#include "vl_mpeg12_decoder.h"
#include "vl_defines.h"
-#define SCALE_FACTOR_16_TO_9 (32768.0f / 256.0f)
+#define SCALE_FACTOR_SNORM (32768.0f / 256.0f)
+#define SCALE_FACTOR_SSCALED (1.0f / 256.0f)
-static const unsigned const_empty_block_mask_420[3][2][2] = {
- { { 0x20, 0x10 }, { 0x08, 0x04 } },
- { { 0x02, 0x02 }, { 0x02, 0x02 } },
- { { 0x01, 0x01 }, { 0x01, 0x01 } }
+static const enum pipe_format const_zscan_source_formats[] = {
+ PIPE_FORMAT_R16_SNORM,
+ PIPE_FORMAT_R16_SSCALED
};
-static void
-map_buffers(struct vl_mpeg12_decoder *ctx, struct vl_mpeg12_buffer *buffer)
+static const unsigned num_zscan_source_formats =
+ sizeof(const_zscan_source_formats) / sizeof(enum pipe_format);
+
+static const enum pipe_format const_idct_source_formats[] = {
+ PIPE_FORMAT_R16G16B16A16_SNORM,
+ PIPE_FORMAT_R16G16B16A16_SSCALED
+};
+
+static const unsigned num_idct_source_formats =
+ sizeof(const_idct_source_formats) / sizeof(enum pipe_format);
+
+static const enum pipe_format const_idct_intermediate_formats[] = {
+ PIPE_FORMAT_R16G16B16A16_FLOAT,
+ PIPE_FORMAT_R16G16B16A16_SNORM,
+ PIPE_FORMAT_R16G16B16A16_SSCALED,
+ PIPE_FORMAT_R32G32B32A32_FLOAT
+};
+
+static const unsigned num_idct_intermediate_formats =
+ sizeof(const_idct_intermediate_formats) / sizeof(enum pipe_format);
+
+static const enum pipe_format const_mc_source_formats[] = {
+ PIPE_FORMAT_R16_SNORM,
+ PIPE_FORMAT_R16_SSCALED
+};
+
+static const unsigned num_mc_source_formats =
+ sizeof(const_mc_source_formats) / sizeof(enum pipe_format);
+
+static bool
+init_zscan_buffer(struct vl_mpeg12_buffer *buffer)
{
- struct pipe_sampler_view **sampler_views;
- struct pipe_resource *tex;
+ enum pipe_format formats[3];
+
+ struct pipe_sampler_view **source;
+ struct pipe_surface **destination;
+
+ struct vl_mpeg12_decoder *dec;
+
unsigned i;
- assert(ctx && buffer);
+ assert(buffer);
+
+ dec = (struct vl_mpeg12_decoder*)buffer->base.decoder;
+
+ formats[0] = formats[1] = formats[2] = dec->zscan_source_format;
+ buffer->zscan_source = vl_video_buffer_init(dec->base.context, dec->pipe,
+ dec->blocks_per_line * BLOCK_WIDTH * BLOCK_HEIGHT,
+ dec->max_blocks / dec->blocks_per_line,
+ 1, PIPE_VIDEO_CHROMA_FORMAT_444,
+ formats, PIPE_USAGE_STATIC);
+ if (!buffer->zscan_source)
+ goto error_source;
- if (ctx->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT)
- sampler_views = buffer->idct_source->get_sampler_views(buffer->idct_source);
+ source = buffer->zscan_source->get_sampler_views(buffer->zscan_source);
+ if (!source)
+ goto error_sampler;
+
+ if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT)
+ destination = buffer->idct_source->get_surfaces(buffer->idct_source);
else
- sampler_views = buffer->mc_source->get_sampler_views(buffer->mc_source);
- assert(sampler_views);
+ destination = buffer->mc_source->get_surfaces(buffer->mc_source);
- for (i = 0; i < VL_MAX_PLANES; ++i) {
- tex = sampler_views[i]->texture;
+ if (!destination)
+ goto error_surface;
- struct pipe_box rect =
- {
- 0, 0, 0,
- tex->width0,
- tex->height0,
- 1
- };
+ for (i = 0; i < VL_MAX_PLANES; ++i)
+ if (!vl_zscan_init_buffer(i == 0 ? &dec->zscan_y : &dec->zscan_c,
+ &buffer->zscan[i], source[i], destination[i]))
+ goto error_plane;
- buffer->tex_transfer[i] = ctx->pipe->get_transfer
- (
- ctx->pipe, tex,
- 0, PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
- &rect
- );
+ return true;
- buffer->texels[i] = ctx->pipe->transfer_map(ctx->pipe, buffer->tex_transfer[i]);
- }
+error_plane:
+ for (; i > 0; --i)
+ vl_zscan_cleanup_buffer(&buffer->zscan[i - 1]);
+
+error_surface:
+error_sampler:
+ buffer->zscan_source->destroy(buffer->zscan_source);
+
+error_source:
+ return false;
}
static void
-upload_block(struct vl_mpeg12_buffer *buffer, unsigned plane, unsigned x, unsigned y, short *block)
+cleanup_zscan_buffer(struct vl_mpeg12_buffer *buffer)
{
- unsigned tex_pitch;
- short *texels;
-
unsigned i;
