/*
* Mesa 3-D graphics library
- * Version: 6.5
+ * Version: 7.1
*
- * Copyright (C) 2006 Tungsten Graphics All Rights Reserved.
+ * Copyright (C) 2007 Tungsten Graphics All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
#include "glheader.h"
#include "macros.h"
#include "enums.h"
-#include "program.h"
-#include "prog_instruction.h"
-#include "prog_parameter.h"
-#include "prog_print.h"
-#include "prog_statevars.h"
+#include "shader/program.h"
+#include "shader/prog_instruction.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+#include "shader/prog_statevars.h"
#include "t_context.h" /* NOTE: very light dependency on this */
#include "t_vp_build.h"
}
+/**
+ * Convert a ureg source register to a prog_src_register.
+ */
static void emit_arg( struct prog_src_register *src,
struct ureg reg )
{
+ assert(reg.file != PROGRAM_OUTPUT);
src->File = reg.file;
src->Index = reg.idx;
src->Swizzle = reg.swz;
src->RelAddr = 0;
}
+/**
+ * Convert a ureg dest register to a prog_dst_register.
+ */
static void emit_dst( struct prog_dst_register *dst,
struct ureg reg, GLuint mask )
{
+ /* Check for legal output register type. UNDEFINED will occur in
+ * instruction that don't produce a result (like END).
+ */
+ assert(reg.file == PROGRAM_TEMPORARY ||
+ reg.file == PROGRAM_OUTPUT ||
+ reg.file == PROGRAM_UNDEFINED);
dst->File = reg.file;
dst->Index = reg.idx;
/* allow zero as a shorthand for xyzw */
dst->WriteMask = mask ? mask : WRITEMASK_XYZW;
- dst->CondMask = COND_TR;
- dst->CondSwizzle = 0;
+ dst->CondMask = COND_TR; /* always pass cond test */
+ dst->CondSwizzle = SWIZZLE_NOOP;
dst->CondSrc = 0;
dst->pad = 0;
}
static void emit_op3fn(struct tnl_program *p,
- GLuint op,
+ enum prog_opcode op,
struct ureg dest,
GLuint mask,
struct ureg src0,
struct ureg rescale = register_param2(p, STATE_INTERNAL,
STATE_NORMAL_SCALE);
- emit_op2( p, OPCODE_MUL, p->eye_normal, 0, normal,
+ emit_op2( p, OPCODE_MUL, p->eye_normal, 0, p->eye_normal,
swizzle1(rescale, X));
}
}
/* Need to add some addtional parameters to allow lighting in object
- * space - STATE_SPOT_DIRECTION and STATE_HALF implicitly assume eye
+ * space - STATE_SPOT_DIRECTION and STATE_HALF_VECTOR implicitly assume eye
* space lighting.
*/
static void build_lighting( struct tnl_program *p )
{
struct ureg shininess = get_material(p, 0, STATE_SHININESS);
emit_op1(p, OPCODE_MOV, dots, WRITEMASK_W, swizzle1(shininess,X));
- release_temp(p, shininess);
_col0 = make_temp(p, get_scenecolor(p, 0));
if (separate)
struct ureg shininess = get_material(p, 1, STATE_SHININESS);
emit_op1(p, OPCODE_MOV, dots, WRITEMASK_Z,
negate(swizzle1(shininess,X)));
- release_temp(p, shininess);
_bfc0 = make_temp(p, get_scenecolor(p, 1));
if (separate)
*/
VPpli = register_param3(p, STATE_LIGHT, i,
STATE_POSITION_NORMALIZED);
- half = register_param3(p, STATE_LIGHT, i, STATE_HALF_VECTOR);
+ if (p->state->light_local_viewer) {
+ struct ureg eye_hat = get_eye_position_normalized(p);
+ half = get_temp(p);
+ emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat);
+ emit_normalize_vec3(p, half, half);
+ } else {
+ half = register_param3(p, STATE_LIGHT, i, STATE_HALF_VECTOR);
+ }
}
else {
struct ureg Ppli = register_param3(p, STATE_LIGHT, i,
STATE_POSITION);
struct ureg V = get_eye_position(p);
struct ureg dist = get_temp(p);
+ struct ureg tmpPpli = get_temp(p);
VPpli = get_temp(p);
- half = get_temp(p);
- /* Calulate VPpli vector
+ /* In homogeneous object coordinates
+ */
+ emit_op1(p, OPCODE_RCP, dist, 0, swizzle1(Ppli, W));
+ emit_op2(p, OPCODE_MUL, tmpPpli, 0, Ppli, dist);
+
+ /* Calculate VPpli vector
*/
- emit_op2(p, OPCODE_SUB, VPpli, 0, Ppli, V);
+ emit_op2(p, OPCODE_SUB, VPpli, 0, tmpPpli, V);
+
+ /* we're done with tmpPpli now */
+ release_temp(p, tmpPpli);
/* Normalize VPpli. The dist value also used in
* attenuation below.
