static int emit_consts( struct svga_context *svga,
int offset,
int unit )
{
struct pipe_transfer *transfer = NULL;
unsigned count;
const float (*data)[4] = NULL;
unsigned i;
int ret = PIPE_OK;
if (svga->curr.cb[unit] == NULL)
goto done;
count = svga->curr.cb[unit]->width0 / (4 * sizeof(float));
data = (const float (*)[4])pipe_buffer_map(&svga->pipe,
svga->curr.cb[unit],
PIPE_TRANSFER_READ,
&transfer);
if (data == NULL) {
ret = PIPE_ERROR_OUT_OF_MEMORY;
goto done;
}
for (i = 0; i < count; i++) {
ret = emit_const( svga, unit, offset + i, data[i] );
if (ret)
goto done;
}
done:
if (data)
pipe_buffer_unmap(&svga->pipe, svga->curr.cb[unit], transfer);
return ret;
}
static void r300_set_constant_buffer(struct pipe_context *pipe,
uint shader, uint index,
const struct pipe_constant_buffer *buf)
{
struct r300_context* r300 = r300_context(pipe);
void *mapped;
if (buf == NULL || buf->buffer->size == 0 ||
(mapped = pipe_buffer_map(pipe->screen, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ)) == NULL)
{
r300->shader_constants[shader].count = 0;
return;
}
assert((buf->buffer->size % 4 * sizeof(float)) == 0);
memcpy(r300->shader_constants[shader].constants, mapped, buf->buffer->size);
r300->shader_constants[shader].count = buf->buffer->size / (4 * sizeof(float));
pipe_buffer_unmap(pipe->screen, buf->buffer);
if (shader == PIPE_SHADER_VERTEX)
r300->dirty_state |= R300_NEW_VERTEX_SHADER_CONSTANTS;
else if (shader == PIPE_SHADER_FRAGMENT)
r300->dirty_state |= R300_NEW_FRAGMENT_SHADER_CONSTANTS;
}
static void *
svga_vbuf_render_map_vertices( struct vbuf_render *render )
{
struct svga_vbuf_render *svga_render = svga_vbuf_render(render);
struct svga_context *svga = svga_render->svga;
if (svga_render->vbuf) {
char *ptr = (char*)pipe_buffer_map(&svga->pipe,
svga_render->vbuf,
PIPE_TRANSFER_WRITE |
PIPE_TRANSFER_FLUSH_EXPLICIT |
PIPE_TRANSFER_DISCARD |
PIPE_TRANSFER_UNSYNCHRONIZED,
&svga_render->vbuf_transfer);
if (ptr)
return ptr + svga_render->vbuf_offset;
else
return NULL;
}
else {
/* we probably ran out of memory when allocating the vertex buffer */
return NULL;
}
}
/**
* Called by VBO to draw arrays when in selection or feedback mode and
* to implement glRasterPos.
* This is very much like the normal draw_vbo() function above.
* Look at code refactoring some day.
*/
void
st_feedback_draw_vbo(struct gl_context *ctx,
const struct gl_client_array **arrays,
const struct _mesa_prim *prims,
GLuint nr_prims,
const struct _mesa_index_buffer *ib,
GLboolean index_bounds_valid,
GLuint min_index,
GLuint max_index,
struct gl_transform_feedback_object *tfb_vertcount)
{
struct st_context *st = st_context(ctx);
struct pipe_context *pipe = st->pipe;
struct draw_context *draw = st->draw;
const struct st_vertex_program *vp;
const struct pipe_shader_state *vs;
struct pipe_vertex_buffer vbuffers[PIPE_MAX_SHADER_INPUTS];
struct pipe_vertex_element velements[PIPE_MAX_ATTRIBS];
struct pipe_index_buffer ibuffer;
struct pipe_transfer *vb_transfer[PIPE_MAX_ATTRIBS];
struct pipe_transfer *ib_transfer = NULL;
GLuint attr, i;
const GLubyte *low_addr = NULL;
const void *mapped_indices = NULL;
assert(draw);
st_validate_state(st);
if (!index_bounds_valid)
vbo_get_minmax_indices(ctx, prims, ib, &min_index, &max_index, nr_prims);
/* must get these after state validation! */
vp = st->vp;
vs = &st->vp_variant->tgsi;
if (!st->vp_variant->draw_shader) {
st->vp_variant->draw_shader = draw_create_vertex_shader(draw, vs);
}
/*
* Set up the draw module's state.
*
* We'd like to do this less frequently, but the normal state-update
* code sends state updates to the pipe, not to our private draw module.
*/
assert(draw);
draw_set_viewport_state(draw, &st->state.viewport);
draw_set_clip_state(draw, &st->state.clip);
draw_set_rasterizer_state(draw, &st->state.rasterizer, NULL);
draw_bind_vertex_shader(draw, st->vp_variant->draw_shader);
set_feedback_vertex_format(ctx);
/* Find the lowest address of the arrays we're drawing */
if (vp->num_inputs) {
low_addr = arrays[vp->index_to_input[0]]->Ptr;
for (attr = 1; attr < vp->num_inputs; attr++) {
const GLubyte *start = arrays[vp->index_to_input[attr]]->Ptr;
low_addr = MIN2(low_addr, start);
}
}
/* loop over TGSI shader inputs to determine vertex buffer
* and attribute info
*/
for (attr = 0; attr < vp->num_inputs; attr++) {
const GLuint mesaAttr = vp->index_to_input[attr];
struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj;
void *map;
if (bufobj && bufobj->Name) {
/* Attribute data is in a VBO.
* Recall that for VBOs, the gl_client_array->Ptr field is
* really an offset from the start of the VBO, not a pointer.
*/
struct st_buffer_object *stobj = st_buffer_object(bufobj);
assert(stobj->buffer);
vbuffers[attr].buffer = NULL;
pipe_resource_reference(&vbuffers[attr].buffer, stobj->buffer);
vbuffers[attr].buffer_offset = pointer_to_offset(low_addr);
velements[attr].src_offset = arrays[mesaAttr]->Ptr - low_addr;
}
else {
/* attribute data is in user-space memory, not a VBO */
uint bytes = (arrays[mesaAttr]->Size
* _mesa_sizeof_type(arrays[mesaAttr]->Type)
* (max_index + 1));
/* wrap user data */
vbuffers[attr].buffer
= pipe_user_buffer_create(pipe->screen, (void *) arrays[mesaAttr]->Ptr,
bytes,
PIPE_BIND_VERTEX_BUFFER);
vbuffers[attr].buffer_offset = 0;
velements[attr].src_offset = 0;
}
/* common-case setup */
vbuffers[attr].stride = arrays[mesaAttr]->StrideB; /* in bytes */
velements[attr].instance_divisor = 0;
velements[attr].vertex_buffer_index = attr;
velements[attr].src_format =
st_pipe_vertex_format(arrays[mesaAttr]->Type,
arrays[mesaAttr]->Size,
arrays[mesaAttr]->Format,
arrays[mesaAttr]->Normalized,
arrays[mesaAttr]->Integer);
assert(velements[attr].src_format);
/* tell draw about this attribute */
#if 0
draw_set_vertex_buffer(draw, attr, &vbuffer[attr]);
#endif
/* map the attrib buffer */
map = pipe_buffer_map(pipe, vbuffers[attr].buffer,
PIPE_TRANSFER_READ,
&vb_transfer[attr]);
draw_set_mapped_vertex_buffer(draw, attr, map);
}
draw_set_vertex_buffers(draw, vp->num_inputs, vbuffers);
draw_set_vertex_elements(draw, vp->num_inputs, velements);
memset(&ibuffer, 0, sizeof(ibuffer));
if (ib) {
struct gl_buffer_object *bufobj = ib->obj;
ibuffer.index_size = vbo_sizeof_ib_type(ib->type);
if (ibuffer.index_size == 0)
goto out_unref_vertex;
if (bufobj && bufobj->Name) {
struct st_buffer_object *stobj = st_buffer_object(bufobj);
pipe_resource_reference(&ibuffer.buffer, stobj->buffer);
ibuffer.offset = pointer_to_offset(ib->ptr);
mapped_indices = pipe_buffer_map(pipe, stobj->buffer,
PIPE_TRANSFER_READ, &ib_transfer);
}
else {
/* skip setting ibuffer.buffer as the draw module does not use it */
mapped_indices = ib->ptr;
}
draw_set_index_buffer(draw, &ibuffer);
draw_set_mapped_index_buffer(draw, mapped_indices);
}
/* set the constant buffer */
draw_set_mapped_constant_buffer(st->draw, PIPE_SHADER_VERTEX, 0,
st->state.constants[PIPE_SHADER_VERTEX].ptr,
st->state.constants[PIPE_SHADER_VERTEX].size);
/* draw here */
for (i = 0; i < nr_prims; i++) {
draw_arrays(draw, prims[i].mode, prims[i].start, prims[i].count);
}
/*
* unmap vertex/index buffers
*/
if (ib) {
draw_set_mapped_index_buffer(draw, NULL);
draw_set_index_buffer(draw, NULL);
if (ib_transfer)
pipe_buffer_unmap(pipe, ib_transfer);
pipe_resource_reference(&ibuffer.buffer, NULL);
}
out_unref_vertex:
for (attr = 0; attr < vp->num_inputs; attr++) {
pipe_buffer_unmap(pipe, vb_transfer[attr]);
draw_set_mapped_vertex_buffer(draw, attr, NULL);
pipe_resource_reference(&vbuffers[attr].buffer, NULL);
}
draw_set_vertex_buffers(draw, 0, NULL);
}
int main(int argc, char **argv)
{
struct fbdemos_scaffold *fbs = 0;
fbdemo_init(&fbs);
int width = fbs->width;
int height = fbs->height;
struct pipe_context *pipe = fbs->pipe;
/* resources */
struct pipe_resource *rt_resource = fbdemo_create_2d(fbs->screen, PIPE_BIND_RENDER_TARGET, PIPE_FORMAT_B8G8R8X8_UNORM, width, height, 0);
struct pipe_resource *z_resource = fbdemo_create_2d(fbs->screen, PIPE_BIND_RENDER_TARGET, PIPE_FORMAT_S8_UINT_Z24_UNORM, width, height, 0);
struct pipe_resource *vtx_resource = pipe_buffer_create(fbs->screen, PIPE_BIND_VERTEX_BUFFER, PIPE_USAGE_IMMUTABLE, VERTEX_BUFFER_SIZE);
struct pipe_resource *idx_resource = pipe_buffer_create(fbs->screen, PIPE_BIND_INDEX_BUFFER, PIPE_USAGE_IMMUTABLE, VERTEX_BUFFER_SIZE);
/* bind render target to framebuffer */
etna_fb_bind_resource(fbs, rt_resource);
/* vertex / index buffer setup */
struct pipe_transfer *vtx_transfer = 0;
float *vtx_logical = pipe_buffer_map(pipe, vtx_resource, PIPE_TRANSFER_WRITE | PIPE_TRANSFER_UNSYNCHRONIZED, &vtx_transfer);
assert(vtx_logical);
for(int vert=0; vert<NUM_VERTICES; ++vert)
{
int dest_idx = vert * 3;
for(int comp=0; comp<3; ++comp)
vtx_logical[dest_idx+comp+0] = vVertices[vert*3 + comp]; /* 0 */
}
pipe_buffer_unmap(pipe, vtx_transfer);
struct pipe_transfer *idx_transfer = 0;
void *idx_logical = pipe_buffer_map(pipe, idx_resource, PIPE_TRANSFER_WRITE | PIPE_TRANSFER_UNSYNCHRONIZED, &idx_transfer);
assert(idx_logical);
memcpy(idx_logical, indices, sizeof(indices));
pipe_buffer_unmap(pipe, idx_transfer);
struct pipe_vertex_buffer vertex_buf_desc = {
.stride = (3)*4,
.buffer_offset = 0,
.buffer = vtx_resource,
.user_buffer = 0
};
struct pipe_vertex_element pipe_vertex_elements[] = {
{ /* positions */
.src_offset = 0,
.instance_divisor = 0,
.vertex_buffer_index = 0,
.src_format = PIPE_FORMAT_R32G32B32_FLOAT
},
};
void *vertex_elements = pipe->create_vertex_elements_state(pipe,
