fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-08 15:38:09 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

[g3dvl] move the rest of the calculations into the vertex shader

Browse source
This commit is contained in:
Christian König 2010-11-11 11:53:12 +01:00
parent 1be1aa7ba1
commit 4c90c039f0
2 changed files with 122 additions and 136 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

254
src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c
View file

@ -78,12 +78,29 @@ enum MACROBLOCK_TYPE
NUM_MACROBLOCK_TYPES
};
/* vertices for two triangles covering four block */
static const struct vertex2f const_rectangle[4][2][3] = {
{
{ {0.0f, 0.0f}, {0.0f, 0.5f}, {0.5f, 0.0f} },
{ {0.5f, 0.0f}, {0.0f, 0.5f}, {0.5f, 0.5f} }
}, {
{ {0.5f, 0.0f}, {0.5f, 0.5f}, {1.0f, 0.0f} },
{ {1.0f, 0.0f}, {0.5f, 0.5f}, {1.0f, 0.5f} }
}, {
{ {0.0f, 0.5f}, {0.0f, 1.0f}, {0.5f, 0.5f} },
{ {0.5f, 0.5f}, {0.0f, 1.0f}, {0.5f, 1.0f} }
}, {
{ {0.5f, 0.5f}, {0.5f, 1.0f}, {1.0f, 0.5f} },
{ {1.0f, 0.5f}, {0.5f, 1.0f}, {1.0f, 1.0f} }
}
};
static void *
create_vert_shader(struct vl_mpeg12_mc_renderer *r, unsigned ref_frames, unsigned mv_per_frame)
{
struct ureg_program *shader;
struct ureg_src norm, mbs;
struct ureg_src vpos, vtex[3], vmv[4];
struct ureg_src vrect, vpos, vtex[3], vmv[4];
struct ureg_dst t_vpos, scale;
struct ureg_dst o_vpos, o_vtex[3], o_vmv[4], o_line;
unsigned i, j, count;
@ -99,11 +116,12 @@ create_vert_shader(struct vl_mpeg12_mc_renderer *r, unsigned ref_frames, unsigne
t_vpos = ureg_DECL_temporary(shader);
scale = ureg_DECL_temporary(shader);
vpos = ureg_DECL_vs_input(shader, 0);
vrect = ureg_DECL_vs_input(shader, 0);
vpos = ureg_DECL_vs_input(shader, 1);
o_vpos = ureg_DECL_output(shader, TGSI_SEMANTIC_POSITION, 0);
for (i = 0; i < 3; ++i) {
vtex[i] = ureg_DECL_vs_input(shader, 1 + i);
vtex[i] = ureg_DECL_vs_input(shader, 2 + i);
o_vtex[i] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, 1 + i);
}
@ -111,13 +129,13 @@ create_vert_shader(struct vl_mpeg12_mc_renderer *r, unsigned ref_frames, unsigne
for (i = 0; i < ref_frames; ++i) {
for (j = 0; j < 2; ++j) {
if(j < mv_per_frame) {
vmv[count] = ureg_DECL_vs_input(shader, 4 + i * 2 + j);
vmv[count] = ureg_DECL_vs_input(shader, 5 + i * 2 + j);
o_vmv[count] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, 4 + count);
count++;
}
/* workaround for r600g */
else if(ref_frames == 2)
ureg_DECL_vs_input(shader, 4 + i * 2 + j);
ureg_DECL_vs_input(shader, 5 + i * 2 + j);
}
}
@ -126,10 +144,12 @@ create_vert_shader(struct vl_mpeg12_mc_renderer *r, unsigned ref_frames, unsigne
/*
* scale = norm * mbs;
*
* t_vpos = vpos * scale
* o_vpos = t_vpos
* t_vpos = (vpos + vrect) * scale
* o_vpos.xy = t_vpos
* o_vpos.zw = vpos
*
* o_vtex[0..2] = vtex[0..2] * scale
* o_vtex[0..2].xy = t_vpos
* o_vtex[0..2].z = vtex[0..2].z
*
