From: Christian König Date: Sat, 19 Mar 2011 00:02:40 +0000 (+0100) Subject: Merge remote branch 'origin/master' into pipe-video X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=2bf95c519e755146704f4942b1703d47d18bfeaa;p=mesa.git Merge remote branch 'origin/master' into pipe-video Conflicts: src/gallium/drivers/r600/r600_asm.c src/gallium/tests/unit/SConscript --- 2bf95c519e755146704f4942b1703d47d18bfeaa diff --cc src/gallium/auxiliary/vl/vl_compositor.c index d1ba5faf788,00000000000..c54e5476f3a mode 100644,000000..100644 --- a/src/gallium/auxiliary/vl/vl_compositor.c +++ b/src/gallium/auxiliary/vl/vl_compositor.c @@@ -1,640 -1,0 +1,640 @@@ +/************************************************************************** + * + * Copyright 2009 Younes Manton. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "vl_compositor.h" +#include "util/u_draw.h" +#include +#include +#include +#include +#include +#include +#include +#include +#include "vl_csc.h" + +struct vertex_shader_consts +{ + struct vertex4f dst_scale; + struct vertex4f dst_trans; + struct vertex4f src_scale; + struct vertex4f src_trans; +}; + +struct fragment_shader_consts +{ + float matrix[16]; +}; + +static bool +u_video_rects_equal(struct pipe_video_rect *a, struct pipe_video_rect *b) +{ + assert(a && b); + + if (a->x != b->x) + return false; + if (a->y != b->y) + return false; + if (a->w != b->w) + return false; + if (a->h != b->h) + return false; + + return true; +} + +static bool +create_vert_shader(struct vl_compositor *c) +{ + struct ureg_program *shader; + struct ureg_src vpos, vtex; + struct ureg_dst o_vpos, o_vtex; + + shader = ureg_create(TGSI_PROCESSOR_VERTEX); + if (!shader) + return false; + + vpos = ureg_DECL_vs_input(shader, 0); + vtex = ureg_DECL_vs_input(shader, 1); + o_vpos = ureg_DECL_output(shader, TGSI_SEMANTIC_POSITION, 0); + o_vtex = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, 1); + + /* + * o_vpos = vpos + * o_vtex = vtex + */ + ureg_MOV(shader, o_vpos, vpos); + ureg_MOV(shader, o_vtex, vtex); + + ureg_END(shader); + + c->vertex_shader = ureg_create_shader_and_destroy(shader, c->pipe); + if (!c->vertex_shader) + return false; + + return true; +} + +static bool +create_frag_shader_ycbcr_2_rgb(struct vl_compositor *c) +{ + struct ureg_program *shader; + struct ureg_src tc; + struct ureg_src csc[4]; + struct ureg_src sampler; + struct ureg_dst texel; + struct ureg_dst fragment; + unsigned i; + + shader = ureg_create(TGSI_PROCESSOR_FRAGMENT); + if (!shader) + return false; + + tc = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, 1, TGSI_INTERPOLATE_LINEAR); + for (i = 0; i < 4; ++i) + csc[i] = ureg_DECL_constant(shader, i); + sampler = ureg_DECL_sampler(shader, 0); + texel = ureg_DECL_temporary(shader); + fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0); + + /* + * texel = tex(tc, sampler) + * fragment = csc * texel + */ + ureg_TEX(shader, texel, TGSI_TEXTURE_2D, tc, sampler); + for (i = 0; i < 4; ++i) + ureg_DP4(shader, ureg_writemask(fragment, TGSI_WRITEMASK_X << i), csc[i], ureg_src(texel)); + + ureg_release_temporary(shader, texel); + ureg_END(shader); + + c->fragment_shader.ycbcr_2_rgb = ureg_create_shader_and_destroy(shader, c->pipe); + if (!c->fragment_shader.ycbcr_2_rgb) + return false; + + return true; +} + +static bool +create_frag_shader_rgb_2_rgb(struct vl_compositor *c) +{ + struct ureg_program *shader; + struct ureg_src tc; + struct ureg_src sampler; + struct ureg_dst fragment; + + shader = ureg_create(TGSI_PROCESSOR_FRAGMENT); + if (!shader) + return false; + + tc = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, 1, TGSI_INTERPOLATE_LINEAR); + sampler = ureg_DECL_sampler(shader, 0); + fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0); + + /* + * fragment = tex(tc, sampler) + */ + ureg_TEX(shader, fragment, TGSI_TEXTURE_2D, tc, sampler); + ureg_END(shader); + + c->fragment_shader.rgb_2_rgb = ureg_create_shader_and_destroy(shader, c->pipe); + if (!c->fragment_shader.rgb_2_rgb) + return false; + + return true; +} + +static bool +init_pipe_state(struct vl_compositor *c) +{ + struct pipe_sampler_state sampler; + + assert(c); + + c->fb_state.nr_cbufs = 1; + c->fb_state.zsbuf = NULL; + + memset(&sampler, 0, sizeof(sampler)); + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR; + sampler.compare_mode = PIPE_TEX_COMPARE_NONE; + sampler.compare_func = PIPE_FUNC_ALWAYS; + sampler.normalized_coords = 1; + /*sampler.lod_bias = ;*/ + /*sampler.min_lod = ;*/ + /*sampler.max_lod = ;*/ + /*sampler.border_color[i] = ;*/ + /*sampler.max_anisotropy = ;*/ + c->sampler = c->pipe->create_sampler_state(c->pipe, &sampler); + + return true; +} + +static void cleanup_pipe_state(struct vl_compositor *c) +{ + assert(c); + + c->pipe->delete_sampler_state(c->pipe, c->sampler); +} + +static bool +init_shaders(struct vl_compositor *c) +{ + assert(c); + + if (!create_vert_shader(c)) { + debug_printf("Unable to create vertex shader.