| /external/XNNPACK/src/f32-ppmm/gen/ |
| D | 3x3-minmax-scalar.c | 46 float vacc0x2 = w[2]; in xnn_f32_ppmm_minmax_ukernel_3x3__scalar() local
49 float vacc1x2 = vacc0x2; in xnn_f32_ppmm_minmax_ukernel_3x3__scalar()
52 float vacc2x2 = vacc0x2; in xnn_f32_ppmm_minmax_ukernel_3x3__scalar()
73 vacc0x2 += va0 * vb2; in xnn_f32_ppmm_minmax_ukernel_3x3__scalar()
87 vacc0x2 = math_min_f32(vacc0x2, vmax); in xnn_f32_ppmm_minmax_ukernel_3x3__scalar()
98 vacc0x2 = math_max_f32(vacc0x2, vmin); in xnn_f32_ppmm_minmax_ukernel_3x3__scalar()
111 c0[2] = vacc0x2; in xnn_f32_ppmm_minmax_ukernel_3x3__scalar()
131 vacc0x0 = vacc0x2; in xnn_f32_ppmm_minmax_ukernel_3x3__scalar()
|
| D | 2x4-minmax-scalar.c | 42 float vacc0x2 = w[2]; in xnn_f32_ppmm_minmax_ukernel_2x4__scalar() local
46 float vacc1x2 = vacc0x2; in xnn_f32_ppmm_minmax_ukernel_2x4__scalar()
66 vacc0x2 += va0 * vb2; in xnn_f32_ppmm_minmax_ukernel_2x4__scalar()
79 vacc0x2 = math_min_f32(vacc0x2, vmax); in xnn_f32_ppmm_minmax_ukernel_2x4__scalar()
89 vacc0x2 = math_max_f32(vacc0x2, vmin); in xnn_f32_ppmm_minmax_ukernel_2x4__scalar()
101 c0[2] = vacc0x2; in xnn_f32_ppmm_minmax_ukernel_2x4__scalar()
118 vacc0x0 = vacc0x2; in xnn_f32_ppmm_minmax_ukernel_2x4__scalar()
|
| D | 4x4-minmax-scalar.c | 50 float vacc0x2 = w[2]; in xnn_f32_ppmm_minmax_ukernel_4x4__scalar() local
54 float vacc1x2 = vacc0x2; in xnn_f32_ppmm_minmax_ukernel_4x4__scalar()
58 float vacc2x2 = vacc0x2; in xnn_f32_ppmm_minmax_ukernel_4x4__scalar()
62 float vacc3x2 = vacc0x2; in xnn_f32_ppmm_minmax_ukernel_4x4__scalar()
88 vacc0x2 += va0 * vb2; in xnn_f32_ppmm_minmax_ukernel_4x4__scalar()
109 vacc0x2 = math_min_f32(vacc0x2, vmax); in xnn_f32_ppmm_minmax_ukernel_4x4__scalar()
127 vacc0x2 = math_max_f32(vacc0x2, vmin); in xnn_f32_ppmm_minmax_ukernel_4x4__scalar()
151 c0[2] = vacc0x2; in xnn_f32_ppmm_minmax_ukernel_4x4__scalar()
176 vacc0x0 = vacc0x2; in xnn_f32_ppmm_minmax_ukernel_4x4__scalar()
|
| /external/XNNPACK/src/f32-vmulcaddc/gen/ |
| D | c4-minmax-wasm-2x.c | 55 float vacc0x2 = i0[2]; in xnn_f32_vmulcaddc_minmax_ukernel_c4__wasm_2x() local
71 vacc0x2 = vacc0x2 * vscale2 + vbias2; in xnn_f32_vmulcaddc_minmax_ukernel_c4__wasm_2x()
80 vacc0x2 = __builtin_wasm_max_f32(vacc0x2, vmin); in xnn_f32_vmulcaddc_minmax_ukernel_c4__wasm_2x()
89 vacc0x2 = __builtin_wasm_min_f32(vacc0x2, vmax); in xnn_f32_vmulcaddc_minmax_ukernel_c4__wasm_2x()
98 o0[2] = vacc0x2; in xnn_f32_vmulcaddc_minmax_ukernel_c4__wasm_2x()
|
| D | c4-minmax-scalar-2x.c | 55 float vacc0x2 = i0[2]; in xnn_f32_vmulcaddc_minmax_ukernel_c4__scalar_2x() local
71 vacc0x2 = vacc0x2 * vscale2 + vbias2; in xnn_f32_vmulcaddc_minmax_ukernel_c4__scalar_2x()
80 vacc0x2 = math_max_f32(vacc0x2, vmin); in xnn_f32_vmulcaddc_minmax_ukernel_c4__scalar_2x()
89 vacc0x2 = math_min_f32(vacc0x2, vmax); in xnn_f32_vmulcaddc_minmax_ukernel_c4__scalar_2x()
98 o0[2] = vacc0x2; in xnn_f32_vmulcaddc_minmax_ukernel_c4__scalar_2x()
|
| /external/XNNPACK/src/f32-spmm/gen/ |
| D | 8x4-minmax-scalar.c | 57 float vacc0x2 = *w++; in xnn_f32_spmm_minmax_ukernel_8x4__scalar() local
58 float vacc1x2 = vacc0x2; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
59 float vacc2x2 = vacc0x2; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
60 float vacc3x2 = vacc0x2; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
61 float vacc4x2 = vacc0x2; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
62 float vacc5x2 = vacc0x2; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
63 float vacc6x2 = vacc0x2; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
64 float vacc7x2 = vacc0x2; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
105 vacc0x2 += vi0 * vw2; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
139 float vout0x2 = math_min_f32(vacc0x2, vmax); in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
[all …]
|
| /external/XNNPACK/src/qs8-gemm/gen/ |
| D | 1x4c8-minmax-wasmsimd-ld64.c | 47 v128_t vacc0x2 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[2]); in xnn_qs8_gemm_minmax_ukernel_1x4c8__wasmsimd_ld64() local
69 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_low_i16x8(vprod0x2)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__wasmsimd_ld64()
