| /external/XNNPACK/src/f32-ppmm/gen/ |
| D | 8x8-minmax-neonfma.c | 88 const float32x4_t vb0123 = vld1q_f32(w); w += 4; in xnn_f32_ppmm_minmax_ukernel_8x8__neonfma() local
92 vacc0x0123 = vfmaq_laneq_f32(vacc0x0123, vb0123, va0123, 0); in xnn_f32_ppmm_minmax_ukernel_8x8__neonfma()
93 vacc1x0123 = vfmaq_laneq_f32(vacc1x0123, vb0123, va0123, 1); in xnn_f32_ppmm_minmax_ukernel_8x8__neonfma()
94 vacc2x0123 = vfmaq_laneq_f32(vacc2x0123, vb0123, va0123, 2); in xnn_f32_ppmm_minmax_ukernel_8x8__neonfma()
95 vacc3x0123 = vfmaq_laneq_f32(vacc3x0123, vb0123, va0123, 3); in xnn_f32_ppmm_minmax_ukernel_8x8__neonfma()
96 vacc4x0123 = vfmaq_laneq_f32(vacc4x0123, vb0123, va4567, 0); in xnn_f32_ppmm_minmax_ukernel_8x8__neonfma()
97 vacc5x0123 = vfmaq_laneq_f32(vacc5x0123, vb0123, va4567, 1); in xnn_f32_ppmm_minmax_ukernel_8x8__neonfma()
98 vacc6x0123 = vfmaq_laneq_f32(vacc6x0123, vb0123, va4567, 2); in xnn_f32_ppmm_minmax_ukernel_8x8__neonfma()
99 vacc7x0123 = vfmaq_laneq_f32(vacc7x0123, vb0123, va4567, 3); in xnn_f32_ppmm_minmax_ukernel_8x8__neonfma()
118 vacc0x0123 = vfmaq_f32(vacc0x0123, va0000, vb0123); in xnn_f32_ppmm_minmax_ukernel_8x8__neonfma()
[all …]
|
| D | 4x8-minmax-neonfma.c | 63 const float32x4_t vb0123 = vld1q_f32(w); w += 4; in xnn_f32_ppmm_minmax_ukernel_4x8__neonfma() local
67 vacc0x0123 = vfmaq_laneq_f32(vacc0x0123, vb0123, va0123, 0); in xnn_f32_ppmm_minmax_ukernel_4x8__neonfma()
68 vacc1x0123 = vfmaq_laneq_f32(vacc1x0123, vb0123, va0123, 1); in xnn_f32_ppmm_minmax_ukernel_4x8__neonfma()
69 vacc2x0123 = vfmaq_laneq_f32(vacc2x0123, vb0123, va0123, 2); in xnn_f32_ppmm_minmax_ukernel_4x8__neonfma()
70 vacc3x0123 = vfmaq_laneq_f32(vacc3x0123, vb0123, va0123, 3); in xnn_f32_ppmm_minmax_ukernel_4x8__neonfma()
81 vacc0x0123 = vfmaq_f32(vacc0x0123, va0000, vb0123); in xnn_f32_ppmm_minmax_ukernel_4x8__neonfma()
82 vacc1x0123 = vfmaq_f32(vacc1x0123, va1111, vb0123); in xnn_f32_ppmm_minmax_ukernel_4x8__neonfma()
83 vacc2x0123 = vfmaq_f32(vacc2x0123, va2222, vb0123); in xnn_f32_ppmm_minmax_ukernel_4x8__neonfma()
84 vacc3x0123 = vfmaq_f32(vacc3x0123, va3333, vb0123); in xnn_f32_ppmm_minmax_ukernel_4x8__neonfma()
|
| /external/XNNPACK/src/f32-vbinary/gen/ |
| D | vmin-sse-x8.c | 38 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vmin_ukernel__sse_x8() local
42 __m128 vy0123 = _mm_min_ps(va0123, vb0123); in xnn_f32_vmin_ukernel__sse_x8()
55 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vmin_ukernel__sse_x8() local
58 __m128 vy0123 = _mm_min_ps(va0123, vb0123); in xnn_f32_vmin_ukernel__sse_x8()
64 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vmin_ukernel__sse_x8() local
66 __m128 vy0123 = _mm_min_ps(va0123, vb0123); in xnn_f32_vmin_ukernel__sse_x8()
|
| D | vmax-sse-x8.c | 38 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vmax_ukernel__sse_x8() local
42 __m128 vy0123 = _mm_max_ps(va0123, vb0123); in xnn_f32_vmax_ukernel__sse_x8()
