| /external/XNNPACK/src/f32-ibilinear/gen/ |
| D | scalar-c4.c | 35 const float valphah = weights[0]; in xnn_f32_ibilinear_ukernel__scalar_c4() local
71 const float vt0 = vtl0 + vtd0 * valphah; in xnn_f32_ibilinear_ukernel__scalar_c4()
72 const float vb0 = vbl0 + vbd0 * valphah; in xnn_f32_ibilinear_ukernel__scalar_c4()
73 const float vt1 = vtl1 + vtd1 * valphah; in xnn_f32_ibilinear_ukernel__scalar_c4()
74 const float vb1 = vbl1 + vbd1 * valphah; in xnn_f32_ibilinear_ukernel__scalar_c4()
75 const float vt2 = vtl2 + vtd2 * valphah; in xnn_f32_ibilinear_ukernel__scalar_c4()
76 const float vb2 = vbl2 + vbd2 * valphah; in xnn_f32_ibilinear_ukernel__scalar_c4()
77 const float vt3 = vtl3 + vtd3 * valphah; in xnn_f32_ibilinear_ukernel__scalar_c4()
78 const float vb3 = vbl3 + vbd3 * valphah; in xnn_f32_ibilinear_ukernel__scalar_c4()
105 const float vt = vtl + vtd * valphah; in xnn_f32_ibilinear_ukernel__scalar_c4()
[all …]
|
| D | scalar-c2.c | 35 const float valphah = weights[0]; in xnn_f32_ibilinear_ukernel__scalar_c2() local
59 const float vt0 = vtl0 + vtd0 * valphah; in xnn_f32_ibilinear_ukernel__scalar_c2()
60 const float vb0 = vbl0 + vbd0 * valphah; in xnn_f32_ibilinear_ukernel__scalar_c2()
61 const float vt1 = vtl1 + vtd1 * valphah; in xnn_f32_ibilinear_ukernel__scalar_c2()
62 const float vb1 = vbl1 + vbd1 * valphah; in xnn_f32_ibilinear_ukernel__scalar_c2()
83 const float vt = vtl + vtd * valphah; in xnn_f32_ibilinear_ukernel__scalar_c2()
84 const float vb = vbl + vbd * valphah; in xnn_f32_ibilinear_ukernel__scalar_c2()
|
| D | wasmsimd-c8.c | 37 const v128_t valphah = wasm_v32x4_load_splat(weights); in xnn_f32_ibilinear_ukernel__wasmsimd_c8() local
61 const v128_t vt0123 = wasm_f32x4_add(vtl0123, wasm_f32x4_mul(vtd0123, valphah)); in xnn_f32_ibilinear_ukernel__wasmsimd_c8()
62 const v128_t vb0123 = wasm_f32x4_add(vbl0123, wasm_f32x4_mul(vbd0123, valphah)); in xnn_f32_ibilinear_ukernel__wasmsimd_c8()
63 const v128_t vt4567 = wasm_f32x4_add(vtl4567, wasm_f32x4_mul(vtd4567, valphah)); in xnn_f32_ibilinear_ukernel__wasmsimd_c8()
64 const v128_t vb4567 = wasm_f32x4_add(vbl4567, wasm_f32x4_mul(vbd4567, valphah)); in xnn_f32_ibilinear_ukernel__wasmsimd_c8()
88 const v128_t vt = wasm_f32x4_add(vtl, wasm_f32x4_mul(vtd, valphah)); in xnn_f32_ibilinear_ukernel__wasmsimd_c8()
89 const v128_t vb = wasm_f32x4_add(vbl, wasm_f32x4_mul(vbd, valphah)); in xnn_f32_ibilinear_ukernel__wasmsimd_c8()
104 const v128_t vt = wasm_f32x4_add(vtl, wasm_f32x4_mul(vtd, valphah)); in xnn_f32_ibilinear_ukernel__wasmsimd_c8()
105 const v128_t vb = wasm_f32x4_add(vbl, wasm_f32x4_mul(vbd, valphah)); in xnn_f32_ibilinear_ukernel__wasmsimd_c8()
|
| D | sse-c8.c | 39 const __m128 valphah = _mm_movelh_ps(valphahv, valphahv); in xnn_f32_ibilinear_ukernel__sse_c8() local
