/external/XNNPACK/src/f32-dwconv2d-chw/gen/ |
D | 3x3s2p1-minmax-wasmsimd-x86-splat-1x4-acc2.c | 76 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc2() local 84 const v128_t vi2x8ACE = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc2() 85 const v128_t vi2x9BDF = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc2() 128 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc2() local 135 …const v128_t vi2x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc2() 136 …const v128_t vi2x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc2()
|
D | 3x3s2p1-minmax-wasmsimd-arm-loadsplat-1x4-acc2.c | 86 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc2() local 94 const v128_t vi2x8ACE = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc2() 95 const v128_t vi2x9BDF = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc2() 138 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc2() local 145 …const v128_t vi2x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc2() 146 …const v128_t vi2x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc2()
|
D | 3x3s2p1-minmax-wasmsimd-arm-splat-1x4-acc2.c | 76 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc2() local 84 const v128_t vi2x8ACE = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc2() 85 const v128_t vi2x9BDF = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc2() 128 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc2() local 135 …const v128_t vi2x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc2() 136 …const v128_t vi2x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc2()
|
D | 3x3s2p1-minmax-wasmsimd-arm-splat-1x4-acc4.c | 76 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc4() local 84 const v128_t vi2x8ACE = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc4() 85 const v128_t vi2x9BDF = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc4() 130 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc4() local 137 …const v128_t vi2x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc4() 138 …const v128_t vi2x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc4()
|
D | 3x3s2p1-minmax-wasmsimd-x86-loadsplat-1x4.c | 86 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4() local 94 const v128_t vi2x8ACE = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4() 95 const v128_t vi2x9BDF = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4() 137 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4() local 144 …const v128_t vi2x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4() 145 …const v128_t vi2x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4()
|
D | 3x3s2p1-minmax-wasmsimd-x86-loadsplat-1x4-acc2.c | 86 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc2() local 94 const v128_t vi2x8ACE = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc2() 95 const v128_t vi2x9BDF = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc2() 138 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc2() local 145 …const v128_t vi2x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc2() 146 …const v128_t vi2x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc2()
|
D | 3x3s2p1-minmax-wasmsimd-arm-loadsplat-1x4.c | 86 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4() local 94 const v128_t vi2x8ACE = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4() 95 const v128_t vi2x9BDF = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4() 137 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4() local 144 …const v128_t vi2x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4() 145 …const v128_t vi2x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4()
|
D | 3x3s2p1-minmax-wasmsimd-arm-splat-1x4.c | 76 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4() local 84 const v128_t vi2x8ACE = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4() 85 const v128_t vi2x9BDF = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4() 127 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4() local 134 …const v128_t vi2x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4() 135 …const v128_t vi2x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4()
|
D | 3x3s2p1-minmax-wasmsimd-x86-splat-1x4-acc3.c | 76 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc3() local 84 const v128_t vi2x8ACE = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc3() 85 const v128_t vi2x9BDF = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc3() 129 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc3() local 136 …const v128_t vi2x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc3() 137 …const v128_t vi2x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc3()
|
D | 3x3s2p1-minmax-wasmsimd-x86-splat-1x4.c | 76 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4() local 84 const v128_t vi2x8ACE = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4() 85 const v128_t vi2x9BDF = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4() 127 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4() local 134 …const v128_t vi2x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4() 135 …const v128_t vi2x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4()
|
D | 3x3s2p1-minmax-wasmsimd-x86-splat-1x4-acc4.c | 76 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc4() local 84 const v128_t vi2x8ACE = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc4() 85 const v128_t vi2x9BDF = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc4() 130 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc4() local 137 …const v128_t vi2x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc4() 138 …const v128_t vi2x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc4()
|
