teambeef/ioq3quest
0
0
Fork 0
mirror of https://github.com/DrBeef/ioq3quest.git synced 2024-11-22 20:11:40 +00:00
Code Issues Projects Releases Packages Wiki Activity

Isolate the Altivec code so non-Altivec PPC targets can use the same binary.

Browse source 
Moved all the code using Altivec intrinsics to separate files. This 
means we can optionally use GCC's -maltivec on just these files, which
are chosen at runtime if the CPU supports Altivec, and compile the rest
without it, making a single binary that has Altivec optimizations but
can still work on G3.

Unlike SSE and similar extensions on x86, there does not seem to be
a way to enable conditional, targeted use of Altivec based on runtime
detection (which is what ioquake3 wants to do) without also giving the
compiler permission to use Altivec in code generation; so to not crash
on CPUs that do not implement Altivec, we'll have to turn it off
altogether, except in translation units that are only entered when
runtime Altivec detection is successful.

This has been tested on Linux PPC (on an Altivec-enabled CPU),
but we may need further work after testing trickles out to other
PowerPC devices and ancient Mac OS X builds.

I did a little work on this patch, but the majority of the effort belongs 
to Simon McVittie (thanks!).
This commit is contained in:
Simon McVittie 2018-05-12 19:14:47 +01:00 committed by Ryan C. Gordon
parent 2326a060b9
commit 5909b9a1cf
12 changed files with 686 additions and 607 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

28
Makefile
View file

@ -348,11 +348,11 @@ ifneq (,$(findstring "$(PLATFORM)", "linux" "gnu_kfreebsd" "kfreebsd-gnu" "gnu")
HAVE_VM_COMPILED=true
else
ifeq ($(ARCH),ppc)
BASE_CFLAGS += -maltivec
ALTIVEC_CFLAGS = -maltivec
HAVE_VM_COMPILED=true
endif
ifeq ($(ARCH),ppc64)
BASE_CFLAGS += -maltivec
ALTIVEC_CFLAGS = -maltivec
HAVE_VM_COMPILED=true
endif
ifeq ($(ARCH),sparc)
@ -440,10 +440,12 @@ ifeq ($(PLATFORM),darwin)
-DMAC_OS_X_VERSION_MIN_REQUIRED=$(MAC_OS_X_VERSION_MIN_REQUIRED)
ifeq ($(ARCH),ppc)
BASE_CFLAGS += -arch ppc -faltivec
BASE_CFLAGS += -arch ppc
ALTIVEC_CFLAGS = -faltivec
endif
ifeq ($(ARCH),ppc64)
BASE_CFLAGS += -arch ppc64 -faltivec
BASE_CFLAGS += -arch ppc64
ALTIVEC_CFLAGS = -faltivec
endif
ifeq ($(ARCH),x86)
OPTIMIZEVM += -march=prescott -mfpmath=sse
@ -1189,11 +1191,21 @@ $(echo_cmd) "CC $<"
$(Q)$(CC) $(NOTSHLIBCFLAGS) $(CFLAGS) $(CLIENT_CFLAGS) $(OPTIMIZE) -o $@ -c $<
endef
define DO_CC_ALTIVEC
$(echo_cmd) "CC $<"
$(Q)$(CC) $(NOTSHLIBCFLAGS) $(CFLAGS) $(CLIENT_CFLAGS) $(OPTIMIZE) $(ALTIVEC_CFLAGS) -o $@ -c $<
endef
define DO_REF_CC
$(echo_cmd) "REF_CC $<"
$(Q)$(CC) $(SHLIBCFLAGS) $(CFLAGS) $(CLIENT_CFLAGS) $(OPTIMIZE) -o $@ -c $<
endef
define DO_REF_CC_ALTIVEC
$(echo_cmd) "REF_CC $<"
$(Q)$(CC) $(SHLIBCFLAGS) $(CFLAGS) $(CLIENT_CFLAGS) $(OPTIMIZE) $(ALTIVEC_CFLAGS) -o $@ -c $<
endef
define DO_REF_STR
$(echo_cmd) "REF_STR $<"
$(Q)rm -f $@
@ -1664,6 +1676,7 @@ Q3OBJ = \
$(B)/client/net_ip.o \
$(B)/client/huffman.o \
\
$(B)/client/snd_altivec.o \
$(B)/client/snd_adpcm.o \
$(B)/client/snd_dma.o \
$(B)/client/snd_mem.o \
@ -1817,6 +1830,7 @@ Q3R2STRINGOBJ = \
$(B)/renderergl2/glsl/tonemap_vp.o
Q3ROBJ = \
$(B)/renderergl1/tr_altivec.o \
$(B)/renderergl1/tr_animation.o \
$(B)/renderergl1/tr_backend.o \
$(B)/renderergl1/tr_bsp.o \
@ -2627,6 +2641,9 @@ $(B)/client/%.o: $(ASMDIR)/%.s
$(B)/client/%.o: $(ASMDIR)/%.c
$(DO_CC) -march=k8
$(B)/client/snd_altivec.o: $(CDIR)/snd_altivec.c
$(DO_CC_ALTIVEC)
$(B)/client/%.o: $(CDIR)/%.c
$(DO_CC)
@ -2691,6 +2708,9 @@ $(B)/renderergl1/%.o: $(RCOMMONDIR)/%.c
$(B)/renderergl1/%.o: $(RGL1DIR)/%.c
$(DO_REF_CC)
$(B)/renderergl1/tr_altivec.o: $(RGL1DIR)/tr_altivec.c
$(DO_REF_CC_ALTIVEC)
$(B)/renderergl2/glsl/%.c: $(RGL2DIR)/glsl/%.glsl
$(DO_REF_STR)

