*
* Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
* All Rights Reserved.
+ * Copyright 2009 VMware, Inc. All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the
**************************************************************************/
#include "glheader.h"
-#include "macros.h"
-#include "enums.h"
+#include "imports.h"
#include "shader/program.h"
#include "shader/prog_parameter.h"
#include "shader/prog_cache.h"
#include "texenvprogram.h"
+/*
+ * Note on texture units:
+ *
+ * The number of texture units supported by fixed-function fragment
+ * processing is MAX_TEXTURE_COORD_UNITS, not MAX_TEXTURE_IMAGE_UNITS.
+ * That's because there's a one-to-one correspondence between texture
+ * coordinates and samplers in fixed-function processing.
+ *
+ * Since fixed-function vertex processing is limited to MAX_TEXTURE_COORD_UNITS
+ * sets of texcoords, so is fixed-function fragment processing.
+ *
+ * We can safely use ctx->Const.MaxTextureUnits for loop bounds.
+ */
+
+
struct texenvprog_cache_item
{
GLuint hash;
struct texenvprog_cache_item *next;
};
+static GLboolean
+texenv_doing_secondary_color(GLcontext *ctx)
+{
+ if (ctx->Light.Enabled &&
+ (ctx->Light.Model.ColorControl == GL_SEPARATE_SPECULAR_COLOR))
+ return GL_TRUE;
+
+ if (ctx->Fog.ColorSumEnabled)
+ return GL_TRUE;
+
+ return GL_FALSE;
+}
/**
* Up to nine instructions per tex unit, plus fog, specular color.
*/
-#define MAX_INSTRUCTIONS ((MAX_TEXTURE_UNITS * 9) + 12)
+#define MAX_INSTRUCTIONS ((MAX_TEXTURE_COORD_UNITS * 9) + 12)
#define DISASSEM (MESA_VERBOSE & VERBOSE_DISASSEM)
struct mode_opt {
- GLubyte Source:4;
- GLubyte Operand:3;
+#ifdef __GNUC__
+ __extension__ GLubyte Source:4; /**< SRC_x */
+ __extension__ GLubyte Operand:3; /**< OPR_x */
+#else
+ GLuint Source; /**< SRC_x */
+ GLuint Operand; /**< OPR_x */
+#endif
};
struct state_key {
GLuint enabled_units:8;
GLuint separate_specular:1;
GLuint fog_enabled:1;
- GLuint fog_mode:2;
+ GLuint fog_mode:2; /**< FOG_x */
GLuint inputs_available:12;
struct {
GLuint enabled:1;
- GLuint source_index:3; /* one of TEXTURE_1D/2D/3D/CUBE/RECT_INDEX */
+ GLuint source_index:3; /**< TEXTURE_x_INDEX */
GLuint shadow:1;
GLuint ScaleShiftRGB:2;
GLuint ScaleShiftA:2;
- GLuint NumArgsRGB:2;
- GLuint ModeRGB:4;
- GLuint NumArgsA:2;
- GLuint ModeA:4;
+ GLuint NumArgsRGB:3; /**< up to MAX_COMBINER_TERMS */
+ GLuint ModeRGB:5; /**< MODE_x */
+
+ GLuint NumArgsA:3; /**< up to MAX_COMBINER_TERMS */
+ GLuint ModeA:5; /**< MODE_x */
- struct mode_opt OptRGB[3];
- struct mode_opt OptA[3];
- } unit[8];
+ struct mode_opt OptRGB[MAX_COMBINER_TERMS];
+ struct mode_opt OptA[MAX_COMBINER_TERMS];
+ } unit[MAX_TEXTURE_UNITS];
};
#define FOG_LINEAR 0
case GL_ONE_MINUS_SRC_ALPHA: return OPR_ONE_MINUS_SRC_ALPHA;
case GL_ZERO: return OPR_ZERO;
case GL_ONE: return OPR_ONE;
- default: return OPR_UNKNOWN;
