aboutsummaryrefslogtreecommitdiff
path: root/mesalib/src/mesa
diff options
context:
space:
mode:
Diffstat (limited to 'mesalib/src/mesa')
-rw-r--r--mesalib/src/mesa/main/ff_fragment_shader.cpp1435
-rw-r--r--mesalib/src/mesa/main/mtypes.h1
-rw-r--r--mesalib/src/mesa/main/state.c19
-rw-r--r--mesalib/src/mesa/main/texenvprogram.h2
-rw-r--r--mesalib/src/mesa/program/program.c2
-rw-r--r--mesalib/src/mesa/state_tracker/st_cb_clear.c1180
-rw-r--r--mesalib/src/mesa/state_tracker/st_cb_texture.c2
7 files changed, 1386 insertions, 1255 deletions
diff --git a/mesalib/src/mesa/main/ff_fragment_shader.cpp b/mesalib/src/mesa/main/ff_fragment_shader.cpp
index ed513397a..0bc534df5 100644
--- a/mesalib/src/mesa/main/ff_fragment_shader.cpp
+++ b/mesalib/src/mesa/main/ff_fragment_shader.cpp
@@ -3,7 +3,6 @@
* Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
* All Rights Reserved.
* Copyright 2009 VMware, Inc. All Rights Reserved.
- * Copyright © 2010 Intel Corporation
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the
@@ -31,8 +30,6 @@ extern "C" {
#include "glheader.h"
#include "imports.h"
#include "mtypes.h"
-#include "main/uniforms.h"
-#include "main/macros.h"
#include "program/program.h"
#include "program/prog_parameter.h"
#include "program/prog_cache.h"
@@ -42,13 +39,6 @@ extern "C" {
#include "program/programopt.h"
#include "texenvprogram.h"
}
-#include "../glsl/glsl_types.h"
-#include "../glsl/ir.h"
-#include "../glsl/glsl_symbol_table.h"
-#include "../glsl/glsl_parser_extras.h"
-#include "../glsl/ir_optimization.h"
-#include "../glsl/ir_print_visitor.h"
-#include "../program/ir_to_mesa.h"
/*
* Note on texture units:
@@ -69,7 +59,7 @@ struct texenvprog_cache_item
{
GLuint hash;
void *key;
- struct gl_shader_program *data;
+ struct gl_fragment_program *data;
struct texenvprog_cache_item *next;
};
@@ -86,6 +76,13 @@ texenv_doing_secondary_color(struct gl_context *ctx)
return GL_FALSE;
}
+/**
+ * Up to nine instructions per tex unit, plus fog, specular color.
+ */
+#define MAX_INSTRUCTIONS ((MAX_TEXTURE_COORD_UNITS * 9) + 12)
+
+#define DISASSEM (MESA_VERBOSE & VERBOSE_DISASSEM)
+
struct mode_opt {
#ifdef __GNUC__
__extension__ GLubyte Source:4; /**< SRC_x */
@@ -119,6 +116,8 @@ struct state_key {
GLuint NumArgsA:3; /**< up to MAX_COMBINER_TERMS */
GLuint ModeA:5; /**< MODE_x */
+ GLuint texture_cyl_wrap:1; /**< For gallium test/debug only */
+
struct mode_opt OptRGB[MAX_COMBINER_TERMS];
struct mode_opt OptA[MAX_COMBINER_TERMS];
} unit[MAX_TEXTURE_UNITS];
@@ -471,6 +470,10 @@ static GLuint make_state_key( struct gl_context *ctx, struct state_key *key )
key->unit[i].OptRGB[1].Operand = OPR_SRC_COLOR;
key->unit[i].OptRGB[1].Source = texUnit->BumpTarget - GL_TEXTURE0 + SRC_TEXTURE0;
}
+
+ /* this is a back-door for enabling cylindrical texture wrap mode */
+ if (texObj->Priority == 0.125)
+ key->unit[i].texture_cyl_wrap = 1;
}
/* _NEW_LIGHT | _NEW_FOG */
@@ -499,15 +502,40 @@ static GLuint make_state_key( struct gl_context *ctx, struct state_key *key )
}
+/**
+ * Use uregs to represent registers internally, translate to Mesa's
+ * expected formats on emit.
+ *
+ * NOTE: These are passed by value extensively in this file rather
+ * than as usual by pointer reference. If this disturbs you, try
+ * remembering they are just 32bits in size.
+ *
+ * GCC is smart enough to deal with these dword-sized structures in
+ * much the same way as if I had defined them as dwords and was using
+ * macros to access and set the fields. This is much nicer and easier
+ * to evolve.
+ */
+struct ureg {
+ GLuint file:4;
+ GLuint idx:8;
+ GLuint negatebase:1;
+ GLuint swz:12;
+ GLuint pad:7;
+};
+
+static const struct ureg undef = {
+ PROGRAM_UNDEFINED,
+ 255,
+ 0,
+ 0,
+ 0
+};
+
+
/** State used to build the fragment program:
*/
struct texenv_fragment_program {
- struct gl_shader_program *shader_program;
- struct gl_shader *shader;
struct gl_fragment_program *program;
- exec_list *instructions;
- exec_list *top_instructions;
- void *mem_ctx;
struct state_key *state;
GLbitfield alu_temps; /**< Track texture indirections, see spec. */
@@ -515,35 +543,385 @@ struct texenv_fragment_program {
GLbitfield temp_in_use; /**< Tracks temporary regs which are in use. */
GLboolean error;
- ir_variable *src_texture[MAX_TEXTURE_COORD_UNITS];
+ struct ureg src_texture[MAX_TEXTURE_COORD_UNITS];
/* Reg containing each texture unit's sampled texture color,
* else undef.
*/
- /* Texcoord override from bumpmapping. */
- struct ir_variable *texcoord_tex[MAX_TEXTURE_COORD_UNITS];
-
+ struct ureg texcoord_tex[MAX_TEXTURE_COORD_UNITS];
/* Reg containing texcoord for a texture unit,
* needed for bump mapping, else undef.
*/
- ir_rvalue *src_previous; /**< Reg containing color from previous
+ struct ureg src_previous; /**< Reg containing color from previous
* stage. May need to be decl'd.
*/
GLuint last_tex_stage; /**< Number of last enabled texture unit */
+
+ struct ureg half;
+ struct ureg one;
+ struct ureg zero;
};
-static ir_rvalue *
-get_source(struct texenv_fragment_program *p,
- GLuint src, GLuint unit)
+
+
+static struct ureg make_ureg(GLuint file, GLuint idx)
+{
+ struct ureg reg;
+ reg.file = file;
+ reg.idx = idx;
+ reg.negatebase = 0;
+ reg.swz = SWIZZLE_NOOP;
+ reg.pad = 0;
+ return reg;
+}
+
+static struct ureg swizzle( struct ureg reg, int x, int y, int z, int w )
+{
+ reg.swz = MAKE_SWIZZLE4(GET_SWZ(reg.swz, x),
+ GET_SWZ(reg.swz, y),
+ GET_SWZ(reg.swz, z),
+ GET_SWZ(reg.swz, w));
+
+ return reg;
+}
+
+static struct ureg swizzle1( struct ureg reg, int x )
+{
+ return swizzle(reg, x, x, x, x);
+}
+
+static struct ureg negate( struct ureg reg )
+{
+ reg.negatebase ^= 1;
+ return reg;
+}
+
+static GLboolean is_undef( struct ureg reg )
+{
+ return reg.file == PROGRAM_UNDEFINED;
+}
+
+
+static struct ureg get_temp( struct texenv_fragment_program *p )
+{
+ GLint bit;
+
+ /* First try and reuse temps which have been used already:
+ */
+ bit = _mesa_ffs( ~p->temp_in_use & p->alu_temps );
+
+ /* Then any unused temporary:
+ */
+ if (!bit)
+ bit = _mesa_ffs( ~p->temp_in_use );
+
+ if (!bit) {
+ _mesa_problem(NULL, "%s: out of temporaries\n", __FILE__);
+ exit(1);
+ }
+
+ if ((GLuint) bit > p->program->Base.NumTemporaries)
+ p->program->Base.NumTemporaries = bit;
+
+ p->temp_in_use |= 1<<(bit-1);
+ return make_ureg(PROGRAM_TEMPORARY, (bit-1));
+}
+
+static struct ureg get_tex_temp( struct texenv_fragment_program *p )
+{
+ int bit;
+
+ /* First try to find available temp not previously used (to avoid
+ * starting a new texture indirection). According to the spec, the
+ * ~p->temps_output isn't necessary, but will keep it there for
+ * now:
+ */
+ bit = _mesa_ffs( ~p->temp_in_use & ~p->alu_temps & ~p->temps_output );
+
+ /* Then any unused temporary:
+ */
+ if (!bit)
+ bit = _mesa_ffs( ~p->temp_in_use );
+
+ if (!bit) {
+ _mesa_problem(NULL, "%s: out of temporaries\n", __FILE__);
+ exit(1);
+ }
+
+ if ((GLuint) bit > p->program->Base.NumTemporaries)
+ p->program->Base.NumTemporaries = bit;
+
+ p->temp_in_use |= 1<<(bit-1);
+ return make_ureg(PROGRAM_TEMPORARY, (bit-1));
+}
+
+
+/** Mark a temp reg as being no longer allocatable. */
+static void reserve_temp( struct texenv_fragment_program *p, struct ureg r )
+{
+ if (r.file == PROGRAM_TEMPORARY)
+ p->temps_output |= (1 << r.idx);
+}
+
+
+static void release_temps(struct gl_context *ctx, struct texenv_fragment_program *p )
+{
+ GLuint max_temp = ctx->Const.FragmentProgram.MaxTemps;
+
+ /* KW: To support tex_env_crossbar, don't release the registers in
+ * temps_output.
+ */
+ if (max_temp >= sizeof(int) * 8)
+ p->temp_in_use = p->temps_output;
+ else
+ p->temp_in_use = ~((1<<max_temp)-1) | p->temps_output;
+}
+
+
+static struct ureg register_param5( struct texenv_fragment_program *p,
+ GLint s0,
+ GLint s1,
+ GLint s2,
+ GLint s3,
+ GLint s4)
+{
+ int tokens[STATE_LENGTH];
+ GLuint idx;
+ tokens[0] = s0;
+ tokens[1] = s1;
+ tokens[2] = s2;
+ tokens[3] = s3;
+ tokens[4] = s4;
+ idx = _mesa_add_state_reference(p->program->Base.Parameters,
+ (gl_state_index *)tokens);
+ return make_ureg(PROGRAM_STATE_VAR, idx);
+}
+
+
+#define register_param1(p,s0) register_param5(p,s0,0,0,0,0)
+#define register_param2(p,s0,s1) register_param5(p,s0,s1,0,0,0)
+#define register_param3(p,s0,s1,s2) register_param5(p,s0,s1,s2,0,0)
+#define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
+
+static GLuint frag_to_vert_attrib( GLuint attrib )
+{
+ switch (attrib) {
+ case FRAG_ATTRIB_COL0: return VERT_ATTRIB_COLOR0;
+ case FRAG_ATTRIB_COL1: return VERT_ATTRIB_COLOR1;
+ default:
+ assert(attrib >= FRAG_ATTRIB_TEX0);
+ assert(attrib <= FRAG_ATTRIB_TEX7);
+ return attrib - FRAG_ATTRIB_TEX0 + VERT_ATTRIB_TEX0;
+ }
+}
+
+
+static struct ureg register_input( struct texenv_fragment_program *p, GLuint input )
+{
+ if (p->state->inputs_available & (1<<input)) {
+ p->program->Base.InputsRead |= (1 << input);
+ return make_ureg(PROGRAM_INPUT, input);
+ }
+ else {
+ GLuint idx = frag_to_vert_attrib( input );
+ return register_param3( p, STATE_INTERNAL, STATE_CURRENT_ATTRIB, idx );
+ }
+}
+
+
+static void emit_arg( struct prog_src_register *reg,
+ struct ureg ureg )
+{
+ reg->File = ureg.file;
+ reg->Index = ureg.idx;
+ reg->Swizzle = ureg.swz;
+ reg->Negate = ureg.negatebase ? NEGATE_XYZW : NEGATE_NONE;
+ reg->Abs = GL_FALSE;
+}
+
+static void emit_dst( struct prog_dst_register *dst,
+ struct ureg ureg, GLuint mask )
+{
+ dst->File = ureg.file;
+ dst->Index = ureg.idx;
+ dst->WriteMask = mask;
+ dst->CondMask = COND_TR; /* always pass cond test */
+ dst->CondSwizzle = SWIZZLE_NOOP;
+}
+
+static struct prog_instruction *
+emit_op(struct texenv_fragment_program *p,
+ enum prog_opcode op,
+ struct ureg dest,
+ GLuint mask,
+ GLboolean saturate,
+ struct ureg src0,
+ struct ureg src1,
+ struct ureg src2 )
+{
+ const GLuint nr = p->program->Base.NumInstructions++;
+ struct prog_instruction *inst = &p->program->Base.Instructions[nr];
+
+ assert(nr < MAX_INSTRUCTIONS);
+
+ _mesa_init_instructions(inst, 1);
+ inst->Opcode = op;
+
+ emit_arg( &inst->SrcReg[0], src0 );
+ emit_arg( &inst->SrcReg[1], src1 );
+ emit_arg( &inst->SrcReg[2], src2 );
+
+ inst->SaturateMode = saturate ? SATURATE_ZERO_ONE : SATURATE_OFF;
+
+ emit_dst( &inst->DstReg, dest, mask );
+
+#if 0
+ /* Accounting for indirection tracking:
+ */
+ if (dest.file == PROGRAM_TEMPORARY)
+ p->temps_output |= 1 << dest.idx;
+#endif
+
+ return inst;
+}
+
+
+static struct ureg emit_arith( struct texenv_fragment_program *p,
+ enum prog_opcode op,
+ struct ureg dest,
+ GLuint mask,
+ GLboolean saturate,
+ struct ureg src0,
+ struct ureg src1,
+ struct ureg src2 )
+{
+ emit_op(p, op, dest, mask, saturate, src0, src1, src2);
+
+ /* Accounting for indirection tracking:
+ */
+ if (src0.file == PROGRAM_TEMPORARY)
+ p->alu_temps |= 1 << src0.idx;
+
+ if (!is_undef(src1) && src1.file == PROGRAM_TEMPORARY)
+ p->alu_temps |= 1 << src1.idx;
+
+ if (!is_undef(src2) && src2.file == PROGRAM_TEMPORARY)
+ p->alu_temps |= 1 << src2.idx;
+
+ if (dest.file == PROGRAM_TEMPORARY)
+ p->alu_temps |= 1 << dest.idx;
+
+ p->program->Base.NumAluInstructions++;
+ return dest;
+}
+
+static struct ureg emit_texld( struct texenv_fragment_program *p,
+ enum prog_opcode op,
+ struct ureg dest,
+ GLuint destmask,
+ GLuint tex_unit,
+ GLuint tex_idx,
+ GLuint tex_shadow,
+ struct ureg coord )
+{
+ struct prog_instruction *inst = emit_op( p, op,
+ dest, destmask,
+ GL_FALSE, /* don't saturate? */
+ coord, /* arg 0? */
+ undef,
+ undef);
+
+ inst->TexSrcTarget = tex_idx;
+ inst->TexSrcUnit = tex_unit;
+ inst->TexShadow = tex_shadow;
+
+ p->program->Base.NumTexInstructions++;
+
+ /* Accounting for indirection tracking:
+ */
+ reserve_temp(p, dest);
+
+#if 0
+ /* Is this a texture indirection?