229
code/client/snd_altivec.c Normal file
View file

@ -0,0 +1,229 @@
/*
===========================================================================
Copyright (C) 1999-2005 Id Software, Inc.
This file is part of Quake III Arena source code.
Quake III Arena source code is free software; you can redistribute it
and/or modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the License,
or (at your option) any later version.
Quake III Arena source code is distributed in the hope that it will be
useful, but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with Quake III Arena source code; if not, write to the Free Software
Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
===========================================================================
*/
/* This file is only compiled for PowerPC builds with Altivec support.
Altivec intrinsics need to be in a separate file, so GCC's -maltivec
command line can enable them, but give us the option to _not_ use that
on other files, where the compiler might then generate Altivec
instructions for normal floating point, crashing on G3 (etc) processors. */
#include "client.h"
#include "snd_local.h"
#if idppc_altivec
#if !defined(__APPLE__)
#include <altivec.h>
#endif
void S_PaintChannelFrom16_altivec( portable_samplepair_t paintbuffer[PAINTBUFFER_SIZE], int snd_vol, channel_t *ch, const sfx_t *sc, int count, int sampleOffset, int bufferOffset ) {
int data, aoff, boff;
int leftvol, rightvol;
int i, j;
portable_samplepair_t *samp;
sndBuffer *chunk;
short *samples;
float ooff, fdata[2], fdiv, fleftvol, frightvol;
if (sc->soundChannels <= 0) {
return;
}
samp = &paintbuffer[ bufferOffset ];
if (ch->doppler) {
sampleOffset = sampleOffset*ch->oldDopplerScale;
}
if ( sc->soundChannels == 2 ) {
sampleOffset *= sc->soundChannels;
if ( sampleOffset & 1 ) {
sampleOffset &= ~1;
}
}
chunk = sc->soundData;
while (sampleOffset>=SND_CHUNK_SIZE) {
chunk = chunk->next;
sampleOffset -= SND_CHUNK_SIZE;
if (!chunk) {
chunk = sc->soundData;
}
}
if (!ch->doppler || ch->dopplerScale==1.0f) {
vector signed short volume_vec;
vector unsigned int volume_shift;
int vectorCount, samplesLeft, chunkSamplesLeft;
leftvol = ch->leftvol*snd_vol;
rightvol = ch->rightvol*snd_vol;
samples = chunk->sndChunk;
((short *)&volume_vec)[0] = leftvol;
((short *)&volume_vec)[1] = leftvol;
((short *)&volume_vec)[4] = leftvol;
((short *)&volume_vec)[5] = leftvol;
((short *)&volume_vec)[2] = rightvol;
((short *)&volume_vec)[3] = rightvol;
((short *)&volume_vec)[6] = rightvol;
((short *)&volume_vec)[7] = rightvol;
volume_shift = vec_splat_u32(8);
i = 0;
while(i < count) {
/* Try to align destination to 16-byte boundary */
while(i < count && (((unsigned long)&samp[i] & 0x1f) || ((count-i) < 8) || ((SND_CHUNK_SIZE - sampleOffset) < 8))) {
data = samples[sampleOffset++];
samp[i].left += (data * leftvol)>>8;
if ( sc->soundChannels == 2 ) {
data = samples[sampleOffset++];
}
samp[i].right += (data * rightvol)>>8;
if (sampleOffset == SND_CHUNK_SIZE) {
chunk = chunk->next;
samples = chunk->sndChunk;
sampleOffset = 0;
}
i++;
}
/* Destination is now aligned. Process as many 8-sample
chunks as we can before we run out of room from the current
sound chunk. We do 8 per loop to avoid extra source data reads. */
samplesLeft = count - i;
chunkSamplesLeft = SND_CHUNK_SIZE - sampleOffset;
if(samplesLeft > chunkSamplesLeft)
samplesLeft = chunkSamplesLeft;
vectorCount = samplesLeft / 8;
if(vectorCount)
{
vector unsigned char tmp;
vector short s0, s1, sampleData0, sampleData1;
vector signed int merge0, merge1;
vector signed int d0, d1, d2, d3;
vector unsigned char samplePermute0 =
VECCONST_UINT8(0, 1, 4, 5, 0, 1, 4, 5, 2, 3, 6, 7, 2, 3, 6, 7);
vector unsigned char samplePermute1 =
VECCONST_UINT8(8, 9, 12, 13, 8, 9, 12, 13, 10, 11, 14, 15, 10, 11, 14, 15);
vector unsigned char loadPermute0, loadPermute1;
// Rather than permute the vectors after we load them to do the sample
// replication and rearrangement, we permute the alignment vector so
// we do everything in one step below and avoid data shuffling.
tmp = vec_lvsl(0,&samples[sampleOffset]);
loadPermute0 = vec_perm(tmp,tmp,samplePermute0);
loadPermute1 = vec_perm(tmp,tmp,samplePermute1);
s0 = *(vector short *)&samples[sampleOffset];
while(vectorCount)
{
/* Load up source (16-bit) sample data */
s1 = *(vector short *)&samples[sampleOffset+7];
/* Load up destination sample data */
d0 = *(vector signed int *)&samp[i];
d1 = *(vector signed int *)&samp[i+2];
d2 = *(vector signed int *)&samp[i+4];
d3 = *(vector signed int *)&samp[i+6];
sampleData0 = vec_perm(s0,s1,loadPermute0);
sampleData1 = vec_perm(s0,s1,loadPermute1);
merge0 = vec_mule(sampleData0,volume_vec);
merge0 = vec_sra(merge0,volume_shift); /* Shift down to proper range */
merge1 = vec_mulo(sampleData0,volume_vec);
merge1 = vec_sra(merge1,volume_shift);
d0 = vec_add(merge0,d0);
d1 = vec_add(merge1,d1);
merge0 = vec_mule(sampleData1,volume_vec);
merge0 = vec_sra(merge0,volume_shift); /* Shift down to proper range */
merge1 = vec_mulo(sampleData1,volume_vec);
merge1 = vec_sra(merge1,volume_shift);
d2 = vec_add(merge0,d2);
d3 = vec_add(merge1,d3);
/* Store destination sample data */
*(vector signed int *)&samp[i] = d0;
*(vector signed int *)&samp[i+2] = d1;
*(vector signed int *)&samp[i+4] = d2;
*(vector signed int *)&samp[i+6] = d3;
i += 8;
vectorCount--;
s0 = s1;
sampleOffset += 8;
}
if (sampleOffset == SND_CHUNK_SIZE) {
chunk = chunk->next;
samples = chunk->sndChunk;
sampleOffset = 0;
}
}
}
} else {
fleftvol = ch->leftvol*snd_vol;
frightvol = ch->rightvol*snd_vol;
ooff = sampleOffset;
samples = chunk->sndChunk;
for ( i=0 ; i<count ; i++ ) {
aoff = ooff;
ooff = ooff + ch->dopplerScale * sc->soundChannels;
boff = ooff;
fdata[0] = fdata[1] = 0;
for (j=aoff; j<boff; j += sc->soundChannels) {
if (j == SND_CHUNK_SIZE) {
chunk = chunk->next;
if (!chunk) {
chunk = sc->soundData;
}
samples = chunk->sndChunk;
ooff -= SND_CHUNK_SIZE;
}
if ( sc->soundChannels == 2 ) {
fdata[0] += samples[j&(SND_CHUNK_SIZE-1)];
fdata[1] += samples[(j+1)&(SND_CHUNK_SIZE-1)];
} else {
fdata[0] += samples[j&(SND_CHUNK_SIZE-1)];
fdata[1] += samples[j&(SND_CHUNK_SIZE-1)];
}
}
fdiv = 256 * (boff-aoff) / sc->soundChannels;
samp[i].left += (fdata[0] * fleftvol)/fdiv;
samp[i].right += (fdata[1] * frightvol)/fdiv;
}
}
}
#endif

4
code/client/snd_local.h
View file

@ -264,3 +264,7 @@ typedef enum
typedef int srcHandle_t;
qboolean S_AL_Init( soundInterface_t *si );
#ifdef idppc_altivec
void S_PaintChannelFrom16_altivec( portable_samplepair_t paintbuffer[PAINTBUFFER_SIZE], int snd_vol, channel_t *ch, const sfx_t *sc, int count, int sampleOffset, int bufferOffset );
#endif

