// This texture mapper uses floating point extensively and writes 8 pixels at once, so it likely works // best on 64 bit RISC processors. // WARNING: it is not endian clean. For big endian, reverse the shift counts in the unrolled loops. I // have no means to test that, so I didn't try it. Please tell me if you get this to work on a big // endian machine. // If you're using an Alpha, use the Compaq compiler for this file for quite some fps more. // Unfortunately, it won't compile the whole source, so simply compile everything, change the // compiler to ccc, remove scanline.o and compile again. // Please send comments/suggestions to [email protected]. void c_fp_tmap_scanline_per() { ubyte *dest; uint c; int x, j, index = fx_xleft + (bytes_per_row * fx_y); double u, v, z, l, dudx, dvdx, dzdx, dldx, rec_z; u_int64_t destlong; u = f2db(fx_u); v = f2db(fx_v) * 64.0; z = f2db(fx_z); l = f2db(fx_l); dudx = f2db(fx_du_dx); dvdx = f2db(fx_dv_dx) * 64.0; dzdx = f2db(fx_dz_dx); dldx = f2db(fx_dl_dx); rec_z = 1.0 / z; // gcc 2.95.2 is won't do this optimization itself dest = (ubyte *) (write_buffer + fx_xleft + (bytes_per_row * fx_y)); x = fx_xright - fx_xleft + 1; if (!Transparency_on) { if (x >= 8) { if ((j = (size_t) dest & 7) != 0) { j = 8 - j; while (j > 0) { if (++index >= SWIDTH*SHEIGHT) return; *dest++ = gr_fade_table[((int) fabs(l)) * 256 + (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]]; l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; x--; j--; } } j = x; while (j >= 8) { destlong = (u_int64_t) gr_fade_table[((int) fabs(l)) * 256 + (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]]; l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; destlong |= (u_int64_t) gr_fade_table[((int) fabs(l)) * 256 + (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]] << 8; l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; destlong |= (u_int64_t) gr_fade_table[((int) fabs(l)) * 256 + (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]] << 16; l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; destlong |= (u_int64_t) gr_fade_table[((int) fabs(l)) * 256 + (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]] << 24; l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; destlong |= (u_int64_t) gr_fade_table[((int) fabs(l)) * 256 + (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]] << 32; l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; destlong |= (u_int64_t) gr_fade_table[((int) fabs(l)) * 256 + (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]] << 40; l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; destlong |= (u_int64_t) gr_fade_table[((int) fabs(l)) * 256 + (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]] << 48; l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; destlong |= (u_int64_t) gr_fade_table[((int) fabs(l)) * 256 + (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]] << 56; l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; *((u_int64_t *) dest) = destlong; dest += 8; x -= 8; j -= 8; index += 8; if (index+8 >= SWIDTH*SHEIGHT) return; } } while (x-- > 0) { if (++index >= SWIDTH*SHEIGHT) return; *dest++ = gr_fade_table[((int) fabs(l)) * 256 + (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]]; l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; } } else { if (x >= 8) { if ((j = (size_t) dest & 7) != 0) { j = 8 - j; while (j > 0) { if (++index >= SWIDTH*SHEIGHT) return; c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]; if (c != 255) *dest = gr_fade_table[((int) fabs(l)) * 256 + c]; dest++; l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; x--; j--; } } j = x; while (j >= 8) { destlong = *((u_int64_t *) dest); c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]; if (c != 255) { destlong &= ~(u_int64_t)0xFF; destlong |= (u_int64_t) gr_fade_table[((int) fabs(l)) * 256 + c]; } l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]; if (c != 255) { destlong &= ~((u_int64_t)0xFF << 8); destlong |= (u_int64_t) gr_fade_table[((int) fabs(l)) * 256 + c] << 8; } l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]; if (c != 255) { destlong &= ~((u_int64_t)0xFF << 16); destlong |= (u_int64_t) gr_fade_table[((int) fabs(l)) * 256 + c] << 16; } l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]; if (c != 255) { destlong &= ~((u_int64_t)0xFF << 24); destlong |= (u_int64_t) gr_fade_table[((int) fabs(l)) * 256 + c] << 24; } l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]; if (c != 255) { destlong &= ~((u_int64_t)0xFF << 32); destlong |= (u_int64_t) gr_fade_table[((int) fabs(l)) * 256 + c] << 32; } l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]; if (c != 255) { destlong &= ~((u_int64_t)0xFF << 40); destlong |= (u_int64_t) gr_fade_table[((int) fabs(l)) * 256 + c] << 40; } l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]; if (c != 255) { destlong &= ~((u_int64_t)0xFF << 48); destlong |= (u_int64_t) gr_fade_table[((int) fabs(l)) * 256 + c] << 48; } l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]; if (c != 255) { destlong &= ~((u_int64_t)0xFF << 56); destlong |= (u_int64_t) gr_fade_table[((int) fabs(l)) * 256 + c] << 56; } l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; *((u_int64_t *) dest) = destlong; dest += 8; x -= 8; j -= 8; index += 8; if (index+8 >= SWIDTH*SHEIGHT) return; } } while (x-- > 0) { if (++index >= SWIDTH*SHEIGHT) return; c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]; if (c != 255) *dest = gr_fade_table[((int) fabs(l)) * 256 + c]; dest++; l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; } } }
// Used for energy centers. See comments for c_tmap_scanline_per().