55 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vmax_ukernel__sse_x8() local
58 __m128 vy0123 = _mm_max_ps(va0123, vb0123); in xnn_f32_vmax_ukernel__sse_x8()
64 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vmax_ukernel__sse_x8() local
66 __m128 vy0123 = _mm_max_ps(va0123, vb0123); in xnn_f32_vmax_ukernel__sse_x8()
|
| D | vmin-neon-x8.c | 34 const float32x4_t vb0123 = vld1q_f32(b); b += 4; in xnn_f32_vmin_ukernel__neon_x8() local
38 float32x4_t vy0123 = vminq_f32(va0123, vb0123); in xnn_f32_vmin_ukernel__neon_x8()
48 const float32x4_t vb0123 = vld1q_f32(b); b += 4; in xnn_f32_vmin_ukernel__neon_x8() local
50 float32x4_t vy0123 = vminq_f32(va0123, vb0123); in xnn_f32_vmin_ukernel__neon_x8()
55 const float32x4_t vb0123 = vld1q_f32(b); in xnn_f32_vmin_ukernel__neon_x8() local
57 float32x4_t vy0123 = vminq_f32(va0123, vb0123); in xnn_f32_vmin_ukernel__neon_x8()
|
| D | vmax-neon-x8.c | 34 const float32x4_t vb0123 = vld1q_f32(b); b += 4; in xnn_f32_vmax_ukernel__neon_x8() local
38 float32x4_t vy0123 = vmaxq_f32(va0123, vb0123); in xnn_f32_vmax_ukernel__neon_x8()
48 const float32x4_t vb0123 = vld1q_f32(b); b += 4; in xnn_f32_vmax_ukernel__neon_x8() local
50 float32x4_t vy0123 = vmaxq_f32(va0123, vb0123); in xnn_f32_vmax_ukernel__neon_x8()
55 const float32x4_t vb0123 = vld1q_f32(b); in xnn_f32_vmax_ukernel__neon_x8() local
57 float32x4_t vy0123 = vmaxq_f32(va0123, vb0123); in xnn_f32_vmax_ukernel__neon_x8()
|
| D | vsqrdiff-sse-x8.c | 38 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vsqrdiff_ukernel__sse_x8() local
42 __m128 vy0123 = _mm_sub_ps(va0123, vb0123); in xnn_f32_vsqrdiff_ukernel__sse_x8()
57 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vsqrdiff_ukernel__sse_x8() local
60 __m128 vy0123 = _mm_sub_ps(va0123, vb0123); in xnn_f32_vsqrdiff_ukernel__sse_x8()
67 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vsqrdiff_ukernel__sse_x8() local
69 __m128 vy0123 = _mm_sub_ps(va0123, vb0123); in xnn_f32_vsqrdiff_ukernel__sse_x8()
|
| D | vsqrdiff-neon-x8.c | 34 const float32x4_t vb0123 = vld1q_f32(b); b += 4; in xnn_f32_vsqrdiff_ukernel__neon_x8() local
38 float32x4_t vy0123 = vsubq_f32(va0123, vb0123); in xnn_f32_vsqrdiff_ukernel__neon_x8()
50 const float32x4_t vb0123 = vld1q_f32(b); b += 4; in xnn_f32_vsqrdiff_ukernel__neon_x8() local
52 float32x4_t vy0123 = vsubq_f32(va0123, vb0123); in xnn_f32_vsqrdiff_ukernel__neon_x8()
58 const float32x4_t vb0123 = vld1q_f32(b); in xnn_f32_vsqrdiff_ukernel__neon_x8() local
60 float32x4_t vy0123 = vsubq_f32(va0123, vb0123); in xnn_f32_vsqrdiff_ukernel__neon_x8()
|
| D | vmul-minmax-neon-x8.c | 36 const float32x4_t vb0123 = vld1q_f32(b); b += 4; in xnn_f32_vmul_minmax_ukernel__neon_x8() local
40 float32x4_t vy0123 = vmulq_f32(va0123, vb0123); in xnn_f32_vmul_minmax_ukernel__neon_x8()
55 const float32x4_t vb0123 = vld1q_f32(b); b += 4; in xnn_f32_vmul_minmax_ukernel__neon_x8() local
