Apply the transformations of the previous commit to tslopevlin().

This doesn't give much, maybe 0.5 fps more, probably because the
loop body is pretty long.

git-svn-id: https://svn.eduke32.com/eduke32@2406 1a8010ca-5511-0410-912e-c29ae57300e0
This commit is contained in:
helixhorned 2012-03-04 20:12:46 +00:00
parent 792ac2fdf0
commit 823d37fb51

View file

@ -4069,34 +4069,43 @@ static void setupslopevlin_alsotrans(int32_t logylogx, intptr_t bufplc, int32_t
ggpal = palookup[globalpal] + (getpalookup(0,globalshade)<<8); ggpal = palookup[globalpal] + (getpalookup(0,globalshade)<<8);
} }
// cnt iterations
static void tslopevlin(uint8_t *p, int32_t i, const intptr_t *slopalptr, int32_t cnt, int32_t bx, int32_t by) static void tslopevlin(uint8_t *p, int32_t i, const intptr_t *slopalptr, int32_t cnt, int32_t bx, int32_t by)
{ {
int32_t bz, bzinc; const char *const buf = ggbuf;
const char *const pal = ggpal;
const char *const trans = transluc;
const int32_t bzinc = (asm1>>3), pinc = ggpinc;
const int32_t transmode = (globalorientation&128);
const int32_t xtou = globalx3, ytov = globaly3;
const int32_t logx = gglogx, logy = gglogy;
int32_t bz = asm3;
do
{
uint8_t ch;
uint32_t u, v; uint32_t u, v;
uint8_t ch;
bz = asm3; bzinc = (asm1>>3);
for (; cnt>0; cnt--)
{
i = krecipasm(bz>>6); bz += bzinc; i = krecipasm(bz>>6); bz += bzinc;
u = bx+globalx3*i; u = bx + xtou*i;
v = by+globaly3*i; v = by + ytov*i;
ch = *(uint8_t *)(slopalptr[0] + ggbuf[((u>>(32-gglogx))<<gglogy)+(v>>(32-gglogy))]); ch = *(uint8_t *)(slopalptr[0] + buf[((u>>(32-logx))<<logy)+(v>>(32-logy))]);
if (globalorientation&128) if (transmode)
{ {
if (ch != 255) *p = transluc[*p+(ggpal[ch]<<8)]; if (ch != 255) *p = trans[*p|(pal[ch]<<8)];
} }
else else
{ {
if (ch != 255) *p = transluc[(*p<<8)+ggpal[ch]]; if (ch != 255) *p = trans[(*p<<8)|pal[ch]];
} }
slopalptr--; slopalptr--;
p += ggpinc; p += pinc;
} }
while (--cnt);
} }
// //