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

Optimized transform.scale2x() #2859

Merged
merged 4 commits into from
Jun 30, 2024
Merged

Optimized transform.scale2x() #2859

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
39539df
optimized all Bpp cases of scale2x
itzpr3d4t0r May 18, 2024
a4cbd0c
finish 3Bpp case
itzpr3d4t0r May 18, 2024
4e41d84
removed restrict keyword, added Bpp var
itzpr3d4t0r May 28, 2024
8acbe81
format
itzpr3d4t0r May 28, 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
252 changes: 143 additions & 109 deletions src_c/scale2x.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,13 +34,19 @@
#define MAX(a, b) (((a) > (b)) ? (a) : (b))
#define MIN(a, b) (((a) < (b)) ? (a) : (b))

#define READINT24(x) ((x)[0] << 16 | (x)[1] << 8 | (x)[2])
#define WRITEINT24(x, i) \
{ \
(x)[0] = i >> 16; \
(x)[1] = (i >> 8) & 0xff; \
x[2] = i & 0xff; \
}
static inline int
read_int24(const Uint8 *x)
{
return (x[0] << 16 | x[1] << 8 | x[2]);
}

static inline void
store_int24(Uint8 *x, int i)
{
x[0] = i >> 16;
x[1] = (i >> 8) & 0xff;
x[2] = i & 0xff;
}

/*
this requires a destination surface already setup to be twice as
Expand All @@ -62,137 +68,165 @@ scale2x(SDL_Surface *src, SDL_Surface *dst)
const int height = src->h;

#if SDL_VERSION_ATLEAST(3, 0, 0)
switch (src->format->bytes_per_pixel) {
const Uint8 Bpp = src->format->bytes_per_pixel;
#else
switch (src->format->BytesPerPixel) {
const Uint8 Bpp = src->format->BytesPerPixel;
#endif

switch (Bpp) {
case 1: {
Uint8 E0, E1, E2, E3, B, D, E, F, H;
for (looph = 0; looph < height; ++looph) {
Uint8 *src_row = srcpix + looph * srcpitch;
Uint8 *dst_row0 = dstpix + looph * 2 * dstpitch;
Uint8 *dst_row1 = dstpix + (looph * 2 + 1) * dstpitch;

Uint8 *src_row_prev = srcpix + MAX(0, looph - 1) * srcpitch;
Uint8 *src_row_next =
srcpix + MIN(height - 1, looph + 1) * srcpitch;

for (loopw = 0; loopw < width; ++loopw) {
B = *(Uint8 *)(srcpix + (MAX(0, looph - 1) * srcpitch) +
(1 * loopw));
D = *(Uint8 *)(srcpix + (looph * srcpitch) +
(1 * MAX(0, loopw - 1)));
E = *(Uint8 *)(srcpix + (looph * srcpitch) + (1 * loopw));
F = *(Uint8 *)(srcpix + (looph * srcpitch) +
(1 * MIN(width - 1, loopw + 1)));
H = *(Uint8 *)(srcpix +
(MIN(height - 1, looph + 1) * srcpitch) +
(1 * loopw));

E0 = D == B && B != F && D != H ? D : E;
E1 = B == F && B != D && F != H ? F : E;
E2 = D == H && D != B && H != F ? D : E;
E3 = H == F && D != H && B != F ? F : E;

*(Uint8 *)(dstpix + looph * 2 * dstpitch + loopw * 2 * 1) =
E0;
*(Uint8 *)(dstpix + looph * 2 * dstpitch +
(loopw * 2 + 1) * 1) = E1;
*(Uint8 *)(dstpix + (looph * 2 + 1) * dstpitch +
loopw * 2 * 1) = E2;
*(Uint8 *)(dstpix + (looph * 2 + 1) * dstpitch +
(loopw * 2 + 1) * 1) = E3;
B = *(Uint8 *)(src_row_prev + loopw);
D = *(Uint8 *)(src_row + MAX(0, loopw - 1));
E = *(Uint8 *)(src_row + loopw);
F = *(Uint8 *)(src_row + MIN(width - 1, loopw + 1));
H = *(Uint8 *)(src_row_next + loopw);

if (B != H && D != F) {
E0 = (D == B) ? D : E;
E1 = (B == F) ? F : E;
E2 = (D == H) ? D : E;
E3 = (H == F) ? F : E;
}
else {
E0 = E;
E1 = E;
E2 = E;
E3 = E;
}