p->state->unit[i].light_attenuated) {
att = calculate_light_attenuation(p, i, VPpli, dist);
}
+
+ /* We're done with dist now */
+ release_temp(p, dist);
/* Calculate viewer direction, or use infinite viewer:
*/
+ half = get_temp(p);
if (p->state->light_local_viewer) {
struct ureg eye_hat = get_eye_position_normalized(p);
emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat);
}
emit_normalize_vec3(p, half, half);
-
- release_temp(p, dist);
}
/* Calculate dot products:
emit_op2(p, OPCODE_DP3, dots, WRITEMASK_X, normal, VPpli);
emit_op2(p, OPCODE_DP3, dots, WRITEMASK_Y, normal, half);
+ /* we're done with VPpli and half now, so free them as to not drive up
+ our temp usage unnecessary */
+ release_temp(p, VPpli);
+ release_temp(p, half);
/* Front face lighting:
*/
release_temp(p, specular);
}
- release_temp(p, half);
- release_temp(p, VPpli);
release_temp(p, att);
}
}
{
struct ureg fog = register_output(p, VERT_RESULT_FOGC);
struct ureg input;
-
+
if (p->state->fog_source_is_depth) {
input = swizzle1(get_eye_position(p), Z);
}
input = swizzle1(register_input(p, VERT_ATTRIB_FOG), X);
}
- if (p->state->tnl_do_vertex_fog) {
+ if (p->state->fog_mode && p->state->tnl_do_vertex_fog) {
struct ureg params = register_param2(p, STATE_INTERNAL,
STATE_FOG_PARAMS_OPTIMIZED);
struct ureg tmp = get_temp(p);
+ GLboolean useabs = (p->state->fog_mode != FOG_EXP2);
+
+ if (useabs) {
+ emit_op1(p, OPCODE_ABS, tmp, 0, input);
+ }
switch (p->state->fog_mode) {
case FOG_LINEAR: {
struct ureg id = get_identity_param(p);
- emit_op3(p, OPCODE_MAD, tmp, 0, input, swizzle1(params,X), swizzle1(params,Y));
+ emit_op3(p, OPCODE_MAD, tmp, 0, useabs ? tmp : input,
+ swizzle1(params,X), swizzle1(params,Y));
emit_op2(p, OPCODE_MAX, tmp, 0, tmp, swizzle1(id,X)); /* saturate */
emit_op2(p, OPCODE_MIN, fog, WRITEMASK_X, tmp, swizzle1(id,W));
break;
}
case FOG_EXP:
- emit_op1(p, OPCODE_ABS, tmp, 0, input);
- emit_op2(p, OPCODE_MUL, tmp, 0, tmp, swizzle1(params,Z));
+ emit_op2(p, OPCODE_MUL, tmp, 0, useabs ? tmp : input,
+ swizzle1(params,Z));
emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, negate(tmp));
break;
case FOG_EXP2:
emit_op2(p, OPCODE_MUL, tmp, 0, input, swizzle1(params,W));
- emit_op2(p, OPCODE_MUL, tmp, 0, tmp, tmp);
+ emit_op2(p, OPCODE_MUL, tmp, 0, tmp, tmp);
emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, negate(tmp));
break;
}
-
+
release_temp(p, tmp);
}
else {
/* results = incoming fog coords (compute fog per-fragment later)
*
* KW: Is it really necessary to do anything in this case?
+ * BP: Yes, we always need to compute the absolute value, unless
+ * we want to push that down into the fragment program...