* if(count > 0) { // Apply motion vectors
* scale = norm * 0.5;
@ -143,12 +163,13 @@ create_vert_shader(struct vl_mpeg12_mc_renderer *r, unsigned ref_frames, unsigne
*/
ureg_MUL(shader, ureg_writemask(scale, TGSI_WRITEMASK_XY), norm, mbs);
ureg_MUL(shader, ureg_writemask(t_vpos, TGSI_WRITEMASK_XY), vpos, ureg_src(scale));
ureg_ADD(shader, ureg_writemask(t_vpos, TGSI_WRITEMASK_XY), vpos, vrect);
ureg_MUL(shader, ureg_writemask(t_vpos, TGSI_WRITEMASK_XY), ureg_src(t_vpos), ureg_src(scale));
ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_XY), ureg_src(t_vpos));
ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_ZW), vpos);
for (i = 0; i < 3; ++i) {
ureg_MUL(shader, ureg_writemask(o_vtex[i], TGSI_WRITEMASK_XY), vpos, ureg_src(scale));
ureg_MOV(shader, ureg_writemask(o_vtex[i], TGSI_WRITEMASK_XY), ureg_src(t_vpos));
ureg_MOV(shader, ureg_writemask(o_vtex[i], TGSI_WRITEMASK_Z), ureg_scalar(vtex[i], TGSI_SWIZZLE_X));
}
@ -160,7 +181,7 @@ create_vert_shader(struct vl_mpeg12_mc_renderer *r, unsigned ref_frames, unsigne
if (interlaced) {
ureg_MOV(shader, ureg_writemask(o_line, TGSI_WRITEMASK_X), ureg_imm1f(shader, 1.0f));
ureg_MUL(shader, ureg_writemask(o_line, TGSI_WRITEMASK_Y), vpos, ureg_imm1f(shader, MACROBLOCK_HEIGHT / 2));
ureg_MUL(shader, ureg_writemask(o_line, TGSI_WRITEMASK_Y), vrect, ureg_imm1f(shader, MACROBLOCK_HEIGHT / 2));
}
ureg_release_temporary(shader, t_vpos);
@ -584,7 +605,7 @@ static bool
init_buffers(struct vl_mpeg12_mc_renderer *r)
{
struct pipe_resource template;
struct pipe_vertex_element vertex_elems[8];
struct pipe_vertex_element vertex_elems[9];
struct pipe_sampler_view sampler_view;
const unsigned mbw =
@ -642,6 +663,16 @@ init_buffers(struct vl_mpeg12_mc_renderer *r)
r->sampler_views.all[i] = r->pipe->create_sampler_view(r->pipe, r->textures.all[i], &sampler_view);
}
r->vertex_bufs.individual.rect.stride = sizeof(struct vertex2f);
r->vertex_bufs.individual.rect.max_index = 24 * r->macroblocks_per_batch - 1;
r->vertex_bufs.individual.rect.buffer_offset = 0;
r->vertex_bufs.individual.rect.buffer = pipe_buffer_create
(
r->pipe->screen,
PIPE_BIND_VERTEX_BUFFER,
sizeof(struct vertex2f) * 24 * r->macroblocks_per_batch
);
r->vertex_bufs.individual.ycbcr.stride = sizeof(struct vert_stream_0);
r->vertex_bufs.individual.ycbcr.max_index = 24 * r->macroblocks_per_batch - 1;
r->vertex_bufs.individual.ycbcr.buffer_offset = 0;
@ -653,12 +684,12 @@ init_buffers(struct vl_mpeg12_mc_renderer *r)
sizeof(struct vert_stream_0) * 24 * r->macroblocks_per_batch
);
for (i = 1; i < 3; ++i) {
r->vertex_bufs.all[i].stride = sizeof(struct vertex2f) * 2;
r->vertex_bufs.all[i].max_index = 24 * r->macroblocks_per_batch - 1;
r->vertex_bufs.all[i].buffer_offset = 0;
for (i = 0; i < 2; ++i) {
r->vertex_bufs.individual.ref[i].stride = sizeof(struct vertex2f) * 2;
