/* Perform the fetch from API vertex elements & vertex buffers, to a
* contiguous set of float[4] attributes as required for the
* vertex_shader->run_linear() method.
*
* This is used in all cases except pure passthrough
* (draw_pt_fetch_emit.c) which has its own version to translate
* directly to hw vertices.
*
*/
void draw_pt_fetch_prepare( struct pt_fetch *fetch,
unsigned vs_input_count,
unsigned vertex_size )
{
struct draw_context *draw = fetch->draw;
unsigned nr_inputs;
unsigned i, nr = 0;
unsigned dst_offset = 0;
struct translate_key key;
fetch->vertex_size = vertex_size;
/* Always emit/leave space for a vertex header.
*
* It's worth considering whether the vertex headers should contain
* a pointer to the 'data', rather than having it inline.
* Something to look at after we've fully switched over to the pt
* paths.
*/
{
/* Need to set header->vertex_id = 0xffff somehow.
*/
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;
key.element[nr].output_format = PIPE_FORMAT_R32_FLOAT;
key.element[nr].output_offset = dst_offset;
dst_offset += 1 * sizeof(float);
nr++;
/* Just leave the clip[] array untouched.
*/
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 );
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].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
key.element[nr].output_offset = dst_offset;
dst_offset += 4 * sizeof(float);
nr++;
}
assert(dst_offset <= vertex_size);
key.nr_elements = nr;
key.output_stride = vertex_size;
if (!fetch->translate ||
translate_key_compare(&fetch->translate->key, &key) != 0)
{
translate_key_sanitize(&key);
fetch->translate = translate_cache_find(fetch->cache, &key);
{
static struct vertex_header vh = { 0, 1, 0, UNDEFINED_VERTEX_ID, { .0f, .0f, .0f, .0f } };
fetch->translate->set_buffer(fetch->translate,
draw->pt.nr_vertex_buffers,
&vh,
0);
}
}
fetch->need_edgeflags = ((draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL ||
draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) &&
draw->pt.user.edgeflag);
}
static int update_zero_stride( struct svga_context *svga,
unsigned dirty )
{
unsigned i;
svga->curr.zero_stride_vertex_elements = 0;
svga->curr.num_zero_stride_vertex_elements = 0;
for (i = 0; i < svga->curr.num_vertex_elements; i++) {
const struct pipe_vertex_element *vel = &svga->curr.ve[i];
const struct pipe_vertex_buffer *vbuffer = &svga->curr.vb[
vel->vertex_buffer_index];
if (vbuffer->stride == 0) {
unsigned const_idx =
svga->curr.num_zero_stride_vertex_elements;
struct translate *translate;
struct translate_key key;
void *mapped_buffer;
svga->curr.zero_stride_vertex_elements |= (1 << i);
++svga->curr.num_zero_stride_vertex_elements;
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;
key.element[0].input_offset = vel->src_offset;
key.element[0].instance_divisor = vel->instance_divisor;
key.element[0].output_offset = const_idx * 4 * sizeof(float);
translate_key_sanitize(&key);
/* translate_generic_create is technically private but
* we don't want to code-generate, just want generic
* translation */
translate = translate_generic_create(&key);
assert(vel->src_offset == 0);
mapped_buffer = pipe_buffer_map_range(svga->pipe.screen,
vbuffer->buffer,
vel->src_offset,
util_format_get_blocksize(vel->src_format),
PIPE_BUFFER_USAGE_CPU_READ);
translate->set_buffer(translate, vel->vertex_buffer_index,
mapped_buffer,
vbuffer->stride);
translate->run(translate, 0, 1, 0,
svga->curr.zero_stride_constants);
pipe_buffer_unmap(svga->pipe.screen,
vbuffer->buffer);
translate->release(translate);
}
}
if (svga->curr.num_zero_stride_vertex_elements)
svga->dirty |= SVGA_NEW_ZERO_STRIDE;
return 0;
}
void
draw_pt_emit_prepare(struct pt_emit *emit,
unsigned prim,
unsigned *max_vertices)
{
struct draw_context *draw = emit->draw;
const struct vertex_info *vinfo;
unsigned dst_offset;
struct translate_key hw_key;
unsigned i;
/* XXX: need to flush to get prim_vbuf.c to release its allocation??