70 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_high_i16x8(vprod0x2)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__wasmsimd_ld64()
81 …02 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x0, vacc0x2, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x0, … in xnn_qs8_gemm_minmax_ukernel_1x4c8__wasmsimd_ld64()
|
| D | 1x4c8-minmax-wasmsimd-ld128.c | 47 v128_t vacc0x2 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[2]); in xnn_qs8_gemm_minmax_ukernel_1x4c8__wasmsimd_ld128() local
73 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_low_i16x8(vprod0x2)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__wasmsimd_ld128()
77 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_high_i16x8(vprod0x2)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__wasmsimd_ld128()
85 …02 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x0, vacc0x2, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x0, … in xnn_qs8_gemm_minmax_ukernel_1x4c8__wasmsimd_ld128()
|
| D | 1x4c8-xw-minmax-wasmsimd.c | 47 v128_t vacc0x2 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[2]); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__wasmsimd() local
69 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_low_i16x8(vprod0x2)); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__wasmsimd()
70 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_high_i16x8(vprod0x2)); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__wasmsimd()
81 …02 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x0, vacc0x2, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x0, … in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__wasmsimd()
|
| D | 2x4c8-xw-minmax-wasmsimd.c | 53 v128_t vacc0x2 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[2]); in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__wasmsimd() local
57 v128_t vacc1x2 = vacc0x2; in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__wasmsimd()
87 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_low_i16x8(vprod0x2)); in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__wasmsimd()
88 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_high_i16x8(vprod0x2)); in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__wasmsimd()
105 …02 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x0, vacc0x2, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x0, … in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__wasmsimd()
|
| D | 2x4c8-minmax-wasmsimd-ld128.c | 53 v128_t vacc0x2 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[2]); in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld128() local
57 v128_t vacc1x2 = vacc0x2; in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld128()
91 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_low_i16x8(vprod0x2)); in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld128()
97 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_high_i16x8(vprod0x2)); in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld128()
109 …02 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x0, vacc0x2, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x0, … in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld128()
|
| D | 2x4c8-minmax-wasmsimd-ld64.c | 53 v128_t vacc0x2 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[2]); in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld64() local
57 v128_t vacc1x2 = vacc0x2; in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld64()
87 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_low_i16x8(vprod0x2)); in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld64()
88 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_high_i16x8(vprod0x2)); in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld64()
105 …02 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x0, vacc0x2, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x0, … in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld64()
|
| D | 3x4c8-xw-minmax-wasmsimd.c | 59 v128_t vacc0x2 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[2]); in xnn_qs8_gemm_xw_minmax_ukernel_3x4c8__wasmsimd() local
63 v128_t vacc1x2 = vacc0x2; in xnn_qs8_gemm_xw_minmax_ukernel_3x4c8__wasmsimd()
67 v128_t vacc2x2 = vacc0x2; in xnn_qs8_gemm_xw_minmax_ukernel_3x4c8__wasmsimd()
