mirrors/rpcs3
0
0
Fork 0
mirror of https://github.com/RPCS3/rpcs3.git synced 2025-08-09 01:29:23 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

rsx: Make input_is_swizzled a template parameter

Browse source 
This lowers the relative cost of this function from ~2.25% to ~1.80% on
gcc 9 which I found quite surprising, some of it probably gets inlined
better in the callers, but I haven’t been able to isolate which parts.
This commit is contained in:
Emmanuel Gil Peyrot 2019-10-27 22:55:19 +01:00 • committed by kd-11
commit 69e9ee26f6
2 changed files with 7 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

6
rpcs3/Emu/RSX/rsx_methods.cpp
View file

@ -1263,13 +1263,13 @@ namespace rsx
switch (out_bpp) switch (out_bpp)
{ {
case 1: case 1:
convert_linear_swizzle<u8>(linear_pixels, swizzled_pixels, sw_width, sw_height, in_pitch, false); convert_linear_swizzle<u8, false>(linear_pixels, swizzled_pixels, sw_width, sw_height, in_pitch);
break; break;
case 2: case 2:
convert_linear_swizzle<u16>(linear_pixels, swizzled_pixels, sw_width, sw_height, in_pitch, false); convert_linear_swizzle<u16, false>(linear_pixels, swizzled_pixels, sw_width, sw_height, in_pitch);
break; break;
case 4: case 4:
convert_linear_swizzle<u32>(linear_pixels, swizzled_pixels, sw_width, sw_height, in_pitch, false); convert_linear_swizzle<u32, false>(linear_pixels, swizzled_pixels, sw_width, sw_height, in_pitch);
break; break;
} }
} }

8
rpcs3/Emu/RSX/rsx_utils.h
View file

@ -330,8 +330,8 @@ namespace rsx
* Restriction: It has mixed results if the height or width is not a power of 2 * Restriction: It has mixed results if the height or width is not a power of 2
* Restriction: Only works with 2D surfaces * Restriction: Only works with 2D surfaces
*/ */
template<typename T> template<typename T, bool input_is_swizzled>
void convert_linear_swizzle(void* input_pixels, void* output_pixels, u16 width, u16 height, u32 pitch, bool input_is_swizzled) void convert_linear_swizzle(void* input_pixels, void* output_pixels, u16 width, u16 height, u32 pitch)
{ {
u32 log2width = ceil_log2(width); u32 log2width = ceil_log2(width);
u32 log2height = ceil_log2(height); u32 log2height = ceil_log2(height);
@ -357,7 +357,7 @@ namespace rsx
u32 adv = pitch / sizeof(T); u32 adv = pitch / sizeof(T);
if (!input_is_swizzled) if constexpr (!input_is_swizzled)
{ {
for (int y = 0; y < height; ++y) for (int y = 0; y < height; ++y)
{ {
@ -414,7 +414,7 @@ namespace rsx
{ {
if (depth == 1) if (depth == 1)
{ {
convert_linear_swizzle<T>(input_pixels, output_pixels, width, height, width * sizeof(T), true); convert_linear_swizzle<T, true>(input_pixels, output_pixels, width, height, width * sizeof(T));
return; return;
} }