| /external/XNNPACK/src/f32-ppmm/gen/ |
| D | 2x4-minmax-scalar.c | 43 float vacc0x3 = w[3]; in xnn_f32_ppmm_minmax_ukernel_2x4__scalar() local
47 float vacc1x3 = vacc0x3; in xnn_f32_ppmm_minmax_ukernel_2x4__scalar()
68 vacc0x3 += va0 * vb3; in xnn_f32_ppmm_minmax_ukernel_2x4__scalar()
81 vacc0x3 = math_min_f32(vacc0x3, vmax); in xnn_f32_ppmm_minmax_ukernel_2x4__scalar()
91 vacc0x3 = math_max_f32(vacc0x3, vmin); in xnn_f32_ppmm_minmax_ukernel_2x4__scalar()
102 c0[3] = vacc0x3; in xnn_f32_ppmm_minmax_ukernel_2x4__scalar()
|
| D | 4x4-minmax-scalar.c | 51 float vacc0x3 = w[3]; in xnn_f32_ppmm_minmax_ukernel_4x4__scalar() local
55 float vacc1x3 = vacc0x3; in xnn_f32_ppmm_minmax_ukernel_4x4__scalar()
59 float vacc2x3 = vacc0x3; in xnn_f32_ppmm_minmax_ukernel_4x4__scalar()
63 float vacc3x3 = vacc0x3; in xnn_f32_ppmm_minmax_ukernel_4x4__scalar()
92 vacc0x3 += va0 * vb3; in xnn_f32_ppmm_minmax_ukernel_4x4__scalar()
113 vacc0x3 = math_min_f32(vacc0x3, vmax); in xnn_f32_ppmm_minmax_ukernel_4x4__scalar()
131 vacc0x3 = math_max_f32(vacc0x3, vmin); in xnn_f32_ppmm_minmax_ukernel_4x4__scalar()
152 c0[3] = vacc0x3; in xnn_f32_ppmm_minmax_ukernel_4x4__scalar()
|
| /external/XNNPACK/src/f32-vmulcaddc/gen/ |
| D | c4-minmax-wasm-2x.c | 56 float vacc0x3 = i0[3]; in xnn_f32_vmulcaddc_minmax_ukernel_c4__wasm_2x() local
72 vacc0x3 = vacc0x3 * vscale3 + vbias3; in xnn_f32_vmulcaddc_minmax_ukernel_c4__wasm_2x()
81 vacc0x3 = __builtin_wasm_max_f32(vacc0x3, vmin); in xnn_f32_vmulcaddc_minmax_ukernel_c4__wasm_2x()
90 vacc0x3 = __builtin_wasm_min_f32(vacc0x3, vmax); in xnn_f32_vmulcaddc_minmax_ukernel_c4__wasm_2x()
99 o0[3] = vacc0x3; in xnn_f32_vmulcaddc_minmax_ukernel_c4__wasm_2x()
|
| D | c4-minmax-scalar-2x.c | 56 float vacc0x3 = i0[3]; in xnn_f32_vmulcaddc_minmax_ukernel_c4__scalar_2x() local
72 vacc0x3 = vacc0x3 * vscale3 + vbias3; in xnn_f32_vmulcaddc_minmax_ukernel_c4__scalar_2x()
81 vacc0x3 = math_max_f32(vacc0x3, vmin); in xnn_f32_vmulcaddc_minmax_ukernel_c4__scalar_2x()
90 vacc0x3 = math_min_f32(vacc0x3, vmax); in xnn_f32_vmulcaddc_minmax_ukernel_c4__scalar_2x()
99 o0[3] = vacc0x3; in xnn_f32_vmulcaddc_minmax_ukernel_c4__scalar_2x()
|
| /external/XNNPACK/src/f32-spmm/gen/ |
| D | 8x4-minmax-scalar.c | 65 float vacc0x3 = *w++; in xnn_f32_spmm_minmax_ukernel_8x4__scalar() local
66 float vacc1x3 = vacc0x3; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
67 float vacc2x3 = vacc0x3; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
68 float vacc3x3 = vacc0x3; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
69 float vacc4x3 = vacc0x3; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
70 float vacc5x3 = vacc0x3; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
71 float vacc6x3 = vacc0x3; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
72 float vacc7x3 = vacc0x3; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
113 vacc0x3 += vi0 * vw3; in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
147 float vout0x3 = math_min_f32(vacc0x3, vmax); in xnn_f32_spmm_minmax_ukernel_8x4__scalar()
[all …]
|
| /external/XNNPACK/src/qs8-gemm/gen/ |
| D | 1x4c8-minmax-wasmsimd-ld64.c | 48 v128_t vacc0x3 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[3]); in xnn_qs8_gemm_minmax_ukernel_1x4c8__wasmsimd_ld64() local
74 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_low_i16x8(vprod0x3)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__wasmsimd_ld64()
75 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_high_i16x8(vprod0x3)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__wasmsimd_ld64()
82 …13 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x1, vacc0x3, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x1, … in xnn_qs8_gemm_minmax_ukernel_1x4c8__wasmsimd_ld64()
|
| D | 1x4c8-minmax-wasmsimd-ld128.c | 48 v128_t vacc0x3 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[3]); in xnn_qs8_gemm_minmax_ukernel_1x4c8__wasmsimd_ld128() local
76 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_low_i16x8(vprod0x3)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__wasmsimd_ld128()