+ */
+ if ((coord.file == PROGRAM_TEMPORARY &&
+ (p->temps_output & (1<<coord.idx))) ||
+ (dest.file == PROGRAM_TEMPORARY &&
+ (p->alu_temps & (1<<dest.idx)))) {
+ p->program->Base.NumTexIndirections++;
+ p->temps_output = 1<<coord.idx;
+ p->alu_temps = 0;
+ assert(0); /* KW: texture env crossbar */
+ }
+#endif
+
+ return dest;
+}
+
+
+static struct ureg register_const4f( struct texenv_fragment_program *p,
+ GLfloat s0,
+ GLfloat s1,
+ GLfloat s2,
+ GLfloat s3)
+{
+ GLfloat values[4];
+ GLuint idx, swizzle;
+ struct ureg r;
+ values[0] = s0;
+ values[1] = s1;
+ values[2] = s2;
+ values[3] = s3;
+ idx = _mesa_add_unnamed_constant( p->program->Base.Parameters, values, 4,
+ &swizzle );
+ r = make_ureg(PROGRAM_CONSTANT, idx);
+ r.swz = swizzle;
+ return r;
+}
+
+#define register_scalar_const(p, s0) register_const4f(p, s0, s0, s0, s0)
+#define register_const1f(p, s0) register_const4f(p, s0, 0, 0, 1)
+#define register_const2f(p, s0, s1) register_const4f(p, s0, s1, 0, 1)
+#define register_const3f(p, s0, s1, s2) register_const4f(p, s0, s1, s2, 1)
+
+
+static struct ureg get_one( struct texenv_fragment_program *p )
+{
+ if (is_undef(p->one))
+ p->one = register_scalar_const(p, 1.0);
+ return p->one;
+}
+
+static struct ureg get_half( struct texenv_fragment_program *p )
{
- ir_variable *var;
- ir_dereference *deref;
+ if (is_undef(p->half))
+ p->half = register_scalar_const(p, 0.5);
+ return p->half;
+}
+
+static struct ureg get_zero( struct texenv_fragment_program *p )
+{
+ if (is_undef(p->zero))
+ p->zero = register_scalar_const(p, 0.0);
+ return p->zero;
+}
+
+
+static void program_error( struct texenv_fragment_program *p, const char *msg )
+{
+ _mesa_problem(NULL, "%s", msg);
+ p->error = 1;
+}
+static struct ureg get_source( struct texenv_fragment_program *p,
+ GLuint src, GLuint unit )
+{
switch (src) {
case SRC_TEXTURE:
- return new(p->mem_ctx) ir_dereference_variable(p->src_texture[unit]);
+ assert(!is_undef(p->src_texture[unit]));
+ return p->src_texture[unit];
case SRC_TEXTURE0:
case SRC_TEXTURE1:
@@ -553,69 +931,66 @@ get_source(struct texenv_fragment_program *p,
case SRC_TEXTURE5:
case SRC_TEXTURE6:
case SRC_TEXTURE7:
- return new(p->mem_ctx)
- ir_dereference_variable(p->src_texture[src - SRC_TEXTURE0]);
+ assert(!is_undef(p->src_texture[src - SRC_TEXTURE0]));
+ return p->src_texture[src - SRC_TEXTURE0];
case SRC_CONSTANT:
- var = p->shader->symbols->get_variable("gl_TextureEnvColor");
- assert(var);
- deref = new(p->mem_ctx) ir_dereference_variable(var);
- var->max_array_access = MAX2(var->max_array_access, unit);
- return new(p->mem_ctx) ir_dereference_array(deref,
- new(p->mem_ctx) ir_constant(unit));
+ return register_param2(p, STATE_TEXENV_COLOR, unit);
case SRC_PRIMARY_COLOR:
- var = p->shader->symbols->get_variable("gl_Color");
- assert(var);
- return new(p->mem_ctx) ir_dereference_variable(var);
+ return register_input(p, FRAG_ATTRIB_COL0);
case SRC_ZERO:
- return new(p->mem_ctx) ir_constant(0.0f);
+ return get_zero(p);
case SRC_PREVIOUS:
- if (!p->src_previous) {
- var = p->shader->symbols->get_variable("gl_Color");
- assert(var);
- return new(p->mem_ctx) ir_dereference_variable(var);
- } else {
- return p->src_previous->clone(p->mem_ctx, NULL);
- }
+ if (is_undef(p->src_previous))
+ return register_input(p, FRAG_ATTRIB_COL0);
+ else
+ return p->src_previous;
default:
assert(0);
- return NULL;
+ return undef;
}
}
-static ir_rvalue *
-emit_combine_source(struct texenv_fragment_program *p,
- GLuint unit,
- GLuint source,
- GLuint operand)
+static struct ureg emit_combine_source( struct texenv_fragment_program *p,
+ GLuint mask,
+ GLuint unit,
+ GLuint source,
+ GLuint operand )
{
- ir_rvalue *src;
+ struct ureg arg, src, one;
src = get_source(p, source, unit);
switch (operand) {
case OPR_ONE_MINUS_SRC_COLOR:
- return new(p->mem_ctx) ir_expression(ir_binop_sub,
- new(p->mem_ctx) ir_constant(1.0f),
- src);
+ /* Get unused tmp,
+ * Emit tmp = 1.0 - arg.xyzw
+ */
+ arg = get_temp( p );
+ one = get_one( p );
+ return emit_arith( p, OPCODE_SUB, arg, mask, 0, one, src, undef);
case OPR_SRC_ALPHA:
- return new(p->mem_ctx) ir_swizzle(src, 3, 3, 3, 3, 1);
-
+ if (mask == WRITEMASK_W)
+ return src;
+ else
+ return swizzle1( src, SWIZZLE_W );
case OPR_ONE_MINUS_SRC_ALPHA:
- return new(p->mem_ctx) ir_expression(ir_binop_sub,
- new(p->mem_ctx) ir_constant(1.0f),
- new(p->mem_ctx) ir_swizzle(src,
- 3, 3,
- 3, 3, 1));
+ /* Get unused tmp,
+ * Emit tmp = 1.0 - arg.wwww
+ */
+ arg = get_temp(p);
+ one = get_one(p);
+ return emit_arith(p, OPCODE_SUB, arg, mask, 0,
+ one, swizzle1(src, SWIZZLE_W), undef);
case OPR_ZERO:
- return new(p->mem_ctx) ir_constant(0.0f);
+ return get_zero(p);
case OPR_ONE:
- return new(p->mem_ctx) ir_constant(1.0f);
+ return get_one(p);
case OPR_SRC_COLOR:
return src;
default:
@@ -664,104 +1039,112 @@ static GLboolean args_match( const struct state_key *key, GLuint unit )
return GL_TRUE;
}
-static ir_rvalue *
-smear(struct texenv_fragment_program *p, ir_rvalue *val)
-{
- if (!val->type->is_scalar())
- return val;
-
- return new(p->mem_ctx) ir_swizzle(val, 0, 0, 0, 0, 4);
-}
-
-static ir_rvalue *
-emit_combine(struct texenv_fragment_program *p,
- GLuint unit,
- GLuint nr,
- GLuint mode,
- const struct mode_opt *opt)
+static struct ureg emit_combine( struct texenv_fragment_program *p,
+ struct ureg dest,
+ GLuint mask,
+ GLboolean saturate,
+ GLuint unit,
+ GLuint nr,
+ GLuint mode,
+ const struct mode_opt *opt)
{
- ir_rvalue *src[MAX_COMBINER_TERMS];
- ir_rvalue *tmp0, *tmp1;
+ struct ureg src[MAX_COMBINER_TERMS];
+ struct ureg tmp, half;
GLuint i;
assert(nr <= MAX_COMBINER_TERMS);
for (i = 0; i < nr; i++)
- src[i] = emit_combine_source( p, unit, opt[i].Source, opt[i].Operand );
+ src[i] = emit_combine_source( p, mask, unit, opt[i].Source, opt[i].Operand );
switch (mode) {
case MODE_REPLACE:
- return src[0];
-
+ if (mask == WRITEMASK_XYZW && !saturate)
+ return src[0];
+ else
+ return emit_arith( p, OPCODE_MOV, dest, mask, saturate, src[0], undef, undef );
case MODE_MODULATE:
- return new(p->mem_ctx) ir_expression(ir_binop_mul, src[0], src[1]);
-
+ return emit_arith( p, OPCODE_MUL, dest, mask, saturate,
+ src[0], src[1], undef );
case MODE_ADD:
- return new(p->mem_ctx) ir_expression(ir_binop_add, src[0], src[1]);
-
+ return emit_arith( p, OPCODE_ADD, dest, mask, saturate,
+ src[0], src[1], undef );
case MODE_ADD_SIGNED:
- tmp0 = new(p->mem_ctx) ir_expression(ir_binop_add, src[0], src[1]);
- return new(p->mem_ctx) ir_expression(ir_binop_add, tmp0,
- new(p->mem_ctx) ir_constant(-0.5f));
-
+ /* tmp = arg0 + arg1
+ * result = tmp - .5
+ */
+ half = get_half(p);
+ tmp = get_temp( p );
+ emit_arith( p, OPCODE_ADD, tmp, mask, 0, src[0], src[1], undef );
+ emit_arith( p, OPCODE_SUB, dest, mask, saturate, tmp, half, undef );
+ return dest;
case MODE_INTERPOLATE:
- /* Arg0 * (Arg2) + Arg1 * (1-Arg2) */
- tmp0 = new(p->mem_ctx) ir_expression(ir_binop_mul, src[0], src[2]);
-
- tmp1 = new(p->mem_ctx) ir_expression(ir_binop_sub,
- new(p->mem_ctx) ir_constant(1.0f),
- src[2]->clone(p->mem_ctx, NULL));
- tmp1 = new(p->mem_ctx) ir_expression(ir_binop_mul, src[1], tmp1);
-
- return new(p->mem_ctx) ir_expression(ir_binop_add, tmp0, tmp1);
+ /* Arg0 * (Arg2) + Arg1 * (1-Arg2) -- note arguments are reordered:
+ */
+ return emit_arith( p, OPCODE_LRP, dest, mask, saturate, src[2], src[0], src[1] );
case MODE_SUBTRACT:
- return new(p->mem_ctx) ir_expression(ir_binop_sub, src[0], src[1]);
+ return emit_arith( p, OPCODE_SUB, dest, mask, saturate, src[0], src[1], undef );
case MODE_DOT3_RGBA:
case MODE_DOT3_RGBA_EXT:
case MODE_DOT3_RGB_EXT:
case MODE_DOT3_RGB: {
- tmp0 = new(p->mem_ctx) ir_expression(ir_binop_mul, src[0],
- new(p->mem_ctx) ir_constant(2.0f));
- tmp0 = new(p->mem_ctx) ir_expression(ir_binop_add, tmp0,
- new(p->mem_ctx) ir_constant(-1.0f));
- tmp0 = new(p->mem_ctx) ir_swizzle(smear(p, tmp0), 0, 1, 2, 3, 3);
-
- tmp1 = new(p->mem_ctx) ir_expression(ir_binop_mul, src[1],
- new(p->mem_ctx) ir_constant(2.0f));
- tmp1 = new(p->mem_ctx) ir_expression(ir_binop_add, tmp1,
- new(p->mem_ctx) ir_constant(-1.0f));
- tmp1 = new(p->mem_ctx) ir_swizzle(smear(p, tmp1), 0, 1, 2, 3, 3);
-
- return new(p->mem_ctx) ir_expression(ir_binop_dot, tmp0, tmp1);
- }
- case MODE_MODULATE_ADD_ATI:
- tmp0 = new(p->mem_ctx) ir_expression(ir_binop_mul, src[0], src[2]);
- return new(p->mem_ctx) ir_expression(ir_binop_add, tmp0, src[1]);
+ struct ureg tmp0 = get_temp( p );
+ struct ureg tmp1 = get_temp( p );
+ struct ureg neg1 = register_scalar_const(p, -1);
+ struct ureg two = register_scalar_const(p, 2);
- case MODE_MODULATE_SIGNED_ADD_ATI:
- tmp0 = new(p->mem_ctx) ir_expression(ir_binop_mul, src[0], src[2]);
- tmp0 = new(p->mem_ctx) ir_expression(ir_binop_add, tmp0, src[1]);
- return new(p->mem_ctx) ir_expression(ir_binop_add, tmp0,
- new(p->mem_ctx) ir_constant(-0.5f));
+ /* tmp0 = 2*src0 - 1
+ * tmp1 = 2*src1 - 1
+ *
+ * dst = tmp0 dot3 tmp1
+ */
+ emit_arith( p, OPCODE_MAD, tmp0, WRITEMASK_XYZW, 0,
+ two, src[0], neg1);
+ if (memcmp(&src[0], &src[1], sizeof(struct ureg)) == 0)
+ tmp1 = tmp0;
+ else
+ emit_arith( p, OPCODE_MAD, tmp1, WRITEMASK_XYZW, 0,
+ two, src[1], neg1);
+ emit_arith( p, OPCODE_DP3, dest, mask, saturate, tmp0, tmp1, undef);
+ return dest;
+ }
+ case MODE_MODULATE_ADD_ATI:
+ /* Arg0 * Arg2 + Arg1 */
+ return emit_arith( p, OPCODE_MAD, dest, mask, saturate,
+ src[0], src[2], src[1] );
+ case MODE_MODULATE_SIGNED_ADD_ATI: {
+ /* Arg0 * Arg2 + Arg1 - 0.5 */
+ struct ureg tmp0 = get_temp(p);
+ half = get_half(p);
+ emit_arith( p, OPCODE_MAD, tmp0, mask, 0, src[0], src[2], src[1] );
+ emit_arith( p, OPCODE_SUB, dest, mask, saturate, tmp0, half, undef );
+ return dest;
+ }
case MODE_MODULATE_SUBTRACT_ATI:
- tmp0 = new(p->mem_ctx) ir_expression(ir_binop_mul, src[0], src[2]);
- return new(p->mem_ctx) ir_expression(ir_binop_sub, tmp0, src[1]);
-
+ /* Arg0 * Arg2 - Arg1 */
+ emit_arith( p, OPCODE_MAD, dest, mask, 0, src[0], src[2], negate(src[1]) );
+ return dest;
case MODE_ADD_PRODUCTS:
- tmp0 = new(p->mem_ctx) ir_expression(ir_binop_mul, src[0], src[1]);
- tmp1 = new(p->mem_ctx) ir_expression(ir_binop_mul, src[2], src[3]);
- return new(p->mem_ctx) ir_expression(ir_binop_add, tmp0, tmp1);
-
+ /* 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:
- tmp0 = new(p->mem_ctx) ir_expression(ir_binop_mul, src[0], src[1]);
- tmp1 = new(p->mem_ctx) ir_expression(ir_binop_mul, src[2], src[3]);
- tmp0 = new(p->mem_ctx) ir_expression(ir_binop_add, tmp0, tmp1);
- return new(p->mem_ctx) ir_expression(ir_binop_add, tmp0,
- new(p->mem_ctx) ir_constant(-0.5f));
-
+ /* 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);
@@ -772,24 +1155,17 @@ emit_combine(struct texenv_fragment_program *p,
}
}
-static ir_rvalue *
-saturate(struct texenv_fragment_program *p, ir_rvalue *val)
-{
- val = new(p->mem_ctx) ir_expression(ir_binop_min, val,
- new(p->mem_ctx) ir_constant(1.0f));
- return new(p->mem_ctx) ir_expression(ir_binop_max, val,
- new(p->mem_ctx) ir_constant(0.0f));
-}
/**
* Generate instructions for one texture unit's env/combiner mode.