sizeof(pipe_vertex_elements)/sizeof(pipe_vertex_elements[0]), pipe_vertex_elements);
struct pipe_index_buffer index_buf_desc = {
.index_size = 1,
.offset = 0,
.buffer = idx_resource,
.user_buffer = 0
};
/* compile gallium3d states */
void *blend = pipe->create_blend_state(pipe, &(struct pipe_blend_state) {
.rt[0] = {
.blend_enable = 0,
.rgb_func = PIPE_BLEND_ADD,
.rgb_src_factor = PIPE_BLENDFACTOR_SRC_ALPHA,
.rgb_dst_factor = PIPE_BLENDFACTOR_INV_SRC_ALPHA,
.alpha_func = PIPE_BLEND_ADD,
.alpha_src_factor = PIPE_BLENDFACTOR_SRC_ALPHA,
.alpha_dst_factor = PIPE_BLENDFACTOR_INV_SRC_ALPHA,
.colormask = 0xf
}
});
void *sampler = pipe->create_sampler_state(pipe, &(struct pipe_sampler_state) {
.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE,
.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE,
.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE,
.min_img_filter = PIPE_TEX_FILTER_LINEAR,
.min_mip_filter = PIPE_TEX_MIPFILTER_LINEAR,
.mag_img_filter = PIPE_TEX_FILTER_LINEAR,
.normalized_coords = 1,
.lod_bias = 0.0f,
.min_lod = 0.0f, .max_lod=1000.0f
});
void
nv30_render_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info)
{
struct nv30_context *nv30 = nv30_context(pipe);
struct draw_context *draw = nv30->draw;
struct pipe_transfer *transfer[PIPE_MAX_ATTRIBS] = {NULL};
struct pipe_transfer *transferi = NULL;
int i;
nv30_render_validate(nv30);
if (nv30->draw_dirty & NV30_NEW_VIEWPORT)
draw_set_viewport_states(draw, 0, 1, &nv30->viewport);
if (nv30->draw_dirty & NV30_NEW_RASTERIZER)
draw_set_rasterizer_state(draw, &nv30->rast->pipe, NULL);
if (nv30->draw_dirty & NV30_NEW_CLIP)
draw_set_clip_state(draw, &nv30->clip);
if (nv30->draw_dirty & NV30_NEW_ARRAYS) {
draw_set_vertex_buffers(draw, 0, nv30->num_vtxbufs, nv30->vtxbuf);
draw_set_vertex_elements(draw, nv30->vertex->num_elements, nv30->vertex->pipe);
}
if (nv30->draw_dirty & NV30_NEW_FRAGPROG) {
struct nv30_fragprog *fp = nv30->fragprog.program;
if (!fp->draw)
fp->draw = draw_create_fragment_shader(draw, &fp->pipe);
draw_bind_fragment_shader(draw, fp->draw);
}
if (nv30->draw_dirty & NV30_NEW_VERTPROG) {
struct nv30_vertprog *vp = nv30->vertprog.program;
if (!vp->draw)
vp->draw = draw_create_vertex_shader(draw, &vp->pipe);
draw_bind_vertex_shader(draw, vp->draw);
}
if (nv30->draw_dirty & NV30_NEW_VERTCONST) {
if (nv30->vertprog.constbuf) {
void *map = nv04_resource(nv30->vertprog.constbuf)->data;
draw_set_mapped_constant_buffer(draw, PIPE_SHADER_VERTEX, 0,
map, nv30->vertprog.constbuf_nr * 16);
} else {
draw_set_mapped_constant_buffer(draw, PIPE_SHADER_VERTEX, 0, NULL, 0);
}
}
for (i = 0; i < nv30->num_vtxbufs; i++) {
const void *map = nv30->vtxbuf[i].user_buffer;
if (!map) {
if (nv30->vtxbuf[i].buffer)
map = pipe_buffer_map(pipe, nv30->vtxbuf[i].buffer,
PIPE_TRANSFER_UNSYNCHRONIZED |
PIPE_TRANSFER_READ, &transfer[i]);
}
draw_set_mapped_vertex_buffer(draw, i, map, ~0);
}
if (info->indexed) {
const void *map = nv30->idxbuf.user_buffer;
if (!map)
map = pipe_buffer_map(pipe, nv30->idxbuf.buffer,
PIPE_TRANSFER_UNSYNCHRONIZED |
PIPE_TRANSFER_READ, &transferi);
draw_set_indexes(draw,
(ubyte *) map + nv30->idxbuf.offset,
nv30->idxbuf.index_size, ~0);
} else {
draw_set_indexes(draw, NULL, 0, 0);
}
draw_vbo(draw, info);
draw_flush(draw);
if (info->indexed && transferi)
pipe_buffer_unmap(pipe, transferi);
for (i = 0; i < nv30->num_vtxbufs; i++)
if (transfer[i])
pipe_buffer_unmap(pipe, transfer[i]);
nv30->draw_dirty = 0;
nv30_state_release(nv30);
}
static void
st_DrawTex(struct gl_context *ctx, GLfloat x, GLfloat y, GLfloat z,
GLfloat width, GLfloat height)
{
struct st_context *st = ctx->st;
struct pipe_context *pipe = st->pipe;
struct cso_context *cso = ctx->st->cso_context;
struct pipe_resource *vbuffer;
struct pipe_transfer *vbuffer_transfer;
GLuint i, numTexCoords, numAttribs;
GLboolean emitColor;
uint semantic_names[2 + MAX_TEXTURE_UNITS];
uint semantic_indexes[2 + MAX_TEXTURE_UNITS];
struct pipe_vertex_element velements[2 + MAX_TEXTURE_UNITS];
GLbitfield inputs = VERT_BIT_POS;
st_validate_state(st);
/* determine if we need vertex color */
if (ctx->FragmentProgram._Current->Base.InputsRead & FRAG_BIT_COL0)
emitColor = GL_TRUE;
else
emitColor = GL_FALSE;
/* determine how many enabled sets of texcoords */
numTexCoords = 0;
for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
if (ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_2D_BIT) {
inputs |= VERT_BIT_TEX(i);
numTexCoords++;
}
}
/* total number of attributes per vertex */
numAttribs = 1 + emitColor + numTexCoords;
/* create the vertex buffer */
vbuffer = pipe_buffer_create(pipe->screen, PIPE_BIND_VERTEX_BUFFER,
PIPE_USAGE_STREAM,
numAttribs * 4 * 4 * sizeof(GLfloat));
/* load vertex buffer */
{
#define SET_ATTRIB(VERT, ATTR, X, Y, Z, W) \
do { \
GLuint k = (((VERT) * numAttribs + (ATTR)) * 4); \
assert(k < 4 * 4 * numAttribs); \
vbuf[k + 0] = X; \
vbuf[k + 1] = Y; \
vbuf[k + 2] = Z; \
vbuf[k + 3] = W; \
} while (0)
const GLfloat x0 = x, y0 = y, x1 = x + width, y1 = y + height;
GLfloat *vbuf = (GLfloat *) pipe_buffer_map(pipe, vbuffer,
PIPE_TRANSFER_WRITE,
&vbuffer_transfer);
GLuint attr;
z = CLAMP(z, 0.0f, 1.0f);
/* positions (in clip coords) */
{
const struct gl_framebuffer *fb = st->ctx->DrawBuffer;
const GLfloat fb_width = (GLfloat)fb->Width;
const GLfloat fb_height = (GLfloat)fb->Height;
const GLfloat clip_x0 = (GLfloat)(x0 / fb_width * 2.0 - 1.0);
const GLfloat clip_y0 = (GLfloat)(y0 / fb_height * 2.0 - 1.0);
const GLfloat clip_x1 = (GLfloat)(x1 / fb_width * 2.0 - 1.0);
const GLfloat clip_y1 = (GLfloat)(y1 / fb_height * 2.0 - 1.0);
SET_ATTRIB(0, 0, clip_x0, clip_y0, z, 1.0f); /* lower left */
SET_ATTRIB(1, 0, clip_x1, clip_y0, z, 1.0f); /* lower right */
SET_ATTRIB(2, 0, clip_x1, clip_y1, z, 1.0f); /* upper right */
SET_ATTRIB(3, 0, clip_x0, clip_y1, z, 1.0f); /* upper left */
semantic_names[0] = TGSI_SEMANTIC_POSITION;
semantic_indexes[0] = 0;
}
/* colors */
if (emitColor) {
const GLfloat *c = ctx->Current.Attrib[VERT_ATTRIB_COLOR0];
SET_ATTRIB(0, 1, c[0], c[1], c[2], c[3]);
SET_ATTRIB(1, 1, c[0], c[1], c[2], c[3]);
SET_ATTRIB(2, 1, c[0], c[1], c[2], c[3]);
SET_ATTRIB(3, 1, c[0], c[1], c[2], c[3]);
semantic_names[1] = TGSI_SEMANTIC_COLOR;
semantic_indexes[1] = 0;
attr = 2;
}
else {
attr = 1;
}
/* texcoords */
for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
if (ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_2D_BIT) {
struct gl_texture_object *obj = ctx->Texture.Unit[i]._Current;
struct gl_texture_image *img = obj->Image[0][obj->BaseLevel];
const GLfloat wt = (GLfloat) img->Width;
const GLfloat ht = (GLfloat) img->Height;
const GLfloat s0 = obj->CropRect[0] / wt;
const GLfloat t0 = obj->CropRect[1] / ht;
const GLfloat s1 = (obj->CropRect[0] + obj->CropRect[2]) / wt;
const GLfloat t1 = (obj->CropRect[1] + obj->CropRect[3]) / ht;
/*printf("crop texcoords: %g, %g .. %g, %g\n", s0, t0, s1, t1);*/
SET_ATTRIB(0, attr, s0, t0, 0.0f, 1.0f); /* lower left */
SET_ATTRIB(1, attr, s1, t0, 0.0f, 1.0f); /* lower right */
SET_ATTRIB(2, attr, s1, t1, 0.0f, 1.0f); /* upper right */
SET_ATTRIB(3, attr, s0, t1, 0.0f, 1.0f); /* upper left */
semantic_names[attr] = TGSI_SEMANTIC_GENERIC;
semantic_indexes[attr] = 0;
attr++;
}
}
pipe_buffer_unmap(pipe, vbuffer_transfer);
#undef SET_ATTRIB
}
cso_save_viewport(cso);
cso_save_vertex_shader(cso);
cso_save_vertex_elements(cso);
cso_save_vertex_buffers(cso);
{
void *vs = lookup_shader(pipe, numAttribs,
semantic_names, semantic_indexes);
cso_set_vertex_shader_handle(cso, vs);
}
for (i = 0; i < numAttribs; i++) {
velements[i].src_offset = i * 4 * sizeof(float);
velements[i].instance_divisor = 0;
velements[i].vertex_buffer_index = 0;
velements[i].src_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
}
cso_set_vertex_elements(cso, numAttribs, velements);
/* viewport state: viewport matching window dims */
{
const struct gl_framebuffer *fb = st->ctx->DrawBuffer;
const GLboolean invert = (st_fb_orientation(fb) == Y_0_TOP);
const GLfloat width = (GLfloat)fb->Width;
const GLfloat height = (GLfloat)fb->Height;
struct pipe_viewport_state vp;
vp.scale[0] = 0.5f * width;
vp.scale[1] = height * (invert ? -0.5f : 0.5f);
vp.scale[2] = 1.0f;
vp.scale[3] = 1.0f;
vp.translate[0] = 0.5f * width;
vp.translate[1] = 0.5f * height;
vp.translate[2] = 0.0f;
vp.translate[3] = 0.0f;
cso_set_viewport(cso, &vp);
}
util_draw_vertex_buffer(pipe, cso, vbuffer,
0, /* offset */
PIPE_PRIM_TRIANGLE_FAN,
4, /* verts */
numAttribs); /* attribs/vert */
pipe_resource_reference(&vbuffer, NULL);
/* restore state */
cso_restore_viewport(cso);
cso_restore_vertex_shader(cso);
cso_restore_vertex_elements(cso);
cso_restore_vertex_buffers(cso);
}
int main(int argc, char **argv)
{
struct fbdemos_scaffold *fbs = 0;
fbdemo_init(&fbs);
int width = fbs->width;
int height = fbs->height;
struct pipe_context *pipe = fbs->pipe;
/* resources */
struct pipe_resource *rt_resource = fbdemo_create_2d(fbs->screen, PIPE_BIND_RENDER_TARGET, PIPE_FORMAT_B8G8R8X8_UNORM, width, height, 0);
struct pipe_resource *z_resource = fbdemo_create_2d(fbs->screen, PIPE_BIND_RENDER_TARGET, PIPE_FORMAT_Z16_UNORM, width, height, 0);
struct pipe_resource *vtx_resource = pipe_buffer_create(fbs->screen, PIPE_BIND_VERTEX_BUFFER, PIPE_USAGE_IMMUTABLE, VERTEX_BUFFER_SIZE);
struct pipe_resource *idx_resource = pipe_buffer_create(fbs->screen, PIPE_BIND_INDEX_BUFFER, PIPE_USAGE_IMMUTABLE, VERTEX_BUFFER_SIZE);
/* bind render target to framebuffer */
etna_fb_bind_resource(&fbs->fb, rt_resource);
/* Phew, now we got all the memory we need.
