/external/XNNPACK/src/f32-prelu/gen/ |
D | wasmsimd-minmax-1x16.c | 50 v128_t vi0x89AB = wasm_v128_load(i0 + 8); in xnn_f32_prelu_ukernel__wasmsimd_minmax_1x16() local 58 v128_t vacc0x89AB = wasm_i32x4_max(vi0x89AB, vzero); in xnn_f32_prelu_ukernel__wasmsimd_minmax_1x16() 59 vi0x89AB = wasm_i32x4_min(vi0x89AB, vzero); in xnn_f32_prelu_ukernel__wasmsimd_minmax_1x16() 65 vacc0x89AB = wasm_f32x4_add(vacc0x89AB, wasm_f32x4_mul(vi0x89AB, vw89AB)); in xnn_f32_prelu_ukernel__wasmsimd_minmax_1x16()
|
D | wasmsimd-bitselect-1x16.c | 50 const v128_t vi0x89AB = wasm_v128_load(i0 + 8); in xnn_f32_prelu_ukernel__wasmsimd_bitselect_1x16() local 58 v128_t vacc0x89AB = wasm_f32x4_mul(vi0x89AB, vw89AB); in xnn_f32_prelu_ukernel__wasmsimd_bitselect_1x16() 59 const v128_t vmask0x89AB = wasm_i32x4_lt(vi0x89AB, vzero); in xnn_f32_prelu_ukernel__wasmsimd_bitselect_1x16() 65 vacc0x89AB = wasm_v128_bitselect(vacc0x89AB, vi0x89AB, vmask0x89AB); in xnn_f32_prelu_ukernel__wasmsimd_bitselect_1x16()
|
D | neon-1x16.c | 48 const float32x4_t vi0x89AB = vld1q_f32(i0); i0 += 4; in xnn_f32_prelu_ukernel__neon_1x16() local 55 float32x4_t vacc0x89AB = vmulq_f32(vi0x89AB, vw89AB); in xnn_f32_prelu_ukernel__neon_1x16() 56 const uint32x4_t vm0x89AB = vcltq_s32(vreinterpretq_s32_f32(vi0x89AB), vmovq_n_s32(0)); in xnn_f32_prelu_ukernel__neon_1x16() 62 vacc0x89AB = vbslq_f32(vm0x89AB, vacc0x89AB, vi0x89AB); in xnn_f32_prelu_ukernel__neon_1x16()
|
D | wasmsimd-minmax-2x16.c | 56 v128_t vi0x89AB = wasm_v128_load(i0 + 8); in xnn_f32_prelu_ukernel__wasmsimd_minmax_2x16() local 69 v128_t vacc0x89AB = wasm_i32x4_max(vi0x89AB, vzero); in xnn_f32_prelu_ukernel__wasmsimd_minmax_2x16() 70 vi0x89AB = wasm_i32x4_min(vi0x89AB, vzero); in xnn_f32_prelu_ukernel__wasmsimd_minmax_2x16() 84 vacc0x89AB = wasm_f32x4_add(vacc0x89AB, wasm_f32x4_mul(vi0x89AB, vw89AB)); in xnn_f32_prelu_ukernel__wasmsimd_minmax_2x16()
|
/external/XNNPACK/src/f32-dwconv2d-chw/gen/ |
D | 3x3s2p1-minmax-wasmsimd-arm-splat-1x4.c | 72 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4() local 82 const v128_t vi0x8ACE = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4() 83 const v128_t vi0x9BDF = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4() 125 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4() local 132 …const v128_t vi0x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4() 133 …const v128_t vi0x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4()
|
D | 3x3s2p1-minmax-wasmsimd-x86-splat-1x4.c | 72 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4() local 82 const v128_t vi0x8ACE = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4() 83 const v128_t vi0x9BDF = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4() 125 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4() local 132 …const v128_t vi0x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4() 133 …const v128_t vi0x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4()
|
D | 3x3s2p1-minmax-wasmsimd-x86-splat-1x4-acc2.c | 72 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc2() local 82 const v128_t vi0x8ACE = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc2() 83 const v128_t vi0x9BDF = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc2() 126 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc2() local 133 …const v128_t vi0x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc2() 134 …const v128_t vi0x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc2()
|
D | 3x3s2p1-minmax-wasmsimd-arm-loadsplat-1x4-acc2.c | 82 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc2() local 92 const v128_t vi0x8ACE = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc2() 93 const v128_t vi0x9BDF = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc2() 136 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc2() local 143 …const v128_t vi0x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc2() 144 …const v128_t vi0x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc2()
|
D | 3x3s2p1-minmax-wasmsimd-arm-splat-1x4-acc3.c | 72 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc3() local 82 const v128_t vi0x8ACE = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc3() 83 const v128_t vi0x9BDF = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc3() 127 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc3() local 134 …const v128_t vi0x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc3() 135 …const v128_t vi0x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc3()
|
D | 3x3s2p1-minmax-wasmsimd-arm-splat-1x4-acc2.c | 72 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc2() local 82 const v128_t vi0x8ACE = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc2() 83 const v128_t vi0x9BDF = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc2() 126 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc2() local 133 …const v128_t vi0x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc2() 134 …const v128_t vi0x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc2()
|