*/
-static ir_rvalue *
+static struct ureg
emit_texenv(struct texenv_fragment_program *p, GLuint unit)
{
const struct state_key *key = p->state;
GLboolean rgb_saturate, alpha_saturate;
GLuint rgb_shift, alpha_shift;
+ struct ureg out, dest;
if (!key->unit[unit].enabled) {
return get_source(p, SRC_PREVIOUS, 0);
@@ -831,232 +1207,129 @@ emit_texenv(struct texenv_fragment_program *p, GLuint unit)
else
alpha_saturate = GL_FALSE;
- ir_variable *temp_var = new(p->mem_ctx) ir_variable(glsl_type::vec4_type,
- "texenv_combine",
- ir_var_temporary);
- p->instructions->push_tail(temp_var);
-
- ir_dereference *deref;
- ir_assignment *assign;
- ir_rvalue *val;
+ /* If this is the very last calculation (and various other conditions
+ * are met), emit directly to the color output register. Otherwise,
+ * emit to a temporary register.
+ */
+ if (key->separate_specular ||
+ unit != p->last_tex_stage ||
+ alpha_shift ||
+ key->num_draw_buffers != 1 ||
+ rgb_shift)
+ dest = get_temp( p );
+ else
+ dest = make_ureg(PROGRAM_OUTPUT, FRAG_RESULT_COLOR);
/* Emit the RGB and A combine ops
*/
if (key->unit[unit].ModeRGB == key->unit[unit].ModeA &&
args_match(key, unit)) {
- val = emit_combine(p, unit,
- key->unit[unit].NumArgsRGB,
- key->unit[unit].ModeRGB,
- key->unit[unit].OptRGB);
- val = smear(p, val);
- if (rgb_saturate)
- val = saturate(p, val);
-
- deref = new(p->mem_ctx) ir_dereference_variable(temp_var);
- assign = new(p->mem_ctx) ir_assignment(deref, val, NULL);
- p->instructions->push_tail(assign);
+ out = emit_combine( p, dest, WRITEMASK_XYZW, rgb_saturate,
+ unit,
+ key->unit[unit].NumArgsRGB,
+ key->unit[unit].ModeRGB,
+ key->unit[unit].OptRGB);
}
else if (key->unit[unit].ModeRGB == MODE_DOT3_RGBA_EXT ||
key->unit[unit].ModeRGB == MODE_DOT3_RGBA) {
- ir_rvalue *val = emit_combine(p, unit,
- key->unit[unit].NumArgsRGB,
- key->unit[unit].ModeRGB,
- key->unit[unit].OptRGB);
- val = smear(p, val);
- if (rgb_saturate)
- val = saturate(p, val);
- deref = new(p->mem_ctx) ir_dereference_variable(temp_var);
- assign = new(p->mem_ctx) ir_assignment(deref, val, NULL);
- p->instructions->push_tail(assign);
+ out = emit_combine( p, dest, WRITEMASK_XYZW, rgb_saturate,
+ unit,
+ key->unit[unit].NumArgsRGB,
+ key->unit[unit].ModeRGB,
+ key->unit[unit].OptRGB);
}
else {
/* Need to do something to stop from re-emitting identical
* argument calculations here:
*/
- val = emit_combine(p, unit,
- key->unit[unit].NumArgsRGB,
- key->unit[unit].ModeRGB,
- key->unit[unit].OptRGB);
- val = smear(p, val);
- val = new(p->mem_ctx) ir_swizzle(val, 0, 1, 2, 3, 3);
- if (rgb_saturate)
- val = saturate(p, val);
- deref = new(p->mem_ctx) ir_dereference_variable(temp_var);
- assign = new(p->mem_ctx) ir_assignment(deref, val, NULL, WRITEMASK_XYZ);
- p->instructions->push_tail(assign);
-
- val = emit_combine(p, unit,
- key->unit[unit].NumArgsA,
- key->unit[unit].ModeA,
- key->unit[unit].OptA);
- val = smear(p, val);
- val = new(p->mem_ctx) ir_swizzle(val, 3, 3, 3, 3, 1);
- if (alpha_saturate)
- val = saturate(p, val);
- deref = new(p->mem_ctx) ir_dereference_variable(temp_var);
- assign = new(p->mem_ctx) ir_assignment(deref, val, NULL, WRITEMASK_W);
- p->instructions->push_tail(assign);
+ out = emit_combine( p, dest, WRITEMASK_XYZ, rgb_saturate,
+ unit,
+ key->unit[unit].NumArgsRGB,
+ key->unit[unit].ModeRGB,
+ key->unit[unit].OptRGB);
+ out = emit_combine( p, dest, WRITEMASK_W, alpha_saturate,
+ unit,
+ key->unit[unit].NumArgsA,
+ key->unit[unit].ModeA,
+ key->unit[unit].OptA);
}
- deref = new(p->mem_ctx) ir_dereference_variable(temp_var);
-
/* Deal with the final shift:
*/
if (alpha_shift || rgb_shift) {
- ir_constant *shift;
+ struct ureg shift;
+ GLboolean saturate = GL_TRUE; /* always saturate at this point */
if (rgb_shift == alpha_shift) {
- shift = new(p->mem_ctx) ir_constant((float)(1 << rgb_shift));
+ shift = register_scalar_const(p, (GLfloat)(1<<rgb_shift));
}
else {
- float const_data[4] = {
- 1 << rgb_shift,
- 1 << rgb_shift,
- 1 << rgb_shift,
- 1 << alpha_shift
- };
- shift = new(p->mem_ctx) ir_constant(glsl_type::vec4_type,
- (ir_constant_data *)const_data);
+ shift = register_const4f(p,
+ (GLfloat)(1<<rgb_shift),
+ (GLfloat)(1<<rgb_shift),
+ (GLfloat)(1<<rgb_shift),
+ (GLfloat)(1<<alpha_shift));
}
-
- return saturate(p, new(p->mem_ctx) ir_expression(ir_binop_mul,
- deref, shift));
+ return emit_arith( p, OPCODE_MUL, dest, WRITEMASK_XYZW,
+ saturate, out, shift, undef );
}
else
- return deref;
+ return out;
}
/**
* Generate instruction for getting a texture source term.
*/
- static void load_texture( struct texenv_fragment_program *p, GLuint unit )
- {
- ir_dereference *deref;
- ir_assignment *assign;
-
- if (p->src_texture[unit])
- return;
-
- const GLuint texTarget = p->state->unit[unit].source_index;
- ir_rvalue *texcoord;
-
- if (p->texcoord_tex[unit]) {
- texcoord = new(p->mem_ctx) ir_dereference_variable(p->texcoord_tex[unit]);
- }
- else {
- ir_variable *tc_array = p->shader->symbols->get_variable("gl_TexCoord");
- assert(tc_array);
- texcoord = new(p->mem_ctx) ir_dereference_variable(tc_array);
- ir_rvalue *index = new(p->mem_ctx) ir_constant(unit);
- texcoord = new(p->mem_ctx) ir_dereference_array(texcoord, index);
- tc_array->max_array_access = MAX2(tc_array->max_array_access, unit);
- }
-
- if (!p->state->unit[unit].enabled) {
- p->src_texture[unit] = new(p->mem_ctx) ir_variable(glsl_type::vec4_type,
- "dummy_tex",
- ir_var_temporary);
- p->instructions->push_tail(p->src_texture[unit]);
-
- deref = new(p->mem_ctx) ir_dereference_variable(p->src_texture[unit]);
- assign = new(p->mem_ctx) ir_assignment(deref,
- new(p->mem_ctx) ir_constant(0.0f),
- NULL);
- p->instructions->push_tail(assign);
- return ;
- }
-
- const glsl_type *sampler_type = NULL;
- int coords = 0;
-
- switch (texTarget) {
- case TEXTURE_1D_INDEX:
- if (p->state->unit[unit].shadow)
- sampler_type = p->shader->symbols->get_type("sampler1DShadow");
- else
- sampler_type = p->shader->symbols->get_type("sampler1D");
- coords = 1;
- break;
- case TEXTURE_1D_ARRAY_INDEX:
- if (p->state->unit[unit].shadow)
- sampler_type = p->shader->symbols->get_type("sampler1DArrayShadow");
- else
- sampler_type = p->shader->symbols->get_type("sampler1DArray");
- coords = 2;
- break;
- case TEXTURE_2D_INDEX:
- if (p->state->unit[unit].shadow)
- sampler_type = p->shader->symbols->get_type("sampler2DShadow");
- else
- sampler_type = p->shader->symbols->get_type("sampler2D");
- coords = 2;
- break;
- case TEXTURE_2D_ARRAY_INDEX:
- if (p->state->unit[unit].shadow)
- sampler_type = p->shader->symbols->get_type("sampler2DArrayShadow");
- else
- sampler_type = p->shader->symbols->get_type("sampler2DArray");
- coords = 3;
- break;
- case TEXTURE_RECT_INDEX:
- if (p->state->unit[unit].shadow)
- sampler_type = p->shader->symbols->get_type("sampler2DRectShadow");
- else
- sampler_type = p->shader->symbols->get_type("sampler2DRect");
- coords = 2;
- break;
- case TEXTURE_3D_INDEX:
- assert(!p->state->unit[unit].shadow);
- sampler_type = p->shader->symbols->get_type("sampler3D");
- coords = 3;
- break;
- case TEXTURE_CUBE_INDEX:
- if (p->state->unit[unit].shadow)
- sampler_type = p->shader->symbols->get_type("samplerCubeShadow");
- else
- sampler_type = p->shader->symbols->get_type("samplerCube");
- coords = 3;
- break;
- }
-
- p->src_texture[unit] = new(p->mem_ctx) ir_variable(glsl_type::vec4_type,
- "tex",
- ir_var_temporary);
- p->instructions->push_tail(p->src_texture[unit]);
-
- ir_texture *tex = new(p->mem_ctx) ir_texture(ir_tex);
-
-
- char *sampler_name = ralloc_asprintf(p->mem_ctx, "sampler_%d", unit);
- ir_variable *sampler = new(p->mem_ctx) ir_variable(sampler_type,
- sampler_name,
- ir_var_uniform);
- p->top_instructions->push_head(sampler);
- deref = new(p->mem_ctx) ir_dereference_variable(sampler);
- tex->set_sampler(deref);
-
- tex->coordinate = new(p->mem_ctx) ir_swizzle(texcoord, 0, 1, 2, 3, coords);
-
- if (p->state->unit[unit].shadow) {
- texcoord = texcoord->clone(p->mem_ctx, NULL);
- tex->shadow_comparitor = new(p->mem_ctx) ir_swizzle(texcoord,
- coords, 0, 0, 0,
- 1);
- coords++;
- }
-
- texcoord = texcoord->clone(p->mem_ctx, NULL);
- tex->projector = new(p->mem_ctx) ir_swizzle(texcoord, 3, 0, 0, 0, 1);
-
- deref = new(p->mem_ctx) ir_dereference_variable(p->src_texture[unit]);
- assign = new(p->mem_ctx) ir_assignment(deref, tex, NULL);
- p->instructions->push_tail(assign);
- }
+static void load_texture( struct texenv_fragment_program *p, GLuint unit )
+{
+ if (is_undef(p->src_texture[unit])) {
+ const GLuint texTarget = p->state->unit[unit].source_index;
+ struct ureg texcoord;
+ struct ureg tmp = get_tex_temp( p );
-static void
-load_texenv_source(struct texenv_fragment_program *p,
- GLuint src, GLuint unit)
+ 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) {
+ GLboolean shadow = GL_FALSE;
+
+ 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
+ * (see _mesa_init_program_struct()) but let's be safe.
