| /external/XNNPACK/src/f32-dwconv/gen/ |
| D | up8x4-wasmsimd.c | 75 const v128_t vk1x0123 = wasm_v128_load(w + 16); in xnn_f32_dwconv_ukernel_up8x4__wasmsimd() local
77 vacc0123p0 = wasm_f32x4_add(vacc0123p0, wasm_f32x4_mul(vi1x0123, vk1x0123)); in xnn_f32_dwconv_ukernel_up8x4__wasmsimd()
120 const v128_t vk1x0123 = wasm_v128_load(w + 16); in xnn_f32_dwconv_ukernel_up8x4__wasmsimd() local
121 vacc0123p0 = wasm_f32x4_add(vacc0123p0, wasm_f32x4_mul(vi1x0123, vk1x0123)); in xnn_f32_dwconv_ukernel_up8x4__wasmsimd()
151 const v128_t vk1x0123 = wasm_v128_load(w + 16); in xnn_f32_dwconv_ukernel_up8x4__wasmsimd() local
152 vacc0123p0 = wasm_f32x4_add(vacc0123p0, wasm_f32x4_mul(vi1x0123, vk1x0123)); in xnn_f32_dwconv_ukernel_up8x4__wasmsimd()
|
| D | up8x4-minmax-sse-acc2.c | 77 const __m128 vk1x0123 = _mm_load_ps(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__sse_acc2() local
79 __m128 vacc0123p1 = _mm_mul_ps(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__sse_acc2()
127 const __m128 vk1x0123 = _mm_load_ps(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__sse_acc2() local
128 __m128 vacc0123p1 = _mm_mul_ps(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__sse_acc2()
161 const __m128 vk1x0123 = _mm_load_ps(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__sse_acc2() local
162 __m128 vacc0123p1 = _mm_mul_ps(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__sse_acc2()
|
| D | up8x4-minmax-neon.c | 74 const float32x4_t vk1x0123 = vld1q_f32(w); w += 4; in xnn_f32_dwconv_minmax_ukernel_up8x4__neon() local
76 vacc0123p0 = vmlaq_f32(vacc0123p0, vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__neon()
111 const float32x4_t vk1x0123 = vld1q_f32(w + 12); in xnn_f32_dwconv_minmax_ukernel_up8x4__neon() local
112 vacc0123p0 = vmlaq_f32(vacc0123p0, vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__neon()
137 const float32x4_t vk1x0123 = vld1q_f32(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__neon() local
138 vacc0123p0 = vmlaq_f32(vacc0123p0, vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__neon()
|
| D | up8x4-minmax-wasmsimd-arm-acc2.c | 77 const v128_t vk1x0123 = wasm_v128_load(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_arm_acc2() local
79 v128_t vacc0123p1 = wasm_f32x4_mul(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_arm_acc2()
128 const v128_t vk1x0123 = wasm_v128_load(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_arm_acc2() local
129 v128_t vacc0123p1 = wasm_f32x4_mul(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_arm_acc2()
162 const v128_t vk1x0123 = wasm_v128_load(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_arm_acc2() local
163 v128_t vacc0123p1 = wasm_f32x4_mul(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_arm_acc2()
|
| D | up8x4-minmax-neonfma.c | 74 const float32x4_t vk1x0123 = vld1q_f32(w); w += 4; in xnn_f32_dwconv_minmax_ukernel_up8x4__neonfma() local
76 vacc0123p0 = vfmaq_f32(vacc0123p0, vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__neonfma()
111 const float32x4_t vk1x0123 = vld1q_f32(w + 12); in xnn_f32_dwconv_minmax_ukernel_up8x4__neonfma() local
112 vacc0123p0 = vfmaq_f32(vacc0123p0, vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__neonfma()
137 const float32x4_t vk1x0123 = vld1q_f32(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__neonfma() local
138 vacc0123p0 = vfmaq_f32(vacc0123p0, vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__neonfma()
|
| D | up8x4-minmax-neon-acc2.c | 74 const float32x4_t vk1x0123 = vld1q_f32(w); w += 4; in xnn_f32_dwconv_minmax_ukernel_up8x4__neon_acc2() local
76 float32x4_t vacc0123p1 = vmulq_f32(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__neon_acc2()
114 const float32x4_t vk1x0123 = vld1q_f32(w + 12); in xnn_f32_dwconv_minmax_ukernel_up8x4__neon_acc2() local