assert(buffer);
- assert(block);
-
- tex_pitch = buffer->tex_transfer[plane]->stride / sizeof(short);
- texels = buffer->texels[plane] + y * tex_pitch * BLOCK_HEIGHT + x * BLOCK_WIDTH;
- for (i = 0; i < BLOCK_HEIGHT; ++i)
- memcpy(texels + i * tex_pitch, block + i * BLOCK_WIDTH, BLOCK_WIDTH * sizeof(short));
+ for (i = 0; i < VL_MAX_PLANES; ++i)
+ vl_zscan_cleanup_buffer(&buffer->zscan[i]);
+ buffer->zscan_source->destroy(buffer->zscan_source);
}
-static void
-upload_buffer(struct vl_mpeg12_decoder *ctx,
- struct vl_mpeg12_buffer *buffer,
- struct pipe_mpeg12_macroblock *mb)
+static bool
+init_idct_buffer(struct vl_mpeg12_buffer *buffer)
{
- short *blocks;
- unsigned tb, x, y;
+ enum pipe_format formats[3];
+
+ struct pipe_sampler_view **idct_source_sv, **idct_intermediate_sv;
+ struct pipe_surface **idct_surfaces;
+
+ struct vl_mpeg12_decoder *dec;
+
+ unsigned i;
- assert(ctx);
assert(buffer);
- assert(mb);
- blocks = mb->blocks;
+ dec = (struct vl_mpeg12_decoder*)buffer->base.decoder;
- for (y = 0; y < 2; ++y) {
- for (x = 0; x < 2; ++x, ++tb) {
- if (mb->cbp & (*ctx->empty_block_mask)[0][y][x]) {
- upload_block(buffer, 0, mb->mbx * 2 + x, mb->mby * 2 + y, blocks);
- blocks += BLOCK_WIDTH * BLOCK_HEIGHT;
- }
- }
- }
+ formats[0] = formats[1] = formats[2] = dec->idct_source_format;
+ buffer->idct_source = vl_video_buffer_init(dec->base.context, dec->pipe,
+ dec->base.width / 4, dec->base.height, 1,
+ dec->base.chroma_format,
+ formats, PIPE_USAGE_STATIC);
+ if (!buffer->idct_source)
+ goto error_source;
- /* TODO: Implement 422, 444 */
- assert(ctx->base.chroma_format == PIPE_VIDEO_CHROMA_FORMAT_420);
+ formats[0] = formats[1] = formats[2] = dec->idct_intermediate_format;
+ buffer->idct_intermediate = vl_video_buffer_init(dec->base.context, dec->pipe,
+ dec->base.width / dec->nr_of_idct_render_targets,
+ dec->base.height / 4, dec->nr_of_idct_render_targets,
+ dec->base.chroma_format,
+ formats, PIPE_USAGE_STATIC);
- for (tb = 1; tb < 3; ++tb) {
- if (mb->cbp & (*ctx->empty_block_mask)[tb][0][0]) {
- upload_block(buffer, tb, mb->mbx, mb->mby, blocks);
- blocks += BLOCK_WIDTH * BLOCK_HEIGHT;
- }
- }
-}
+ if (!buffer->idct_intermediate)
+ goto error_intermediate;
-static void
-unmap_buffers(struct vl_mpeg12_decoder *ctx, struct vl_mpeg12_buffer *buffer)
-{
- unsigned i;
+ idct_source_sv = buffer->idct_source->get_sampler_views(buffer->idct_source);
+ if (!idct_source_sv)
+ goto error_source_sv;
- assert(ctx && buffer);
+ idct_intermediate_sv = buffer->idct_intermediate->get_sampler_views(buffer->idct_intermediate);
+ if (!idct_intermediate_sv)
+ goto error_intermediate_sv;
- for (i = 0; i < VL_MAX_PLANES; ++i) {
- ctx->pipe->transfer_unmap(ctx->pipe, buffer->tex_transfer[i]);
- ctx->pipe->transfer_destroy(ctx->pipe, buffer->tex_transfer[i]);
- }
+ idct_surfaces = buffer->mc_source->get_surfaces(buffer->mc_source);
+ if (!idct_surfaces)
+ goto error_surfaces;
+
+ for (i = 0; i < 3; ++i)
+ if (!vl_idct_init_buffer(i == 0 ? &dec->idct_y : &dec->idct_c,
+ &buffer->idct[i], idct_source_sv[i],
+ idct_intermediate_sv[i], idct_surfaces[i]))
+ goto error_plane;
+
+ return true;
+
+error_plane:
+ for (; i > 0; --i)
+ vl_idct_cleanup_buffer(i == 1 ? &dec->idct_c : &dec->idct_y, &buffer->idct[i - 1]);
+
+error_surfaces:
+error_intermediate_sv:
+error_source_sv:
+ buffer->idct_intermediate->destroy(buffer->idct_intermediate);
+
+error_intermediate:
+ buffer->idct_source->destroy(buffer->idct_source);
+
+error_source:
+ return false;
}
static void
dec = (struct vl_mpeg12_decoder*)buf->base.decoder;
assert(dec);
- buf->idct_source->destroy(buf->idct_source);
- buf->idct_intermediate->destroy(buf->idct_intermediate);
vl_idct_cleanup_buffer(&dec->idct_y, &buf->idct[0]);
vl_idct_cleanup_buffer(&dec->idct_c, &buf->idct[1]);
vl_idct_cleanup_buffer(&dec->idct_c, &buf->idct[2]);
+ buf->idct_source->destroy(buf->idct_source);