79 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_high_i16x8(vprod0x3)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__wasmsimd_ld128()
86 …13 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x1, vacc0x3, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x1, … in xnn_qs8_gemm_minmax_ukernel_1x4c8__wasmsimd_ld128()
|
| D | 1x4c8-xw-minmax-wasmsimd.c | 48 v128_t vacc0x3 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[3]); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__wasmsimd() local
74 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_low_i16x8(vprod0x3)); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__wasmsimd()
75 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_high_i16x8(vprod0x3)); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__wasmsimd()
82 …13 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x1, vacc0x3, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x1, … in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__wasmsimd()
|
| D | 2x4c8-xw-minmax-wasmsimd.c | 54 v128_t vacc0x3 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[3]); in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__wasmsimd() local
58 v128_t vacc1x3 = vacc0x3; in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__wasmsimd()
95 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_low_i16x8(vprod0x3)); in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__wasmsimd()
96 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_high_i16x8(vprod0x3)); in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__wasmsimd()
106 …13 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x1, vacc0x3, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x1, … in xnn_qs8_gemm_xw_minmax_ukernel_2x4c8__wasmsimd()
|
| D | 2x4c8-minmax-wasmsimd-ld128.c | 54 v128_t vacc0x3 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[3]); in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld128() local
58 v128_t vacc1x3 = vacc0x3; in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld128()
96 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_low_i16x8(vprod0x3)); in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld128()
102 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_high_i16x8(vprod0x3)); in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld128()
110 …13 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x1, vacc0x3, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x1, … in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld128()
|
| D | 2x4c8-minmax-wasmsimd-ld64.c | 54 v128_t vacc0x3 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[3]); in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld64() local
58 v128_t vacc1x3 = vacc0x3; in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld64()
95 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_low_i16x8(vprod0x3)); in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld64()
96 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_high_i16x8(vprod0x3)); in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld64()
106 …13 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x1, vacc0x3, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x1, … in xnn_qs8_gemm_minmax_ukernel_2x4c8__wasmsimd_ld64()
|
| D | 3x4c8-xw-minmax-wasmsimd.c | 60 v128_t vacc0x3 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[3]); in xnn_qs8_gemm_xw_minmax_ukernel_3x4c8__wasmsimd() local
64 v128_t vacc1x3 = vacc0x3; in xnn_qs8_gemm_xw_minmax_ukernel_3x4c8__wasmsimd()
68 v128_t vacc2x3 = vacc0x3; in xnn_qs8_gemm_xw_minmax_ukernel_3x4c8__wasmsimd()
116 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_low_i16x8(vprod0x3)); in xnn_qs8_gemm_xw_minmax_ukernel_3x4c8__wasmsimd()
117 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_high_i16x8(vprod0x3)); in xnn_qs8_gemm_xw_minmax_ukernel_3x4c8__wasmsimd()
130 …13 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x1, vacc0x3, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x1, … in xnn_qs8_gemm_xw_minmax_ukernel_3x4c8__wasmsimd()
|
| D | 3x4c8-minmax-wasmsimd-ld128.c | 60 v128_t vacc0x3 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[3]); in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld128() local
