#include "tgsi/tgsi_info.h"
#include "util/u_math.h"
#include "util/u_memory.h"
-#include "st_glsl_types.h"
#include "st_program.h"
#include "st_mesa_to_tgsi.h"
#include "st_format.h"
virtual void visit(ir_barrier *);
/*@}*/
- void visit_expression(ir_expression *, st_src_reg *) ATTRIBUTE_NOINLINE;
+ void ATTRIBUTE_NOINLINE visit_expression(ir_expression *, st_src_reg *);
void visit_atomic_counter_intrinsic(ir_call *);
void visit_ssbo_intrinsic(ir_call *);
emit_asm(ir, TGSI_OPCODE_DFRACEXP, undef_dst, result_dst, op[0]);
break;
- case ir_unop_noise: {
- /* At some point, a motivated person could add a better
- * implementation of noise. Currently not even the nvidia
- * binary drivers do anything more than this. In any case, the
- * place to do this is in the GL state tracker, not the poor
- * driver.
- */
- emit_asm(ir, TGSI_OPCODE_MOV, result_dst, st_src_reg_for_float(0.5));
- break;
- }
-
case ir_binop_add:
emit_asm(ir, TGSI_OPCODE_ADD, result_dst, op[0], op[1]);
break;
emit_asm(ir, TGSI_OPCODE_NOT, result_dst, op[0]);
break;
}
+ /* fallthrough */
case ir_unop_u2f:
if (native_integers) {
emit_asm(ir, TGSI_OPCODE_U2F, result_dst, op[0]);
break;
}
+ /* fallthrough */
case ir_binop_lshift:
case ir_binop_rshift:
if (native_integers) {
emit_asm(ir, opcode, result_dst, op[0], count);
break;
}
+ /* fallthrough */
case ir_binop_bit_and:
if (native_integers) {
emit_asm(ir, TGSI_OPCODE_AND, result_dst, op[0], op[1]);
break;
}
+ /* fallthrough */
case ir_binop_bit_xor:
if (native_integers) {
emit_asm(ir, TGSI_OPCODE_XOR, result_dst, op[0], op[1]);
break;
}
+ /* fallthrough */
case ir_binop_bit_or:
if (native_integers) {
emit_asm(ir, TGSI_OPCODE_OR, result_dst, op[0], op[1]);
case ir_unop_get_buffer_size: {
ir_constant *const_offset = ir->operands[0]->as_constant();
- int buf_base = ctx->st->has_hw_atomics
- ? 0 : ctx->Const.Program[shader->Stage].MaxAtomicBuffers;
st_src_reg buffer(
PROGRAM_BUFFER,
- buf_base + (const_offset ? const_offset->value.u[0] : 0),
+ const_offset ? const_offset->value.u[0] : 0,
GLSL_TYPE_UINT);
if (!const_offset) {
buffer.reladdr = ralloc(mem_ctx, st_src_reg);
case ir_unop_ssbo_unsized_array_length:
case ir_unop_atan:
case ir_binop_atan2:
+ case ir_unop_clz:
+ case ir_binop_add_sat:
+ case ir_binop_sub_sat:
+ case ir_binop_abs_sub:
+ case ir_binop_avg:
+ case ir_binop_avg_round:
+ case ir_binop_mul_32x16:
+ case ir_unop_f162f:
+ case ir_unop_f2f16:
+ case ir_unop_f2fmp:
+ case ir_unop_f162b:
+ case ir_unop_b2f16:
+ case ir_unop_i2i:
+ case ir_unop_i2imp:
+ case ir_unop_u2u:
+ case ir_unop_u2ump:
/* This operation is not supported, or should have already been handled.
*/
assert(!"Invalid ir opcode in glsl_to_tgsi_visitor::visit()");
{
st_src_reg src;
int i;
- int swizzle[4];
+ int swizzle[4] = {0};
/* Note that this is only swizzles in expressions, not those on the left
* hand side of an assignment, which do write masking. See ir_assignment
if (handle_bound_deref(ir->as_dereference()))
return;
- /* We only need the logic provided by st_glsl_storage_type_size()
+ /* We only need the logic provided by count_vec4_slots()
* for arrays of structs. Indirect sampler and image indexing is handled
* elsewhere.
