| Christophe Fontaine | 33e8195 | 2016-12-19 14:41:52 +0100 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (c) 2015 Cisco and/or its affiliates. |
| 3 | * Licensed under the Apache License, Version 2.0 (the "License"); |
| 4 | * you may not use this file except in compliance with the License. |
| 5 | * You may obtain a copy of the License at: |
| 6 | * |
| 7 | * http://www.apache.org/licenses/LICENSE-2.0 |
| 8 | * |
| 9 | * Unless required by applicable law or agreed to in writing, software |
| 10 | * distributed under the License is distributed on an "AS IS" BASIS, |
| 11 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 12 | * See the License for the specific language governing permissions and |
| 13 | * limitations under the License. |
| 14 | */ |
| 15 | |
| 16 | #ifndef included_vector_neon_h |
| 17 | #define included_vector_neon_h |
| 18 | #include <arm_neon.h> |
| 19 | |
| 20 | /* Splats. */ |
| 21 | |
| 22 | #define u8x16_splat(i) vdupq_n_u8(i) |
| 23 | #define u16x8_splat(i) vdupq_n_u16(i) |
| 24 | #define i16x8_splat(i) vdupq_n_s16(i) |
| 25 | #define u32x4_splat(i) vdupq_n_u32(i) |
| 26 | #define i32x4_splat(i) vdupq_n_s32(i) |
| 27 | |
| 28 | /* Arithmetic */ |
| 29 | #define u16x8_add(a,b) vaddq_u16(a,b) |
| 30 | #define i16x8_add(a,b) vaddq_s16(a,b) |
| 31 | #define u16x8_sub_saturate(a,b) vsubq_u16(a,b) |
| 32 | #define i16x8_sub_saturate(a,b) vsubq_s16(a,b) |
| 33 | |
| 34 | #define u16x8_is_equal(a,b) vceqq_u16(a,b) |
| 35 | #define i16x8_is_equal(a,b) vceqq_i16(a,b) |
| 36 | |
| 37 | always_inline u32 |
| 38 | u16x8_zero_byte_mask (u16x8 input) |
| 39 | { |
| 40 | u8x16 vall_one = vdupq_n_u8 (0x0); |
| 41 | u8x16 res_values = { 0x01, 0x02, 0x04, 0x08, |
| 42 | 0x10, 0x20, 0x40, 0x80, |
| 43 | 0x01, 0x02, 0x04, 0x08, |
| 44 | 0x10, 0x20, 0x40, 0x80 |
| 45 | }; |
| 46 | |
| 47 | /* input --> [0x80, 0x40, 0x01, 0xf0, ... ] */ |
| 48 | u8x16 test_result = |
| 49 | vreinterpretq_u8_u16 (vceqq_u16 (input, vreinterpretq_u16_u8 (vall_one))); |
| 50 | u8x16 before_merge = vminq_u8 (test_result, res_values); |
| 51 | /*before_merge--> [0x80, 0x00, 0x00, 0x10, ... ] */ |
| 52 | /* u8x16 --> [a,b,c,d, e,f,g,h, i,j,k,l, m,n,o,p] */ |
| 53 | /* pair add until we have 2 uint64_t */ |
| 54 | u16x8 merge1 = vpaddlq_u8 (before_merge); |
| 55 | /* u16x8--> [a+b,c+d, e+f,g+h, i+j,k+l, m+n,o+p] */ |
| 56 | u32x4 merge2 = vpaddlq_u16 (merge1); |
| 57 | /* u32x4--> [a+b+c+d, e+f+g+h, i+j+k+l, m+n+o+p] */ |
| 58 | u64x2 merge3 = vpaddlq_u32 (merge2); |
| 59 | /* u64x2--> [a+b+c+d+e+f+g+h, i+j+k+l+m+n+o+p] */ |
| 60 | return (u32) (vgetq_lane_u64 (merge3, 1) << 8) + vgetq_lane_u64 (merge3, 0); |
| 61 | } |
| 62 | |
| 63 | #endif /* included_vector_neon_h */ |
| 64 | |
| 65 | /* |
| 66 | * fd.io coding-style-patch-verification: ON |
| 67 | * |
| 68 | * Local Variables: |
| 69 | * eval: (c-set-style "gnu") |
| 70 | * End: |
| 71 | */ |