63 const __m128 vt0123 = _mm_add_ps(vtl0123, _mm_mul_ps(vtd0123, valphah)); in xnn_f32_ibilinear_ukernel__sse_c8()
64 const __m128 vb0123 = _mm_add_ps(vbl0123, _mm_mul_ps(vbd0123, valphah)); in xnn_f32_ibilinear_ukernel__sse_c8()
65 const __m128 vt4567 = _mm_add_ps(vtl4567, _mm_mul_ps(vtd4567, valphah)); in xnn_f32_ibilinear_ukernel__sse_c8()
66 const __m128 vb4567 = _mm_add_ps(vbl4567, _mm_mul_ps(vbd4567, valphah)); in xnn_f32_ibilinear_ukernel__sse_c8()
91 const __m128 vt0123 = _mm_add_ps(vtl0123, _mm_mul_ps(vtd0123, valphah)); in xnn_f32_ibilinear_ukernel__sse_c8()
92 const __m128 vb0123 = _mm_add_ps(vbl0123, _mm_mul_ps(vbd0123, valphah)); in xnn_f32_ibilinear_ukernel__sse_c8()
110 const __m128 vt0123 = _mm_add_ps(vtl0123, _mm_mul_ps(vtd0123, valphah)); in xnn_f32_ibilinear_ukernel__sse_c8()
111 const __m128 vb0123 = _mm_add_ps(vbl0123, _mm_mul_ps(vbd0123, valphah)); in xnn_f32_ibilinear_ukernel__sse_c8()
|
| D | neonfma-c8.c | 40 const float32x4_t valphah = vdupq_lane_f32(valphahv, 0); in xnn_f32_ibilinear_ukernel__neonfma_c8() local
61 const float32x4_t vt0123 = vfmaq_f32(vtl0123, vtd0123, valphah); in xnn_f32_ibilinear_ukernel__neonfma_c8()
62 const float32x4_t vb0123 = vfmaq_f32(vbl0123, vbd0123, valphah); in xnn_f32_ibilinear_ukernel__neonfma_c8()
63 const float32x4_t vt4567 = vfmaq_f32(vtl4567, vtd4567, valphah); in xnn_f32_ibilinear_ukernel__neonfma_c8()
64 const float32x4_t vb4567 = vfmaq_f32(vbl4567, vbd4567, valphah); in xnn_f32_ibilinear_ukernel__neonfma_c8()
96 const float32x4_t vt0123 = vfmaq_f32(vtl0123, vtd0123, valphah); in xnn_f32_ibilinear_ukernel__neonfma_c8()
97 const float32x4_t vb0123 = vfmaq_f32(vbl0123, vbd0123, valphah); in xnn_f32_ibilinear_ukernel__neonfma_c8()
124 const float32x4_t vt0123 = vfmaq_f32(vtl0123, vtd0123, valphah); in xnn_f32_ibilinear_ukernel__neonfma_c8()
125 const float32x4_t vb0123 = vfmaq_f32(vbl0123, vbd0123, valphah); in xnn_f32_ibilinear_ukernel__neonfma_c8()
|
| D | wasmsimd-c4.c | 37 const v128_t valphah = wasm_v32x4_load_splat(weights); in xnn_f32_ibilinear_ukernel__wasmsimd_c4() local
54 const v128_t vt = wasm_f32x4_add(vtl, wasm_f32x4_mul(vtd, valphah)); in xnn_f32_ibilinear_ukernel__wasmsimd_c4()
55 const v128_t vb = wasm_f32x4_add(vbl, wasm_f32x4_mul(vbd, valphah)); in xnn_f32_ibilinear_ukernel__wasmsimd_c4()
70 const v128_t vt = wasm_f32x4_add(vtl, wasm_f32x4_mul(vtd, valphah)); in xnn_f32_ibilinear_ukernel__wasmsimd_c4()
71 const v128_t vb = wasm_f32x4_add(vbl, wasm_f32x4_mul(vbd, valphah)); in xnn_f32_ibilinear_ukernel__wasmsimd_c4()
|
| D | sse-c4.c | 39 const __m128 valphah = _mm_movelh_ps(valphahv, valphahv); in xnn_f32_ibilinear_ukernel__sse_c4() local
57 const __m128 vt0123 = _mm_add_ps(vtl0123, _mm_mul_ps(vtd0123, valphah)); in xnn_f32_ibilinear_ukernel__sse_c4()
