ROMManager: optimise ROMIcon function

Precompute all 16 5-bit RGB palette colours into 8-bit RGBA to avoid
repeated and superfluous calculation within the nested loop at the
point of index lookup.

A speedup was observed, from ~7ms, to a consistent 1ms
(i.e. now practically instantaneous) through timing with
std::chrono::high_resolution_clock.

Also improve comprehensibility, by using meaningful names, where
appropriate, for loop counter variables.
This commit is contained in:
Rayyan Ansari 2024-04-16 23:06:44 +01:00
parent 95adc87f6d
commit 5a852cb00d
No known key found for this signature in database
GPG Key ID: 46A8D18E5BC49D84
1 changed files with 18 additions and 13 deletions

View File

@ -1572,23 +1572,28 @@ QString GBACartLabel()
void ROMIcon(const u8 (&data)[512], const u16 (&palette)[16], u32 (&iconRef)[32*32])
{
int index = 0;
for (int i = 0; i < 4; i++)
u32 paletteRGBA[16];
for (int i = 0; i < 16; i++)
{
for (int j = 0; j < 4; j++)
u8 r = ((palette[i] >> 0) & 0x1F) * 255 / 31;
u8 g = ((palette[i] >> 5) & 0x1F) * 255 / 31;
u8 b = ((palette[i] >> 10) & 0x1F) * 255 / 31;
u8 a = i ? 255 : 0;
paletteRGBA[i] = r | (g << 8) | (b << 16) | (a << 24);
}
int count = 0;
for (int ytile = 0; ytile < 4; ytile++)
{
for (int k = 0; k < 8; k++)
for (int xtile = 0; xtile < 4; xtile++)
{
for (int l = 0; l < 8; l++)
for (int ypixel = 0; ypixel < 8; ypixel++)
{
u8 pal_index = index % 2 ? data[index/2] >> 4 : data[index/2] & 0x0F;
u8 r = ((palette[pal_index] >> 0) & 0x1F) * 255 / 31;
u8 g = ((palette[pal_index] >> 5) & 0x1F) * 255 / 31;
u8 b = ((palette[pal_index] >> 10) & 0x1F) * 255 / 31;
u8 a = pal_index ? 255: 0;
u32* row = &iconRef[256 * i + 32 * k + 8 * j];
row[l] = r | (g << 8) | (b << 16) | (a << 24);
index++;
for (int xpixel = 0; xpixel < 8; xpixel++)
{
u8 pal_index = count % 2 ? data[count/2] >> 4 : data[count/2] & 0x0F;
iconRef[ytile*256 + ypixel*32 + xtile*8 + xpixel] = paletteRGBA[pal_index];
count++;
}
}
}