D | 3x3s2p1-minmax-wasmsimd-arm-loadsplat-1x4.c | 82 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4() local 92 const v128_t vi0x8ACE = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4() 93 const v128_t vi0x9BDF = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4() 135 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4() local 142 …const v128_t vi0x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4() 143 …const v128_t vi0x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4()
|
D | 3x3s2p1-minmax-wasmsimd-arm-splat-1x4-acc4.c | 72 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc4() local 82 const v128_t vi0x8ACE = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc4() 83 const v128_t vi0x9BDF = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc4() 128 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc4() local 135 …const v128_t vi0x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc4() 136 …const v128_t vi0x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_splat_1x4_acc4()
|
D | 3x3s2p1-minmax-wasmsimd-x86-splat-1x4-acc4.c | 72 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc4() local 82 const v128_t vi0x8ACE = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc4() 83 const v128_t vi0x9BDF = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc4() 128 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc4() local 135 …const v128_t vi0x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc4() 136 …const v128_t vi0x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc4()
|
D | 3x3s2p1-minmax-wasmsimd-x86-loadsplat-1x4.c | 82 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4() local 92 const v128_t vi0x8ACE = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4() 93 const v128_t vi0x9BDF = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4() 135 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4() local 142 …const v128_t vi0x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4() 143 …const v128_t vi0x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4()
|
D | 3x3s2p1-minmax-wasmsimd-x86-loadsplat-1x4-acc2.c | 82 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc2() local 92 const v128_t vi0x8ACE = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc2() 93 const v128_t vi0x9BDF = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc2() 136 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc2() local 143 …const v128_t vi0x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc2() 144 …const v128_t vi0x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc2()
|
D | 3x3s2p1-minmax-wasmsimd-arm-loadsplat-1x4-acc3.c | 82 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc3() local 92 const v128_t vi0x8ACE = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc3() 93 const v128_t vi0x9BDF = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc3() 137 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc3() local 144 …const v128_t vi0x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc3() 145 …const v128_t vi0x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc3()
|
D | 3x3s2p1-minmax-sse-1x4-acc2.c | 74 const __m128 vi0x89AB = _mm_loadu_ps(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc2() local 84 const __m128 vi0x8ACE = _mm_shuffle_ps(vi0x89AB, vi0xCDEF, _MM_SHUFFLE(2, 0, 2, 0)); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc2() 85 const __m128 vi0x9BDF = _mm_shuffle_ps(vi0x89AB, vi0xCDEF, _MM_SHUFFLE(3, 1, 3, 1)); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc2() 127 const __m128 vi0x89AB = _mm_loadu_ps(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc2() local 134 …const __m128 vi0x8ACE = _mm_and_ps(vmask_even, _mm_shuffle_ps(vi0x89AB, vi0xCDEF, _MM_SHUFFLE(2, 0… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc2() 135 …const __m128 vi0x9BDF = _mm_and_ps(vmask_odd, _mm_shuffle_ps(vi0x89AB, vi0xCDEF, _MM_SHUFFLE(3, 1… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc2()
|
D | 3x3s2p1-minmax-sse-1x4-acc3.c | 74 const __m128 vi0x89AB = _mm_loadu_ps(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc3() local 84 const __m128 vi0x8ACE = _mm_shuffle_ps(vi0x89AB, vi0xCDEF, _MM_SHUFFLE(2, 0, 2, 0)); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc3() 85 const __m128 vi0x9BDF = _mm_shuffle_ps(vi0x89AB, vi0xCDEF, _MM_SHUFFLE(3, 1, 3, 1)); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc3() 128 const __m128 vi0x89AB = _mm_loadu_ps(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc3() local 135 …const __m128 vi0x8ACE = _mm_and_ps(vmask_even, _mm_shuffle_ps(vi0x89AB, vi0xCDEF, _MM_SHUFFLE(2, 0… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc3() 136 …const __m128 vi0x9BDF = _mm_and_ps(vmask_odd, _mm_shuffle_ps(vi0x89AB, vi0xCDEF, _MM_SHUFFLE(3, 1… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc3()
|
D | 3x3s2p1-minmax-wasmsimd-x86-loadsplat-1x4-acc3.c | 82 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc3() local 92 const v128_t vi0x8ACE = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc3() 93 const v128_t vi0x9BDF = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc3() 137 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc3() local 144 …const v128_t vi0x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc3() 145 …const v128_t vi0x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc3()
