Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Optimize color conversions in non-JIT vertex decoder #19241

Merged
merged 4 commits into from
Jun 4, 2024
Merged

Optimize color conversions in non-JIT vertex decoder #19241

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
f2b4ad0
Write new 16-bit to 32-bit color conversion utility functions
hrydgard Jun 4, 2024
5b14cf9
Switch to the new functions.
hrydgard Jun 4, 2024
9ac7054
Vertex decoder (non-JIT): Optimize 16-bit color decoders.
hrydgard Jun 4, 2024
2174a89
Clarify operator precedence
hrydgard Jun 4, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
26 changes: 14 additions & 12 deletions Common/Data/Convert/ColorConv.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -64,26 +64,28 @@ inline u32 RGBA4444ToRGBA8888(u16 src) {
const u32 g = (src & 0x00F0) << 4;
const u32 b = (src & 0x0F00) << 8;
const u32 a = (src & 0xF000) << 12;

const u32 c = r | g | b | a;
return c | (c << 4);
}

inline u32 RGBA5551ToRGBA8888(u16 src) {
u8 r = Convert5To8((src >> 0) & 0x1F);
u8 g = Convert5To8((src >> 5) & 0x1F);
u8 b = Convert5To8((src >> 10) & 0x1F);
u8 a = (src >> 15) & 0x1;
a = (a) ? 0xff : 0;
return (a << 24) | (b << 16) | (g << 8) | r;
u32 dark = ((src & 0x1F) << 3) | ((src & 0x3E0) << 6) | ((src & 0x7C00) << 9);
// Replicate the top 3 upper bits into the missing lower bits.
u32 full = (dark | ((dark >> 5) & 0x070707));
if (src >> 15) {
full |= 0xFF000000;
}
return full;
}

inline u32 RGB565ToRGBA8888(u16 src) {
u8 r = Convert5To8((src >> 0) & 0x1F);
u8 g = Convert6To8((src >> 5) & 0x3F);
u8 b = Convert5To8((src >> 11) & 0x1F);
u8 a = 0xFF;
return (a << 24) | (b << 16) | (g << 8) | r;
u32 dark_rb = ((src & 0x1F) << 3) | ((src & 0xF800) << 8);
// Replicate the top 3 upper bits into the missing lower bits.
u32 full_rb = (dark_rb | ((dark_rb >> 5) & 0x070007));
// Add in green (6 bits instead of 5).
u32 dark_g = ((src & 0x7E0) << 5);
u32 full_g = dark_g | ((dark_g >> 6) & 0x300);
return full_rb | full_g | 0xFF000000;
}

inline u16 RGBA8888ToRGB565(u32 value) {
Expand Down
33 changes: 16 additions & 17 deletions GPU/Common/VertexDecoderCommon.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -506,40 +506,39 @@ void VertexDecoder::Step_ColorInvalid() const

void VertexDecoder::Step_Color565() const
{
u8 *c = decoded_ + decFmt.c0off;
u16 cdata = *(const u16_le *)(ptr_ + coloff);
c[0] = Convert5To8(cdata & 0x1f);
c[1] = Convert6To8((cdata >> 5) & 0x3f);
c[2] = Convert5To8((cdata >> 11) & 0x1f);
c[3] = 255;
// Always full alpha.
u32 *c = (u32 *)(decoded_ + decFmt.c0off);
*c = RGB565ToRGBA8888(cdata);
}

void VertexDecoder::Step_Color5551() const
{
u8 *c = decoded_ + decFmt.c0off;
u16 cdata = *(const u16_le *)(ptr_ + coloff);
gstate_c.vertexFullAlpha = gstate_c.vertexFullAlpha && (cdata >> 15) != 0;
c[0] = Convert5To8(cdata & 0x1f);
c[1] = Convert5To8((cdata >> 5) & 0x1f);
c[2] = Convert5To8((cdata >> 10) & 0x1f);
c[3] = (cdata >> 15) ? 255 : 0;
u32 *c = (u32 *)(decoded_ + decFmt.c0off);
int alpha = (cdata >> 15);
if (!alpha) {
gstate_c.vertexFullAlpha = false;
}
*c = RGBA5551ToRGBA8888(cdata);
}

void VertexDecoder::Step_Color4444() const
{
u8 *c = decoded_ + decFmt.c0off;
u16 cdata = *(const u16_le *)(ptr_ + coloff);
gstate_c.vertexFullAlpha = gstate_c.vertexFullAlpha && (cdata >> 12) == 0xF;
for (int j = 0; j < 4; j++)
c[j] = Convert4To8((cdata >> (j * 4)) & 0xF);
u32 *c = (u32 *)(decoded_ + decFmt.c0off);
if ((cdata >> 12) != 0xF) {
gstate_c.vertexFullAlpha = false;
}
*c = RGBA4444ToRGBA8888(cdata);
}

void VertexDecoder::Step_Color8888() const
{
u8 *c = decoded_ + decFmt.c0off;
const u8 *cdata = (const u8*)(ptr_ + coloff);
gstate_c.vertexFullAlpha = gstate_c.vertexFullAlpha && cdata[3] == 255;
if (cdata[3] != 255) {
gstate_c.vertexFullAlpha = false;
}
memcpy(c, cdata, sizeof(u8) * 4);
}

Expand Down
36 changes: 36 additions & 0 deletions unittest/UnitTest.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -67,6 +67,7 @@
#include "Common/Log.h"
#include "Common/StringUtils.h"
#include "Core/Config.h"
#include "Common/Data/Convert/ColorConv.h"
#include "Common/File/VFS/VFS.h"
#include "Common/File/VFS/DirectoryReader.h"
#include "Core/FileSystems/ISOFileSystem.h"
Expand Down Expand Up @@ -997,6 +998,40 @@ bool TestIniFile() {
return true;
}

inline u32 ReferenceRGBA5551ToRGBA8888(u16 src) {
u8 r = Convert5To8((src >> 0) & 0x1F);
u8 g = Convert5To8((src >> 5) & 0x1F);
u8 b = Convert5To8((src >> 10) & 0x1F);
u8 a = (src >> 15) & 0x1;
a = (a) ? 0xff : 0;
return (a << 24) | (b << 16) | (g << 8) | r;
}

inline u32 ReferenceRGB565ToRGBA8888(u16 src) {
u8 r = Convert5To8((src >> 0) & 0x1F);
u8 g = Convert6To8((src >> 5) & 0x3F);
u8 b = Convert5To8((src >> 11) & 0x1F);
u8 a = 0xFF;
return (a << 24) | (b << 16) | (g << 8) | r;
}

bool TestColorConv() {
// Can exhaustively test the 16->32 conversions.
for (int i = 0; i < 65536; i++) {
u16 col16 = i;

u32 reference = ReferenceRGBA5551ToRGBA8888(col16);
u32 value = RGBA5551ToRGBA8888(col16);
EXPECT_EQ_INT(reference, value);

reference = ReferenceRGB565ToRGBA8888(col16);
value = RGB565ToRGBA8888(col16);
EXPECT_EQ_INT(reference, value);
}

return true;
}

typedef bool (*TestFunc)();
struct TestItem {
const char *name;
Expand Down Expand Up @@ -1056,6 +1091,7 @@ TestItem availableTests[] = {
TEST_ITEM(VFS),
TEST_ITEM(Substitutions),
TEST_ITEM(IniFile),
TEST_ITEM(ColorConv),
};

int main(int argc, const char *argv[]) {
Expand Down
Loading