Home
last modified time | relevance | path

Searched refs:vbias (Results 1 – 25 of 25) sorted by relevance

/external/XNNPACK/src/f32-vmulcaddc/gen/
Dc1-scalar-2x.c53 const float vbias = w[1]; in xnn_f32_vmulcaddc_ukernel_c1__scalar_2x() local
55 vacc0 = vacc0 * vscale + vbias; in xnn_f32_vmulcaddc_ukernel_c1__scalar_2x()
56 vacc1 = vacc1 * vscale + vbias; in xnn_f32_vmulcaddc_ukernel_c1__scalar_2x()
Dc1-wasm-2x.c53 const float vbias = w[1]; in xnn_f32_vmulcaddc_ukernel_c1__wasm_2x() local
55 vacc0 = vacc0 * vscale + vbias; in xnn_f32_vmulcaddc_ukernel_c1__wasm_2x()
56 vacc1 = vacc1 * vscale + vbias; in xnn_f32_vmulcaddc_ukernel_c1__wasm_2x()
Dc2-scalar-2x.c92 const float vbias = w[1]; in xnn_f32_vmulcaddc_ukernel_c2__scalar_2x() local
94 vacc0 = vacc0 * vscale + vbias; in xnn_f32_vmulcaddc_ukernel_c2__scalar_2x()
95 vacc1 = vacc1 * vscale + vbias; in xnn_f32_vmulcaddc_ukernel_c2__scalar_2x()
Dc2-wasm-2x.c92 const float vbias = w[1]; in xnn_f32_vmulcaddc_ukernel_c2__wasm_2x() local
94 vacc0 = vacc0 * vscale + vbias; in xnn_f32_vmulcaddc_ukernel_c2__wasm_2x()
95 vacc1 = vacc1 * vscale + vbias; in xnn_f32_vmulcaddc_ukernel_c2__wasm_2x()
Dc4-scalar-2x.c116 const float vbias = w[3]; in xnn_f32_vmulcaddc_ukernel_c4__scalar_2x() local
118 vacc0 = vacc0 * vscale + vbias; in xnn_f32_vmulcaddc_ukernel_c4__scalar_2x()
119 vacc1 = vacc1 * vscale + vbias; in xnn_f32_vmulcaddc_ukernel_c4__scalar_2x()
Dc4-wasm-2x.c116 const float vbias = w[3]; in xnn_f32_vmulcaddc_ukernel_c4__wasm_2x() local
118 vacc0 = vacc0 * vscale + vbias; in xnn_f32_vmulcaddc_ukernel_c4__wasm_2x()
119 vacc1 = vacc1 * vscale + vbias; in xnn_f32_vmulcaddc_ukernel_c4__wasm_2x()
/external/XNNPACK/src/f32-vmulcaddc/
Dscalar.c.in66 const float vbias${ABC[C]} = w[${C + CHANNEL_TILE}];
70 vacc${M}x${ABC[C]} = vacc${M}x${ABC[C]} * vscale${ABC[C]} + vbias${ABC[C]};
94 const float vbias = w[${CHANNEL_TILE - 1}];
97 vacc${M} = vacc${M} * vscale + vbias;
118 const float vbias = w[1];
121 vacc${M} = vacc${M} * vscale + vbias;
Dneon.c.in70 const float32x4_t vbias${ABC[C:C+4]} = vld1q_f32(w); w += 4;
75 vacc${M}x${ABC[C:C+4]} = vaddq_f32(vacc${M}x${ABC[C:C+4]}, vbias${ABC[C:C+4]});
79 …vacc${M}x${ABC[C:C+4]} = vfmaq_f32(vbias${ABC[C:C+4]}, vscale${ABC[C:C+4]}, vacc${M}x${ABC[C:C+4]}…
Dpsimd.c.in68 const psimd_f32 vbias${ABC[C:C+4]} = psimd_load_f32(w + ${C + CHANNEL_TILE});
72 …vacc${M}x${ABC[C:C+4]} = psimd_qfma_f32(vbias${ABC[C:C+4]}, vscale${ABC[C:C+4]}, vacc${M}x${ABC[C:…
Dsse.c.in72 const __m128 vbias${ABC[C:C+4]} = _mm_load_ps(w + ${C + CHANNEL_TILE});
76 vacc${M}x${ABC[C:C+4]} = _mm_add_ps(vacc${M}x${ABC[C:C+4]}, vbias${ABC[C:C+4]});
/external/XNNPACK/src/q8-avgpool/