*/
draw_do_flush( draw, DRAW_FLUSH_BACKEND );
/* XXX: may need to defensively reset this later on as clipping can
* clobber this state in the render backend.
*/
emit->prim = prim;
draw->render->set_primitive(draw->render, emit->prim);
/* Must do this after set_primitive() above:
*/
emit->vinfo = vinfo = draw->render->get_vertex_info(draw->render);
/* Translate from pipeline vertices to hw vertices.
*/
dst_offset = 0;
for (i = 0; i < vinfo->num_attribs; i++) {
unsigned emit_sz = 0;
unsigned src_buffer = 0;
unsigned output_format;
unsigned src_offset = (vinfo->attrib[i].src_index * 4 * sizeof(float) );
output_format = draw_translate_vinfo_format(vinfo->attrib[i].emit);
emit_sz = draw_translate_vinfo_size(vinfo->attrib[i].emit);
/* doesn't handle EMIT_OMIT */
assert(emit_sz != 0);
if (vinfo->attrib[i].emit == EMIT_1F_PSIZE) {
src_buffer = 1;
src_offset = 0;
}
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;
hw_key.element[i].instance_divisor = 0;
hw_key.element[i].output_format = output_format;
hw_key.element[i].output_offset = dst_offset;
dst_offset += emit_sz;
}
hw_key.nr_elements = vinfo->num_attribs;
hw_key.output_stride = vinfo->size * 4;
if (!emit->translate ||
translate_key_compare(&emit->translate->key, &hw_key) != 0) {
translate_key_sanitize(&hw_key);
emit->translate = translate_cache_find(emit->cache, &hw_key);
}
*max_vertices = (draw->render->max_vertex_buffer_bytes /
(vinfo->size * 4));
}
static void fetch_emit_prepare( struct draw_pt_middle_end *middle,
unsigned prim,
unsigned opt,
unsigned *max_vertices )
{
struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
struct draw_context *draw = feme->draw;
const struct vertex_info *vinfo;
unsigned i, dst_offset;
boolean ok;
struct translate_key key;
ok = draw->render->set_primitive( draw->render,
prim );
if (!ok) {
assert(0);
return;
}
/* Must do this after set_primitive() above:
*/
vinfo = feme->vinfo = draw->render->get_vertex_info(draw->render);
/* Transform from API vertices to HW vertices, skipping the
* pipeline_vertex intermediate step.
*/
dst_offset = 0;
memset(&key, 0, sizeof(key));
for (i = 0; i < vinfo->num_attribs; i++) {
const struct pipe_vertex_element *src = &draw->pt.vertex_element[vinfo->attrib[i].src_index];
unsigned emit_sz = 0;
unsigned input_format = src->src_format;
unsigned input_buffer = src->vertex_buffer_index;
unsigned input_offset = src->src_offset;
unsigned output_format;
switch (vinfo->attrib[i].emit) {
case EMIT_4UB:
output_format = PIPE_FORMAT_R8G8B8A8_UNORM;
emit_sz = 4 * sizeof(unsigned char);
break;
case EMIT_4F:
output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
emit_sz = 4 * sizeof(float);
break;
case EMIT_3F:
output_format = PIPE_FORMAT_R32G32B32_FLOAT;
emit_sz = 3 * sizeof(float);
break;
case EMIT_2F:
output_format = PIPE_FORMAT_R32G32_FLOAT;
emit_sz = 2 * sizeof(float);
break;
case EMIT_1F:
output_format = PIPE_FORMAT_R32_FLOAT;
emit_sz = 1 * sizeof(float);
break;
case EMIT_1F_PSIZE:
input_format = PIPE_FORMAT_R32_FLOAT;
input_buffer = draw->pt.nr_vertex_buffers;
input_offset = 0;
output_format = PIPE_FORMAT_R32_FLOAT;
emit_sz = 1 * sizeof(float);
break;
case EMIT_OMIT:
continue;
default:
assert(0);
output_format = PIPE_FORMAT_NONE;
emit_sz = 0;
continue;
}
key.element[i].input_format = input_format;
key.element[i].input_buffer = input_buffer;
key.element[i].input_offset = input_offset;
key.element[i].output_format = output_format;
key.element[i].output_offset = dst_offset;
dst_offset += emit_sz;
}
key.nr_elements = vinfo->num_attribs;
key.output_stride = vinfo->size * 4;
/* Don't bother with caching at this stage:
*/
if (!feme->translate ||
translate_key_compare(&feme->translate->key, &key) != 0)
{
translate_key_sanitize(&key);
feme->translate = translate_cache_find(feme->cache,
&key);
feme->translate->set_buffer(feme->translate,
draw->pt.nr_vertex_buffers,
&feme->point_size,
0);
}
feme->point_size = draw->rasterizer->point_size;
for (i = 0; i < draw->pt.nr_vertex_buffers; i++) {
feme->translate->set_buffer(feme->translate,
i,
((char *)draw->pt.user.vbuffer[i] +
draw->pt.vertex_buffer[i].buffer_offset),
draw->pt.vertex_buffer[i].stride );
}
*max_vertices = (draw->render->max_vertex_buffer_bytes /
(vinfo->size * 4));
/* Return an even number of verts.
* This prevents "parity" errors when splitting long triangle strips which
* can lead to front/back culling mix-ups.
* Every other triangle in a strip has an alternate front/back orientation
* so splitting at an odd position can cause the orientation of subsequent
* triangles to get reversed.
*/
*max_vertices = *max_vertices & ~1;
}
static void fetch_emit_prepare( struct draw_pt_middle_end *middle,
unsigned prim,
unsigned opt,
unsigned *max_vertices )
{
struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
struct draw_context *draw = feme->draw;
const struct vertex_info *vinfo;
unsigned i, dst_offset;
boolean ok;
struct translate_key key;
unsigned gs_out_prim = (draw->gs.geometry_shader ?
draw->gs.geometry_shader->output_primitive :
prim);
ok = draw->render->set_primitive( draw->render,
gs_out_prim );
if (!ok) {
assert(0);
return;
}
/* Must do this after set_primitive() above:
*/
vinfo = feme->vinfo = draw->render->get_vertex_info(draw->render);
/* Transform from API vertices to HW vertices, skipping the
* pipeline_vertex intermediate step.
*/
dst_offset = 0;
memset(&key, 0, sizeof(key));
for (i = 0; i < vinfo->num_attribs; i++) {
const struct pipe_vertex_element *src = &draw->pt.vertex_element[vinfo->attrib[i].src_index];
unsigned emit_sz = 0;
unsigned input_format = src->src_format;
unsigned input_buffer = src->vertex_buffer_index;
unsigned input_offset = src->src_offset;
unsigned output_format;
output_format = draw_translate_vinfo_format(vinfo->attrib[i].emit);
emit_sz = draw_translate_vinfo_size(vinfo->attrib[i].emit);
if (vinfo->attrib[i].emit == EMIT_OMIT)
continue;
if (vinfo->attrib[i].emit == EMIT_1F_PSIZE) {
input_format = PIPE_FORMAT_R32_FLOAT;
input_buffer = draw->pt.nr_vertex_buffers;
input_offset = 0;
}
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;
key.element[i].instance_divisor = src->instance_divisor;
key.element[i].output_format = output_format;
key.element[i].output_offset = dst_offset;
dst_offset += emit_sz;
}
key.nr_elements = vinfo->num_attribs;
key.output_stride = vinfo->size * 4;
/* Don't bother with caching at this stage:
*/
if (!feme->translate ||
translate_key_compare(&feme->translate->key, &key) != 0)
{
translate_key_sanitize(&key);
feme->translate = translate_cache_find(feme->cache,
&key);
feme->translate->set_buffer(feme->translate,
draw->pt.nr_vertex_buffers,
&feme->point_size,
0,
~0);
}
feme->point_size = draw->rasterizer->point_size;
for (i = 0; i < draw->pt.nr_vertex_buffers; i++) {
feme->translate->set_buffer(feme->translate,
i,
((char *)draw->pt.user.vbuffer[i] +
draw->pt.vertex_buffer[i].buffer_offset),
draw->pt.vertex_buffer[i].stride,
draw->pt.max_index);
}
*max_vertices = (draw->render->max_vertex_buffer_bytes /
(vinfo->size * 4));
}
/**
* Perform the fetch from API vertex elements & vertex buffers, to a
* contiguous set of float[4] attributes as required for the
* vertex_shader->run_linear() method.