+ */
+ p->program->Base.SamplerUnits[unit] = unit;
+ }
+ else
+ p->src_texture[unit] = get_zero(p);
+
+ if (p->state->unit[unit].texture_cyl_wrap) {
+ /* set flag which is checked by Mesa->Gallium program translation */
+ p->program->Base.InputFlags[0] |= PROG_PARAM_BIT_CYL_WRAP;
+ }
+
+ }
+}
+
+static GLboolean load_texenv_source( struct texenv_fragment_program *p,
+ GLuint src, GLuint unit )
{
switch (src) {
case SRC_TEXTURE:
@@ -1078,6 +1351,8 @@ load_texenv_source(struct texenv_fragment_program *p,
/* not a texture src - do nothing */
break;
}
+
+ return GL_TRUE;
}
@@ -1104,214 +1379,108 @@ load_texunit_sources( struct texenv_fragment_program *p, GLuint unit )
/**
* Generate instructions for loading bump map textures.
*/
-static void
+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;
- ir_rvalue *bump;
- ir_rvalue *texcoord;
- ir_variable *rot_mat_0_var, *rot_mat_1_var;
- ir_dereference_variable *rot_mat_0, *rot_mat_1;
-
- rot_mat_0_var = p->shader->symbols->get_variable("gl_MESABumpRotMatrix0");
- rot_mat_1_var = p->shader->symbols->get_variable("gl_MESABumpRotMatrix1");
- rot_mat_0 = new(p->mem_ctx) ir_dereference_variable(rot_mat_0_var);
- rot_mat_1 = new(p->mem_ctx) ir_dereference_variable(rot_mat_1_var);
-
- ir_variable *tc_array = p->shader->symbols->get_variable("gl_TexCoord");
- assert(tc_array);
- texcoord = new(p->mem_ctx) ir_dereference_variable(tc_array);
- ir_rvalue *index = new(p->mem_ctx) ir_constant(bumpedUnitNr);
- texcoord = new(p->mem_ctx) ir_dereference_array(texcoord, index);
- tc_array->max_array_access = MAX2(tc_array->max_array_access, unit);
+ 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 = Arg1 + (Arg0.xx * rotMat0) + (Arg0.yy * rotMat1)
+ * dest = (Arg0.xxxx * rotMat0 + Arg1) + (Arg0.yyyy * rotMat1)
* note only 2 coords are affected the rest are left unchanged (mul by 0)
*/
- ir_dereference *deref;
- ir_assignment *assign;
- ir_rvalue *bump_x, *bump_y;
-
- texcoord = smear(p, texcoord);
-
- /* bump_texcoord = texcoord */
- ir_variable *bumped = new(p->mem_ctx) ir_variable(texcoord->type,
- "bump_texcoord",
- ir_var_temporary);
- p->instructions->push_tail(bumped);
-
- deref = new(p->mem_ctx) ir_dereference_variable(bumped);
- assign = new(p->mem_ctx) ir_assignment(deref, texcoord, NULL);
- p->instructions->push_tail(assign);
-
- /* bump_texcoord.xy += arg0.x * rotmat0 + arg0.y * rotmat1 */
- bump = get_source(p, key->unit[unit].OptRGB[0].Source, unit);
- bump_x = new(p->mem_ctx) ir_swizzle(bump, 0, 0, 0, 0, 1);
- bump = bump->clone(p->mem_ctx, NULL);
- bump_y = new(p->mem_ctx) ir_swizzle(bump, 1, 0, 0, 0, 1);
-
- bump_x = new(p->mem_ctx) ir_expression(ir_binop_mul, bump_x, rot_mat_0);
- bump_y = new(p->mem_ctx) ir_expression(ir_binop_mul, bump_y, rot_mat_1);
-
- ir_expression *expr;
- expr = new(p->mem_ctx) ir_expression(ir_binop_add, bump_x, bump_y);
-
- deref = new(p->mem_ctx) ir_dereference_variable(bumped);
- expr = new(p->mem_ctx) ir_expression(ir_binop_add,
- new(p->mem_ctx) ir_swizzle(deref,
- 0, 1, 1, 1,
- 2),
- expr);
-
- deref = new(p->mem_ctx) ir_dereference_variable(bumped);
- assign = new(p->mem_ctx) ir_assignment(deref, expr, NULL, WRITEMASK_XY);
- p->instructions->push_tail(assign);
-
- p->texcoord_tex[bumpedUnitNr] = bumped;
+ 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;
}
/**
- * Applies the fog calculations.
- *
- * This is basically like the ARB_fragment_prorgam fog options. Note
- * that ffvertex_prog.c produces fogcoord for us when
- * GL_FOG_COORDINATE_EXT is set to GL_FRAGMENT_DEPTH_EXT.
+ * Generate a new fragment program which implements the context's
+ * current texture env/combine mode.
*/
-static ir_rvalue *
-emit_fog_instructions(struct texenv_fragment_program *p,
- ir_rvalue *fragcolor)
+static void
+create_new_program(struct gl_context *ctx, struct state_key *key,
+ struct gl_fragment_program *program)
{
- struct state_key *key = p->state;
- ir_rvalue *f, *temp;
- ir_variable *params, *oparams;
- ir_variable *fogcoord;
- ir_assignment *assign;
-
- /* Temporary storage for the whole fog result. Fog calculations
- * only affect rgb so we're hanging on to the .a value of fragcolor
- * this way.
- */
- ir_variable *fog_result = new(p->mem_ctx) ir_variable(glsl_type::vec4_type,
- "fog_result",
- ir_var_auto);
- p->instructions->push_tail(fog_result);
- temp = new(p->mem_ctx) ir_dereference_variable(fog_result);
- assign = new(p->mem_ctx) ir_assignment(temp, fragcolor, NULL);
- p->instructions->push_tail(assign);
-
- temp = new(p->mem_ctx) ir_dereference_variable(fog_result);
- fragcolor = new(p->mem_ctx) ir_swizzle(temp, 0, 1, 2, 3, 3);
-
- oparams = p->shader->symbols->get_variable("gl_MESAFogParamsOptimized");
- fogcoord = p->shader->symbols->get_variable("gl_FogFragCoord");
- params = p->shader->symbols->get_variable("gl_Fog");
- f = new(p->mem_ctx) ir_dereference_variable(fogcoord);
-
- ir_variable *f_var = new(p->mem_ctx) ir_variable(glsl_type::float_type,
- "fog_factor", ir_var_auto);
- p->instructions->push_tail(f_var);
-
- switch (key->fog_mode) {
- case FOG_LINEAR:
- /* f = (end - z) / (end - start)
- *
- * gl_MesaFogParamsOptimized gives us (-1 / (end - start)) and
- * (end / (end - start)) so we can generate a single MAD.
- */
- temp = new(p->mem_ctx) ir_dereference_variable(oparams);
- temp = new(p->mem_ctx) ir_swizzle(temp, 0, 0, 0, 0, 1);
- f = new(p->mem_ctx) ir_expression(ir_binop_mul, f, temp);
-
- temp = new(p->mem_ctx) ir_dereference_variable(oparams);
- temp = new(p->mem_ctx) ir_swizzle(temp, 1, 0, 0, 0, 1);
- f = new(p->mem_ctx) ir_expression(ir_binop_add, f, temp);
- break;
- case FOG_EXP:
- /* f = e^(-(density * fogcoord))
- *
- * gl_MesaFogParamsOptimized gives us density/ln(2) so we can
- * use EXP2 which is generally the native instruction without
- * having to do any further math on the fog density uniform.
- */
- temp = new(p->mem_ctx) ir_dereference_variable(oparams);
- temp = new(p->mem_ctx) ir_swizzle(temp, 2, 0, 0, 0, 1);
- f = new(p->mem_ctx) ir_expression(ir_binop_mul, f, temp);
- f = new(p->mem_ctx) ir_expression(ir_unop_neg, f);
- f = new(p->mem_ctx) ir_expression(ir_unop_exp2, f);
- break;
- case FOG_EXP2:
- /* f = e^(-(density * fogcoord)^2)
- *
- * gl_MesaFogParamsOptimized gives us density/sqrt(ln(2)) so we
- * can do this like FOG_EXP but with a squaring after the
- * multiply by density.
- */
- ir_variable *temp_var = new(p->mem_ctx) ir_variable(glsl_type::float_type,
- "fog_temp",
- ir_var_auto);
- p->instructions->push_tail(temp_var);
-
- temp = new(p->mem_ctx) ir_dereference_variable(oparams);
- temp = new(p->mem_ctx) ir_swizzle(temp, 3, 0, 0, 0, 1);
- f = new(p->mem_ctx) ir_expression(ir_binop_mul,
- f, temp);
-
- temp = new(p->mem_ctx) ir_dereference_variable(temp_var);
- ir_assignment *assign = new(p->mem_ctx) ir_assignment(temp, f, NULL);
- p->instructions->push_tail(assign);
-
- f = new(p->mem_ctx) ir_dereference_variable(temp_var);
- temp = new(p->mem_ctx) ir_dereference_variable(temp_var);
- f = new(p->mem_ctx) ir_expression(ir_binop_mul, f, temp);
- f = new(p->mem_ctx) ir_expression(ir_unop_neg, f);
- f = new(p->mem_ctx) ir_expression(ir_unop_exp2, f);
- break;
- }
-
- f = saturate(p, f);
+ struct prog_instruction instBuffer[MAX_INSTRUCTIONS];
+ struct texenv_fragment_program p;
+ GLuint unit;
+ struct ureg cf, out;
+ int i;
- temp = new(p->mem_ctx) ir_dereference_variable(f_var);
- assign = new(p->mem_ctx) ir_assignment(temp, f, NULL);
- p->instructions->push_tail(assign);
+ memset(&p, 0, sizeof(p));
+ p.state = key;
+ p.program = program;
- f = new(p->mem_ctx) ir_dereference_variable(f_var);
- f = new(p->mem_ctx) ir_expression(ir_binop_sub,
- new(p->mem_ctx) ir_constant(1.0f),
- f);
- temp = new(p->mem_ctx) ir_dereference_variable(params);
- temp = new(p->mem_ctx) ir_dereference_record(temp, "color");
- temp = new(p->mem_ctx) ir_swizzle(temp, 0, 1, 2, 3, 3);
- temp = new(p->mem_ctx) ir_expression(ir_binop_mul, temp, f);
+ /* During code generation, use locally-allocated instruction buffer,
+ * then alloc dynamic storage below.
+ */
+ p.program->Base.Instructions = instBuffer;
+ p.program->Base.Target = GL_FRAGMENT_PROGRAM_ARB;
+ 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.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;
+
+ if (key->num_draw_buffers == 1)
+ p.program->Base.OutputsWritten = 1 << FRAG_RESULT_COLOR;
+ else {
+ for (i = 0; i < key->num_draw_buffers; i++)
+ p.program->Base.OutputsWritten |= (1 << (FRAG_RESULT_DATA0 + i));
+ }
- f = new(p->mem_ctx) ir_dereference_variable(f_var);
- f = new(p->mem_ctx) ir_expression(ir_binop_mul, fragcolor, f);
- f = new(p->mem_ctx) ir_expression(ir_binop_add, temp, f);
+ for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) {
+ p.src_texture[unit] = undef;
+ p.texcoord_tex[unit] = undef;
+ }
- ir_dereference *deref = new(p->mem_ctx) ir_dereference_variable(fog_result);
- assign = new(p->mem_ctx) ir_assignment(deref, f, NULL, WRITEMASK_XYZ);
- p->instructions->push_tail(assign);
+ p.src_previous = undef;
+ p.half = undef;
+ p.zero = undef;
+ p.one = undef;
- return new(p->mem_ctx) ir_dereference_variable(fog_result);
-}
+ p.last_tex_stage = 0;
+ release_temps(ctx, &p);
-static void
-emit_instructions(struct texenv_fragment_program *p)
-{
- struct state_key *key = p->state;
- GLuint unit;
+ if (key->enabled_units && key->num_draw_buffers) {
+ GLboolean needbumpstage = GL_FALSE;
- if (key->enabled_units) {
/* Zeroth pass - bump map textures first */
- for (unit = 0; unit < key->nr_enabled_units; unit++) {
+ for (unit = 0; unit < key->nr_enabled_units; unit++)
if (key->unit[unit].enabled &&
key->unit[unit].ModeRGB == MODE_BUMP_ENVMAP_ATI) {
- load_texunit_bumpmap(p, unit);
+ 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
@@ -1319,157 +1488,104 @@ emit_instructions(struct texenv_fragment_program *p)
*/
for (unit = 0; unit < key->nr_enabled_units; unit++)
if (key->unit[unit].enabled) {
- load_texunit_sources(p, unit);
- p->last_tex_stage = unit;
+ load_texunit_sources( &p, unit );
+ p.last_tex_stage = unit;
}
/* Second pass - emit combine instructions to build final color:
*/
- for (unit = 0; unit < key->nr_enabled_units; unit++) {
+ for (unit = 0; unit < key->nr_enabled_units; unit++)
if (key->unit[unit].enabled) {
- p->src_previous = emit_texenv(p, unit);
+ 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 */
}
- }
}
- ir_rvalue *cf = get_source(p, SRC_PREVIOUS, 0);
- ir_dereference_variable *deref;
- ir_assignment *assign;
-
- if (key->separate_specular) {
- ir_rvalue *tmp0, *tmp1;
- ir_variable *spec_result = new(p->mem_ctx) ir_variable(glsl_type::vec4_type,
- "specular_add",
- ir_var_temporary);
-
- p->instructions->push_tail(spec_result);
-
- deref = new(p->mem_ctx) ir_dereference_variable(spec_result);
- assign = new(p->mem_ctx) ir_assignment(deref, cf, NULL);
- p->instructions->push_tail(assign);
-
- deref = new(p->mem_ctx) ir_dereference_variable(spec_result);
- tmp0 = new(p->mem_ctx) ir_swizzle(deref, 0, 1, 2, 3, 3);
+ cf = get_source( &p, SRC_PREVIOUS, 0 );
- ir_variable *secondary =
- p->shader->symbols->get_variable("gl_SecondaryColor");
- assert(secondary);
- deref = new(p->mem_ctx) ir_dereference_variable(secondary);
- tmp1 = new(p->mem_ctx) ir_swizzle(deref, 0, 1, 2, 3, 3);
-
- tmp0 = new(p->mem_ctx) ir_expression(ir_binop_add,
- tmp0, tmp1);
-
- deref = new(p->mem_ctx) ir_dereference_variable(spec_result);
- assign = new(p->mem_ctx) ir_assignment(deref, tmp0, NULL, WRITEMASK_XYZ);
- p->instructions->push_tail(assign);
+ for (i = 0; i < key->num_draw_buffers; i++) {
+ if (key->num_draw_buffers == 1)
+ out = make_ureg( PROGRAM_OUTPUT, FRAG_RESULT_COLOR );
+ else {
+ out = make_ureg( PROGRAM_OUTPUT, FRAG_RESULT_DATA0 + i );
+ }
- cf = new(p->mem_ctx) ir_dereference_variable(spec_result);
+ if (key->separate_specular) {
+ /* Emit specular add.