+ buf->idct_intermediate->destroy(buf->idct_intermediate);
+}
+
+static bool
+init_mc_buffer(struct vl_mpeg12_buffer *buf)
+{
+ struct vl_mpeg12_decoder *dec;
+ enum pipe_format formats[3];
+ struct pipe_sampler_view **mc_source_sv;
+
+ assert(buf);
+
+ dec = (struct vl_mpeg12_decoder*)buf->base.decoder;
+ assert(dec);
+
+ formats[0] = formats[1] = formats[2] =dec->mc_source_format;
+ buf->mc_source = vl_video_buffer_init(dec->base.context, dec->pipe,
+ dec->base.width, dec->base.height, 1,
+ dec->base.chroma_format,
+ formats, PIPE_USAGE_STATIC);
+
+ if (!buf->mc_source)
+ goto error_mc_source;
+
+ mc_source_sv = buf->mc_source->get_sampler_views(buf->mc_source);
+ if (!mc_source_sv)
+ goto error_mc_source_sv;
+
+ if(!vl_mc_init_buffer(&dec->mc_y, &buf->mc[0], mc_source_sv[0]))
+ goto error_mc_y;
+
+ if(!vl_mc_init_buffer(&dec->mc_c, &buf->mc[1], mc_source_sv[1]))
+ goto error_mc_cb;
+
+ if(!vl_mc_init_buffer(&dec->mc_c, &buf->mc[2], mc_source_sv[2]))
+ goto error_mc_cr;
+
+ return true;
+
+error_mc_cr:
+ vl_mc_cleanup_buffer(&buf->mc[1]);
+
+error_mc_cb:
+ vl_mc_cleanup_buffer(&buf->mc[0]);
+
+error_mc_y:
+error_mc_source_sv:
+ buf->mc_source->destroy(buf->mc_source);
+
+error_mc_source:
+ return false;
+}
+
+static void
+cleanup_mc_buffer(struct vl_mpeg12_buffer *buf)
+{
+ unsigned i;
+
+ assert(buf);
+
+ for (i = 0; i < VL_MAX_PLANES; ++i)
+ vl_mc_cleanup_buffer(&buf->mc[i]);
+
+ buf->mc_source->destroy(buf->mc_source);
}
static void
{
struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)buffer;
struct vl_mpeg12_decoder *dec;
+
assert(buf);
dec = (struct vl_mpeg12_decoder*)buf->base.decoder;
assert(dec);
+ cleanup_zscan_buffer(buf);
+
if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT)
cleanup_idct_buffer(buf);
- buf->mc_source->destroy(buf->mc_source);
+ cleanup_mc_buffer(buf);
+
vl_vb_cleanup(&buf->vertex_stream);
- vl_mpeg12_mc_cleanup_buffer(&buf->mc[0]);
- vl_mpeg12_mc_cleanup_buffer(&buf->mc[1]);
- vl_mpeg12_mc_cleanup_buffer(&buf->mc[2]);
FREE(buf);
}
{
struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)buffer;
struct vl_mpeg12_decoder *dec;
+
+ struct pipe_sampler_view **sampler_views;
+ unsigned i;
+
assert(buf);
dec = (struct vl_mpeg12_decoder *)buf->base.decoder;
assert(dec);
vl_vb_map(&buf->vertex_stream, dec->pipe);
- map_buffers(dec, buf);
+
+ sampler_views = buf->zscan_source->get_sampler_views(buf->zscan_source);
+
+ assert(sampler_views);
+
+ for (i = 0; i < VL_MAX_PLANES; ++i) {
+ struct pipe_resource *tex = sampler_views[i]->texture;
+ struct pipe_box rect =
+ {
+ 0, 0, 0,
+ tex->width0,
+ tex->height0,
+ 1
+ };
+
+ buf->tex_transfer[i] = dec->pipe->get_transfer
+ (
+ dec->pipe, tex,
+ 0, PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
+ &rect
+ );
+
+ buf->texels[i] = dec->pipe->transfer_map(dec->pipe, buf->tex_transfer[i]);
+ }
}
-static void
-vl_mpeg12_buffer_add_macroblocks(struct pipe_video_decode_buffer *buffer,
- unsigned num_macroblocks,
- struct pipe_macroblock *macroblocks)
+static struct pipe_ycbcr_block *
+vl_mpeg12_buffer_get_ycbcr_stream(struct pipe_video_decode_buffer *buffer, int component)
{
- struct pipe_mpeg12_macroblock *mb = (struct pipe_mpeg12_macroblock*)macroblocks;
struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)buffer;
- struct vl_mpeg12_decoder *dec;
- unsigned i;
assert(buf);
- dec = (struct vl_mpeg12_decoder*)buf->base.decoder;
- assert(dec);
+ return vl_vb_get_ycbcr_stream(&buf->vertex_stream, component);
+}
+
+static short *
+vl_mpeg12_buffer_get_ycbcr_buffer(struct pipe_video_decode_buffer *buffer, int component)
+{
+ struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)buffer;
- assert(num_macroblocks);
- assert(macroblocks);
- assert(macroblocks->codec == PIPE_VIDEO_CODEC_MPEG12);
+ assert(buf);
+ assert(component < VL_MAX_PLANES);
- for ( i = 0; i < num_macroblocks; ++i ) {
- vl_vb_add_block(&buf->vertex_stream, &mb[i], dec->empty_block_mask);