+ default:
+ assert(0);
+ return OPR_UNKNOWN;
}
}
#define SRC_CONSTANT 9
#define SRC_PRIMARY_COLOR 10
#define SRC_PREVIOUS 11
+#define SRC_ZERO 12
#define SRC_UNKNOWN 15
static GLuint translate_source( GLenum src )
case GL_CONSTANT: return SRC_CONSTANT;
case GL_PRIMARY_COLOR: return SRC_PRIMARY_COLOR;
case GL_PREVIOUS: return SRC_PREVIOUS;
- default: return SRC_UNKNOWN;
+ case GL_ZERO:
+ return SRC_ZERO;
+ default:
+ assert(0);
+ return SRC_UNKNOWN;
}
}
-#define MODE_REPLACE 0
-#define MODE_MODULATE 1
-#define MODE_ADD 2
-#define MODE_ADD_SIGNED 3
-#define MODE_INTERPOLATE 4
-#define MODE_SUBTRACT 5
-#define MODE_DOT3_RGB 6
-#define MODE_DOT3_RGB_EXT 7
-#define MODE_DOT3_RGBA 8
-#define MODE_DOT3_RGBA_EXT 9
-#define MODE_MODULATE_ADD_ATI 10
-#define MODE_MODULATE_SIGNED_ADD_ATI 11
-#define MODE_MODULATE_SUBTRACT_ATI 12
-#define MODE_UNKNOWN 15
-
-static GLuint translate_mode( GLenum mode )
+#define MODE_REPLACE 0 /* r = a0 */
+#define MODE_MODULATE 1 /* r = a0 * a1 */
+#define MODE_ADD 2 /* r = a0 + a1 */
+#define MODE_ADD_SIGNED 3 /* r = a0 + a1 - 0.5 */
+#define MODE_INTERPOLATE 4 /* r = a0 * a2 + a1 * (1 - a2) */
+#define MODE_SUBTRACT 5 /* r = a0 - a1 */
+#define MODE_DOT3_RGB 6 /* r = a0 . a1 */
+#define MODE_DOT3_RGB_EXT 7 /* r = a0 . a1 */
+#define MODE_DOT3_RGBA 8 /* r = a0 . a1 */
+#define MODE_DOT3_RGBA_EXT 9 /* r = a0 . a1 */
+#define MODE_MODULATE_ADD_ATI 10 /* r = a0 * a2 + a1 */
+#define MODE_MODULATE_SIGNED_ADD_ATI 11 /* r = a0 * a2 + a1 - 0.5 */
+#define MODE_MODULATE_SUBTRACT_ATI 12 /* r = a0 * a2 - a1 */
+#define MODE_ADD_PRODUCTS 13 /* r = a0 * a1 + a2 * a3 */
+#define MODE_ADD_PRODUCTS_SIGNED 14 /* r = a0 * a1 + a2 * a3 - 0.5 */
+#define MODE_BUMP_ENVMAP_ATI 15 /* special */
+#define MODE_UNKNOWN 16
+
+/**
+ * Translate GL combiner state into a MODE_x value
+ */
+static GLuint translate_mode( GLenum envMode, GLenum mode )
{
switch (mode) {
case GL_REPLACE: return MODE_REPLACE;
case GL_MODULATE: return MODE_MODULATE;
- case GL_ADD: return MODE_ADD;
- case GL_ADD_SIGNED: return MODE_ADD_SIGNED;
+ case GL_ADD:
+ if (envMode == GL_COMBINE4_NV)
+ return MODE_ADD_PRODUCTS;
+ else
+ return MODE_ADD;
+ case GL_ADD_SIGNED:
+ if (envMode == GL_COMBINE4_NV)
+ return MODE_ADD_PRODUCTS_SIGNED;
+ else
+ return MODE_ADD_SIGNED;
case GL_INTERPOLATE: return MODE_INTERPOLATE;
case GL_SUBTRACT: return MODE_SUBTRACT;
case GL_DOT3_RGB: return MODE_DOT3_RGB;
case GL_MODULATE_ADD_ATI: return MODE_MODULATE_ADD_ATI;
case GL_MODULATE_SIGNED_ADD_ATI: return MODE_MODULATE_SIGNED_ADD_ATI;
case GL_MODULATE_SUBTRACT_ATI: return MODE_MODULATE_SUBTRACT_ATI;
- default: return MODE_UNKNOWN;
+ case GL_BUMP_ENVMAP_ATI: return MODE_BUMP_ENVMAP_ATI;
+ default:
+ assert(0);
+ return MODE_UNKNOWN;
}
}
-#define TEXTURE_UNKNOWN_INDEX 7
+
+/**
+ * Translate TEXTURE_x_BIT to TEXTURE_x_INDEX.