+ */
+ struct ureg s = register_input(&p, FRAG_ATTRIB_COL1);
+ emit_arith( &p, OPCODE_ADD, out, WRITEMASK_XYZ, 0, cf, s, undef );
+ emit_arith( &p, OPCODE_MOV, out, WRITEMASK_W, 0, cf, undef, undef );
+ }
+ else if (memcmp(&cf, &out, sizeof(cf)) != 0) {
+ /* Will wind up in here if no texture enabled or a couple of
+ * other scenarios (GL_REPLACE for instance).
+ */
+ emit_arith( &p, OPCODE_MOV, out, WRITEMASK_XYZW, 0, cf, undef, undef );
+ }
}
+ /* Finish up:
+ */
+ emit_arith( &p, OPCODE_END, undef, WRITEMASK_XYZW, 0, undef, undef, undef);
if (key->fog_enabled) {
- cf = emit_fog_instructions(p, cf);
+ /* Pull fog mode from struct gl_context, the value in the state key is
+ * a reduced value and not what is expected in FogOption
+ */
+ p.program->FogOption = ctx->Fog.Mode;
+ p.program->Base.InputsRead |= FRAG_BIT_FOGC;
}
-
- ir_variable *frag_color = p->shader->symbols->get_variable("gl_FragColor");
- assert(frag_color);
- deref = new(p->mem_ctx) ir_dereference_variable(frag_color);
- assign = new(p->mem_ctx) ir_assignment(deref, cf, NULL);
- p->instructions->push_tail(assign);
-}
-
-/**
- * Generate a new fragment program which implements the context's
- * current texture env/combine mode.
- */
-static struct gl_shader_program *
-create_new_program(struct gl_context *ctx, struct state_key *key)
-{
- struct texenv_fragment_program p;
- unsigned int unit;
- _mesa_glsl_parse_state *state;
-
- memset(&p, 0, sizeof(p));
- p.mem_ctx = ralloc_context(NULL);
- p.shader = ctx->Driver.NewShader(ctx, 0, GL_FRAGMENT_SHADER);
- p.shader->ir = new(p.shader) exec_list;
- state = new(p.shader) _mesa_glsl_parse_state(ctx, GL_FRAGMENT_SHADER,
- p.shader);
- p.shader->symbols = state->symbols;
- p.top_instructions = p.shader->ir;
- p.instructions = p.shader->ir;
- p.state = key;
- p.shader_program = ctx->Driver.NewShaderProgram(ctx, 0);
-
- state->language_version = 120;
- _mesa_glsl_initialize_types(state);
- _mesa_glsl_initialize_variables(p.instructions, state);
-
- for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) {
- p.src_texture[unit] = NULL;
- p.texcoord_tex[unit] = NULL;
+ else {
+ p.program->FogOption = GL_NONE;
}
- p.src_previous = NULL;
+ if (p.program->Base.NumTexIndirections > ctx->Const.FragmentProgram.MaxTexIndirections)
+ program_error(&p, "Exceeded max nr indirect texture lookups");
- p.last_tex_stage = 0;
+ if (p.program->Base.NumTexInstructions > ctx->Const.FragmentProgram.MaxTexInstructions)
+ program_error(&p, "Exceeded max TEX instructions");
- ir_function *main_f = new(p.mem_ctx) ir_function("main");
- p.instructions->push_tail(main_f);
- state->symbols->add_function(main_f);
+ if (p.program->Base.NumAluInstructions > ctx->Const.FragmentProgram.MaxAluInstructions)
+ program_error(&p, "Exceeded max ALU instructions");
- ir_function_signature *main_sig =
- new(p.mem_ctx) ir_function_signature(p.shader->symbols->get_type("void"));
- main_sig->is_defined = true;
- main_f->add_signature(main_sig);
+ ASSERT(p.program->Base.NumInstructions <= MAX_INSTRUCTIONS);
- p.instructions = &main_sig->body;
- if (key->num_draw_buffers)
- emit_instructions(&p);
-
- validate_ir_tree(p.shader->ir);
-
- while (do_common_optimization(p.shader->ir, false, 32))
- ;
- reparent_ir(p.shader->ir, p.shader->ir);
+ /* Allocate final instruction array */
+ p.program->Base.Instructions
+ = _mesa_alloc_instructions(p.program->Base.NumInstructions);
+ if (!p.program->Base.Instructions) {
+ _mesa_error(ctx, GL_OUT_OF_MEMORY,
+ "generating tex env program");
+ return;
+ }
+ _mesa_copy_instructions(p.program->Base.Instructions, instBuffer,
+ p.program->Base.NumInstructions);
- p.shader->CompileStatus = true;
- p.shader->Version = state->language_version;
- p.shader->num_builtins_to_link = state->num_builtins_to_link;
- p.shader_program->Shaders =
- (gl_shader **)malloc(sizeof(*p.shader_program->Shaders));
- p.shader_program->Shaders[0] = p.shader;
- p.shader_program->NumShaders = 1;
+ if (key->num_draw_buffers && p.program->FogOption) {
+ _mesa_append_fog_code(ctx, p.program);
+ p.program->FogOption = GL_NONE;
+ }
- _mesa_glsl_link_shader(ctx, p.shader_program);
- /* Set the sampler uniforms, and relink to get them into the linked
- * program.
+ /* Notify driver the fragment program has (actually) changed.
*/
- struct gl_fragment_program *fp = p.shader_program->FragmentProgram;
- for (unsigned int i = 0; i < MAX_TEXTURE_UNITS; i++) {
- char *name = ralloc_asprintf(p.mem_ctx, "sampler_%d", i);
- int loc = _mesa_get_uniform_location(ctx, p.shader_program, name);
- if (loc != -1) {
- /* Avoid using _mesa_uniform() because it flags state
- * updates, so if we're generating this shader_program in a
- * state update, we end up recursing. Instead, just set the
- * value, which is picked up at re-link.
- */
- loc = (loc & 0xffff) + (loc >> 16);
- int sampler = fp->Base.Parameters->ParameterValues[loc][0];
- fp->Base.SamplerUnits[sampler] = i;
- }
+ if (ctx->Driver.ProgramStringNotify) {
+ GLboolean ok = ctx->Driver.ProgramStringNotify(ctx,
+ GL_FRAGMENT_PROGRAM_ARB,
+ &p.program->Base);
+ /* Driver should be able to handle any texenv programs as long as
+ * the driver correctly reported max number of texture units correctly,
+ * etc.
+ */
+ ASSERT(ok);
+ (void) ok; /* silence unused var warning */
}
- _mesa_update_shader_textures_used(&fp->Base);
- (void) ctx->Driver.ProgramStringNotify(ctx, fp->Base.Target, &fp->Base);
- if (!p.shader_program->LinkStatus)
- _mesa_problem(ctx, "Failed to link fixed function fragment shader: %s\n",
- p.shader_program->InfoLog);
-
- ralloc_free(p.mem_ctx);
- return p.shader_program;
+ if (DISASSEM) {
+ _mesa_print_program(&p.program->Base);
+ printf("\n");
+ }
}
extern "C" {
@@ -1478,27 +1594,30 @@ extern "C" {
* Return a fragment program which implements the current
* fixed-function texture, fog and color-sum operations.
*/
-struct gl_shader_program *
+struct gl_fragment_program *
_mesa_get_fixed_func_fragment_program(struct gl_context *ctx)
{
- struct gl_shader_program *shader_program;
+ struct gl_fragment_program *prog;
struct state_key key;
GLuint keySize;
-
+
keySize = make_state_key(ctx, &key);
-
- shader_program = (struct gl_shader_program *)
+
+ prog = (struct gl_fragment_program *)
_mesa_search_program_cache(ctx->FragmentProgram.Cache,
&key, keySize);
- if (!shader_program) {
- shader_program = create_new_program(ctx, &key);
+ if (!prog) {
+ prog = (struct gl_fragment_program *)
+ ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0);
+
+ create_new_program(ctx, &key, prog);
- _mesa_shader_cache_insert(ctx, ctx->FragmentProgram.Cache,
- &key, keySize, shader_program);
+ _mesa_program_cache_insert(ctx, ctx->FragmentProgram.Cache,
+ &key, keySize, &prog->Base);
}
- return shader_program;
+ return prog;
}
}
diff --git a/mesalib/src/mesa/main/mtypes.h b/mesalib/src/mesa/main/mtypes.h
index 520d96689..39b6f72cc 100644
--- a/mesalib/src/mesa/main/mtypes.h
+++ b/mesalib/src/mesa/main/mtypes.h
@@ -2191,7 +2191,6 @@ struct gl_shader_state
struct gl_shader_program *CurrentVertexProgram;
struct gl_shader_program *CurrentGeometryProgram;
struct gl_shader_program *CurrentFragmentProgram;
- struct gl_shader_program *_CurrentFragmentProgram;
/**
* Program used by glUniform calls.
diff --git a/mesalib/src/mesa/main/state.c b/mesalib/src/mesa/main/state.c
index 5651e3263..f50f2af1e 100644
--- a/mesalib/src/mesa/main/state.c
+++ b/mesalib/src/mesa/main/state.c
@@ -43,7 +43,6 @@
#include "pixel.h"
#include "program/program.h"
#include "program/prog_parameter.h"
-#include "shaderobj.h"
#include "state.h"
#include "stencil.h"
#include "texenvprogram.h"
@@ -250,7 +249,7 @@ update_program(struct gl_context *ctx)
{
const struct gl_shader_program *vsProg = ctx->Shader.CurrentVertexProgram;
const struct gl_shader_program *gsProg = ctx->Shader.CurrentGeometryProgram;
- struct gl_shader_program *fsProg = ctx->Shader.CurrentFragmentProgram;
+ const struct gl_shader_program *fsProg = ctx->Shader.CurrentFragmentProgram;
const struct gl_vertex_program *prevVP = ctx->VertexProgram._Current;
const struct gl_fragment_program *prevFP = ctx->FragmentProgram._Current;
const struct gl_geometry_program *prevGP = ctx->GeometryProgram._Current;
@@ -276,31 +275,23 @@ update_program(struct gl_context *ctx)
/* Use shader programs */
_mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current,
fsProg->FragmentProgram);
- _mesa_reference_shader_program(ctx, &ctx->Shader._CurrentFragmentProgram,
- fsProg);
}
else if (ctx->FragmentProgram._Enabled) {
- /* use user-defined fragment program */
+ /* use user-defined vertex program */
_mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current,
ctx->FragmentProgram.Current);
- _mesa_reference_shader_program(ctx, &ctx->Shader._CurrentFragmentProgram,
- NULL);
}
else if (ctx->FragmentProgram._MaintainTexEnvProgram) {
/* Use fragment program generated from fixed-function state.
*/
- struct gl_shader_program *f = _mesa_get_fixed_func_fragment_program(ctx);
- _mesa_reference_shader_program(ctx,
- &ctx->Shader._CurrentFragmentProgram, f);
-
_mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current,
- f->FragmentProgram);
+ _mesa_get_fixed_func_fragment_program(ctx));
+ _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._TexEnvProgram,
+ ctx->FragmentProgram._Current);
}
else {
/* no fragment program */
_mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, NULL);
- _mesa_reference_shader_program(ctx, &ctx->Shader._CurrentFragmentProgram,
- NULL);
}
if (gsProg && gsProg->LinkStatus && gsProg->GeometryProgram) {
diff --git a/mesalib/src/mesa/main/texenvprogram.h b/mesalib/src/mesa/main/texenvprogram.h
index dba775feb..0895ebacb 100644
--- a/mesalib/src/mesa/main/texenvprogram.h
+++ b/mesalib/src/mesa/main/texenvprogram.h
@@ -29,7 +29,7 @@
struct gl_context;
-extern struct gl_shader_program *
+extern struct gl_fragment_program *
_mesa_get_fixed_func_fragment_program(struct gl_context *ctx);
#endif
diff --git a/mesalib/src/mesa/program/program.c b/mesalib/src/mesa/program/program.c
index 43f894a9b..6c97787e8 100644
--- a/mesalib/src/mesa/program/program.c
+++ b/mesalib/src/mesa/program/program.c
@@ -140,7 +140,7 @@ _mesa_free_program_data(struct gl_context *ctx)
#endif
#if FEATURE_NV_fragment_program || FEATURE_ARB_fragment_program
_mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, NULL);
- _mesa_delete_shader_cache(ctx, ctx->FragmentProgram.Cache);
+ _mesa_delete_program_cache(ctx, ctx->FragmentProgram.Cache);
#endif
#if FEATURE_ARB_geometry_shader4
_mesa_reference_geomprog(ctx, &ctx->GeometryProgram.Current, NULL);
diff --git a/mesalib/src/mesa/state_tracker/st_cb_clear.c b/mesalib/src/mesa/state_tracker/st_cb_clear.c
index 0e0c4326e..0130c7a5a 100644
--- a/mesalib/src/mesa/state_tracker/st_cb_clear.c
+++ b/mesalib/src/mesa/state_tracker/st_cb_clear.c
@@ -1,579 +1,601 @@
-/**************************************************************************
- *
- * 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
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- **************************************************************************/
-
- /*
- * Authors:
- * Keith Whitwell <keith@tungstengraphics.com>
- * Brian Paul
- * Michel Dänzer
- */
-
-#include "main/glheader.h"
-#include "main/formats.h"
-#include "main/macros.h"
-#include "program/prog_instruction.h"
-#include "st_context.h"
-#include "st_atom.h"
-#include "st_cb_accum.h"
-#include "st_cb_clear.h"
-#include "st_cb_fbo.h"
-#include "st_format.h"
-#include "st_program.h"
-
-#include "pipe/p_context.h"
-#include "pipe/p_shader_tokens.h"
-#include "pipe/p_state.h"
-#include "pipe/p_defines.h"
-#include "util/u_format.h"
-#include "util/u_inlines.h"
-#include "util/u_simple_shaders.h"
-#include "util/u_draw_quad.h"
-
-#include "cso_cache/cso_context.h"
-
-
-/**
- * Do per-context initialization for glClear.