void c_fp_tmap_scanline_per_nolight()
{
ubyte *dest;
uint c;
int x, j, index = fx_xleft + (bytes_per_row * fx_y);
double u, v, z, dudx, dvdx, dzdx, rec_z;
u_int64_t destlong;
u = f2db(fx_u);
v = f2db(fx_v) * 64.0;
z = f2db(fx_z);
dudx = f2db(fx_du_dx);
dvdx = f2db(fx_dv_dx) * 64.0;
dzdx = f2db(fx_dz_dx);
rec_z = 1.0 / z;
dest = (ubyte *) (write_buffer + fx_xleft + (bytes_per_row * fx_y));
x = fx_xright - fx_xleft + 1;
if (!Transparency_on) {
if (x >= 8) {
if ((j = (size_t) dest & 7) != 0) {
j = 8 - j;
while (j > 0) {
if (++index >= SWIDTH*SHEIGHT) return;
*dest++ =
(uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)];
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
x--;
j--;
}
}
while (j >= 8) {
destlong =
(u_int64_t) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)];
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
destlong |=
(u_int64_t) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)] << 8;
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
destlong |=
(u_int64_t) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)] << 16;
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
destlong |=
(u_int64_t) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)] << 24;
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
destlong |=
(u_int64_t) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)] << 32;
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
destlong |=
(u_int64_t) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)] << 40;
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
destlong |=
(u_int64_t) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)] << 48;
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
destlong |=
(u_int64_t) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)] << 56;
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
*((u_int64_t *) dest) = destlong;
dest += 8;
x -= 8;
j -= 8;
index +=8;
if (index+8 >= SWIDTH*SHEIGHT) return;
}
}
while (x-- > 0) {
if (++index >= SWIDTH*SHEIGHT) return;
*dest++ =
(u_int64_t) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)];
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
}
} else {
x = fx_xright - fx_xleft + 1;
if (x >= 8) {
if ((j = (size_t) dest & 7) != 0) {
j = 8 - j;
while (j > 0) {
if (++index >= SWIDTH*SHEIGHT) return;
c =
(uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)];
if (c != 255)
*dest = c;
dest++;
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
x--;
j--;
}
}
j = x;
while (j >= 8) {
destlong = *((u_int64_t *) dest);
c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)];
if (c != 255) {
destlong &= ~(u_int64_t)0xFF;
destlong |= (u_int64_t) c;
}
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)];
if (c != 255) {
destlong &= ~((u_int64_t)0xFF << 8);
destlong |= (u_int64_t) c << 8;
}
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)];
if (c != 255) {
destlong &= ~((u_int64_t)0xFF << 16);
destlong |= (u_int64_t) c << 16;
}
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)];
if (c != 255) {
destlong &= ~((u_int64_t)0xFF << 24);
destlong |= (u_int64_t) c << 24;
}
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)];
if (c != 255) {
destlong &= ~((u_int64_t)0xFF << 32);
destlong |= (u_int64_t) c << 32;
}
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)];
if (c != 255) {
destlong &= ~((u_int64_t)0xFF << 40);
destlong |= (u_int64_t) c << 40;
}
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)];
if (c != 255) {
destlong &= ~((u_int64_t)0xFF << 48);
destlong |= (u_int64_t) c << 48;
}
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)];
if (c != 255) {
destlong &= ~((u_int64_t)0xFF << 56);
destlong |= (u_int64_t) c << 56;
}
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
*((u_int64_t *) dest) = destlong;
dest += 8;
x -= 8;
j -= 8;
index += 8;
if (index+8 >= SWIDTH*SHEIGHT) return;
}
}
while (x-- > 0) {
if (++index >= SWIDTH*SHEIGHT) return;
c = (uint) pixptr[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)];
if (c != 255)
*dest = c;
dest++;
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
}
}
}