* Write interleaved attribute vertex stream.
* Unlike the GL example we only do this once, not every time glDrawArrays is called, the same would be accomplished
* from GL by using a vertex buffer object.
*/
float *vVertices;
float *vNormals;
float *vTexCoords;
uint16_t *vIndices;
int numVertices = 0;
int numIndices = esGenSphere(80, 1.0f, &vVertices, &vNormals,
&vTexCoords, &vIndices, &numVertices);
unsigned vtxStride = 3+3+2;
assert((numVertices * vtxStride*4) < VERTEX_BUFFER_SIZE);
struct pipe_transfer *vtx_transfer = 0;
float *vtx_logical = pipe_buffer_map(pipe, vtx_resource, PIPE_TRANSFER_WRITE | PIPE_TRANSFER_UNSYNCHRONIZED, &vtx_transfer);
for(int vert=0; vert<numVertices; ++vert)
{
int dest_idx = vert * vtxStride;
for(int comp=0; comp<3; ++comp)
vtx_logical[dest_idx+comp+0] = vVertices[vert*3 + comp]; /* 0 */
for(int comp=0; comp<3; ++comp)
vtx_logical[dest_idx+comp+3] = vNormals[vert*3 + comp]; /* 1 */
for(int comp=0; comp<2; ++comp)
vtx_logical[dest_idx+comp+6] = vTexCoords[vert*2 + comp]; /* 2 */
}
pipe_buffer_unmap(pipe, vtx_transfer);
assert((numIndices * 2) < VERTEX_BUFFER_SIZE);
struct pipe_transfer *idx_transfer = 0;
void *idx_logical = pipe_buffer_map(pipe, idx_resource, PIPE_TRANSFER_WRITE | PIPE_TRANSFER_UNSYNCHRONIZED, &idx_transfer);
memcpy(idx_logical, vIndices, numIndices*sizeof(uint16_t));
pipe_buffer_unmap(pipe, idx_transfer);
/* compile gallium3d states */
void *blend = pipe->create_blend_state(pipe, &(struct pipe_blend_state) {
.rt[0] = {
.blend_enable = 0,
.rgb_func = PIPE_BLEND_ADD,
.rgb_src_factor = PIPE_BLENDFACTOR_ONE,
.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO,
.alpha_func = PIPE_BLEND_ADD,
.alpha_src_factor = PIPE_BLENDFACTOR_ONE,
.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO,
.colormask = 0xf
}
});
/**
* Draw vertex arrays, with optional indexing.
* Basically, map the vertex buffers (and drawing surfaces), then hand off
* the drawing to the 'draw' module.
*
* XXX should the element buffer be specified/bound with a separate function?
*/
static boolean
cell_draw_range_elements(struct pipe_context *pipe,
struct pipe_buffer *indexBuffer,
unsigned indexSize,
unsigned min_index,
unsigned max_index,
unsigned mode, unsigned start, unsigned count)
{
struct cell_context *sp = cell_context(pipe);
struct draw_context *draw = sp->draw;
unsigned i;
if (sp->dirty)
cell_update_derived( sp );
#if 0
cell_map_surfaces(sp);
#endif
cell_map_constant_buffers(sp);
/*
* Map vertex buffers
*/
for (i = 0; i < sp->num_vertex_buffers; i++) {
void *buf = pipe_buffer_map(pipe->screen,
sp->vertex_buffer[i].buffer,
PIPE_BUFFER_USAGE_CPU_READ);
cell_flush_buffer_range(sp, buf, sp->vertex_buffer[i].buffer->size);
draw_set_mapped_vertex_buffer(draw, i, buf);
}
/* Map index buffer, if present */
if (indexBuffer) {
void *mapped_indexes = pipe_buffer_map(pipe->screen,
indexBuffer,
PIPE_BUFFER_USAGE_CPU_READ);
draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes);
}
else {
/* no index/element buffer */
draw_set_mapped_element_buffer(draw, 0, NULL);
}
/* draw! */
draw_arrays(draw, mode, start, count);
/*
* unmap vertex/index buffers - will cause draw module to flush
*/
for (i = 0; i < sp->num_vertex_buffers; i++) {
draw_set_mapped_vertex_buffer(draw, i, NULL);
pipe_buffer_unmap(pipe->screen, sp->vertex_buffer[i].buffer);
}
if (indexBuffer) {
draw_set_mapped_element_buffer(draw, 0, NULL);
pipe_buffer_unmap(pipe->screen, indexBuffer);
}
/* Note: leave drawing surfaces mapped */
cell_unmap_constant_buffers(sp);
return TRUE;
}
int main(int argc, char **argv)
{
struct fbdemos_scaffold *fbs = 0;
fbdemo_init(&fbs);
int width = fbs->width;
int height = fbs->height;
struct pipe_context *pipe = fbs->pipe;
dds_texture *dds = 0;
if(argc<2 || !dds_load(argv[1], &dds))
{
printf("Error loading texture\n");
exit(1);
}
uint32_t tex_format = 0;
uint32_t tex_base_width = dds->slices[0][0].width;
uint32_t tex_base_height = dds->slices[0][0].height;
switch(dds->fmt)
{
case FMT_A8R8G8B8: tex_format = PIPE_FORMAT_B8G8R8A8_UNORM; break;
case FMT_X8R8G8B8: tex_format = PIPE_FORMAT_B8G8R8X8_UNORM; break;
case FMT_DXT1: tex_format = PIPE_FORMAT_DXT1_RGB; break;
case FMT_DXT3: tex_format = PIPE_FORMAT_DXT3_RGBA; break;
case FMT_DXT5: tex_format = PIPE_FORMAT_DXT5_RGBA; break;
case FMT_ETC1: tex_format = PIPE_FORMAT_ETC1_RGB8; break;
case FMT_A8: tex_format = PIPE_FORMAT_A8_UNORM; break;
case FMT_L8: tex_format = PIPE_FORMAT_L8_UNORM; break;
case FMT_A8L8: tex_format = PIPE_FORMAT_L8A8_UNORM; break;
default:
printf("Unknown texture format\n");
exit(1);
}
struct pipe_resource *tex_resource = fbdemo_create_2d(fbs->screen, PIPE_BIND_SAMPLER_VIEW, tex_format, tex_base_width, tex_base_height,
dds->num_mipmaps - 1);
printf("Loading compressed texture (format %i, %ix%i)\n", dds->fmt, tex_base_width, tex_base_height);
for(int ix=0; ix<dds->num_mipmaps; ++ix)
{
printf("%08x: Uploading mipmap %i (%ix%i)\n", dds->slices[0][ix].offset, ix, dds->slices[0][ix].width, dds->slices[0][ix].height);
etna_pipe_inline_write(pipe, tex_resource, 0, ix, dds->slices[0][ix].data, dds->slices[0][ix].size);
}
/* resources */
struct pipe_resource *rt_resource = fbdemo_create_2d(fbs->screen, PIPE_BIND_RENDER_TARGET, PIPE_FORMAT_B8G8R8X8_UNORM, width, height, 0);
struct pipe_resource *z_resource = fbdemo_create_2d(fbs->screen, PIPE_BIND_RENDER_TARGET, PIPE_FORMAT_Z16_UNORM, width, height, 0);
struct pipe_resource *vtx_resource = pipe_buffer_create(fbs->screen, PIPE_BIND_VERTEX_BUFFER, PIPE_USAGE_IMMUTABLE, VERTEX_BUFFER_SIZE);
/* bind render target to framebuffer */
etna_fb_bind_resource(&fbs->fb, rt_resource);
/* Phew, now we got all the memory we need.
* Write interleaved attribute vertex stream.
* Unlike the GL example we only do this once, not every time glDrawArrays is called, the same would be accomplished
* from GL by using a vertex buffer object.
