static void *
nv30_vertex_state_create(struct pipe_context *pipe, unsigned num_elements,
const struct pipe_vertex_element *elements)
{
struct nv30_vertex_stateobj *so;
struct translate_key transkey;
unsigned i;
assert(num_elements);
so = MALLOC(sizeof(*so) + sizeof(*so->element) * num_elements);
if (!so)
return NULL;
memcpy(so->pipe, elements, sizeof(*elements) * num_elements);
so->num_elements = num_elements;
so->need_conversion = FALSE;
transkey.nr_elements = 0;
transkey.output_stride = 0;
for (i = 0; i < num_elements; i++) {
const struct pipe_vertex_element *ve = &elements[i];
const unsigned vbi = ve->vertex_buffer_index;
enum pipe_format fmt = ve->src_format;
so->element[i].state = nv30_vtxfmt(pipe->screen, fmt)->hw;
if (!so->element[i].state) {
switch (util_format_get_nr_components(fmt)) {
case 1: fmt = PIPE_FORMAT_R32_FLOAT; break;
case 2: fmt = PIPE_FORMAT_R32G32_FLOAT; break;
case 3: fmt = PIPE_FORMAT_R32G32B32_FLOAT; break;
case 4: fmt = PIPE_FORMAT_R32G32B32A32_FLOAT; break;
default:
assert(0);
return NULL;
}
so->element[i].state = nv30_vtxfmt(pipe->screen, fmt)->hw;
so->need_conversion = TRUE;
}
if (1) {
unsigned j = transkey.nr_elements++;
transkey.element[j].type = TRANSLATE_ELEMENT_NORMAL;
transkey.element[j].input_format = ve->src_format;
transkey.element[j].input_buffer = vbi;
transkey.element[j].input_offset = ve->src_offset;
transkey.element[j].instance_divisor = ve->instance_divisor;
transkey.element[j].output_format = fmt;
transkey.element[j].output_offset = transkey.output_stride;
transkey.output_stride += (util_format_get_stride(fmt, 1) + 3) & ~3;
}
}
so->translate = translate_create(&transkey);
so->vtx_size = transkey.output_stride / 4;
so->vtx_per_packet_max = NV04_PFIFO_MAX_PACKET_LEN / MAX2(so->vtx_size, 1);
return so;
}
struct translate * translate_cache_find(struct translate_cache *cache,
struct translate_key *key)
{
unsigned hash_key = create_key(key);
struct translate *translate = (struct translate*)
cso_hash_find_data_from_template(cache->hash,
hash_key,
key, sizeof(*key));
if (!translate) {
/* create/insert */
translate = translate_create(key);
cso_hash_insert(cache->hash, hash_key, translate);
}
return translate;
}
static void *
nvfx_vtxelts_state_create(struct pipe_context *pipe,
unsigned num_elements,
const struct pipe_vertex_element *elements)
{
struct nvfx_vtxelt_state *cso = CALLOC_STRUCT(nvfx_vtxelt_state);
struct translate_key transkey;
unsigned per_vertex_size[16];
unsigned vb_compacted_index[16];
if(num_elements > 16)
{
_debug_printf("Error: application attempted to use %u vertex elements, but only 16 are supported: ignoring the rest\n", num_elements);
num_elements = 16;
}
memset(per_vertex_size, 0, sizeof(per_vertex_size));
memcpy(cso->pipe, elements, num_elements * sizeof(elements[0]));
cso->num_elements = num_elements;
cso->needs_translate = FALSE;
transkey.nr_elements = 0;
transkey.output_stride = 0;
for(unsigned i = 0; i < num_elements; ++i)
{
const struct pipe_vertex_element* ve = &elements[i];
if(!ve->instance_divisor)
per_vertex_size[ve->vertex_buffer_index] += util_format_get_stride(ve->src_format, 1);
}
for(unsigned i = 0; i < 16; ++i)
{
if(per_vertex_size[i])
{
unsigned idx = cso->num_per_vertex_buffer_infos++;
cso->per_vertex_buffer_info[idx].vertex_buffer_index = i;
cso->per_vertex_buffer_info[idx].per_vertex_size = per_vertex_size[i];
vb_compacted_index[i] = idx;
}
}
for(unsigned i = 0; i < num_elements; ++i)
{
const struct pipe_vertex_element* ve = &elements[i];
unsigned type = nvfx_vertex_formats[ve->src_format];
unsigned ncomp = util_format_get_nr_components(ve->src_format);
//if(ve->frequency != PIPE_ELEMENT_FREQUENCY_PER_VERTEX)
if(ve->instance_divisor)
{
struct nvfx_low_frequency_element* lfve;
cso->vtxfmt[i] = NV30_3D_VTXFMT_TYPE_V32_FLOAT;
//if(ve->frequency == PIPE_ELEMENT_FREQUENCY_CONSTANT)
if(0)
lfve = &cso->constant[cso->num_constant++];
else
{
lfve = &cso->per_instance[cso->num_per_instance++].base;
((struct nvfx_per_instance_element*)lfve)->instance_divisor = ve->instance_divisor;
}
lfve->idx = i;
lfve->vertex_buffer_index = ve->vertex_buffer_index;
lfve->src_offset = ve->src_offset;
lfve->fetch_rgba_float = util_format_description(ve->src_format)->fetch_rgba_float;
lfve->ncomp = ncomp;
}
else
{
unsigned idx;
idx = cso->num_per_vertex++;
cso->per_vertex[idx].idx = i;
cso->per_vertex[idx].vertex_buffer_index = ve->vertex_buffer_index;
cso->per_vertex[idx].src_offset = ve->src_offset;
idx = transkey.nr_elements++;
transkey.element[idx].input_format = ve->src_format;
transkey.element[idx].input_buffer = vb_compacted_index[ve->vertex_buffer_index];
transkey.element[idx].input_offset = ve->src_offset;
transkey.element[idx].instance_divisor = 0;
transkey.element[idx].type = TRANSLATE_ELEMENT_NORMAL;
if(type)
{
transkey.element[idx].output_format = ve->src_format;
cso->vtxfmt[i] = (ncomp << NV30_3D_VTXFMT_SIZE__SHIFT) | type;
}
else
{
unsigned float32[4] = {PIPE_FORMAT_R32_FLOAT, PIPE_FORMAT_R32G32_FLOAT, PIPE_FORMAT_R32G32B32_FLOAT, PIPE_FORMAT_R32G32B32A32_FLOAT};
transkey.element[idx].output_format = float32[ncomp - 1];
cso->needs_translate = TRUE;
cso->vtxfmt[i] = (ncomp << NV30_3D_VTXFMT_SIZE__SHIFT) | NV30_3D_VTXFMT_TYPE_V32_FLOAT;
}
transkey.element[idx].output_offset = transkey.output_stride;
transkey.output_stride += (util_format_get_stride(transkey.element[idx].output_format, 1) + 3) & ~3;
}
}
cso->translate = translate_create(&transkey);
cso->vertex_length = transkey.output_stride >> 2;
cso->max_vertices_per_packet = 2047 / MAX2(cso->vertex_length, 1);
return (void *)cso;
}