- */
-void
-st_init_clear(struct st_context *st)
-{
- struct pipe_context *pipe = st->pipe;
- struct pipe_screen *pscreen = st->pipe->screen;
-
- memset(&st->clear, 0, sizeof(st->clear));
-
- st->clear.raster.gl_rasterization_rules = 1;
- st->clear.enable_ds_separate = pscreen->get_param(pscreen, PIPE_CAP_DEPTHSTENCIL_CLEAR_SEPARATE);
-
- /* fragment shader state: color pass-through program */
- st->clear.fs = util_make_fragment_passthrough_shader(pipe);
-
- /* vertex shader state: color/position pass-through */
- {
- const uint semantic_names[] = { TGSI_SEMANTIC_POSITION,
- TGSI_SEMANTIC_COLOR };
- const uint semantic_indexes[] = { 0, 0 };
- st->clear.vs = util_make_vertex_passthrough_shader(pipe, 2,
- semantic_names,
- semantic_indexes);
- }
-}
-
-
-/**
- * Free per-context state for glClear.
- */
-void
-st_destroy_clear(struct st_context *st)
-{
- if (st->clear.fs) {
- cso_delete_fragment_shader(st->cso_context, st->clear.fs);
- st->clear.fs = NULL;
- }
- if (st->clear.vs) {
- cso_delete_vertex_shader(st->cso_context, st->clear.vs);
- st->clear.vs = NULL;
- }
- if (st->clear.vbuf) {
- pipe_resource_reference(&st->clear.vbuf, NULL);
- st->clear.vbuf = NULL;
- }
-}
-
-
-/**
- * Draw a screen-aligned quadrilateral.
- * Coords are clip coords with y=0=bottom.
- */
-static void
-draw_quad(struct st_context *st,
- float x0, float y0, float x1, float y1, GLfloat z,
- const GLfloat color[4])
-{
- struct pipe_context *pipe = st->pipe;
-
- /* XXX: Need to improve buffer_write to allow NO_WAIT (as well as
- * no_flush) updates to buffers where we know there is no conflict
- * with previous data. Currently using max_slots > 1 will cause
- * synchronous rendering if the driver flushes its command buffers
- * between one bitmap and the next. Our flush hook below isn't
- * sufficient to catch this as the driver doesn't tell us when it
- * flushes its own command buffers. Until this gets fixed, pay the
- * price of allocating a new buffer for each bitmap cache-flush to
- * avoid synchronous rendering.
- */
- const GLuint max_slots = 1; /* 1024 / sizeof(st->clear.vertices); */
- GLuint i;
-
- if (st->clear.vbuf_slot >= max_slots) {
- pipe_resource_reference(&st->clear.vbuf, NULL);
- st->clear.vbuf_slot = 0;
- }
-
- if (!st->clear.vbuf) {
- st->clear.vbuf = pipe_buffer_create(pipe->screen,
- PIPE_BIND_VERTEX_BUFFER,
- PIPE_USAGE_STREAM,
- max_slots * sizeof(st->clear.vertices));
- }
-
- /* positions */
- st->clear.vertices[0][0][0] = x0;
- st->clear.vertices[0][0][1] = y0;
-
- st->clear.vertices[1][0][0] = x1;
- st->clear.vertices[1][0][1] = y0;
-
- st->clear.vertices[2][0][0] = x1;
- st->clear.vertices[2][0][1] = y1;
-
- st->clear.vertices[3][0][0] = x0;
- st->clear.vertices[3][0][1] = y1;
-
- /* same for all verts: */
- for (i = 0; i < 4; i++) {
- st->clear.vertices[i][0][2] = z;
- st->clear.vertices[i][0][3] = 1.0;
- st->clear.vertices[i][1][0] = color[0];
- st->clear.vertices[i][1][1] = color[1];
- st->clear.vertices[i][1][2] = color[2];
- st->clear.vertices[i][1][3] = color[3];
- }
-
- /* put vertex data into vbuf */
- pipe_buffer_write_nooverlap(st->pipe, st->clear.vbuf,
- st->clear.vbuf_slot
- * sizeof(st->clear.vertices),
- sizeof(st->clear.vertices),
- st->clear.vertices);
-
- /* draw */
- util_draw_vertex_buffer(pipe,
- st->cso_context,
- st->clear.vbuf,
- st->clear.vbuf_slot * sizeof(st->clear.vertices),
- PIPE_PRIM_TRIANGLE_FAN,
- 4, /* verts */
- 2); /* attribs/vert */
-
- /* Increment slot */
- st->clear.vbuf_slot++;
-}
-
-
-
-/**
- * Do glClear by drawing a quadrilateral.
- * The vertices of the quad will be computed from the
- * ctx->DrawBuffer->_X/Ymin/max fields.
- */
-static void
-clear_with_quad(struct gl_context *ctx,
- GLboolean color, GLboolean depth, GLboolean stencil)
-{
- struct st_context *st = st_context(ctx);
- const struct gl_framebuffer *fb = ctx->DrawBuffer;
- const GLfloat fb_width = (GLfloat) fb->Width;
- const GLfloat fb_height = (GLfloat) fb->Height;
- const GLfloat x0 = (GLfloat) ctx->DrawBuffer->_Xmin / fb_width * 2.0f - 1.0f;
- const GLfloat x1 = (GLfloat) ctx->DrawBuffer->_Xmax / fb_width * 2.0f - 1.0f;
- const GLfloat y0 = (GLfloat) ctx->DrawBuffer->_Ymin / fb_height * 2.0f - 1.0f;
- const GLfloat y1 = (GLfloat) ctx->DrawBuffer->_Ymax / fb_height * 2.0f - 1.0f;
- float clearColor[4];
-
- /*
- printf("%s %s%s%s %f,%f %f,%f\n", __FUNCTION__,
- color ? "color, " : "",
- depth ? "depth, " : "",
- stencil ? "stencil" : "",
- x0, y0,
- x1, y1);
- */
-
- cso_save_blend(st->cso_context);
- cso_save_stencil_ref(st->cso_context);
- cso_save_depth_stencil_alpha(st->cso_context);
- cso_save_rasterizer(st->cso_context);
- cso_save_viewport(st->cso_context);
- cso_save_clip(st->cso_context);
- cso_save_fragment_shader(st->cso_context);
- cso_save_vertex_shader(st->cso_context);
- cso_save_vertex_elements(st->cso_context);
- cso_save_vertex_buffers(st->cso_context);
-
- /* blend state: RGBA masking */
- {
- struct pipe_blend_state blend;
- memset(&blend, 0, sizeof(blend));
- blend.rt[0].rgb_src_factor = PIPE_BLENDFACTOR_ONE;
- blend.rt[0].alpha_src_factor = PIPE_BLENDFACTOR_ONE;
- blend.rt[0].rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
- blend.rt[0].alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
- if (color) {
- if (ctx->Color.ColorMask[0][0])
- blend.rt[0].colormask |= PIPE_MASK_R;
- if (ctx->Color.ColorMask[0][1])
- blend.rt[0].colormask |= PIPE_MASK_G;
- if (ctx->Color.ColorMask[0][2])
- blend.rt[0].colormask |= PIPE_MASK_B;
- if (ctx->Color.ColorMask[0][3])
- blend.rt[0].colormask |= PIPE_MASK_A;
- if (st->ctx->Color.DitherFlag)
- blend.dither = 1;
- }
- cso_set_blend(st->cso_context, &blend);
- }
-
- /* depth_stencil state: always pass/set to ref value */
- {
- struct pipe_depth_stencil_alpha_state depth_stencil;
- memset(&depth_stencil, 0, sizeof(depth_stencil));
- if (depth) {
- depth_stencil.depth.enabled = 1;
- depth_stencil.depth.writemask = 1;
- depth_stencil.depth.func = PIPE_FUNC_ALWAYS;
- }
-
- if (stencil) {
- struct pipe_stencil_ref stencil_ref;
- memset(&stencil_ref, 0, sizeof(stencil_ref));
- depth_stencil.stencil[0].enabled = 1;
- depth_stencil.stencil[0].func = PIPE_FUNC_ALWAYS;
- depth_stencil.stencil[0].fail_op = PIPE_STENCIL_OP_REPLACE;
- depth_stencil.stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE;
- depth_stencil.stencil[0].zfail_op = PIPE_STENCIL_OP_REPLACE;
- depth_stencil.stencil[0].valuemask = 0xff;
- depth_stencil.stencil[0].writemask = ctx->Stencil.WriteMask[0] & 0xff;
- stencil_ref.ref_value[0] = ctx->Stencil.Clear;
- cso_set_stencil_ref(st->cso_context, &stencil_ref);
- }
-
- cso_set_depth_stencil_alpha(st->cso_context, &depth_stencil);
- }
-
- cso_set_vertex_elements(st->cso_context, 2, st->velems_util_draw);
-
- cso_set_rasterizer(st->cso_context, &st->clear.raster);
-
- /* viewport state: viewport matching window dims */
- {
- const GLboolean invert = (st_fb_orientation(fb) == Y_0_TOP);
- struct pipe_viewport_state vp;
- vp.scale[0] = 0.5f * fb_width;
- vp.scale[1] = fb_height * (invert ? -0.5f : 0.5f);
- vp.scale[2] = 1.0f;
- vp.scale[3] = 1.0f;
- vp.translate[0] = 0.5f * fb_width;
- vp.translate[1] = 0.5f * fb_height;
- vp.translate[2] = 0.0f;
- vp.translate[3] = 0.0f;
- cso_set_viewport(st->cso_context, &vp);
- }
-
- cso_set_clip(st->cso_context, &st->clear.clip);
- cso_set_fragment_shader_handle(st->cso_context, st->clear.fs);
- cso_set_vertex_shader_handle(st->cso_context, st->clear.vs);
-
- if (ctx->DrawBuffer->_ColorDrawBuffers[0]) {
- st_translate_color(ctx->Color.ClearColor,
- ctx->DrawBuffer->_ColorDrawBuffers[0]->_BaseFormat,
- clearColor);
- }
-
- /* draw quad matching scissor rect */
- draw_quad(st, x0, y0, x1, y1, (GLfloat) ctx->Depth.Clear, clearColor);
-
- /* Restore pipe state */
- cso_restore_blend(st->cso_context);
- cso_restore_stencil_ref(st->cso_context);
- cso_restore_depth_stencil_alpha(st->cso_context);
- cso_restore_rasterizer(st->cso_context);
- cso_restore_viewport(st->cso_context);
- cso_restore_clip(st->cso_context);
- cso_restore_fragment_shader(st->cso_context);
- cso_restore_vertex_shader(st->cso_context);
- cso_restore_vertex_elements(st->cso_context);
- cso_restore_vertex_buffers(st->cso_context);
-}
-
-
-/**
- * Determine if we need to clear the depth buffer by drawing a quad.
- */
-static INLINE GLboolean
-check_clear_color_with_quad(struct gl_context *ctx, struct gl_renderbuffer *rb)
-{
- if (ctx->Scissor.Enabled &&
- (ctx->Scissor.X != 0 ||
- ctx->Scissor.Y != 0 ||
- ctx->Scissor.Width < rb->Width ||
- ctx->Scissor.Height < rb->Height))
- return GL_TRUE;
-
- if (!ctx->Color.ColorMask[0][0] ||
- !ctx->Color.ColorMask[0][1] ||
- !ctx->Color.ColorMask[0][2] ||
- !ctx->Color.ColorMask[0][3])
- return GL_TRUE;
-
- return GL_FALSE;
-}
-
-
-/**
- * Determine if we need to clear the combiend depth/stencil buffer by
- * drawing a quad.
- */
-static INLINE GLboolean
-check_clear_depth_stencil_with_quad(struct gl_context *ctx, struct gl_renderbuffer *rb)
-{
- const GLuint stencilMax = 0xff;
- GLboolean maskStencil
- = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax;
-
- assert(rb->Format == MESA_FORMAT_S8 ||
- rb->Format == MESA_FORMAT_Z24_S8 ||
- rb->Format == MESA_FORMAT_S8_Z24);
-
- if (ctx->Scissor.Enabled &&
- (ctx->Scissor.X != 0 ||
- ctx->Scissor.Y != 0 ||
- ctx->Scissor.Width < rb->Width ||
- ctx->Scissor.Height < rb->Height))
- return GL_TRUE;
-
- if (maskStencil)
- return GL_TRUE;
-
- return GL_FALSE;
-}
-
-
-/**
- * Determine if we need to clear the depth buffer by drawing a quad.
- */
-static INLINE GLboolean
-check_clear_depth_with_quad(struct gl_context *ctx, struct gl_renderbuffer *rb,
- boolean ds_separate)
-{
- const struct st_renderbuffer *strb = st_renderbuffer(rb);
- const GLboolean isDS = util_format_is_depth_and_stencil(strb->surface->format);
-
- if (ctx->Scissor.Enabled &&
- (ctx->Scissor.X != 0 ||
- ctx->Scissor.Y != 0 ||
- ctx->Scissor.Width < rb->Width ||
- ctx->Scissor.Height < rb->Height))
- return GL_TRUE;
-
- if (!ds_separate && isDS && ctx->DrawBuffer->Visual.stencilBits > 0)
- return GL_TRUE;
-
- return GL_FALSE;
-}
-
-
-/**
- * Determine if we need to clear the stencil buffer by drawing a quad.
- */
-static INLINE GLboolean
-check_clear_stencil_with_quad(struct gl_context *ctx, struct gl_renderbuffer *rb,
- boolean ds_separate)
-{
- const struct st_renderbuffer *strb = st_renderbuffer(rb);
- const GLboolean isDS = util_format_is_depth_and_stencil(strb->surface->format);
- const GLuint stencilMax = 0xff;
- const GLboolean maskStencil
- = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax;
-
- assert(rb->Format == MESA_FORMAT_S8 ||
- rb->Format == MESA_FORMAT_Z24_S8 ||
- rb->Format == MESA_FORMAT_S8_Z24);
-
- if (maskStencil)
- return GL_TRUE;
-
- if (ctx->Scissor.Enabled &&
- (ctx->Scissor.X != 0 ||
- ctx->Scissor.Y != 0 ||
- ctx->Scissor.Width < rb->Width ||
- ctx->Scissor.Height < rb->Height))
- return GL_TRUE;
-
- /* This is correct, but it is necessary to look at the depth clear
- * value held in the surface when it comes time to issue the clear,
- * rather than taking depth and stencil clear values from the
- * current state.