*/
struct pipe_transfer *vtx_transfer = 0;
float *vtx_logical = pipe_buffer_map(pipe, vtx_resource, PIPE_TRANSFER_WRITE | PIPE_TRANSFER_UNSYNCHRONIZED, &vtx_transfer);
assert(vtx_logical);
for(int vert=0; vert<NUM_VERTICES; ++vert)
{
int dest_idx = vert * (3 + 3 + 2);
for(int comp=0; comp<3; ++comp)
vtx_logical[dest_idx+comp+0] = vVertices[vert*3 + comp]; /* 0 */
for(int comp=0; comp<3; ++comp)
vtx_logical[dest_idx+comp+3] = vNormals[vert*3 + comp]; /* 1 */
for(int comp=0; comp<2; ++comp)
vtx_logical[dest_idx+comp+6] = vTexCoords[vert*2 + comp]; /* 2 */
}
pipe_buffer_unmap(pipe, vtx_transfer);
/* compile gallium3d states */
void *blend = NULL;
if(tex_format == PIPE_FORMAT_A8_UNORM || tex_format == PIPE_FORMAT_L8A8_UNORM) /* if alpha texture, enable blending */
{
blend = pipe->create_blend_state(pipe, &(struct pipe_blend_state) {
.rt[0] = {
.blend_enable = 1,
.rgb_func = PIPE_BLEND_ADD,
.rgb_src_factor = PIPE_BLENDFACTOR_SRC_ALPHA,
.rgb_dst_factor = PIPE_BLENDFACTOR_INV_SRC_ALPHA,
.alpha_func = PIPE_BLEND_ADD,
.alpha_src_factor = PIPE_BLENDFACTOR_SRC_ALPHA,
.alpha_dst_factor = PIPE_BLENDFACTOR_INV_SRC_ALPHA,
.colormask = 0xf
}
});
static void r300_emit_draw_arrays_immediate(struct r300_context *r300,
unsigned mode,
unsigned start,
unsigned count)
{
struct pipe_vertex_element* velem;
struct pipe_vertex_buffer* vbuf;
unsigned vertex_element_count = r300->velems->count;
unsigned i, v, vbi;
/* Size of the vertex, in dwords. */
unsigned vertex_size = r300->velems->vertex_size_dwords;
/* The number of dwords for this draw operation. */
unsigned dwords = 9 + count * vertex_size;
/* Size of the vertex element, in dwords. */
unsigned size[PIPE_MAX_ATTRIBS];
/* Stride to the same attrib in the next vertex in the vertex buffer,
* in dwords. */
unsigned stride[PIPE_MAX_ATTRIBS];
/* Mapped vertex buffers. */
uint32_t* map[PIPE_MAX_ATTRIBS];
uint32_t* mapelem[PIPE_MAX_ATTRIBS];
struct pipe_transfer* transfer[PIPE_MAX_ATTRIBS] = {0};
CS_LOCALS(r300);
if (!r300_prepare_for_rendering(r300, PREP_FIRST_DRAW, NULL, dwords, 0, 0))
return;
/* Calculate the vertex size, offsets, strides etc. and map the buffers. */
for (i = 0; i < vertex_element_count; i++) {
velem = &r300->velems->velem[i];
size[i] = r300->velems->hw_format_size[i] / 4;
vbi = velem->vertex_buffer_index;
vbuf = &r300->vertex_buffer[vbi];
stride[i] = vbuf->stride / 4;
/* Map the buffer. */
if (!transfer[vbi]) {
map[vbi] = (uint32_t*)pipe_buffer_map(&r300->context,
vbuf->buffer,
PIPE_TRANSFER_READ,
&transfer[vbi]);
map[vbi] += (vbuf->buffer_offset / 4) + stride[i] * start;
}
mapelem[i] = map[vbi] + (velem->src_offset / 4);
}
BEGIN_CS(dwords);
OUT_CS_REG(R300_GA_COLOR_CONTROL,
r300_provoking_vertex_fixes(r300, mode));
OUT_CS_REG(R300_VAP_VTX_SIZE, vertex_size);
OUT_CS_REG_SEQ(R300_VAP_VF_MAX_VTX_INDX, 2);
OUT_CS(count - 1);
OUT_CS(0);
OUT_CS_PKT3(R300_PACKET3_3D_DRAW_IMMD_2, count * vertex_size);
OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_EMBEDDED | (count << 16) |
r300_translate_primitive(mode));
/* Emit vertices. */
for (v = 0; v < count; v++) {
for (i = 0; i < vertex_element_count; i++) {
OUT_CS_TABLE(&mapelem[i][stride[i] * v], size[i]);
}
}
END_CS;
/* Unmap buffers. */
for (i = 0; i < vertex_element_count; i++) {
vbi = r300->velems->velem[i].vertex_buffer_index;
if (transfer[vbi]) {
vbuf = &r300->vertex_buffer[vbi];
pipe_buffer_unmap(&r300->context, vbuf->buffer, transfer[vbi]);
transfer[vbi] = NULL;
}
}
}
/* SW TCL elements, using Draw. */
static void r300_swtcl_draw_vbo(struct pipe_context* pipe,
const struct pipe_draw_info *info)
{
struct r300_context* r300 = r300_context(pipe);
struct pipe_transfer *vb_transfer[PIPE_MAX_ATTRIBS];
struct pipe_transfer *ib_transfer = NULL;
unsigned count = info->count;
int i;
void *indices = NULL;
boolean indexed = info->indexed && r300->index_buffer.buffer;
if (r300->skip_rendering) {
return;
}
if (!u_trim_pipe_prim(info->mode, &count)) {
return;
}
r300_update_derived_state(r300);
r300_reserve_cs_dwords(r300,
PREP_FIRST_DRAW | PREP_EMIT_AOS_SWTCL |
(indexed ? PREP_INDEXED : 0),
indexed ? 256 : 6);
for (i = 0; i < r300->vertex_buffer_count; i++) {
if (r300->vertex_buffer[i].buffer) {
void *buf = pipe_buffer_map(pipe,
r300->vertex_buffer[i].buffer,
PIPE_TRANSFER_READ,
&vb_transfer[i]);
draw_set_mapped_vertex_buffer(r300->draw, i, buf);
}
}
if (indexed) {
indices = pipe_buffer_map(pipe, r300->index_buffer.buffer,
PIPE_TRANSFER_READ, &ib_transfer);
}
draw_set_mapped_index_buffer(r300->draw, indices);
r300->draw_vbo_locked = TRUE;
r300->draw_first_emitted = FALSE;
draw_vbo(r300->draw, info);
draw_flush(r300->draw);
r300->draw_vbo_locked = FALSE;
for (i = 0; i < r300->vertex_buffer_count; i++) {
if (r300->vertex_buffer[i].buffer) {
pipe_buffer_unmap(pipe, r300->vertex_buffer[i].buffer,
vb_transfer[i]);
draw_set_mapped_vertex_buffer(r300->draw, i, NULL);
}
}
if (indexed) {
pipe_buffer_unmap(pipe, r300->index_buffer.buffer, ib_transfer);
draw_set_mapped_index_buffer(r300->draw, NULL);
}
}
struct pipe_context *
dd_context_create(struct dd_screen *dscreen, struct pipe_context *pipe)
{
struct dd_context *dctx;
if (!pipe)
return NULL;
dctx = CALLOC_STRUCT(dd_context);
if (!dctx)
goto fail;
dctx->pipe = pipe;
dctx->base.priv = pipe->priv; /* expose wrapped priv data */
dctx->base.screen = &dscreen->base;
dctx->base.stream_uploader = pipe->stream_uploader;
dctx->base.const_uploader = pipe->const_uploader;
dctx->base.destroy = dd_context_destroy;
CTX_INIT(render_condition);
CTX_INIT(create_query);
CTX_INIT(create_batch_query);
CTX_INIT(destroy_query);
CTX_INIT(begin_query);
CTX_INIT(end_query);
CTX_INIT(get_query_result);
CTX_INIT(set_active_query_state);
CTX_INIT(create_blend_state);
CTX_INIT(bind_blend_state);
CTX_INIT(delete_blend_state);
CTX_INIT(create_sampler_state);
CTX_INIT(bind_sampler_states);
CTX_INIT(delete_sampler_state);
CTX_INIT(create_rasterizer_state);
CTX_INIT(bind_rasterizer_state);
CTX_INIT(delete_rasterizer_state);
CTX_INIT(create_depth_stencil_alpha_state);
CTX_INIT(bind_depth_stencil_alpha_state);
CTX_INIT(delete_depth_stencil_alpha_state);
CTX_INIT(create_fs_state);
CTX_INIT(bind_fs_state);
CTX_INIT(delete_fs_state);
CTX_INIT(create_vs_state);
CTX_INIT(bind_vs_state);
CTX_INIT(delete_vs_state);
CTX_INIT(create_gs_state);
CTX_INIT(bind_gs_state);
CTX_INIT(delete_gs_state);
CTX_INIT(create_tcs_state);
CTX_INIT(bind_tcs_state);
CTX_INIT(delete_tcs_state);
CTX_INIT(create_tes_state);
CTX_INIT(bind_tes_state);
CTX_INIT(delete_tes_state);
CTX_INIT(create_compute_state);
CTX_INIT(bind_compute_state);
CTX_INIT(delete_compute_state);
CTX_INIT(create_vertex_elements_state);
CTX_INIT(bind_vertex_elements_state);
CTX_INIT(delete_vertex_elements_state);
CTX_INIT(set_blend_color);
CTX_INIT(set_stencil_ref);
CTX_INIT(set_sample_mask);
CTX_INIT(set_min_samples);
CTX_INIT(set_clip_state);
CTX_INIT(set_constant_buffer);
CTX_INIT(set_framebuffer_state);
CTX_INIT(set_polygon_stipple);
CTX_INIT(set_scissor_states);
CTX_INIT(set_viewport_states);
CTX_INIT(set_sampler_views);
CTX_INIT(set_tess_state);
CTX_INIT(set_shader_buffers);
CTX_INIT(set_shader_images);
CTX_INIT(set_vertex_buffers);
CTX_INIT(create_stream_output_target);
CTX_INIT(stream_output_target_destroy);
CTX_INIT(set_stream_output_targets);
CTX_INIT(create_sampler_view);
CTX_INIT(sampler_view_destroy);
CTX_INIT(create_surface);
CTX_INIT(surface_destroy);
CTX_INIT(transfer_map);
CTX_INIT(transfer_flush_region);
CTX_INIT(transfer_unmap);
CTX_INIT(buffer_subdata);
CTX_INIT(texture_subdata);
CTX_INIT(texture_barrier);
CTX_INIT(memory_barrier);
CTX_INIT(resource_commit);
/* create_video_codec */
/* create_video_buffer */
/* set_compute_resources */
/* set_global_binding */
CTX_INIT(get_sample_position);
CTX_INIT(invalidate_resource);
CTX_INIT(get_device_reset_status);
CTX_INIT(set_device_reset_callback);
CTX_INIT(dump_debug_state);
CTX_INIT(emit_string_marker);
CTX_INIT(create_texture_handle);
CTX_INIT(delete_texture_handle);
CTX_INIT(make_texture_handle_resident);
CTX_INIT(create_image_handle);
CTX_INIT(delete_image_handle);
CTX_INIT(make_image_handle_resident);
dd_init_draw_functions(dctx);
u_log_context_init(&dctx->log);
if (pipe->set_log_context)
pipe->set_log_context(pipe, &dctx->log);
dctx->draw_state.sample_mask = ~0;
if (dscreen->mode == DD_DETECT_HANGS_PIPELINED) {
dctx->fence = pipe_buffer_create(dscreen->screen, PIPE_BIND_CUSTOM,
PIPE_USAGE_STAGING, 4);
if (!dctx->fence)
goto fail;
dctx->mapped_fence = pipe_buffer_map(pipe, dctx->fence,
PIPE_TRANSFER_READ_WRITE |
PIPE_TRANSFER_PERSISTENT |
PIPE_TRANSFER_COHERENT,
&dctx->fence_transfer);
if (!dctx->mapped_fence)
goto fail;
*dctx->mapped_fence = 0;
(void) mtx_init(&dctx->mutex, mtx_plain);
dctx->thread = u_thread_create(dd_thread_pipelined_hang_detect, dctx);
if (!dctx->thread) {
mtx_destroy(&dctx->mutex);
goto fail;
}
}
return &dctx->base;
fail:
if (dctx) {
if (dctx->mapped_fence)
pipe_transfer_unmap(pipe, dctx->fence_transfer);
pipe_resource_reference(&dctx->fence, NULL);
FREE(dctx);
}
pipe->destroy(pipe);
return NULL;
}
void vl_compositor_render(struct vl_compositor *compositor,
/*struct pipe_texture *backround,
struct pipe_video_rect *backround_area,*/
struct pipe_texture *src_surface,
enum pipe_mpeg12_picture_type picture_type,
/*unsigned num_past_surfaces,
struct pipe_texture *past_surfaces,
unsigned num_future_surfaces,
struct pipe_texture *future_surfaces,*/
struct pipe_video_rect *src_area,
struct pipe_texture *dst_surface,
struct pipe_video_rect *dst_area,
/*unsigned num_layers,
struct pipe_texture *layers,
struct pipe_video_rect *layer_src_areas,
struct pipe_video_rect *layer_dst_areas*/
struct pipe_fence_handle **fence)
{
struct vertex_shader_consts *vs_consts;
assert(compositor);
assert(src_surface);
assert(src_area);
assert(dst_surface);
assert(dst_area);
assert(picture_type == PIPE_MPEG12_PICTURE_TYPE_FRAME);
compositor->fb_state.width = dst_surface->width[0];
compositor->fb_state.height = dst_surface->height[0];
compositor->fb_state.cbufs[0] = compositor->pipe->screen->get_tex_surface
(
compositor->pipe->screen,
dst_surface,
0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE
);
compositor->viewport.scale[0] = compositor->fb_state.width;
compositor->viewport.scale[1] = compositor->fb_state.height;
compositor->viewport.scale[2] = 1;
compositor->viewport.scale[3] = 1;
compositor->viewport.translate[0] = 0;
compositor->viewport.translate[1] = 0;
compositor->viewport.translate[2] = 0;
compositor->viewport.translate[3] = 0;
compositor->scissor.maxx = compositor->fb_state.width;
compositor->scissor.maxy = compositor->fb_state.height;
compositor->pipe->set_framebuffer_state(compositor->pipe, &compositor->fb_state);
compositor->pipe->set_viewport_state(compositor->pipe, &compositor->viewport);
compositor->pipe->set_scissor_state(compositor->pipe, &compositor->scissor);
compositor->pipe->bind_sampler_states(compositor->pipe, 1, &compositor->sampler);
compositor->pipe->set_sampler_textures(compositor->pipe, 1, &src_surface);
compositor->pipe->bind_vs_state(compositor->pipe, compositor->vertex_shader);
compositor->pipe->bind_fs_state(compositor->pipe, compositor->fragment_shader);
compositor->pipe->set_vertex_buffers(compositor->pipe, 2, compositor->vertex_bufs);
compositor->pipe->set_vertex_elements(compositor->pipe, 2, compositor->vertex_elems);
compositor->pipe->set_constant_buffer(compositor->pipe, PIPE_SHADER_VERTEX, 0, &compositor->vs_const_buf);
compositor->pipe->set_constant_buffer(compositor->pipe, PIPE_SHADER_FRAGMENT, 0, &compositor->fs_const_buf);
vs_consts = pipe_buffer_map
(
compositor->pipe->screen,
compositor->vs_const_buf.buffer,
PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_DISCARD
);
vs_consts->dst_scale.x = dst_area->w / (float)compositor->fb_state.cbufs[0]->width;
vs_consts->dst_scale.y = dst_area->h / (float)compositor->fb_state.cbufs[0]->height;
vs_consts->dst_scale.z = 1;
vs_consts->dst_scale.w = 1;
vs_consts->dst_trans.x = dst_area->x / (float)compositor->fb_state.cbufs[0]->width;
vs_consts->dst_trans.y = dst_area->y / (float)compositor->fb_state.cbufs[0]->height;
vs_consts->dst_trans.z = 0;
vs_consts->dst_trans.w = 0;
vs_consts->src_scale.x = src_area->w / (float)src_surface->width[0];
vs_consts->src_scale.y = src_area->h / (float)src_surface->height[0];
vs_consts->src_scale.z = 1;
vs_consts->src_scale.w = 1;
vs_consts->src_trans.x = src_area->x / (float)src_surface->width[0];
vs_consts->src_trans.y = src_area->y / (float)src_surface->height[0];
vs_consts->src_trans.z = 0;
vs_consts->src_trans.w = 0;
pipe_buffer_unmap(compositor->pipe->screen, compositor->vs_const_buf.buffer);
compositor->pipe->draw_arrays(compositor->pipe, PIPE_PRIM_TRIANGLE_STRIP, 0, 4);
compositor->pipe->flush(compositor->pipe, PIPE_FLUSH_RENDER_CACHE, fence);
pipe_surface_reference(&compositor->fb_state.cbufs[0], NULL);
}
static bool
init_buffers(struct vl_compositor *c)
{
struct fragment_shader_consts fsc;
assert(c);
/*
* Create our vertex buffer and vertex buffer element
* VB contains 4 vertices that render a quad covering the entire window
* to display a rendered surface
* Quad is rendered as a tri strip
*/
c->vertex_bufs[0].stride = sizeof(struct vertex2f);
c->vertex_bufs[0].max_index = 3;
c->vertex_bufs[0].buffer_offset = 0;
c->vertex_bufs[0].buffer = pipe_buffer_create
(
c->pipe->screen,
1,
PIPE_BUFFER_USAGE_VERTEX,
sizeof(struct vertex2f) * 4
);
memcpy
(
pipe_buffer_map(c->pipe->screen, c->vertex_bufs[0].buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
surface_verts,
sizeof(struct vertex2f) * 4
);
pipe_buffer_unmap(c->pipe->screen, c->vertex_bufs[0].buffer);
c->vertex_elems[0].src_offset = 0;
c->vertex_elems[0].vertex_buffer_index = 0;
c->vertex_elems[0].nr_components = 2;
c->vertex_elems[0].src_format = PIPE_FORMAT_R32G32_FLOAT;
/*
* Create our texcoord buffer and texcoord buffer element
* Texcoord buffer contains the TCs for mapping the rendered surface to the 4 vertices
*/
c->vertex_bufs[1].stride = sizeof(struct vertex2f);
c->vertex_bufs[1].max_index = 3;
c->vertex_bufs[1].buffer_offset = 0;
c->vertex_bufs[1].buffer = pipe_buffer_create
(
c->pipe->screen,
1,
PIPE_BUFFER_USAGE_VERTEX,
sizeof(struct vertex2f) * 4
);
memcpy
(
pipe_buffer_map(c->pipe->screen, c->vertex_bufs[1].buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
surface_texcoords,
sizeof(struct vertex2f) * 4
);
pipe_buffer_unmap(c->pipe->screen, c->vertex_bufs[1].buffer);
c->vertex_elems[1].src_offset = 0;
c->vertex_elems[1].vertex_buffer_index = 1;
c->vertex_elems[1].nr_components = 2;
c->vertex_elems[1].src_format = PIPE_FORMAT_R32G32_FLOAT;
/*
* Create our vertex shader's constant buffer
* Const buffer contains scaling and translation vectors
*/
c->vs_const_buf.buffer = pipe_buffer_create
(
c->pipe->screen,
1,
PIPE_BUFFER_USAGE_CONSTANT | PIPE_BUFFER_USAGE_DISCARD,
sizeof(struct vertex_shader_consts)
);
/*
* Create our fragment shader's constant buffer
* Const buffer contains the color conversion matrix and bias vectors
*/
c->fs_const_buf.buffer = pipe_buffer_create
(
c->pipe->screen,
1,
PIPE_BUFFER_USAGE_CONSTANT,
sizeof(struct fragment_shader_consts)
);
vl_csc_get_matrix(VL_CSC_COLOR_STANDARD_IDENTITY, NULL, true, fsc.matrix);
vl_compositor_set_csc_matrix(c, fsc.matrix);
return true;
}
pipe->set_viewport_state(pipe, &(struct pipe_viewport_state){
.scale = {width/2.0f, height/2.0f, 0.5f, 1.0f},
.translate = {width/2.0f, height/2.0f, 0.5f, 1.0f}
});
pipe->set_fragment_sampler_views(pipe, 1, &sampler_view);
pipe->set_vertex_buffers(pipe, 0, 1, &vertex_buffer_desc);
pipe->set_index_buffer(pipe, NULL);
void *vtx_shader = graw_parse_vertex_shader(pipe, particle_system_vert);
void *frag_shader = graw_parse_fragment_shader(pipe, particle_system_frag);
pipe->bind_vs_state(pipe, vtx_shader);
pipe->bind_fs_state(pipe, frag_shader);
/* Fill in particle data array */
struct pipe_transfer *vtx_transfer = 0;
float *vtx_logical = pipe_buffer_map(pipe, vtx_resource, PIPE_TRANSFER_WRITE | PIPE_TRANSFER_UNSYNCHRONIZED, &vtx_transfer);
srand(0);
for(int i = 0; i < NUM_PARTICLES; i++)
{
float *particleData = &vtx_logical[i * PARTICLE_SIZE];
// Lifetime of particle
(*particleData++) = ( (float)(rand() % 10000) / 10000.0f );
// End position of particle
(*particleData++) = ( (float)(rand() % 10000) / 5000.0f ) - 1.0f;
(*particleData++) = ( (float)(rand() % 10000) / 5000.0f ) - 1.0f;
(*particleData++) = ( (float)(rand() % 10000) / 5000.0f ) - 1.0f;
// Start position of particle
(*particleData++) = ( (float)(rand() % 10000) / 40000.0f ) - 0.125f;
int main(int argc, char **argv)
{
struct fbdemos_scaffold *fbs = 0;
fbdemo_init(&fbs);
int width = fbs->width;
int height = fbs->height;
struct pipe_context *pipe = fbs->pipe;
/* Convert and upload embedded texture */
struct pipe_resource *tex_resource = fbdemo_create_2d(fbs->screen, PIPE_BIND_SAMPLER_VIEW, PIPE_FORMAT_B8G8R8X8_UNORM,
COMPANION_TEXTURE_WIDTH, COMPANION_TEXTURE_HEIGHT, 0);
void *temp = malloc(COMPANION_TEXTURE_WIDTH * COMPANION_TEXTURE_HEIGHT * 4);
etna_convert_r8g8b8_to_b8g8r8x8(temp, (const uint8_t*)companion_texture, COMPANION_TEXTURE_WIDTH * COMPANION_TEXTURE_HEIGHT);
etna_pipe_inline_write(pipe, tex_resource, 0, 0, temp, COMPANION_TEXTURE_WIDTH * COMPANION_TEXTURE_HEIGHT * 4);
free(temp);
/* resources */
struct pipe_resource *rt_resource = fbdemo_create_2d(fbs->screen, PIPE_BIND_RENDER_TARGET, PIPE_FORMAT_B8G8R8X8_UNORM, width, height, 0);
struct pipe_resource *z_resource = fbdemo_create_2d(fbs->screen, PIPE_BIND_RENDER_TARGET, PIPE_FORMAT_Z16_UNORM, width, height, 0);
/* bind render target to framebuffer */
etna_fb_bind_resource(&fbs->fb, rt_resource);
/* geometry */
struct pipe_resource *vtx_resource = pipe_buffer_create(fbs->screen, PIPE_BIND_VERTEX_BUFFER, PIPE_USAGE_IMMUTABLE, VERTEX_BUFFER_SIZE);
struct pipe_resource *idx_resource = pipe_buffer_create(fbs->screen, PIPE_BIND_INDEX_BUFFER, PIPE_USAGE_IMMUTABLE, INDEX_BUFFER_SIZE);
struct pipe_transfer *vtx_transfer = 0;
float *vtx_logical = pipe_buffer_map(pipe, vtx_resource, PIPE_TRANSFER_WRITE | PIPE_TRANSFER_UNSYNCHRONIZED, &vtx_transfer);
assert(vtx_logical);
struct pipe_transfer *idx_transfer = 0;
float *idx_logical = pipe_buffer_map(pipe, idx_resource, PIPE_TRANSFER_WRITE | PIPE_TRANSFER_UNSYNCHRONIZED, &idx_transfer);
assert(idx_logical);
#ifndef INDEXED
printf("Interleaving vertices...\n");
float *vertices_array = companion_vertices_array();
float *texture_coordinates_array =
companion_texture_coordinates_array();
float *normals_array = companion_normals_array();
assert(COMPANION_ARRAY_COUNT*(3+3+2)*sizeof(float) < VERTEX_BUFFER_SIZE);
for(int vert=0; vert<COMPANION_ARRAY_COUNT; ++vert)
{
int dest_idx = vert * (3 + 3 + 2);
for(int comp=0; comp<3; ++comp)
((float*)vtx_logical)[dest_idx+comp+0] = vertices_array[vert*3 + comp]; /* 0 */
for(int comp=0; comp<3; ++comp)
((float*)vtx_logical)[dest_idx+comp+3] = normals_array[vert*3 + comp]; /* 1 */
for(int comp=0; comp<2; ++comp)
((float*)vtx_logical)[dest_idx+comp+6] = texture_coordinates_array[vert*2 + comp]; /* 2 */
}
#else
printf("Interleaving vertices and copying index buffer...\n");
assert(COMPANION_VERTEX_COUNT*(3+3+2)*sizeof(float) < VERTEX_BUFFER_SIZE);
for(int vert=0; vert<COMPANION_VERTEX_COUNT; ++vert)
{
int dest_idx = vert * (3 + 3 + 2);
for(int comp=0; comp<3; ++comp)
((float*)vtx_logical)[dest_idx+comp+0] = companion_vertices[vert][comp]; /* 0 */
for(int comp=0; comp<3; ++comp)
((float*)vtx_logical)[dest_idx+comp+3] = companion_normals[vert][comp]; /* 1 */
for(int comp=0; comp<2; ++comp)
((float*)vtx_logical)[dest_idx+comp+6] = companion_texture_coordinates[vert][comp]; /* 2 */
}
assert(COMPANION_TRIANGLE_COUNT*3*sizeof(unsigned short) < INDEX_BUFFER_SIZE);
memcpy(idx_logical, &companion_triangles[0][0], COMPANION_TRIANGLE_COUNT*3*sizeof(unsigned short));
#endif
pipe_buffer_unmap(pipe, vtx_transfer);
pipe_buffer_unmap(pipe, idx_transfer);
struct pipe_vertex_buffer vertex_buffer_desc = {
.stride = (3 + 3 + 2)*4,
.buffer_offset = 0,
.buffer = vtx_resource,
.user_buffer = 0
};
struct pipe_index_buffer index_buffer_desc = {
.index_size = sizeof(unsigned short),
.offset = 0,
.buffer = idx_resource,
.user_buffer = 0
};
struct pipe_vertex_element pipe_vertex_elements[] = {
{ /* positions */
.src_offset = 0,
.instance_divisor = 0,
.vertex_buffer_index = 0,
.src_format = PIPE_FORMAT_R32G32B32_FLOAT
},
{ /* normals */
.src_offset = 0xc,
.instance_divisor = 0,
.vertex_buffer_index = 0,
.src_format = PIPE_FORMAT_R32G32B32_FLOAT
},
{ /* texture coord */
.src_offset = 0x18,
.instance_divisor = 0,
.vertex_buffer_index = 0,
.src_format = PIPE_FORMAT_R32G32_FLOAT
}
};
/**
* Called by VBO to draw arrays when in selection or feedback mode and
* to implement glRasterPos.
