It's all screaming for integer support -- fake it with float for now.
emit_sz = 0;
break;
}
-
+
+ hw_key.element[i].type = TRANSLATE_ELEMENT_NORMAL;
hw_key.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
hw_key.element[i].input_buffer = src_buffer;
hw_key.element[i].input_offset = src_offset;
struct pt_fetch;
void draw_pt_fetch_prepare( struct pt_fetch *fetch,
unsigned vertex_input_count,
- unsigned vertex_size );
+ unsigned vertex_size,
+ unsigned instance_id_index );
void draw_pt_fetch_run( struct pt_fetch *fetch,
const unsigned *elts,
emit_sz = 0;
break;
}
-
+
+ hw_key.element[i].type = TRANSLATE_ELEMENT_NORMAL;
hw_key.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
hw_key.element[i].input_buffer = src_buffer;
hw_key.element[i].input_offset = src_offset;
*/
void draw_pt_fetch_prepare( struct pt_fetch *fetch,
unsigned vs_input_count,
- unsigned vertex_size )
+ unsigned vertex_size,
+ unsigned instance_id_index )
{
struct draw_context *draw = fetch->draw;
unsigned nr_inputs;
- unsigned i, nr = 0;
+ unsigned i, nr = 0, ei = 0;
unsigned dst_offset = 0;
+ unsigned num_extra_inputs = 0;
struct translate_key key;
fetch->vertex_size = vertex_size;
{
/* Need to set header->vertex_id = 0xffff somehow.
*/
+ key.element[nr].type = TRANSLATE_ELEMENT_NORMAL;
key.element[nr].input_format = PIPE_FORMAT_R32_FLOAT;
key.element[nr].input_buffer = draw->pt.nr_vertex_buffers;
key.element[nr].input_offset = 0;
*/
dst_offset += 4 * sizeof(float);
}
-
- assert( draw->pt.nr_vertex_elements >= vs_input_count );
- nr_inputs = MIN2( vs_input_count, draw->pt.nr_vertex_elements );
+ if (instance_id_index != ~0) {
+ num_extra_inputs++;
+ }
+
+ assert(draw->pt.nr_vertex_elements + num_extra_inputs >= vs_input_count);
+
+ nr_inputs = MIN2(vs_input_count, draw->pt.nr_vertex_elements + num_extra_inputs);
for (i = 0; i < nr_inputs; i++) {
- key.element[nr].input_format = draw->pt.vertex_element[i].src_format;
- key.element[nr].input_buffer = draw->pt.vertex_element[i].vertex_buffer_index;
- key.element[nr].input_offset = draw->pt.vertex_element[i].src_offset;
- key.element[nr].instance_divisor = draw->pt.vertex_element[i].instance_divisor;
+ if (i == instance_id_index) {
+ key.element[nr].type = TRANSLATE_ELEMENT_INSTANCE_ID;
+ key.element[nr].input_format = PIPE_FORMAT_R32_FLOAT; /* XXX: Make it UINT. */
+ } else {
+ key.element[nr].type = TRANSLATE_ELEMENT_NORMAL;
+ key.element[nr].input_format = draw->pt.vertex_element[ei].src_format;
+ key.element[nr].input_buffer = draw->pt.vertex_element[ei].vertex_buffer_index;
+ key.element[nr].input_offset = draw->pt.vertex_element[ei].src_offset;
+ key.element[nr].instance_divisor = draw->pt.vertex_element[ei].instance_divisor;
+ ei++;
+ }
key.element[nr].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
key.element[nr].output_offset = dst_offset;
continue;
}
+ key.element[i].type = TRANSLATE_ELEMENT_NORMAL;
key.element[i].input_format = input_format;
key.element[i].input_buffer = input_buffer;
key.element[i].input_offset = input_offset;
struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
struct draw_context *draw = fpme->draw;
struct draw_vertex_shader *vs = draw->vs.vertex_shader;
+ unsigned i;
+ boolean instance_id_index = ~0;
/* Add one to num_outputs because the pipeline occasionally tags on
* an additional texcoord, eg for AA lines.
unsigned nr = MAX2( vs->info.num_inputs,
vs->info.num_outputs + 1 );
+ /* Scan for instanceID system value.
