int samplers_used;
bool indirect_addr_temps;
bool indirect_addr_consts;
+ int num_clip_distances;
int glsl_version;
bool native_integers;
/** List of immediate_storage */
exec_list immediates;
- int num_immediates;
+ unsigned num_immediates;
/** List of function_entry */
exec_list function_signatures;
bool process_move_condition(ir_rvalue *ir);
- void remove_output_reads(gl_register_file type);
void simplify_cmp(void);
void rename_temp_register(int index, int new_index);
case3(SLT, ISLT, USLT);
case2iu(ISHR, USHR);
+
+ case2fi(SSG, ISSG);
+ case3(ABS, IABS, IABS);
default: break;
}
}
break;
case ir_unop_neg:
- assert(result_dst.type == GLSL_TYPE_FLOAT || result_dst.type == GLSL_TYPE_INT);
- if (result_dst.type == GLSL_TYPE_INT)
+ if (result_dst.type == GLSL_TYPE_INT || result_dst.type == GLSL_TYPE_UINT)
emit(ir, TGSI_OPCODE_INEG, result_dst, op[0]);
else {
op[0].negate = ~op[0].negate;
}
break;
case ir_unop_abs:
- assert(result_dst.type == GLSL_TYPE_FLOAT);
emit(ir, TGSI_OPCODE_ABS, result_dst, op[0]);
break;
case ir_unop_sign:
gl_type = native_integers ? GL_BOOL : GL_FLOAT;
for (i = 0; i < ir->type->vector_elements; i++) {
if (native_integers)
- values[i].b = ir->value.b[i];
+ values[i].u = ir->value.b[i] ? ~0 : 0;
else
values[i].f = ir->value.b[i];
}
ir->shadow_comparitor->accept(this);
/* XXX This will need to be updated for cubemap array samplers. */
- if (sampler_type->sampler_dimensionality == GLSL_SAMPLER_DIM_2D &&
- sampler_type->sampler_array) {
+ if ((sampler_type->sampler_dimensionality == GLSL_SAMPLER_DIM_2D &&
+ sampler_type->sampler_array) ||
+ sampler_type->sampler_dimensionality == GLSL_SAMPLER_DIM_CUBE) {
coord_dst.writemask = WRITEMASK_W;
} else {
coord_dst.writemask = WRITEMASK_Z;
if (is_tex_instruction(inst->op)) {
v->samplers_used |= 1 << inst->sampler;
- prog->SamplerTargets[inst->sampler] =
- (gl_texture_index)inst->tex_target;
if (inst->tex_shadow) {
prog->ShadowSamplers |= 1 << inst->sampler;
}
}
prog->SamplersUsed = v->samplers_used;
- _mesa_update_shader_textures_used(prog);
+
+ if (v->shader_program != NULL)
+ _mesa_update_shader_textures_used(v->shader_program, prog);
}
static void
}
}
-/*
- * Scan/rewrite program to remove reads of custom (output) registers.
- * The passed type has to be either PROGRAM_OUTPUT or PROGRAM_VARYING
- * (for vertex shaders).
- * In GLSL shaders, varying vars can be read and written.
- * On some hardware, trying to read an output register causes trouble.
- * So, rewrite the program to use a temporary register in this case.
- *
- * Based on _mesa_remove_output_reads from programopt.c.