115 float32x4_t vacc0123p1 = vmulq_f32(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__neon_acc2()
142 const float32x4_t vk1x0123 = vld1q_f32(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__neon_acc2() local
143 float32x4_t vacc0123p1 = vmulq_f32(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__neon_acc2()
|
| D | up8x4-minmax-neonfma-acc2.c | 74 const float32x4_t vk1x0123 = vld1q_f32(w); w += 4; in xnn_f32_dwconv_minmax_ukernel_up8x4__neonfma_acc2() local
76 float32x4_t vacc0123p1 = vmulq_f32(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__neonfma_acc2()
114 const float32x4_t vk1x0123 = vld1q_f32(w + 12); in xnn_f32_dwconv_minmax_ukernel_up8x4__neonfma_acc2() local
115 float32x4_t vacc0123p1 = vmulq_f32(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__neonfma_acc2()
142 const float32x4_t vk1x0123 = vld1q_f32(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__neonfma_acc2() local
143 float32x4_t vacc0123p1 = vmulq_f32(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__neonfma_acc2()
|
| D | up8x4-minmax-wasmsimd-arm.c | 77 const v128_t vk1x0123 = wasm_v128_load(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_arm() local
79 vacc0123p0 = wasm_f32x4_add(vacc0123p0, wasm_f32x4_mul(vi1x0123, vk1x0123)); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_arm()
125 const v128_t vk1x0123 = wasm_v128_load(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_arm() local
126 vacc0123p0 = wasm_f32x4_add(vacc0123p0, wasm_f32x4_mul(vi1x0123, vk1x0123)); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_arm()
157 const v128_t vk1x0123 = wasm_v128_load(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_arm() local
158 vacc0123p0 = wasm_f32x4_add(vacc0123p0, wasm_f32x4_mul(vi1x0123, vk1x0123)); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_arm()
|
| D | up8x4-minmax-wasmsimd-x86.c | 77 const v128_t vk1x0123 = wasm_v128_load(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_x86() local
79 vacc0123p0 = wasm_f32x4_add(vacc0123p0, wasm_f32x4_mul(vi1x0123, vk1x0123)); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_x86()
125 const v128_t vk1x0123 = wasm_v128_load(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_x86() local
126 vacc0123p0 = wasm_f32x4_add(vacc0123p0, wasm_f32x4_mul(vi1x0123, vk1x0123)); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_x86()
157 const v128_t vk1x0123 = wasm_v128_load(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_x86() local
158 vacc0123p0 = wasm_f32x4_add(vacc0123p0, wasm_f32x4_mul(vi1x0123, vk1x0123)); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_x86()
|
| D | up8x4-minmax-sse.c | 77 const __m128 vk1x0123 = _mm_load_ps(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__sse() local
79 vacc0123p0 = _mm_add_ps(vacc0123p0, _mm_mul_ps(vi1x0123, vk1x0123)); in xnn_f32_dwconv_minmax_ukernel_up8x4__sse()
124 const __m128 vk1x0123 = _mm_load_ps(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__sse() local
125 vacc0123p0 = _mm_add_ps(vacc0123p0, _mm_mul_ps(vi1x0123, vk1x0123)); in xnn_f32_dwconv_minmax_ukernel_up8x4__sse()
156 const __m128 vk1x0123 = _mm_load_ps(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__sse() local
157 vacc0123p0 = _mm_add_ps(vacc0123p0, _mm_mul_ps(vi1x0123, vk1x0123)); in xnn_f32_dwconv_minmax_ukernel_up8x4__sse()
|
| D | up8x4-minmax-wasmsimd-x86-acc2.c | 77 const v128_t vk1x0123 = wasm_v128_load(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_x86_acc2() local
79 v128_t vacc0123p1 = wasm_f32x4_mul(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_x86_acc2()
128 const v128_t vk1x0123 = wasm_v128_load(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_x86_acc2() local
129 v128_t vacc0123p1 = wasm_f32x4_mul(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_x86_acc2()