r->vertex_bufs.individual.ref[i].max_index = 24 * r->macroblocks_per_batch - 1;
r->vertex_bufs.individual.ref[i].buffer_offset = 0;
/* XXX: Create with usage DYNAMIC or STREAM */
r->vertex_bufs.all[i].buffer = pipe_buffer_create
r->vertex_bufs.individual.ref[i].buffer = pipe_buffer_create
(
r->pipe->screen,
PIPE_BIND_VERTEX_BUFFER,
@ -668,57 +699,63 @@ init_buffers(struct vl_mpeg12_mc_renderer *r)
memset(&vertex_elems, 0, sizeof(vertex_elems));
/* Position element */
/* Rectangle element */
vertex_elems[0].src_offset = 0;
vertex_elems[0].instance_divisor = 0;
vertex_elems[0].vertex_buffer_index = 0;
vertex_elems[0].src_format = PIPE_FORMAT_R32G32_FLOAT;
/* Luma, texcoord element */
vertex_elems[1].src_offset = sizeof(struct vertex2f);
/* Position element */
vertex_elems[1].src_offset = 0;
vertex_elems[1].instance_divisor = 0;
vertex_elems[1].vertex_buffer_index = 0;
vertex_elems[1].src_format = PIPE_FORMAT_R32_FLOAT;
vertex_elems[1].vertex_buffer_index = 1;
vertex_elems[1].src_format = PIPE_FORMAT_R32G32_FLOAT;
/* Chroma Cr texcoord element */
vertex_elems[2].src_offset = sizeof(struct vertex2f) + sizeof(float);
/* Luma, texcoord element */
vertex_elems[2].src_offset = sizeof(struct vertex2f);
vertex_elems[2].instance_divisor = 0;
vertex_elems[2].vertex_buffer_index = 0;
vertex_elems[2].vertex_buffer_index = 1;
vertex_elems[2].src_format = PIPE_FORMAT_R32_FLOAT;
/* Chroma Cb texcoord element */
vertex_elems[3].src_offset = sizeof(struct vertex2f) + sizeof(float) * 2;
/* Chroma Cr texcoord element */
vertex_elems[3].src_offset = sizeof(struct vertex2f) + sizeof(float);
vertex_elems[3].instance_divisor = 0;
vertex_elems[3].vertex_buffer_index = 0;
vertex_elems[3].vertex_buffer_index = 1;
vertex_elems[3].src_format = PIPE_FORMAT_R32_FLOAT;
/* First ref surface top field texcoord element */
vertex_elems[4].src_offset = 0;
/* Chroma Cb texcoord element */
vertex_elems[4].src_offset = sizeof(struct vertex2f) + sizeof(float) * 2;
vertex_elems[4].instance_divisor = 0;
vertex_elems[4].vertex_buffer_index = 1;
vertex_elems[4].src_format = PIPE_FORMAT_R32G32_FLOAT;
vertex_elems[4].src_format = PIPE_FORMAT_R32_FLOAT;
/* First ref surface bottom field texcoord element */
vertex_elems[5].src_offset = sizeof(struct vertex2f);
/* First ref surface top field texcoord element */
vertex_elems[5].src_offset = 0;
vertex_elems[5].instance_divisor = 0;
vertex_elems[5].vertex_buffer_index = 1;
vertex_elems[5].vertex_buffer_index = 2;
vertex_elems[5].src_format = PIPE_FORMAT_R32G32_FLOAT;
/* Second ref surface top field texcoord element */
vertex_elems[6].src_offset = 0;
/* First ref surface bottom field texcoord element */
vertex_elems[6].src_offset = sizeof(struct vertex2f);
vertex_elems[6].instance_divisor = 0;
vertex_elems[6].vertex_buffer_index = 2;
vertex_elems[6].src_format = PIPE_FORMAT_R32G32_FLOAT;