198
code/client/snd_mix.c
View file

@ -23,9 +23,6 @@ Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
#include "client.h"
#include "snd_local.h"
#if idppc_altivec && !defined(__APPLE__)
#include <altivec.h>
#endif
static portable_samplepair_t paintbuffer[PAINTBUFFER_SIZE];
static int snd_vol;
@ -241,197 +238,6 @@ CHANNEL MIXING
===============================================================================
*/
#if idppc_altivec
static void S_PaintChannelFrom16_altivec( channel_t *ch, const sfx_t *sc, int count, int sampleOffset, int bufferOffset ) {
int data, aoff, boff;
int leftvol, rightvol;
int i, j;
portable_samplepair_t *samp;
sndBuffer *chunk;
short *samples;
float ooff, fdata[2], fdiv, fleftvol, frightvol;
if (sc->soundChannels <= 0) {
return;
}
samp = &paintbuffer[ bufferOffset ];
if (ch->doppler) {
sampleOffset = sampleOffset*ch->oldDopplerScale;
}
if ( sc->soundChannels == 2 ) {
sampleOffset *= sc->soundChannels;
if ( sampleOffset & 1 ) {
sampleOffset &= ~1;
}
}
chunk = sc->soundData;
while (sampleOffset>=SND_CHUNK_SIZE) {
chunk = chunk->next;
sampleOffset -= SND_CHUNK_SIZE;
if (!chunk) {
chunk = sc->soundData;
}
}
if (!ch->doppler || ch->dopplerScale==1.0f) {
vector signed short volume_vec;
vector unsigned int volume_shift;
int vectorCount, samplesLeft, chunkSamplesLeft;
leftvol = ch->leftvol*snd_vol;
rightvol = ch->rightvol*snd_vol;
samples = chunk->sndChunk;
((short *)&volume_vec)[0] = leftvol;
((short *)&volume_vec)[1] = leftvol;
((short *)&volume_vec)[4] = leftvol;
((short *)&volume_vec)[5] = leftvol;
((short *)&volume_vec)[2] = rightvol;
((short *)&volume_vec)[3] = rightvol;
((short *)&volume_vec)[6] = rightvol;
((short *)&volume_vec)[7] = rightvol;
volume_shift = vec_splat_u32(8);
i = 0;
while(i < count) {
/* Try to align destination to 16-byte boundary */
while(i < count && (((unsigned long)&samp[i] & 0x1f) || ((count-i) < 8) || ((SND_CHUNK_SIZE - sampleOffset) < 8))) {
data = samples[sampleOffset++];
samp[i].left += (data * leftvol)>>8;
if ( sc->soundChannels == 2 ) {
data = samples[sampleOffset++];
}
samp[i].right += (data * rightvol)>>8;
if (sampleOffset == SND_CHUNK_SIZE) {
chunk = chunk->next;
samples = chunk->sndChunk;
sampleOffset = 0;
}
i++;
}
/* Destination is now aligned. Process as many 8-sample
chunks as we can before we run out of room from the current
sound chunk. We do 8 per loop to avoid extra source data reads. */
samplesLeft = count - i;
chunkSamplesLeft = SND_CHUNK_SIZE - sampleOffset;
if(samplesLeft > chunkSamplesLeft)
samplesLeft = chunkSamplesLeft;
vectorCount = samplesLeft / 8;
if(vectorCount)
{
vector unsigned char tmp;
vector short s0, s1, sampleData0, sampleData1;
vector signed int merge0, merge1;
vector signed int d0, d1, d2, d3;
vector unsigned char samplePermute0 =
VECCONST_UINT8(0, 1, 4, 5, 0, 1, 4, 5, 2, 3, 6, 7, 2, 3, 6, 7);
vector unsigned char samplePermute1 =
VECCONST_UINT8(8, 9, 12, 13, 8, 9, 12, 13, 10, 11, 14, 15, 10, 11, 14, 15);
vector unsigned char loadPermute0, loadPermute1;
// Rather than permute the vectors after we load them to do the sample
// replication and rearrangement, we permute the alignment vector so
// we do everything in one step below and avoid data shuffling.
tmp = vec_lvsl(0,&samples[sampleOffset]);
loadPermute0 = vec_perm(tmp,tmp,samplePermute0);
loadPermute1 = vec_perm(tmp,tmp,samplePermute1);
s0 = *(vector short *)&samples[sampleOffset];
while(vectorCount)
{
/* Load up source (16-bit) sample data */
s1 = *(vector short *)&samples[sampleOffset+7];
/* Load up destination sample data */
d0 = *(vector signed int *)&samp[i];
d1 = *(vector signed int *)&samp[i+2];
d2 = *(vector signed int *)&samp[i+4];
d3 = *(vector signed int *)&samp[i+6];
sampleData0 = vec_perm(s0,s1,loadPermute0);
sampleData1 = vec_perm(s0,s1,loadPermute1);
merge0 = vec_mule(sampleData0,volume_vec);
merge0 = vec_sra(merge0,volume_shift); /* Shift down to proper range */
merge1 = vec_mulo(sampleData0,volume_vec);
merge1 = vec_sra(merge1,volume_shift);
d0 = vec_add(merge0,d0);
d1 = vec_add(merge1,d1);
merge0 = vec_mule(sampleData1,volume_vec);
merge0 = vec_sra(merge0,volume_shift); /* Shift down to proper range */
merge1 = vec_mulo(sampleData1,volume_vec);
merge1 = vec_sra(merge1,volume_shift);
d2 = vec_add(merge0,d2);
d3 = vec_add(merge1,d3);
/* Store destination sample data */
*(vector signed int *)&samp[i] = d0;
*(vector signed int *)&samp[i+2] = d1;
*(vector signed int *)&samp[i+4] = d2;
*(vector signed int *)&samp[i+6] = d3;
i += 8;
vectorCount--;
s0 = s1;
sampleOffset += 8;
}
if (sampleOffset == SND_CHUNK_SIZE) {
chunk = chunk->next;
samples = chunk->sndChunk;
sampleOffset = 0;
}
}
}
} else {
fleftvol = ch->leftvol*snd_vol;
frightvol = ch->rightvol*snd_vol;
ooff = sampleOffset;
samples = chunk->sndChunk;
for ( i=0 ; i<count ; i++ ) {
aoff = ooff;
ooff = ooff + ch->dopplerScale * sc->soundChannels;
boff = ooff;
fdata[0] = fdata[1] = 0;
for (j=aoff; j<boff; j += sc->soundChannels) {
if (j == SND_CHUNK_SIZE) {
chunk = chunk->next;
if (!chunk) {
chunk = sc->soundData;
}
samples = chunk->sndChunk;
ooff -= SND_CHUNK_SIZE;
}
if ( sc->soundChannels == 2 ) {
fdata[0] += samples[j&(SND_CHUNK_SIZE-1)];
fdata[1] += samples[(j+1)&(SND_CHUNK_SIZE-1)];
} else {
fdata[0] += samples[j&(SND_CHUNK_SIZE-1)];
fdata[1] += samples[j&(SND_CHUNK_SIZE-1)];
}
}
fdiv = 256 * (boff-aoff) / sc->soundChannels;
samp[i].left += (fdata[0] * fleftvol)/fdiv;
samp[i].right += (fdata[1] * frightvol)/fdiv;
}
}
}
#endif
static void S_PaintChannelFrom16_scalar( channel_t *ch, const sfx_t *sc, int count, int sampleOffset, int bufferOffset ) {
int data, aoff, boff;
int leftvol, rightvol;
@ -530,8 +336,8 @@ static void S_PaintChannelFrom16_scalar( channel_t *ch, const sfx_t *sc, int cou
static void S_PaintChannelFrom16( channel_t *ch, const sfx_t *sc, int count, int sampleOffset, int bufferOffset ) {
#if idppc_altivec
if (com_altivec->integer) {
// must be in a separate function or G3 systems will crash.
S_PaintChannelFrom16_altivec( ch, sc, count, sampleOffset, bufferOffset );
// must be in a separate translation unit or G3 systems will crash.
S_PaintChannelFrom16_altivec( paintbuffer, snd_vol, ch, sc, count, sampleOffset, bufferOffset );
return;
}
#endif

