Opencl-sobel detection and opencl-sobel

Source: Internet
Author: User

Opencl-sobel detection and opencl-sobel

Comparison of the time of the C version, neon, and GPU detected by sobel.


Platform: LG G3, Adreno 330, img size 3264x2448


Sobel:

C code

Neon

GPU

73

13

42 + 3.7 + 6.6

Unit: ms GPU time = memorytime + Queued time + Run time




Sobel org

Sobel vector

Sobel vector + mem_fence

Queued time

4.6

7.2

2.8

Wait time

0.07

0.09

0.07

Run time

66.9

7.3

6.6






Typedef unsigned char BYTE; void sobel (BYTE * src, int w, int h, BYTE * Ix, BYTE * Iy) {int src_step = w; int dst_step = w; int x, height = h-2; BYTE * dstX = Ix + dst_step; BYTE * dstY = Iy + dst_step; for (; height --; src + = src_step, dstX + = dst_step, dstY + = dst_step) {const BYTE * src2 = src + src_step; const BYTE * src3 = src + src_step * 2; for (x = 1; x <W-1; x ++) {short t0 = 0; short t1 = 0; t0 =-src [x-1] + src [x + 1]; t1 = src [x-1] + (src [x] <1) + src [x + 1]; t0 + = (-src2 [x-1] + src2 [x + 1]) <1); t0 + =-src3 [x-1] + src3 [x + 1]; t1-= (src3[ x-1] + (src3 [x] <1) + src3 [x + 1]); dstX [x] = t0> 3; dstY [x] = t1> 3 ;}} void sobel_neon (BYTE * src, int w, int h, BYTE * Ix, BYTE * Iy) {int src_step = w; int dst_step = w; int x, height = h-2; BYTE * dstX = Ix + dst_step; BYTE * dstY = Iy + dst_step; for (; height --; src + = src_step, dstX + = dst_step, dstY + = dst_step) {const BYTE * src2 = src + src_step; const BYTE * src3 = src + src_step * 2; x = 1; while (x + 8) <= W-1) {uint8x8_t left = vld1_u8 (src + x-1); uint8x8_t mid = vld1_u8 (src + x ); scheme right = vld1_u8 (src + x + 1); int16x8_t t0 = Scheme (vsubl_u8 (right, left); int16x8_t t1 = vaddq_s16 (scheme (vaddl_u8 (left, right )), values (vshll_n_u8 (mid, 1); left = vld1_u8 (src2 + x-1); right = vld1_u8 (src2 + x + 1); int16x8_t temp = values (vsubl_u8 (right, left); t0 = vaddq_s16 (t0, vshlq_n_s16 (temp, 1); left = vld1_u8 (src3 + x-1); mid = vld1_u8 (src3 + x ); right = vld1_u8 (src3 + x + 1); t0 = vaddq_s16 (t0, clerk (vsubl_u8 (right, left); temp = vaddq_s16 (Clerk (vaddl_u8 (left, right), round (vshll_n_u8 (mid, 1); t1 = vsubq_s16 (t1, temp); vst1_s8 (int8_t *) dstX + x, vshrn_n_s16 (t0, (3); vst1_s8 (int8_t *) dstY + x, vshrn_n_s16 (t1, 3); x + = 8;} while (x) <W-1) {short t0 = 0; short t1 = 0; t0 =-src [x-1] + src [x + 1]; t1 = src [x-1] + (src [x] <1) + src [x + 1]; t0 + = (-src2 [x-1] + src2 [x + 1]) <1); t0 + =-src3 [x-1] + src3 [x + 1]; t1-= (src3[ x-1] + (src3 [x] <1) + src3 [x + 1]); dstX [x] = t0> 3; dstY [x] = t1> 3; x ++ ;}}}View Code

 

Contact Us

The content source of this page is from Internet, which doesn't represent Alibaba Cloud's opinion; products and services mentioned on that page don't have any relationship with Alibaba Cloud. If the content of the page makes you feel confusing, please write us an email, we will handle the problem within 5 days after receiving your email.

If you find any instances of plagiarism from the community, please send an email to: info-contact@alibabacloud.com and provide relevant evidence. A staff member will contact you within 5 working days.

A Free Trial That Lets You Build Big!

Start building with 50+ products and up to 12 months usage for Elastic Compute Service

  • Sales Support

    1 on 1 presale consultation

  • After-Sales Support

    24/7 Technical Support 6 Free Tickets per Quarter Faster Response

  • Alibaba Cloud offers highly flexible support services tailored to meet your exact needs.