- upload_buffer(dec, buf, &mb[i]);
- }
+ return buf->texels[component];
+}
+
+static unsigned
+vl_mpeg12_buffer_get_mv_stream_stride(struct pipe_video_decode_buffer *buffer)
+{
+ struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)buffer;
+
+ assert(buf);
+
+ return vl_vb_get_mv_stream_stride(&buf->vertex_stream);
+}
+
+static struct pipe_motionvector *
+vl_mpeg12_buffer_get_mv_stream(struct pipe_video_decode_buffer *buffer, int ref_frame)
+{
+ struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)buffer;
+
+ assert(buf);
+
+ return vl_vb_get_mv_stream(&buf->vertex_stream, ref_frame);
}
static void
{
struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)buffer;
struct vl_mpeg12_decoder *dec;
+ unsigned i;
+
assert(buf);
dec = (struct vl_mpeg12_decoder *)buf->base.decoder;
assert(dec);
vl_vb_unmap(&buf->vertex_stream, dec->pipe);
- unmap_buffers(dec, buf);
+
+ for (i = 0; i < VL_MAX_PLANES; ++i) {
+ dec->pipe->transfer_unmap(dec->pipe, buf->tex_transfer[i]);
+ dec->pipe->transfer_destroy(dec->pipe, buf->tex_transfer[i]);
+ }
}
static void
dec->pipe->bind_vs_state(dec->pipe, NULL);
dec->pipe->bind_fs_state(dec->pipe, NULL);
- dec->pipe->delete_blend_state(dec->pipe, dec->blend);
dec->pipe->delete_depth_stencil_alpha_state(dec->pipe, dec->dsa);
- vl_mpeg12_mc_renderer_cleanup(&dec->mc);
+ vl_mc_cleanup(&dec->mc_y);
+ vl_mc_cleanup(&dec->mc_c);
+
if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT) {
vl_idct_cleanup(&dec->idct_y);
vl_idct_cleanup(&dec->idct_c);
}
- dec->pipe->delete_vertex_elements_state(dec->pipe, dec->ves[0]);
- dec->pipe->delete_vertex_elements_state(dec->pipe, dec->ves[1]);
- dec->pipe->delete_vertex_elements_state(dec->pipe, dec->ves[2]);
- pipe_resource_reference(&dec->quads.buffer, NULL);
-
- FREE(dec);
-}
-
-static bool
-init_idct_buffer(struct vl_mpeg12_buffer *buffer)
-{
- const enum pipe_format idct_source_formats[3] = {
- PIPE_FORMAT_R16G16B16A16_SNORM,
- PIPE_FORMAT_R16G16B16A16_SNORM,
- PIPE_FORMAT_R16G16B16A16_SNORM
- };
-
- struct pipe_sampler_view **idct_source_sv, **idct_intermediate_sv;
- struct pipe_surface **idct_surfaces;
-
- struct vl_mpeg12_decoder *dec;
-
- unsigned i;
-
- assert(buffer);
-
- dec = (struct vl_mpeg12_decoder*)buffer->base.decoder;
-
- buffer->idct_source = vl_video_buffer_init(dec->base.context, dec->pipe,
- dec->base.width / 4, dec->base.height, 1,
- dec->base.chroma_format, 3,
- idct_source_formats,
- PIPE_USAGE_STREAM);
- if (!buffer->idct_source)
- goto error_source;
-
- buffer->idct_intermediate = vl_video_buffer_init(dec->base.context, dec->pipe,
- dec->base.width / dec->nr_of_idct_render_targets,
- dec->base.height / 4, dec->nr_of_idct_render_targets,
- dec->base.chroma_format, 3,
- idct_source_formats,
- PIPE_USAGE_STATIC);
- if (!buffer->idct_intermediate)
- goto error_intermediate;
+ vl_zscan_cleanup(&dec->zscan_y);
+ vl_zscan_cleanup(&dec->zscan_c);
- idct_source_sv = buffer->idct_source->get_sampler_views(buffer->idct_source);
- if (!idct_source_sv)
- goto error_source_sv;
-
- idct_intermediate_sv = buffer->idct_intermediate->get_sampler_views(buffer->idct_intermediate);
- if (!idct_intermediate_sv)
- goto error_intermediate_sv;
+ dec->pipe->delete_vertex_elements_state(dec->pipe, dec->ves_ycbcr);
+ dec->pipe->delete_vertex_elements_state(dec->pipe, dec->ves_mv);
- idct_surfaces = buffer->mc_source->get_surfaces(buffer->mc_source);
- if (!idct_surfaces)
- goto error_surfaces;
-
- for (i = 0; i < 3; ++i)
- if (!vl_idct_init_buffer(i == 0 ? &dec->idct_y : &dec->idct_c,
- &buffer->idct[i],
- idct_source_sv[i],
- idct_intermediate_sv[i],
- idct_surfaces[i]))
- goto error_plane;
-
- return true;
-
-error_plane:
- for (; i > 0; --i)
- vl_idct_cleanup_buffer(i == 1 ? &dec->idct_c : &dec->idct_y, &buffer->idct[i - 1]);
-
-error_surfaces:
-error_intermediate_sv:
-error_source_sv:
- buffer->idct_intermediate->destroy(buffer->idct_intermediate);