414
code/renderergl1/tr_altivec.c Normal file
View file

@ -0,0 +1,414 @@
/*
===========================================================================
Copyright (C) 1999-2005 Id Software, Inc.
This file is part of Quake III Arena source code.
Quake III Arena source code is free software; you can redistribute it
and/or modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the License,
or (at your option) any later version.
Quake III Arena source code is distributed in the hope that it will be
useful, but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with Quake III Arena source code; if not, write to the Free Software
Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
===========================================================================
*/
/* This file is only compiled for PowerPC builds with Altivec support.
Altivec intrinsics need to be in a separate file, so GCC's -maltivec
command line can enable them, but give us the option to _not_ use that
on other files, where the compiler might then generate Altivec
instructions for normal floating point, crashing on G3 (etc) processors. */
#include "tr_local.h"
#if idppc_altivec
#if !defined(__APPLE__)
#include <altivec.h>
#endif
void ProjectDlightTexture_altivec( void ) {
int i, l;
vec_t origin0, origin1, origin2;
float texCoords0, texCoords1;
vector float floatColorVec0, floatColorVec1;
vector float modulateVec, colorVec, zero;
vector short colorShort;
vector signed int colorInt;
vector unsigned char floatColorVecPerm, modulatePerm, colorChar;
vector unsigned char vSel = VECCONST_UINT8(0x00, 0x00, 0x00, 0xff,
0x00, 0x00, 0x00, 0xff,
0x00, 0x00, 0x00, 0xff,
0x00, 0x00, 0x00, 0xff);
float *texCoords;
byte *colors;
byte clipBits[SHADER_MAX_VERTEXES];
float texCoordsArray[SHADER_MAX_VERTEXES][2];
byte colorArray[SHADER_MAX_VERTEXES][4];
glIndex_t hitIndexes[SHADER_MAX_INDEXES];
int numIndexes;
float scale;
float radius;
vec3_t floatColor;
float modulate = 0.0f;
if ( !backEnd.refdef.num_dlights ) {
return;
}
// There has to be a better way to do this so that floatColor
// and/or modulate are already 16-byte aligned.
floatColorVecPerm = vec_lvsl(0,(float *)floatColor);
modulatePerm = vec_lvsl(0,(float *)&modulate);
modulatePerm = (vector unsigned char)vec_splat((vector unsigned int)modulatePerm,0);
zero = (vector float)vec_splat_s8(0);
for ( l = 0 ; l < backEnd.refdef.num_dlights ; l++ ) {
dlight_t *dl;
if ( !( tess.dlightBits & ( 1 << l ) ) ) {
continue; // this surface definitely doesn't have any of this light
}
texCoords = texCoordsArray[0];
colors = colorArray[0];
dl = &backEnd.refdef.dlights[l];
origin0 = dl->transformed[0];
origin1 = dl->transformed[1];
origin2 = dl->transformed[2];
radius = dl->radius;
scale = 1.0f / radius;
if(r_greyscale->integer)
{
float luminance;
luminance = LUMA(dl->color[0], dl->color[1], dl->color[2]) * 255.0f;
floatColor[0] = floatColor[1] = floatColor[2] = luminance;
}
else if(r_greyscale->value)
{
float luminance;
luminance = LUMA(dl->color[0], dl->color[1], dl->color[2]) * 255.0f;
floatColor[0] = LERP(dl->color[0] * 255.0f, luminance, r_greyscale->value);
floatColor[1] = LERP(dl->color[1] * 255.0f, luminance, r_greyscale->value);
floatColor[2] = LERP(dl->color[2] * 255.0f, luminance, r_greyscale->value);
}
else
{
floatColor[0] = dl->color[0] * 255.0f;
floatColor[1] = dl->color[1] * 255.0f;
floatColor[2] = dl->color[2] * 255.0f;
}
floatColorVec0 = vec_ld(0, floatColor);
floatColorVec1 = vec_ld(11, floatColor);
floatColorVec0 = vec_perm(floatColorVec0,floatColorVec0,floatColorVecPerm);
for ( i = 0 ; i < tess.numVertexes ; i++, texCoords += 2, colors += 4 ) {
int clip = 0;
vec_t dist0, dist1, dist2;
dist0 = origin0 - tess.xyz[i][0];
dist1 = origin1 - tess.xyz[i][1];
dist2 = origin2 - tess.xyz[i][2];
backEnd.pc.c_dlightVertexes++;
texCoords0 = 0.5f + dist0 * scale;
texCoords1 = 0.5f + dist1 * scale;
if( !r_dlightBacks->integer &&
// dist . tess.normal[i]
( dist0 * tess.normal[i][0] +
dist1 * tess.normal[i][1] +
dist2 * tess.normal[i][2] ) < 0.0f ) {
clip = 63;
} else {
if ( texCoords0 < 0.0f ) {
clip |= 1;
} else if ( texCoords0 > 1.0f ) {
clip |= 2;
}
if ( texCoords1 < 0.0f ) {
clip |= 4;
} else if ( texCoords1 > 1.0f ) {
clip |= 8;
}
texCoords[0] = texCoords0;
texCoords[1] = texCoords1;
// modulate the strength based on the height and color
if ( dist2 > radius ) {
clip |= 16;
modulate = 0.0f;
} else if ( dist2 < -radius ) {
clip |= 32;
modulate = 0.0f;
} else {
dist2 = Q_fabs(dist2);
if ( dist2 < radius * 0.5f ) {
modulate = 1.0f;
} else {
modulate = 2.0f * (radius - dist2) * scale;
}
}
}
clipBits[i] = clip;
modulateVec = vec_ld(0,(float *)&modulate);
modulateVec = vec_perm(modulateVec,modulateVec,modulatePerm);
colorVec = vec_madd(floatColorVec0,modulateVec,zero);
colorInt = vec_cts(colorVec,0); // RGBx
colorShort = vec_pack(colorInt,colorInt); // RGBxRGBx
colorChar = vec_packsu(colorShort,colorShort); // RGBxRGBxRGBxRGBx
colorChar = vec_sel(colorChar,vSel,vSel); // RGBARGBARGBARGBA replace alpha with 255
vec_ste((vector unsigned int)colorChar,0,(unsigned int *)colors); // store color
}
// build a list of triangles that need light
numIndexes = 0;
for ( i = 0 ; i < tess.numIndexes ; i += 3 ) {
int a, b, c;
a = tess.indexes[i];
b = tess.indexes[i+1];
c = tess.indexes[i+2];
if ( clipBits[a] & clipBits[b] & clipBits[c] ) {
continue; // not lighted
}
hitIndexes[numIndexes] = a;
hitIndexes[numIndexes+1] = b;
hitIndexes[numIndexes+2] = c;
numIndexes += 3;
}
if ( !numIndexes ) {
continue;
}
qglEnableClientState( GL_TEXTURE_COORD_ARRAY );
qglTexCoordPointer( 2, GL_FLOAT, 0, texCoordsArray[0] );
qglEnableClientState( GL_COLOR_ARRAY );
qglColorPointer( 4, GL_UNSIGNED_BYTE, 0, colorArray );
GL_Bind( tr.dlightImage );
// include GLS_DEPTHFUNC_EQUAL so alpha tested surfaces don't add light
// where they aren't rendered
if ( dl->additive ) {
GL_State( GLS_SRCBLEND_ONE | GLS_DSTBLEND_ONE | GLS_DEPTHFUNC_EQUAL );
}
else {
GL_State( GLS_SRCBLEND_DST_COLOR | GLS_DSTBLEND_ONE | GLS_DEPTHFUNC_EQUAL );
}
R_DrawElements( numIndexes, hitIndexes );
backEnd.pc.c_totalIndexes += numIndexes;
backEnd.pc.c_dlightIndexes += numIndexes;
}
}
void RB_CalcDiffuseColor_altivec( unsigned char *colors )
{
int i;
float *v, *normal;
trRefEntity_t *ent;
int ambientLightInt;
vec3_t lightDir;
int numVertexes;
vector unsigned char vSel = VECCONST_UINT8(0x00, 0x00, 0x00, 0xff,
0x00, 0x00, 0x00, 0xff,
0x00, 0x00, 0x00, 0xff,
0x00, 0x00, 0x00, 0xff);
vector float ambientLightVec;
vector float directedLightVec;
vector float lightDirVec;
vector float normalVec0, normalVec1;