- */
-void
-glsl_to_tgsi_visitor::remove_output_reads(gl_register_file type)
-{
- GLuint i;
- GLint outputMap[VERT_RESULT_MAX];
- GLint outputTypes[VERT_RESULT_MAX];
- GLuint numVaryingReads = 0;
- GLboolean *usedTemps;
- GLuint firstTemp = 0;
-
- usedTemps = new GLboolean[MAX_TEMPS];
- if (!usedTemps) {
- return;
- }
- _mesa_find_used_registers(prog, PROGRAM_TEMPORARY,
- usedTemps, MAX_TEMPS);
-
- assert(type == PROGRAM_VARYING || type == PROGRAM_OUTPUT);
- assert(prog->Target == GL_VERTEX_PROGRAM_ARB || type != PROGRAM_VARYING);
-
- for (i = 0; i < VERT_RESULT_MAX; i++)
- outputMap[i] = -1;
-
- /* look for instructions which read from varying vars */
- foreach_iter(exec_list_iterator, iter, this->instructions) {
- glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *)iter.get();
- const GLuint numSrc = num_inst_src_regs(inst->op);
- GLuint j;
- for (j = 0; j < numSrc; j++) {
- if (inst->src[j].file == type) {
- /* replace the read with a temp reg */
- const GLuint var = inst->src[j].index;
- if (outputMap[var] == -1) {
- numVaryingReads++;
- outputMap[var] = _mesa_find_free_register(usedTemps,
- MAX_TEMPS,
- firstTemp);
- outputTypes[var] = inst->src[j].type;
- firstTemp = outputMap[var] + 1;
- }
- inst->src[j].file = PROGRAM_TEMPORARY;
- inst->src[j].index = outputMap[var];
- }
- }
- }
-
- delete [] usedTemps;
-
- if (numVaryingReads == 0)
- return; /* nothing to be done */
-
- /* look for instructions which write to the varying vars identified above */
- foreach_iter(exec_list_iterator, iter, this->instructions) {
- glsl_to_tgsi_instruction *inst = (glsl_to_tgsi_instruction *)iter.get();
- if (inst->dst.file == type && outputMap[inst->dst.index] >= 0) {
- /* change inst to write to the temp reg, instead of the varying */
- inst->dst.file = PROGRAM_TEMPORARY;
- inst->dst.index = outputMap[inst->dst.index];
- }
- }
-
- /* insert new MOV instructions at the end */
- for (i = 0; i < VERT_RESULT_MAX; i++) {
- if (outputMap[i] >= 0) {
- /* MOV VAR[i], TEMP[tmp]; */
- st_src_reg src = st_src_reg(PROGRAM_TEMPORARY, outputMap[i], outputTypes[i]);
- st_dst_reg dst = st_dst_reg(type, WRITEMASK_XYZW, outputTypes[i]);
- dst.index = i;
- this->emit(NULL, TGSI_OPCODE_MOV, dst, src);
- }
- }
-}
-
/**
* Returns the mask of channels (bitmask of WRITEMASK_X,Y,Z,W) which
* are read from the given src in this instruction
if (!tempWrites) {
return;
}
- memset(tempWrites, 0, sizeof(tempWrites));
+ memset(tempWrites, 0, sizeof(unsigned) * MAX_TEMPS);
memset(outputWrites, 0, sizeof(outputWrites));
foreach_iter(exec_list_iterator, iter, this->instructions) {
v->indirect_addr_temps = original->indirect_addr_temps;
v->indirect_addr_consts = original->indirect_addr_consts;
memcpy(&v->immediates, &original->immediates, sizeof(v->immediates));
+ v->num_immediates = original->num_immediates;
/*
* Get initial pixel color from the texture.
v->indirect_addr_temps = original->indirect_addr_temps;
v->indirect_addr_consts = original->indirect_addr_consts;
memcpy(&v->immediates, &original->immediates, sizeof(v->immediates));
+ v->num_immediates = original->num_immediates;
/* TEX tmp0, fragment.texcoord[0], texture[0], 2D; */
coord = st_src_reg(PROGRAM_INPUT, FRAG_ATTRIB_TEX0, glsl_type::vec2_type);
struct ureg_src samplers[PIPE_MAX_SAMPLERS];
struct ureg_src systemValues[SYSTEM_VALUE_MAX];
