/external/XNNPACK/src/f32-conv-hwc2chw/ |
D | 3x3s2p1c3x4-neonfma-2x2.c | 242 const float32x4_t vk11c2 = vld1q_f32(w + 68); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neonfma_2x2() local 244 vo0x0 = vfmaq_laneq_f32(vo0x0, vk11c2, vi1x1, 2); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neonfma_2x2() 245 vo1x0 = vfmaq_laneq_f32(vo1x0, vk11c2, vi3x1, 2); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neonfma_2x2() 246 vo0x1 = vfmaq_laneq_f32(vo0x1, vk11c2, vi1x3, 0); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neonfma_2x2() 247 vo1x1 = vfmaq_laneq_f32(vo1x1, vk11c2, vi3x3, 0); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neonfma_2x2() 523 const float32x4_t vk11c2 = vld1q_f32(w + 68); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neonfma_2x2() local 525 vo0x0 = vfmaq_laneq_f32(vo0x0, vk11c2, vi1x1, 2); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neonfma_2x2() 526 vo1x0 = vfmaq_laneq_f32(vo1x0, vk11c2, vi3x1, 2); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neonfma_2x2() 527 vo0x1 = vfmaq_laneq_f32(vo0x1, vk11c2, vi1x3, 0); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neonfma_2x2() 528 vo1x1 = vfmaq_laneq_f32(vo1x1, vk11c2, vi3x3, 0); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neonfma_2x2()
|
D | 3x3s2p1c3x4-neon-2x2.c | 242 const float32x4_t vk11c2 = vld1q_f32(w + 68); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neon_2x2() local 244 vo0x0 = vmlaq_lane_f32(vo0x0, vk11c2, vget_high_f32(vi1x1), 0); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neon_2x2() 245 vo1x0 = vmlaq_lane_f32(vo1x0, vk11c2, vget_high_f32(vi3x1), 0); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neon_2x2() 246 vo0x1 = vmlaq_lane_f32(vo0x1, vk11c2, vget_low_f32(vi1x3), 0); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neon_2x2() 247 vo1x1 = vmlaq_lane_f32(vo1x1, vk11c2, vget_low_f32(vi3x3), 0); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neon_2x2() 521 const float32x4_t vk11c2 = vld1q_f32(w + 68); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neon_2x2() local 523 vo0x0 = vmlaq_lane_f32(vo0x0, vk11c2, vget_high_f32(vi1x1), 0); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neon_2x2() 524 vo1x0 = vmlaq_lane_f32(vo1x0, vk11c2, vget_high_f32(vi3x1), 0); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neon_2x2() 525 vo0x1 = vmlaq_lane_f32(vo0x1, vk11c2, vget_low_f32(vi1x3), 0); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neon_2x2() 526 vo1x1 = vmlaq_lane_f32(vo1x1, vk11c2, vget_low_f32(vi3x3), 0); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__neon_2x2()
|
D | 3x3s2p1c3x4-wasmsimd-2x2.c | 243 const v128_t vk11c2 = wasm_v128_load(w + 68); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__wasmsimd_2x2() local 245 …vo0x0 = wasm_f32x4_add(vo0x0, wasm_f32x4_mul(vk11c2, wasm_v32x4_shuffle(vi1x1, vi1x1, 2, 2, 2, 2))… in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__wasmsimd_2x2() 246 …vo1x0 = wasm_f32x4_add(vo1x0, wasm_f32x4_mul(vk11c2, wasm_v32x4_shuffle(vi3x1, vi3x1, 2, 2, 2, 2))… in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__wasmsimd_2x2() 247 …vo0x1 = wasm_f32x4_add(vo0x1, wasm_f32x4_mul(vk11c2, wasm_v32x4_shuffle(vi1x3, vi1x3, 0, 0, 0, 0))… in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__wasmsimd_2x2() 248 …vo1x1 = wasm_f32x4_add(vo1x1, wasm_f32x4_mul(vk11c2, wasm_v32x4_shuffle(vi3x3, vi3x3, 0, 0, 0, 0))… in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__wasmsimd_2x2() 523 const v128_t vk11c2 = wasm_v128_load(w + 68); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__wasmsimd_2x2() local 525 …vo0x0 = wasm_f32x4_add(vo0x0, wasm_f32x4_mul(vk11c2, wasm_v32x4_shuffle(vi1x1, vi1x1, 2, 2, 2, 2))… in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__wasmsimd_2x2() 526 …vo1x0 = wasm_f32x4_add(vo1x0, wasm_f32x4_mul(vk11c2, wasm_v32x4_shuffle(vi3x1, vi3x1, 2, 2, 2, 2))… in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__wasmsimd_2x2() 527 …vo0x1 = wasm_f32x4_add(vo0x1, wasm_f32x4_mul(vk11c2, wasm_v32x4_shuffle(vi1x3, vi1x3, 0, 0, 0, 0))… in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__wasmsimd_2x2() 528 …vo1x1 = wasm_f32x4_add(vo1x1, wasm_f32x4_mul(vk11c2, wasm_v32x4_shuffle(vi3x3, vi3x3, 0, 0, 0, 0))… in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__wasmsimd_2x2()
|
D | 3x3s2p1c3x4-sse-2x2.c | 242 const __m128 vk11c2 = _mm_load_ps(w + 68); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2() local 244 …vo0x0 = _mm_add_ps(vo0x0, _mm_mul_ps(vk11c2, _mm_shuffle_ps(vi1x1, vi1x1, _MM_SHUFFLE(2, 2, 2, 2))… in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2() 245 …vo1x0 = _mm_add_ps(vo1x0, _mm_mul_ps(vk11c2, _mm_shuffle_ps(vi3x1, vi3x1, _MM_SHUFFLE(2, 2, 2, 2))… in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2() 246 …vo0x1 = _mm_add_ps(vo0x1, _mm_mul_ps(vk11c2, _mm_shuffle_ps(vi1x3, vi1x3, _MM_SHUFFLE(0, 0, 0, 0))… in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2() 247 …vo1x1 = _mm_add_ps(vo1x1, _mm_mul_ps(vk11c2, _mm_shuffle_ps(vi3x3, vi3x3, _MM_SHUFFLE(0, 0, 0, 0))… in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2() 523 const __m128 vk11c2 = _mm_load_ps(w + 68); in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2() local 525 …vo0x0 = _mm_add_ps(vo0x0, _mm_mul_ps(vk11c2, _mm_shuffle_ps(vi1x1, vi1x1, _MM_SHUFFLE(2, 2, 2, 2))… in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2() 526 …vo1x0 = _mm_add_ps(vo1x0, _mm_mul_ps(vk11c2, _mm_shuffle_ps(vi3x1, vi3x1, _MM_SHUFFLE(2, 2, 2, 2))… in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2() 527 …vo0x1 = _mm_add_ps(vo0x1, _mm_mul_ps(vk11c2, _mm_shuffle_ps(vi1x3, vi1x3, _MM_SHUFFLE(0, 0, 0, 0))… in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2() 528 …vo1x1 = _mm_add_ps(vo1x1, _mm_mul_ps(vk11c2, _mm_shuffle_ps(vi3x3, vi3x3, _MM_SHUFFLE(0, 0, 0, 0))… in xnn_f32_conv_hwc2chw_ukernel_3x3s2p1c3x4__sse_2x2()
|