*/
- emit_op1(p, OPCODE_MOV, fog, WRITEMASK_X, input);
+ GLboolean useabs = GL_TRUE;
+ emit_op1(p, useabs ? OPCODE_ABS : OPCODE_MOV, fog, WRITEMASK_X, input);
}
}
struct ureg out = register_output(p, VERT_RESULT_PSIZ);
struct ureg ut = get_temp(p);
+ /* dist = |eyez| */
+ emit_op1(p, OPCODE_ABS, ut, WRITEMASK_Y, swizzle1(eye, Z));
/* p1 + dist * (p2 + dist * p3); */
- emit_op3(p, OPCODE_MAD, ut, 0, negate(swizzle1(eye, Z)),
+ emit_op3(p, OPCODE_MAD, ut, WRITEMASK_X, swizzle1(ut, Y),
swizzle1(state_attenuation, Z), swizzle1(state_attenuation, Y));
- emit_op3(p, OPCODE_MAD, ut, 0, negate(swizzle1(eye, Z)),
+ emit_op3(p, OPCODE_MAD, ut, WRITEMASK_X, swizzle1(ut, Y),
ut, swizzle1(state_attenuation, X));
/* 1 / sqrt(factor) */
- emit_op1(p, OPCODE_RSQ, ut, 0, ut );
+ emit_op1(p, OPCODE_RSQ, ut, WRITEMASK_X, ut );
#if 1
/* out = pointSize / sqrt(factor) */
#else
/* not sure, might make sense to do clamping here,
but it's not done in t_vb_points neither */
- emit_op2(p, OPCODE_MUL, ut, 0, ut, state_size);
- emit_op2(p, OPCODE_MAX, ut, 0, ut, swizzle1(state_size, Y));
+ emit_op2(p, OPCODE_MUL, ut, WRITEMASK_X, ut, state_size);
+ emit_op2(p, OPCODE_MAX, ut, WRITEMASK_X, ut, swizzle1(state_size, Y));
emit_op2(p, OPCODE_MIN, out, WRITEMASK_X, ut, swizzle1(state_size, Z));
#endif
else
p.temp_reserved = ~((1<<max_temps)-1);
- p.program->Base.Instructions
- = (struct prog_instruction*) MALLOC(sizeof(struct prog_instruction) * MAX_INSN);
- p.program->Base.String = 0;
+ p.program->Base.Instructions = _mesa_alloc_instructions(MAX_INSN);
+ p.program->Base.String = NULL;
p.program->Base.NumInstructions =
p.program->Base.NumTemporaries =
p.program->Base.NumParameters =
build_tnl_program( &p );
}
-static void *search_cache( struct tnl_cache *cache,
- GLuint hash,
- const void *key,
- GLuint keysize)
+
+static struct gl_vertex_program *
+search_cache(struct tnl_cache *cache, GLuint hash,
+ const void *key, GLuint keysize)
{
struct tnl_cache_item *c;
for (c = cache->items[hash % cache->size]; c; c = c->next) {
if (c->hash == hash && _mesa_memcmp(c->key, key, keysize) == 0)
- return c->data;
+ return c->prog;
}
return NULL;
}
+
static void rehash( struct tnl_cache *cache )
{
struct tnl_cache_item **items;
cache->size = size;
}
-static void cache_item( struct tnl_cache *cache,
+static void cache_item( GLcontext *ctx,
+ struct tnl_cache *cache,
GLuint hash,
void *key,
- void *data )
+ struct gl_vertex_program *prog )
{
- struct tnl_cache_item *c = (struct tnl_cache_item*) _mesa_malloc(sizeof(*c));
+ struct tnl_cache_item *c = CALLOC_STRUCT(tnl_cache_item);
c->hash = hash;
c->key = key;
- c->data = data;
+
+ c->prog = prog;
if (++cache->n_items > cache->size * 1.5)
rehash(cache);
GLuint hash;
const struct gl_vertex_program *prev = ctx->VertexProgram._Current;
- if (!ctx->VertexProgram._Current) {
+ if (!ctx->VertexProgram._Current ||
+ ctx->VertexProgram._Current == ctx->VertexProgram._TnlProgram) {
+ struct gl_vertex_program *newProg;
+
/* Grab all the relevent state and put it in a single structure:
*/
key = make_state_key(ctx);
/* Look for an already-prepared program for this state:
*/
- ctx->VertexProgram._TnlProgram = (struct gl_vertex_program *)
- search_cache( tnl->vp_cache, hash, key, sizeof(*key) );
+ newProg = search_cache( tnl->vp_cache, hash, key, sizeof(*key));
/* OK, we'll have to build a new one:
*/
- if (!ctx->VertexProgram._TnlProgram) {
+ if (!newProg) {
+
if (0)
_mesa_printf("Build new TNL program\n");
- ctx->VertexProgram._TnlProgram = (struct gl_vertex_program *)
+ newProg = (struct gl_vertex_program *)
ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0);
- create_new_program( key, ctx->VertexProgram._TnlProgram,
- ctx->Const.VertexProgram.MaxTemps );
+ create_new_program( key, newProg, ctx->Const.VertexProgram.MaxTemps );
if (ctx->Driver.ProgramStringNotify)
ctx->Driver.ProgramStringNotify( ctx, GL_VERTEX_PROGRAM_ARB,
- &ctx->VertexProgram._TnlProgram->Base );
+ &newProg->Base );
- cache_item(tnl->vp_cache, hash, key, ctx->VertexProgram._TnlProgram );
+ /* Our ownership of newProg is transferred to the cache */
+ cache_item(ctx, tnl->vp_cache, hash, key, newProg);
}
else {
FREE(key);
- if (0)
- _mesa_printf("Found existing TNL program for key %x\n", hash);
}
- ctx->VertexProgram._Current = ctx->VertexProgram._TnlProgram;
+
+ _mesa_reference_vertprog(ctx, &ctx->VertexProgram._TnlProgram, newProg);
+ _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current, newProg);
}
/* Tell the driver about the change. Could define a new target for
for (c = tnl->vp_cache->items[i]; c; c = next) {
next = c->next;
FREE(c->key);
- FREE(c->data);
+ _mesa_reference_vertprog(ctx, &c->prog, NULL);
FREE(c);
}