+ */
static GLuint translate_tex_src_bit( GLbitfield bit )
{
- switch (bit) {
- case TEXTURE_1D_BIT: return TEXTURE_1D_INDEX;
- case TEXTURE_2D_BIT: return TEXTURE_2D_INDEX;
- case TEXTURE_RECT_BIT: return TEXTURE_RECT_INDEX;
- case TEXTURE_3D_BIT: return TEXTURE_3D_INDEX;
- case TEXTURE_CUBE_BIT: return TEXTURE_CUBE_INDEX;
- default: return TEXTURE_UNKNOWN_INDEX;
- }
+ ASSERT(bit);
+ return _mesa_ffs(bit) - 1;
}
+
#define VERT_BIT_TEX_ANY (0xff << VERT_ATTRIB_TEX0)
#define VERT_RESULT_TEX_ANY (0xff << VERT_RESULT_TEX0)
*/
static GLbitfield get_fp_input_mask( GLcontext *ctx )
{
+ /* _NEW_PROGRAM */
+ const GLboolean vertexShader = (ctx->Shader.CurrentProgram &&
+ ctx->Shader.CurrentProgram->LinkStatus &&
+ ctx->Shader.CurrentProgram->VertexProgram);
+ const GLboolean vertexProgram = ctx->VertexProgram._Enabled;
GLbitfield fp_inputs = 0x0;
if (ctx->VertexProgram._Overriden) {
fp_inputs = ~0;
}
else if (ctx->RenderMode == GL_FEEDBACK) {
+ /* _NEW_RENDERMODE */
fp_inputs = (FRAG_BIT_COL0 | FRAG_BIT_TEX0);
}
- else if (!ctx->VertexProgram._Current) {
- /* Fixed function logic */
+ else if (!(vertexProgram || vertexShader) ||
+ !ctx->VertexProgram._Current) {
+ /* Fixed function vertex logic */
+ /* _NEW_ARRAY */
GLbitfield varying_inputs = ctx->varying_vp_inputs;
/* These get generated in the setup routine regardless of the
* vertex program:
*/
+ /* _NEW_POINT */
if (ctx->Point.PointSprite)
varying_inputs |= FRAG_BITS_TEX_ANY;
/* First look at what values may be computed by the generated
* vertex program:
*/
+ /* _NEW_LIGHT */
if (ctx->Light.Enabled) {
fp_inputs |= FRAG_BIT_COL0;
- if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR)
+ if (texenv_doing_secondary_color(ctx))
fp_inputs |= FRAG_BIT_COL1;
}
+ /* _NEW_TEXTURE */
fp_inputs |= (ctx->Texture._TexGenEnabled |
ctx->Texture._TexMatEnabled) << FRAG_ATTRIB_TEX0;
/* Then look at what might be varying as a result of enabled
* arrays, etc:
*/
- if (varying_inputs & VERT_BIT_COLOR0) fp_inputs |= FRAG_BIT_COL0;
- if (varying_inputs & VERT_BIT_COLOR1) fp_inputs |= FRAG_BIT_COL1;
+ if (varying_inputs & VERT_BIT_COLOR0)
+ fp_inputs |= FRAG_BIT_COL0;
+ if (varying_inputs & VERT_BIT_COLOR1)
+ fp_inputs |= FRAG_BIT_COL1;
fp_inputs |= (((varying_inputs & VERT_BIT_TEX_ANY) >> VERT_ATTRIB_TEX0)
<< FRAG_ATTRIB_TEX0);
}
else {
/* calculate from vp->outputs */
- GLbitfield vp_outputs = ctx->VertexProgram._Current->Base.OutputsWritten;
+ struct gl_vertex_program *vprog;
+ GLbitfield vp_outputs;
+
+ /* Choose GLSL vertex shader over ARB vertex program. Need this
+ * since vertex shader state validation comes after fragment state
+ * validation (see additional comments in state.c).