*
* This is used in all cases except pure passthrough
* (draw_pt_fetch_emit.c) which has its own version to translate
* directly to hw vertices.
*
*/
void
draw_pt_fetch_prepare(struct pt_fetch *fetch,
unsigned vs_input_count,
unsigned vertex_size,
unsigned instance_id_index)
{
struct draw_context *draw = fetch->draw;
unsigned nr_inputs;
unsigned i, nr = 0, ei = 0;
unsigned dst_offset = 0;
unsigned num_extra_inputs = 0;
struct translate_key key;
fetch->vertex_size = vertex_size;
/* Leave the clipmask/edgeflags/pad/vertex_id untouched
*/
dst_offset += 1 * sizeof(float);
/* Just leave the clip[] and pre_clip_pos[] array untouched.
*/
dst_offset += 8 * sizeof(float);
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++) {
if (i == instance_id_index) {
key.element[nr].type = TRANSLATE_ELEMENT_INSTANCE_ID;
key.element[nr].input_format = PIPE_FORMAT_R32_USCALED;
key.element[nr].output_format = PIPE_FORMAT_R32_USCALED;
key.element[nr].output_offset = dst_offset;
dst_offset += sizeof(uint);
} else if (util_format_is_pure_sint(draw->pt.vertex_element[i].src_format)) {
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;
key.element[nr].output_format = PIPE_FORMAT_R32G32B32A32_SINT;
key.element[nr].output_offset = dst_offset;
ei++;
dst_offset += 4 * sizeof(int);
} else if (util_format_is_pure_uint(draw->pt.vertex_element[i].src_format)) {
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;
key.element[nr].output_format = PIPE_FORMAT_R32G32B32A32_UINT;
key.element[nr].output_offset = dst_offset;
ei++;
dst_offset += 4 * sizeof(unsigned);
} 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;
key.element[nr].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
key.element[nr].output_offset = dst_offset;
ei++;
dst_offset += 4 * sizeof(float);
}
nr++;
}
assert(dst_offset <= vertex_size);
key.nr_elements = nr;
key.output_stride = vertex_size;
if (!fetch->translate ||
translate_key_compare(&fetch->translate->key, &key) != 0)
{
translate_key_sanitize(&key);
fetch->translate = translate_cache_find(fetch->cache, &key);
}
}
/**
* Set the prim type for subsequent vertices.
* This may result in a new vertex size. The existing vbuffer (if any)
* will be flushed if needed and a new one allocated.
*/
static void
vbuf_start_prim( struct vbuf_stage *vbuf, uint prim )
{
struct translate_key hw_key;
unsigned dst_offset;
unsigned i;
vbuf->render->set_primitive(vbuf->render, prim);
/* Must do this after set_primitive() above:
*
* XXX: need some state managment to track when this needs to be
* recalculated. The driver should tell us whether there was a
* state change.
*/
vbuf->vinfo = vbuf->render->get_vertex_info(vbuf->render);
vbuf->vertex_size = vbuf->vinfo->size * sizeof(float);
/* Translate from pipeline vertices to hw vertices.