|
D | 3x3s2p1-minmax-sse-1x4.c | 74 const __m128 vi0x89AB = _mm_loadu_ps(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4() local 84 const __m128 vi0x8ACE = _mm_shuffle_ps(vi0x89AB, vi0xCDEF, _MM_SHUFFLE(2, 0, 2, 0)); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4() 85 const __m128 vi0x9BDF = _mm_shuffle_ps(vi0x89AB, vi0xCDEF, _MM_SHUFFLE(3, 1, 3, 1)); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4() 126 const __m128 vi0x89AB = _mm_loadu_ps(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4() local 133 …const __m128 vi0x8ACE = _mm_and_ps(vmask_even, _mm_shuffle_ps(vi0x89AB, vi0xCDEF, _MM_SHUFFLE(2, 0… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4() 134 …const __m128 vi0x9BDF = _mm_and_ps(vmask_odd, _mm_shuffle_ps(vi0x89AB, vi0xCDEF, _MM_SHUFFLE(3, 1… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4()
|
D | 3x3s2p1-minmax-wasmsimd-x86-splat-1x4-acc3.c | 72 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc3() local 82 const v128_t vi0x8ACE = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc3() 83 const v128_t vi0x9BDF = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc3() 127 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc3() local 134 …const v128_t vi0x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc3() 135 …const v128_t vi0x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_splat_1x4_acc3()
|
D | 3x3s2p1-minmax-wasmsimd-arm-loadsplat-1x4-acc4.c | 82 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc4() local 92 const v128_t vi0x8ACE = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc4() 93 const v128_t vi0x9BDF = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc4() 138 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc4() local 145 …const v128_t vi0x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc4() 146 …const v128_t vi0x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_arm_loadsplat_1x4_acc4()
|
D | 3x3s2p1-minmax-wasmsimd-x86-loadsplat-1x4-acc4.c | 82 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc4() local 92 const v128_t vi0x8ACE = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, 6); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc4() 93 const v128_t vi0x9BDF = wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, 7); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc4() 138 const v128_t vi0x89AB = wasm_v128_load(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc4() local 145 …const v128_t vi0x8ACE = wasm_v128_and(vmask_even, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 0, 2, 4, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc4() 146 …const v128_t vi0x9BDF = wasm_v128_and(vmask_odd, wasm_v32x4_shuffle(vi0x89AB, vi0xCDEF, 1, 3, 5, … in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__wasmsimd_x86_loadsplat_1x4_acc4()
|
D | 3x3s2p1-minmax-sse-1x4-acc4.c | 74 const __m128 vi0x89AB = _mm_loadu_ps(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc4() local 84 const __m128 vi0x8ACE = _mm_shuffle_ps(vi0x89AB, vi0xCDEF, _MM_SHUFFLE(2, 0, 2, 0)); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc4() 85 const __m128 vi0x9BDF = _mm_shuffle_ps(vi0x89AB, vi0xCDEF, _MM_SHUFFLE(3, 1, 3, 1)); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc4() 129 const __m128 vi0x89AB = _mm_loadu_ps(i0); in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc4() local 136 …const __m128 vi0x8ACE = _mm_and_ps(vmask_even, _mm_shuffle_ps(vi0x89AB, vi0xCDEF, _MM_SHUFFLE(2, 0… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc4() 137 …const __m128 vi0x9BDF = _mm_and_ps(vmask_odd, _mm_shuffle_ps(vi0x89AB, vi0xCDEF, _MM_SHUFFLE(3, 1… in xnn_f32_dwconv2d_chw_ukernel_3x3s2p1__sse_1x4_acc4()
|
D | 5x5p2-minmax-neonfma-1x4-acc3.c | 80 const float32x4_t vi0x89AB = vld1q_f32(i0); i0 += 4; in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4_acc3() local 133 const float32x4_t vi0x5678 = vextq_f32(vi0x4567, vi0x89AB, 1); in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4_acc3() 149 const float32x4_t vi0x6789 = vextq_f32(vi0x4567, vi0x89AB, 2); in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4_acc3() 150 vi0x4567 = vi0x89AB; in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4_acc3() 183 float32x4_t vi0x89AB = vld1q_f32(i0); i0 += 4; in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4_acc3() local 189 vi0x89AB = vreinterpretq_f32_u32(vandq_u32(vmask, vreinterpretq_u32_f32(vi0x89AB))); in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4_acc3() 242 const float32x4_t vi0x5678 = vextq_f32(vi0x4567, vi0x89AB, 1); in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4_acc3() 258 const float32x4_t vi0x6789 = vextq_f32(vi0x4567, vi0x89AB, 2); in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4_acc3() 259 vi0x4567 = vi0x89AB; in xnn_f32_dwconv2d_chw_ukernel_5x5p2__neonfma_1x4_acc3()
|