-
-error_intermediate:
- buffer->idct_source->destroy(buffer->idct_source);
+ pipe_resource_reference(&dec->quads.buffer, NULL);
+ pipe_resource_reference(&dec->pos.buffer, NULL);
-error_source:
- return false;
+ FREE(dec);
}
static struct pipe_video_decode_buffer *
vl_mpeg12_create_buffer(struct pipe_video_decoder *decoder)
{
- const enum pipe_format mc_source_formats[3] = {
- PIPE_FORMAT_R16_SNORM,
- PIPE_FORMAT_R16_SNORM,
- PIPE_FORMAT_R16_SNORM
- };
-
struct vl_mpeg12_decoder *dec = (struct vl_mpeg12_decoder*)decoder;
struct vl_mpeg12_buffer *buffer;
- struct pipe_sampler_view **mc_source_sv;
-
assert(dec);
buffer = CALLOC_STRUCT(vl_mpeg12_buffer);
buffer->base.decoder = decoder;
buffer->base.destroy = vl_mpeg12_buffer_destroy;
buffer->base.map = vl_mpeg12_buffer_map;
- buffer->base.add_macroblocks = vl_mpeg12_buffer_add_macroblocks;
+ buffer->base.get_ycbcr_stream = vl_mpeg12_buffer_get_ycbcr_stream;
+ buffer->base.get_ycbcr_buffer = vl_mpeg12_buffer_get_ycbcr_buffer;
+ buffer->base.get_mv_stream_stride = vl_mpeg12_buffer_get_mv_stream_stride;
+ buffer->base.get_mv_stream = vl_mpeg12_buffer_get_mv_stream;
buffer->base.unmap = vl_mpeg12_buffer_unmap;
- buffer->vertex_bufs.individual.quad.stride = dec->quads.stride;
- buffer->vertex_bufs.individual.quad.buffer_offset = dec->quads.buffer_offset;
- pipe_resource_reference(&buffer->vertex_bufs.individual.quad.buffer, dec->quads.buffer);
-
- buffer->vertex_bufs.individual.stream = vl_vb_init(&buffer->vertex_stream, dec->pipe,
- dec->base.width / MACROBLOCK_WIDTH *
- dec->base.height / MACROBLOCK_HEIGHT);
- if (!buffer->vertex_bufs.individual.stream.buffer)
- goto error_vertex_stream;
-
- buffer->mc_source = vl_video_buffer_init(dec->base.context, dec->pipe,
- dec->base.width, dec->base.height, 1,
- dec->base.chroma_format, 3,
- mc_source_formats,
- PIPE_USAGE_STATIC);
+ if (!vl_vb_init(&buffer->vertex_stream, dec->pipe,
+ dec->base.width / MACROBLOCK_WIDTH,
+ dec->base.height / MACROBLOCK_HEIGHT))
+ goto error_vertex_buffer;
- if (!buffer->mc_source)
- goto error_mc_source;
+ if (!init_mc_buffer(buffer))
+ goto error_mc;
if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT)
if (!init_idct_buffer(buffer))
goto error_idct;
- mc_source_sv = buffer->mc_source->get_sampler_views(buffer->mc_source);
- if (!mc_source_sv)
- goto error_mc_source_sv;
-
- if(!vl_mpeg12_mc_init_buffer(&dec->mc, &buffer->mc[0], mc_source_sv[0]))
- goto error_mc_y;
-
- if(!vl_mpeg12_mc_init_buffer(&dec->mc, &buffer->mc[1], mc_source_sv[1]))
- goto error_mc_cb;
-
- if(!vl_mpeg12_mc_init_buffer(&dec->mc, &buffer->mc[2], mc_source_sv[2]))
- goto error_mc_cr;
+ if (!init_zscan_buffer(buffer))
+ goto error_zscan;
return &buffer->base;
-error_mc_cr:
- vl_mpeg12_mc_cleanup_buffer(&buffer->mc[1]);
-
-error_mc_cb:
- vl_mpeg12_mc_cleanup_buffer(&buffer->mc[0]);
-
-error_mc_y:
-error_mc_source_sv:
+error_zscan:
if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT)
cleanup_idct_buffer(buffer);
error_idct:
- buffer->mc_source->destroy(buffer->mc_source);
+ cleanup_mc_buffer(buffer);
-error_mc_source:
+error_mc:
vl_vb_cleanup(&buffer->vertex_stream);
-error_vertex_stream:
+error_vertex_buffer:
FREE(buffer);
return NULL;
}
static void
vl_mpeg12_decoder_flush_buffer(struct pipe_video_decode_buffer *buffer,
+ unsigned num_ycbcr_blocks[3],
struct pipe_video_buffer *refs[2],
struct pipe_video_buffer *dst,
struct pipe_fence_handle **fence)
struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer *)buffer;
struct vl_mpeg12_decoder *dec;
- struct pipe_sampler_view **sv_past;
- struct pipe_sampler_view **sv_future;
+ struct pipe_sampler_view **sv[2];
struct pipe_surface **surfaces;
- struct pipe_sampler_view *sv_refs[2];
- unsigned ne_start, ne_num, e_start, e_num;
- unsigned i;
+ struct pipe_vertex_buffer vb[3];
+
+ unsigned i, j;
assert(buf);
dec = (struct vl_mpeg12_decoder *)buf->base.decoder;