* This is very much like the normal draw_vbo() function above.
* Look at code refactoring some day.
* Might move this into the failover module some day.
*/
void
st_feedback_draw_vbo(GLcontext *ctx,
const struct gl_client_array **arrays,
const struct _mesa_prim *prims,
GLuint nr_prims,
const struct _mesa_index_buffer *ib,
GLboolean index_bounds_valid,
GLuint min_index,
GLuint max_index)
{
struct st_context *st = ctx->st;
struct pipe_context *pipe = st->pipe;
struct draw_context *draw = st->draw;
const struct st_vertex_program *vp;
const struct pipe_shader_state *vs;
struct pipe_buffer *index_buffer_handle = 0;
struct pipe_vertex_buffer vbuffers[PIPE_MAX_SHADER_INPUTS];
struct pipe_vertex_element velements[PIPE_MAX_ATTRIBS];
GLuint attr, i;
ubyte *mapped_constants;
assert(draw);
st_validate_state(ctx->st);
if (!index_bounds_valid)
vbo_get_minmax_index(ctx, prims, ib, &min_index, &max_index);
/* must get these after state validation! */
vp = ctx->st->vp;
vs = &st->vp->state;
if (!st->vp->draw_shader) {
st->vp->draw_shader = draw_create_vertex_shader(draw, vs);
}
/*
* Set up the draw module's state.
*
* We'd like to do this less frequently, but the normal state-update
* code sends state updates to the pipe, not to our private draw module.
*/
assert(draw);
draw_set_viewport_state(draw, &st->state.viewport);
draw_set_clip_state(draw, &st->state.clip);
draw_set_rasterizer_state(draw, &st->state.rasterizer);
draw_bind_vertex_shader(draw, st->vp->draw_shader);
set_feedback_vertex_format(ctx);
/* loop over TGSI shader inputs to determine vertex buffer
* and attribute info
*/
for (attr = 0; attr < vp->num_inputs; attr++) {
const GLuint mesaAttr = vp->index_to_input[attr];
struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj;
void *map;
if (bufobj && bufobj->Name) {
/* Attribute data is in a VBO.
* Recall that for VBOs, the gl_client_array->Ptr field is
* really an offset from the start of the VBO, not a pointer.
*/
struct st_buffer_object *stobj = st_buffer_object(bufobj);
assert(stobj->buffer);
vbuffers[attr].buffer = NULL;
pipe_buffer_reference(&vbuffers[attr].buffer, stobj->buffer);
vbuffers[attr].buffer_offset = pointer_to_offset(arrays[0]->Ptr);
velements[attr].src_offset = arrays[mesaAttr]->Ptr - arrays[0]->Ptr;
}
else {
/* attribute data is in user-space memory, not a VBO */
uint bytes = (arrays[mesaAttr]->Size
* _mesa_sizeof_type(arrays[mesaAttr]->Type)
* (max_index + 1));
/* wrap user data */
vbuffers[attr].buffer
= pipe_user_buffer_create(pipe->screen, (void *) arrays[mesaAttr]->Ptr,
bytes);
vbuffers[attr].buffer_offset = 0;
velements[attr].src_offset = 0;
}
/* common-case setup */
vbuffers[attr].stride = arrays[mesaAttr]->StrideB; /* in bytes */
vbuffers[attr].max_index = max_index;
velements[attr].vertex_buffer_index = attr;
velements[attr].nr_components = arrays[mesaAttr]->Size;
velements[attr].src_format =
st_pipe_vertex_format(arrays[mesaAttr]->Type,
arrays[mesaAttr]->Size,
arrays[mesaAttr]->Format,
arrays[mesaAttr]->Normalized);
assert(velements[attr].src_format);
/* tell draw about this attribute */
#if 0
draw_set_vertex_buffer(draw, attr, &vbuffer[attr]);
#endif
/* map the attrib buffer */
map = pipe_buffer_map(pipe->screen, vbuffers[attr].buffer,
PIPE_BUFFER_USAGE_CPU_READ);
draw_set_mapped_vertex_buffer(draw, attr, map);
}
draw_set_vertex_buffers(draw, vp->num_inputs, vbuffers);
draw_set_vertex_elements(draw, vp->num_inputs, velements);
if (ib) {
struct gl_buffer_object *bufobj = ib->obj;
unsigned indexSize;
void *map;
switch (ib->type) {
case GL_UNSIGNED_INT:
indexSize = 4;
break;
case GL_UNSIGNED_SHORT:
indexSize = 2;
break;
default:
assert(0);
return;
}
if (bufobj && bufobj->Name) {
struct st_buffer_object *stobj = st_buffer_object(bufobj);
index_buffer_handle = stobj->buffer;
map = pipe_buffer_map(pipe->screen, index_buffer_handle,
PIPE_BUFFER_USAGE_CPU_READ);
draw_set_mapped_element_buffer(draw, indexSize, map);
}
else {
draw_set_mapped_element_buffer(draw, indexSize, (void *) ib->ptr);
}
}
else {
/* no index/element buffer */
draw_set_mapped_element_buffer(draw, 0, NULL);
}
/* map constant buffers */
mapped_constants = pipe_buffer_map(pipe->screen,
st->state.constants[PIPE_SHADER_VERTEX].buffer,
PIPE_BUFFER_USAGE_CPU_READ);
draw_set_mapped_constant_buffer(st->draw, mapped_constants,
st->state.constants[PIPE_SHADER_VERTEX].buffer->size);
/* draw here */
for (i = 0; i < nr_prims; i++) {
draw_arrays(draw, prims[i].mode, prims[i].start, prims[i].count);
}
/* unmap constant buffers */
pipe_buffer_unmap(pipe->screen, st->state.constants[PIPE_SHADER_VERTEX].buffer);
/*
* unmap vertex/index buffers
*/
for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
if (draw->pt.vertex_buffer[i].buffer) {
pipe_buffer_unmap(pipe->screen, draw->pt.vertex_buffer[i].buffer);
pipe_buffer_reference(&draw->pt.vertex_buffer[i].buffer, NULL);
draw_set_mapped_vertex_buffer(draw, i, NULL);
}
}
if (index_buffer_handle) {
pipe_buffer_unmap(pipe->screen, index_buffer_handle);
draw_set_mapped_element_buffer(draw, 0, NULL);
}
}
int main(int argc, char **argv)
{
int width = 1920;
int height = 1080;
bool do_clear = false;
bool super_tiled = false;
bool enable_ts = true;
bool early_z = false;
int num_frames = 2000;
unsigned fmt_rt = PIPE_FORMAT_B8G8R8X8_UNORM;
unsigned fmt_zs = PIPE_FORMAT_S8_UINT_Z24_UNORM;
int opt;
int error = 0;
while ((opt = getopt(argc, argv, "w:h:l:s:t:e:f:d:c:")) != -1) {
switch(opt)
{
case 'w': width = atoi(optarg); break;
case 'h': height = atoi(optarg); break;
case 'l': do_clear = atoi(optarg); break;
case 's': super_tiled = atoi(optarg); break;
case 't': enable_ts = atoi(optarg); break;
case 'e': early_z = atoi(optarg); break;
case 'f': num_frames = atoi(optarg); break;
case 'd':
switch(atoi(optarg))
{
case 0: fmt_zs = PIPE_FORMAT_NONE; break;
case 16: fmt_zs = PIPE_FORMAT_Z16_UNORM; break;
case 32: fmt_zs = PIPE_FORMAT_S8_UINT_Z24_UNORM; break;
default:
printf("Invalid depth stencil surface depth %s\n", optarg);
error = 1;
}
break;
case 'c':
switch(atoi(optarg))
{
case 0: fmt_rt = PIPE_FORMAT_NONE; break;
case 16: fmt_rt = PIPE_FORMAT_B5G6R5_UNORM; break;
case 32: fmt_rt = PIPE_FORMAT_B8G8R8X8_UNORM; break;
default:
printf("Invalid color surface depth %s\n", optarg);
error = 1;
}
break;
default:
printf("Unknown argument %c\n", opt);
error = 1;
}
}
if(error)
{
printf("Usage:\n");
printf(" %s [-w <width>] [-h <height>] [-l <0/1>] [-s <0/1>] [-t <0/1>] [-e <0/1>] [-f <frames>] [-d <0/16/32>] [-c <16/32>]\n", argv[0]);
printf("\n");
printf(" -w <width> Width of surface (default is 1920)\n");
printf(" -h <height> Height of surface (default is 1080)\n");
printf(" -l <0/1> Clear surface every frame (0=no, 1=yes, default is 0)\n");
printf(" -s <0/1> Use supertile layout (0=no, 1=yes, default is 0)\n");
printf(" -t <0/1> Enable TS (0=no, 1=yes, default is 1)\n");
printf(" -e <0/1> Enable early Z (0=no, 1=yes, default is 0)\n");
printf(" -f <frames> Number of frames to render (default is 2000)\n");
printf(" -d <0/16/32> Depth/stencil surface depth\n");
printf(" -c <16/32> Color surface depth\n");
exit(1);
}
struct fbdemos_scaffold *fbs = 0;
fbdemo_init(&fbs);
struct pipe_context *pipe = fbs->pipe;
/* resources */
struct pipe_resource *rt_resource = NULL;
struct pipe_resource *z_resource = NULL;
if(!super_tiled)
etna_mesa_debug |= ETNA_DBG_NO_SUPERTILE;
if(!enable_ts)
etna_mesa_debug |= ETNA_DBG_NO_TS;
if(!early_z)
etna_mesa_debug |= ETNA_DBG_NO_EARLY_Z;
if(fmt_rt != PIPE_FORMAT_NONE)
rt_resource = fbdemo_create_2d(fbs->screen, PIPE_BIND_RENDER_TARGET, fmt_rt, width, height, 0);
if(fmt_zs != PIPE_FORMAT_NONE)
z_resource = fbdemo_create_2d(fbs->screen, PIPE_BIND_RENDER_TARGET, fmt_zs, width, height, 0);
struct pipe_resource *vtx_resource = pipe_buffer_create(fbs->screen, PIPE_BIND_VERTEX_BUFFER, PIPE_USAGE_IMMUTABLE, sizeof(vVertices));
/* vertex / index buffer setup */
struct pipe_transfer *vtx_transfer = 0;
float *vtx_logical = pipe_buffer_map(pipe, vtx_resource, PIPE_TRANSFER_WRITE | PIPE_TRANSFER_UNSYNCHRONIZED, &vtx_transfer);
assert(vtx_logical);
memcpy(vtx_logical, vVertices, sizeof(vVertices));
pipe_buffer_unmap(pipe, vtx_transfer);
struct pipe_vertex_buffer vertex_buf_desc = {