+ */
+ if (vertexShader)
+ vprog = ctx->Shader.CurrentProgram->VertexProgram;
+ else
+ vprog = ctx->VertexProgram.Current;
+
+ vp_outputs = vprog->Base.OutputsWritten;
/* These get generated in the setup routine regardless of the
* vertex program:
*/
+ /* _NEW_POINT */
if (ctx->Point.PointSprite)
vp_outputs |= FRAG_BITS_TEX_ANY;
- if (vp_outputs & (1 << VERT_RESULT_COL0)) fp_inputs |= FRAG_BIT_COL0;
- if (vp_outputs & (1 << VERT_RESULT_COL1)) fp_inputs |= FRAG_BIT_COL1;
+ if (vp_outputs & (1 << VERT_RESULT_COL0))
+ fp_inputs |= FRAG_BIT_COL0;
+ if (vp_outputs & (1 << VERT_RESULT_COL1))
+ fp_inputs |= FRAG_BIT_COL1;
fp_inputs |= (((vp_outputs & VERT_RESULT_TEX_ANY) >> VERT_RESULT_TEX0)
<< FRAG_ATTRIB_TEX0);
{
GLuint i, j;
GLbitfield inputs_referenced = FRAG_BIT_COL0;
- GLbitfield inputs_available = get_fp_input_mask( ctx );
+ const GLbitfield inputs_available = get_fp_input_mask( ctx );
memset(key, 0, sizeof(*key));
- for (i=0;i<MAX_TEXTURE_UNITS;i++) {
+ /* _NEW_TEXTURE */
+ for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i];
+ const struct gl_texture_object *texObj = texUnit->_Current;
+ const struct gl_tex_env_combine_state *comb = texUnit->_CurrentCombine;
GLenum format;
if (!texUnit->_ReallyEnabled || !texUnit->Enabled)
continue;
- format = texUnit->_Current->Image[0][texUnit->_Current->BaseLevel]->_BaseFormat;
+ format = texObj->Image[0][texObj->BaseLevel]->_BaseFormat;
key->unit[i].enabled = 1;
key->enabled_units |= (1<<i);
key->nr_enabled_units = i+1;
inputs_referenced |= FRAG_BIT_TEX(i);
- key->unit[i].source_index =
- translate_tex_src_bit(texUnit->_ReallyEnabled);
- key->unit[i].shadow = ((texUnit->_Current->CompareMode == GL_COMPARE_R_TO_TEXTURE) &&
+ key->unit[i].source_index =
+ translate_tex_src_bit(texUnit->_ReallyEnabled);
+
+ key->unit[i].shadow = ((texObj->CompareMode == GL_COMPARE_R_TO_TEXTURE) &&
((format == GL_DEPTH_COMPONENT) ||
(format == GL_DEPTH_STENCIL_EXT)));
- key->unit[i].NumArgsRGB = texUnit->_CurrentCombine->_NumArgsRGB;
- key->unit[i].NumArgsA = texUnit->_CurrentCombine->_NumArgsA;
+ key->unit[i].NumArgsRGB = comb->_NumArgsRGB;
+ key->unit[i].NumArgsA = comb->_NumArgsA;
key->unit[i].ModeRGB =
- translate_mode(texUnit->_CurrentCombine->ModeRGB);
+ translate_mode(texUnit->EnvMode, comb->ModeRGB);
key->unit[i].ModeA =
- translate_mode(texUnit->_CurrentCombine->ModeA);
-
- key->unit[i].ScaleShiftRGB = texUnit->_CurrentCombine->ScaleShiftRGB;
- key->unit[i].ScaleShiftA = texUnit->_CurrentCombine->ScaleShiftA;
-
- for (j=0;j<3;j++) {
- key->unit[i].OptRGB[j].Operand =
- translate_operand(texUnit->_CurrentCombine->OperandRGB[j]);
- key->unit[i].OptA[j].Operand =
- translate_operand(texUnit->_CurrentCombine->OperandA[j]);
- key->unit[i].OptRGB[j].Source =
- translate_source(texUnit->_CurrentCombine->SourceRGB[j]);
- key->unit[i].OptA[j].Source =
- translate_source(texUnit->_CurrentCombine->SourceA[j]);
+ translate_mode(texUnit->EnvMode, comb->ModeA);
+
+ key->unit[i].ScaleShiftRGB = comb->ScaleShiftRGB;
+ key->unit[i].ScaleShiftA = comb->ScaleShiftA;
+
+ for (j = 0; j < MAX_COMBINER_TERMS; j++) {
+ key->unit[i].OptRGB[j].Operand = translate_operand(comb->OperandRGB[j]);
+ key->unit[i].OptA[j].Operand = translate_operand(comb->OperandA[j]);
+ key->unit[i].OptRGB[j].Source = translate_source(comb->SourceRGB[j]);
+ key->unit[i].OptA[j].Source = translate_source(comb->SourceA[j]);
}
+
+ if (key->unit[i].ModeRGB == MODE_BUMP_ENVMAP_ATI) {
+ /* requires some special translation */
+ key->unit[i].NumArgsRGB = 2;
+ key->unit[i].ScaleShiftRGB = 0;
+ key->unit[i].OptRGB[0].Operand = OPR_SRC_COLOR;
+ key->unit[i].OptRGB[0].Source = SRC_TEXTURE;
+ key->unit[i].OptRGB[1].Operand = OPR_SRC_COLOR;
+ key->unit[i].OptRGB[1].Source = texUnit->BumpTarget - GL_TEXTURE0 + SRC_TEXTURE0;
+ }
}
-
- if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR) {
+
+ /* _NEW_LIGHT | _NEW_FOG */
+ if (texenv_doing_secondary_color(ctx)) {
key->separate_specular = 1;
inputs_referenced |= FRAG_BIT_COL1;
}
+ /* _NEW_FOG */
if (ctx->Fog.Enabled) {
key->fog_enabled = 1;
key->fog_mode = translate_fog_mode(ctx->Fog.Mode);
GLuint file:4;
GLuint idx:8;
GLuint negatebase:1;
- GLuint abs:1;
- GLuint negateabs:1;
GLuint swz:12;
- GLuint pad:5;
+ GLuint pad:7;
};
static const struct ureg undef = {
~0,
0,
0,
- 0,
- 0,
0
};
*/
struct texenv_fragment_program {
struct gl_fragment_program *program;
- GLcontext *ctx;
struct state_key *state;
GLbitfield alu_temps; /**< Track texture indirections, see spec. */
GLbitfield temp_in_use; /**< Tracks temporary regs which are in use. */
GLboolean error;
- struct ureg src_texture[MAX_TEXTURE_UNITS];
+ struct ureg src_texture[MAX_TEXTURE_COORD_UNITS];
/* Reg containing each texture unit's sampled texture color,
* else undef.