+ */
+ for (i = 0; i < vs->info.num_inputs; i++) {
+ if (vs->info.input_semantic_name[i] == TGSI_SEMANTIC_INSTANCEID) {
+ instance_id_index = i;
+ break;
+ }
+ }
+
fpme->prim = prim;
fpme->opt = opt;
draw_pt_fetch_prepare( fpme->fetch,
vs->info.num_inputs,
- fpme->vertex_size );
+ fpme->vertex_size,
+ instance_id_index );
/* XXX: it's not really gl rasterization rules we care about here,
* but gl vs dx9 clip spaces.
*/
fetch.nr_elements = key->nr_inputs;
fetch.output_stride = vsvg->temp_vertex_stride;
for (i = 0; i < key->nr_inputs; i++) {
+ fetch.element[i].type = TRANSLATE_ELEMENT_NORMAL;
fetch.element[i].input_format = key->element[i].in.format;
fetch.element[i].input_buffer = key->element[i].in.buffer;
fetch.element[i].input_offset = key->element[i].in.offset;
for (i = 0; i < key->nr_outputs; i++) {
if (key->element[i].out.format != EMIT_1F_PSIZE)
{
+ emit.element[i].type = TRANSLATE_ELEMENT_NORMAL;
emit.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
emit.element[i].input_buffer = 0;
emit.element[i].input_offset = key->element[i].out.vs_output * 4 * sizeof(float);
assert(emit.element[i].input_offset <= fetch.output_stride);
}
else {
+ emit.element[i].type = TRANSLATE_ELEMENT_NORMAL;
emit.element[i].input_format = PIPE_FORMAT_R32_FLOAT;
emit.element[i].input_buffer = 1;
emit.element[i].input_offset = 0;
#include "pipe/p_format.h"
#include "pipe/p_state.h"
+enum translate_element_type {
+ TRANSLATE_ELEMENT_NORMAL,
+ TRANSLATE_ELEMENT_INSTANCE_ID
+};
+
struct translate_element
{
+ enum translate_element_type type;
enum pipe_format input_format;
enum pipe_format output_format;
unsigned input_buffer:8;
struct translate translate;
struct {
+ enum translate_element_type type;
+
fetch_func fetch;
unsigned buffer;
unsigned input_offset;
for (attr = 0; attr < nr_attrs; attr++) {
float data[4];
- const char *src;
char *dst = (vert +
tg->attrib[attr].output_offset);
- if (tg->attrib[attr].instance_divisor) {
- src = tg->attrib[attr].input_ptr +
- tg->attrib[attr].input_stride *
- (instance_id / tg->attrib[attr].instance_divisor);
+ if (tg->attrib[attr].type == TRANSLATE_ELEMENT_NORMAL) {
+ const char *src;
+
+ if (tg->attrib[attr].instance_divisor) {
+ src = tg->attrib[attr].input_ptr +
+ tg->attrib[attr].input_stride *
+ (instance_id / tg->attrib[attr].instance_divisor);
+ } else {
+ src = tg->attrib[attr].input_ptr +
+ tg->attrib[attr].input_stride * elt;
+ }
+
+ tg->attrib[attr].fetch( src, data );
} else {
- src = tg->attrib[attr].input_ptr +
- tg->attrib[attr].input_stride * elt;
+ data[0] = (float)instance_id;
}
- tg->attrib[attr].fetch( src, data );
-
if (0) debug_printf("vert %d attr %d: %f %f %f %f\n",
i, attr, data[0], data[1], data[2], data[3]);
tg->translate.run = generic_run;
for (i = 0; i < key->nr_elements; i++) {
+ tg->attrib[i].type = key->element[i].type;
tg->attrib[i].fetch = get_fetch_func(key->element[i].input_format);
tg->attrib[i].buffer = key->element[i].input_buffer;
unsigned start,
unsigned count,
unsigned instance_id,
- void *output_buffer );
+ void *output_buffer,
+ float instance_id_float );
typedef void (PIPE_CDECL *run_elts_func)( struct translate *translate,
const unsigned *elts,
unsigned count,