Dup9-neon.c33 const int32x4_t vbias = vld1q_dup_s32(&params->neon.bias); in xnn_q8_avgpool_ukernel_up9__neon() local
101 int32x4_t vacc_lo = vaddw_s16(vbias, vreinterpret_s16_u16(vget_low_u16(vsum))); in xnn_q8_avgpool_ukernel_up9__neon()
102 int32x4_t vacc_hi = vaddw_s16(vbias, vreinterpret_s16_u16(vget_high_u16(vsum))); in xnn_q8_avgpool_ukernel_up9__neon()
174 int32x4_t vacc_lo = vaddw_s16(vbias, vreinterpret_s16_u16(vget_low_u16(vsum))); in xnn_q8_avgpool_ukernel_up9__neon()
175 int32x4_t vacc_hi = vaddw_s16(vbias, vreinterpret_s16_u16(vget_high_u16(vsum))); in xnn_q8_avgpool_ukernel_up9__neon()
Dup9-sse2.c32 const __m128i vbias = _mm_load_si128((const __m128i*) &params->sse2.bias); in xnn_q8_avgpool_ukernel_up9__sse2() local
105 const __m128i vacc_lo = _mm_add_epi32(vbias, _mm_unpacklo_epi16(vsum, vzero)); in xnn_q8_avgpool_ukernel_up9__sse2()
106 const __m128i vacc_hi = _mm_add_epi32(vbias, _mm_unpackhi_epi16(vsum, vzero)); in xnn_q8_avgpool_ukernel_up9__sse2()
180 const __m128i vacc_lo = _mm_add_epi32(vbias, _mm_unpacklo_epi16(vsum, vzero)); in xnn_q8_avgpool_ukernel_up9__sse2()
181 const __m128i vacc_hi = _mm_add_epi32(vbias, _mm_unpackhi_epi16(vsum, vzero)); in xnn_q8_avgpool_ukernel_up9__sse2()
Dup9-scalar.c31 const int32_t vbias = params->scalar.bias; in xnn_q8_avgpool_ukernel_up9__scalar() local
95 const int32_t vacc = vbias + (int32_t) vsum; in xnn_q8_avgpool_ukernel_up9__scalar()
Dmp9p8q-scalar.c31 const int32_t vbias = params->scalar.bias; in xnn_q8_avgpool_ukernel_mp9p8q__scalar() local
71 int32_t vacc = vbias + (int32_t) vsum2345; in xnn_q8_avgpool_ukernel_mp9p8q__scalar()
Dmp9p8q-neon.c33 const int32x4_t vbias = vld1q_dup_s32(&params->neon.bias); in xnn_q8_avgpool_ukernel_mp9p8q__neon() local
77 const int32x4_t vacc_lo = vaddw_s16(vbias, vreinterpret_s16_u16(vget_low_u16(vsum))); in xnn_q8_avgpool_ukernel_mp9p8q__neon()
78 const int32x4_t vacc_hi = vaddw_s16(vbias, vreinterpret_s16_u16(vget_high_u16(vsum))); in xnn_q8_avgpool_ukernel_mp9p8q__neon()
Dmp9p8q-sse2.c32 const __m128i vbias = _mm_load_si128((const __m128i*) &params->sse2.bias); in xnn_q8_avgpool_ukernel_mp9p8q__sse2() local
81 const __m128i vacc_lo = _mm_add_epi32(vbias, _mm_unpacklo_epi16(vsum, vzero)); in xnn_q8_avgpool_ukernel_mp9p8q__sse2()
82 const __m128i vacc_hi = _mm_add_epi32(vbias, _mm_unpackhi_epi16(vsum, vzero)); in xnn_q8_avgpool_ukernel_mp9p8q__sse2()
/external/XNNPACK/src/q8-gavgpool/
Dup7-sse2.c55 const __m128i vbias = _mm_load_si128((const __m128i*) &params->sse2.bias); in xnn_q8_gavgpool_ukernel_up7__sse2() local
86 __m128i vacc_lo = _mm_add_epi32(vbias, _mm_unpacklo_epi16(vsum, vzero)); in xnn_q8_gavgpool_ukernel_up7__sse2()