D | 3x3s2p1-minmax-wasmsimd-arm-splat-1x4-acc3.c | 76 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc3() local 84 const v128_t vi2x8ACE = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc3() 85 const v128_t vi2x9BDF = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc3() 129 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc3() local 136 …const v128_t vi2x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc3() 137 …const v128_t vi2x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc3()
|
D | 3x3s2p1-minmax-wasmsimd-x86-loadsplat-1x4-acc3.c | 86 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc3() local 94 const v128_t vi2x8ACE = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc3() 95 const v128_t vi2x9BDF = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc3() 139 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc3() local 146 …const v128_t vi2x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc3() 147 …const v128_t vi2x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc3()
|
D | 3x3s2p1-minmax-sse-1x4.c | 80 const __m128 vi2x89AB = _mm_loadu_ps(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4() local 88 const __m128 vi2x8ACE = _mm_shuffle_ps(vi2x89AB, vi2xCDEF, _MM_SHUFFLE(2, 0, 2, 0)); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4() 89 const __m128 vi2x9BDF = _mm_shuffle_ps(vi2x89AB, vi2xCDEF, _MM_SHUFFLE(3, 1, 3, 1)); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4() 130 const __m128 vi2x89AB = _mm_loadu_ps(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4() local 137 …const __m128 vi2x8ACE = _mm_and_ps(vmask_even, _mm_shuffle_ps(vi2x89AB, vi2xCDEF, _MM_SHUFFLE(2, 0… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4() 138 …const __m128 vi2x9BDF = _mm_and_ps(vmask_odd, _mm_shuffle_ps(vi2x89AB, vi2xCDEF, _MM_SHUFFLE(3, 1… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4()
|
D | 3x3s2p1-minmax-sse-1x4-acc3.c | 80 const __m128 vi2x89AB = _mm_loadu_ps(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc3() local 88 const __m128 vi2x8ACE = _mm_shuffle_ps(vi2x89AB, vi2xCDEF, _MM_SHUFFLE(2, 0, 2, 0)); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc3() 89 const __m128 vi2x9BDF = _mm_shuffle_ps(vi2x89AB, vi2xCDEF, _MM_SHUFFLE(3, 1, 3, 1)); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc3() 132 const __m128 vi2x89AB = _mm_loadu_ps(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc3() local 139 …const __m128 vi2x8ACE = _mm_and_ps(vmask_even, _mm_shuffle_ps(vi2x89AB, vi2xCDEF, _MM_SHUFFLE(2, 0… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc3() 140 …const __m128 vi2x9BDF = _mm_and_ps(vmask_odd, _mm_shuffle_ps(vi2x89AB, vi2xCDEF, _MM_SHUFFLE(3, 1… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc3()
|
D | 3x3s2p1-minmax-wasmsimd-arm-loadsplat-1x4-acc4.c | 86 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc4() local 94 const v128_t vi2x8ACE = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc4() 95 const v128_t vi2x9BDF = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc4() 140 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc4() local 147 …const v128_t vi2x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc4() 148 …const v128_t vi2x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc4()
|
D | 3x3s2p1-minmax-wasmsimd-arm-loadsplat-1x4-acc3.c | 86 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc3() local 94 const v128_t vi2x8ACE = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc3() 95 const v128_t vi2x9BDF = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc3() 139 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc3() local 146 …const v128_t vi2x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc3() 147 …const v128_t vi2x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc3()
|
D | 3x3s2p1-minmax-wasmsimd-x86-loadsplat-1x4-acc4.c | 86 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc4() local 94 const v128_t vi2x8ACE = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc4() 95 const v128_t vi2x9BDF = wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc4() 140 const v128_t vi2x89AB = wasm_v128_load(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc4() local 147 …const v128_t vi2x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc4() 148 …const v128_t vi2x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi2x89AB, vi2xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc4()
|
D | 3x3s2p1-minmax-sse-1x4-acc2.c | 80 const __m128 vi2x89AB = _mm_loadu_ps(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc2() local 88 const __m128 vi2x8ACE = _mm_shuffle_ps(vi2x89AB, vi2xCDEF, _MM_SHUFFLE(2, 0, 2, 0)); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc2() 89 const __m128 vi2x9BDF = _mm_shuffle_ps(vi2x89AB, vi2xCDEF, _MM_SHUFFLE(3, 1, 3, 1)); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc2() 131 const __m128 vi2x89AB = _mm_loadu_ps(i2); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc2() local 138 …const __m128 vi2x8ACE = _mm_and_ps(vmask_even, _mm_shuffle_ps(vi2x89AB, vi2xCDEF, _MM_SHUFFLE(2, 0… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc2() 139 …const __m128 vi2x9BDF = _mm_and_ps(vmask_odd, _mm_shuffle_ps(vi2x89AB, vi2xCDEF, _MM_SHUFFLE(3, 1… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc2()
|
D | 5x5p2-minmax-neonfma-1x4.c | 82 const float32x4_t vi2x89AB = vld1q_f32(i2); i2 += 4; in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4() local 135 const float32x4_t vi2x5678 = vextq_f32(vi2x4567, vi2x89AB, 1); in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4() 153 const float32x4_t vi2x6789 = vextq_f32(vi2x4567, vi2x89AB, 2); in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4() 154 vi2x4567 = vi2x89AB; in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4() 183 float32x4_t vi2x89AB = vld1q_f32(i2); i2 += 4; in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4() local 189 vi2x89AB = vreinterpretq_f32_u32(vandq_u32(vmask, vreinterpretq_u32_f32(vi2x89AB))); in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4() 242 const float32x4_t vi2x5678 = vextq_f32(vi2x4567, vi2x89AB, 1); in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4() 260 const float32x4_t vi2x6789 = vextq_f32(vi2x4567, vi2x89AB, 2); in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4() 261 vi2x4567 = vi2x89AB; in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4()