*/
+ struct ureg texcoord_tex[MAX_TEXTURE_COORD_UNITS];
+ /* Reg containing texcoord for a texture unit,
+ * needed for bump mapping, else undef.
+ */
+
struct ureg src_previous; /**< Reg containing color from previous
* stage. May need to be decl'd.
*/
reg.file = file;
reg.idx = idx;
reg.negatebase = 0;
- reg.abs = 0;
- reg.negateabs = 0;
reg.swz = SWIZZLE_NOOP;
reg.pad = 0;
return reg;
reg->File = ureg.file;
reg->Index = ureg.idx;
reg->Swizzle = ureg.swz;
- reg->NegateBase = ureg.negatebase ? 0xf : 0x0;
- reg->Abs = ureg.abs;
- reg->NegateAbs = ureg.negateabs;
+ reg->Negate = ureg.negatebase ? NEGATE_XYZW : NEGATE_NONE;
+ reg->Abs = GL_FALSE;
}
static void emit_dst( struct prog_dst_register *dst,
struct ureg src1,
struct ureg src2 )
{
- GLuint nr = p->program->Base.NumInstructions++;
+ const GLuint nr = p->program->Base.NumInstructions++;
struct prog_instruction *inst = &p->program->Base.Instructions[nr];
assert(nr < MAX_INSTRUCTIONS);
GLuint destmask,
GLuint tex_unit,
GLuint tex_idx,
+ GLuint tex_shadow,
struct ureg coord )
{
struct prog_instruction *inst = emit_op( p, op,
inst->TexSrcTarget = tex_idx;
inst->TexSrcUnit = tex_unit;
+ inst->TexShadow = tex_shadow;
p->program->Base.NumTexInstructions++;
*/
reserve_temp(p, dest);
+#if 0
/* Is this a texture indirection?
*/
if ((coord.file == PROGRAM_TEMPORARY &&
p->alu_temps = 0;
assert(0); /* KW: texture env crossbar */
}
+#endif
return dest;
}
case SRC_PRIMARY_COLOR:
return register_input(p, FRAG_ATTRIB_COL0);
+ case SRC_ZERO:
+ return get_zero(p);
+
case SRC_PREVIOUS:
- default:
if (is_undef(p->src_previous))
return register_input(p, FRAG_ATTRIB_COL0);
else
return p->src_previous;
+
+ default:
+ assert(0);
+ return undef;
}
}
case OPR_ONE:
return get_one(p);
case OPR_SRC_COLOR:
+ return src;
default:
+ assert(0);
return src;
}
}
-static GLboolean args_match( struct state_key *key, GLuint unit )
+/**
+ * Check if the RGB and Alpha sources and operands match for the given
+ * texture unit's combinder state. When the RGB and A sources and
+ * operands match, we can emit fewer instructions.