87 __m128i vacc_hi = _mm_add_epi32(vbias, _mm_unpackhi_epi16(vsum, vzero)); in xnn_q8_gavgpool_ukernel_up7__sse2()
155 __m128i vacc_lo = _mm_add_epi32(vbias, _mm_unpacklo_epi16(vsum, vzero)); in xnn_q8_gavgpool_ukernel_up7__sse2()
156 __m128i vacc_hi = _mm_add_epi32(vbias, _mm_unpackhi_epi16(vsum, vzero)); in xnn_q8_gavgpool_ukernel_up7__sse2()
Dup7-neon.c56 const int32x4_t vbias = vld1q_dup_s32(&params->neon.bias); in xnn_q8_gavgpool_ukernel_up7__neon() local
83 int32x4_t vacc_lo = vaddw_s16(vbias, vget_low_s16(vsum)); in xnn_q8_gavgpool_ukernel_up7__neon()
84 int32x4_t vacc_hi = vaddw_s16(vbias, vget_high_s16(vsum)); in xnn_q8_gavgpool_ukernel_up7__neon()
153 int32x4_t vacc_lo = vaddw_s16(vbias, vget_low_s16(vsum)); in xnn_q8_gavgpool_ukernel_up7__neon()
154 int32x4_t vacc_hi = vaddw_s16(vbias, vget_high_s16(vsum)); in xnn_q8_gavgpool_ukernel_up7__neon()
Dup7-scalar.c51 const int32_t vbias = params->scalar.bias; in xnn_q8_gavgpool_ukernel_up7__scalar() local
75 const int32_t vacc = vbias + (int32_t) vsum; in xnn_q8_gavgpool_ukernel_up7__scalar()
Dmp7p7q-scalar.c36 const int32_t vbias = params->scalar.bias; in xnn_q8_gavgpool_ukernel_mp7p7q__scalar() local
57 const int32_t vacc = vbias + (int32_t) vsum; in xnn_q8_gavgpool_ukernel_mp7p7q__scalar()
Dmp7p7q-neon.c40 const int32x4_t vbias = vld1q_dup_s32(&params->neon.bias); in xnn_q8_gavgpool_ukernel_mp7p7q__neon() local
61 const int32x4_t vacc_lo = vaddw_s16(vbias, vget_low_s16(vsum)); in xnn_q8_gavgpool_ukernel_mp7p7q__neon()
62 const int32x4_t vacc_hi = vaddw_s16(vbias, vget_high_s16(vsum)); in xnn_q8_gavgpool_ukernel_mp7p7q__neon()
Dmp7p7q-sse2.c39 const __m128i vbias = _mm_load_si128((const __m128i*) &params->sse2.bias); in xnn_q8_gavgpool_ukernel_mp7p7q__sse2() local
68 const __m128i vacc_lo = _mm_add_epi32(vbias, _mm_unpacklo_epi16(vsum, vzero)); in xnn_q8_gavgpool_ukernel_mp7p7q__sse2()
69 const __m128i vacc_hi = _mm_add_epi32(vbias, _mm_unpackhi_epi16(vsum, vzero)); in xnn_q8_gavgpool_ukernel_mp7p7q__sse2()
/external/XNNPACK/src/f32-dwconv-spchw/
D3x3p1-sse.c40 const __m128 vbias = _mm_load1_ps(weights); in xnn_f32_dwconv_spchw_ukernel_3x3p1__sse() local
70 __m128 vo4567p0 = vbias; in xnn_f32_dwconv_spchw_ukernel_3x3p1__sse()
143 __m128 vo4567p0 = vbias; in xnn_f32_dwconv_spchw_ukernel_3x3p1__sse()
D3x3s2p1-sse.c41 const __m128 vbias = _mm_load1_ps(weights); in xnn_f32_dwconv_spchw_ukernel_3x3s2p1__sse() local
59 __m128 vo8ACEp0 = vbias; in xnn_f32_dwconv_spchw_ukernel_3x3s2p1__sse()
118 __m128 vo8ACEp0 = vbias; in xnn_f32_dwconv_spchw_ukernel_3x3s2p1__sse()
/external/XNNPACK/src/f32-dwconv/
Dup4x9-aarch64-neonfma-cortex-a55.S57 # Load vbias.lo
60 # Load vbias.hi
351 # Load vbias.lo
354 # Load vbias.hi