vector float incomingVec0, incomingVec1, incomingVec2;
vector float zero, jVec;
vector signed int jVecInt;
vector signed short jVecShort;
vector unsigned char jVecChar, normalPerm;
ent = backEnd.currentEntity;
ambientLightInt = ent->ambientLightInt;
// A lot of this could be simplified if we made sure
// entities light info was 16-byte aligned.
jVecChar = vec_lvsl(0, ent->ambientLight);
ambientLightVec = vec_ld(0, (vector float *)ent->ambientLight);
jVec = vec_ld(11, (vector float *)ent->ambientLight);
ambientLightVec = vec_perm(ambientLightVec,jVec,jVecChar);
jVecChar = vec_lvsl(0, ent->directedLight);
directedLightVec = vec_ld(0,(vector float *)ent->directedLight);
jVec = vec_ld(11,(vector float *)ent->directedLight);
directedLightVec = vec_perm(directedLightVec,jVec,jVecChar);
jVecChar = vec_lvsl(0, ent->lightDir);
lightDirVec = vec_ld(0,(vector float *)ent->lightDir);
jVec = vec_ld(11,(vector float *)ent->lightDir);
lightDirVec = vec_perm(lightDirVec,jVec,jVecChar);
zero = (vector float)vec_splat_s8(0);
VectorCopy( ent->lightDir, lightDir );
v = tess.xyz[0];
normal = tess.normal[0];
normalPerm = vec_lvsl(0,normal);
numVertexes = tess.numVertexes;
for (i = 0 ; i < numVertexes ; i++, v += 4, normal += 4) {
normalVec0 = vec_ld(0,(vector float *)normal);
normalVec1 = vec_ld(11,(vector float *)normal);
normalVec0 = vec_perm(normalVec0,normalVec1,normalPerm);
incomingVec0 = vec_madd(normalVec0, lightDirVec, zero);
incomingVec1 = vec_sld(incomingVec0,incomingVec0,4);
incomingVec2 = vec_add(incomingVec0,incomingVec1);
incomingVec1 = vec_sld(incomingVec1,incomingVec1,4);
incomingVec2 = vec_add(incomingVec2,incomingVec1);
incomingVec0 = vec_splat(incomingVec2,0);
incomingVec0 = vec_max(incomingVec0,zero);
normalPerm = vec_lvsl(12,normal);
jVec = vec_madd(incomingVec0, directedLightVec, ambientLightVec);
jVecInt = vec_cts(jVec,0); // RGBx
jVecShort = vec_pack(jVecInt,jVecInt); // RGBxRGBx
jVecChar = vec_packsu(jVecShort,jVecShort); // RGBxRGBxRGBxRGBx
jVecChar = vec_sel(jVecChar,vSel,vSel); // RGBARGBARGBARGBA replace alpha with 255
vec_ste((vector unsigned int)jVecChar,0,(unsigned int *)&colors[i*4]); // store color
}
}
void LerpMeshVertexes_altivec(md3Surface_t *surf, float backlerp)
{
short *oldXyz, *newXyz, *oldNormals, *newNormals;
float *outXyz, *outNormal;
float oldXyzScale QALIGN(16);
float newXyzScale QALIGN(16);
float oldNormalScale QALIGN(16);
float newNormalScale QALIGN(16);
int vertNum;
unsigned lat, lng;
int numVerts;
outXyz = tess.xyz[tess.numVertexes];
outNormal = tess.normal[tess.numVertexes];
newXyz = (short *)((byte *)surf + surf->ofsXyzNormals)
+ (backEnd.currentEntity->e.frame * surf->numVerts * 4);
newNormals = newXyz + 3;
newXyzScale = MD3_XYZ_SCALE * (1.0 - backlerp);
newNormalScale = 1.0 - backlerp;
numVerts = surf->numVerts;
if ( backlerp == 0 ) {
vector signed short newNormalsVec0;
vector signed short newNormalsVec1;
vector signed int newNormalsIntVec;
vector float newNormalsFloatVec;
vector float newXyzScaleVec;
vector unsigned char newNormalsLoadPermute;
vector unsigned char newNormalsStorePermute;
vector float zero;
newNormalsStorePermute = vec_lvsl(0,(float *)&newXyzScaleVec);
newXyzScaleVec = *(vector float *)&newXyzScale;
newXyzScaleVec = vec_perm(newXyzScaleVec,newXyzScaleVec,newNormalsStorePermute);
newXyzScaleVec = vec_splat(newXyzScaleVec,0);
newNormalsLoadPermute = vec_lvsl(0,newXyz);
newNormalsStorePermute = vec_lvsr(0,outXyz);
zero = (vector float)vec_splat_s8(0);
//
// just copy the vertexes
//
for (vertNum=0 ; vertNum < numVerts ; vertNum++,
newXyz += 4, newNormals += 4,
outXyz += 4, outNormal += 4)
{
newNormalsLoadPermute = vec_lvsl(0,newXyz);
newNormalsStorePermute = vec_lvsr(0,outXyz);
newNormalsVec0 = vec_ld(0,newXyz);
newNormalsVec1 = vec_ld(16,newXyz);
newNormalsVec0 = vec_perm(newNormalsVec0,newNormalsVec1,newNormalsLoadPermute);
newNormalsIntVec = vec_unpackh(newNormalsVec0);
newNormalsFloatVec = vec_ctf(newNormalsIntVec,0);
newNormalsFloatVec = vec_madd(newNormalsFloatVec,newXyzScaleVec,zero);
newNormalsFloatVec = vec_perm(newNormalsFloatVec,newNormalsFloatVec,newNormalsStorePermute);
//outXyz[0] = newXyz[0] * newXyzScale;
//outXyz[1] = newXyz[1] * newXyzScale;
//outXyz[2] = newXyz[2] * newXyzScale;
lat = ( newNormals[0] >> 8 ) & 0xff;
lng = ( newNormals[0] & 0xff );
lat *= (FUNCTABLE_SIZE/256);
lng *= (FUNCTABLE_SIZE/256);
// decode X as cos( lat ) * sin( long )
// decode Y as sin( lat ) * sin( long )
// decode Z as cos( long )
outNormal[0] = tr.sinTable[(lat+(FUNCTABLE_SIZE/4))&FUNCTABLE_MASK] * tr.sinTable[lng];
outNormal[1] = tr.sinTable[lat] * tr.sinTable[lng];
outNormal[2] = tr.sinTable[(lng+(FUNCTABLE_SIZE/4))&FUNCTABLE_MASK];
vec_ste(newNormalsFloatVec,0,outXyz);
vec_ste(newNormalsFloatVec,4,outXyz);
vec_ste(newNormalsFloatVec,8,outXyz);
}
} else {
//
// interpolate and copy the vertex and normal
//
oldXyz = (short *)((byte *)surf + surf->ofsXyzNormals)
+ (backEnd.currentEntity->e.oldframe * surf->numVerts * 4);
oldNormals = oldXyz + 3;
oldXyzScale = MD3_XYZ_SCALE * backlerp;
oldNormalScale = backlerp;
for (vertNum=0 ; vertNum < numVerts ; vertNum++,
oldXyz += 4, newXyz += 4, oldNormals += 4, newNormals += 4,
outXyz += 4, outNormal += 4)
{
vec3_t uncompressedOldNormal, uncompressedNewNormal;
// interpolate the xyz
outXyz[0] = oldXyz[0] * oldXyzScale + newXyz[0] * newXyzScale;
outXyz[1] = oldXyz[1] * oldXyzScale + newXyz[1] * newXyzScale;
outXyz[2] = oldXyz[2] * oldXyzScale + newXyz[2] * newXyzScale;
// FIXME: interpolate lat/long instead?
lat = ( newNormals[0] >> 8 ) & 0xff;
lng = ( newNormals[0] & 0xff );
lat *= 4;
lng *= 4;
uncompressedNewNormal[0] = tr.sinTable[(lat+(FUNCTABLE_SIZE/4))&FUNCTABLE_MASK] * tr.sinTable[lng];
uncompressedNewNormal[1] = tr.sinTable[lat] * tr.sinTable[lng];
uncompressedNewNormal[2] = tr.sinTable[(lng+(FUNCTABLE_SIZE/4))&FUNCTABLE_MASK];
lat = ( oldNormals[0] >> 8 ) & 0xff;
lng = ( oldNormals[0] & 0xff );
lat *= 4;
lng *= 4;
uncompressedOldNormal[0] = tr.sinTable[(lat+(FUNCTABLE_SIZE/4))&FUNCTABLE_MASK] * tr.sinTable[lng];
uncompressedOldNormal[1] = tr.sinTable[lat] * tr.sinTable[lng];
uncompressedOldNormal[2] = tr.sinTable[(lng+(FUNCTABLE_SIZE/4))&FUNCTABLE_MASK];
outNormal[0] = uncompressedOldNormal[0] * oldNormalScale + uncompressedNewNormal[0] * newNormalScale;
outNormal[1] = uncompressedOldNormal[1] * oldNormalScale + uncompressedNewNormal[1] * newNormalScale;
outNormal[2] = uncompressedOldNormal[2] * oldNormalScale + uncompressedNewNormal[2] * newNormalScale;
// VectorNormalize (outNormal);
}
VectorArrayNormalize((vec4_t *)tess.normal[tess.numVertexes], numVerts);
}
}
#endif