assert(dec);
- sv_past = refs[0] ? refs[0]->get_sampler_views(refs[0]) : NULL;
- sv_future = refs[1] ? refs[1]->get_sampler_views(refs[1]) : NULL;
+ for (i = 0; i < 2; ++i)
+ sv[i] = refs[i] ? refs[i]->get_sampler_views(refs[i]) : NULL;
surfaces = dst->get_surfaces(dst);
- vl_vb_restart(&buf->vertex_stream, &ne_start, &ne_num, &e_start, &e_num);
-
- dec->pipe->set_vertex_buffers(dec->pipe, 2, buf->vertex_bufs.all);
- dec->pipe->bind_blend_state(dec->pipe, dec->blend);
+ vb[0] = dec->quads;
+ vb[1] = dec->pos;
+ dec->pipe->bind_vertex_elements_state(dec->pipe, dec->ves_mv);
for (i = 0; i < VL_MAX_PLANES; ++i) {
- dec->pipe->bind_vertex_elements_state(dec->pipe, dec->ves[i]);
+ vl_mc_set_surface(&buf->mc[i], surfaces[i]);
- if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT)
- vl_idct_flush(i == 0 ? &dec->idct_y : &dec->idct_c, &buf->idct[i], ne_num);
+ for (j = 0; j < 2; ++j) {
+ if (sv[j] == NULL) continue;
- sv_refs[0] = sv_past ? sv_past[i] : NULL;
- sv_refs[1] = sv_future ? sv_future[i] : NULL;
+ vb[2] = vl_vb_get_mv(&buf->vertex_stream, j);;
+ dec->pipe->set_vertex_buffers(dec->pipe, 3, vb);
- vl_mpeg12_mc_renderer_flush(&dec->mc, &buf->mc[i], surfaces[i], sv_refs,
- ne_start, ne_num, e_start, e_num, fence);
+ vl_mc_render_ref(&buf->mc[i], sv[j][i]);
+ }
}
-}
-static void
-vl_mpeg12_decoder_clear_buffer(struct pipe_video_decode_buffer *buffer)
-{
- struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer *)buffer;
- unsigned ne_start, ne_num, e_start, e_num;
+ dec->pipe->bind_vertex_elements_state(dec->pipe, dec->ves_ycbcr);
+ for (i = 0; i < VL_MAX_PLANES; ++i) {
+ if (num_ycbcr_blocks[i] == 0) continue;
- assert(buf);
+ vb[1] = vl_vb_get_ycbcr(&buf->vertex_stream, i);
+ dec->pipe->set_vertex_buffers(dec->pipe, 2, vb);
+
+ vl_zscan_render(&buf->zscan[i] , num_ycbcr_blocks[i]);
- vl_vb_restart(&buf->vertex_stream, &ne_start, &ne_num, &e_start, &e_num);
+ if (dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT)
+ vl_idct_flush(i == 0 ? &dec->idct_y : &dec->idct_c, &buf->idct[i], num_ycbcr_blocks[i]);
+
+ vl_mc_render_ycbcr(&buf->mc[i], num_ycbcr_blocks[i]);
+ }
+
+ dec->pipe->flush(dec->pipe, fence);
}
static bool
init_pipe_state(struct vl_mpeg12_decoder *dec)
{
- struct pipe_blend_state blend;
struct pipe_depth_stencil_alpha_state dsa;
unsigned i;
assert(dec);
- memset(&blend, 0, sizeof blend);
-
- blend.independent_blend_enable = 0;
- blend.rt[0].blend_enable = 0;
- blend.rt[0].rgb_func = PIPE_BLEND_ADD;
- blend.rt[0].rgb_src_factor = PIPE_BLENDFACTOR_ONE;
- blend.rt[0].rgb_dst_factor = PIPE_BLENDFACTOR_ONE;
- blend.rt[0].alpha_func = PIPE_BLEND_ADD;
- blend.rt[0].alpha_src_factor = PIPE_BLENDFACTOR_ONE;
- blend.rt[0].alpha_dst_factor = PIPE_BLENDFACTOR_ONE;
- blend.logicop_enable = 0;
- blend.logicop_func = PIPE_LOGICOP_CLEAR;
- /* Needed to allow color writes to FB, even if blending disabled */
- blend.rt[0].colormask = PIPE_MASK_RGBA;
- blend.dither = 0;
- dec->blend = dec->pipe->create_blend_state(dec->pipe, &blend);
-
memset(&dsa, 0, sizeof dsa);
dsa.depth.enabled = 0;
dsa.depth.writemask = 0;
return true;
}
+static enum pipe_format
+find_first_supported_format(struct vl_mpeg12_decoder *dec,
+ const enum pipe_format formats[],
+ unsigned num_formats,
+ enum pipe_texture_target target)
+{
+ struct pipe_screen *screen;
+ unsigned i;
+
+ assert(dec);
+
+ screen = dec->pipe->screen;
+
+ for (i = 0; i < num_formats; ++i)
+ if (screen->is_format_supported(dec->pipe->screen, formats[i], target, 1,
+ PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET))
+ return formats[i];
+
+ return PIPE_FORMAT_NONE;
+}
+
static bool
-init_idct(struct vl_mpeg12_decoder *dec, unsigned buffer_width, unsigned buffer_height)
+init_zscan(struct vl_mpeg12_decoder *dec)
{
- unsigned chroma_width, chroma_height, chroma_blocks_x, chroma_blocks_y;
- struct pipe_sampler_view *idct_matrix;
+ struct pipe_sampler_view *layout;
