Skip to content

D
dlib
Commit 61c2a5eb authored by Davis King's avatar Davis King
Browse files
Options

Sped up the image filtering functions by switching them to use AVX 

instructions when available.
parent 80d36f43
Hide whitespace changes
Inline Side-by-side
Showing with 7 additions and 7 deletions
dlib/image_transforms/spatial_filtering.h
View file @ 61c2a5eb
...@@ -166,10 +166,10 @@ namespace dlib ...@@ -166,10 +166,10 @@ namespace dlib
for (long r = first_row; r < last_row; ++r) for (long r = first_row; r < last_row; ++r)
{ {
long c = first_col; long c = first_col;
for (; c < last_col-3; c+=4) for (; c < last_col-7; c+=8)
{ {
simd4f p,p2,p3; simd8f p,p2,p3;
simd4f temp = 0, temp2=0, temp3=0; simd8f temp = 0, temp2=0, temp3=0;
for (long m = 0; m < filter.nr(); ++m) for (long m = 0; m < filter.nr(); ++m)
{ {
long n = 0; long n = 0;
...@@ -587,9 +587,9 @@ namespace dlib ...@@ -587,9 +587,9 @@ namespace dlib
for (long r = 0; r < in_img.nr(); ++r) for (long r = 0; r < in_img.nr(); ++r)
{ {
long c = first_col; long c = first_col;
for (; c < last_col-3; c+=4) for (; c < last_col-7; c+=8)
{ {
simd4f p,p2,p3, temp = 0, temp2=0, temp3=0; simd8f p,p2,p3, temp = 0, temp2=0, temp3=0;
long n = 0; long n = 0;
for (; n < row_filter.size()-2; n+=3) for (; n < row_filter.size()-2; n+=3)
{ {
...@@ -628,9 +628,9 @@ namespace dlib ...@@ -628,9 +628,9 @@ namespace dlib
for (long r = first_row; r < last_row; ++r) for (long r = first_row; r < last_row; ++r)
{ {
long c = first_col; long c = first_col;
for (; c < last_col-3; c+=4) for (; c < last_col-7; c+=8)
{ {
simd4f p, p2, p3, temp = 0, temp2 = 0, temp3 = 0; simd8f p, p2, p3, temp = 0, temp2 = 0, temp3 = 0;
long m = 0; long m = 0;
for (; m < col_filter.size()-2; m+=3) for (; m < col_filter.size()-2; m+=3)
{ {
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment