fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2025-12-23 19:50:11 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

vc4: Move shader record setup before the draw call.

Browse source 
The flush only happens after both are written, so we can do them in either
order.  This will let me compute max_index during the shader record setup.
This commit is contained in:
Eric Anholt 2014-08-12 14:57:39 -07:00
parent ba0c0a186d
commit 61cb08ab4f
1 changed files with 38 additions and 38 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

76
src/gallium/drivers/vc4/vc4_draw.c
View file

@ -96,49 +96,14 @@ vc4_draw_vbo(struct pipe_context *pctx, const struct pipe_draw_info *info)
return; return;
} }
struct vc4_vertex_stateobj *vtx = vc4->vtx;
struct vc4_vertexbuf_stateobj *vertexbuf = &vc4->vertexbuf;
vc4_start_draw(vc4); vc4_start_draw(vc4);
vc4_update_compiled_shaders(vc4, info->mode); vc4_update_compiled_shaders(vc4, info->mode);
vc4_emit_state(pctx); vc4_emit_state(pctx);
/* the actual draw call. */
struct vc4_vertex_stateobj *vtx = vc4->vtx;
struct vc4_vertexbuf_stateobj *vertexbuf = &vc4->vertexbuf;
cl_u8(&vc4->bcl, VC4_PACKET_GL_SHADER_STATE);
assert(vtx->num_elements <= 8);
/* Note that number of attributes == 0 in the packet means 8
* attributes. This field also contains the offset into shader_rec.
*/
cl_u32(&vc4->bcl, vtx->num_elements & 0x7);
/* Note that the primitive type fields match with OpenGL/gallium
* definitions, up to but not including QUADS.
*/
if (info->indexed) {
struct vc4_resource *rsc = vc4_resource(vc4->indexbuf.buffer);
assert(vc4->indexbuf.index_size == 1 ||
vc4->indexbuf.index_size == 2);
cl_start_reloc(&vc4->bcl, 1);
cl_u8(&vc4->bcl, VC4_PACKET_GL_INDEXED_PRIMITIVE);
cl_u8(&vc4->bcl,
info->mode |
(vc4->indexbuf.index_size == 2 ?
VC4_INDEX_BUFFER_U16:
VC4_INDEX_BUFFER_U8));
cl_u32(&vc4->bcl, info->count);
cl_reloc(vc4, &vc4->bcl, rsc->bo, vc4->indexbuf.offset);
cl_u32(&vc4->bcl, info->max_index);
} else {
cl_u8(&vc4->bcl, VC4_PACKET_GL_ARRAY_PRIMITIVE);
cl_u8(&vc4->bcl, info->mode);
cl_u32(&vc4->bcl, info->count);
cl_u32(&vc4->bcl, info->start);
}
// Shader Record
vc4_write_uniforms(vc4, vc4->prog.fs, vc4_write_uniforms(vc4, vc4->prog.fs,
&vc4->constbuf[PIPE_SHADER_FRAGMENT], &vc4->constbuf[PIPE_SHADER_FRAGMENT],
&vc4->fragtex, &vc4->fragtex,
@ -152,6 +117,7 @@ vc4_draw_vbo(struct pipe_context *pctx, const struct pipe_draw_info *info)
&vc4->verttex, &vc4->verttex,
1); 1);
/* Emit the shader record. */
cl_start_shader_reloc(&vc4->shader_rec, 3 + vtx->num_elements); cl_start_shader_reloc(&vc4->shader_rec, 3 + vtx->num_elements);
cl_u16(&vc4->shader_rec, cl_u16(&vc4->shader_rec,
VC4_SHADER_FLAG_ENABLE_CLIPPING | VC4_SHADER_FLAG_ENABLE_CLIPPING |
@ -191,6 +157,40 @@ vc4_draw_vbo(struct pipe_context *pctx, const struct pipe_draw_info *info)
cl_u8(&vc4->shader_rec, i * 16); /* CS VPM offset */ cl_u8(&vc4->shader_rec, i * 16); /* CS VPM offset */
} }
/* the actual draw call. */
cl_u8(&vc4->bcl, VC4_PACKET_GL_SHADER_STATE);
assert(vtx->num_elements <= 8);
/* Note that number of attributes == 0 in the packet means 8
* attributes. This field also contains the offset into shader_rec.
*/
cl_u32(&vc4->bcl, vtx->num_elements & 0x7);
/* Note that the primitive type fields match with OpenGL/gallium
* definitions, up to but not including QUADS.
*/
if (info->indexed) {
struct vc4_resource *rsc = vc4_resource(vc4->indexbuf.buffer);
assert(vc4->indexbuf.index_size == 1 ||
vc4->indexbuf.index_size == 2);
cl_start_reloc(&vc4->bcl, 1);
cl_u8(&vc4->bcl, VC4_PACKET_GL_INDEXED_PRIMITIVE);
cl_u8(&vc4->bcl,
info->mode |
(vc4->indexbuf.index_size == 2 ?
VC4_INDEX_BUFFER_U16:
VC4_INDEX_BUFFER_U8));
cl_u32(&vc4->bcl, info->count);
cl_reloc(vc4, &vc4->bcl, rsc->bo, vc4->indexbuf.offset);
cl_u32(&vc4->bcl, info->max_index);
} else {
cl_u8(&vc4->bcl, VC4_PACKET_GL_ARRAY_PRIMITIVE);
cl_u8(&vc4->bcl, info->mode);
cl_u32(&vc4->bcl, info->count);
cl_u32(&vc4->bcl, info->start);
}
if (vc4->zsa && vc4->zsa->base.depth.enabled) { if (vc4->zsa && vc4->zsa->base.depth.enabled) {
vc4->resolve |= PIPE_CLEAR_DEPTH; vc4->resolve |= PIPE_CLEAR_DEPTH;
} }