8
code/renderergl1/tr_local.h
View file

@ -1583,5 +1583,13 @@ size_t RE_SaveJPGToBuffer(byte *buffer, size_t bufSize, int quality,
void RE_TakeVideoFrame( int width, int height,
byte *captureBuffer, byte *encodeBuffer, qboolean motionJpeg );
void R_DrawElements( int numIndexes, const glIndex_t *indexes );
void VectorArrayNormalize( vec4_t *normals, unsigned int count );
#ifdef idppc_altivec
void LerpMeshVertexes_altivec( md3Surface_t *surf, float backlerp );
void ProjectDlightTexture_altivec( void );
void RB_CalcDiffuseColor_altivec( unsigned char *colors );
#endif
#endif //TR_LOCAL_H

190
code/renderergl1/tr_shade.c
View file

@ -22,9 +22,6 @@ Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
// tr_shade.c
#include "tr_local.h"
#if idppc_altivec && !defined(__APPLE__)
#include <altivec.h>
#endif
/*
@ -163,7 +160,7 @@ instead of using the single glDrawElements call that may be inefficient
without compiled vertex arrays.
==================
*/
static void R_DrawElements( int numIndexes, const glIndex_t *indexes ) {
void R_DrawElements( int numIndexes, const glIndex_t *indexes ) {
int primitives;
primitives = r_primitives->integer;
@ -407,189 +404,6 @@ ProjectDlightTexture
Perform dynamic lighting with another rendering pass
===================
*/
#if idppc_altivec
static void ProjectDlightTexture_altivec( void ) {
int i, l;
vec_t origin0, origin1, origin2;
float texCoords0, texCoords1;
vector float floatColorVec0, floatColorVec1;
vector float modulateVec, colorVec, zero;
vector short colorShort;
vector signed int colorInt;
vector unsigned char floatColorVecPerm, modulatePerm, colorChar;
vector unsigned char vSel = VECCONST_UINT8(0x00, 0x00, 0x00, 0xff,
0x00, 0x00, 0x00, 0xff,
0x00, 0x00, 0x00, 0xff,
0x00, 0x00, 0x00, 0xff);
float *texCoords;
byte *colors;
byte clipBits[SHADER_MAX_VERTEXES];
float texCoordsArray[SHADER_MAX_VERTEXES][2];
byte colorArray[SHADER_MAX_VERTEXES][4];
glIndex_t hitIndexes[SHADER_MAX_INDEXES];
int numIndexes;
float scale;
float radius;
vec3_t floatColor;
float modulate = 0.0f;
if ( !backEnd.refdef.num_dlights ) {
return;
}
// There has to be a better way to do this so that floatColor
// and/or modulate are already 16-byte aligned.
floatColorVecPerm = vec_lvsl(0,(float *)floatColor);
modulatePerm = vec_lvsl(0,(float *)&modulate);
modulatePerm = (vector unsigned char)vec_splat((vector unsigned int)modulatePerm,0);
zero = (vector float)vec_splat_s8(0);
for ( l = 0 ; l < backEnd.refdef.num_dlights ; l++ ) {
dlight_t *dl;
if ( !( tess.dlightBits & ( 1 << l ) ) ) {
continue; // this surface definitely doesn't have any of this light
}
texCoords = texCoordsArray[0];
colors = colorArray[0];
dl = &backEnd.refdef.dlights[l];
origin0 = dl->transformed[0];
origin1 = dl->transformed[1];
origin2 = dl->transformed[2];
radius = dl->radius;
scale = 1.0f / radius;
if(r_greyscale->integer)
{
float luminance;
luminance = LUMA(dl->color[0], dl->color[1], dl->color[2]) * 255.0f;
floatColor[0] = floatColor[1] = floatColor[2] = luminance;
}
else if(r_greyscale->value)
{
float luminance;
luminance = LUMA(dl->color[0], dl->color[1], dl->color[2]) * 255.0f;
floatColor[0] = LERP(dl->color[0] * 255.0f, luminance, r_greyscale->value);
floatColor[1] = LERP(dl->color[1] * 255.0f, luminance, r_greyscale->value);
floatColor[2] = LERP(dl->color[2] * 255.0f, luminance, r_greyscale->value);
}
else
{
floatColor[0] = dl->color[0] * 255.0f;
floatColor[1] = dl->color[1] * 255.0f;
floatColor[2] = dl->color[2] * 255.0f;
}
floatColorVec0 = vec_ld(0, floatColor);
floatColorVec1 = vec_ld(11, floatColor);
floatColorVec0 = vec_perm(floatColorVec0,floatColorVec0,floatColorVecPerm);
for ( i = 0 ; i < tess.numVertexes ; i++, texCoords += 2, colors += 4 ) {
int clip = 0;
vec_t dist0, dist1, dist2;
dist0 = origin0 - tess.xyz[i][0];
dist1 = origin1 - tess.xyz[i][1];
dist2 = origin2 - tess.xyz[i][2];
backEnd.pc.c_dlightVertexes++;
texCoords0 = 0.5f + dist0 * scale;
texCoords1 = 0.5f + dist1 * scale;
if( !r_dlightBacks->integer &&
// dist . tess.normal[i]
( dist0 * tess.normal[i][0] +
dist1 * tess.normal[i][1] +
dist2 * tess.normal[i][2] ) < 0.0f ) {
clip = 63;
} else {
if ( texCoords0 < 0.0f ) {
clip |= 1;
} else if ( texCoords0 > 1.0f ) {
clip |= 2;
}
if ( texCoords1 < 0.0f ) {
clip |= 4;
} else if ( texCoords1 > 1.0f ) {
clip |= 8;
}
texCoords[0] = texCoords0;
texCoords[1] = texCoords1;
// modulate the strength based on the height and color
if ( dist2 > radius ) {
clip |= 16;
modulate = 0.0f;
} else if ( dist2 < -radius ) {
clip |= 32;
modulate = 0.0f;
} else {
dist2 = Q_fabs(dist2);
if ( dist2 < radius * 0.5f ) {
modulate = 1.0f;
} else {
modulate = 2.0f * (radius - dist2) * scale;
}
}
}
clipBits[i] = clip;
modulateVec = vec_ld(0,(float *)&modulate);
modulateVec = vec_perm(modulateVec,modulateVec,modulatePerm);
colorVec = vec_madd(floatColorVec0,modulateVec,zero);
colorInt = vec_cts(colorVec,0); // RGBx
colorShort = vec_pack(colorInt,colorInt); // RGBxRGBx
colorChar = vec_packsu(colorShort,colorShort); // RGBxRGBxRGBxRGBx
colorChar = vec_sel(colorChar,vSel,vSel); // RGBARGBARGBARGBA replace alpha with 255
vec_ste((vector unsigned int)colorChar,0,(unsigned int *)colors); // store color
}
// build a list of triangles that need light
numIndexes = 0;
for ( i = 0 ; i < tess.numIndexes ; i += 3 ) {
int a, b, c;
a = tess.indexes[i];
b = tess.indexes[i+1];
c = tess.indexes[i+2];
if ( clipBits[a] & clipBits[b] & clipBits[c] ) {
continue; // not lighted
}
hitIndexes[numIndexes] = a;
hitIndexes[numIndexes+1] = b;
hitIndexes[numIndexes+2] = c;
numIndexes += 3;
}
if ( !numIndexes ) {
continue;
}
qglEnableClientState( GL_TEXTURE_COORD_ARRAY );
qglTexCoordPointer( 2, GL_FLOAT, 0, texCoordsArray[0] );
qglEnableClientState( GL_COLOR_ARRAY );
qglColorPointer( 4, GL_UNSIGNED_BYTE, 0, colorArray );
GL_Bind( tr.dlightImage );
// include GLS_DEPTHFUNC_EQUAL so alpha tested surfaces don't add light
// where they aren't rendered
if ( dl->additive ) {
GL_State( GLS_SRCBLEND_ONE | GLS_DSTBLEND_ONE | GLS_DEPTHFUNC_EQUAL );
}
else {
GL_State( GLS_SRCBLEND_DST_COLOR | GLS_DSTBLEND_ONE | GLS_DEPTHFUNC_EQUAL );
}
R_DrawElements( numIndexes, hitIndexes );
backEnd.pc.c_totalIndexes += numIndexes;
backEnd.pc.c_dlightIndexes += numIndexes;
}
}
#endif
static void ProjectDlightTexture_scalar( void ) {
int i, l;
vec3_t origin;
@ -745,7 +559,7 @@ static void ProjectDlightTexture_scalar( void ) {
static void ProjectDlightTexture( void ) {
#if idppc_altivec
if (com_altivec->integer) {
// must be in a separate function or G3 systems will crash.
// must be in a separate translation unit or G3 systems will crash.
ProjectDlightTexture_altivec();
return;
}