*/
int element_size = ir->type->without_array()->is_struct() ?
- st_glsl_storage_type_size(ir->type, var->data.bindless) :
+ ir->type->count_vec4_slots(false, var->data.bindless) :
type_size(ir->type);
index = ir->array_index->constant_expression_value(ralloc_parent(ir));
if (i == (unsigned) ir->field_idx)
break;
const glsl_type *member_type = struct_type->fields.structure[i].type;
- offset += st_glsl_storage_type_size(member_type, var->data.bindless);
+ offset += member_type->count_vec4_slots(false, var->data.bindless);
}
/* If the type is smaller than a vec4, replicate the last channel out. */
resource = buffer;
} else {
- st_src_reg buffer(PROGRAM_BUFFER, location->data.binding,
+ st_src_reg buffer(PROGRAM_BUFFER,
+ prog->info.num_ssbos +
+ location->data.binding,
GLSL_TYPE_ATOMIC_UINT);
if (offset.file != PROGRAM_UNDEFINED) {
ir_rvalue *offset = ((ir_instruction *)param)->as_rvalue();
ir_constant *const_block = block->as_constant();
- int buf_base = st_context(ctx)->has_hw_atomics
- ? 0 : ctx->Const.Program[shader->Stage].MaxAtomicBuffers;
st_src_reg buffer(
PROGRAM_BUFFER,
- buf_base + (const_block ? const_block->value.u[0] : 0),
+ const_block ? const_block->value.u[0] : 0,
GLSL_TYPE_UINT);
if (!const_block) {
get_image_qualifiers(ir_dereference *ir, const glsl_type **type,
bool *memory_coherent, bool *memory_volatile,
bool *memory_restrict, bool *memory_read_only,
- unsigned *image_format)
+ enum pipe_format *image_format)
{
switch (ir->ir_type) {
unsigned sampler_array_size = 1, sampler_base = 0;
bool memory_coherent = false, memory_volatile = false,
memory_restrict = false, memory_read_only = false;
- unsigned image_format = 0;
+ enum pipe_format image_format = PIPE_FORMAT_NONE;
const glsl_type *type = NULL;
get_image_qualifiers(img, &type, &memory_coherent, &memory_volatile,
}
inst->tex_target = type->sampler_index();
- inst->image_format = st_mesa_format_to_pipe_format(st_context(ctx),
- _mesa_get_shader_image_format(image_format));
+ inst->image_format = image_format;
inst->read_only = memory_read_only;
if (memory_coherent)
enum tgsi_opcode opcode = TGSI_OPCODE_NOP;
const glsl_type *sampler_type = ir->sampler->type;
unsigned sampler_array_size = 1, sampler_base = 0;
- bool is_cube_array = false, is_cube_shadow = false;
+ bool is_cube_array = false;
ir_variable *var = ir->sampler->variable_referenced();
unsigned i;
/* if we are a cube array sampler or a cube shadow */
if (sampler_type->sampler_dimensionality == GLSL_SAMPLER_DIM_CUBE) {
is_cube_array = sampler_type->sampler_array;
- is_cube_shadow = sampler_type->sampler_shadow;
}
if (ir->coordinate) {
}
break;
case ir_txb:
- if (is_cube_array || is_cube_shadow) {
+ if (is_cube_array ||
+ (sampler_type->sampler_shadow && sampler_type->coordinate_components() >= 3)) {
opcode = TGSI_OPCODE_TXB2;
}
else {
if (this->has_tex_txf_lz && ir->lod_info.lod->is_zero()) {
opcode = TGSI_OPCODE_TEX_LZ;
} else {
- opcode = is_cube_array ? TGSI_OPCODE_TXL2 : TGSI_OPCODE_TXL;
+ opcode = (is_cube_array || (sampler_type->sampler_shadow && sampler_type->coordinate_components() >= 3)) ? TGSI_OPCODE_TXL2 : TGSI_OPCODE_TXL;
ir->lod_info.lod->accept(this);
lod_info = this->result;
}
ir->shadow_comparator->accept(this);
if (is_cube_array) {
- cube_sc = get_temp(glsl_type::float_type);
- cube_sc_dst = st_dst_reg(cube_sc);
- cube_sc_dst.writemask = WRITEMASK_X;
+ if (lod_info.file != PROGRAM_UNDEFINED) {
+ // If we have both a cube array *and* a bias/lod, stick the
+ // comparator into the .Y of the second argument.