- /* Extra info for handling point size clamping in vertex shader */
- struct ureg_dst pointSizeResult; /**< Actual point size output register */
- struct ureg_src pointSizeConst; /**< Point size range constant register */
- GLint pointSizeOutIndex; /**< Temp point size output register */
- GLboolean prevInstWrotePointSize;
-
const GLuint *inputMapping;
const GLuint *outputMapping;
return t->temps[index];
case PROGRAM_OUTPUT:
- if (t->procType == TGSI_PROCESSOR_VERTEX && index == VERT_RESULT_PSIZ)
- t->prevInstWrotePointSize = GL_TRUE;
-
if (t->procType == TGSI_PROCESSOR_VERTEX)
assert(index < VERT_RESULT_MAX);
else if (t->procType == TGSI_PROCESSOR_FRAGMENT)
static struct ureg_dst
translate_dst(struct st_translate *t,
const st_dst_reg *dst_reg,
- bool saturate)
+ bool saturate, bool clamp_color)
{
struct ureg_dst dst = dst_register(t,
dst_reg->file,
if (saturate)
dst = ureg_saturate(dst);
+ else if (clamp_color && dst_reg->file == PROGRAM_OUTPUT) {
+ /* Clamp colors for ARB_color_buffer_float. */
+ switch (t->procType) {
+ case TGSI_PROCESSOR_VERTEX:
+ /* XXX if the geometry shader is present, this must be done there
+ * instead of here. */
+ if (dst_reg->index == VERT_RESULT_COL0 ||
+ dst_reg->index == VERT_RESULT_COL1 ||
+ dst_reg->index == VERT_RESULT_BFC0 ||
+ dst_reg->index == VERT_RESULT_BFC1) {
+ dst = ureg_saturate(dst);
+ }
+ break;
+
+ case TGSI_PROCESSOR_FRAGMENT:
+ if (dst_reg->index >= FRAG_RESULT_COLOR) {
+ dst = ureg_saturate(dst);
+ }
+ break;
+ }
+ }
if (dst_reg->reladdr != NULL)
dst = ureg_dst_indirect(dst, ureg_src(t->address[0]));
static void
compile_tgsi_instruction(struct st_translate *t,
- const glsl_to_tgsi_instruction *inst)
+ const glsl_to_tgsi_instruction *inst,
+ bool clamp_dst_color_output)
{
struct ureg_program *ureg = t->ureg;
GLuint i;
if (num_dst)
dst[0] = translate_dst(t,
&inst->dst,
- inst->saturate);
+ inst->saturate,
+ clamp_dst_color_output);
for (i = 0; i < num_src; i++)
src[i] = translate_src(t, &inst->src[i]);
ureg_tex_insn(ureg,
inst->op,
dst, num_dst,
- translate_texture_target(inst->tex_target, inst->tex_shadow),
+ st_translate_texture_target(inst->tex_target, inst->tex_shadow),
texoffsets, inst->tex_offset_num_offset,
src, num_src);
return;
const GLuint outputMapping[],
const ubyte outputSemanticName[],
const ubyte outputSemanticIndex[],
- boolean passthrough_edgeflags)
+ boolean passthrough_edgeflags,
+ boolean clamp_color)
{
struct st_translate *t;
unsigned i;
t->inputMapping = inputMapping;
t->outputMapping = outputMapping;
t->ureg = ureg;
- t->pointSizeOutIndex = -1;
- t->prevInstWrotePointSize = GL_FALSE;
if (program->shader_program) {
for (i = 0; i < program->shader_program->NumUserUniformStorage; i++) {
}
for (i = 0; i < numOutputs; i++) {
- t->outputs[i] = ureg_DECL_output(ureg,
- outputSemanticName[i],
- outputSemanticIndex[i]);
- if ((outputSemanticName[i] == TGSI_SEMANTIC_PSIZE) && proginfo->Id) {
- /* Writing to the point size result register requires special
- * handling to implement clamping.
- */
- static const gl_state_index pointSizeClampState[STATE_LENGTH]
- = { STATE_INTERNAL, STATE_POINT_SIZE_IMPL_CLAMP, (gl_state_index)0, (gl_state_index)0, (gl_state_index)0 };
- /* XXX: note we are modifying the incoming shader here! Need to
- * do this before emitting the constant decls below, or this
- * will be missed.