// This texture mapper uses floating point extensively and writes 8 pixels at once, so it likely works // best on 64 bit RISC processors. // WARNING: it is not endian clean. For big endian, reverse the shift counts in the unrolled loops. I // have no means to test that, so I didn't try it. Please tell me if you get this to work on a big // endian machine. // If you're using an Alpha, use the Compaq compiler for this file for quite some fps more. // Unfortunately, it won't compile the whole source, so simply compile everything, change the // compiler to ccc, remove scanline.o and compile again. // Please send comments/suggestions to [email protected]. void c_fp_tmap_scanline_per() { ubyte *dest; ubyte c; int x; double u, v, z, dudx, dvdx, dzdx, rec_z; double ubyz, vbyz, ubyz0, vbyz0, ubyz8, vbyz8, du1, dv1; double dudx8, dvdx8, dzdx8; fix l, dldx; u_int64_t destlong;//, destmask; // give dumb compilers a chance to put these global pointers into registers or at least have // nicer names :) ubyte *texmap = pixptr, *fadetable = gr_fade_table; #ifdef CYCLECOUNT unsigned long start, stop, time; static unsigned long sum, count; #endif // v is pre-scaled by 64 to avoid the multiplication when accessing the 64x64 texture array u = f2db(fx_u); v = f2db(fx_v) * 64.0; z = f2db(fx_z); l = fx_l >> 8; dudx = f2db(fx_du_dx); dvdx = f2db(fx_dv_dx) * 64.0; dzdx = f2db(fx_dz_dx); dldx = fx_dl_dx >> 8; dudx8 = dudx * 8.0; dvdx8 = dvdx * 8.0; dzdx8 = dzdx * 8.0; rec_z = 1.0 / z; // multiplication is often faster than division dest = (ubyte *) (write_buffer + fx_xleft + (bytes_per_row * fx_y)); x = fx_xright - fx_xleft + 1; if (!Transparency_on) { if (x >= 8) { // draw till we are on a 8-byte aligned address for ( ; (size_t) dest & 7; --x) { *dest++ = fadetable[(l & 0x7f00) + (uint) texmap[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]]; l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; } // Now draw 8 pixels at once, interpolating 1/z linearly. Artifacts of the // interpolation aren't really noticeable; many games even interpolate over 16 // pixels. // We do these calculations once before and then at the end of the loop instead // of simply at the start of the loop, because he scheduler can then interleave // them with the texture accesses. Silly, but gains a few fps. ubyz0 = u * rec_z; vbyz0 = v * rec_z; u += dudx8; v += dvdx8; z += dzdx8; rec_z = 1.0 / z; ubyz8 = u * rec_z; vbyz8 = v * rec_z; du1 = (ubyz8 - ubyz0) / 8.0; dv1 = (vbyz8 - vbyz0) / 8.0; ubyz = ubyz0; vbyz = vbyz0; // This loop is the "hot spot" of the game; it takes about 70% of the time. The // major weak point are the many integer casts, which have to go through memory // on processors < 21264. But when using integers, one needs to compensate for // inexactness, and the code ends up being not really faster. for ( ; x >= 8; x -= 8) { #ifdef CYCLECOUNT start = virtcc(); #endif destlong = (u_int64_t) fadetable[(l & 0x7f00) + (uint) texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)]]; l += dldx; ubyz += du1; vbyz += dv1; destlong |= (u_int64_t) fadetable[(l & 0x7f00) + (uint) texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)]] << 8; l += dldx; ubyz += du1; vbyz += dv1; destlong |= (u_int64_t) fadetable[(l & 0x7f00) + (uint) texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)]] << 16; l += dldx; ubyz += du1; vbyz += dv1; destlong |= (u_int64_t) fadetable[(l & 0x7f00) + (uint) texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)]] << 24; l += dldx; ubyz += du1; vbyz += dv1; destlong |= (u_int64_t) fadetable[(l & 0x7f00) + (uint) texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)]] << 32; l += dldx; ubyz += du1; vbyz += dv1; destlong |= (u_int64_t) fadetable[(l & 0x7f00) + (uint) texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)]] << 40; l += dldx; ubyz += du1; vbyz += dv1; destlong |= (u_int64_t) fadetable[(l & 0x7f00) + (uint) texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)]] << 48; l += dldx; ubyz += du1; vbyz += dv1; destlong |= (u_int64_t) fadetable[(l & 0x7f00) + (uint) texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)]] << 56; l += dldx; ubyz0 = ubyz8; vbyz0 = vbyz8; u += dudx8; v += dvdx8; z += dzdx8; rec_z = 1.0 / z; ubyz8 = u * rec_z; vbyz8 = v * rec_z; du1 = (ubyz8 - ubyz0) / 8.0; dv1 = (vbyz8 - vbyz0) / 8.0; ubyz = ubyz0; vbyz = vbyz0; *((u_int64_t *) dest) = destlong; dest += 8; #ifdef CYCLECOUNT stop = virtcc(); #endif } // compensate for being calculated once too often u -= dudx8; v -= dvdx8; z -= dzdx8; #ifdef CYCLECOUNT time = stop - start; if (time > 10 && time < 900) { sum += time; ++count; if (count % 10000 == 1) printf("%f %d\n", (double) sum / (double) count, time); } #endif } // Draw the last few (<8) pixels. rec_z = 1.0 / z; for ( ; x > 0; x--) { *dest++ = fadetable[(l & 0x7f00) + (uint) texmap[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]]; l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; } } else { // Transparency_on if (x >= 8) { for ( ; (size_t) dest & 7; --x) { c = (uint) texmap[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]; if (c != TRANSPARENCY_COLOR) *dest = fadetable[(l & 0x7f00) + c]; dest++; l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; } ubyz0 = u * rec_z; vbyz0 = v * rec_z; u += dudx8; v += dvdx8; z += dzdx8; rec_z = 1.0 / z; ubyz8 = u * rec_z; vbyz8 = v * rec_z; du1 = (ubyz8 - ubyz0) / 8.0; dv1 = (vbyz8 - vbyz0) / 8.0; ubyz = ubyz0; vbyz = vbyz0; for ( ; x >= 8; x -= 8) { destlong = *((u_int64_t *) dest); c = texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)]; if (c != TRANSPARENCY_COLOR) { destlong &= ~((u_int64_t) 0xFF); destlong |= (u_int64_t) fadetable[(l & 0x7f00) + c]; } l += dldx; ubyz += du1; vbyz += dv1; c = texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)]; if (c != TRANSPARENCY_COLOR) { destlong &= ~((u_int64_t) 0xFF << 8); destlong |= (u_int64_t) fadetable[(l & 0x7f00) + c] << 8; } l += dldx; ubyz += du1; vbyz += dv1; c = texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)]; if (c != TRANSPARENCY_COLOR) { destlong &= ~((u_int64_t) 0xFF << 16); destlong |= (u_int64_t) fadetable[(l & 0x7f00) + c] << 16; } l += dldx; ubyz += du1; vbyz += dv1; c = texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)]; if (c != TRANSPARENCY_COLOR) { destlong &= ~((u_int64_t) 0xFF << 24); destlong |= (u_int64_t) fadetable[(l & 0x7f00) + c] << 24; } l += dldx; ubyz += du1; vbyz += dv1; c = texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)]; if (c != TRANSPARENCY_COLOR) { destlong &= ~((u_int64_t) 0xFF << 32); destlong |= (u_int64_t) fadetable[(l & 0x7f00) + c] << 32; } l += dldx; ubyz += du1; vbyz += dv1; c = texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)]; if (c != TRANSPARENCY_COLOR) { destlong &= ~((u_int64_t) 0xFF << 40); destlong |= (u_int64_t) fadetable[(l & 0x7f00) + c] << 40; } l += dldx; ubyz += du1; vbyz += dv1; c = texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)]; if (c != TRANSPARENCY_COLOR) { destlong &= ~((u_int64_t) 0xFF << 48); destlong |= (u_int64_t) fadetable[(l & 0x7f00) + c] << 48; } l += dldx; ubyz += du1; vbyz += dv1; c = texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)]; if (c != TRANSPARENCY_COLOR) { destlong &= ~((u_int64_t) 0xFF << 56); destlong |= (u_int64_t) fadetable[(l & 0x7f00) + c] << 56; } l += dldx; *((u_int64_t *) dest) = destlong; dest += 8; ubyz0 = ubyz8; vbyz0 = vbyz8; u += dudx8; v += dvdx8; z += dzdx8; rec_z = 1.0 / z; ubyz8 = u * rec_z; vbyz8 = v * rec_z; du1 = (ubyz8 - ubyz0) / 8.0; dv1 = (vbyz8 - vbyz0) / 8.0; ubyz = ubyz0; vbyz = vbyz0; } u -= dudx8; v -= dvdx8; z -= dzdx8; } rec_z = 1.0 / z; for ( ; x > 0; x--) { c = (uint) texmap[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)]; if (c != TRANSPARENCY_COLOR) *dest = fadetable[(l & 0x7f00) + c]; dest++; l += dldx; u += dudx; v += dvdx; z += dzdx; rec_z = 1.0 / z; } } }
void ComputeLightMaps (int segNum)
{
tSegment *segP;
tSide *sideP;
tLightMap *lmapP;
int sideNum, lastSeg, mapNum;
short sideVerts [4];
#if 1
# define Xs 8
# define Ys 8
#else
int Xs = 8, Ys = 8;
#endif
int x, y, xy;
int v0, v1, v2, v3;
GLfloat *pTexColor;// = {0.0, 0.0, 0.0, 1.0};
GLfloat texColor [Xs][Ys][4];
#if 1
# define pixelOffset 0.0
#else
double pixelOffset = 0; //0.5
#endif
int l, s, nMethod, sideRad;
GLfloat tempBright = 0;
vmsVector OffsetU, OffsetV, pixelPos [Xs][Ys], *pPixelPos, rayVec, faceNorm, sidePos;
double brightPrct, pixelDist;
double delta;
double f_offset [8] = {
0.0 / (Xs - 1), 1.0 / (Xs - 1), 2.0 / (Xs - 1), 3.0 / (Xs - 1),
4.0 / (Xs - 1), 5.0 / (Xs - 1), 6.0 / (Xs - 1), 7.0 / (Xs - 1)
};
#if LMAP_REND2TEX
ubyte brightMap [512];
ubyte lightMap [512*3];
tUVL lMapUVL [4];
fix nDist, nMinDist;
GLuint lightMapId;
int bStart;
#endif
if (segNum <= 0) {
DestroyLightMaps ();
if (!InitLightData ())
return;
#if LMAP_REND2TEX
InitBrightMap (brightMap);
memset (&lMapUVL, 0, sizeof (lMapUVL));
#endif
}
INIT_PROGRESS_LOOP (segNum, lastSeg, gameData.segs.nSegments);
//Next Go through each surface and create a lightmap for it.