162 const v128_t vk1x0123 = wasm_v128_load(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_x86_acc2() local
163 v128_t vacc0123p1 = wasm_f32x4_mul(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x4__wasmsimd_x86_acc2()
|
| D | up4x4-wasmsimd.c | 70 const v128_t vk1x0123 = wasm_v128_load(w + 8); in xnn_f32_dwconv_ukernel_up4x4__wasmsimd() local
71 vacc0123p0 = wasm_f32x4_add(vacc0123p0, wasm_f32x4_mul(vi1x0123, vk1x0123)); in xnn_f32_dwconv_ukernel_up4x4__wasmsimd()
101 const v128_t vk1x0123 = wasm_v128_load(w + 8); in xnn_f32_dwconv_ukernel_up4x4__wasmsimd() local
102 vacc0123p0 = wasm_f32x4_add(vacc0123p0, wasm_f32x4_mul(vi1x0123, vk1x0123)); in xnn_f32_dwconv_ukernel_up4x4__wasmsimd()
|
| D | up4x4-minmax-sse.c | 72 const __m128 vk1x0123 = _mm_load_ps(w + 8); in xnn_f32_dwconv_minmax_ukernel_up4x4__sse() local
73 vacc0123p0 = _mm_add_ps(vacc0123p0, _mm_mul_ps(vi1x0123, vk1x0123)); in xnn_f32_dwconv_minmax_ukernel_up4x4__sse()
104 const __m128 vk1x0123 = _mm_load_ps(w + 8); in xnn_f32_dwconv_minmax_ukernel_up4x4__sse() local
105 vacc0123p0 = _mm_add_ps(vacc0123p0, _mm_mul_ps(vi1x0123, vk1x0123)); in xnn_f32_dwconv_minmax_ukernel_up4x4__sse()
|
| D | up4x4-minmax-wasmsimd-arm.c | 72 const v128_t vk1x0123 = wasm_v128_load(w + 8); in xnn_f32_dwconv_minmax_ukernel_up4x4__wasmsimd_arm() local
73 vacc0123p0 = wasm_f32x4_add(vacc0123p0, wasm_f32x4_mul(vi1x0123, vk1x0123)); in xnn_f32_dwconv_minmax_ukernel_up4x4__wasmsimd_arm()
105 const v128_t vk1x0123 = wasm_v128_load(w + 8); in xnn_f32_dwconv_minmax_ukernel_up4x4__wasmsimd_arm() local
106 vacc0123p0 = wasm_f32x4_add(vacc0123p0, wasm_f32x4_mul(vi1x0123, vk1x0123)); in xnn_f32_dwconv_minmax_ukernel_up4x4__wasmsimd_arm()
|
| D | up4x4-minmax-wasmsimd-arm-acc2.c | 72 const v128_t vk1x0123 = wasm_v128_load(w + 8); in xnn_f32_dwconv_minmax_ukernel_up4x4__wasmsimd_arm_acc2() local
73 v128_t vacc0123p1 = wasm_f32x4_mul(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up4x4__wasmsimd_arm_acc2()
107 const v128_t vk1x0123 = wasm_v128_load(w + 8); in xnn_f32_dwconv_minmax_ukernel_up4x4__wasmsimd_arm_acc2() local
108 v128_t vacc0123p1 = wasm_f32x4_mul(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up4x4__wasmsimd_arm_acc2()
|
| D | up4x4-minmax-sse-acc2.c | 72 const __m128 vk1x0123 = _mm_load_ps(w + 8); in xnn_f32_dwconv_minmax_ukernel_up4x4__sse_acc2() local
73 __m128 vacc0123p1 = _mm_mul_ps(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up4x4__sse_acc2()
106 const __m128 vk1x0123 = _mm_load_ps(w + 8); in xnn_f32_dwconv_minmax_ukernel_up4x4__sse_acc2() local
107 __m128 vacc0123p1 = _mm_mul_ps(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up4x4__sse_acc2()
|
| D | up4x4-minmax-neon.c | 69 const float32x4_t vk1x0123 = vld1q_f32(w); w += 4; in xnn_f32_dwconv_minmax_ukernel_up4x4__neon() local
70 vacc0123p0 = vmlaq_f32(vacc0123p0, vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up4x4__neon()
95 const float32x4_t vk1x0123 = vld1q_f32(w); w += 4; in xnn_f32_dwconv_minmax_ukernel_up4x4__neon() local
96 vacc0123p0 = vmlaq_f32(vacc0123p0, vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up4x4__neon()
|
| D | up4x4-minmax-neonfma-acc2.c | 69 const float32x4_t vk1x0123 = vld1q_f32(w); w += 4; in xnn_f32_dwconv_minmax_ukernel_up4x4__neonfma_acc2() local
70 float32x4_t vacc0123p1 = vmulq_f32(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up4x4__neonfma_acc2()
97 const float32x4_t vk1x0123 = vld1q_f32(w); w += 4; in xnn_f32_dwconv_minmax_ukernel_up4x4__neonfma_acc2() local
98 float32x4_t vacc0123p1 = vmulq_f32(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up4x4__neonfma_acc2()
|
| D | up4x4-minmax-wasmsimd-x86.c | 72 const v128_t vk1x0123 = wasm_v128_load(w + 8); in xnn_f32_dwconv_minmax_ukernel_up4x4__wasmsimd_x86() local
73 vacc0123p0 = wasm_f32x4_add(vacc0123p0, wasm_f32x4_mul(vi1x0123, vk1x0123)); in xnn_f32_dwconv_minmax_ukernel_up4x4__wasmsimd_x86()