- */
- if (!ds_separate && isDS && ctx->DrawBuffer->Visual.depthBits > 0)
- return GL_TRUE;
-
- return GL_FALSE;
-}
-
-
-
-/**
- * Called when we need to flush.
- */
-void
-st_flush_clear(struct st_context *st)
-{
- /* Release vertex buffer to avoid synchronous rendering if we were
- * to map it in the next frame.
- */
- pipe_resource_reference(&st->clear.vbuf, NULL);
- st->clear.vbuf_slot = 0;
-}
-
-
-
-/**
- * Called via ctx->Driver.Clear()
- */
-static void
-st_Clear(struct gl_context *ctx, GLbitfield mask)
-{
- static const GLbitfield BUFFER_BITS_DS
- = (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL);
- struct st_context *st = st_context(ctx);
- struct gl_renderbuffer *depthRb
- = ctx->DrawBuffer->Attachment[BUFFER_DEPTH].Renderbuffer;
- struct gl_renderbuffer *stencilRb
- = ctx->DrawBuffer->Attachment[BUFFER_STENCIL].Renderbuffer;
- GLbitfield quad_buffers = 0x0;
- GLbitfield clear_buffers = 0x0;
- GLuint i;
-
- /* This makes sure the pipe has the latest scissor, etc values */
- st_validate_state( st );
-
- if (mask & BUFFER_BITS_COLOR) {
- for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++) {
- GLuint b = ctx->DrawBuffer->_ColorDrawBufferIndexes[i];
-
- if (mask & (1 << b)) {
- struct gl_renderbuffer *rb
- = ctx->DrawBuffer->Attachment[b].Renderbuffer;
- struct st_renderbuffer *strb = st_renderbuffer(rb);
-
- if (!strb || !strb->surface)
- continue;
-
- if (check_clear_color_with_quad( ctx, rb ))
- quad_buffers |= PIPE_CLEAR_COLOR;
- else
- clear_buffers |= PIPE_CLEAR_COLOR;
- }
- }
- }
-
- if ((mask & BUFFER_BITS_DS) == BUFFER_BITS_DS && depthRb == stencilRb) {
- /* clearing combined depth + stencil */
- struct st_renderbuffer *strb = st_renderbuffer(depthRb);
-
- if (strb->surface) {
- if (check_clear_depth_stencil_with_quad(ctx, depthRb))
- quad_buffers |= PIPE_CLEAR_DEPTHSTENCIL;
- else
- clear_buffers |= PIPE_CLEAR_DEPTHSTENCIL;
- }
- }
- else {
- /* separate depth/stencil clears */
- /* I don't think truly separate buffers are actually possible in gallium or hw? */
- if (mask & BUFFER_BIT_DEPTH) {
- struct st_renderbuffer *strb = st_renderbuffer(depthRb);
-
- if (strb->surface) {
- if (check_clear_depth_with_quad(ctx, depthRb,
- st->clear.enable_ds_separate))
- quad_buffers |= PIPE_CLEAR_DEPTH;
- else
- clear_buffers |= PIPE_CLEAR_DEPTH;
- }
- }
- if (mask & BUFFER_BIT_STENCIL) {
- struct st_renderbuffer *strb = st_renderbuffer(stencilRb);
-
- if (strb->surface) {
- if (check_clear_stencil_with_quad(ctx, stencilRb,
- st->clear.enable_ds_separate))
- quad_buffers |= PIPE_CLEAR_STENCIL;
- else
- clear_buffers |= PIPE_CLEAR_STENCIL;
- }
- }
- }
-
- /*
- * If we're going to use clear_with_quad() for any reason, use it for
- * everything possible.
- */
- if (quad_buffers) {
- quad_buffers |= clear_buffers;
- clear_with_quad(ctx,
- quad_buffers & PIPE_CLEAR_COLOR,
- quad_buffers & PIPE_CLEAR_DEPTH,
- quad_buffers & PIPE_CLEAR_STENCIL);
- } else if (clear_buffers) {
- /* driver cannot know it can clear everything if the buffer
- * is a combined depth/stencil buffer but this wasn't actually
- * required from the visual. Hence fix this up to avoid potential
- * read-modify-write in the driver.
- */
- float clearColor[4];
-
- if ((clear_buffers & PIPE_CLEAR_DEPTHSTENCIL) &&
- ((clear_buffers & PIPE_CLEAR_DEPTHSTENCIL) != PIPE_CLEAR_DEPTHSTENCIL) &&
- (depthRb == stencilRb) &&
- (ctx->DrawBuffer->Visual.depthBits == 0 ||
- ctx->DrawBuffer->Visual.stencilBits == 0))
- clear_buffers |= PIPE_CLEAR_DEPTHSTENCIL;
-
- if (ctx->DrawBuffer->_ColorDrawBuffers[0]) {
- st_translate_color(ctx->Color.ClearColor,
- ctx->DrawBuffer->_ColorDrawBuffers[0]->_BaseFormat,
- clearColor);
- }
-
- st->pipe->clear(st->pipe, clear_buffers, ctx->Color.ClearColor,
- ctx->Depth.Clear, ctx->Stencil.Clear);
- }
- if (mask & BUFFER_BIT_ACCUM)
- st_clear_accum_buffer(ctx,
- ctx->DrawBuffer->Attachment[BUFFER_ACCUM].Renderbuffer);
-}
-
-
-void
-st_init_clear_functions(struct dd_function_table *functions)
-{
- functions->Clear = st_Clear;
-}
+/**************************************************************************
+ *
+ * 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
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+ /*
+ * Authors:
+ * Keith Whitwell <keith@tungstengraphics.com>
+ * Brian Paul
+ * Michel Dänzer
+ */
+
+#include "main/glheader.h"
+#include "main/formats.h"
+#include "main/macros.h"
+#include "program/prog_instruction.h"
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_cb_accum.h"
+#include "st_cb_clear.h"
+#include "st_cb_fbo.h"
+#include "st_format.h"
+#include "st_program.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_shader_tokens.h"
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "util/u_format.h"
+#include "util/u_inlines.h"
+#include "util/u_simple_shaders.h"
+#include "util/u_draw_quad.h"
+
+#include "cso_cache/cso_context.h"
+
+
+/**
+ * Do per-context initialization for glClear.
+ */
+void
+st_init_clear(struct st_context *st)
+{
+ struct pipe_screen *pscreen = st->pipe->screen;
+
+ memset(&st->clear, 0, sizeof(st->clear));
+
+ st->clear.raster.gl_rasterization_rules = 1;
+ st->clear.enable_ds_separate = pscreen->get_param(pscreen, PIPE_CAP_DEPTHSTENCIL_CLEAR_SEPARATE);
+}
+
+
+/**
+ * Free per-context state for glClear.
+ */
+void
+st_destroy_clear(struct st_context *st)
+{
+ if (st->clear.fs) {
+ cso_delete_fragment_shader(st->cso_context, st->clear.fs);
+ st->clear.fs = NULL;
+ }
+ if (st->clear.vs) {
+ cso_delete_vertex_shader(st->cso_context, st->clear.vs);
+ st->clear.vs = NULL;
+ }
+ if (st->clear.vbuf) {
+ pipe_resource_reference(&st->clear.vbuf, NULL);
+ st->clear.vbuf = NULL;
+ }
+}
+
+
+/**
+ * Helper function to set the fragment shaders.
+ */
+static INLINE void
+set_fragment_shader(struct st_context *st)
+{
+ if (!st->clear.fs)
+ st->clear.fs = util_make_fragment_passthrough_shader(st->pipe);
+
+ cso_set_fragment_shader_handle(st->cso_context, st->clear.fs);
+}
+
+
+/**
+ * Helper function to set the vertex shader.
+ */
+static INLINE void
+set_vertex_shader(struct st_context *st)
+{
+ /* vertex shader - still required to provide the linkage between
+ * fragment shader input semantics and vertex_element/buffers.
+ */
+ if (!st->clear.vs)
+ {
+ const uint semantic_names[] = { TGSI_SEMANTIC_POSITION,
+ TGSI_SEMANTIC_COLOR };
+ const uint semantic_indexes[] = { 0, 0 };
+ st->clear.vs = util_make_vertex_passthrough_shader(st->pipe, 2,
+ semantic_names,
+ semantic_indexes);
+ }
+
+ cso_set_vertex_shader_handle(st->cso_context, st->clear.vs);
+}
+
+
+/**
+ * Draw a screen-aligned quadrilateral.
+ * Coords are clip coords with y=0=bottom.
+ */
+static void
+draw_quad(struct st_context *st,
+ float x0, float y0, float x1, float y1, GLfloat z,
+ const GLfloat color[4])
+{
+ struct pipe_context *pipe = st->pipe;
+
+ /* XXX: Need to improve buffer_write to allow NO_WAIT (as well as
+ * no_flush) updates to buffers where we know there is no conflict
+ * with previous data. Currently using max_slots > 1 will cause
+ * synchronous rendering if the driver flushes its command buffers
+ * between one bitmap and the next. Our flush hook below isn't
+ * sufficient to catch this as the driver doesn't tell us when it
+ * flushes its own command buffers. Until this gets fixed, pay the
+ * price of allocating a new buffer for each bitmap cache-flush to
+ * avoid synchronous rendering.
+ */
+ const GLuint max_slots = 1; /* 1024 / sizeof(st->clear.vertices); */
+ GLuint i;
+
+ if (st->clear.vbuf_slot >= max_slots) {
+ pipe_resource_reference(&st->clear.vbuf, NULL);
+ st->clear.vbuf_slot = 0;
+ }
+
+ if (!st->clear.vbuf) {
+ st->clear.vbuf = pipe_buffer_create(pipe->screen,
+ PIPE_BIND_VERTEX_BUFFER,
+ PIPE_USAGE_STREAM,
+ max_slots * sizeof(st->clear.vertices));
+ }
+
+ /* positions */
+ st->clear.vertices[0][0][0] = x0;
+ st->clear.vertices[0][0][1] = y0;
+
+ st->clear.vertices[1][0][0] = x1;
+ st->clear.vertices[1][0][1] = y0;
+
+ st->clear.vertices[2][0][0] = x1;
+ st->clear.vertices[2][0][1] = y1;
+
+ st->clear.vertices[3][0][0] = x0;
+ st->clear.vertices[3][0][1] = y1;
+
+ /* same for all verts: */
+ for (i = 0; i < 4; i++) {
+ st->clear.vertices[i][0][2] = z;
+ st->clear.vertices[i][0][3] = 1.0;
+ st->clear.vertices[i][1][0] = color[0];
+ st->clear.vertices[i][1][1] = color[1];
+ st->clear.vertices[i][1][2] = color[2];
+ st->clear.vertices[i][1][3] = color[3];
+ }
+
+ /* put vertex data into vbuf */
+ pipe_buffer_write_nooverlap(st->pipe, st->clear.vbuf,
+ st->clear.vbuf_slot
+ * sizeof(st->clear.vertices),
+ sizeof(st->clear.vertices),
+ st->clear.vertices);
+
+ /* draw */
+ util_draw_vertex_buffer(pipe,
+ st->cso_context,
+ st->clear.vbuf,
+ st->clear.vbuf_slot * sizeof(st->clear.vertices),
+ PIPE_PRIM_TRIANGLE_FAN,
+ 4, /* verts */
+ 2); /* attribs/vert */
+
+ /* Increment slot */
+ st->clear.vbuf_slot++;
+}
+
+
+
+/**
+ * Do glClear by drawing a quadrilateral.
+ * The vertices of the quad will be computed from the
+ * ctx->DrawBuffer->_X/Ymin/max fields.