58 const __m128 vb0123 = _mm_add_ps(vbl0123, _mm_mul_ps(vbd0123, valphah)); in xnn_f32_ibilinear_ukernel__sse_c4()
76 const __m128 vt0123 = _mm_add_ps(vtl0123, _mm_mul_ps(vtd0123, valphah)); in xnn_f32_ibilinear_ukernel__sse_c4()
77 const __m128 vb0123 = _mm_add_ps(vbl0123, _mm_mul_ps(vbd0123, valphah)); in xnn_f32_ibilinear_ukernel__sse_c4()
|
| D | neonfma-c4.c | 40 const float32x4_t valphah = vdupq_lane_f32(valphahv, 0); in xnn_f32_ibilinear_ukernel__neonfma_c4() local
55 const float32x4_t vt0123 = vfmaq_f32(vtl0123, vtd0123, valphah); in xnn_f32_ibilinear_ukernel__neonfma_c4()
56 const float32x4_t vb0123 = vfmaq_f32(vbl0123, vbd0123, valphah); in xnn_f32_ibilinear_ukernel__neonfma_c4()
82 const float32x4_t vt0123 = vfmaq_f32(vtl0123, vtd0123, valphah); in xnn_f32_ibilinear_ukernel__neonfma_c4()
83 const float32x4_t vb0123 = vfmaq_f32(vbl0123, vbd0123, valphah); in xnn_f32_ibilinear_ukernel__neonfma_c4()
|
| D | scalar-c1.c | 35 const float valphah = weights[0]; in xnn_f32_ibilinear_ukernel__scalar_c1() local
49 const float vt = vtl + vtd * valphah; in xnn_f32_ibilinear_ukernel__scalar_c1()
50 const float vb = vbl + vbd * valphah; in xnn_f32_ibilinear_ukernel__scalar_c1()
|
| /external/XNNPACK/src/f32-ibilinear-chw/ |
| D | scalar.c.in | 40 const float valphah${ABC[P]} = w[${P * 2}];
55 const float vt${ABC[P]} = vtl${ABC[P]} + vtd${ABC[P]} * valphah${ABC[P]};
56 const float vb${ABC[P]} = vbl${ABC[P]} + vbd${ABC[P]} * valphah${ABC[P]};
74 const float valphah = w[0]; variable
86 const float vt = vtl + vtd * valphah;
87 const float vb = vbl + vbd * valphah;
101 const float valphah = w[0]; variable
113 const float vt = vtl + vtd * valphah;
114 const float vb = vbl + vbd * valphah;
|
| D | wasmsimd.c.in | 52 …const v128_t valphah${ABC[P:P+4]} = wasm_v32x4_shuffle(vw${ABC[P:P+4]}p0, vw${ABC[P:P+4]}p1, 0, 2,…
78 …C[P:P+4]} = wasm_f32x4_add(vl${ABC[P:P+4]}, wasm_f32x4_mul(vd${ABC[P:P+4]}, valphah${ABC[P:P+4]}));
101 const v128_t valphah = wasm_v32x4_shuffle(vw0, vw1, 0, 2, 4, 6); variable
121 const v128_t vo = wasm_f32x4_add(vl, wasm_f32x4_mul(vd, valphah));
132 const v128_t valphah = wasm_v32x4_shuffle(vw, vw, 0, 2, 0, 2); variable
154 const v128_t vo = wasm_f32x4_add(vl, wasm_f32x4_mul(vd, valphah));
|
| D | neon.c.in | 51 const float32x4_t valphah${ABC[P:P+4]} = vw${ABC[P:P+4]}.val[0];
79 …at32x4_t vo${ABC[P:P+4]} = ${VMULADDQ_F32}(vl${ABC[P:P+4]}, vd${ABC[P:P+4]}, valphah${ABC[P:P+4]});
99 const float32x4_t valphah = vw.val[0];
121 const float32x4_t vo = ${VMULADDQ_F32}(vl, vd, valphah);
132 const float32x2_t valphah = vw.val[0];
159 const float32x2_t vo = ${VMULADD_F32}(vl, vd, valphah);
|
| /external/XNNPACK/src/f32-ibilinear/ |
| D | scalar.c.in | 34 const float valphah = weights[0];
