx = 0; in getInstanceLenient() local
45 if (vidx == 3) { in getInstanceLenient()
50 vidx++; in getInstanceLenient()
54 ver[vidx] = ver[vidx] * 10 + (c - '0'); in getInstanceLenient()
55 if (ver[vidx] > 255) { in getInstanceLenient()
59 ver[vidx] = 0; in getInstanceLenient()
64 ver[vidx] = c - '0'; in getInstanceLenient()
|
| /external/icu/android_icu4j/src/main/java/android/icu/impl/ |
| D | ICUDebug.java | 44 int i = 0, vidx = 0; in getInstanceLenient() local
49 if (vidx == 3) { in getInstanceLenient()
54 vidx++; in getInstanceLenient()
58 ver[vidx] = ver[vidx] * 10 + (c - '0'); in getInstanceLenient()
59 if (ver[vidx] > 255) { in getInstanceLenient()
63 ver[vidx] = 0; in getInstanceLenient()
68 ver[vidx] = c - '0'; in getInstanceLenient()
|
| /external/XNNPACK/src/math/ |
| D | expm1minus-sse2-rr2-lut16-p3.c | 71 const __m128i vidx = _mm_slli_epi32(_mm_and_si128(_mm_castps_si128(vn), vindex_mask), 2); in xnn_math_f32_expm1minus__sse2_rr2_lut16_p3() local
73 const uint64_t vidx_lo = (uint64_t) _mm_cvtsi128_si64(vidx); in xnn_math_f32_expm1minus__sse2_rr2_lut16_p3()
74 const uint64_t vidx_hi = (uint64_t) _mm_cvtsi128_si64(_mm_unpackhi_epi64(vidx, vidx)); in xnn_math_f32_expm1minus__sse2_rr2_lut16_p3()
80 const uint32_t vidx0 = (uint32_t) _mm_cvtsi128_si32(vidx); in xnn_math_f32_expm1minus__sse2_rr2_lut16_p3()
81 const uint32_t vidx1 = (uint32_t) _mm_extract_epi16(vidx, 2); in xnn_math_f32_expm1minus__sse2_rr2_lut16_p3()
82 const uint32_t vidx2 = (uint32_t) _mm_extract_epi16(vidx, 4); in xnn_math_f32_expm1minus__sse2_rr2_lut16_p3()
83 const uint32_t vidx3 = (uint32_t) _mm_extract_epi16(vidx, 6); in xnn_math_f32_expm1minus__sse2_rr2_lut16_p3()
|
| D | sigmoid-sse2-rr2-lut64-p2-div.c | 76 const __m128i vidx = _mm_slli_epi32(_mm_and_si128(_mm_castps_si128(vn), vindex_mask), 2); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_div() local
78 const uint64_t vidx_lo = (uint64_t) _mm_cvtsi128_si64(vidx); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_div()
79 const uint64_t vidx_hi = (uint64_t) _mm_cvtsi128_si64(_mm_unpackhi_epi64(vidx, vidx)); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_div()
85 const uint32_t vidx0 = (uint32_t) _mm_cvtsi128_si32(vidx); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_div()
86 const uint32_t vidx1 = (uint32_t) _mm_extract_epi16(vidx, 2); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_div()
87 const uint32_t vidx2 = (uint32_t) _mm_extract_epi16(vidx, 4); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_div()
88 const uint32_t vidx3 = (uint32_t) _mm_extract_epi16(vidx, 6); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_div()
|
| D | sigmoid-sse2-rr2-lut64-p2-nr2.c | 77 const __m128i vidx = _mm_slli_epi32(_mm_and_si128(_mm_castps_si128(vn), vindex_mask), 2); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_nr2() local
79 const uint64_t vidx_lo = (uint64_t) _mm_cvtsi128_si64(vidx); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_nr2()
80 const uint64_t vidx_hi = (uint64_t) _mm_cvtsi128_si64(_mm_unpackhi_epi64(vidx, vidx)); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_nr2()
86 const uint32_t vidx0 = (uint32_t) _mm_cvtsi128_si32(vidx); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_nr2()
87 const uint32_t vidx1 = (uint32_t) _mm_extract_epi16(vidx, 2); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_nr2()
88 const uint32_t vidx2 = (uint32_t) _mm_extract_epi16(vidx, 4); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_nr2()
89 const uint32_t vidx3 = (uint32_t) _mm_extract_epi16(vidx, 6); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_nr2()
|
| D | exp-sse2-rr2-lut64-p2.c | 69 const __m128i vidx = _mm_slli_epi32(_mm_and_si128(_mm_castps_si128(vn), vindex_mask), 2); in xnn_math_f32_exp__sse2_rr2_lut64_p2() local
71 const uint64_t vidx01 = (uint64_t) _mm_cvtsi128_si64(vidx); in xnn_math_f32_exp__sse2_rr2_lut64_p2()
72 const uint64_t vidx23 = (uint64_t) _mm_cvtsi128_si64(_mm_unpackhi_epi64(vidx, vidx)); in xnn_math_f32_exp__sse2_rr2_lut64_p2()
78 const uint32_t vidx0 = (uint32_t) _mm_cvtsi128_si32(vidx); in xnn_math_f32_exp__sse2_rr2_lut64_p2()
79 const uint32_t vidx1 = (uint32_t) _mm_extract_epi16(vidx, 2); in xnn_math_f32_exp__sse2_rr2_lut64_p2()
80 const uint32_t vidx2 = (uint32_t) _mm_extract_epi16(vidx, 4); in xnn_math_f32_exp__sse2_rr2_lut64_p2()
81 const uint32_t vidx3 = (uint32_t) _mm_extract_epi16(vidx, 6); in xnn_math_f32_exp__sse2_rr2_lut64_p2()
|
| D | sigmoid-sse2-rr2-lut64-p2-nr1.c | 77 const __m128i vidx = _mm_slli_epi32(_mm_and_si128(_mm_castps_si128(vn), vindex_mask), 2); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_nr1() local
79 const uint64_t vidx_lo = (uint64_t) _mm_cvtsi128_si64(vidx); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_nr1()
80 const uint64_t vidx_hi = (uint64_t) _mm_cvtsi128_si64(_mm_unpackhi_epi64(vidx, vidx)); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_nr1()
86 const uint32_t vidx0 = (uint32_t) _mm_cvtsi128_si32(vidx); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_nr1()
87 const uint32_t vidx1 = (uint32_t) _mm_extract_epi16(vidx, 2); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_nr1()
88 const uint32_t vidx2 = (uint32_t) _mm_extract_epi16(vidx, 4); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_nr1()
89 const uint32_t vidx3 = (uint32_t) _mm_extract_epi16(vidx, 6); in xnn_math_f32_sigmoid__sse2_rr2_lut64_p2_nr1()
|