105 const v128_t vk1x0123 = wasm_v128_load(w + 8); in xnn_f32_dwconv_minmax_ukernel_up4x4__wasmsimd_x86() local
106 vacc0123p0 = wasm_f32x4_add(vacc0123p0, wasm_f32x4_mul(vi1x0123, vk1x0123)); in xnn_f32_dwconv_minmax_ukernel_up4x4__wasmsimd_x86()
|
| D | up4x4-minmax-neonfma.c | 69 const float32x4_t vk1x0123 = vld1q_f32(w); w += 4; in xnn_f32_dwconv_minmax_ukernel_up4x4__neonfma() local
70 vacc0123p0 = vfmaq_f32(vacc0123p0, vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up4x4__neonfma()
95 const float32x4_t vk1x0123 = vld1q_f32(w); w += 4; in xnn_f32_dwconv_minmax_ukernel_up4x4__neonfma() local
96 vacc0123p0 = vfmaq_f32(vacc0123p0, vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up4x4__neonfma()
|
| D | up4x4-minmax-neon-acc2.c | 69 const float32x4_t vk1x0123 = vld1q_f32(w); w += 4; in xnn_f32_dwconv_minmax_ukernel_up4x4__neon_acc2() local
70 float32x4_t vacc0123p1 = vmulq_f32(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up4x4__neon_acc2()
97 const float32x4_t vk1x0123 = vld1q_f32(w); w += 4; in xnn_f32_dwconv_minmax_ukernel_up4x4__neon_acc2() local
98 float32x4_t vacc0123p1 = vmulq_f32(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up4x4__neon_acc2()
|
| D | up4x4-minmax-wasmsimd-x86-acc2.c | 72 const v128_t vk1x0123 = wasm_v128_load(w + 8); in xnn_f32_dwconv_minmax_ukernel_up4x4__wasmsimd_x86_acc2() local
73 v128_t vacc0123p1 = wasm_f32x4_mul(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up4x4__wasmsimd_x86_acc2()
107 const v128_t vk1x0123 = wasm_v128_load(w + 8); in xnn_f32_dwconv_minmax_ukernel_up4x4__wasmsimd_x86_acc2() local
108 v128_t vacc0123p1 = wasm_f32x4_mul(vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up4x4__wasmsimd_x86_acc2()
|
| D | up8x9-minmax-sse.c | 102 const __m128 vk1x0123 = _mm_load_ps(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x9__sse() local
104 vacc0123p0 = _mm_add_ps(vacc0123p0, _mm_mul_ps(vi1x0123, vk1x0123)); in xnn_f32_dwconv_minmax_ukernel_up8x9__sse()
194 const __m128 vk1x0123 = _mm_load_ps(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x9__sse() local
195 vacc0123p0 = _mm_add_ps(vacc0123p0, _mm_mul_ps(vi1x0123, vk1x0123)); in xnn_f32_dwconv_minmax_ukernel_up8x9__sse()
256 const __m128 vk1x0123 = _mm_load_ps(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x9__sse() local
257 vacc0123p0 = _mm_add_ps(vacc0123p0, _mm_mul_ps(vi1x0123, vk1x0123)); in xnn_f32_dwconv_minmax_ukernel_up8x9__sse()
|
| D | up8x9-wasmsimd.c | 100 const v128_t vk1x0123 = wasm_v128_load(w + 16); in xnn_f32_dwconv_ukernel_up8x9__wasmsimd() local
102 vacc0123p0 = wasm_f32x4_add(vacc0123p0, wasm_f32x4_mul(vi1x0123, vk1x0123)); in xnn_f32_dwconv_ukernel_up8x9__wasmsimd()
190 const v128_t vk1x0123 = wasm_v128_load(w + 16); in xnn_f32_dwconv_ukernel_up8x9__wasmsimd() local
191 vacc0123p0 = wasm_f32x4_add(vacc0123p0, wasm_f32x4_mul(vi1x0123, vk1x0123)); in xnn_f32_dwconv_ukernel_up8x9__wasmsimd()
251 const v128_t vk1x0123 = wasm_v128_load(w + 16); in xnn_f32_dwconv_ukernel_up8x9__wasmsimd() local
252 vacc0123p0 = wasm_f32x4_add(vacc0123p0, wasm_f32x4_mul(vi1x0123, vk1x0123)); in xnn_f32_dwconv_ukernel_up8x9__wasmsimd()
|
| D | up8x9-minmax-neon.c | 99 const float32x4_t vk1x0123 = vld1q_f32(w); w += 4; in xnn_f32_dwconv_minmax_ukernel_up8x9__neon() local
101 vacc0123p0 = vmlaq_f32(vacc0123p0, vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x9__neon()
171 const float32x4_t vk1x0123 = vld1q_f32(w + 12); in xnn_f32_dwconv_minmax_ukernel_up8x9__neon() local
172 vacc0123p0 = vmlaq_f32(vacc0123p0, vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x9__neon()
217 const float32x4_t vk1x0123 = vld1q_f32(w + 16); in xnn_f32_dwconv_minmax_ukernel_up8x9__neon() local
218 vacc0123p0 = vmlaq_f32(vacc0123p0, vi1x0123, vk1x0123); in xnn_f32_dwconv_minmax_ukernel_up8x9__neon()
|