57 float32x4_t vy0123 = vmulq_f32(va0123, vb0123); in xnn_f32_vmul_minmax_ukernel__neon_x8()
64 const float32x4_t vb0123 = vld1q_f32(b); in xnn_f32_vmul_minmax_ukernel__neon_x8() local
66 float32x4_t vy0123 = vmulq_f32(va0123, vb0123); in xnn_f32_vmul_minmax_ukernel__neon_x8()
|
| D | vadd-minmax-sse-x8.c | 40 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vadd_minmax_ukernel__sse_x8() local
44 __m128 vy0123 = _mm_add_ps(va0123, vb0123); in xnn_f32_vadd_minmax_ukernel__sse_x8()
62 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vadd_minmax_ukernel__sse_x8() local
65 __m128 vy0123 = _mm_add_ps(va0123, vb0123); in xnn_f32_vadd_minmax_ukernel__sse_x8()
73 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vadd_minmax_ukernel__sse_x8() local
75 __m128 vy0123 = _mm_add_ps(va0123, vb0123); in xnn_f32_vadd_minmax_ukernel__sse_x8()
|
| D | vdiv-minmax-neon-x8.c | 36 const float32x4_t vb0123 = vld1q_f32(b); b += 4; in xnn_f32_vdiv_minmax_ukernel__neon_x8() local
40 float32x4_t vy0123 = vdivq_f32(va0123, vb0123); in xnn_f32_vdiv_minmax_ukernel__neon_x8()
55 const float32x4_t vb0123 = vld1q_f32(b); b += 4; in xnn_f32_vdiv_minmax_ukernel__neon_x8() local
57 float32x4_t vy0123 = vdivq_f32(va0123, vb0123); in xnn_f32_vdiv_minmax_ukernel__neon_x8()
64 const float32x4_t vb0123 = vld1q_f32(b); in xnn_f32_vdiv_minmax_ukernel__neon_x8() local
66 float32x4_t vy0123 = vdivq_f32(va0123, vb0123); in xnn_f32_vdiv_minmax_ukernel__neon_x8()
|
| D | vadd-minmax-neon-x8.c | 36 const float32x4_t vb0123 = vld1q_f32(b); b += 4; in xnn_f32_vadd_minmax_ukernel__neon_x8() local
40 float32x4_t vy0123 = vaddq_f32(va0123, vb0123); in xnn_f32_vadd_minmax_ukernel__neon_x8()
55 const float32x4_t vb0123 = vld1q_f32(b); b += 4; in xnn_f32_vadd_minmax_ukernel__neon_x8() local
57 float32x4_t vy0123 = vaddq_f32(va0123, vb0123); in xnn_f32_vadd_minmax_ukernel__neon_x8()
64 const float32x4_t vb0123 = vld1q_f32(b); in xnn_f32_vadd_minmax_ukernel__neon_x8() local
66 float32x4_t vy0123 = vaddq_f32(va0123, vb0123); in xnn_f32_vadd_minmax_ukernel__neon_x8()
|
| D | vmul-minmax-sse-x8.c | 40 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vmul_minmax_ukernel__sse_x8() local
44 __m128 vy0123 = _mm_mul_ps(va0123, vb0123); in xnn_f32_vmul_minmax_ukernel__sse_x8()
62 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vmul_minmax_ukernel__sse_x8() local
65 __m128 vy0123 = _mm_mul_ps(va0123, vb0123); in xnn_f32_vmul_minmax_ukernel__sse_x8()
73 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vmul_minmax_ukernel__sse_x8() local
75 __m128 vy0123 = _mm_mul_ps(va0123, vb0123); in xnn_f32_vmul_minmax_ukernel__sse_x8()
|
| D | vdiv-minmax-sse-x8.c | 40 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vdiv_minmax_ukernel__sse_x8() local
44 __m128 vy0123 = _mm_div_ps(va0123, vb0123); in xnn_f32_vdiv_minmax_ukernel__sse_x8()
62 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vdiv_minmax_ukernel__sse_x8() local