76
code/renderergl1/tr_shade_calc.c
View file

@ -22,9 +22,6 @@ Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
// tr_shade_calc.c
#include "tr_local.h"
#if idppc_altivec && !defined(__APPLE__)
#include <altivec.h>
#endif
#define WAVEVALUE( table, base, amplitude, phase, freq ) ((base) + table[ ( (int64_t) ( ( (phase) + tess.shaderTime * (freq) ) * FUNCTABLE_SIZE ) ) & FUNCTABLE_MASK ] * (amplitude))
@ -1082,77 +1079,6 @@ void RB_CalcSpecularAlpha( unsigned char *alphas ) {
**
** The basic vertex lighting calc
*/
#if idppc_altivec
static void RB_CalcDiffuseColor_altivec( unsigned char *colors )
{
int i;
float *v, *normal;
trRefEntity_t *ent;
int ambientLightInt;
vec3_t lightDir;
int numVertexes;
vector unsigned char vSel = VECCONST_UINT8(0x00, 0x00, 0x00, 0xff,
0x00, 0x00, 0x00, 0xff,
0x00, 0x00, 0x00, 0xff,
0x00, 0x00, 0x00, 0xff);
vector float ambientLightVec;
vector float directedLightVec;
vector float lightDirVec;
vector float normalVec0, normalVec1;
vector float incomingVec0, incomingVec1, incomingVec2;
vector float zero, jVec;
vector signed int jVecInt;
vector signed short jVecShort;
vector unsigned char jVecChar, normalPerm;
ent = backEnd.currentEntity;
ambientLightInt = ent->ambientLightInt;
// A lot of this could be simplified if we made sure
// entities light info was 16-byte aligned.
jVecChar = vec_lvsl(0, ent->ambientLight);
ambientLightVec = vec_ld(0, (vector float *)ent->ambientLight);
jVec = vec_ld(11, (vector float *)ent->ambientLight);
ambientLightVec = vec_perm(ambientLightVec,jVec,jVecChar);
jVecChar = vec_lvsl(0, ent->directedLight);
directedLightVec = vec_ld(0,(vector float *)ent->directedLight);
jVec = vec_ld(11,(vector float *)ent->directedLight);
directedLightVec = vec_perm(directedLightVec,jVec,jVecChar);
jVecChar = vec_lvsl(0, ent->lightDir);
lightDirVec = vec_ld(0,(vector float *)ent->lightDir);
jVec = vec_ld(11,(vector float *)ent->lightDir);
lightDirVec = vec_perm(lightDirVec,jVec,jVecChar);
zero = (vector float)vec_splat_s8(0);
VectorCopy( ent->lightDir, lightDir );
v = tess.xyz[0];
normal = tess.normal[0];
normalPerm = vec_lvsl(0,normal);
numVertexes = tess.numVertexes;
for (i = 0 ; i < numVertexes ; i++, v += 4, normal += 4) {
normalVec0 = vec_ld(0,(vector float *)normal);
normalVec1 = vec_ld(11,(vector float *)normal);
normalVec0 = vec_perm(normalVec0,normalVec1,normalPerm);
incomingVec0 = vec_madd(normalVec0, lightDirVec, zero);
incomingVec1 = vec_sld(incomingVec0,incomingVec0,4);
incomingVec2 = vec_add(incomingVec0,incomingVec1);
incomingVec1 = vec_sld(incomingVec1,incomingVec1,4);
incomingVec2 = vec_add(incomingVec2,incomingVec1);
incomingVec0 = vec_splat(incomingVec2,0);
incomingVec0 = vec_max(incomingVec0,zero);
normalPerm = vec_lvsl(12,normal);
jVec = vec_madd(incomingVec0, directedLightVec, ambientLightVec);
jVecInt = vec_cts(jVec,0); // RGBx
jVecShort = vec_pack(jVecInt,jVecInt); // RGBxRGBx
jVecChar = vec_packsu(jVecShort,jVecShort); // RGBxRGBxRGBxRGBx
jVecChar = vec_sel(jVecChar,vSel,vSel); // RGBARGBARGBARGBA replace alpha with 255
vec_ste((vector unsigned int)jVecChar,0,(unsigned int *)&colors[i*4]); // store color
}
}
#endif
static void RB_CalcDiffuseColor_scalar( unsigned char *colors )
{
int i, j;
@ -1206,7 +1132,7 @@ void RB_CalcDiffuseColor( unsigned char *colors )
{
#if idppc_altivec
if (com_altivec->integer) {
// must be in a separate function or G3 systems will crash.
// must be in a separate translation unit or G3 systems will crash.
RB_CalcDiffuseColor_altivec( colors );
return;
}

