static void tgsi_gs_prepare(struct draw_geometry_shader *shader,
const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
const unsigned constants_size[PIPE_MAX_CONSTANT_BUFFERS])
{
struct tgsi_exec_machine *machine = shader->machine;
tgsi_exec_set_constant_buffers(machine, PIPE_MAX_CONSTANT_BUFFERS,
constants, constants_size);
}
static void tgsi_gs_prepare(struct draw_geometry_shader *shader,
const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
const unsigned constants_size[PIPE_MAX_CONSTANT_BUFFERS])
{
struct tgsi_exec_machine *machine = shader->machine;
int j;
tgsi_exec_set_constant_buffers(machine, PIPE_MAX_CONSTANT_BUFFERS,
constants, constants_size);
if (shader->info.uses_invocationid) {
unsigned i = machine->SysSemanticToIndex[TGSI_SEMANTIC_INVOCATIONID];
for (j = 0; j < TGSI_QUAD_SIZE; j++)
machine->SystemValue[i].i[j] = shader->invocation_id;
}
}
/**
* Shade/write an array of quads
* Called via quad_stage::run()
*/
static void
shade_quads(struct quad_stage *qs,
struct quad_header *quads[],
unsigned nr)
{
struct softpipe_context *softpipe = qs->softpipe;
struct tgsi_exec_machine *machine = softpipe->fs_machine;
unsigned i, nr_quads = 0;
tgsi_exec_set_constant_buffers(machine, PIPE_MAX_CONSTANT_BUFFERS,
softpipe->mapped_constants[PIPE_SHADER_FRAGMENT],
softpipe->const_buffer_size[PIPE_SHADER_FRAGMENT]);
machine->InterpCoefs = quads[0]->coef;
for (i = 0; i < nr; i++) {
/* Only omit this quad from the output list if all the fragments
* are killed _AND_ it's not the first quad in the list.
* The first quad is special in the (optimized) depth-testing code:
* the quads' Z coordinates are step-wise interpolated with respect
* to the first quad in the list.
* For multi-pass algorithms we need to produce exactly the same
* Z values in each pass. If interpolation starts with different quads
* we can get different Z values for the same (x,y).
*/
if (!shade_quad(qs, quads[i]) && i > 0)
continue; /* quad totally culled/killed */
if (/*do_coverage*/ 0)
coverage_quad( qs, quads[i] );
quads[nr_quads++] = quads[i];
}
if (nr_quads)
qs->next->run(qs->next, quads, nr_quads);
}
/**
* Execute geometry shader using TGSI interpreter.
*/
int draw_geometry_shader_run(struct draw_geometry_shader *shader,
const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
const unsigned constants_size[PIPE_MAX_CONSTANT_BUFFERS],
const struct draw_vertex_info *input_verts,
const struct draw_prim_info *input_prim,
struct draw_vertex_info *output_verts,
struct draw_prim_info *output_prims )
{
const float (*input)[4] = (const float (*)[4])input_verts->verts->data;
unsigned input_stride = input_verts->vertex_size;
unsigned vertex_size = input_verts->vertex_size;
struct tgsi_exec_machine *machine = shader->machine;
unsigned num_input_verts = input_prim->linear ?