64 v128_t vacc1x3 = vacc0x3; in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
68 v128_t vacc2x3 = vacc0x3; in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
116 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_low_i16x8(vprod0x3)); in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
125 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_high_i16x8(vprod0x3)); in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
134 …13 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x1, vacc0x3, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x1, … in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
|
| D | 3x4c8-minmax-wasmsimd-ld64.c | 60 v128_t vacc0x3 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[3]); in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld64() local
64 v128_t vacc1x3 = vacc0x3; in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
68 v128_t vacc2x3 = vacc0x3; in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
116 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_low_i16x8(vprod0x3)); in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
117 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_high_i16x8(vprod0x3)); in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
130 …13 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x1, vacc0x3, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x1, … in xnn_qs8_gemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
|
| D | 1x8c8-minmax-neon-mlal-padal.c | 47 …int32x4_t vacc0x3 = 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
83 vacc0x3 = vpadalq_s16(vacc0x3, vprod0x3); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mlal_padal()
119 vacc0x3 = vpadalq_s16(vacc0x3, vprod0x3); 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()
147 const int32x2_t vpsum0x3 = vadd_s32(vget_low_s32(vacc0x3), vget_high_s32(vacc0x3)); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mlal_padal()
|
| D | 1x4c8-xw-minmax-sse2.c | 47 __m128i vacc0x3 = _mm_cvtsi32_si128((int) ((const int32_t*) w)[3]); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__sse2() local
67 vacc0x3 = _mm_add_epi32(vacc0x3, _mm_madd_epi16(vxa0, vxb3)); in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__sse2()
74 … vacc0x13 = _mm_add_epi32(_mm_unpacklo_epi32(vacc0x1, vacc0x3), _mm_unpackhi_epi32(vacc0x1, vacc0x… in xnn_qs8_gemm_xw_minmax_ukernel_1x4c8__sse2()
|
| D | 1x4c8-minmax-sse2-ld128.c | 47 __m128i vacc0x3 = _mm_cvtsi32_si128((int) ((const int32_t*) w)[3]); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld128() local
69 vacc0x3 = _mm_add_epi32(vacc0x3, _mm_madd_epi16(vxa0, vxb3)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld128()
76 … vacc0x13 = _mm_add_epi32(_mm_unpacklo_epi32(vacc0x1, vacc0x3), _mm_unpackhi_epi32(vacc0x1, vacc0x… in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld128()
|
| D | 1x4c8-minmax-sse2-ld64.c | 47 __m128i vacc0x3 = _mm_cvtsi32_si128((int) ((const int32_t*) w)[3]); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld64() local
71 vacc0x3 = _mm_add_epi32(vacc0x3, _mm_madd_epi16(vxa0, vxb3)); in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld64()
78 … vacc0x13 = _mm_add_epi32(_mm_unpacklo_epi32(vacc0x1, vacc0x3), _mm_unpackhi_epi32(vacc0x1, vacc0x… in xnn_qs8_gemm_minmax_ukernel_1x4c8__sse2_ld64()
|
| D | 1x8c8-minmax-neon-mull-padal.c | 47 …int32x4_t vacc0x3 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mull_padal() local
70 vacc0x3 = vpadalq_s16(vacc0x3, vprod0x3); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mull_padal()
89 const int32x4_t vsum0x23 = vpaddq_s32(vacc0x2, vacc0x3); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mull_padal()
98 const int32x2_t vpsum0x3 = vadd_s32(vget_low_s32(vacc0x3), vget_high_s32(vacc0x3)); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mull_padal()
|
| /external/XNNPACK/src/qs8-igemm/gen/ |
| D | 1x4c8-minmax-wasmsimd-ld64.c | 51 v128_t vacc0x3 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[3]); in xnn_qs8_igemm_minmax_ukernel_1x4c8__wasmsimd_ld64() local