- dec->nr_of_idct_render_targets = 4;
+ unsigned num_channels;
- if (!(idct_matrix = vl_idct_upload_matrix(dec->pipe, sqrt(SCALE_FACTOR_16_TO_9))))
- goto error_idct_matrix;
+ assert(dec);
- if (!vl_idct_init(&dec->idct_y, dec->pipe, buffer_width, buffer_height,
- 2, 2, dec->nr_of_idct_render_targets, idct_matrix))
- goto error_idct_y;
+ dec->blocks_per_line = 4;
+ dec->max_blocks =
+ (dec->base.width * dec->base.height) /
+ (BLOCK_WIDTH * BLOCK_HEIGHT);
- if (dec->base.chroma_format == PIPE_VIDEO_CHROMA_FORMAT_420) {
- chroma_width = buffer_width / 2;
- chroma_height = buffer_height / 2;
- chroma_blocks_x = 1;
- chroma_blocks_y = 1;
- } else if (dec->base.chroma_format == PIPE_VIDEO_CHROMA_FORMAT_422) {
- chroma_width = buffer_width;
- chroma_height = buffer_height / 2;
- chroma_blocks_x = 2;
- chroma_blocks_y = 1;
- } else {
- chroma_width = buffer_width;
- chroma_height = buffer_height;
- chroma_blocks_x = 2;
- chroma_blocks_y = 2;
+ dec->zscan_source_format = find_first_supported_format(dec, const_zscan_source_formats,
+ num_zscan_source_formats, PIPE_TEXTURE_2D);
+
+ if (dec->zscan_source_format == PIPE_FORMAT_NONE)
+ return false;
+
+ layout = vl_zscan_linear(dec->pipe, dec->blocks_per_line);
+
+ num_channels = dec->base.entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT ? 4 : 1;
+
+ if (!vl_zscan_init(&dec->zscan_y, dec->pipe, dec->base.width, dec->base.height,
+ dec->blocks_per_line, dec->max_blocks, num_channels))
+ return false;
+
+ vl_zscan_set_layout(&dec->zscan_y, layout);
+
+ if (!vl_zscan_init(&dec->zscan_c, dec->pipe, dec->chroma_width, dec->chroma_height,
+ dec->blocks_per_line, dec->max_blocks, num_channels))
+ return false;
+
+ vl_zscan_set_layout(&dec->zscan_c, layout);
+
+ return true;
+}
+
+static bool
+init_idct(struct vl_mpeg12_decoder *dec)
+{
+ struct pipe_sampler_view *matrix, *transpose = NULL;
+ float matrix_scale, transpose_scale;
+
+ dec->nr_of_idct_render_targets = dec->pipe->screen->get_param(dec->pipe->screen, PIPE_CAP_MAX_RENDER_TARGETS);
+
+ // more than 4 render targets usually doesn't makes any seens
+ dec->nr_of_idct_render_targets = MIN2(dec->nr_of_idct_render_targets, 4);
+
+ dec->idct_source_format = find_first_supported_format(dec, const_idct_source_formats,
+ num_idct_source_formats, PIPE_TEXTURE_2D);
+
+ if (dec->idct_source_format == PIPE_FORMAT_NONE)
+ return false;
+
+ dec->idct_intermediate_format = find_first_supported_format(dec, const_idct_intermediate_formats,
+ num_idct_intermediate_formats, PIPE_TEXTURE_3D);
+
+ if (dec->idct_intermediate_format == PIPE_FORMAT_NONE)
+ return false;
+
+ switch (dec->idct_source_format) {
+ case PIPE_FORMAT_R16G16B16A16_SSCALED:
+ matrix_scale = SCALE_FACTOR_SSCALED;
+ break;
+
+ case PIPE_FORMAT_R16G16B16A16_SNORM:
+ matrix_scale = SCALE_FACTOR_SNORM;
+ break;
+
+ default:
+ assert(0);
+ return false;
}
- if(!vl_idct_init(&dec->idct_c, dec->pipe, chroma_width, chroma_height,
- chroma_blocks_x, chroma_blocks_y,
- dec->nr_of_idct_render_targets, idct_matrix))
- goto error_idct_c;
+ if (dec->idct_intermediate_format == PIPE_FORMAT_R16G16B16A16_FLOAT ||
+ dec->idct_intermediate_format == PIPE_FORMAT_R32G32B32A32_FLOAT)
+ transpose_scale = 1.0f;
+ else
+ transpose_scale = matrix_scale = sqrt(matrix_scale);
+
+ if (dec->mc_source_format == PIPE_FORMAT_R16_SSCALED)
+ transpose_scale /= SCALE_FACTOR_SSCALED;
+
+ if (!(matrix = vl_idct_upload_matrix(dec->pipe, matrix_scale)))
+ goto error_matrix;
+
+ if (matrix_scale != transpose_scale) {
+ if (!(transpose = vl_idct_upload_matrix(dec->pipe, transpose_scale)))
+ goto error_transpose;
+ } else
+ pipe_sampler_view_reference(&transpose, matrix);
- pipe_sampler_view_reference(&idct_matrix, NULL);
+ if (!vl_idct_init(&dec->idct_y, dec->pipe, dec->base.width, dec->base.height,
+ dec->nr_of_idct_render_targets, matrix, transpose))
+ goto error_y;
+