+ */
+static GLboolean args_match( const struct state_key *key, GLuint unit )
{
- GLuint i, nr = key->unit[unit].NumArgsRGB;
+ GLuint i, numArgs = key->unit[unit].NumArgsRGB;
- for (i = 0 ; i < nr ; i++) {
+ for (i = 0; i < numArgs; i++) {
if (key->unit[unit].OptA[i].Source != key->unit[unit].OptRGB[i].Source)
return GL_FALSE;
- switch(key->unit[unit].OptA[i].Operand) {
+ switch (key->unit[unit].OptA[i].Operand) {
case OPR_SRC_ALPHA:
- switch(key->unit[unit].OptRGB[i].Operand) {
+ switch (key->unit[unit].OptRGB[i].Operand) {
case OPR_SRC_COLOR:
case OPR_SRC_ALPHA:
break;
}
break;
case OPR_ONE_MINUS_SRC_ALPHA:
- switch(key->unit[unit].OptRGB[i].Operand) {
+ switch (key->unit[unit].OptRGB[i].Operand) {
case OPR_ONE_MINUS_SRC_COLOR:
case OPR_ONE_MINUS_SRC_ALPHA:
break;
GLuint mode,
const struct mode_opt *opt)
{
- struct ureg src[3];
+ struct ureg src[MAX_COMBINER_TERMS];
struct ureg tmp, half;
GLuint i;
+ assert(nr <= MAX_COMBINER_TERMS);
+
tmp = undef; /* silence warning (bug 5318) */
for (i = 0; i < nr; i++)
/* Arg0 * Arg2 - Arg1 */
emit_arith( p, OPCODE_MAD, dest, mask, 0, src[0], src[2], negate(src[1]) );
return dest;
+ case MODE_ADD_PRODUCTS:
+ /* Arg0 * Arg1 + Arg2 * Arg3 */
+ {
+ struct ureg tmp0 = get_temp(p);
+ emit_arith( p, OPCODE_MUL, tmp0, mask, 0, src[0], src[1], undef );
+ emit_arith( p, OPCODE_MAD, dest, mask, saturate, src[2], src[3], tmp0 );
+ }
+ return dest;
+ case MODE_ADD_PRODUCTS_SIGNED:
+ /* Arg0 * Arg1 + Arg2 * Arg3 - 0.5 */
+ {
+ struct ureg tmp0 = get_temp(p);
+ half = get_half(p);
+ emit_arith( p, OPCODE_MUL, tmp0, mask, 0, src[0], src[1], undef );
+ emit_arith( p, OPCODE_MAD, tmp0, mask, 0, src[2], src[3], tmp0 );
+ emit_arith( p, OPCODE_SUB, dest, mask, saturate, tmp0, half, undef );
+ }
+ return dest;
+ case MODE_BUMP_ENVMAP_ATI:
+ /* special - not handled here */
+ assert(0);
+ return src[0];
default:
+ assert(0);
return src[0];
}
}
static struct ureg
emit_texenv(struct texenv_fragment_program *p, GLuint unit)
{
- struct state_key *key = p->state;
- GLboolean saturate = (unit < p->last_tex_stage);
+ const struct state_key *key = p->state;
+ GLboolean saturate;
GLuint rgb_shift, alpha_shift;
- struct ureg out, shift;
- struct ureg dest;
+ struct ureg out, dest;
if (!key->unit[unit].enabled) {
return get_source(p, SRC_PREVIOUS, 0);
}
+ if (key->unit[unit].ModeRGB == MODE_BUMP_ENVMAP_ATI) {
+ /* this isn't really a env stage delivering a color and handled elsewhere */
+ return get_source(p, SRC_PREVIOUS, 0);
+ }
switch (key->unit[unit].ModeRGB) {
case MODE_DOT3_RGB_EXT:
break;
}
+ /* If we'll do rgb/alpha shifting don't saturate in emit_combine().
+ * We don't want to clamp twice.
+ */
+ saturate = !(rgb_shift || alpha_shift);
+
/* If this is the very last calculation, emit direct to output reg:
*/
if (key->separate_specular ||
rgb_shift)
dest = get_temp( p );
else
- dest = make_ureg(PROGRAM_OUTPUT, FRAG_RESULT_COLR);
+ dest = make_ureg(PROGRAM_OUTPUT, FRAG_RESULT_COLOR);
/* Emit the RGB and A combine ops
*/
}
else if (key->unit[unit].ModeRGB == MODE_DOT3_RGBA_EXT ||
key->unit[unit].ModeRGB == MODE_DOT3_RGBA) {
-
out = emit_combine( p, dest, WRITEMASK_XYZW, saturate,
unit,
key->unit[unit].NumArgsRGB,
/* Deal with the final shift:
*/
if (alpha_shift || rgb_shift) {
+ struct ureg shift;
+
+ saturate = GL_TRUE; /* always saturate at this point */
+
if (rgb_shift == alpha_shift) {
shift = register_scalar_const(p, (GLfloat)(1<<rgb_shift));
}
static void load_texture( struct texenv_fragment_program *p, GLuint unit )
{
if (is_undef(p->src_texture[unit])) {
- GLuint dim = p->state->unit[unit].source_index;
- struct ureg texcoord = register_input(p, FRAG_ATTRIB_TEX0+unit);
+ const GLuint texTarget = p->state->unit[unit].source_index;
+ struct ureg texcoord;
struct ureg tmp = get_tex_temp( p );
- if (dim == TEXTURE_UNKNOWN_INDEX)
- program_error(p, "TexSrcBit");
-
+ if (is_undef(p->texcoord_tex[unit])) {
+ texcoord = register_input(p, FRAG_ATTRIB_TEX0+unit);
+ }
+ else {
+ /* might want to reuse this reg for tex output actually */
+ texcoord = p->texcoord_tex[unit];
+ }
+
/* TODO: Use D0_MASK_XY where possible.