+ */
+static void
+clear_with_quad(struct gl_context *ctx,
+ GLboolean color, GLboolean depth, GLboolean stencil)
+{
+ struct st_context *st = st_context(ctx);
+ const struct gl_framebuffer *fb = ctx->DrawBuffer;
+ const GLfloat fb_width = (GLfloat) fb->Width;
+ const GLfloat fb_height = (GLfloat) fb->Height;
+ const GLfloat x0 = (GLfloat) ctx->DrawBuffer->_Xmin / fb_width * 2.0f - 1.0f;
+ const GLfloat x1 = (GLfloat) ctx->DrawBuffer->_Xmax / fb_width * 2.0f - 1.0f;
+ const GLfloat y0 = (GLfloat) ctx->DrawBuffer->_Ymin / fb_height * 2.0f - 1.0f;
+ const GLfloat y1 = (GLfloat) ctx->DrawBuffer->_Ymax / fb_height * 2.0f - 1.0f;
+ float clearColor[4];
+
+ /*
+ printf("%s %s%s%s %f,%f %f,%f\n", __FUNCTION__,
+ color ? "color, " : "",
+ depth ? "depth, " : "",
+ stencil ? "stencil" : "",
+ x0, y0,
+ x1, y1);
+ */
+
+ cso_save_blend(st->cso_context);
+ cso_save_stencil_ref(st->cso_context);
+ cso_save_depth_stencil_alpha(st->cso_context);
+ cso_save_rasterizer(st->cso_context);
+ cso_save_viewport(st->cso_context);
+ cso_save_clip(st->cso_context);
+ cso_save_fragment_shader(st->cso_context);
+ cso_save_vertex_shader(st->cso_context);
+ cso_save_vertex_elements(st->cso_context);
+ cso_save_vertex_buffers(st->cso_context);
+
+ /* blend state: RGBA masking */
+ {
+ struct pipe_blend_state blend;
+ memset(&blend, 0, sizeof(blend));
+ blend.rt[0].rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+ blend.rt[0].alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+ blend.rt[0].rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+ blend.rt[0].alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+ if (color) {
+ if (ctx->Color.ColorMask[0][0])
+ blend.rt[0].colormask |= PIPE_MASK_R;
+ if (ctx->Color.ColorMask[0][1])
+ blend.rt[0].colormask |= PIPE_MASK_G;
+ if (ctx->Color.ColorMask[0][2])
+ blend.rt[0].colormask |= PIPE_MASK_B;
+ if (ctx->Color.ColorMask[0][3])
+ blend.rt[0].colormask |= PIPE_MASK_A;
+ if (st->ctx->Color.DitherFlag)
+ blend.dither = 1;
+ }
+ cso_set_blend(st->cso_context, &blend);
+ }
+
+ /* depth_stencil state: always pass/set to ref value */
+ {
+ struct pipe_depth_stencil_alpha_state depth_stencil;
+ memset(&depth_stencil, 0, sizeof(depth_stencil));
+ if (depth) {
+ depth_stencil.depth.enabled = 1;
+ depth_stencil.depth.writemask = 1;
+ depth_stencil.depth.func = PIPE_FUNC_ALWAYS;
+ }
+
+ if (stencil) {
+ struct pipe_stencil_ref stencil_ref;
+ memset(&stencil_ref, 0, sizeof(stencil_ref));
+ depth_stencil.stencil[0].enabled = 1;
+ depth_stencil.stencil[0].func = PIPE_FUNC_ALWAYS;
+ depth_stencil.stencil[0].fail_op = PIPE_STENCIL_OP_REPLACE;
+ depth_stencil.stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE;
+ depth_stencil.stencil[0].zfail_op = PIPE_STENCIL_OP_REPLACE;
+ depth_stencil.stencil[0].valuemask = 0xff;
+ depth_stencil.stencil[0].writemask = ctx->Stencil.WriteMask[0] & 0xff;
+ stencil_ref.ref_value[0] = ctx->Stencil.Clear;
+ cso_set_stencil_ref(st->cso_context, &stencil_ref);
+ }
+
+ cso_set_depth_stencil_alpha(st->cso_context, &depth_stencil);
+ }
+
+ cso_set_vertex_elements(st->cso_context, 2, st->velems_util_draw);
+
+ cso_set_rasterizer(st->cso_context, &st->clear.raster);
+
+ /* viewport state: viewport matching window dims */
+ {
+ const GLboolean invert = (st_fb_orientation(fb) == Y_0_TOP);
+ struct pipe_viewport_state vp;
+ vp.scale[0] = 0.5f * fb_width;
+ vp.scale[1] = fb_height * (invert ? -0.5f : 0.5f);
+ vp.scale[2] = 1.0f;
+ vp.scale[3] = 1.0f;
+ vp.translate[0] = 0.5f * fb_width;
+ vp.translate[1] = 0.5f * fb_height;
+ vp.translate[2] = 0.0f;
+ vp.translate[3] = 0.0f;
+ cso_set_viewport(st->cso_context, &vp);
+ }
+
+ cso_set_clip(st->cso_context, &st->clear.clip);
+ set_fragment_shader(st);
+ set_vertex_shader(st);
+
+ if (ctx->DrawBuffer->_ColorDrawBuffers[0]) {
+ st_translate_color(ctx->Color.ClearColor,
+ ctx->DrawBuffer->_ColorDrawBuffers[0]->_BaseFormat,
+ clearColor);
+ }
+
+ /* draw quad matching scissor rect */
+ draw_quad(st, x0, y0, x1, y1, (GLfloat) ctx->Depth.Clear, clearColor);
+
+ /* Restore pipe state */
+ cso_restore_blend(st->cso_context);
+ cso_restore_stencil_ref(st->cso_context);
+ cso_restore_depth_stencil_alpha(st->cso_context);
+ cso_restore_rasterizer(st->cso_context);
+ cso_restore_viewport(st->cso_context);
+ cso_restore_clip(st->cso_context);
+ cso_restore_fragment_shader(st->cso_context);
+ cso_restore_vertex_shader(st->cso_context);
+ cso_restore_vertex_elements(st->cso_context);
+ cso_restore_vertex_buffers(st->cso_context);
+}
+
+
+/**
+ * Determine if we need to clear the depth buffer by drawing a quad.
+ */
+static INLINE GLboolean
+check_clear_color_with_quad(struct gl_context *ctx, struct gl_renderbuffer *rb)
+{
+ if (ctx->Scissor.Enabled &&
+ (ctx->Scissor.X != 0 ||
+ ctx->Scissor.Y != 0 ||
+ ctx->Scissor.Width < rb->Width ||
+ ctx->Scissor.Height < rb->Height))
+ return GL_TRUE;
+
+ if (!ctx->Color.ColorMask[0][0] ||
+ !ctx->Color.ColorMask[0][1] ||
+ !ctx->Color.ColorMask[0][2] ||
+ !ctx->Color.ColorMask[0][3])
+ return GL_TRUE;
+
+ return GL_FALSE;
+}
+
+
+/**
+ * Determine if we need to clear the combiend depth/stencil buffer by
+ * drawing a quad.
+ */
+static INLINE GLboolean
+check_clear_depth_stencil_with_quad(struct gl_context *ctx, struct gl_renderbuffer *rb)
+{
+ const GLuint stencilMax = 0xff;
+ GLboolean maskStencil
+ = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax;
+
+ assert(rb->Format == MESA_FORMAT_S8 ||
+ rb->Format == MESA_FORMAT_Z24_S8 ||
+ rb->Format == MESA_FORMAT_S8_Z24);
+
+ if (ctx->Scissor.Enabled &&
+ (ctx->Scissor.X != 0 ||
+ ctx->Scissor.Y != 0 ||
+ ctx->Scissor.Width < rb->Width ||
+ ctx->Scissor.Height < rb->Height))
+ return GL_TRUE;
+
+ if (maskStencil)
+ return GL_TRUE;
+
+ return GL_FALSE;
+}
+
+
+/**
+ * Determine if we need to clear the depth buffer by drawing a quad.
+ */
+static INLINE GLboolean
+check_clear_depth_with_quad(struct gl_context *ctx, struct gl_renderbuffer *rb,
+ boolean ds_separate)
+{
+ const struct st_renderbuffer *strb = st_renderbuffer(rb);
+ const GLboolean isDS = util_format_is_depth_and_stencil(strb->surface->format);
+
+ if (ctx->Scissor.Enabled &&
+ (ctx->Scissor.X != 0 ||
+ ctx->Scissor.Y != 0 ||
+ ctx->Scissor.Width < rb->Width ||
+ ctx->Scissor.Height < rb->Height))
+ return GL_TRUE;
+
+ if (!ds_separate && isDS && ctx->DrawBuffer->Visual.stencilBits > 0)
+ return GL_TRUE;
+
+ return GL_FALSE;
+}
+
+
+/**
+ * Determine if we need to clear the stencil buffer by drawing a quad.
+ */
+static INLINE GLboolean
+check_clear_stencil_with_quad(struct gl_context *ctx, struct gl_renderbuffer *rb,
+ boolean ds_separate)
+{
+ const struct st_renderbuffer *strb = st_renderbuffer(rb);
+ const GLboolean isDS = util_format_is_depth_and_stencil(strb->surface->format);
+ const GLuint stencilMax = 0xff;
+ const GLboolean maskStencil
+ = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax;
+
+ assert(rb->Format == MESA_FORMAT_S8 ||
+ rb->Format == MESA_FORMAT_Z24_S8 ||
+ rb->Format == MESA_FORMAT_S8_Z24);
+
+ if (maskStencil)
+ return GL_TRUE;
+
+ if (ctx->Scissor.Enabled &&
+ (ctx->Scissor.X != 0 ||
+ ctx->Scissor.Y != 0 ||
+ ctx->Scissor.Width < rb->Width ||
+ ctx->Scissor.Height < rb->Height))
+ return GL_TRUE;
+
+ /* This is correct, but it is necessary to look at the depth clear
+ * value held in the surface when it comes time to issue the clear,
+ * rather than taking depth and stencil clear values from the
+ * current state.
+ */
+ if (!ds_separate && isDS && ctx->DrawBuffer->Visual.depthBits > 0)
+ return GL_TRUE;
+
+ return GL_FALSE;
+}
+
+
+
+/**
+ * Called when we need to flush.
+ */
+void
+st_flush_clear(struct st_context *st)
+{
+ /* Release vertex buffer to avoid synchronous rendering if we were
+ * to map it in the next frame.
+ */
+ pipe_resource_reference(&st->clear.vbuf, NULL);
+ st->clear.vbuf_slot = 0;
+}
+
+
+
+/**
+ * Called via ctx->Driver.Clear()
+ */
+static void
+st_Clear(struct gl_context *ctx, GLbitfield mask)
+{
+ static const GLbitfield BUFFER_BITS_DS
+ = (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL);
+ struct st_context *st = st_context(ctx);
+ struct gl_renderbuffer *depthRb
+ = ctx->DrawBuffer->Attachment[BUFFER_DEPTH].Renderbuffer;
+ struct gl_renderbuffer *stencilRb
+ = ctx->DrawBuffer->Attachment[BUFFER_STENCIL].Renderbuffer;
+ GLbitfield quad_buffers = 0x0;
+ GLbitfield clear_buffers = 0x0;
+ GLuint i;
+
+ /* This makes sure the pipe has the latest scissor, etc values */
+ st_validate_state( st );
+
+ if (mask & BUFFER_BITS_COLOR) {
+ for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++) {
+ GLuint b = ctx->DrawBuffer->_ColorDrawBufferIndexes[i];
+
+ if (mask & (1 << b)) {
+ struct gl_renderbuffer *rb
+ = ctx->DrawBuffer->Attachment[b].Renderbuffer;
+ struct st_renderbuffer *strb = st_renderbuffer(rb);
+
+ if (!strb || !strb->surface)
+ continue;
+
+ if (check_clear_color_with_quad( ctx, rb ))
+ quad_buffers |= PIPE_CLEAR_COLOR;
+ else
+ clear_buffers |= PIPE_CLEAR_COLOR;
+ }
+ }
+ }
+
+ if ((mask & BUFFER_BITS_DS) == BUFFER_BITS_DS && depthRb == stencilRb) {
+ /* clearing combined depth + stencil */
+ struct st_renderbuffer *strb = st_renderbuffer(depthRb);
+
+ if (strb->surface) {
+ if (check_clear_depth_stencil_with_quad(ctx, depthRb))
+ quad_buffers |= PIPE_CLEAR_DEPTHSTENCIL;
+ else
+ clear_buffers |= PIPE_CLEAR_DEPTHSTENCIL;
+ }
+ }
+ else {
+ /* separate depth/stencil clears */
+ /* I don't think truly separate buffers are actually possible in gallium or hw? */
+ if (mask & BUFFER_BIT_DEPTH) {
+ struct st_renderbuffer *strb = st_renderbuffer(depthRb);
+
+ if (strb->surface) {
+ if (check_clear_depth_with_quad(ctx, depthRb,
+ st->clear.enable_ds_separate))
+ quad_buffers |= PIPE_CLEAR_DEPTH;
+ else
+ clear_buffers |= PIPE_CLEAR_DEPTH;
+ }
+ }
+ if (mask & BUFFER_BIT_STENCIL) {
+ struct st_renderbuffer *strb = st_renderbuffer(stencilRb);
+
+ if (strb->surface) {
+ if (check_clear_stencil_with_quad(ctx, stencilRb,
+ st->clear.enable_ds_separate))
+ quad_buffers |= PIPE_CLEAR_STENCIL;
+ else
+ clear_buffers |= PIPE_CLEAR_STENCIL;
+ }
+ }
+ }
+
+ /*
+ * If we're going to use clear_with_quad() for any reason, use it for
+ * everything possible.
+ */
+ if (quad_buffers) {
+ quad_buffers |= clear_buffers;
+ clear_with_quad(ctx,
+ quad_buffers & PIPE_CLEAR_COLOR,
+ quad_buffers & PIPE_CLEAR_DEPTH,
+ quad_buffers & PIPE_CLEAR_STENCIL);
+ } else if (clear_buffers) {
+ /* driver cannot know it can clear everything if the buffer
+ * is a combined depth/stencil buffer but this wasn't actually
+ * required from the visual. Hence fix this up to avoid potential
+ * read-modify-write in the driver.
+ */
+ float clearColor[4];
+
+ if ((clear_buffers & PIPE_CLEAR_DEPTHSTENCIL) &&
+ ((clear_buffers & PIPE_CLEAR_DEPTHSTENCIL) != PIPE_CLEAR_DEPTHSTENCIL) &&
+ (depthRb == stencilRb) &&
+ (ctx->DrawBuffer->Visual.depthBits == 0 ||
+ ctx->DrawBuffer->Visual.stencilBits == 0))
+ clear_buffers |= PIPE_CLEAR_DEPTHSTENCIL;
+
+ if (ctx->DrawBuffer->_ColorDrawBuffers[0]) {
+ st_translate_color(ctx->Color.ClearColor,
+ ctx->DrawBuffer->_ColorDrawBuffers[0]->_BaseFormat,
+ clearColor);
+ }
+
+ st->pipe->clear(st->pipe, clear_buffers, ctx->Color.ClearColor,
+ ctx->Depth.Clear, ctx->Stencil.Clear);
+ }
+ if (mask & BUFFER_BIT_ACCUM)
+ st_clear_accum_buffer(ctx,
+ ctx->DrawBuffer->Attachment[BUFFER_ACCUM].Renderbuffer);
+}
+
+
+void
+st_init_clear_functions(struct dd_function_table *functions)
+{
+ functions->Clear = st_Clear;
+}
diff --git a/mesalib/src/mesa/state_tracker/st_cb_texture.c b/mesalib/src/mesa/state_tracker/st_cb_texture.c
index 8bdb3c801..1b824c0de 100644
--- a/mesalib/src/mesa/state_tracker/st_cb_texture.c
+++ b/mesalib/src/mesa/state_tracker/st_cb_texture.c
@@ -1530,7 +1530,7 @@ st_copy_texsubimage(struct gl_context *ctx,
GLint srcY0, srcY1;
struct pipe_surface surf_tmpl;
memset(&surf_tmpl, 0, sizeof(surf_tmpl));
- surf_tmpl.format = stImage->pt->format;
+ surf_tmpl.format = util_format_linear(stImage->pt->format);
surf_tmpl.usage = PIPE_BIND_RENDER_TARGET;
surf_tmpl.u.tex.level = stImage->level;
surf_tmpl.u.tex.first_layer = stImage->face + destZ;