65 __m128 vy0123 = _mm_div_ps(va0123, vb0123); in xnn_f32_vdiv_minmax_ukernel__sse_x8()
73 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vdiv_minmax_ukernel__sse_x8() local
75 __m128 vy0123 = _mm_div_ps(va0123, vb0123); in xnn_f32_vdiv_minmax_ukernel__sse_x8()
|
| D | vsub-minmax-sse-x8.c | 40 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vsub_minmax_ukernel__sse_x8() local
44 __m128 vy0123 = _mm_sub_ps(va0123, vb0123); in xnn_f32_vsub_minmax_ukernel__sse_x8()
62 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vsub_minmax_ukernel__sse_x8() local
65 __m128 vy0123 = _mm_sub_ps(va0123, vb0123); in xnn_f32_vsub_minmax_ukernel__sse_x8()
73 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vsub_minmax_ukernel__sse_x8() local
75 __m128 vy0123 = _mm_sub_ps(va0123, vb0123); in xnn_f32_vsub_minmax_ukernel__sse_x8()
|
| D | vsub-minmax-neon-x8.c | 36 const float32x4_t vb0123 = vld1q_f32(b); b += 4; in xnn_f32_vsub_minmax_ukernel__neon_x8() local
40 float32x4_t vy0123 = vsubq_f32(va0123, vb0123); in xnn_f32_vsub_minmax_ukernel__neon_x8()
55 const float32x4_t vb0123 = vld1q_f32(b); b += 4; in xnn_f32_vsub_minmax_ukernel__neon_x8() local
57 float32x4_t vy0123 = vsubq_f32(va0123, vb0123); in xnn_f32_vsub_minmax_ukernel__neon_x8()
64 const float32x4_t vb0123 = vld1q_f32(b); in xnn_f32_vsub_minmax_ukernel__neon_x8() local
66 float32x4_t vy0123 = vsubq_f32(va0123, vb0123); in xnn_f32_vsub_minmax_ukernel__neon_x8()
|
| D | vmax-neon-x4.c | 34 const float32x4_t vb0123 = vld1q_f32(b); b += 4; in xnn_f32_vmax_ukernel__neon_x4() local
36 float32x4_t vy0123 = vmaxq_f32(va0123, vb0123); in xnn_f32_vmax_ukernel__neon_x4()
44 const float32x4_t vb0123 = vld1q_f32(b); in xnn_f32_vmax_ukernel__neon_x4() local
46 float32x4_t vy0123 = vmaxq_f32(va0123, vb0123); in xnn_f32_vmax_ukernel__neon_x4()
|
| D | vmax-sse-x4.c | 37 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vmax_ukernel__sse_x4() local
40 __m128 vy0123 = _mm_max_ps(va0123, vb0123); in xnn_f32_vmax_ukernel__sse_x4()
49 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vmax_ukernel__sse_x4() local
51 __m128 vy0123 = _mm_max_ps(va0123, vb0123); in xnn_f32_vmax_ukernel__sse_x4()
|
| D | vmin-neon-x4.c | 34 const float32x4_t vb0123 = vld1q_f32(b); b += 4; in xnn_f32_vmin_ukernel__neon_x4() local
36 float32x4_t vy0123 = vminq_f32(va0123, vb0123); in xnn_f32_vmin_ukernel__neon_x4()
44 const float32x4_t vb0123 = vld1q_f32(b); in xnn_f32_vmin_ukernel__neon_x4() local
46 float32x4_t vy0123 = vminq_f32(va0123, vb0123); in xnn_f32_vmin_ukernel__neon_x4()
|
| D | vmin-sse-x4.c | 37 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vmin_ukernel__sse_x4() local
40 __m128 vy0123 = _mm_min_ps(va0123, vb0123); in xnn_f32_vmin_ukernel__sse_x4()
49 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vmin_ukernel__sse_x4() local
51 __m128 vy0123 = _mm_min_ps(va0123, vb0123); in xnn_f32_vmin_ukernel__sse_x4()
|