/* Second ref surface bottom field texcoord element */
vertex_elems[7].src_offset = sizeof(struct vertex2f);
/* Second ref surface top field texcoord element */
vertex_elems[7].src_offset = 0;
vertex_elems[7].instance_divisor = 0;
vertex_elems[7].vertex_buffer_index = 2;
vertex_elems[7].vertex_buffer_index = 3;
vertex_elems[7].src_format = PIPE_FORMAT_R32G32_FLOAT;
r->vertex_elems_state.individual.i = r->pipe->create_vertex_elements_state(r->pipe, 4, vertex_elems);
r->vertex_elems_state.individual.p = r->pipe->create_vertex_elements_state(r->pipe, 6, vertex_elems);
r->vertex_elems_state.individual.b = r->pipe->create_vertex_elements_state(r->pipe, 8, vertex_elems);
/* Second ref surface bottom field texcoord element */
vertex_elems[8].src_offset = sizeof(struct vertex2f);
vertex_elems[8].instance_divisor = 0;
vertex_elems[8].vertex_buffer_index = 3;
vertex_elems[8].src_format = PIPE_FORMAT_R32G32_FLOAT;
r->vertex_elems_state.individual.i = r->pipe->create_vertex_elements_state(r->pipe, 5, vertex_elems);
r->vertex_elems_state.individual.p = r->pipe->create_vertex_elements_state(r->pipe, 7, vertex_elems);
r->vertex_elems_state.individual.b = r->pipe->create_vertex_elements_state(r->pipe, 9, vertex_elems);
r->vs_const_buf = pipe_buffer_create
(
@ -776,87 +813,40 @@ get_macroblock_type(struct pipe_mpeg12_macroblock *mb)
void
gen_block_verts(struct vert_stream_0 *vb, struct pipe_mpeg12_macroblock *mb,
const struct vertex2f *offset,
unsigned luma_mask, unsigned cb_mask, unsigned cr_mask,
bool use_zeroblocks, struct vertex2f *zero_blocks)
unsigned luma_mask, unsigned cb_mask, unsigned cr_mask)
{
unsigned cbp = mb->cbp;
unsigned mbx = mb->mbx;
unsigned mby = mb->mby;
const struct vertex2f half =
{
0.5f,
0.5f
};
struct vertex2f v;
unsigned i;
struct vert_stream_0 v;
assert(vb);
assert(offset);
assert(zero_blocks || !use_zeroblocks);
/* Generate vertices for two triangles covering a block */
v.x = mbx + offset->x;
v.y = mby + offset->y;
v.pos.x = mb->mbx;
v.pos.y = mb->mby;
vb[0].pos.x = v.x;
vb[0].pos.y = v.y;
vb[1].pos.x = v.x;
vb[1].pos.y = v.y + half.y;
vb[2].pos.x = v.x + half.x;
vb[2].pos.y = v.y;
vb[3].pos.x = v.x + half.x;
vb[3].pos.y = v.y;
vb[4].pos.x = v.x;
vb[4].pos.y = v.y + half.y;
vb[5].pos.x = v.x + half.x;
vb[5].pos.y = v.y + half.y;
/* Generate texcoords for the triangles, either pointing to the correct area on the luma/chroma texture
or if zero blocks are being used, to the zero block if the appropriate CBP bits aren't set (i.e. no data
for this channel is defined for this block) */
if (!use_zeroblocks || cbp & luma_mask || mb->dct_type == PIPE_MPEG12_DCT_TYPE_FIELD) {
v.x = 0.0f;
if (cbp & luma_mask || mb->dct_type == PIPE_MPEG12_DCT_TYPE_FIELD) {
v.luma_eb = 0.0f;
}
else {
v.x = -1.0f;
v.luma_eb = -1.0f;
}
vb[0].luma_eb = v.x;
vb[1].luma_eb = v.x;
vb[2].luma_eb = v.x;