*/
dst_offset = 0;
for (i = 0; i < vbuf->vinfo->num_attribs; i++) {
unsigned emit_sz = 0;
unsigned src_buffer = 0;
unsigned output_format;
unsigned src_offset = (vbuf->vinfo->attrib[i].src_index * 4 * sizeof(float) );
switch (vbuf->vinfo->attrib[i].emit) {
case EMIT_4F:
output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
emit_sz = 4 * sizeof(float);
break;
case EMIT_3F:
output_format = PIPE_FORMAT_R32G32B32_FLOAT;
emit_sz = 3 * sizeof(float);
break;
case EMIT_2F:
output_format = PIPE_FORMAT_R32G32_FLOAT;
emit_sz = 2 * sizeof(float);
break;
case EMIT_1F:
output_format = PIPE_FORMAT_R32_FLOAT;
emit_sz = 1 * sizeof(float);
break;
case EMIT_1F_PSIZE:
output_format = PIPE_FORMAT_R32_FLOAT;
emit_sz = 1 * sizeof(float);
src_buffer = 1;
src_offset = 0;
break;
case EMIT_4UB:
output_format = PIPE_FORMAT_B8G8R8A8_UNORM;
emit_sz = 4 * sizeof(ubyte);
break;
default:
assert(0);
output_format = PIPE_FORMAT_NONE;
emit_sz = 0;
break;
}
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;
hw_key.element[i].output_format = output_format;
hw_key.element[i].output_offset = dst_offset;
dst_offset += emit_sz;
}
hw_key.nr_elements = vbuf->vinfo->num_attribs;
hw_key.output_stride = vbuf->vinfo->size * 4;
/* Don't bother with caching at this stage:
*/
if (!vbuf->translate ||
translate_key_compare(&vbuf->translate->key, &hw_key) != 0)
{
translate_key_sanitize(&hw_key);
vbuf->translate = translate_cache_find(vbuf->cache, &hw_key);
vbuf->translate->set_buffer(vbuf->translate, 1, &vbuf->point_size, 0);
}
vbuf->point_size = vbuf->stage.draw->rasterizer->point_size;
/* Allocate new buffer?
*/
assert(vbuf->vertices == NULL);
vbuf_alloc_vertices(vbuf);
}
/**
* Set the prim type for subsequent vertices.
* This may result in a new vertex size. The existing vbuffer (if any)
* will be flushed if needed and a new one allocated.
*/
static void
vbuf_start_prim( struct vbuf_stage *vbuf, uint prim )
{
struct translate_key hw_key;
unsigned dst_offset;
unsigned i;
const struct vertex_info *vinfo;
vbuf->render->set_primitive(vbuf->render, prim);
/* Must do this after set_primitive() above:
*
* XXX: need some state managment to track when this needs to be
* recalculated. The driver should tell us whether there was a
* state change.
*/
vbuf->vinfo = vbuf->render->get_vertex_info(vbuf->render);
vinfo = vbuf->vinfo;
vbuf->vertex_size = vinfo->size * sizeof(float);
/* Translate from pipeline vertices to hw vertices.
*/
dst_offset = 0;
for (i = 0; i < vinfo->num_attribs; i++) {
unsigned emit_sz = 0;
unsigned src_buffer = 0;
enum pipe_format output_format;
unsigned src_offset = (vinfo->attrib[i].src_index * 4 * sizeof(float) );
output_format = draw_translate_vinfo_format(vinfo->attrib[i].emit);
emit_sz = draw_translate_vinfo_size(vinfo->attrib[i].emit);
/* doesn't handle EMIT_OMIT */
assert(emit_sz != 0);
if (vinfo->attrib[i].emit == EMIT_1F_PSIZE) {
src_buffer = 1;
src_offset = 0;
}
else if (vinfo->attrib[i].src_index == DRAW_ATTR_NONEXIST) {
/* elements which don't exist will get assigned zeros */
src_buffer = 2;
src_offset = 0;
}
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;
hw_key.element[i].instance_divisor = 0;
hw_key.element[i].output_format = output_format;
hw_key.element[i].output_offset = dst_offset;
dst_offset += emit_sz;
}
hw_key.nr_elements = vinfo->num_attribs;
hw_key.output_stride = vbuf->vertex_size;
/* Don't bother with caching at this stage:
*/
if (!vbuf->translate ||
translate_key_compare(&vbuf->translate->key, &hw_key) != 0)
{
translate_key_sanitize(&hw_key);
vbuf->translate = translate_cache_find(vbuf->cache, &hw_key);
vbuf->translate->set_buffer(vbuf->translate, 1, &vbuf->point_size, 0, ~0);
vbuf->translate->set_buffer(vbuf->translate, 2, &vbuf->zero4[0], 0, ~0);
}
vbuf->point_size = vbuf->stage.draw->rasterizer->point_size;
/* Allocate new buffer?
*/
assert(vbuf->vertices == NULL);
vbuf_alloc_vertices(vbuf);
}