for (mapNum = 6 * segNum, segP = gameData.segs.segments + segNum;
segNum < lastSeg;
segNum++, segP++) {
for (sideNum = 0, sideP = segP->sides; sideNum < 6; sideNum++, mapNum++, sideP++) {
#if TEXTURE_CHECK
if ((segP->children [sideNum] >= 0) && !IS_WALL (WallNumS (sideP)))
continue; //skip open sides
#endif
GetSideVerts (sideVerts, segNum, sideNum);
#if LMAP_REND2TEX
OglCreateFBuffer (&lightMaps [mapNum].fbuffer, 64, 64);
OglEnableFBuffer (&lightMaps [mapNum].fbuffer);
#else
lightMaps [mapNum].handle = EmptyTexture (Xs, Ys);
OGL_BINDTEX (lightMaps [mapNum].handle);
glTexParameteri (GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP);
glTexParameteri (GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP);
nMethod = (sideP->nType == SIDE_IS_QUAD) || (sideP->nType == SIDE_IS_TRI_02);
pPixelPos = &pixelPos [0][0];
for (x = 0; x < Xs; x++) {
for (y = 0; y < Ys; y++, pPixelPos++) {
if (nMethod) {
v0 = sideVerts [0];
v2 = sideVerts [2];
if (x >= y) {
v1 = sideVerts [1];
//Next calculate this pixel's place in the world (tricky stuff)
FindOffset (&OffsetU, gameData.segs.vertices [v0], gameData.segs.vertices [v1], f_offset [x]); //(((double) x) + pixelOffset) / (Xs - 1)); //took me forever to figure out this should be an inverse thingy
FindOffset (&OffsetV, gameData.segs.vertices [v1], gameData.segs.vertices [v2], f_offset [y]); //(((double) y) + pixelOffset) / (Ys - 1));
VmVecAdd (pPixelPos, &OffsetU, &OffsetV);
VmVecInc (pPixelPos, gameData.segs.vertices + v0); //This should be the real world position of the pixel.
//Find Normal
VmVecNormal (&faceNorm, gameData.segs.vertices + v0, gameData.segs.vertices + v2, gameData.segs.vertices + v1);
}
else {
//Next calculate this pixel's place in the world (tricky stuff)
v3 = sideVerts [3];
FindOffset (&OffsetV, gameData.segs.vertices [v0], gameData.segs.vertices [v3], f_offset [y]); //(((double) y) + pixelOffset) / (Xs - 1)); //Notice y/x and OffsetU/OffsetV are swapped from above
FindOffset (&OffsetU, gameData.segs.vertices [v3], gameData.segs.vertices [v2], f_offset [x]); //(((double) x) + pixelOffset) / (Ys - 1));
VmVecAdd (pPixelPos, &OffsetU, &OffsetV);
VmVecInc (pPixelPos, gameData.segs.vertices + v0); //This should be the real world position of the pixel.
VmVecNormal (&faceNorm, gameData.segs.vertices + v0, gameData.segs.vertices + v3, gameData.segs.vertices + v2);
}
}
else {//SIDE_IS_TRI_02
v1 = sideVerts [1];
v3 = sideVerts [3];
if (Xs - x >= y) {
v0 = sideVerts [0];
FindOffset (&OffsetU, gameData.segs.vertices [v0], gameData.segs.vertices [v1], f_offset [x]); //(((double) x) + pixelOffset) / (Xs - 1));
FindOffset (&OffsetV, gameData.segs.vertices [v0], gameData.segs.vertices [v3], f_offset [y]); //(((double) y) + pixelOffset) / (Xs - 1));
VmVecAdd (pPixelPos, &OffsetU, &OffsetV);
VmVecInc (pPixelPos, gameData.segs.vertices + v0); //This should be the real world position of the pixel.
}
else {
v2 = sideVerts [2];
//Not certain this is correct, may need to subtract something
FindOffset (&OffsetV, gameData.segs.vertices [v2], gameData.segs.vertices [v1], f_offset [Xs - 1 - y]); //((double) ((Xs - 1) - y) + pixelOffset) / (Xs - 1));
FindOffset (&OffsetU, gameData.segs.vertices [v2], gameData.segs.vertices [v3], f_offset [Xs - 1 - x]); //((double) ((Xs - 1) - x) + pixelOffset) / (Xs - 1));
VmVecAdd (pPixelPos, &OffsetU, &OffsetV);
VmVecInc (pPixelPos, gameData.segs.vertices + v2); //This should be the real world position of the pixel.