- */
- unsigned pointSizeClampConst =
- _mesa_add_state_reference(proginfo->Parameters,
- pointSizeClampState);
- struct ureg_dst psizregtemp = ureg_DECL_temporary(ureg);
- t->pointSizeConst = ureg_DECL_constant(ureg, pointSizeClampConst);
- t->pointSizeResult = t->outputs[i];
- t->pointSizeOutIndex = i;
- t->outputs[i] = psizregtemp;
+ if (outputSemanticName[i] == TGSI_SEMANTIC_CLIPDIST) {
+ int mask = ((1 << (program->num_clip_distances - 4*outputSemanticIndex[i])) - 1) & TGSI_WRITEMASK_XYZW;
+ t->outputs[i] = ureg_DECL_output_masked(ureg,
+ outputSemanticName[i],
+ outputSemanticIndex[i],
+ mask);
+ } else {
+ t->outputs[i] = ureg_DECL_output(ureg,
+ outputSemanticName[i],
+ outputSemanticIndex[i]);
}
}
if (passthrough_edgeflags)
i = 0;
foreach_iter(exec_list_iterator, iter, program->immediates) {
immediate_storage *imm = (immediate_storage *)iter.get();
+ assert(i < program->num_immediates);
t->immediates[i++] = emit_immediate(t, imm->values, imm->type, imm->size);
}
+ assert(i == program->num_immediates);
/* texture samplers */
for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
*/
foreach_iter(exec_list_iterator, iter, program->instructions) {
set_insn_start(t, ureg_get_instruction_number(ureg));
- compile_tgsi_instruction(t, (glsl_to_tgsi_instruction *)iter.get());
-
- if (t->prevInstWrotePointSize && proginfo->Id) {
- /* The previous instruction wrote to the (fake) vertex point size
- * result register. Now we need to clamp that value to the min/max
- * point size range, putting the result into the real point size
- * register.
- * Note that we can't do this easily at the end of program due to
- * possible early return.
- */
- set_insn_start(t, ureg_get_instruction_number(ureg));
- ureg_MAX(t->ureg,
- ureg_writemask(t->outputs[t->pointSizeOutIndex], WRITEMASK_X),
- ureg_src(t->outputs[t->pointSizeOutIndex]),
- ureg_swizzle(t->pointSizeConst, 1,1,1,1));
- ureg_MIN(t->ureg, ureg_writemask(t->pointSizeResult, WRITEMASK_X),
- ureg_src(t->outputs[t->pointSizeOutIndex]),
- ureg_swizzle(t->pointSizeConst, 2,2,2,2));
- }
- t->prevInstWrotePointSize = GL_FALSE;
+ compile_tgsi_instruction(t, (glsl_to_tgsi_instruction *)iter.get(),
+ clamp_color);
}
/* Fix up all emitted labels:
static struct gl_program *
get_mesa_program(struct gl_context *ctx,
struct gl_shader_program *shader_program,
- struct gl_shader *shader)
+ struct gl_shader *shader,
+ int num_clip_distances)
{
glsl_to_tgsi_visitor* v = new glsl_to_tgsi_visitor();
struct gl_program *prog;
- struct pipe_screen * screen = st_context(ctx)->pipe->screen;
- unsigned pipe_shader_type;
GLenum target;
const char *target_string;
bool progress;
case GL_VERTEX_SHADER:
target = GL_VERTEX_PROGRAM_ARB;
target_string = "vertex";
- pipe_shader_type = PIPE_SHADER_VERTEX;
break;
case GL_FRAGMENT_SHADER:
target = GL_FRAGMENT_PROGRAM_ARB;
target_string = "fragment";
- pipe_shader_type = PIPE_SHADER_FRAGMENT;
break;
case GL_GEOMETRY_SHADER:
target = GL_GEOMETRY_PROGRAM_NV;
target_string = "geometry";
- pipe_shader_type = PIPE_SHADER_GEOMETRY;
break;
default:
assert(!"should not be reached");
v->options = options;
v->glsl_version = ctx->Const.GLSLVersion;
v->native_integers = ctx->Const.NativeIntegers;
+ v->num_clip_distances = num_clip_distances;
_mesa_generate_parameters_list_for_uniforms(shader_program, shader,
prog->Parameters);
+ /* Remove reads from output registers. */
+ lower_output_reads(shader->ir);
+
/* Emit intermediate IR for main(). */
visit_exec_list(shader->ir, v);
}
#endif
- if (!screen->get_shader_param(screen, pipe_shader_type,
- PIPE_SHADER_CAP_OUTPUT_READ)) {
- /* Remove reads to output registers, and to varyings in vertex shaders. */
- v->remove_output_reads(PROGRAM_OUTPUT);
- if (target == GL_VERTEX_PROGRAM_ARB)
- v->remove_output_reads(PROGRAM_VARYING);
- }
-
/* Perform optimizations on the instructions in the glsl_to_tgsi_visitor. */
v->simplify_cmp();
v->copy_propagate();
return prog;
}
+/**
+ * Searches through the IR for a declaration of gl_ClipDistance and returns the
+ * declared size of the gl_ClipDistance array. Returns 0 if gl_ClipDistance is
+ * not declared in the IR.