| D | vsqrdiff-sse-x4.c | 37 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vsqrdiff_ukernel__sse_x4() local
40 __m128 vy0123 = _mm_sub_ps(va0123, vb0123); in xnn_f32_vsqrdiff_ukernel__sse_x4()
50 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vsqrdiff_ukernel__sse_x4() local
52 __m128 vy0123 = _mm_sub_ps(va0123, vb0123); in xnn_f32_vsqrdiff_ukernel__sse_x4()
|
| D | vsqrdiff-neon-x4.c | 34 const float32x4_t vb0123 = vld1q_f32(b); b += 4; in xnn_f32_vsqrdiff_ukernel__neon_x4() local
36 float32x4_t vy0123 = vsubq_f32(va0123, vb0123); in xnn_f32_vsqrdiff_ukernel__neon_x4()
45 const float32x4_t vb0123 = vld1q_f32(b); in xnn_f32_vsqrdiff_ukernel__neon_x4() local
47 float32x4_t vy0123 = vsubq_f32(va0123, vb0123); in xnn_f32_vsqrdiff_ukernel__neon_x4()
|
| D | vsub-minmax-sse-x4.c | 39 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vsub_minmax_ukernel__sse_x4() local
42 __m128 vy0123 = _mm_sub_ps(va0123, vb0123); in xnn_f32_vsub_minmax_ukernel__sse_x4()
54 const __m128 vb0123 = _mm_loadu_ps(b); in xnn_f32_vsub_minmax_ukernel__sse_x4() local
56 __m128 vy0123 = _mm_sub_ps(va0123, vb0123); in xnn_f32_vsub_minmax_ukernel__sse_x4()
|
| /external/XNNPACK/src/f32-ibilinear/gen/ |
| D | neonfma-c8.c | 62 const float32x4_t vb0123 = vfmaq_f32(vbl0123, vbd0123, valphah); in xnn_f32_ibilinear_ukernel__neonfma_c8() local
67 const float32x4_t vb0123 = vfmaq_lane_f32(vbl0123, vbd0123, valphahv, 0); in xnn_f32_ibilinear_ukernel__neonfma_c8() local
72 const float32x4_t vd0123 = vsubq_f32(vb0123, vt0123); in xnn_f32_ibilinear_ukernel__neonfma_c8()
97 const float32x4_t vb0123 = vfmaq_f32(vbl0123, vbd0123, valphah); in xnn_f32_ibilinear_ukernel__neonfma_c8() local
100 const float32x4_t vb0123 = vfmaq_lane_f32(vbl0123, vbd0123, valphahv, 0); in xnn_f32_ibilinear_ukernel__neonfma_c8() local
103 const float32x4_t vd0123 = vsubq_f32(vb0123, vt0123); in xnn_f32_ibilinear_ukernel__neonfma_c8()
125 const float32x4_t vb0123 = vfmaq_f32(vbl0123, vbd0123, valphah); in xnn_f32_ibilinear_ukernel__neonfma_c8() local
128 const float32x4_t vb0123 = vfmaq_lane_f32(vbl0123, vbd0123, valphahv, 0); in xnn_f32_ibilinear_ukernel__neonfma_c8() local
131 const float32x4_t vd0123 = vsubq_f32(vb0123, vt0123); in xnn_f32_ibilinear_ukernel__neonfma_c8()
|
| D | neonfma-c4.c | 56 const float32x4_t vb0123 = vfmaq_f32(vbl0123, vbd0123, valphah); in xnn_f32_ibilinear_ukernel__neonfma_c4() local
59 const float32x4_t vb0123 = vfmaq_lane_f32(vbl0123, vbd0123, valphahv, 0); in xnn_f32_ibilinear_ukernel__neonfma_c4() local
62 const float32x4_t vd0123 = vsubq_f32(vb0123, vt0123); in xnn_f32_ibilinear_ukernel__neonfma_c4()
83 const float32x4_t vb0123 = vfmaq_f32(vbl0123, vbd0123, valphah); in xnn_f32_ibilinear_ukernel__neonfma_c4() local
86 const float32x4_t vb0123 = vfmaq_lane_f32(vbl0123, vbd0123, valphahv, 0); in xnn_f32_ibilinear_ukernel__neonfma_c4() local
89 const float32x4_t vd0123 = vsubq_f32(vb0123, vt0123); in xnn_f32_ibilinear_ukernel__neonfma_c4()
|