}
}
}
}
#endif
//Calculate LightVal
//Next iterate through all the lights and add the light to the pixel every iteration.
sideRad = (int) (SideRad (segNum, sideNum) + 0.5);
VmVecAvg4 (
&sidePos,
&pixelPos [0][0],
&pixelPos [Xs-1][0],
&pixelPos [Xs-1][Ys-1],
&pixelPos [0][Ys-1]);
#if 1
pTexColor = texColor [0][0] + 3;
memset (texColor, 0, sizeof (texColor));
for (xy = Xs * Ys; xy; xy--, pTexColor += 4)
*pTexColor = 1;
#else
pTexColor = texColor [0][0];
for (x = 0; x < Xs; x++) {
for (y = 0; y < Ys; y++, pTexColor += 4) {
pTexColor [0] =
pTexColor [1] =
pTexColor [2] = 0;
pTexColor [3] = 1;
}
}
#endif
#if LMAP_REND2TEX
bStart = 1;
#endif
for (l = 0, lmapP = lightData; l < numLightMaps; l++, lmapP++) {
#if LMAP_REND2TEX
nMinDist = 0x7FFFFFFF;
// get the distances of all 4 tSide corners to the light source center
// scaled by the light source range
for (i = 0; i < 4; i++) {
int svi = sideVerts [i];
sidePos.x = gameData.segs.vertices [svi].x;
sidePos.y = gameData.segs.vertices [svi].y;
sidePos.z = gameData.segs.vertices [svi].z;
nDist = f2i (VmVecDist (&sidePos, &lmapP->pos)); // calc distance
if (nMinDist > nDist)
nMinDist = nDist;
lMapUVL [i].u = F1_0 * (double) nDist / (double) lmapP->range; // scale distance
}
if ((lmapP->color [0] + lmapP->color [1] + lmapP->color [2] < 3) &&
(nMinDist < lmapP->range + sideRad)) {
// create and initialize an OpenGL texture for the lightmap
InitLightMap (lightMap, brightMap, lmapP->color);
glGenTextures (1, &lightMapId);
glTexImage1D (GL_TEXTURE_1D, 0, GL_RGB, 512, 1, GL_RGB, GL_UNSIGNED_BYTE, lightMap);
OglActiveTexture (GL_TEXTURE0_ARB);
glEnable (GL_TEXTURE_1D);
glEnable (GL_BLEND);
glBlendFunc (GL_ONE, bStart ? GL_ZERO : GL_ONE);
// If processing first light, set the lightmap, else modify it
glTexEnvi (GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, bStart ? GL_REPLACE : GL_ADD);
glTexEnvi (GL_TEXTURE_ENV, GL_TEXTURE_ENV_COLOR, GL_RGBA);
glBindTexture (GL_TEXTURE_1D, lightMapId);
// extend the lightmap to the texture edges
glTexParameteri (GL_TEXTURE_1D, GL_TEXTURE_WRAP_S, GL_CLAMP);
glTexParameteri (GL_TEXTURE_1D, GL_TEXTURE_WRAP_T, GL_CLAMP);
glBegin (GL_QUADS);
glColor4f (1.0f, 1.0f, 1.0f, 1.0f);
for (i = 0; i < 4; i++) {
glMultiTexCoord2f (GL_TEXTURE0_ARB, f2fl (lMapUVL [i].u), f2fl (lMapUVL [i].v));
glVertex3f (f2fl (gameData.segs.vertices [sideVerts [i]].x),
f2fl (gameData.segs.vertices [sideVerts [i]].y),
f2fl (gameData.segs.vertices [sideVerts [i]].z));
}
glEnd ();
glDisable (GL_BLEND);
glDisable (GL_TEXTURE_1D);
glDeleteTextures (1, &lightMapId);
bStart = 0;
}
#else
if (f2i (VmVecDist (&sidePos, &lmapP->pos)) < lmapP->range + sideRad) {
pPixelPos = &pixelPos [0][0];
pTexColor = texColor [0][0];
#if 1
for (xy = Xs * Ys; xy; xy--, pPixelPos++, pTexColor += 4) {
#else
for (x = 0; x < Xs; x++)
for (y = 0; y < Ys; y++, pPixelPos++, pTexColor += 4) {
#endif
//Find angle to this light.