+ */
+int get_clip_distance_size(exec_list *ir)
+{
+ foreach_iter (exec_list_iterator, iter, *ir) {
+ ir_instruction *inst = (ir_instruction *)iter.get();
+ ir_variable *var = inst->as_variable();
+ if (var == NULL) continue;
+ if (!strcmp(var->name, "gl_ClipDistance")) {
+ return var->type->length;
+ }
+ }
+
+ return 0;
+}
+
extern "C" {
struct gl_shader *
GLboolean
st_link_shader(struct gl_context *ctx, struct gl_shader_program *prog)
{
+ int num_clip_distances[MESA_SHADER_TYPES];
assert(prog->LinkStatus);
for (unsigned i = 0; i < MESA_SHADER_TYPES; i++) {
const struct gl_shader_compiler_options *options =
&ctx->ShaderCompilerOptions[_mesa_shader_type_to_index(prog->_LinkedShaders[i]->Type)];
+ /* We have to determine the length of the gl_ClipDistance array before
+ * the array is lowered to two vec4s by lower_clip_distance().
+ */
+ num_clip_distances[i] = get_clip_distance_size(ir);
+
do {
+ unsigned what_to_lower = MOD_TO_FRACT | DIV_TO_MUL_RCP |
+ EXP_TO_EXP2 | LOG_TO_LOG2;
+ if (options->EmitNoPow)
+ what_to_lower |= POW_TO_EXP2;
+ if (!ctx->Const.NativeIntegers)
+ what_to_lower |= INT_DIV_TO_MUL_RCP;
+
progress = false;
/* Lowering */
do_mat_op_to_vec(ir);
- lower_instructions(ir, (MOD_TO_FRACT | DIV_TO_MUL_RCP | EXP_TO_EXP2
- | LOG_TO_LOG2 | INT_DIV_TO_MUL_RCP
- | ((options->EmitNoPow) ? POW_TO_EXP2 : 0)));
+ lower_instructions(ir, what_to_lower);
progress = do_lower_jumps(ir, true, true, options->EmitNoMainReturn, options->EmitNoCont, options->EmitNoLoops) || progress;
|| progress;
progress = lower_quadop_vector(ir, false) || progress;
+ progress = lower_clip_distance(ir) || progress;
if (options->MaxIfDepth == 0)
progress = lower_discard(ir) || progress;
if (prog->_LinkedShaders[i] == NULL)
continue;
- linked_prog = get_mesa_program(ctx, prog, prog->_LinkedShaders[i]);
+ linked_prog = get_mesa_program(ctx, prog, prog->_LinkedShaders[i],
+ num_clip_distances[i]);
if (linked_prog) {
static const GLenum targets[] = {
}
void
-st_translate_stream_output_info(struct glsl_to_tgsi_visitor *glsl_to_tgsi,
+st_translate_stream_output_info(glsl_to_tgsi_visitor *glsl_to_tgsi,
const GLuint outputMapping[],
struct pipe_stream_output_info *so)
{
- static unsigned comps_to_mask[] = {
- 0,
- TGSI_WRITEMASK_X,
- TGSI_WRITEMASK_XY,
- TGSI_WRITEMASK_XYZ,
- TGSI_WRITEMASK_XYZW
- };
unsigned i;
struct gl_transform_feedback_info *info =
&glsl_to_tgsi->shader_program->LinkedTransformFeedback;
for (i = 0; i < info->NumOutputs; i++) {
- assert(info->Outputs[i].NumComponents < Elements(comps_to_mask));
so->output[i].register_index =
outputMapping[info->Outputs[i].OutputRegister];
- so->output[i].register_mask =
- comps_to_mask[info->Outputs[i].NumComponents];
+ so->output[i].start_component = info->Outputs[i].ComponentOffset;
+ so->output[i].num_components = info->Outputs[i].NumComponents;
so->output[i].output_buffer = info->Outputs[i].OutputBuffer;
+ so->output[i].dst_offset = info->Outputs[i].DstOffset;
+ }
+
+ for (i = 0; i < PIPE_MAX_SO_BUFFERS; i++) {
+ so->stride[i] = info->BufferStride[i];
}
so->num_outputs = info->NumOutputs;
}
projects / mesa.git / blobdiff