vb[3].luma_eb = v.x;
vb[4].luma_eb = v.x;
vb[5].luma_eb = v.x;
if (!use_zeroblocks || cbp & cb_mask) {
v.x = 0.0f;
if (cbp & cb_mask) {
v.cb_eb = 0.0f;
}
else {
v.x = -1.0f;
v.cb_eb = -1.0f;
}
vb[0].cb_eb = v.x;
vb[1].cb_eb = v.x;
vb[2].cb_eb = v.x;
vb[3].cb_eb = v.x;
vb[4].cb_eb = v.x;
vb[5].cb_eb = v.x;
if (!use_zeroblocks || cbp & cr_mask) {
v.x = 0.0f;
if (cbp & cr_mask) {
v.cr_eb = 0.0f;
}
else {
v.x = -1.0f;
v.cr_eb = -1.0f;
}
vb[0].cr_eb = v.x;
vb[1].cr_eb = v.x;
vb[2].cr_eb = v.x;
vb[3].cr_eb = v.x;
vb[4].cr_eb = v.x;
vb[5].cr_eb = v.x;
for ( i = 0; i < 6; ++i )
memcpy(vb + i, &v, sizeof(v));
}
void
@ -970,30 +960,12 @@ gen_macroblock_verts(struct vl_mpeg12_mc_renderer *r,
}
case PIPE_MPEG12_MACROBLOCK_TYPE_INTRA:
{
const struct vertex2f offsets[2][2] =
{
{
{0, 0}, {0, 0.5f}
},
{
{0.5f, 0}, {0.5f, 0.5f}
}
};
const bool use_zb = r->eb_handling == VL_MPEG12_MC_RENDERER_EMPTY_BLOCK_XFER_ONE;
struct vert_stream_0 *vb = ycbcr_vb + pos * 24;
gen_block_verts(vb, mb, &offsets[0][0],
32, 2, 1, use_zb, r->zero_block);
gen_block_verts(vb + 6, mb, &offsets[1][0],
16, 2, 1, use_zb, r->zero_block);
gen_block_verts(vb + 12, mb, &offsets[0][1],
8, 2, 1, use_zb, r->zero_block);
gen_block_verts(vb + 18, mb, &offsets[1][1],
4, 2, 1, use_zb, r->zero_block);
gen_block_verts(vb, mb, 32, 2, 1);
gen_block_verts(vb + 6, mb, 16, 2, 1);
gen_block_verts(vb + 12, mb, 8, 2, 1);
gen_block_verts(vb + 18, mb, 4, 2, 1);
break;
}
@ -1090,11 +1062,25 @@ flush(struct vl_mpeg12_mc_renderer *r)
unsigned vb_start = 0;
struct vertex_shader_consts *vs_consts;
struct pipe_transfer *buf_transfer;
struct vertex2f *rect;
unsigned i;
assert(r);
assert(r->num_macroblocks == r->macroblocks_per_batch);
rect = pipe_buffer_map
(
r->pipe,
r->vertex_bufs.individual.rect.buffer,
PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
&buf_transfer
);
for ( i = 0; i < r->macroblocks_per_batch; ++i)
memcpy(rect + i * 24, &const_rectangle, sizeof(const_rectangle));
pipe_buffer_unmap(r->pipe, r->vertex_bufs.individual.rect.buffer, buf_transfer);
gen_macroblock_stream(r, num_macroblocks);
r->fb_state.cbufs[0] = r->surface;
@ -1118,7 +1104,7 @@ flush(struct vl_mpeg12_mc_renderer *r)
r->vs_const_buf);
if (num_macroblocks[MACROBLOCK_TYPE_INTRA] > 0) {
r->pipe->set_vertex_buffers(r->pipe, 1, r->vertex_bufs.all);
r->pipe->set_vertex_buffers(r->pipe, 2, r->vertex_bufs.all);
r->pipe->bind_vertex_elements_state(r->pipe, r->vertex_elems_state.individual.i);
r->pipe->set_fragment_sampler_views(r->pipe, 3, r->sampler_views.all);
r->pipe->bind_fragment_sampler_states(r->pipe, 3, r->samplers.all);
@ -1131,7 +1117,7 @@ flush(struct vl_mpeg12_mc_renderer *r)
}
if (num_macroblocks[MACROBLOCK_TYPE_FWD_FRAME_PRED] > 0) {
r->pipe->set_vertex_buffers(r->pipe, 2, r->vertex_bufs.all);
r->pipe->set_vertex_buffers(r->pipe, 3, r->vertex_bufs.all);