137
code/renderergl1/tr_surface.c
View file

@ -21,9 +21,6 @@ Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
*/
// tr_surf.c
#include "tr_local.h"
#if idppc_altivec && !defined(__APPLE__)
#include <altivec.h>
#endif
/*
@ -557,7 +554,7 @@ static void RB_SurfaceLightningBolt( void ) {
* The inputs to this routing seem to always be close to length = 1.0 (about 0.6 to 2.0)
* This means that we don't have to worry about zero length or enormously long vectors.
*/
static void VectorArrayNormalize(vec4_t *normals, unsigned int count)
void VectorArrayNormalize(vec4_t *normals, unsigned int count)
{
// assert(count);
@ -612,136 +609,6 @@ static void VectorArrayNormalize(vec4_t *normals, unsigned int count)
/*
** LerpMeshVertexes
*/
#if idppc_altivec
static void LerpMeshVertexes_altivec(md3Surface_t *surf, float backlerp)
{
short *oldXyz, *newXyz, *oldNormals, *newNormals;
float *outXyz, *outNormal;
float oldXyzScale QALIGN(16);
float newXyzScale QALIGN(16);
float oldNormalScale QALIGN(16);
float newNormalScale QALIGN(16);
int vertNum;
unsigned lat, lng;
int numVerts;
outXyz = tess.xyz[tess.numVertexes];
outNormal = tess.normal[tess.numVertexes];
newXyz = (short *)((byte *)surf + surf->ofsXyzNormals)
+ (backEnd.currentEntity->e.frame * surf->numVerts * 4);
newNormals = newXyz + 3;
newXyzScale = MD3_XYZ_SCALE * (1.0 - backlerp);
newNormalScale = 1.0 - backlerp;
numVerts = surf->numVerts;
if ( backlerp == 0 ) {
vector signed short newNormalsVec0;
vector signed short newNormalsVec1;
vector signed int newNormalsIntVec;
vector float newNormalsFloatVec;
vector float newXyzScaleVec;
vector unsigned char newNormalsLoadPermute;
vector unsigned char newNormalsStorePermute;
vector float zero;
newNormalsStorePermute = vec_lvsl(0,(float *)&newXyzScaleVec);
newXyzScaleVec = *(vector float *)&newXyzScale;
newXyzScaleVec = vec_perm(newXyzScaleVec,newXyzScaleVec,newNormalsStorePermute);
newXyzScaleVec = vec_splat(newXyzScaleVec,0);
newNormalsLoadPermute = vec_lvsl(0,newXyz);
newNormalsStorePermute = vec_lvsr(0,outXyz);
zero = (vector float)vec_splat_s8(0);
//
// just copy the vertexes
//
for (vertNum=0 ; vertNum < numVerts ; vertNum++,
newXyz += 4, newNormals += 4,
outXyz += 4, outNormal += 4)
{
newNormalsLoadPermute = vec_lvsl(0,newXyz);
newNormalsStorePermute = vec_lvsr(0,outXyz);
newNormalsVec0 = vec_ld(0,newXyz);
newNormalsVec1 = vec_ld(16,newXyz);
newNormalsVec0 = vec_perm(newNormalsVec0,newNormalsVec1,newNormalsLoadPermute);
newNormalsIntVec = vec_unpackh(newNormalsVec0);
newNormalsFloatVec = vec_ctf(newNormalsIntVec,0);
newNormalsFloatVec = vec_madd(newNormalsFloatVec,newXyzScaleVec,zero);
newNormalsFloatVec = vec_perm(newNormalsFloatVec,newNormalsFloatVec,newNormalsStorePermute);
//outXyz[0] = newXyz[0] * newXyzScale;
//outXyz[1] = newXyz[1] * newXyzScale;
//outXyz[2] = newXyz[2] * newXyzScale;
lat = ( newNormals[0] >> 8 ) & 0xff;
lng = ( newNormals[0] & 0xff );
lat *= (FUNCTABLE_SIZE/256);
lng *= (FUNCTABLE_SIZE/256);
// decode X as cos( lat ) * sin( long )
// decode Y as sin( lat ) * sin( long )
// decode Z as cos( long )
outNormal[0] = tr.sinTable[(lat+(FUNCTABLE_SIZE/4))&FUNCTABLE_MASK] * tr.sinTable[lng];
outNormal[1] = tr.sinTable[lat] * tr.sinTable[lng];
outNormal[2] = tr.sinTable[(lng+(FUNCTABLE_SIZE/4))&FUNCTABLE_MASK];
vec_ste(newNormalsFloatVec,0,outXyz);
vec_ste(newNormalsFloatVec,4,outXyz);
vec_ste(newNormalsFloatVec,8,outXyz);
}
} else {
//
// interpolate and copy the vertex and normal
//
oldXyz = (short *)((byte *)surf + surf->ofsXyzNormals)
+ (backEnd.currentEntity->e.oldframe * surf->numVerts * 4);
oldNormals = oldXyz + 3;
oldXyzScale = MD3_XYZ_SCALE * backlerp;
oldNormalScale = backlerp;
for (vertNum=0 ; vertNum < numVerts ; vertNum++,
oldXyz += 4, newXyz += 4, oldNormals += 4, newNormals += 4,
outXyz += 4, outNormal += 4)
{
vec3_t uncompressedOldNormal, uncompressedNewNormal;
// interpolate the xyz
outXyz[0] = oldXyz[0] * oldXyzScale + newXyz[0] * newXyzScale;
outXyz[1] = oldXyz[1] * oldXyzScale + newXyz[1] * newXyzScale;
outXyz[2] = oldXyz[2] * oldXyzScale + newXyz[2] * newXyzScale;
// FIXME: interpolate lat/long instead?
lat = ( newNormals[0] >> 8 ) & 0xff;
lng = ( newNormals[0] & 0xff );
lat *= 4;
lng *= 4;
uncompressedNewNormal[0] = tr.sinTable[(lat+(FUNCTABLE_SIZE/4))&FUNCTABLE_MASK] * tr.sinTable[lng];
uncompressedNewNormal[1] = tr.sinTable[lat] * tr.sinTable[lng];
uncompressedNewNormal[2] = tr.sinTable[(lng+(FUNCTABLE_SIZE/4))&FUNCTABLE_MASK];
lat = ( oldNormals[0] >> 8 ) & 0xff;
lng = ( oldNormals[0] & 0xff );
lat *= 4;
lng *= 4;
uncompressedOldNormal[0] = tr.sinTable[(lat+(FUNCTABLE_SIZE/4))&FUNCTABLE_MASK] * tr.sinTable[lng];
uncompressedOldNormal[1] = tr.sinTable[lat] * tr.sinTable[lng];
uncompressedOldNormal[2] = tr.sinTable[(lng+(FUNCTABLE_SIZE/4))&FUNCTABLE_MASK];
outNormal[0] = uncompressedOldNormal[0] * oldNormalScale + uncompressedNewNormal[0] * newNormalScale;
outNormal[1] = uncompressedOldNormal[1] * oldNormalScale + uncompressedNewNormal[1] * newNormalScale;
outNormal[2] = uncompressedOldNormal[2] * oldNormalScale + uncompressedNewNormal[2] * newNormalScale;
// VectorNormalize (outNormal);
}
VectorArrayNormalize((vec4_t *)tess.normal[tess.numVertexes], numVerts);
}
}
#endif
static void LerpMeshVertexes_scalar(md3Surface_t *surf, float backlerp)
{
short *oldXyz, *newXyz, *oldNormals, *newNormals;
@ -844,7 +711,7 @@ static void LerpMeshVertexes(md3Surface_t *surf, float backlerp)
{
#if idppc_altivec
if (com_altivec->integer) {
// must be in a separate function or G3 systems will crash.
// must be in a separate translation unit or G3 systems will crash.
LerpMeshVertexes_altivec( surf, backlerp );
return;
}

3
code/renderergl2/tr_shade.c
View file

@ -22,9 +22,6 @@ Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
// tr_shade.c
#include "tr_local.h"
#if idppc_altivec && !defined(__APPLE__)
#include <altivec.h>
#endif
/*

3
code/renderergl2/tr_shade_calc.c
View file

@ -22,9 +22,6 @@ Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
// tr_shade_calc.c
#include "tr_local.h"
#if idppc_altivec && !defined(__APPLE__)
#include <altivec.h>
#endif
#define WAVEVALUE( table, base, amplitude, phase, freq ) ((base) + table[ ( (int64_t) ( ( (phase) + tess.shaderTime * (freq) ) * FUNCTABLE_SIZE ) ) & FUNCTABLE_MASK ] * (amplitude))

3
code/renderergl2/tr_surface.c
View file

@ -21,9 +21,6 @@ Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
*/
// tr_surf.c
#include "tr_local.h"
#if idppc_altivec && !defined(__APPLE__)
#include <altivec.h>
#endif
/*