*(Uint8 *)(dst_row0 + loopw * 2) = E0;
*(Uint8 *)(dst_row0 + loopw * 2 + 1) = E1;
*(Uint8 *)(dst_row1 + loopw * 2) = E2;
*(Uint8 *)(dst_row1 + loopw * 2 + 1) = E3;
}
}
break;
}
case 2: {
Uint16 E0, E1, E2, E3, B, D, E, F, H;
for (looph = 0; looph < height; ++looph) {
Uint8 *src_row = srcpix + looph * srcpitch;
Uint8 *dst_row0 = dstpix + looph * 2 * dstpitch;
Uint8 *dst_row1 = dstpix + (looph * 2 + 1) * dstpitch;

Uint8 *src_row_prev = srcpix + MAX(0, looph - 1) * srcpitch;
Uint8 *src_row_next =
srcpix + MIN(height - 1, looph + 1) * srcpitch;
Comment on lines +119 to +125
Copy link
Member

@Starbuck5 Starbuck5 May 28, 2024 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

So in your PR description you said you had optimized using pointers and using functions instead of macros, but reading this code it seems like the main optimization is actually some manual Loop-invariant code motion.

Very clever! This reminds me of your Rect union PR, also a great way of looking at existing code and making it faster in a straightforward way.

It’s a shame the compiler doesn’t do that automatically in these cases.


for (loopw = 0; loopw < width; ++loopw) {
B = *(Uint16 *)(srcpix + (MAX(0, looph - 1) * srcpitch) +
(2 * loopw));
D = *(Uint16 *)(srcpix + (looph * srcpitch) +
(2 * MAX(0, loopw - 1)));
E = *(Uint16 *)(srcpix + (looph * srcpitch) + (2 * loopw));
F = *(Uint16 *)(srcpix + (looph * srcpitch) +
(2 * MIN(width - 1, loopw + 1)));
H = *(Uint16 *)(srcpix +
(MIN(height - 1, looph + 1) * srcpitch) +
(2 * loopw));

E0 = D == B && B != F && D != H ? D : E;
E1 = B == F && B != D && F != H ? F : E;
E2 = D == H && D != B && H != F ? D : E;
E3 = H == F && D != H && B != F ? F : E;

*(Uint16 *)(dstpix + looph * 2 * dstpitch +
loopw * 2 * 2) = E0;
*(Uint16 *)(dstpix + looph * 2 * dstpitch +
(loopw * 2 + 1) * 2) = E1;
*(Uint16 *)(dstpix + (looph * 2 + 1) * dstpitch +
loopw * 2 * 2) = E2;
*(Uint16 *)(dstpix + (looph * 2 + 1) * dstpitch +
(loopw * 2 + 1) * 2) = E3;
B = *(Uint16 *)(src_row_prev + 2 * loopw);
D = *(Uint16 *)(src_row + 2 * MAX(0, loopw - 1));
E = *(Uint16 *)(src_row + 2 * loopw);
F = *(Uint16 *)(src_row + 2 * MIN(width - 1, loopw + 1));
H = *(Uint16 *)(src_row_next + 2 * loopw);

if (B != H && D != F) {
E0 = (D == B) ? D : E;
E1 = (B == F) ? F : E;
E2 = (D == H) ? D : E;
E3 = (H == F) ? F : E;
}
else {
E0 = E;
E1 = E;
E2 = E;
E3 = E;
}

*(Uint16 *)(dst_row0 + loopw * 2 * 2) = E0;
*(Uint16 *)(dst_row0 + (loopw * 2 + 1) * 2) = E1;
*(Uint16 *)(dst_row1 + loopw * 2 * 2) = E2;
*(Uint16 *)(dst_row1 + (loopw * 2 + 1) * 2) = E3;
}
}
break;
}
case 3: {
int E0, E1, E2, E3, B, D, E, F, H;
for (looph = 0; looph < height; ++looph) {
Uint8 *src_row = srcpix + looph * srcpitch;
Uint8 *dst_row0 = dstpix + looph * 2 * dstpitch;
Uint8 *dst_row1 = dstpix + (looph * 2 + 1) * dstpitch;

Uint8 *src_row_prev = srcpix + MAX(0, looph - 1) * srcpitch;
Uint8 *src_row_next =
srcpix + MIN(height - 1, looph + 1) * srcpitch;

for (loopw = 0; loopw < width; ++loopw) {
B = READINT24(srcpix + (MAX(0, looph - 1) * srcpitch) +
(3 * loopw));
D = READINT24(srcpix + (looph * srcpitch) +
(3 * MAX(0, loopw - 1)));
E = READINT24(srcpix + (looph * srcpitch) + (3 * loopw));
F = READINT24(srcpix + (looph * srcpitch) +
(3 * MIN(width - 1, loopw + 1)));
H = READINT24(srcpix +
(MIN(height - 1, looph + 1) * srcpitch) +
(3 * loopw));