105 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_low_i16x8(vprod0x2)); in xnn_qs8_gemm_xw_minmax_ukernel_3x4c8__wasmsimd()
106 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_high_i16x8(vprod0x2)); in xnn_qs8_gemm_xw_minmax_ukernel_3x4c8__wasmsimd()
129 …02 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x0, vacc0x2, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x0, … in xnn_qs8_gemm_xw_minmax_ukernel_3x4c8__wasmsimd()
|
| D | 3x4c8-minmax-wasmsimd-ld128.c | 59 v128_t vacc0x2 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[2]); in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld128() local
63 v128_t vacc1x2 = vacc0x2; in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
67 v128_t vacc2x2 = vacc0x2; in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
109 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_low_i16x8(vprod0x2)); in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
117 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_high_i16x8(vprod0x2)); in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
133 …02 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x0, vacc0x2, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x0, … in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
|
| D | 3x4c8-minmax-wasmsimd-ld64.c | 59 v128_t vacc0x2 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[2]); in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld64() local
63 v128_t vacc1x2 = vacc0x2; in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
67 v128_t vacc2x2 = vacc0x2; in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
105 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_low_i16x8(vprod0x2)); in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
106 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_high_i16x8(vprod0x2)); in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
129 …02 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x0, vacc0x2, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x0, … in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
|
| D | 1x8c8-minmax-neon-mlal-padal.c | 46 …int32x4_t vacc0x2 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mlal_padal() local
79 vacc0x2 = vpadalq_s16(vacc0x2, vprod0x2); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mlal_padal()
116 vacc0x2 = vpadalq_s16(vacc0x2, vprod0x2); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mlal_padal()
138 const int32x4_t vsum0x23 = vpaddq_s32(vacc0x2, vacc0x3); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mlal_padal()
146 const int32x2_t vpsum0x2 = vadd_s32(vget_low_s32(vacc0x2), vget_high_s32(vacc0x2)); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mlal_padal()
|
| D | 1x4c8-xw-minmax-sse2.c | 46 __m128i vacc0x2 = _mm_cvtsi32_si128((int) ((const int32_t*) w)[2]); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__sse2() local
64 vacc0x2 = _mm_add_epi32(vacc0x2, _mm_madd_epi16(vxa0, vxb2)); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__sse2()
73 … vacc0x02 = _mm_add_epi32(_mm_unpacklo_epi32(vacc0x0, vacc0x2), _mm_unpackhi_epi32(vacc0x0, vacc0x… in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__sse2()
|
| D | 1x4c8-minmax-sse2-ld128.c | 46 __m128i vacc0x2 = _mm_cvtsi32_si128((int) ((const int32_t*) w)[2]); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld128() local
68 vacc0x2 = _mm_add_epi32(vacc0x2, _mm_madd_epi16(vxa0, vxb2)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld128()
75 … vacc0x02 = _mm_add_epi32(_mm_unpacklo_epi32(vacc0x0, vacc0x2), _mm_unpackhi_epi32(vacc0x0, vacc0x… in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld128()
|
| D | 1x4c8-minmax-sse2-ld64.c | 46 __m128i vacc0x2 = _mm_cvtsi32_si128((int) ((const int32_t*) w)[2]); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld64() local
67 vacc0x2 = _mm_add_epi32(vacc0x2, _mm_madd_epi16(vxa0, vxb2)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld64()
77 … vacc0x02 = _mm_add_epi32(_mm_unpacklo_epi32(vacc0x0, vacc0x2), _mm_unpackhi_epi32(vacc0x0, vacc0x… in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld64()
|
| /external/XNNPACK/src/qs8-igemm/gen/ |
| D | 1x4c8-minmax-wasmsimd-ld64.c | 50 v128_t vacc0x2 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[2]); in xnn_qs8_igemm_minmax_ukernel_1x4c8__wasmsimd_ld64() local