+ if(!vl_idct_init(&dec->idct_c, dec->pipe, dec->chroma_width, dec->chroma_height,
+ dec->nr_of_idct_render_targets, matrix, transpose))
+ goto error_c;
+
+ pipe_sampler_view_reference(&matrix, NULL);
+ pipe_sampler_view_reference(&transpose, NULL);
return true;
-error_idct_c:
+error_c:
vl_idct_cleanup(&dec->idct_y);
-error_idct_y:
- pipe_sampler_view_reference(&idct_matrix, NULL);
+error_y:
+ pipe_sampler_view_reference(&transpose, NULL);
+
+error_transpose:
+ pipe_sampler_view_reference(&matrix, NULL);
-error_idct_matrix:
+error_matrix:
return false;
}
unsigned width, unsigned height)
{
struct vl_mpeg12_decoder *dec;
- unsigned i;
+ float mc_scale;
assert(u_reduce_video_profile(profile) == PIPE_VIDEO_CODEC_MPEG12);
dec->base.destroy = vl_mpeg12_destroy;
dec->base.create_buffer = vl_mpeg12_create_buffer;
dec->base.flush_buffer = vl_mpeg12_decoder_flush_buffer;
- dec->base.clear_buffer = vl_mpeg12_decoder_clear_buffer;
+
+ dec->base.width = align(width, MACROBLOCK_WIDTH);
+ dec->base.height = align(height, MACROBLOCK_HEIGHT);
dec->pipe = pipe;
- dec->quads = vl_vb_upload_quads(dec->pipe, 2, 2);
- for (i = 0; i < VL_MAX_PLANES; ++i)
- dec->ves[i] = vl_vb_get_elems_state(dec->pipe, i);
+ dec->quads = vl_vb_upload_quads(dec->pipe);
+ dec->pos = vl_vb_upload_pos(
+ dec->pipe,
+ dec->base.width / MACROBLOCK_WIDTH,
+ dec->base.height / MACROBLOCK_HEIGHT
+ );
- dec->base.width = align(width, MACROBLOCK_WIDTH);
- dec->base.height = align(height, MACROBLOCK_HEIGHT);
+ dec->ves_ycbcr = vl_vb_get_ves_ycbcr(dec->pipe);
+ dec->ves_mv = vl_vb_get_ves_mv(dec->pipe);
/* TODO: Implement 422, 444 */
assert(dec->base.chroma_format == PIPE_VIDEO_CHROMA_FORMAT_420);
- dec->empty_block_mask = &const_empty_block_mask_420;
- if (entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT)
- if (!init_idct(dec, dec->base.width, dec->base.height))
+ dec->mc_source_format = find_first_supported_format(dec, const_mc_source_formats,
+ num_mc_source_formats, PIPE_TEXTURE_3D);
+
+ if (dec->mc_source_format == PIPE_FORMAT_NONE)
+ return NULL;
+
+ if (dec->base.chroma_format == PIPE_VIDEO_CHROMA_FORMAT_420) {
+ dec->chroma_width = dec->base.width / 2;
+ dec->chroma_height = dec->base.height / 2;
+ } else if (dec->base.chroma_format == PIPE_VIDEO_CHROMA_FORMAT_422) {
+ dec->chroma_width = dec->base.width;
+ dec->chroma_height = dec->base.height / 2;
+ } else {
+ dec->chroma_width = dec->base.width;
+ dec->chroma_height = dec->base.height;
+ }
+
+ if (!init_zscan(dec))
+ goto error_zscan;
+
+ if (entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT) {
+ if (!init_idct(dec))
goto error_idct;
+ if (dec->mc_source_format == PIPE_FORMAT_R16_SSCALED)
+ mc_scale = SCALE_FACTOR_SSCALED;
+ else
+ mc_scale = 1.0f;
+ } else {
+ switch (dec->mc_source_format) {
+ case PIPE_FORMAT_R16_SNORM:
+ mc_scale = SCALE_FACTOR_SNORM;
+ break;
+
+ case PIPE_FORMAT_R16_SSCALED:
+ mc_scale = SCALE_FACTOR_SSCALED;
+ break;
+
+ default:
+ assert(0);
+ return NULL;
+ }
+ }
- if (!vl_mpeg12_mc_renderer_init(&dec->mc, dec->pipe, dec->base.width, dec->base.height,
- entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT ? 1.0f : SCALE_FACTOR_16_TO_9))
- goto error_mc;
+ if (!vl_mc_init(&dec->mc_y, dec->pipe, dec->base.width, dec->base.height, MACROBLOCK_HEIGHT, mc_scale))
+ goto error_mc_y;
+
+ // TODO
+ if (!vl_mc_init(&dec->mc_c, dec->pipe, dec->base.width, dec->base.height, BLOCK_HEIGHT, mc_scale))
+ goto error_mc_c;
if (!init_pipe_state(dec))
goto error_pipe_state;
return &dec->base;
error_pipe_state:
- vl_mpeg12_mc_renderer_cleanup(&dec->mc);
+ vl_mc_cleanup(&dec->mc_c);
-error_mc:
+error_mc_c:
+ vl_mc_cleanup(&dec->mc_y);
+
+error_mc_y:
if (entrypoint <= PIPE_VIDEO_ENTRYPOINT_IDCT) {
vl_idct_cleanup(&dec->idct_y);
vl_idct_cleanup(&dec->idct_c);
}
error_idct:
+ vl_zscan_cleanup(&dec->zscan_y);
+ vl_zscan_cleanup(&dec->zscan_c);
+
+error_zscan:
FREE(dec);
return NULL;
}
projects / mesa.git / blobdiff