pixelDist = f2i (VmVecDist (pPixelPos, &lmapP->pos));
if (pixelDist >= lmapP->range)
continue;
VmVecSub (&rayVec, &lmapP->pos, pPixelPos);
delta = f2db (VmVecDeltaAng (&lmapP->dir, &rayVec, NULL));
if (delta < 0)
delta = -delta;
brightPrct = 1 - (pixelDist / lmapP->range);
brightPrct *= brightPrct; //square result
if (delta < 0.245)
brightPrct /= 4;
pTexColor [0] += (GLfloat) (brightPrct * lmapP->color [0]);
pTexColor [1] += (GLfloat) (brightPrct * lmapP->color [1]);
pTexColor [2] += (GLfloat) (brightPrct * lmapP->color [2]);
}
}
#endif
}
#if LMAP_REND2TEX
lightMaps [mapNum].handle = lightMaps [mapNum].fbuffer.texId;
lightMaps [mapNum].fbuffer.texId = 0;
OglDestroyFBuffer (&lightMaps [mapNum].fbuffer);
#else
pPixelPos = &pixelPos [0][0];
pTexColor = texColor [0][0];
for (x = 0; x < Xs; x++)
for (y = 0; y < Ys; y++, pPixelPos++, pTexColor += 4) {
tempBright = 0;
for (s = 0; s < 3; s++)
if (pTexColor [s] > tempBright)
tempBright = pTexColor [s];
if (tempBright > 1.0)
for (s = 0; s < 3; s++)
pTexColor [s] /= tempBright;
glTexSubImage2D (GL_TEXTURE_2D, 0, x, y, 1, 1, GL_RGBA, GL_FLOAT, pTexColor);
}
#endif
}
}
}
//------------------------------------------------------------------------------
int HaveLightMaps (void)
{
return (lightData != NULL);
}
//------------------------------------------------------------------------------
static int segNum = 0;
static void CreateLightMapsPoll (int nItems, tMenuItem *m, int *key, int cItem)
{
GrPaletteStepLoad (NULL);
if (segNum < gameData.segs.nSegments) {
ComputeLightMaps (segNum);
segNum += PROGRESS_INCR;
}
else {
*key = -2;
GrPaletteStepLoad (NULL);
return;
}
m [0].value++;
m [0].rebuild = 1;
*key = 0;
GrPaletteStepLoad (NULL);
return;
}
// Used for energy centers. See comments for c_tmap_scanline_per().
void c_fp_tmap_scanline_per_nolight()
{
ubyte *dest;
ubyte c;
int x;
double u, v, z, dudx, dvdx, dzdx, rec_z;
double ubyz, vbyz, ubyz0, vbyz0, ubyz8, vbyz8, du1, dv1;
double dudx8, dvdx8, dzdx8;
u_int64_t destlong;//, destmask;
ubyte *texmap = pixptr;//, *fadetable = gr_fade_table;
u = f2db(fx_u);
v = f2db(fx_v) * 64.0;
z = f2db(fx_z);
dudx = f2db(fx_du_dx);
dvdx = f2db(fx_dv_dx) * 64.0;
dzdx = f2db(fx_dz_dx);
dudx8 = dudx * 8.0;
dvdx8 = dvdx * 8.0;
dzdx8 = dzdx * 8.0;
rec_z = 1.0 / z;
dest = (ubyte *) (write_buffer + fx_xleft + (bytes_per_row * fx_y));
x = fx_xright - fx_xleft + 1;
if (!Transparency_on) { // I'm not sure this is ever used (energy texture is transparent)
if (x >= 8) {
for ( ; (size_t) dest & 7; --x) {
*dest++ = (uint) texmap[(((int) (v * rec_z)) & (64 * 63)) +
(((int) (u * rec_z)) & 63)];
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
}
ubyz0 = u * rec_z;
vbyz0 = v * rec_z;
u += dudx8;
v += dvdx8;
z += dzdx8;
rec_z = 1.0 / z;
ubyz8 = u * rec_z;
vbyz8 = v * rec_z;
du1 = (ubyz8 - ubyz0) / 8.0;
dv1 = (vbyz8 - vbyz0) / 8.0;
ubyz = ubyz0;
vbyz = vbyz0;
for ( ; x >= 8; x -= 8) {
destlong = (u_int64_t) texmap[(((int) vbyz) & (64 * 63)) +
(((int) ubyz) & 63)];
ubyz += du1;
vbyz += dv1;
destlong |= (u_int64_t) texmap[(((int) vbyz) & (64 * 63)) +
(((int) ubyz) & 63)] << 8;
ubyz += du1;
vbyz += dv1;
destlong |= (u_int64_t) texmap[(((int) vbyz) & (64 * 63)) +
(((int) ubyz) & 63)] << 16;
ubyz += du1;
vbyz += dv1;
destlong |= (u_int64_t) texmap[(((int) vbyz) & (64 * 63)) +
(((int) ubyz) & 63)] << 24;
ubyz += du1;
vbyz += dv1;
destlong |= (u_int64_t) texmap[(((int) vbyz) & (64 * 63)) +