input_verts->count :
input_prim->count;
unsigned num_in_primitives =
MAX2(u_gs_prims_for_vertices(input_prim->prim, num_input_verts),
u_gs_prims_for_vertices(shader->input_primitive, num_input_verts));
unsigned max_out_prims = u_gs_prims_for_vertices(shader->output_primitive,
shader->max_output_vertices)
* num_in_primitives;
output_verts->vertex_size = input_verts->vertex_size;
output_verts->stride = input_verts->vertex_size;
output_verts->verts =
(struct vertex_header *)MALLOC(input_verts->vertex_size *
num_in_primitives *
shader->max_output_vertices);
#if 0
debug_printf("%s count = %d (in prims # = %d)\n",
__FUNCTION__, num_input_verts, num_in_primitives);
debug_printf("\tlinear = %d, prim_info->count = %d\n",
input_prim->linear, input_prim->count);
debug_printf("\tprimt pipe = %d, shader in = %d, shader out = %d, max out = %d\n",
input_prim->prim, shader->input_primitive,
shader->output_primitive,
shader->max_output_vertices);
#endif
shader->emitted_vertices = 0;
shader->emitted_primitives = 0;
shader->vertex_size = vertex_size;
shader->tmp_output = (float (*)[4])output_verts->verts->data;
shader->in_prim_idx = 0;
shader->input_vertex_stride = input_stride;
shader->input = input;
if (shader->primitive_lengths) {
FREE(shader->primitive_lengths);
}
shader->primitive_lengths = MALLOC(max_out_prims * sizeof(unsigned));
tgsi_exec_set_constant_buffers(machine, PIPE_MAX_CONSTANT_BUFFERS,
constants, constants_size);
if (input_prim->linear)
gs_run(shader, input_prim, input_verts,
output_prims, output_verts);
else
gs_run_elts(shader, input_prim, input_verts,
output_prims, output_verts);
/* Update prim_info:
*/
output_prims->linear = TRUE;
output_prims->elts = NULL;
output_prims->start = 0;
output_prims->count = shader->emitted_vertices;
output_prims->prim = shader->output_primitive;
output_prims->flags = 0x0;
output_prims->primitive_lengths = shader->primitive_lengths;
output_prims->primitive_count = shader->emitted_primitives;
output_verts->count = shader->emitted_vertices;
#if 0
debug_printf("GS finished, prims = %d, verts = %d\n",
output_prims->primitive_count,
output_verts->count);
#endif
return shader->emitted_vertices;
}
void
softpipe_launch_grid(struct pipe_context *context,
const struct pipe_grid_info *info)
{
struct softpipe_context *softpipe = softpipe_context(context);
struct sp_compute_shader *cs = softpipe->cs;
int num_threads_in_group;
struct tgsi_exec_machine **machines;
int bwidth, bheight, bdepth;
int w, h, d, i;
int g_w, g_h, g_d;
uint32_t grid_size[3];
void *local_mem = NULL;
softpipe_update_compute_samplers(softpipe);
bwidth = cs->info.properties[TGSI_PROPERTY_CS_FIXED_BLOCK_WIDTH];
bheight = cs->info.properties[TGSI_PROPERTY_CS_FIXED_BLOCK_HEIGHT];
bdepth = cs->info.properties[TGSI_PROPERTY_CS_FIXED_BLOCK_DEPTH];
num_threads_in_group = bwidth * bheight * bdepth;
fill_grid_size(context, info, grid_size);
if (cs->shader.req_local_mem) {
local_mem = CALLOC(1, cs->shader.req_local_mem);
}
machines = CALLOC(sizeof(struct tgsi_exec_machine *), num_threads_in_group);
if (!machines) {
FREE(local_mem);
return;
}
/* initialise machines + GRID_SIZE + THREAD_ID + BLOCK_SIZE */
for (d = 0; d < bdepth; d++) {
for (h = 0; h < bheight; h++) {
for (w = 0; w < bwidth; w++) {
int idx = w + (h * bwidth) + (d * bheight * bwidth);
machines[idx] = tgsi_exec_machine_create(PIPE_SHADER_COMPUTE);
machines[idx]->LocalMem = local_mem;
machines[idx]->LocalMemSize = cs->shader.req_local_mem;
cs_prepare(cs, machines[idx],
w, h, d,
grid_size[0], grid_size[1], grid_size[2],
bwidth, bheight, bdepth,
(struct tgsi_sampler *)softpipe->tgsi.sampler[PIPE_SHADER_COMPUTE],
(struct tgsi_image *)softpipe->tgsi.image[PIPE_SHADER_COMPUTE],
(struct tgsi_buffer *)softpipe->tgsi.buffer[PIPE_SHADER_COMPUTE]);
tgsi_exec_set_constant_buffers(machines[idx], PIPE_MAX_CONSTANT_BUFFERS,
softpipe->mapped_constants[PIPE_SHADER_COMPUTE],
softpipe->const_buffer_size[PIPE_SHADER_COMPUTE]);
}
}
}
for (g_d = 0; g_d < grid_size[2]; g_d++) {
for (g_h = 0; g_h < grid_size[1]; g_h++) {
for (g_w = 0; g_w < grid_size[0]; g_w++) {
run_workgroup(cs, g_w, g_h, g_d, num_threads_in_group, machines);
}
}
}
for (i = 0; i < num_threads_in_group; i++) {
cs_delete(cs, machines[i]);
tgsi_exec_machine_destroy(machines[i]);
}
FREE(local_mem);
FREE(machines);
}