.stride = VERTEX_STRIDE*4,
.buffer_offset = 0,
.buffer = vtx_resource,
.user_buffer = 0
};
struct pipe_vertex_element pipe_vertex_elements[] = {
{ /* positions */
.src_offset = 0*4,
.instance_divisor = 0,
.vertex_buffer_index = 0,
.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT
},
{ /* texcoord */
.src_offset = 4*4,
.instance_divisor = 0,
.vertex_buffer_index = 0,
.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT
},
};
enum pipe_error
svga_swtnl_draw_range_elements(struct svga_context *svga,
struct pipe_buffer *indexBuffer,
unsigned indexSize,
unsigned min_index,
unsigned max_index,
unsigned prim, unsigned start, unsigned count)
{
struct draw_context *draw = svga->swtnl.draw;
unsigned i;
const void *map;
enum pipe_error ret;
assert(!svga->dirty);
assert(svga->state.sw.need_swtnl);
assert(draw);
ret = svga_update_state(svga, SVGA_STATE_SWTNL_DRAW);
if (ret) {
svga_context_flush(svga, NULL);
ret = svga_update_state(svga, SVGA_STATE_SWTNL_DRAW);
svga->swtnl.new_vbuf = TRUE;
assert(ret == PIPE_OK);
}
/*
* Map vertex buffers
*/
for (i = 0; i < svga->curr.num_vertex_buffers; i++) {
map = pipe_buffer_map(svga->pipe.screen,
svga->curr.vb[i].buffer,
PIPE_BUFFER_USAGE_CPU_READ);
draw_set_mapped_vertex_buffer(draw, i, map);
}
/* Map index buffer, if present */
if (indexBuffer) {
map = pipe_buffer_map(svga->pipe.screen, indexBuffer,
PIPE_BUFFER_USAGE_CPU_READ);
draw_set_mapped_element_buffer_range(draw,
indexSize,
min_index,
max_index,
map);
}
if (svga->curr.cb[PIPE_SHADER_VERTEX]) {
map = pipe_buffer_map(svga->pipe.screen,
svga->curr.cb[PIPE_SHADER_VERTEX],
PIPE_BUFFER_USAGE_CPU_READ);
assert(map);
draw_set_mapped_constant_buffer(
draw, PIPE_SHADER_VERTEX, 0,
map,
svga->curr.cb[PIPE_SHADER_VERTEX]->size);
}
draw_arrays(svga->swtnl.draw, prim, start, count);
draw_flush(svga->swtnl.draw);
/* Ensure the draw module didn't touch this */
assert(i == svga->curr.num_vertex_buffers);
/*
* unmap vertex/index buffers
*/
for (i = 0; i < svga->curr.num_vertex_buffers; i++) {
pipe_buffer_unmap(svga->pipe.screen, svga->curr.vb[i].buffer);
draw_set_mapped_vertex_buffer(draw, i, NULL);
}
if (indexBuffer) {
pipe_buffer_unmap(svga->pipe.screen, indexBuffer);
draw_set_mapped_element_buffer(draw, 0, NULL);
}
if (svga->curr.cb[PIPE_SHADER_VERTEX]) {
pipe_buffer_unmap(svga->pipe.screen,
svga->curr.cb[PIPE_SHADER_VERTEX]);
}
return ret;
}
static boolean
nv30_vertprog_validate(struct nv30_context *nv30)
{
struct pipe_screen *pscreen = nv30->pipe.screen;
struct nouveau_grobj *rankine = nv30->screen->rankine;
struct nv30_vertex_program *vp;
struct pipe_buffer *constbuf;
boolean upload_code = FALSE, upload_data = FALSE;
int i;
vp = nv30->vertprog;
constbuf = nv30->constbuf[PIPE_SHADER_VERTEX];
/* Translate TGSI shader into hw bytecode */
if (!vp->translated) {
nv30_vertprog_translate(nv30, vp);
if (!vp->translated)
return FALSE;
}
/* Allocate hw vtxprog exec slots */
if (!vp->exec) {
struct nouveau_resource *heap = nv30->screen->vp_exec_heap;
struct nouveau_stateobj *so;
uint vplen = vp->nr_insns;
if (nouveau_resource_alloc(heap, vplen, vp, &vp->exec)) {
while (heap->next && heap->size < vplen) {
struct nv30_vertex_program *evict;
evict = heap->next->priv;
nouveau_resource_free(&evict->exec);
}
if (nouveau_resource_alloc(heap, vplen, vp, &vp->exec))
assert(0);
}
so = so_new(2, 0);
so_method(so, rankine, NV34TCL_VP_START_FROM_ID, 1);
so_data (so, vp->exec->start);
so_ref(so, &vp->so);
so_ref(NULL, &so);
upload_code = TRUE;
}
/* Allocate hw vtxprog const slots */
if (vp->nr_consts && !vp->data) {
struct nouveau_resource *heap = nv30->screen->vp_data_heap;
if (nouveau_resource_alloc(heap, vp->nr_consts, vp, &vp->data)) {
while (heap->next && heap->size < vp->nr_consts) {
struct nv30_vertex_program *evict;
evict = heap->next->priv;
nouveau_resource_free(&evict->data);
}
if (nouveau_resource_alloc(heap, vp->nr_consts, vp,
&vp->data))
assert(0);
}
/*XXX: handle this some day */
assert(vp->data->start >= vp->data_start_min);
upload_data = TRUE;
if (vp->data_start != vp->data->start)
upload_code = TRUE;
}
/* If exec or data segments moved we need to patch the program to
* fixup offsets and register IDs.
*/
if (vp->exec_start != vp->exec->start) {
for (i = 0; i < vp->nr_insns; i++) {
struct nv30_vertex_program_exec *vpi = &vp->insns[i];
if (vpi->has_branch_offset) {
assert(0);
}
}
vp->exec_start = vp->exec->start;
}
if (vp->nr_consts && vp->data_start != vp->data->start) {
for (i = 0; i < vp->nr_insns; i++) {
struct nv30_vertex_program_exec *vpi = &vp->insns[i];
if (vpi->const_index >= 0) {
vpi->data[1] &= ~NV30_VP_INST_CONST_SRC_MASK;
vpi->data[1] |=
(vpi->const_index + vp->data->start) <<
NV30_VP_INST_CONST_SRC_SHIFT;
}
}
vp->data_start = vp->data->start;
}
/* Update + Upload constant values */
if (vp->nr_consts) {
float *map = NULL;
if (constbuf) {
map = pipe_buffer_map(pscreen, constbuf,
PIPE_BUFFER_USAGE_CPU_READ);
}
for (i = 0; i < vp->nr_consts; i++) {
struct nv30_vertex_program_data *vpd = &vp->consts[i];
if (vpd->index >= 0) {
if (!upload_data &&
!memcmp(vpd->value, &map[vpd->index * 4],
4 * sizeof(float)))
continue;
memcpy(vpd->value, &map[vpd->index * 4],
4 * sizeof(float));
}
BEGIN_RING(rankine, NV34TCL_VP_UPLOAD_CONST_ID, 5);
OUT_RING (i + vp->data->start);
OUT_RINGp ((uint32_t *)vpd->value, 4);
}
if (constbuf)
pipe_buffer_unmap(pscreen, constbuf);
}
/* Upload vtxprog */
if (upload_code) {
#if 0
for (i = 0; i < vp->nr_insns; i++) {
NOUVEAU_MSG("VP inst %d: 0x%08x 0x%08x 0x%08x 0x%08x\n",
i, vp->insns[i].data[0], vp->insns[i].data[1],
vp->insns[i].data[2], vp->insns[i].data[3]);
}
#endif
BEGIN_RING(rankine, NV34TCL_VP_UPLOAD_FROM_ID, 1);
OUT_RING (vp->exec->start);
for (i = 0; i < vp->nr_insns; i++) {
BEGIN_RING(rankine, NV34TCL_VP_UPLOAD_INST(0), 4);
OUT_RINGp (vp->insns[i].data, 4);
}
}
if (vp->so != nv30->state.hw[NV30_STATE_VERTPROG]) {
so_ref(vp->so, &nv30->state.hw[NV30_STATE_VERTPROG]);
return TRUE;
}
return FALSE;
}
static void
u_vbuf_translate_begin(struct u_vbuf_priv *mgr,
int min_index, int max_index)
{
struct translate_key key;
struct translate_element *te;
unsigned tr_elem_index[PIPE_MAX_ATTRIBS];
struct translate *tr;
boolean vb_translated[PIPE_MAX_ATTRIBS] = {0};
uint8_t *out_map;
struct pipe_transfer *vb_transfer[PIPE_MAX_ATTRIBS] = {0};
struct pipe_resource *out_buffer = NULL;
unsigned i, num_verts, out_offset;
boolean upload_flushed = FALSE;
memset(&key, 0, sizeof(key));
memset(tr_elem_index, 0xff, sizeof(tr_elem_index));
/* Get a new vertex buffer slot. */
mgr->fallback_vb_slot = u_vbuf_get_free_real_vb_slot(mgr);
if (mgr->fallback_vb_slot == ~0) {
return; /* XXX error, not enough attribs */
}
/* Initialize the description of how vertices should be translated. */
for (i = 0; i < mgr->ve->count; i++) {
enum pipe_format output_format = mgr->ve->native_format[i];
unsigned output_format_size = mgr->ve->native_format_size[i];
/* Check for support. */
if (!mgr->ve->incompatible_layout_elem[i] &&
!mgr->incompatible_vb[mgr->ve->ve[i].vertex_buffer_index]) {
continue;
}
/* Workaround for translate: output floats instead of halfs. */
switch (output_format) {
case PIPE_FORMAT_R16_FLOAT:
output_format = PIPE_FORMAT_R32_FLOAT;
output_format_size = 4;
break;
case PIPE_FORMAT_R16G16_FLOAT:
output_format = PIPE_FORMAT_R32G32_FLOAT;
output_format_size = 8;
break;
case PIPE_FORMAT_R16G16B16_FLOAT:
output_format = PIPE_FORMAT_R32G32B32_FLOAT;
output_format_size = 12;
break;
case PIPE_FORMAT_R16G16B16A16_FLOAT:
output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
output_format_size = 16;
break;
default:;
}
/* Add this vertex element. */
te = &key.element[key.nr_elements];
te->type = TRANSLATE_ELEMENT_NORMAL;
te->instance_divisor = 0;
te->input_buffer = mgr->ve->ve[i].vertex_buffer_index;
te->input_format = mgr->ve->ve[i].src_format;
te->input_offset = mgr->ve->ve[i].src_offset;
te->output_format = output_format;
te->output_offset = key.output_stride;
key.output_stride += output_format_size;
vb_translated[mgr->ve->ve[i].vertex_buffer_index] = TRUE;
tr_elem_index[i] = key.nr_elements;
key.nr_elements++;
}
/* Get a translate object. */
tr = translate_cache_find(mgr->translate_cache, &key);
/* Map buffers we want to translate. */
for (i = 0; i < mgr->b.nr_vertex_buffers; i++) {
if (vb_translated[i]) {
struct pipe_vertex_buffer *vb = &mgr->b.vertex_buffer[i];
uint8_t *map = pipe_buffer_map(mgr->pipe, vb->buffer,
PIPE_TRANSFER_READ, &vb_transfer[i]);
tr->set_buffer(tr, i,
map + vb->buffer_offset + vb->stride * min_index,
vb->stride, ~0);
}
}
/* Create and map the output buffer. */
num_verts = max_index + 1 - min_index;
u_upload_alloc(mgr->b.uploader,
key.output_stride * min_index,
key.output_stride * num_verts,
&out_offset, &out_buffer, &upload_flushed,
(void**)&out_map);
out_offset -= key.output_stride * min_index;
/* Translate. */
tr->run(tr, 0, num_verts, 0, out_map);
/* Unmap all buffers. */
for (i = 0; i < mgr->b.nr_vertex_buffers; i++) {
if (vb_translated[i]) {
pipe_buffer_unmap(mgr->pipe, vb_transfer[i]);
}
}
/* Setup the new vertex buffer. */
mgr->b.real_vertex_buffer[mgr->fallback_vb_slot].buffer_offset = out_offset;
mgr->b.real_vertex_buffer[mgr->fallback_vb_slot].stride = key.output_stride;
/* Move the buffer reference. */
pipe_resource_reference(
&mgr->b.real_vertex_buffer[mgr->fallback_vb_slot].buffer, NULL);
mgr->b.real_vertex_buffer[mgr->fallback_vb_slot].buffer = out_buffer;
out_buffer = NULL;
/* Setup new vertex elements. */
for (i = 0; i < mgr->ve->count; i++) {
if (tr_elem_index[i] < key.nr_elements) {
te = &key.element[tr_elem_index[i]];
mgr->fallback_velems[i].instance_divisor = mgr->ve->ve[i].instance_divisor;
mgr->fallback_velems[i].src_format = te->output_format;
mgr->fallback_velems[i].src_offset = te->output_offset;
mgr->fallback_velems[i].vertex_buffer_index = mgr->fallback_vb_slot;
} else {
memcpy(&mgr->fallback_velems[i], &mgr->ve->ve[i],
sizeof(struct pipe_vertex_element));
}
}
mgr->fallback_ve =
mgr->pipe->create_vertex_elements_state(mgr->pipe, mgr->ve->count,
mgr->fallback_velems);
/* Preserve saved_ve. */
mgr->ve_binding_lock = TRUE;
mgr->pipe->bind_vertex_elements_state(mgr->pipe, mgr->fallback_ve);
mgr->ve_binding_lock = FALSE;
}
/**
* Emit all the constants in a constant buffer for a shader stage.
* On VGPU10, emit_consts_vgpu10 is used instead.
*/
static enum pipe_error
emit_consts_vgpu9(struct svga_context *svga, unsigned shader)
{
const struct pipe_constant_buffer *cbuf;
struct svga_screen *ss = svga_screen(svga->pipe.screen);
struct pipe_transfer *transfer = NULL;
unsigned count;
const float (*data)[4] = NULL;
unsigned i;
enum pipe_error ret = PIPE_OK;
const unsigned offset = 0;
assert(shader < PIPE_SHADER_TYPES);
assert(!svga_have_vgpu10(svga));
/* Only one constant buffer per shader is supported before VGPU10.
* This is only an approximate check against that.
*/
assert(svga->curr.constbufs[shader][1].buffer == NULL);
cbuf = &svga->curr.constbufs[shader][0];
if (svga->curr.constbufs[shader][0].buffer) {
/* emit user-provided constants */
data = (const float (*)[4])
pipe_buffer_map(&svga->pipe, svga->curr.constbufs[shader][0].buffer,
PIPE_TRANSFER_READ, &transfer);
if (!data) {
return PIPE_ERROR_OUT_OF_MEMORY;
}
/* sanity check */
assert(cbuf->buffer->width0 >=
cbuf->buffer_size);
/* Use/apply the constant buffer size and offsets here */
count = cbuf->buffer_size / (4 * sizeof(float));
data += cbuf->buffer_offset / (4 * sizeof(float));
if (ss->hw_version >= SVGA3D_HWVERSION_WS8_B1) {
ret = emit_const_range( svga, shader, offset, count, data );
}
else {
for (i = 0; i < count; i++) {
ret = emit_const( svga, shader, offset + i, data[i] );
if (ret != PIPE_OK) {
break;
}
}
}
pipe_buffer_unmap(&svga->pipe, transfer);
if (ret != PIPE_OK) {
return ret;
}
}
/* emit extra shader constants */
{
const struct svga_shader_variant *variant = NULL;
unsigned offset;
float extras[MAX_EXTRA_CONSTS][4];
unsigned count, i;
switch (shader) {
case PIPE_SHADER_VERTEX:
variant = svga->state.hw_draw.vs;
count = svga_get_extra_vs_constants(svga, (float *) extras);
break;
case PIPE_SHADER_FRAGMENT:
variant = svga->state.hw_draw.fs;
count = svga_get_extra_fs_constants(svga, (float *) extras);
break;
default:
assert(!"Unexpected shader type");
count = 0;
}
assert(variant);
offset = variant->shader->info.file_max[TGSI_FILE_CONSTANT] + 1;
assert(count <= ARRAY_SIZE(extras));
if (count > 0) {
if (ss->hw_version >= SVGA3D_HWVERSION_WS8_B1) {
ret = emit_const_range(svga, shader, offset, count,
(const float (*) [4])extras);
}
else {
for (i = 0; i < count; i++) {
ret = emit_const(svga, shader, offset + i, extras[i]);
if (ret != PIPE_OK)
return ret;
}
}
}
}
return ret;
}
enum pipe_error
svga_swtnl_draw_vbo(struct svga_context *svga,
const struct pipe_draw_info *info)
{
struct pipe_transfer *vb_transfer[PIPE_MAX_ATTRIBS];
struct pipe_transfer *ib_transfer = NULL;
struct pipe_transfer *cb_transfer = NULL;
struct draw_context *draw = svga->swtnl.draw;
unsigned i;
const void *map;
enum pipe_error ret;
assert(!svga->dirty);
assert(svga->state.sw.need_swtnl);
assert(draw);
/* Make sure that the need_swtnl flag does not go away */
svga->state.sw.in_swtnl_draw = TRUE;
ret = svga_update_state(svga, SVGA_STATE_SWTNL_DRAW);
if (ret != PIPE_OK) {
svga_context_flush(svga, NULL);
ret = svga_update_state(svga, SVGA_STATE_SWTNL_DRAW);
svga->swtnl.new_vbuf = TRUE;
assert(ret == PIPE_OK);
}
/*
* Map vertex buffers
*/
for (i = 0; i < svga->curr.num_vertex_buffers; i++) {
if (svga->curr.vb[i].buffer) {
map = pipe_buffer_map(&svga->pipe,
svga->curr.vb[i].buffer,
PIPE_TRANSFER_READ,
&vb_transfer[i]);
draw_set_mapped_vertex_buffer(draw, i, map);
}
}
/* Map index buffer, if present */
map = NULL;
if (info->indexed && svga->curr.ib.buffer) {
map = pipe_buffer_map(&svga->pipe, svga->curr.ib.buffer,
PIPE_TRANSFER_READ,
&ib_transfer);
draw_set_indexes(draw,
(const ubyte *) map + svga->curr.ib.offset,
svga->curr.ib.index_size);
}
if (svga->curr.cb[PIPE_SHADER_VERTEX]) {
map = pipe_buffer_map(&svga->pipe,
svga->curr.cb[PIPE_SHADER_VERTEX],
PIPE_TRANSFER_READ,
&cb_transfer);
assert(map);
draw_set_mapped_constant_buffer(
draw, PIPE_SHADER_VERTEX, 0,
map,
svga->curr.cb[PIPE_SHADER_VERTEX]->width0);
}
draw_vbo(draw, info);
draw_flush(svga->swtnl.draw);
/* Ensure the draw module didn't touch this */
assert(i == svga->curr.num_vertex_buffers);
/*
* unmap vertex/index buffers
*/
for (i = 0; i < svga->curr.num_vertex_buffers; i++) {
if (svga->curr.vb[i].buffer) {
pipe_buffer_unmap(&svga->pipe, vb_transfer[i]);
draw_set_mapped_vertex_buffer(draw, i, NULL);
}
}
if (ib_transfer) {
pipe_buffer_unmap(&svga->pipe, ib_transfer);
draw_set_indexes(draw, NULL, 0);
}
if (svga->curr.cb[PIPE_SHADER_VERTEX]) {
pipe_buffer_unmap(&svga->pipe, cb_transfer);
}
/* Now safe to remove the need_swtnl flag in any update_state call */
svga->state.sw.in_swtnl_draw = FALSE;
svga->dirty |= SVGA_NEW_NEED_PIPELINE | SVGA_NEW_NEED_SWVFETCH;
return ret;
}
static boolean
i915_draw_range_elements(struct pipe_context *pipe,
struct pipe_buffer *indexBuffer,
unsigned indexSize,
unsigned min_index,
unsigned max_index,
unsigned prim, unsigned start, unsigned count)
{
struct i915_context *i915 = i915_context(pipe);
struct draw_context *draw = i915->draw;
unsigned i;
if (i915->dirty)
i915_update_derived(i915);
/*
* Map vertex buffers
*/
for (i = 0; i < i915->num_vertex_buffers; i++) {
void *buf = pipe_buffer_map(pipe->screen, i915->vertex_buffer[i].buffer,
PIPE_BUFFER_USAGE_CPU_READ);
draw_set_mapped_vertex_buffer(draw, i, buf);
}
/*
* Map index buffer, if present
*/
if (indexBuffer) {
void *mapped_indexes = pipe_buffer_map(pipe->screen, indexBuffer,
PIPE_BUFFER_USAGE_CPU_READ);
draw_set_mapped_element_buffer_range(draw, indexSize,
min_index,
max_index,
mapped_indexes);
} else {
draw_set_mapped_element_buffer(draw, 0, NULL);
}
draw_set_mapped_constant_buffer(draw,
i915->current.constants[PIPE_SHADER_VERTEX],
(i915->current.num_user_constants[PIPE_SHADER_VERTEX] *
4 * sizeof(float)));
/*
* Do the drawing
*/
draw_arrays(i915->draw, prim, start, count);
/*
* unmap vertex/index buffers
*/
for (i = 0; i < i915->num_vertex_buffers; i++) {
pipe_buffer_unmap(pipe->screen, i915->vertex_buffer[i].buffer);
draw_set_mapped_vertex_buffer(draw, i, NULL);
}
if (indexBuffer) {
pipe_buffer_unmap(pipe->screen, indexBuffer);
draw_set_mapped_element_buffer_range(draw, 0, start, start + count - 1, NULL);
}
return TRUE;
}