unsigned instance_id,
- void *output_buffer );
+ void *output_buffer,
+ float instance_id_float );
struct translate_buffer {
const void *base_ptr;
};
+#define ELEMENT_BUFFER_INSTANCE_ID 1001
+
+
struct translate_sse {
struct translate translate;
boolean use_instancing;
unsigned instance_id;
+ float instance_id_float; /* XXX: needed while no integer support in TGSI */
run_func gen_run;
run_elts_func gen_run_elts;
unsigned var_idx,
struct x86_reg elt )
{
+ if (var_idx == ELEMENT_BUFFER_INSTANCE_ID) {
+ return x86_make_disp(p->machine_EDX,
+ get_offset(p, &p->instance_id_float));
+ }
if (linear && p->nr_buffer_varients == 1) {
return p->idx_EBX;
}
x86_mov(p->func,
x86_make_disp(p->machine_EDX, get_offset(p, &p->instance_id)),
p->tmp_EAX);
+
+ /* XXX: temporary */
+ x86_mov(p->func,
+ p->tmp_EAX,
+ x86_fn_arg(p->func, 6));
+ x86_mov(p->func,
+ x86_make_disp(p->machine_EDX, get_offset(p, &p->instance_id_float)),
+ p->tmp_EAX);
}
/* Get vertex count, compare to zero
elts,
count,
instance_id,
- output_buffer );
+ output_buffer,
+ (float)instance_id );
}
static void PIPE_CDECL translate_sse_run( struct translate *translate,
start,
count,
instance_id,
- output_buffer );
+ output_buffer,
+ (float)instance_id);
}
p->translate.run = translate_sse_run;
for (i = 0; i < key->nr_elements; i++) {
- unsigned j;
+ if (key->element[i].type == TRANSLATE_ELEMENT_NORMAL) {
+ unsigned j;
- p->nr_buffers = MAX2( p->nr_buffers, key->element[i].input_buffer + 1 );
+ p->nr_buffers = MAX2(p->nr_buffers, key->element[i].input_buffer + 1);
- if (key->element[i].instance_divisor) {
- p->use_instancing = TRUE;
- }
+ if (key->element[i].instance_divisor) {
+ p->use_instancing = TRUE;
+ }
- /*
- * Map vertex element to vertex buffer varient.
- */
- for (j = 0; j < p->nr_buffer_varients; j++) {
- if (p->buffer_varient[j].buffer_index == key->element[i].input_buffer &&
- p->buffer_varient[j].instance_divisor == key->element[i].instance_divisor) {
- break;
+ /*
+ * Map vertex element to vertex buffer varient.
+ */
+ for (j = 0; j < p->nr_buffer_varients; j++) {
+ if (p->buffer_varient[j].buffer_index == key->element[i].input_buffer &&
+ p->buffer_varient[j].instance_divisor == key->element[i].instance_divisor) {
+ break;
+ }
}
+ if (j == p->nr_buffer_varients) {
+ p->buffer_varient[j].buffer_index = key->element[i].input_buffer;
+ p->buffer_varient[j].instance_divisor = key->element[i].instance_divisor;
+ p->nr_buffer_varients++;
+ }
+ p->element_to_buffer_varient[i] = j;
+ } else {
+ assert(key->element[i].type == TRANSLATE_ELEMENT_INSTANCE_ID);
+
+ p->element_to_buffer_varient[i] = ELEMENT_BUFFER_INSTANCE_ID;
}
- if (j == p->nr_buffer_varients) {
- p->buffer_varient[j].buffer_index = key->element[i].input_buffer;
- p->buffer_varient[j].instance_divisor = key->element[i].instance_divisor;
- p->nr_buffer_varients++;
- }
- p->element_to_buffer_varient[i] = j;
}
if (0) debug_printf("nr_buffers: %d\n", p->nr_buffers);
key.output_stride = 4 * sizeof(float);
key.nr_elements = 1;
+ key.element[0].type = TRANSLATE_ELEMENT_NORMAL;
key.element[0].input_format = vel->src_format;
key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
key.element[0].input_buffer = vel->vertex_buffer_index;