zdoom/gzdoom-gles
0
0
Fork 0
mirror of https://github.com/ZDoom/gzdoom-gles.git synced 2025-01-31 04:20:34 +00:00
Code Issues Projects Releases Packages Wiki Activity

- Restored DoBlending_MMX for non SSE2 cpus.

Browse source
This commit is contained in:
drfrag666 2018-04-28 20:36:54 +02:00
parent 0c38c72ad7
commit a695491a1f
2 changed files with 75 additions and 8 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

37
src/v_palette.cpp
View file

@ -361,6 +361,7 @@ void InitPalette ()
R_InitColormaps ();
}
void DoBlending_MMX (const PalEntry *from, PalEntry *to, int count, int r, int g, int b, int a);
void DoBlending_SSE2 (const PalEntry *from, PalEntry *to, int count, int r, int g, int b, int a);
void DoBlending (const PalEntry *from, PalEntry *to, int count, int r, int g, int b, int a)
@ -385,17 +386,37 @@ void DoBlending (const PalEntry *from, PalEntry *to, int count, int r, int g, in
return;
}
#if defined(_M_X64) || defined(_M_IX86) || defined(__i386__) || defined(__amd64__)
else if (count >= 4)
else if (CPU.bSSE2)
{
int not3count = count & ~3;
DoBlending_SSE2 (from, to, not3count, r, g, b, a);
count &= 3;
if (count <= 0)
if (count >= 4)
{
return;
int not3count = count & ~3;
DoBlending_SSE2 (from, to, not3count, r, g, b, a);
count &= 3;
if (count <= 0)
{
return;
}
from += not3count;
to += not3count;
}
}
#endif
#if defined(_M_IX86) || defined(__i386__)
else if (CPU.bMMX)
{
if (count >= 4)
{
int not3count = count & ~3;
DoBlending_MMX (from, to, not3count, r, g, b, a);
count &= 3;
if (count <= 0)
{
return;
}
from += not3count;
to += not3count;
}
from += not3count;
to += not3count;
}
#endif
int i, ia;

46
src/x86.cpp
View file

@ -246,6 +246,9 @@ void DumpCPUInfo(const CPUInfo *cpu)
cpu->Family, cpu->Model, cpu->Stepping);
}
Printf(" Features:");
if (cpu->bMMX) Printf(" MMX");
if (cpu->bMMXPlus) Printf(" MMX+");
if (cpu->bSSE) Printf(" SSE");
if (cpu->bSSE2) Printf(" SSE2");
if (cpu->bSSE3) Printf(" SSE3");
if (cpu->bSSSE3) Printf(" SSSE3");
@ -257,6 +260,49 @@ void DumpCPUInfo(const CPUInfo *cpu)
Printf ("\n");
}
}
#if !defined(__amd64__) && !defined(_M_X64)
void DoBlending_MMX(const PalEntry *from, PalEntry *to, int count, int r, int g, int b, int a)
{
__m64 blendcolor;
__m64 blendalpha;
__m64 zero;
__m64 blending256;
__m64 color1;
__m64 color2;
zero = _mm_setzero_si64();
#ifndef __GNUC__
blending256.m64_i64 = 0x10001000100;
#else
blending256 = (__m64)0x10001000100ll;
#endif
blendcolor = _mm_unpacklo_pi8(_m_from_int((r << 16) | (g << 8) | b), zero); // 000000RR 00GG00BB
blendalpha = _mm_unpacklo_pi8(_m_from_int((a << 16) | (a << 8) | a), zero); // 000000AA 00AA00AA
blendcolor = _mm_mullo_pi16(blendcolor, blendalpha); // premultiply blend by alpha
blendalpha = _mm_subs_pu16(blending256, blendalpha); // one minus alpha
// Do two colors per iteration: Count must be even
for (count >>= 1; count > 0; --count)
{
color1 = *(__m64 *)from; // 00r2g2b2 00r1g1b1
from += 2;
color2 = _mm_unpackhi_pi8(color1, zero); // 000000r2 00g200b2
color1 = _mm_unpacklo_pi8(color1, zero); // 000000r1 00g100b1
color1 = _mm_mullo_pi16(blendalpha, color1); // 0000r1rr g1ggb1bb
color2 = _mm_mullo_pi16(blendalpha, color2); // 0000r2rr g2ggb2bb
color1 = _mm_adds_pu16(blendcolor, color1);
color2 = _mm_adds_pu16(blendcolor, color2);
color1 = _mm_srli_pi16(color1, 8);
color2 = _mm_srli_pi16(color2, 8);
*(__m64 *)to = _mm_packs_pu16(color1, color2); // 00r2g2b2 00r1g1b1
to += 2;
}
_mm_empty();
}
#endif
void DoBlending_SSE2(const PalEntry *from, PalEntry *to, int count, int r, int g, int b, int a)