85 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_low_i16x8(vprod0x3)); in xnn_qs8_igemm_minmax_ukernel_1x4c8__wasmsimd_ld64()
86 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_high_i16x8(vprod0x3)); in xnn_qs8_igemm_minmax_ukernel_1x4c8__wasmsimd_ld64()
95 …13 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x1, vacc0x3, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x1, … in xnn_qs8_igemm_minmax_ukernel_1x4c8__wasmsimd_ld64()
|
| D | 1x4c8-minmax-wasmsimd-ld128.c | 51 v128_t vacc0x3 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[3]); in xnn_qs8_igemm_minmax_ukernel_1x4c8__wasmsimd_ld128() local
87 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_low_i16x8(vprod0x3)); in xnn_qs8_igemm_minmax_ukernel_1x4c8__wasmsimd_ld128()
90 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_high_i16x8(vprod0x3)); in xnn_qs8_igemm_minmax_ukernel_1x4c8__wasmsimd_ld128()
99 …13 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x1, vacc0x3, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x1, … in xnn_qs8_igemm_minmax_ukernel_1x4c8__wasmsimd_ld128()
|
| D | 2x4c8-minmax-wasmsimd-ld128.c | 55 v128_t vacc0x3 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[3]); in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld128() local
59 v128_t vacc1x3 = vacc0x3; in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld128()
109 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_low_i16x8(vprod0x3)); in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld128()
115 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_high_i16x8(vprod0x3)); in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld128()
125 …13 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x1, vacc0x3, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x1, … in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld128()
|
| D | 2x4c8-minmax-wasmsimd-ld64.c | 55 v128_t vacc0x3 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[3]); in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld64() local
59 v128_t vacc1x3 = vacc0x3; in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld64()
108 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_low_i16x8(vprod0x3)); in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld64()
109 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_high_i16x8(vprod0x3)); in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld64()
121 …13 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x1, vacc0x3, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x1, … in xnn_qs8_igemm_minmax_ukernel_2x4c8__wasmsimd_ld64()
|
| D | 3x4c8-minmax-wasmsimd-ld128.c | 59 v128_t vacc0x3 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[3]); in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld128() local
63 v128_t vacc1x3 = vacc0x3; in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
67 v128_t vacc2x3 = vacc0x3; in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
131 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_low_i16x8(vprod0x3)); in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
140 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_high_i16x8(vprod0x3)); in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
151 …13 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x1, vacc0x3, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x1, … in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld128()
|
| D | 3x4c8-minmax-wasmsimd-ld64.c | 59 v128_t vacc0x3 = wasm_f32x4_replace_lane(vzero, 0, ((const float*) w)[3]); in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld64() local
63 v128_t vacc1x3 = vacc0x3; in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
67 v128_t vacc2x3 = vacc0x3; in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
131 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_low_i16x8(vprod0x3)); in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
132 vacc0x3 = wasm_i32x4_add(vacc0x3, wasm_i32x4_widen_high_i16x8(vprod0x3)); in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
147 …13 = wasm_i32x4_add(wasm_v32x4_shuffle(vacc0x1, vacc0x3, 0, 4, 1, 5), wasm_v32x4_shuffle(vacc0x1, … in xnn_qs8_igemm_minmax_ukernel_3x4c8__wasmsimd_ld64()
|