r->pipe->bind_vertex_elements_state(r->pipe, r->vertex_elems_state.individual.p);
r->textures.individual.ref[0] = r->past->texture;
r->sampler_views.individual.ref[0] = find_or_create_sampler_view(r, r->past);
@ -1146,7 +1132,7 @@ flush(struct vl_mpeg12_mc_renderer *r)
}
if (num_macroblocks[MACROBLOCK_TYPE_FWD_FIELD_PRED] > 0) {
r->pipe->set_vertex_buffers(r->pipe, 2, r->vertex_bufs.all);
r->pipe->set_vertex_buffers(r->pipe, 3, r->vertex_bufs.all);
r->pipe->bind_vertex_elements_state(r->pipe, r->vertex_elems_state.individual.p);
r->textures.individual.ref[0] = r->past->texture;
r->sampler_views.individual.ref[0] = find_or_create_sampler_view(r, r->past);
@ -1161,7 +1147,7 @@ flush(struct vl_mpeg12_mc_renderer *r)
}
if (num_macroblocks[MACROBLOCK_TYPE_BKWD_FRAME_PRED] > 0) {
r->pipe->set_vertex_buffers(r->pipe, 2, r->vertex_bufs.all);
r->pipe->set_vertex_buffers(r->pipe, 3, r->vertex_bufs.all);
r->pipe->bind_vertex_elements_state(r->pipe, r->vertex_elems_state.individual.p);
r->textures.individual.ref[0] = r->future->texture;
r->sampler_views.individual.ref[0] = find_or_create_sampler_view(r, r->future);
@ -1176,7 +1162,7 @@ flush(struct vl_mpeg12_mc_renderer *r)
}
if (num_macroblocks[MACROBLOCK_TYPE_BKWD_FIELD_PRED] > 0) {
r->pipe->set_vertex_buffers(r->pipe, 2, r->vertex_bufs.all);
r->pipe->set_vertex_buffers(r->pipe, 3, r->vertex_bufs.all);
r->pipe->bind_vertex_elements_state(r->pipe, r->vertex_elems_state.individual.p);
r->textures.individual.ref[0] = r->future->texture;
r->sampler_views.individual.ref[0] = find_or_create_sampler_view(r, r->future);
@ -1191,7 +1177,7 @@ flush(struct vl_mpeg12_mc_renderer *r)
}
if (num_macroblocks[MACROBLOCK_TYPE_BI_FRAME_PRED] > 0) {
r->pipe->set_vertex_buffers(r->pipe, 3, r->vertex_bufs.all);
r->pipe->set_vertex_buffers(r->pipe, 4, r->vertex_bufs.all);
r->pipe->bind_vertex_elements_state(r->pipe, r->vertex_elems_state.individual.b);
r->textures.individual.ref[0] = r->past->texture;
r->textures.individual.ref[1] = r->future->texture;
@ -1208,7 +1194,7 @@ flush(struct vl_mpeg12_mc_renderer *r)
}
if (num_macroblocks[MACROBLOCK_TYPE_BI_FIELD_PRED] > 0) {
r->pipe->set_vertex_buffers(r->pipe, 3, r->vertex_bufs.all);
r->pipe->set_vertex_buffers(r->pipe, 4, r->vertex_bufs.all);
r->pipe->bind_vertex_elements_state(r->pipe, r->vertex_elems_state.individual.b);
r->textures.individual.ref[0] = r->past->texture;
r->textures.individual.ref[1] = r->future->texture;

4
src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.h
View file

@ -94,8 +94,8 @@ struct vl_mpeg12_mc_renderer
union
{
struct pipe_vertex_buffer all[3];
struct { struct pipe_vertex_buffer ycbcr, ref[2]; } individual;
struct pipe_vertex_buffer all[4];
struct { struct pipe_vertex_buffer rect, ycbcr, ref[2]; } individual;
} vertex_bufs;
struct pipe_surface *surface, *past, *future;