+ st_src_reg tmp = get_temp(glsl_type::vec2_type);
+ cube_sc_dst = st_dst_reg(tmp);
+ cube_sc_dst.writemask = WRITEMASK_X;
+ emit_asm(ir, TGSI_OPCODE_MOV, cube_sc_dst, lod_info);
+ lod_info = tmp;
+ cube_sc_dst.writemask = WRITEMASK_Y;
+ } else {
+ cube_sc = get_temp(glsl_type::float_type);
+ cube_sc_dst = st_dst_reg(cube_sc);
+ cube_sc_dst.writemask = WRITEMASK_X;
+ }
emit_asm(ir, TGSI_OPCODE_MOV, cube_sc_dst, this->result);
- cube_sc_dst.writemask = WRITEMASK_X;
}
else {
if ((sampler_type->sampler_dimensionality == GLSL_SAMPLER_DIM_2D &&
bool tg4_component_in_swizzle;
};
-/** Map Mesa's SYSTEM_VALUE_x to TGSI_SEMANTIC_x */
-enum tgsi_semantic
-_mesa_sysval_to_semantic(unsigned sysval)
-{
- switch (sysval) {
- /* Vertex shader */
- case SYSTEM_VALUE_VERTEX_ID:
- return TGSI_SEMANTIC_VERTEXID;
- case SYSTEM_VALUE_INSTANCE_ID:
- return TGSI_SEMANTIC_INSTANCEID;
- case SYSTEM_VALUE_VERTEX_ID_ZERO_BASE:
- return TGSI_SEMANTIC_VERTEXID_NOBASE;
- case SYSTEM_VALUE_BASE_VERTEX:
- return TGSI_SEMANTIC_BASEVERTEX;
- case SYSTEM_VALUE_BASE_INSTANCE:
- return TGSI_SEMANTIC_BASEINSTANCE;
- case SYSTEM_VALUE_DRAW_ID:
- return TGSI_SEMANTIC_DRAWID;
-
- /* Geometry shader */
- case SYSTEM_VALUE_INVOCATION_ID:
- return TGSI_SEMANTIC_INVOCATIONID;
-
- /* Fragment shader */
- case SYSTEM_VALUE_FRAG_COORD:
- return TGSI_SEMANTIC_POSITION;
- case SYSTEM_VALUE_POINT_COORD:
- return TGSI_SEMANTIC_PCOORD;
- case SYSTEM_VALUE_FRONT_FACE:
- return TGSI_SEMANTIC_FACE;
- case SYSTEM_VALUE_SAMPLE_ID:
- return TGSI_SEMANTIC_SAMPLEID;
- case SYSTEM_VALUE_SAMPLE_POS:
- return TGSI_SEMANTIC_SAMPLEPOS;
- case SYSTEM_VALUE_SAMPLE_MASK_IN:
- return TGSI_SEMANTIC_SAMPLEMASK;
- case SYSTEM_VALUE_HELPER_INVOCATION:
- return TGSI_SEMANTIC_HELPER_INVOCATION;
-
- /* Tessellation shader */
- case SYSTEM_VALUE_TESS_COORD:
- return TGSI_SEMANTIC_TESSCOORD;
- case SYSTEM_VALUE_VERTICES_IN:
- return TGSI_SEMANTIC_VERTICESIN;
- case SYSTEM_VALUE_PRIMITIVE_ID:
- return TGSI_SEMANTIC_PRIMID;
- case SYSTEM_VALUE_TESS_LEVEL_OUTER:
- return TGSI_SEMANTIC_TESSOUTER;
- case SYSTEM_VALUE_TESS_LEVEL_INNER:
- return TGSI_SEMANTIC_TESSINNER;
-
- /* Compute shader */
- case SYSTEM_VALUE_LOCAL_INVOCATION_ID:
- return TGSI_SEMANTIC_THREAD_ID;
- case SYSTEM_VALUE_WORK_GROUP_ID:
- return TGSI_SEMANTIC_BLOCK_ID;
- case SYSTEM_VALUE_NUM_WORK_GROUPS:
- return TGSI_SEMANTIC_GRID_SIZE;
- case SYSTEM_VALUE_LOCAL_GROUP_SIZE:
- return TGSI_SEMANTIC_BLOCK_SIZE;
-
- /* ARB_shader_ballot */
- case SYSTEM_VALUE_SUBGROUP_SIZE:
- return TGSI_SEMANTIC_SUBGROUP_SIZE;
- case SYSTEM_VALUE_SUBGROUP_INVOCATION:
- return TGSI_SEMANTIC_SUBGROUP_INVOCATION;
- case SYSTEM_VALUE_SUBGROUP_EQ_MASK:
- return TGSI_SEMANTIC_SUBGROUP_EQ_MASK;
- case SYSTEM_VALUE_SUBGROUP_GE_MASK:
- return TGSI_SEMANTIC_SUBGROUP_GE_MASK;
- case SYSTEM_VALUE_SUBGROUP_GT_MASK:
- return TGSI_SEMANTIC_SUBGROUP_GT_MASK;
- case SYSTEM_VALUE_SUBGROUP_LE_MASK:
- return TGSI_SEMANTIC_SUBGROUP_LE_MASK;
- case SYSTEM_VALUE_SUBGROUP_LT_MASK:
- return TGSI_SEMANTIC_SUBGROUP_LT_MASK;
-
- /* Unhandled */
- case SYSTEM_VALUE_LOCAL_INVOCATION_INDEX:
- case SYSTEM_VALUE_GLOBAL_INVOCATION_ID:
- case SYSTEM_VALUE_VERTEX_CNT:
- case SYSTEM_VALUE_BARYCENTRIC_PIXEL:
- case SYSTEM_VALUE_BARYCENTRIC_SAMPLE:
- case SYSTEM_VALUE_BARYCENTRIC_CENTROID:
- case SYSTEM_VALUE_BARYCENTRIC_SIZE:
- default:
- assert(!"Unexpected SYSTEM_VALUE_ enum");
- return TGSI_SEMANTIC_COUNT;
- }
-}
-
/**
* Map a glsl_to_tgsi constant/immediate to a TGSI immediate.
*/
t->inputs[t->inputMapping[VARYING_SLOT_FACE]] = ureg_src(face_temp);
}
-static void
-emit_compute_block_size(const struct gl_program *prog,
- struct ureg_program *ureg) {
- ureg_property(ureg, TGSI_PROPERTY_CS_FIXED_BLOCK_WIDTH,
- prog->info.cs.local_size[0]);
- ureg_property(ureg, TGSI_PROPERTY_CS_FIXED_BLOCK_HEIGHT,
- prog->info.cs.local_size[1]);
- ureg_property(ureg, TGSI_PROPERTY_CS_FIXED_BLOCK_DEPTH,
- prog->info.cs.local_size[2]);
-}
-
struct sort_inout_decls {
bool operator()(const struct inout_decl &a, const struct inout_decl &b) const {
return mapping[a.mesa_index] < mapping[b.mesa_index];
std::sort(decls, decls + count, sorter);
}
-static enum tgsi_interpolate_mode
-st_translate_interp(enum glsl_interp_mode glsl_qual, GLuint varying)
-{
- switch (glsl_qual) {
- case INTERP_MODE_NONE:
- if (varying == VARYING_SLOT_COL0 || varying == VARYING_SLOT_COL1)
- return TGSI_INTERPOLATE_COLOR;
- return TGSI_INTERPOLATE_PERSPECTIVE;
- case INTERP_MODE_SMOOTH:
- return TGSI_INTERPOLATE_PERSPECTIVE;
- case INTERP_MODE_FLAT:
- return TGSI_INTERPOLATE_CONSTANT;
- case INTERP_MODE_NOPERSPECTIVE:
- return TGSI_INTERPOLATE_LINEAR;
- default:
- assert(0 && "unexpected interp mode in st_translate_interp()");
- return TGSI_INTERPOLATE_PERSPECTIVE;
- }
-}
-
/**
* Translate intermediate IR (glsl_to_tgsi_instruction) to TGSI format.