*/
if (p->state->unit[unit].enabled) {
- p->src_texture[unit] = emit_texld( p, OPCODE_TXP,
- tmp, WRITEMASK_XYZW,
- unit, dim, texcoord );
+ GLboolean shadow = GL_FALSE;
- if (p->state->unit[unit].shadow)
+ if (p->state->unit[unit].shadow) {
p->program->Base.ShadowSamplers |= 1 << unit;
+ shadow = GL_TRUE;
+ }
+
+ p->src_texture[unit] = emit_texld( p, OPCODE_TXP,
+ tmp, WRITEMASK_XYZW,
+ unit, texTarget, shadow,
+ texcoord );
p->program->Base.SamplersUsed |= (1 << unit);
/* This identity mapping should already be in place
break;
default:
+ /* not a texture src - do nothing */
break;
}
* Generate instructions for loading all texture source terms.
*/
static GLboolean
-load_texunit_sources( struct texenv_fragment_program *p, int unit )
+load_texunit_sources( struct texenv_fragment_program *p, GLuint unit )
{
- struct state_key *key = p->state;
+ const struct state_key *key = p->state;
GLuint i;
for (i = 0; i < key->unit[unit].NumArgsRGB; i++) {
- load_texenv_source( p, key->unit[unit].OptRGB[i].Source, unit);
+ load_texenv_source( p, key->unit[unit].OptRGB[i].Source, unit );
}
for (i = 0; i < key->unit[unit].NumArgsA; i++) {
return GL_TRUE;
}
+/**
+ * Generate instructions for loading bump map textures.
+ */
+static GLboolean
+load_texunit_bumpmap( struct texenv_fragment_program *p, GLuint unit )
+{
+ const struct state_key *key = p->state;
+ GLuint bumpedUnitNr = key->unit[unit].OptRGB[1].Source - SRC_TEXTURE0;
+ struct ureg texcDst, bumpMapRes;
+ struct ureg constdudvcolor = register_const4f(p, 0.0, 0.0, 0.0, 1.0);
+ struct ureg texcSrc = register_input(p, FRAG_ATTRIB_TEX0 + bumpedUnitNr);
+ struct ureg rotMat0 = register_param3( p, STATE_INTERNAL, STATE_ROT_MATRIX_0, unit );
+ struct ureg rotMat1 = register_param3( p, STATE_INTERNAL, STATE_ROT_MATRIX_1, unit );
+
+ load_texenv_source( p, unit + SRC_TEXTURE0, unit );
+
+ bumpMapRes = get_source(p, key->unit[unit].OptRGB[0].Source, unit);
+ texcDst = get_tex_temp( p );
+ p->texcoord_tex[bumpedUnitNr] = texcDst;
+
+ /* Apply rot matrix and add coords to be available in next phase.
+ * dest = (Arg0.xxxx * rotMat0 + Arg1) + (Arg0.yyyy * rotMat1)
+ * note only 2 coords are affected the rest are left unchanged (mul by 0)
+ */
+ emit_arith( p, OPCODE_MAD, texcDst, WRITEMASK_XYZW, 0,
+ swizzle1(bumpMapRes, SWIZZLE_X), rotMat0, texcSrc );
+ emit_arith( p, OPCODE_MAD, texcDst, WRITEMASK_XYZW, 0,
+ swizzle1(bumpMapRes, SWIZZLE_Y), rotMat1, texcDst );
+
+ /* Move 0,0,0,1 into bumpmap src if someone (crossbar) is foolish
+ * enough to access this later, should optimize away.