E0 = D == B && B != F && D != H ? D : E;
E1 = B == F && B != D && F != H ? F : E;
E2 = D == H && D != B && H != F ? D : E;
E3 = H == F && D != H && B != F ? F : E;

WRITEINT24((dstpix + looph * 2 * dstpitch + loopw * 2 * 3),
E0);
WRITEINT24(
(dstpix + looph * 2 * dstpitch + (loopw * 2 + 1) * 3),
E1);
WRITEINT24(
(dstpix + (looph * 2 + 1) * dstpitch + loopw * 2 * 3),
E2);
WRITEINT24((dstpix + (looph * 2 + 1) * dstpitch +
(loopw * 2 + 1) * 3),
E3);
B = read_int24(src_row_prev + (3 * loopw));
D = read_int24(src_row + (3 * MAX(0, loopw - 1)));
E = read_int24(src_row + (3 * loopw));
F = read_int24(src_row + (3 * MIN(width - 1, loopw + 1)));
H = read_int24(src_row_next + (3 * loopw));

if (B != H && D != F) {
E0 = (D == B) ? D : E;
E1 = (B == F) ? F : E;
E2 = (D == H) ? D : E;
E3 = (H == F) ? F : E;
}
else {
E0 = E;
E1 = E;
E2 = E;
E3 = E;
}

store_int24(dst_row0 + loopw * 2 * 3, E0);
store_int24(dst_row0 + (loopw * 2 + 1) * 3, E1);
store_int24(dst_row1 + loopw * 2 * 3, E2);
store_int24(dst_row1 + (loopw * 2 + 1) * 3, E3);
}
}
break;
}
default: { /*case 4:*/
default: {
Uint32 E0, E1, E2, E3, B, D, E, F, H;

for (looph = 0; looph < height; ++looph) {
Uint8 *src_row = srcpix + looph * srcpitch;
Uint8 *dst_row0 = dstpix + looph * 2 * dstpitch;
Uint8 *dst_row1 = dstpix + (looph * 2 + 1) * dstpitch;

Uint8 *src_row_prev = srcpix + MAX(0, looph - 1) * srcpitch;
Uint8 *src_row_next =
srcpix + MIN(height - 1, looph + 1) * srcpitch;

for (loopw = 0; loopw < width; ++loopw) {
B = *(Uint32 *)(srcpix + (MAX(0, looph - 1) * srcpitch) +
(4 * loopw));
D = *(Uint32 *)(srcpix + (looph * srcpitch) +
(4 * MAX(0, loopw - 1)));
E = *(Uint32 *)(srcpix + (looph * srcpitch) + (4 * loopw));
F = *(Uint32 *)(srcpix + (looph * srcpitch) +
(4 * MIN(width - 1, loopw + 1)));
H = *(Uint32 *)(srcpix +
(MIN(height - 1, looph + 1) * srcpitch) +
(4 * loopw));

E0 = D == B && B != F && D != H ? D : E;
E1 = B == F && B != D && F != H ? F : E;
E2 = D == H && D != B && H != F ? D : E;
E3 = H == F && D != H && B != F ? F : E;

*(Uint32 *)(dstpix + looph * 2 * dstpitch +
loopw * 2 * 4) = E0;
*(Uint32 *)(dstpix + looph * 2 * dstpitch +
(loopw * 2 + 1) * 4) = E1;
*(Uint32 *)(dstpix + (looph * 2 + 1) * dstpitch +
loopw * 2 * 4) = E2;
*(Uint32 *)(dstpix + (looph * 2 + 1) * dstpitch +
(loopw * 2 + 1) * 4) = E3;
B = *(Uint32 *)(src_row_prev + 4 * loopw);
D = *(Uint32 *)(src_row + 4 * MAX(0, loopw - 1));
E = *(Uint32 *)(src_row + 4 * loopw);
F = *(Uint32 *)(src_row + 4 * MIN(width - 1, loopw + 1));
H = *(Uint32 *)(src_row_next + 4 * loopw);

if (B != H && D != F) {
E0 = (D == B) ? D : E;
E1 = (B == F) ? F : E;
E2 = (D == H) ? D : E;
E3 = (H == F) ? F : E;
}
else {
E0 = E;
E1 = E;
E2 = E;
E3 = E;
}

*(Uint32 *)(dst_row0 + loopw * 2 * 4) = E0;
*(Uint32 *)(dst_row0 + (loopw * 2 + 1) * 4) = E1;
*(Uint32 *)(dst_row1 + loopw * 2 * 4) = E2;
*(Uint32 *)(dst_row1 + (loopw * 2 + 1) * 4) = E3;
}
}
break;
Expand Down
Loading