|
D | 5x5p2-minmax-neonfma-1x4-acc2.c | 82 const float32x4_t vi2x89AB = vld1q_f32(i2); i2 += 4; in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4_acc2() local 135 const float32x4_t vi2x5678 = vextq_f32(vi2x4567, vi2x89AB, 1); in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4_acc2() 153 const float32x4_t vi2x6789 = vextq_f32(vi2x4567, vi2x89AB, 2); in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4_acc2() 154 vi2x4567 = vi2x89AB; in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4_acc2() 184 float32x4_t vi2x89AB = vld1q_f32(i2); i2 += 4; in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4_acc2() local 190 vi2x89AB = vreinterpretq_f32_u32(vandq_u32(vmask, vreinterpretq_u32_f32(vi2x89AB))); in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4_acc2() 243 const float32x4_t vi2x5678 = vextq_f32(vi2x4567, vi2x89AB, 1); in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4_acc2() 261 const float32x4_t vi2x6789 = vextq_f32(vi2x4567, vi2x89AB, 2); in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4_acc2() 262 vi2x4567 = vi2x89AB; in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4_acc2()
|
/external/XNNPACK/src/f16-dwconv2d-chw/gen/ |
D | 5x5p2-minmax-neonfp16arith-1x4-acc4.c | 80 const float16x4_t vi2x89AB = vld1_f16(i2); i2 += 4; in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc4() local 133 const float16x4_t vi2x5678 = vext_f16(vi2x4567, vi2x89AB, 1); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc4() 151 const float16x4_t vi2x6789 = vext_f16(vi2x4567, vi2x89AB, 2); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc4() 152 vi2x4567 = vi2x89AB; in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc4() 184 float16x4_t vi2x89AB = vld1_f16(i2); i2 += 4; in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc4() local 190 vi2x89AB = vreinterpret_f16_u16(vand_u16(vmask, vreinterpret_u16_f16(vi2x89AB))); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc4() 243 const float16x4_t vi2x5678 = vext_f16(vi2x4567, vi2x89AB, 1); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc4() 261 const float16x4_t vi2x6789 = vext_f16(vi2x4567, vi2x89AB, 2); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc4() 262 vi2x4567 = vi2x89AB; in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc4()
|
D | 5x5p2-minmax-neonfp16arith-1x4.c | 80 const float16x4_t vi2x89AB = vld1_f16(i2); i2 += 4; in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4() local 133 const float16x4_t vi2x5678 = vext_f16(vi2x4567, vi2x89AB, 1); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4() 151 const float16x4_t vi2x6789 = vext_f16(vi2x4567, vi2x89AB, 2); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4() 152 vi2x4567 = vi2x89AB; in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4() 181 float16x4_t vi2x89AB = vld1_f16(i2); i2 += 4; in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4() local 187 vi2x89AB = vreinterpret_f16_u16(vand_u16(vmask, vreinterpret_u16_f16(vi2x89AB))); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4() 240 const float16x4_t vi2x5678 = vext_f16(vi2x4567, vi2x89AB, 1); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4() 258 const float16x4_t vi2x6789 = vext_f16(vi2x4567, vi2x89AB, 2); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4() 259 vi2x4567 = vi2x89AB; in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4()
|
D | 5x5p2-minmax-neonfp16arith-1x4-acc2.c | 80 const float16x4_t vi2x89AB = vld1_f16(i2); i2 += 4; in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc2() local 133 const float16x4_t vi2x5678 = vext_f16(vi2x4567, vi2x89AB, 1); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc2() 151 const float16x4_t vi2x6789 = vext_f16(vi2x4567, vi2x89AB, 2); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc2() 152 vi2x4567 = vi2x89AB; in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc2() 182 float16x4_t vi2x89AB = vld1_f16(i2); i2 += 4; in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc2() local 188 vi2x89AB = vreinterpret_f16_u16(vand_u16(vmask, vreinterpret_u16_f16(vi2x89AB))); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc2() 241 const float16x4_t vi2x5678 = vext_f16(vi2x4567, vi2x89AB, 1); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc2() 259 const float16x4_t vi2x6789 = vext_f16(vi2x4567, vi2x89AB, 2); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc2() 260 vi2x4567 = vi2x89AB; in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc2()
|
D | 5x5p2-minmax-neonfp16arith-1x4-acc3.c | 80 const float16x4_t vi2x89AB = vld1_f16(i2); i2 += 4; in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc3() local 133 const float16x4_t vi2x5678 = vext_f16(vi2x4567, vi2x89AB, 1); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc3() 151 const float16x4_t vi2x6789 = vext_f16(vi2x4567, vi2x89AB, 2); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc3() 152 vi2x4567 = vi2x89AB; in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc3() 183 float16x4_t vi2x89AB = vld1_f16(i2); i2 += 4; in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc3() local 189 vi2x89AB = vreinterpret_f16_u16(vand_u16(vmask, vreinterpret_u16_f16(vi2x89AB))); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc3() 242 const float16x4_t vi2x5678 = vext_f16(vi2x4567, vi2x89AB, 1); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc3() 260 const float16x4_t vi2x6789 = vext_f16(vi2x4567, vi2x89AB, 2); in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc3() 261 vi2x4567 = vi2x89AB; in xnn_f16_dwconv2d_chw_ukernel_5x5p2__neonfp16arith_1x4_acc3()
|