mirrors/stb
1
0
Fork 0
mirror of https://github.com/nothings/stb synced 2024-12-15 04:22:35 +03:00
Code Issues Packages Projects Releases Wiki Activity

Reorder these loops because I think we get a cache win if we write the entire ring buffer entry at once.

Browse Source
This commit is contained in:
Jorge Rodriguez 2014-07-26 13:12:48 -07:00
parent 01fb58d6b4
commit 155c71fb90
1 changed files with 9 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
stb_resample.h
View File

@ -777,26 +777,26 @@ static void stbr__resample_vertical_downsample(stbr__info* stbr_info, int n, int
int n0 = vertical_contributors->n0; int n0 = vertical_contributors->n0;
int n1 = vertical_contributors->n1; int n1 = vertical_contributors->n1;
int max_n = stbr__min(n1, output_h - 1);
stbr__output_decode_coefficients output_decode_coefficients_fn = stbr__get_output_decode_coefficients_function(channels); stbr__output_decode_coefficients output_decode_coefficients_fn = stbr__get_output_decode_coefficients_function(channels);
STBR_DEBUG_ASSERT(!stbr__use_height_upsampling(stbr_info)); STBR_DEBUG_ASSERT(!stbr__use_height_upsampling(stbr_info));
STBR_DEBUG_ASSERT(n0 >= in_first_scanline); STBR_DEBUG_ASSERT(n0 >= in_first_scanline);
STBR_DEBUG_ASSERT(n1 <= in_last_scanline); STBR_DEBUG_ASSERT(n1 <= in_last_scanline);
STBR_DEBUG_ASSERT(n1 >= n0);
for (x = 0; x < output_w; x++) // Using min and max to avoid writing into ring buffers that will be thrown out.
for (k = stbr__max(n0, 0); k <= max_n; k++)
{ {
int in_texel_index = x * channels; int coefficient_index = k - n0;
int max_n = stbr__min(n1, output_h-1);
STBR_DEBUG_ASSERT(n1 >= n0); float* ring_buffer_entry = stbr__get_ring_buffer_scanline(k, ring_buffer, ring_buffer_begin_index, ring_buffer_first_scanline, kernel_texel_width, ring_buffer_length);
float coefficient = vertical_coefficients[coefficient_index];
// Using min and max to avoid writing into ring buffers that will be thrown out. for (x = 0; x < output_w; x++)
for (k = stbr__max(n0, 0); k <= max_n; k++)
{ {
int coefficient_index = k - n0; int in_texel_index = x * channels;
float* ring_buffer_entry = stbr__get_ring_buffer_scanline(k, ring_buffer, ring_buffer_begin_index, ring_buffer_first_scanline, kernel_texel_width, ring_buffer_length);
float coefficient = vertical_coefficients[coefficient_index];
output_decode_coefficients_fn(ring_buffer_entry, in_texel_index, horizontal_buffer, in_texel_index, channels, coefficient); output_decode_coefficients_fn(ring_buffer_entry, in_texel_index, horizontal_buffer, in_texel_index, channels, coefficient);
} }