* \param program the program to translate
assert(numOutputs <= ARRAY_SIZE(t->outputs));
ASSERT_BITFIELD_SIZE(st_src_reg, type, GLSL_TYPE_ERROR);
+ ASSERT_BITFIELD_SIZE(st_src_reg, file, PROGRAM_FILE_MAX);
ASSERT_BITFIELD_SIZE(st_dst_reg, type, GLSL_TYPE_ERROR);
+ ASSERT_BITFIELD_SIZE(st_dst_reg, file, PROGRAM_FILE_MAX);
ASSERT_BITFIELD_SIZE(glsl_to_tgsi_instruction, tex_type, GLSL_TYPE_ERROR);
ASSERT_BITFIELD_SIZE(glsl_to_tgsi_instruction, image_format, PIPE_FORMAT_COUNT);
ASSERT_BITFIELD_SIZE(glsl_to_tgsi_instruction, tex_target,
assert(interpMode);
interp_mode = interpMode[slot] != TGSI_INTERPOLATE_COUNT ?
(enum tgsi_interpolate_mode) interpMode[slot] :
- st_translate_interp(decl->interp, inputSlotToAttr[slot]);
+ tgsi_get_interp_mode(decl->interp,
+ inputSlotToAttr[slot] == VARYING_SLOT_COL0 ||
+ inputSlotToAttr[slot] == VARYING_SLOT_COL1);
interp_location = (enum tgsi_interpolate_loc) decl->interp_loc;
}
}
if (procType == PIPE_SHADER_FRAGMENT) {
- if (program->shader->Program->info.fs.early_fragment_tests ||
- program->shader->Program->info.fs.post_depth_coverage) {
- ureg_property(ureg, TGSI_PROPERTY_FS_EARLY_DEPTH_STENCIL, 1);
-
- if (program->shader->Program->info.fs.post_depth_coverage)
- ureg_property(ureg, TGSI_PROPERTY_FS_POST_DEPTH_COVERAGE, 1);
- }
-
if (proginfo->info.inputs_read & VARYING_BIT_POS) {
/* Must do this after setting up t->inputs. */
emit_wpos(st_context(ctx), t, proginfo, ureg,
goto out;
}
}
+
+ if (program->shader->Program->sh.fs.BlendSupport)
+ ureg_property(ureg,
+ TGSI_PROPERTY_FS_BLEND_EQUATION_ADVANCED,
+ program->shader->Program->sh.fs.BlendSupport);
+
}
else if (procType == PIPE_SHADER_VERTEX) {
for (i = 0; i < numOutputs; i++) {
}
}
- if (procType == PIPE_SHADER_COMPUTE) {
- emit_compute_block_size(proginfo, ureg);
- }
-
/* Declare address register.