+ */
+ emit_arith( p, OPCODE_MOV, bumpMapRes, WRITEMASK_XYZW, 0,
+ constdudvcolor, undef, undef );
+
+ return GL_TRUE;
+}
/**
* Generate a new fragment program which implements the context's
struct ureg cf, out;
_mesa_memset(&p, 0, sizeof(p));
- p.ctx = ctx;
p.state = key;
p.program = program;
*/
p.program->Base.Instructions = instBuffer;
p.program->Base.Target = GL_FRAGMENT_PROGRAM_ARB;
- p.program->Base.NumTexIndirections = 1; /* correct? */
+ p.program->Base.String = NULL;
+ p.program->Base.NumTexIndirections = 1; /* is this right? */
p.program->Base.NumTexInstructions = 0;
p.program->Base.NumAluInstructions = 0;
- p.program->Base.String = NULL;
- p.program->Base.NumInstructions =
- p.program->Base.NumTemporaries =
- p.program->Base.NumParameters =
- p.program->Base.NumAttributes = p.program->Base.NumAddressRegs = 0;
+ p.program->Base.NumInstructions = 0;
+ p.program->Base.NumTemporaries = 0;
+ p.program->Base.NumParameters = 0;
+ p.program->Base.NumAttributes = 0;
+ p.program->Base.NumAddressRegs = 0;
p.program->Base.Parameters = _mesa_new_parameter_list();
+ p.program->Base.InputsRead = 0x0;
+ p.program->Base.OutputsWritten = 1 << FRAG_RESULT_COLOR;
- p.program->Base.InputsRead = 0;
- p.program->Base.OutputsWritten = 1 << FRAG_RESULT_COLR;
-
- for (unit = 0; unit < MAX_TEXTURE_UNITS; unit++)
+ for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) {
p.src_texture[unit] = undef;
+ p.texcoord_tex[unit] = undef;
+ }
p.src_previous = undef;
p.half = undef;
release_temps(ctx, &p);
if (key->enabled_units) {
+ GLboolean needbumpstage = GL_FALSE;
+
+ /* Zeroth pass - bump map textures first */
+ for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++)
+ if (key->unit[unit].enabled &&
+ key->unit[unit].ModeRGB == MODE_BUMP_ENVMAP_ATI) {
+ needbumpstage = GL_TRUE;
+ load_texunit_bumpmap( &p, unit );
+ }
+ if (needbumpstage)
+ p.program->Base.NumTexIndirections++;
+
/* First pass - to support texture_env_crossbar, first identify
* all referenced texture sources and emit texld instructions
* for each:
*/
- for (unit = 0 ; unit < ctx->Const.MaxTextureUnits ; unit++)
+ for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++)
if (key->unit[unit].enabled) {
load_texunit_sources( &p, unit );
p.last_tex_stage = unit;
/* Second pass - emit combine instructions to build final color:
*/
- for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++)
- if (key->enabled_units & (1<<unit)) {
+ for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++)
+ if (key->unit[unit].enabled) {
p.src_previous = emit_texenv( &p, unit );
reserve_temp(&p, p.src_previous); /* don't re-use this temp reg */
release_temps(ctx, &p); /* release all temps */
}
cf = get_source( &p, SRC_PREVIOUS, 0 );
- out = make_ureg( PROGRAM_OUTPUT, FRAG_RESULT_COLR );
+ out = make_ureg( PROGRAM_OUTPUT, FRAG_RESULT_COLOR );
if (key->separate_specular) {
/* Emit specular add.
* a reduced value and not what is expected in FogOption
*/
p.program->FogOption = ctx->Fog.Mode;
- p.program->Base.InputsRead |= FRAG_BIT_FOGC; /* XXX new */
- } else
+ p.program->Base.InputsRead |= FRAG_BIT_FOGC;
+ }
+ else {
p.program->FogOption = GL_NONE;
+ }
if (p.program->Base.NumTexIndirections > ctx->Const.FragmentProgram.MaxTexIndirections)
program_error(&p, "Exceeded max nr indirect texture lookups");
return prog;
}
-
-
-
-/**
- * If _MaintainTexEnvProgram is set we'll generate a fragment program that
- * implements the current texture env/combine mode.
- * This function generates that program and puts it into effect.
- *
- * XXX: remove this function. currently only called by some drivers,
- * not by mesa core. We now handle this properly from inside mesa.
- */
-void
-_mesa_UpdateTexEnvProgram( GLcontext *ctx )
-{
- const struct gl_fragment_program *prev = ctx->FragmentProgram._Current;
-
- ASSERT(ctx->FragmentProgram._MaintainTexEnvProgram);
-
- /* If a conventional fragment program/shader isn't in effect... */
- if (!ctx->FragmentProgram._Enabled &&
- (!ctx->Shader.CurrentProgram ||
- !ctx->Shader.CurrentProgram->FragmentProgram) )
- {
- struct gl_fragment_program *newProg;
-
- newProg = _mesa_get_fixed_func_fragment_program(ctx);
-
- _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, newProg);
- _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._TexEnvProgram, newProg);
- }
-
- /* Tell the driver about the change. Could define a new target for
- * this?
- */
- if (ctx->FragmentProgram._Current != prev && ctx->Driver.BindProgram) {
- ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB,
- (struct gl_program *) ctx->FragmentProgram._Current);
- }
-}