56 const float vt${ABC[C]} = vtl${ABC[C]} + vtd${ABC[C]} * valphah;
57 const float vb${ABC[C]} = vbl${ABC[C]} + vbd${ABC[C]} * valphah;
78 const float vt = vtl + vtd * valphah;
79 const float vb = vbl + vbd * valphah;
97 const float vt = vtl + vtd * valphah;
98 const float vb = vbl + vbd * valphah;
|
| D | sse.c.in | 39 const __m128 valphah = _mm_movelh_ps(valphahv, valphahv);
64 …const __m128 vt${ABC[C:C+4]} = _mm_add_ps(vtl${ABC[C:C+4]}, _mm_mul_ps(vtd${ABC[C:C+4]}, valphah));
65 …const __m128 vb${ABC[C:C+4]} = _mm_add_ps(vbl${ABC[C:C+4]}, _mm_mul_ps(vbd${ABC[C:C+4]}, valphah));
92 const __m128 vt0123 = _mm_add_ps(vtl0123, _mm_mul_ps(vtd0123, valphah));
93 const __m128 vb0123 = _mm_add_ps(vbl0123, _mm_mul_ps(vbd0123, valphah));
111 const __m128 vt0123 = _mm_add_ps(vtl0123, _mm_mul_ps(vtd0123, valphah));
112 const __m128 vb0123 = _mm_add_ps(vbl0123, _mm_mul_ps(vbd0123, valphah));
|
| D | wasmsimd.c.in | 37 const v128_t valphah = wasm_v32x4_load_splat(weights);
63 …28_t vt${ABC[C:C+4]} = wasm_f32x4_add(vtl${ABC[C:C+4]}, wasm_f32x4_mul(vtd${ABC[C:C+4]}, valphah));
64 …28_t vb${ABC[C:C+4]} = wasm_f32x4_add(vbl${ABC[C:C+4]}, wasm_f32x4_mul(vbd${ABC[C:C+4]}, valphah));
89 const v128_t vt = wasm_f32x4_add(vtl, wasm_f32x4_mul(vtd, valphah));
90 const v128_t vb = wasm_f32x4_add(vbl, wasm_f32x4_mul(vbd, valphah));
105 const v128_t vt = wasm_f32x4_add(vtl, wasm_f32x4_mul(vtd, valphah));
106 const v128_t vb = wasm_f32x4_add(vbl, wasm_f32x4_mul(vbd, valphah));
|
| D | neon.c.in | 42 const float32x4_t valphah = vdupq_lane_f32(valphahv, 0);
61 … const float32x4_t vt${ABC[C:C+4]} = vfmaq_f32(vtl${ABC[C:C+4]}, vtd${ABC[C:C+4]}, valphah);
62 … const float32x4_t vb${ABC[C:C+4]} = vfmaq_f32(vbl${ABC[C:C+4]}, vbd${ABC[C:C+4]}, valphah);
103 const float32x4_t vt0123 = vfmaq_f32(vtl0123, vtd0123, valphah);
104 const float32x4_t vb0123 = vfmaq_f32(vbl0123, vbd0123, valphah);
138 const float32x4_t vt0123 = vfmaq_f32(vtl0123, vtd0123, valphah);
139 const float32x4_t vb0123 = vfmaq_f32(vbl0123, vbd0123, valphah);
|
| /external/XNNPACK/src/f32-ibilinear-chw/gen/ |
| D | scalar-p1.c | 39 const float valphah = w[0]; in xnn_f32_ibilinear_chw_ukernel__scalar_p1() local
51 const float vt = vtl + vtd * valphah; in xnn_f32_ibilinear_chw_ukernel__scalar_p1()
52 const float vb = vbl + vbd * valphah; in xnn_f32_ibilinear_chw_ukernel__scalar_p1()
|
| D | scalar-p2.c | 82 const float valphah = w[0]; in xnn_f32_ibilinear_chw_ukernel__scalar_p2() local
94 const float vt = vtl + vtd * valphah; in xnn_f32_ibilinear_chw_ukernel__scalar_p2()
95 const float vb = vbl + vbd * valphah; in xnn_f32_ibilinear_chw_ukernel__scalar_p2()
|
| D | wasmsimd-p4.c | 59 const v128_t valphah = wasm_v32x4_shuffle(vw0, vw1, 0, 2, 4, 6); in xnn_f32_ibilinear_chw_ukernel__wasmsimd_p4() local