80 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_low_i16x8(vprod0x2)); in xnn_qs8_igemm_minmax_ukernel_1x4c8__wasmsimd_ld64()
81 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_high_i16x8(vprod0x2)); in xnn_qs8_igemm_minmax_ukernel_1x4c8__wasmsimd_ld64()
94 …02 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x0, vacc0x2, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x0, … in xnn_qs8_igemm_minmax_ukernel_1x4c8__wasmsimd_ld64()
|
| D | 1x4c8-minmax-wasmsimd-ld128.c | 50 v128_t vacc0x2 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[2]); in xnn_qs8_igemm_minmax_ukernel_1x4c8__wasmsimd_ld128() local
84 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_low_i16x8(vprod0x2)); in xnn_qs8_igemm_minmax_ukernel_1x4c8__wasmsimd_ld128()
88 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_high_i16x8(vprod0x2)); in xnn_qs8_igemm_minmax_ukernel_1x4c8__wasmsimd_ld128()
98 …02 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x0, vacc0x2, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x0, … in xnn_qs8_igemm_minmax_ukernel_1x4c8__wasmsimd_ld128()
|
| D | 2x4c8-minmax-wasmsimd-ld128.c | 54 v128_t vacc0x2 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[2]); in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld128() local
58 v128_t vacc1x2 = vacc0x2; in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld128()
104 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_low_i16x8(vprod0x2)); in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld128()
110 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_high_i16x8(vprod0x2)); in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld128()
124 …02 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x0, vacc0x2, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x0, … in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld128()
|
| D | 2x4c8-minmax-wasmsimd-ld64.c | 54 v128_t vacc0x2 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[2]); in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld64() local
58 v128_t vacc1x2 = vacc0x2; in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld64()
100 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_low_i16x8(vprod0x2)); in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld64()
101 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_high_i16x8(vprod0x2)); in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld64()
120 …02 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x0, vacc0x2, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x0, … in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld64()
|
| D | 3x4c8-minmax-wasmsimd-ld128.c | 58 v128_t vacc0x2 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[2]); in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld128() local
62 v128_t vacc1x2 = vacc0x2; in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
66 v128_t vacc2x2 = vacc0x2; in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
124 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_low_i16x8(vprod0x2)); in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
132 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_high_i16x8(vprod0x2)); in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
150 …02 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x0, vacc0x2, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x0, … in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
|
| D | 3x4c8-minmax-wasmsimd-ld64.c | 58 v128_t vacc0x2 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[2]); in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld64() local
62 v128_t vacc1x2 = vacc0x2; in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
66 v128_t vacc2x2 = vacc0x2; in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
120 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_low_i16x8(vprod0x2)); in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
121 vacc0x2 = wasm_i32x4_add(vacc0x2, wasm_i32x4_widen_high_i16x8(vprod0x2)); in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
146 …02 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x0, vacc0x2, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x0, … in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
|