*/
if (program->num_address_regs > 0) {
for (i = 0; sysInputs; i++) {
if (sysInputs & (1ull << i)) {
- enum tgsi_semantic semName = _mesa_sysval_to_semantic(i);
+ enum tgsi_semantic semName = tgsi_get_sysval_semantic(i);
t->systemValues[i] = ureg_DECL_system_value(ureg, semName, 0);
if (!st_context(ctx)->has_hw_atomics) {
for (i = 0; i < prog->info.num_abos; i++) {
- unsigned index = prog->sh.AtomicBuffers[i]->Binding;
- assert(index < frag_const->MaxAtomicBuffers);
+ unsigned index = (prog->info.num_ssbos +
+ prog->sh.AtomicBuffers[i]->Binding);
+ assert(prog->sh.AtomicBuffers[i]->Binding <
+ frag_const->MaxAtomicBuffers);
t->buffers[index] = ureg_DECL_buffer(ureg, index, true);
}
} else {
assert(prog->info.num_ssbos <= frag_const->MaxShaderStorageBlocks);
for (i = 0; i < prog->info.num_ssbos; i++) {
- unsigned index = i;
- if (!st_context(ctx)->has_hw_atomics)
- index += frag_const->MaxAtomicBuffers;
-
- t->buffers[index] = ureg_DECL_buffer(ureg, index, false);
+ t->buffers[i] = ureg_DECL_buffer(ureg, i, false);
}
}
foreach_in_list(glsl_to_tgsi_instruction, inst, &program->instructions)
compile_tgsi_instruction(t, inst);
- /* Set the next shader stage hint for VS and TES. */
- switch (procType) {
- case PIPE_SHADER_VERTEX:
- case PIPE_SHADER_TESS_EVAL:
- if (program->shader_program->SeparateShader)
- break;
-
- for (i = program->shader->Stage+1; i <= MESA_SHADER_FRAGMENT; i++) {
- if (program->shader_program->_LinkedShaders[i]) {
- ureg_set_next_shader_processor(
- ureg, pipe_shader_type_from_mesa((gl_shader_stage)i));
- break;
- }
- }
- break;
- default:
- ; /* nothing - silence compiler warning */
- }
-
out:
if (t) {
free(t->arrays);
return NULL;
}
- struct st_vertex_program *stvp;
- struct st_fragment_program *stfp;
- struct st_common_program *stp;
- struct st_compute_program *stcp;
-
- switch (shader->Stage) {
- case MESA_SHADER_VERTEX:
- stvp = (struct st_vertex_program *)prog;
- stvp->glsl_to_tgsi = v;
- break;
- case MESA_SHADER_FRAGMENT:
- stfp = (struct st_fragment_program *)prog;
- stfp->glsl_to_tgsi = v;
- break;
- case MESA_SHADER_TESS_CTRL:
- case MESA_SHADER_TESS_EVAL:
- case MESA_SHADER_GEOMETRY:
- stp = st_common_program(prog);
- stp->glsl_to_tgsi = v;
- break;
- case MESA_SHADER_COMPUTE:
- stcp = (struct st_compute_program *)prog;
- stcp->glsl_to_tgsi = v;
- break;
- default:
- assert(!"should not be reached");
- return NULL;
- }
+ st_program(prog)->glsl_to_tgsi = v;
PRINT_STATS(v->print_stats());
st_set_prog_affected_state_flags(linked_prog);
if (linked_prog) {
+ /* This is really conservative: */
+ linked_prog->info.writes_memory =
+ linked_prog->info.num_ssbos ||
+ linked_prog->info.num_images ||
+ ctx->Extensions.ARB_bindless_texture ||
+ (linked_prog->sh.LinkedTransformFeedback &&
+ linked_prog->sh.LinkedTransformFeedback->NumVarying);
+
if (!ctx->Driver.ProgramStringNotify(ctx,
_mesa_shader_stage_to_program(i),
linked_prog)) {
return GL_TRUE;
}
-
-extern "C" {
-
-void
-st_translate_stream_output_info(struct gl_transform_feedback_info *info,
- const ubyte outputMapping[],
- struct pipe_stream_output_info *so)
-{
- unsigned i;
-
- if (!info) {
- so->num_outputs = 0;
- return;
- }
-
- for (i = 0; i < info->NumOutputs; i++) {
- so->output[i].register_index =
- outputMapping[info->Outputs[i].OutputRegister];
- 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;
- so->output[i].stream = info->Outputs[i].StreamId;
- }
-
- for (i = 0; i < PIPE_MAX_SO_BUFFERS; i++) {
- so->stride[i] = info->Buffers[i].Stride;
- }
- so->num_outputs = info->NumOutputs;
-}
-
-} /* extern "C" */
projects / mesa.git / blobdiff