80 const v128_t vo = wasm_f32x4_add(vl, wasm_f32x4_mul(vd, valphah)); in xnn_f32_ibilinear_chw_ukernel__wasmsimd_p4()
91 const v128_t valphah = wasm_v32x4_shuffle(vw, vw, 0, 2, 0, 2); in xnn_f32_ibilinear_chw_ukernel__wasmsimd_p4() local
114 const v128_t vo = wasm_f32x4_add(vl, wasm_f32x4_mul(vd, valphah)); in xnn_f32_ibilinear_chw_ukernel__wasmsimd_p4()
|
| D | neonfma-p4.c | 58 const float32x4_t valphah = vw.val[0]; in xnn_f32_ibilinear_chw_ukernel__neonfma_p4() local
81 const float32x4_t vo = vfmaq_f32(vl, vd, valphah); in xnn_f32_ibilinear_chw_ukernel__neonfma_p4()
92 const float32x2_t valphah = vw.val[0]; in xnn_f32_ibilinear_chw_ukernel__neonfma_p4() local
121 const float32x2_t vo = vfma_f32(vl, vd, valphah); in xnn_f32_ibilinear_chw_ukernel__neonfma_p4()
|
| D | neon-p4.c | 58 const float32x4_t valphah = vw.val[0]; in xnn_f32_ibilinear_chw_ukernel__neon_p4() local
81 const float32x4_t vo = vmlaq_f32(vl, vd, valphah); in xnn_f32_ibilinear_chw_ukernel__neon_p4()
92 const float32x2_t valphah = vw.val[0]; in xnn_f32_ibilinear_chw_ukernel__neon_p4() local
121 const float32x2_t vo = vmla_f32(vl, vd, valphah); in xnn_f32_ibilinear_chw_ukernel__neon_p4()
|
| D | scalar-p4.c | 112 const float valphah = w[0]; in xnn_f32_ibilinear_chw_ukernel__scalar_p4() local
124 const float vt = vtl + vtd * valphah; in xnn_f32_ibilinear_chw_ukernel__scalar_p4()
125 const float vb = vbl + vbd * valphah; in xnn_f32_ibilinear_chw_ukernel__scalar_p4()
|
| D | wasmsimd-p8.c | 145 const v128_t valphah = wasm_v32x4_shuffle(vw0, vw1, 0, 2, 4, 6); in xnn_f32_ibilinear_chw_ukernel__wasmsimd_p8() local
166 const v128_t vo = wasm_f32x4_add(vl, wasm_f32x4_mul(vd, valphah)); in xnn_f32_ibilinear_chw_ukernel__wasmsimd_p8()
177 const v128_t valphah = wasm_v32x4_shuffle(vw, vw, 0, 2, 0, 2); in xnn_f32_ibilinear_chw_ukernel__wasmsimd_p8() local
200 const v128_t vo = wasm_f32x4_add(vl, wasm_f32x4_mul(vd, valphah)); in xnn_f32_ibilinear_chw_ukernel__wasmsimd_p8()
|
| D | neonfma-p8.c | 146 const float32x4_t valphah = vw.val[0]; in xnn_f32_ibilinear_chw_ukernel__neonfma_p8() local
169 const float32x4_t vo = vfmaq_f32(vl, vd, valphah); in xnn_f32_ibilinear_chw_ukernel__neonfma_p8()
180 const float32x2_t valphah = vw.val[0]; in xnn_f32_ibilinear_chw_ukernel__neonfma_p8() local
209 const float32x2_t vo = vfma_f32(vl, vd, valphah); in xnn_f32_ibilinear_chw_ukernel__neonfma_p8()
|
| D | neon-p8.c | 146 const float32x4_t valphah = vw.val[0]; in xnn_f32_ibilinear_chw_ukernel__neon_p8() local
169 const float32x4_t vo = vmlaq_f32(vl, vd, valphah); in xnn_f32_ibilinear_chw_ukernel__neon_p8()
180 const float32x2_t valphah = vw.val[0]; in xnn_f32_ibilinear_chw_ukernel__neon_p8() local
209 const float32x2_t vo = vmla_f32(vl, vd, valphah); in xnn_f32_ibilinear_chw_ukernel__neon_p8()
|