(((int) ubyz) & 63)] << 32;
ubyz += du1;
vbyz += dv1;
destlong |= (u_int64_t) texmap[(((int) vbyz) & (64 * 63)) +
(((int) ubyz) & 63)] << 40;
ubyz += du1;
vbyz += dv1;
destlong |= (u_int64_t) texmap[(((int) vbyz) & (64 * 63)) +
(((int) ubyz) & 63)] << 48;
ubyz += du1;
vbyz += dv1;
destlong |= (u_int64_t) texmap[(((int) vbyz) & (64 * 63)) +
(((int) ubyz) & 63)] << 56;
ubyz0 = ubyz8;
vbyz0 = vbyz8;
u += dudx8;
v += dvdx8;
z += dzdx8;
rec_z = 1.0 / z;
ubyz8 = u * rec_z;
vbyz8 = v * rec_z;
du1 = (ubyz8 - ubyz0) / 8.0;
dv1 = (vbyz8 - vbyz0) / 8.0;
ubyz = ubyz0;
vbyz = vbyz0;
*((u_int64_t *) dest) = destlong;
dest += 8;
}
u -= dudx8;
v -= dvdx8;
z -= dzdx8;
}
rec_z = 1.0 / z;
for ( ; x > 0; x--) {
*dest++ = (uint) texmap[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)];
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
}
} else { // Transparency_on
if (x >= 8) {
for ( ; (size_t) dest & 7; --x) {
c = (uint) texmap[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)];
if (c != TRANSPARENCY_COLOR)
*dest = c;
dest++;
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
}
ubyz0 = u * rec_z;
vbyz0 = v * rec_z;
u += dudx8;
v += dvdx8;
z += dzdx8;
rec_z = 1.0 / z;
ubyz8 = u * rec_z;
vbyz8 = v * rec_z;
du1 = (ubyz8 - ubyz0) / 8.0;
dv1 = (vbyz8 - vbyz0) / 8.0;
ubyz = ubyz0;
vbyz = vbyz0;
for ( ; x >= 8; x -= 8) {
destlong = *((u_int64_t *) dest);
c = texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)];
if (c != TRANSPARENCY_COLOR) {
destlong &= ~((u_int64_t) 0xFF);
destlong |= (u_int64_t) c;
}
ubyz += du1;
vbyz += dv1;
c = texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)];
if (c != TRANSPARENCY_COLOR) {
destlong &= ~((u_int64_t) 0xFF << 8);
destlong |= (u_int64_t) c << 8;
}
ubyz += du1;
vbyz += dv1;
c = texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)];
if (c != TRANSPARENCY_COLOR) {
destlong &= ~((u_int64_t) 0xFF << 16);
destlong |= (u_int64_t) c << 16;
}
ubyz += du1;
vbyz += dv1;
c = texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)];
if (c != TRANSPARENCY_COLOR) {
destlong &= ~((u_int64_t) 0xFF << 24);
destlong |= (u_int64_t) c << 24;
}
ubyz += du1;
vbyz += dv1;
c = texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)];
if (c != TRANSPARENCY_COLOR) {
destlong &= ~((u_int64_t) 0xFF << 32);
destlong |= (u_int64_t) c << 32;
}
ubyz += du1;
vbyz += dv1;
c = texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)];
if (c != TRANSPARENCY_COLOR) {
destlong &= ~((u_int64_t) 0xFF << 40);
destlong |= (u_int64_t) c << 40;
}
ubyz += du1;
vbyz += dv1;
c = texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)];
if (c != TRANSPARENCY_COLOR) {
destlong &= ~((u_int64_t) 0xFF << 48);
destlong |= (u_int64_t) c << 48;
}
ubyz += du1;
vbyz += dv1;
c = texmap[(((int) vbyz) & (64 * 63)) + (((int) ubyz) & 63)];
if (c != TRANSPARENCY_COLOR) {
destlong &= ~((u_int64_t) 0xFF << 56);
destlong |= (u_int64_t) c << 56;
}
*((u_int64_t *) dest) = destlong;
dest += 8;
ubyz0 = ubyz8;
vbyz0 = vbyz8;
u += dudx8;
v += dvdx8;
z += dzdx8;
rec_z = 1.0 / z;
ubyz8 = u * rec_z;
vbyz8 = v * rec_z;
du1 = (ubyz8 - ubyz0) / 8.0;
dv1 = (vbyz8 - vbyz0) / 8.0;
ubyz = ubyz0;
vbyz = vbyz0;
}
u -= dudx8;
v -= dvdx8;
z -= dzdx8;
}
rec_z = 1.0 / z;
for ( ; x > 0; x--) {
c = (uint) texmap[(((int) (v * rec_z)) & (64 * 63)) + (((int) (u * rec_z)) & 63)];
if (c != TRANSPARENCY_COLOR)
*dest = c;
dest++;
u += dudx;
v += dvdx;
z += dzdx;
rec_z = 1.0 / z;
}
}
}