\n"); + return false; + } + if (!create_frag_shader_ycbcr_2_rgb(c)) { + debug_printf("Unable to create YCbCr-to-RGB fragment shader.\n"); + return false; + } + if (!create_frag_shader_rgb_2_rgb(c)) { + debug_printf("Unable to create RGB-to-RGB fragment shader.\n"); + return false; + } + + return true; +} + +static void cleanup_shaders(struct vl_compositor *c) +{ + assert(c); + + c->pipe->delete_vs_state(c->pipe, c->vertex_shader); + c->pipe->delete_fs_state(c->pipe, c->fragment_shader.ycbcr_2_rgb); + c->pipe->delete_fs_state(c->pipe, c->fragment_shader.rgb_2_rgb); +} + +static bool +init_buffers(struct vl_compositor *c) +{ + struct fragment_shader_consts fsc; + struct pipe_vertex_element vertex_elems[2]; + + assert(c); + + /* + * Create our vertex buffer and vertex buffer elements + */ + c->vertex_buf.stride = sizeof(struct vertex4f); + c->vertex_buf.buffer_offset = 0; + /* XXX: Create with DYNAMIC or STREAM */ + c->vertex_buf.buffer = pipe_buffer_create + ( + c->pipe->screen, + PIPE_BIND_VERTEX_BUFFER, + PIPE_USAGE_STATIC, + sizeof(struct vertex4f) * (VL_COMPOSITOR_MAX_LAYERS + 2) * 6 + ); + + 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; + vertex_elems[1].src_offset = sizeof(struct vertex2f); + vertex_elems[1].instance_divisor = 0; + vertex_elems[1].vertex_buffer_index = 0; + vertex_elems[1].src_format = PIPE_FORMAT_R32G32_FLOAT; + c->vertex_elems_state = c->pipe->create_vertex_elements_state(c->pipe, 2, vertex_elems); + + /* + * Create our fragment shader's constant buffer + * Const buffer contains the color conversion matrix and bias vectors + */ + /* XXX: Create with IMMUTABLE/STATIC... although it does change every once in a long while... */ + c->fs_const_buf = pipe_buffer_create + ( + c->pipe->screen, + PIPE_BIND_CONSTANT_BUFFER, + PIPE_USAGE_STATIC, + 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; +} + +static void +cleanup_buffers(struct vl_compositor *c) +{ + assert(c); + + c->pipe->delete_vertex_elements_state(c->pipe, c->vertex_elems_state); + pipe_resource_reference(&c->vertex_buf.buffer, NULL); + pipe_resource_reference(&c->fs_const_buf, NULL); +} + +static void +texview_map_delete(const struct keymap *map, + const void *key, void *data, + void *user) +{ + struct pipe_sampler_view *sv = (struct pipe_sampler_view*)data; + + assert(map); + assert(key); + assert(data); + assert(user); + + pipe_sampler_view_reference(&sv, NULL); +} + +bool vl_compositor_init(struct vl_compositor *compositor, struct pipe_context *pipe) +{ + unsigned i; + + assert(compositor); + + memset(compositor, 0, sizeof(struct vl_compositor)); + + compositor->pipe = pipe; + + compositor->texview_map = util_new_keymap(sizeof(struct pipe_surface*), -1, + texview_map_delete); + if (!compositor->texview_map) + return false; + + if (!init_pipe_state(compositor)) { + util_delete_keymap(compositor->texview_map, compositor->pipe); + return false; + } + if (!init_shaders(compositor)) { + util_delete_keymap(compositor->texview_map, compositor->pipe); + cleanup_pipe_state(compositor); + return false; + } + if (!init_buffers(compositor)) { + util_delete_keymap(compositor->texview_map, compositor->pipe); + cleanup_shaders(compositor); + cleanup_pipe_state(compositor); + return false; + } + + compositor->fb_state.width = 0; + compositor->fb_state.height = 0; + compositor->bg = NULL; + compositor->dirty_bg = false; + for (i = 0; i < VL_COMPOSITOR_MAX_LAYERS; ++i) + compositor->layers[i] = NULL; + compositor->dirty_layers = 0; + + return true; +} + +void vl_compositor_cleanup(struct vl_compositor *compositor) +{ + assert(compositor); + + util_delete_keymap(compositor->texview_map, compositor->pipe); + cleanup_buffers(compositor); + cleanup_shaders(compositor); + cleanup_pipe_state(compositor); +} + +void vl_compositor_set_background(struct vl_compositor *compositor, + struct pipe_surface *bg, struct pipe_video_rect *bg_src_rect) +{ + assert(compositor); + assert((bg && bg_src_rect) || (!bg && !bg_src_rect)); + + if (compositor->bg != bg || + !u_video_rects_equal(&compositor->bg_src_rect, bg_src_rect)) { + pipe_surface_reference(&compositor->bg, bg); + /*if (!u_video_rects_equal(&compositor->bg_src_rect, bg_src_rect))*/ + compositor->bg_src_rect = *bg_src_rect; + compositor->dirty_bg = true; + } +} + +void vl_compositor_set_layers(struct vl_compositor *compositor, + struct pipe_surface *layers[], + struct pipe_video_rect *src_rects[], + struct pipe_video_rect *dst_rects[], + unsigned num_layers) +{ + unsigned i; + + assert(compositor); + assert(num_layers <= VL_COMPOSITOR_MAX_LAYERS); + + for (i = 0; i < num_layers; ++i) + { + assert((layers[i] && src_rects[i] && dst_rects[i]) || + (!layers[i] && !src_rects[i] && !dst_rects[i])); + + if (compositor->layers[i] != layers[i] || + !u_video_rects_equal(&compositor->layer_src_rects[i], src_rects[i]) || + !u_video_rects_equal(&compositor->layer_dst_rects[i], dst_rects[i])) + { + pipe_surface_reference(&compositor->layers[i], layers[i]); + /*if (!u_video_rects_equal(&compositor->layer_src_rects[i], src_rects[i]))*/ + compositor->layer_src_rects[i] = *src_rects[i]; + /*if (!u_video_rects_equal(&compositor->layer_dst_rects[i], dst_rects[i]))*/ + compositor->layer_dst_rects[i] = *dst_rects[i]; + compositor->dirty_layers |= 1 << i; + } + + if (layers[i]) + compositor->dirty_layers |= 1 << i; + } + + for (; i < VL_COMPOSITOR_MAX_LAYERS; ++i) + pipe_surface_reference(&compositor->layers[i], NULL); +} + +static void gen_rect_verts(unsigned pos, + struct pipe_video_rect *src_rect, + struct vertex2f *src_inv_size, + struct pipe_video_rect *dst_rect, + struct vertex2f *dst_inv_size, + struct vertex4f *vb) +{ + assert(pos < VL_COMPOSITOR_MAX_LAYERS + 2); + assert(src_rect); + assert(src_inv_size); + assert((dst_rect && dst_inv_size) /*|| (!dst_rect && !dst_inv_size)*/); + assert(vb); + + vb[pos * 6 + 0].x = dst_rect->x * dst_inv_size->x; + vb[pos * 6 + 0].y = dst_rect->y * dst_inv_size->y; + vb[pos * 6 + 0].z = src_rect->x * src_inv_size->x; + vb[pos * 6 + 0].w = src_rect->y * src_inv_size->y; + + vb[pos * 6 + 1].x = dst_rect->x * dst_inv_size->x; + vb[pos * 6 + 1].y = (dst_rect->y + dst_rect->h) * dst_inv_size->y; + vb[pos * 6 + 1].z = src_rect->x * src_inv_size->x; + vb[pos * 6 + 1].w = (src_rect->y + src_rect->h) * src_inv_size->y; + + vb[pos * 6 + 2].x = (dst_rect->x + dst_rect->w) * dst_inv_size->x; + vb[pos * 6 + 2].y = dst_rect->y * dst_inv_size->y; + vb[pos * 6 + 2].z = (src_rect->x + src_rect->w) * src_inv_size->x; + vb[pos * 6 + 2].w = src_rect->y * src_inv_size->y; + + vb[pos * 6 + 3].x = (dst_rect->x + dst_rect->w) * dst_inv_size->x; + vb[pos * 6 + 3].y = dst_rect->y * dst_inv_size->y; + vb[pos * 6 + 3].z = (src_rect->x + src_rect->w) * src_inv_size->x; + vb[pos * 6 + 3].w = src_rect->y * src_inv_size->y; + + vb[pos * 6 + 4].x = dst_rect->x * dst_inv_size->x; + vb[pos * 6 + 4].y = (dst_rect->y + dst_rect->h) * dst_inv_size->y; + vb[pos * 6 + 4].z = src_rect->x * src_inv_size->x; + vb[pos * 6 + 4].w = (src_rect->y + src_rect->h) * src_inv_size->y; + + vb[pos * 6 + 5].x = (dst_rect->x + dst_rect->w) * dst_inv_size->x; + vb[pos * 6 + 5].y = (dst_rect->y + dst_rect->h) * dst_inv_size->y; + vb[pos * 6 + 5].z = (src_rect->x + src_rect->w) * src_inv_size->x; + vb[pos * 6 + 5].w = (src_rect->y + src_rect->h) * src_inv_size->y; +} + +static unsigned gen_data(struct vl_compositor *c, + struct pipe_surface *src_surface, + struct pipe_video_rect *src_rect, + struct pipe_video_rect *dst_rect, + struct pipe_surface **textures, + void **frag_shaders) +{ + void *vb; + struct pipe_transfer *buf_transfer; + unsigned num_rects = 0; + unsigned i; + + assert(c); + assert(src_surface); + assert(src_rect); + assert(dst_rect); + assert(textures); + + vb = pipe_buffer_map(c->pipe, c->vertex_buf.buffer, + PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD, + &buf_transfer); + + if (!vb) + return 0; + + if (c->dirty_bg) { + struct vertex2f bg_inv_size = {1.0f / c->bg->width, 1.0f / c->bg->height}; + gen_rect_verts(num_rects, &c->bg_src_rect, &bg_inv_size, NULL, NULL, vb); + textures[num_rects] = c->bg; + /* XXX: Hack */ + frag_shaders[num_rects] = c->fragment_shader.rgb_2_rgb; + ++num_rects; + c->dirty_bg = false; + } + + { + struct vertex2f src_inv_size = { 1.0f / src_surface->width, 1.0f / src_surface->height}; + gen_rect_verts(num_rects, src_rect, &src_inv_size, dst_rect, &c->fb_inv_size, vb); + textures[num_rects] = src_surface; + /* XXX: Hack, sort of */ + frag_shaders[num_rects] = c->fragment_shader.ycbcr_2_rgb; + ++num_rects; + } + + for (i = 0; c->dirty_layers > 0; i++) { + assert(i < VL_COMPOSITOR_MAX_LAYERS); + + if (c->dirty_layers & (1 << i)) { + struct vertex2f layer_inv_size = {1.0f / c->layers[i]->width, 1.0f / c->layers[i]->height}; + gen_rect_verts(num_rects, &c->layer_src_rects[i], &layer_inv_size, + &c->layer_dst_rects[i], &c->fb_inv_size, vb); + textures[num_rects] = c->layers[i]; + /* XXX: Hack */ + frag_shaders[num_rects] = c->fragment_shader.rgb_2_rgb; + ++num_rects; + c->dirty_layers &= ~(1 << i); + } + } + + pipe_buffer_unmap(c->pipe, buf_transfer); + + return num_rects; +} + +static void draw_layers(struct vl_compositor *c, + struct pipe_surface *src_surface, + struct pipe_video_rect *src_rect, + struct pipe_video_rect *dst_rect) +{ + unsigned num_rects; + struct pipe_surface *src_surfaces[VL_COMPOSITOR_MAX_LAYERS + 2]; + void *frag_shaders[VL_COMPOSITOR_MAX_LAYERS + 2]; + unsigned i; + + assert(c); + assert(src_surface); + assert(src_rect); + assert(dst_rect); + + num_rects = gen_data(c, src_surface, src_rect, dst_rect, src_surfaces, frag_shaders); + + for (i = 0; i < num_rects; ++i) { + boolean delete_view = FALSE; + struct pipe_sampler_view *surface_view = (struct pipe_sampler_view*)util_keymap_lookup(c->texview_map, + &src_surfaces[i]); + if (!surface_view) { + struct pipe_sampler_view templat; + u_sampler_view_default_template(&templat, src_surfaces[i]->texture, + src_surfaces[i]->texture->format); + surface_view = c->pipe->create_sampler_view(c->pipe, src_surfaces[i]->texture, + &templat); + if (!surface_view) + return; + + delete_view = !util_keymap_insert(c->texview_map, &src_surfaces[i], + surface_view, c->pipe); + } + + c->pipe->bind_fs_state(c->pipe, frag_shaders[i]); + c->pipe->set_fragment_sampler_views(c->pipe, 1, &surface_view); + + util_draw_arrays(c->pipe, PIPE_PRIM_TRIANGLES, i * 6, 6); + + if (delete_view) { + pipe_sampler_view_reference(&surface_view, NULL); + } + } +} + +void vl_compositor_render(struct vl_compositor *compositor, + struct pipe_surface *src_surface, + enum pipe_mpeg12_picture_type picture_type, + /*unsigned num_past_surfaces, + struct pipe_surface *past_surfaces, + unsigned num_future_surfaces, + struct pipe_surface *future_surfaces,*/ + struct pipe_video_rect *src_area, + struct pipe_surface *dst_surface, + struct pipe_video_rect *dst_area, + struct pipe_fence_handle **fence) +{ + assert(compositor); + assert(src_surface); + assert(src_area); + assert(dst_surface); + assert(dst_area); + assert(picture_type == PIPE_MPEG12_PICTURE_TYPE_FRAME); + + if (compositor->fb_state.width != dst_surface->width) { + compositor->fb_inv_size.x = 1.0f / dst_surface->width; + compositor->fb_state.width = dst_surface->width; + } + if (compositor->fb_state.height != dst_surface->height) { + compositor->fb_inv_size.y = 1.0f / dst_surface->height; + compositor->fb_state.height = dst_surface->height; + } + + compositor->fb_state.cbufs[0] = dst_surface; + + 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->pipe->set_framebuffer_state(compositor->pipe, &compositor->fb_state); + compositor->pipe->set_viewport_state(compositor->pipe, &compositor->viewport); + compositor->pipe->bind_fragment_sampler_states(compositor->pipe, 1, &compositor->sampler); + compositor->pipe->bind_vs_state(compositor->pipe, compositor->vertex_shader); + compositor->pipe->set_vertex_buffers(compositor->pipe, 1, &compositor->vertex_buf); + compositor->pipe->bind_vertex_elements_state(compositor->pipe, compositor->vertex_elems_state); + compositor->pipe->set_constant_buffer(compositor->pipe, PIPE_SHADER_FRAGMENT, 0, compositor->fs_const_buf); + + draw_layers(compositor, src_surface, src_area, dst_area); + + assert(!compositor->dirty_bg && !compositor->dirty_layers); - compositor->pipe->flush(compositor->pipe, PIPE_FLUSH_RENDER_CACHE, fence); ++ compositor->pipe->flush(compositor->pipe, fence); +} + +void vl_compositor_set_csc_matrix(struct vl_compositor *compositor, const float *mat) +{ + struct pipe_transfer *buf_transfer; + + assert(compositor); + + memcpy + ( + pipe_buffer_map(compositor->pipe, compositor->fs_const_buf, + PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD, + &buf_transfer), + mat, + sizeof(struct fragment_shader_consts) + ); + + pipe_buffer_unmap(compositor->pipe, buf_transfer); +} diff --cc src/gallium/auxiliary/vl/vl_mpeg12_context.c index ce1c158e828,00000000000..74893ab4e3a mode 100644,000000..100644 --- a/src/gallium/auxiliary/vl/vl_mpeg12_context.c +++ b/src/gallium/auxiliary/vl/vl_mpeg12_context.c @@@ -1,742 -1,0 +1,738 @@@ +/************************************************************************** + * + * Copyright 2009 Younes Manton. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "util/u_inlines.h" +#include "util/u_memory.h" + +#include "vl_mpeg12_context.h" +#include "vl_defines.h" +#include +#include +#include +#include +#include +#include +#include + +#define NUM_BUFFERS 2 + +static const unsigned const_empty_block_mask_420[3][2][2] = { + { { 0x20, 0x10 }, { 0x08, 0x04 } }, + { { 0x02, 0x02 }, { 0x02, 0x02 } }, + { { 0x01, 0x01 }, { 0x01, 0x01 } } +}; + +static void +flush_buffer(struct vl_mpeg12_context *ctx) +{ + unsigned ne_start, ne_num, e_start, e_num; + assert(ctx); + + if(ctx->cur_buffer != NULL) { + + vl_vb_unmap(&ctx->cur_buffer->vertex_stream, ctx->pipe); + vl_idct_unmap_buffers(&ctx->idct_y, &ctx->cur_buffer->idct_y); + vl_idct_unmap_buffers(&ctx->idct_cr, &ctx->cur_buffer->idct_cr); + vl_idct_unmap_buffers(&ctx->idct_cb, &ctx->cur_buffer->idct_cb); + vl_vb_restart(&ctx->cur_buffer->vertex_stream, + &ne_start, &ne_num, &e_start, &e_num); + + ctx->pipe->set_vertex_buffers(ctx->pipe, 2, ctx->cur_buffer->vertex_bufs.all); + ctx->pipe->bind_vertex_elements_state(ctx->pipe, ctx->vertex_elems_state); + vl_idct_flush(&ctx->idct_y, &ctx->cur_buffer->idct_y, ne_num); + vl_idct_flush(&ctx->idct_cr, &ctx->cur_buffer->idct_cr, ne_num); + vl_idct_flush(&ctx->idct_cb, &ctx->cur_buffer->idct_cb, ne_num); + vl_mpeg12_mc_renderer_flush(&ctx->mc_renderer, &ctx->cur_buffer->mc, + ne_start, ne_num, e_start, e_num); + + ctx->cur_buffer = NULL; + } +} + +static void +rotate_buffer(struct vl_mpeg12_context *ctx) +{ + struct pipe_resource *y, *cr, *cb; + static unsigned key = 0; + struct vl_mpeg12_buffer *buffer; + + assert(ctx); + + flush_buffer(ctx); + + buffer = (struct vl_mpeg12_buffer*)util_keymap_lookup(ctx->buffer_map, &key); + if (!buffer) { + boolean added_to_map; + + buffer = CALLOC_STRUCT(vl_mpeg12_buffer); + if (buffer == NULL) + return; + + buffer->vertex_bufs.individual.quad.stride = ctx->quads.stride; + buffer->vertex_bufs.individual.quad.buffer_offset = ctx->quads.buffer_offset; + pipe_resource_reference(&buffer->vertex_bufs.individual.quad.buffer, ctx->quads.buffer); + + buffer->vertex_bufs.individual.stream = vl_vb_init(&buffer->vertex_stream, ctx->pipe, + ctx->vertex_buffer_size); + if (!(y = vl_idct_init_buffer(&ctx->idct_y, &buffer->idct_y))) { + FREE(buffer); + return; + } + + if (!(cr = vl_idct_init_buffer(&ctx->idct_cr, &buffer->idct_cr))) { + FREE(buffer); + return; + } + + if (!(cb = vl_idct_init_buffer(&ctx->idct_cb, &buffer->idct_cb))) { + FREE(buffer); + return; + } + + if(!vl_mpeg12_mc_init_buffer(&ctx->mc_renderer, &buffer->mc, y, cr, cb)) { + FREE(buffer); + return; + } + + added_to_map = util_keymap_insert(ctx->buffer_map, &key, buffer, ctx); + assert(added_to_map); + } + ++key; + key %= NUM_BUFFERS; + ctx->cur_buffer = buffer; + + vl_vb_map(&ctx->cur_buffer->vertex_stream, ctx->pipe); + vl_idct_map_buffers(&ctx->idct_y, &ctx->cur_buffer->idct_y); + vl_idct_map_buffers(&ctx->idct_cr, &ctx->cur_buffer->idct_cr); + vl_idct_map_buffers(&ctx->idct_cb, &ctx->cur_buffer->idct_cb); +} + +static void +delete_buffer(const struct keymap *map, + const void *key, void *data, + void *user) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)user; + struct vl_mpeg12_buffer *buf = (struct vl_mpeg12_buffer*)data; + + assert(map); + assert(key); + assert(data); + assert(user); + + vl_vb_cleanup(&buf->vertex_stream); + vl_idct_cleanup_buffer(&ctx->idct_y, &buf->idct_y); + vl_idct_cleanup_buffer(&ctx->idct_cb, &buf->idct_cb); + vl_idct_cleanup_buffer(&ctx->idct_cr, &buf->idct_cr); + vl_mpeg12_mc_cleanup_buffer(&ctx->mc_renderer, &buf->mc); +} + +static void +upload_buffer(struct vl_mpeg12_context *ctx, + struct vl_mpeg12_buffer *buffer, + struct pipe_mpeg12_macroblock *mb) +{ + short *blocks; + unsigned tb, x, y; + + assert(ctx); + assert(buffer); + assert(mb); + + blocks = mb->blocks; + + for (y = 0; y < 2; ++y) { + for (x = 0; x < 2; ++x, ++tb) { + if (mb->cbp & (*ctx->empty_block_mask)[0][y][x]) { + vl_idct_add_block(&buffer->idct_y, mb->mbx * 2 + x, mb->mby * 2 + y, blocks); + blocks += BLOCK_WIDTH * BLOCK_HEIGHT; + } + } + } + + /* TODO: Implement 422, 444 */ + assert(ctx->base.chroma_format == PIPE_VIDEO_CHROMA_FORMAT_420); + + for (tb = 1; tb < 3; ++tb) { + if (mb->cbp & (*ctx->empty_block_mask)[tb][0][0]) { + if(tb == 1) + vl_idct_add_block(&buffer->idct_cb, mb->mbx, mb->mby, blocks); + else + vl_idct_add_block(&buffer->idct_cr, mb->mbx, mb->mby, blocks); + blocks += BLOCK_WIDTH * BLOCK_HEIGHT; + } + } +} + +static void +vl_mpeg12_destroy(struct pipe_video_context *vpipe) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)vpipe; + + assert(vpipe); + + flush_buffer(ctx); + + /* Asserted in softpipe_delete_fs_state() for some reason */ + ctx->pipe->bind_vs_state(ctx->pipe, NULL); + ctx->pipe->bind_fs_state(ctx->pipe, NULL); + + ctx->pipe->delete_blend_state(ctx->pipe, ctx->blend); + ctx->pipe->delete_rasterizer_state(ctx->pipe, ctx->rast); + ctx->pipe->delete_depth_stencil_alpha_state(ctx->pipe, ctx->dsa); + + pipe_surface_reference(&ctx->decode_target, NULL); + vl_compositor_cleanup(&ctx->compositor); + util_delete_keymap(ctx->buffer_map, ctx); + vl_mpeg12_mc_renderer_cleanup(&ctx->mc_renderer); + vl_idct_cleanup(&ctx->idct_y); + vl_idct_cleanup(&ctx->idct_cr); + vl_idct_cleanup(&ctx->idct_cb); + ctx->pipe->delete_vertex_elements_state(ctx->pipe, ctx->vertex_elems_state); + pipe_resource_reference(&ctx->quads.buffer, NULL); + ctx->pipe->destroy(ctx->pipe); + + FREE(ctx); +} + +static int +vl_mpeg12_get_param(struct pipe_video_context *vpipe, int param) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)vpipe; + + assert(vpipe); + + switch (param) { + case PIPE_CAP_NPOT_TEXTURES: + /* XXX: Temporary; not all paths are NPOT-tested */ +#if 0 + return ctx->pipe->screen->get_param(ctx->pipe->screen, param); +#endif + return FALSE; + case PIPE_CAP_DECODE_TARGET_PREFERRED_FORMAT: + return ctx->decode_format; + default: + { + debug_printf("vl_mpeg12_context: Unknown PIPE_CAP %d\n", param); + return 0; + } + } +} + +static struct pipe_surface * +vl_mpeg12_create_surface(struct pipe_video_context *vpipe, + struct pipe_resource *resource, + const struct pipe_surface *templat) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)vpipe; + + assert(vpipe); + + return ctx->pipe->create_surface(ctx->pipe, resource, templat); +} + +static boolean +vl_mpeg12_is_format_supported(struct pipe_video_context *vpipe, + enum pipe_format format, + unsigned usage, + unsigned geom) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)vpipe; + + assert(vpipe); + - /* XXX: Temporary; not all paths are NPOT-tested */ - if (geom & PIPE_TEXTURE_GEOM_NON_POWER_OF_TWO) - return FALSE; - - - return ctx->pipe->screen->is_format_supported(ctx->pipe->screen, format, PIPE_TEXTURE_2D, - 0, usage, geom); ++ return ctx->pipe->screen->is_format_supported(ctx->pipe->screen, format, ++ PIPE_TEXTURE_2D, ++ 0, usage); +} + +static void +vl_mpeg12_decode_macroblocks(struct pipe_video_context *vpipe, + struct pipe_surface *past, + struct pipe_surface *future, + unsigned num_macroblocks, + struct pipe_macroblock *macroblocks, + struct pipe_fence_handle **fence) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)vpipe; + struct pipe_mpeg12_macroblock *mpeg12_macroblocks = (struct pipe_mpeg12_macroblock*)macroblocks; + unsigned i; + + assert(vpipe); + assert(num_macroblocks); + assert(macroblocks); + assert(macroblocks->codec == PIPE_VIDEO_CODEC_MPEG12); + assert(ctx->decode_target); + assert(ctx->cur_buffer); + + for ( i = 0; i < num_macroblocks; ++i ) { + vl_vb_add_block(&ctx->cur_buffer->vertex_stream, &mpeg12_macroblocks[i], + ctx->empty_block_mask); + upload_buffer(ctx, ctx->cur_buffer, &mpeg12_macroblocks[i]); + } + + vl_mpeg12_mc_set_surfaces(&ctx->mc_renderer, &ctx->cur_buffer->mc, + ctx->decode_target, past, future, fence); +} + +static void +vl_mpeg12_clear_render_target(struct pipe_video_context *vpipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + const float *rgba, + unsigned width, unsigned height) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)vpipe; + + assert(vpipe); + assert(dst); + + if (ctx->pipe->clear_render_target) + ctx->pipe->clear_render_target(ctx->pipe, dst, rgba, dstx, dsty, width, height); + else + util_clear_render_target(ctx->pipe, dst, rgba, dstx, dsty, width, height); +} + +static void +vl_mpeg12_resource_copy_region(struct pipe_video_context *vpipe, + struct pipe_resource *dst, + unsigned dstx, unsigned dsty, unsigned dstz, + struct pipe_resource *src, + unsigned srcx, unsigned srcy, unsigned srcz, + unsigned width, unsigned height) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)vpipe; + + assert(vpipe); + assert(dst); + + struct pipe_box box; + box.x = srcx; + box.y = srcy; + box.z = srcz; + box.width = width; + box.height = height; + + if (ctx->pipe->resource_copy_region) + ctx->pipe->resource_copy_region(ctx->pipe, dst, 0, + dstx, dsty, dstz, + src, 0, &box); + else + util_resource_copy_region(ctx->pipe, dst, 0, + dstx, dsty, dstz, + src, 0, &box); +} + +static struct pipe_transfer* +vl_mpeg12_get_transfer(struct pipe_video_context *vpipe, + struct pipe_resource *resource, + unsigned level, + unsigned usage, /* a combination of PIPE_TRANSFER_x */ + const struct pipe_box *box) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)vpipe; + + assert(vpipe); + assert(resource); + assert(box); + + return ctx->pipe->get_transfer(ctx->pipe, resource, level, usage, box); +} + +static void +vl_mpeg12_transfer_destroy(struct pipe_video_context *vpipe, + struct pipe_transfer *transfer) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)vpipe; + + assert(vpipe); + assert(transfer); + + ctx->pipe->transfer_destroy(ctx->pipe, transfer); +} + +static void* +vl_mpeg12_transfer_map(struct pipe_video_context *vpipe, + struct pipe_transfer *transfer) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)vpipe; + + assert(vpipe); + assert(transfer); + + return ctx->pipe->transfer_map(ctx->pipe, transfer); +} + +static void +vl_mpeg12_transfer_flush_region(struct pipe_video_context *vpipe, + struct pipe_transfer *transfer, + const struct pipe_box *box) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)vpipe; + + assert(vpipe); + assert(transfer); + assert(box); + + ctx->pipe->transfer_flush_region(ctx->pipe, transfer, box); +} + +static void +vl_mpeg12_transfer_unmap(struct pipe_video_context *vpipe, + struct pipe_transfer *transfer) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)vpipe; + + assert(vpipe); + assert(transfer); + + ctx->pipe->transfer_unmap(ctx->pipe, transfer); +} + +static void +vl_mpeg12_transfer_inline_write(struct pipe_video_context *vpipe, + struct pipe_resource *resource, + unsigned level, + unsigned usage, /* a combination of PIPE_TRANSFER_x */ + const struct pipe_box *box, + const void *data, + unsigned stride, + unsigned slice_stride) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)vpipe; + + assert(vpipe); + assert(resource); + assert(box); + assert(data); + assert(ctx->pipe->transfer_inline_write); + + ctx->pipe->transfer_inline_write(ctx->pipe, resource, level, usage, + box, data, stride, slice_stride); +} + +static void +vl_mpeg12_render_picture(struct pipe_video_context *vpipe, + struct pipe_surface *src_surface, + enum pipe_mpeg12_picture_type picture_type, + /*unsigned num_past_surfaces, + struct pipe_surface *past_surfaces, + unsigned num_future_surfaces, + struct pipe_surface *future_surfaces,*/ + struct pipe_video_rect *src_area, + struct pipe_surface *dst_surface, + struct pipe_video_rect *dst_area, + struct pipe_fence_handle **fence) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)vpipe; + + assert(vpipe); + assert(src_surface); + assert(src_area); + assert(dst_surface); + assert(dst_area); + + flush_buffer(ctx); + + vl_compositor_render(&ctx->compositor, src_surface, + picture_type, src_area, dst_surface, dst_area, fence); +} + +static void +vl_mpeg12_set_picture_background(struct pipe_video_context *vpipe, + struct pipe_surface *bg, + struct pipe_video_rect *bg_src_rect) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)vpipe; + + assert(vpipe); + assert(bg); + assert(bg_src_rect); + + vl_compositor_set_background(&ctx->compositor, bg, bg_src_rect); +} + +static void +vl_mpeg12_set_picture_layers(struct pipe_video_context *vpipe, + struct pipe_surface *layers[], + struct pipe_video_rect *src_rects[], + struct pipe_video_rect *dst_rects[], + unsigned num_layers) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)vpipe; + + assert(vpipe); + assert((layers && src_rects && dst_rects) || + (!layers && !src_rects && !dst_rects)); + + vl_compositor_set_layers(&ctx->compositor, layers, src_rects, dst_rects, num_layers); +} + +static void +vl_mpeg12_set_decode_target(struct pipe_video_context *vpipe, + struct pipe_surface *dt) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)vpipe; + + assert(vpipe); + assert(dt); + + if (ctx->decode_target != dt || ctx->cur_buffer == NULL) { + rotate_buffer(ctx); + + pipe_surface_reference(&ctx->decode_target, dt); + } +} + +static void +vl_mpeg12_set_csc_matrix(struct pipe_video_context *vpipe, const float *mat) +{ + struct vl_mpeg12_context *ctx = (struct vl_mpeg12_context*)vpipe; + + assert(vpipe); + + vl_compositor_set_csc_matrix(&ctx->compositor, mat); +} + +static bool +init_pipe_state(struct vl_mpeg12_context *ctx) +{ + struct pipe_rasterizer_state rast; + struct pipe_blend_state blend; + struct pipe_depth_stencil_alpha_state dsa; + unsigned i; + + assert(ctx); + + memset(&rast, 0, sizeof rast); + rast.flatshade = 1; + rast.flatshade_first = 0; + rast.light_twoside = 0; + rast.front_ccw = 1; + rast.cull_face = PIPE_FACE_NONE; + rast.fill_back = PIPE_POLYGON_MODE_FILL; + rast.fill_front = PIPE_POLYGON_MODE_FILL; + rast.offset_point = 0; + rast.offset_line = 0; + rast.scissor = 0; + rast.poly_smooth = 0; + rast.poly_stipple_enable = 0; + rast.sprite_coord_enable = 0; + rast.point_size_per_vertex = 0; + rast.multisample = 0; + rast.line_smooth = 0; + rast.line_stipple_enable = 0; + rast.line_stipple_factor = 0; + rast.line_stipple_pattern = 0; + rast.line_last_pixel = 0; + rast.line_width = 1; + rast.point_smooth = 0; + rast.point_quad_rasterization = 0; + rast.point_size_per_vertex = 1; + rast.offset_units = 1; + rast.offset_scale = 1; + rast.gl_rasterization_rules = 1; + + ctx->rast = ctx->pipe->create_rasterizer_state(ctx->pipe, &rast); + ctx->pipe->bind_rasterizer_state(ctx->pipe, ctx->rast); + + 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; + ctx->blend = ctx->pipe->create_blend_state(ctx->pipe, &blend); + ctx->pipe->bind_blend_state(ctx->pipe, ctx->blend); + + memset(&dsa, 0, sizeof dsa); + dsa.depth.enabled = 0; + dsa.depth.writemask = 0; + dsa.depth.func = PIPE_FUNC_ALWAYS; + for (i = 0; i < 2; ++i) { + dsa.stencil[i].enabled = 0; + dsa.stencil[i].func = PIPE_FUNC_ALWAYS; + dsa.stencil[i].fail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[i].zpass_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[i].zfail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[i].valuemask = 0; + dsa.stencil[i].writemask = 0; + } + dsa.alpha.enabled = 0; + dsa.alpha.func = PIPE_FUNC_ALWAYS; + dsa.alpha.ref_value = 0; + ctx->dsa = ctx->pipe->create_depth_stencil_alpha_state(ctx->pipe, &dsa); + ctx->pipe->bind_depth_stencil_alpha_state(ctx->pipe, ctx->dsa); + + return true; +} + +struct pipe_video_context * +vl_create_mpeg12_context(struct pipe_context *pipe, + enum pipe_video_profile profile, + enum pipe_video_chroma_format chroma_format, + unsigned width, unsigned height, + bool pot_buffers, + enum pipe_format decode_format) +{ + struct pipe_resource *idct_matrix; + unsigned buffer_width, buffer_height; + unsigned chroma_width, chroma_height, chroma_blocks_x, chroma_blocks_y; + struct vl_mpeg12_context *ctx; + + assert(u_reduce_video_profile(profile) == PIPE_VIDEO_CODEC_MPEG12); + + ctx = CALLOC_STRUCT(vl_mpeg12_context); + + if (!ctx) + return NULL; + + /* TODO: Non-pot buffers untested, probably doesn't work without changes to texcoord generation, vert shader, etc */ + assert(pot_buffers); + + buffer_width = pot_buffers ? util_next_power_of_two(width) : width; + buffer_height = pot_buffers ? util_next_power_of_two(height) : height; + + ctx->base.profile = profile; + ctx->base.chroma_format = chroma_format; + ctx->base.width = width; + ctx->base.height = height; + + ctx->base.screen = pipe->screen; + + ctx->base.destroy = vl_mpeg12_destroy; + ctx->base.get_param = vl_mpeg12_get_param; + ctx->base.is_format_supported = vl_mpeg12_is_format_supported; + ctx->base.create_surface = vl_mpeg12_create_surface; + ctx->base.decode_macroblocks = vl_mpeg12_decode_macroblocks; + ctx->base.render_picture = vl_mpeg12_render_picture; + ctx->base.clear_render_target = vl_mpeg12_clear_render_target; + ctx->base.resource_copy_region = vl_mpeg12_resource_copy_region; + ctx->base.get_transfer = vl_mpeg12_get_transfer; + ctx->base.transfer_destroy = vl_mpeg12_transfer_destroy; + ctx->base.transfer_map = vl_mpeg12_transfer_map; + ctx->base.transfer_flush_region = vl_mpeg12_transfer_flush_region; + ctx->base.transfer_unmap = vl_mpeg12_transfer_unmap; + if (pipe->transfer_inline_write) + ctx->base.transfer_inline_write = vl_mpeg12_transfer_inline_write; + ctx->base.set_picture_background = vl_mpeg12_set_picture_background; + ctx->base.set_picture_layers = vl_mpeg12_set_picture_layers; + ctx->base.set_decode_target = vl_mpeg12_set_decode_target; + ctx->base.set_csc_matrix = vl_mpeg12_set_csc_matrix; + + ctx->pipe = pipe; + ctx->decode_format = decode_format; + + ctx->quads = vl_vb_upload_quads(ctx->pipe, 2, 2); + ctx->vertex_buffer_size = width / MACROBLOCK_WIDTH * height / MACROBLOCK_HEIGHT; + ctx->vertex_elems_state = vl_vb_get_elems_state(ctx->pipe, true); + + if (ctx->vertex_elems_state == NULL) { + ctx->pipe->destroy(ctx->pipe); + FREE(ctx); + return NULL; + } + + /* TODO: Implement 422, 444 */ + assert(chroma_format == PIPE_VIDEO_CHROMA_FORMAT_420); + ctx->empty_block_mask = &const_empty_block_mask_420; + + if (!(idct_matrix = vl_idct_upload_matrix(ctx->pipe))) + return false; + + if (!vl_idct_init(&ctx->idct_y, ctx->pipe, buffer_width, buffer_height, + 2, 2, TGSI_SWIZZLE_X, idct_matrix)) + return false; + + if (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 (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; + } + + if(!vl_idct_init(&ctx->idct_cr, ctx->pipe, chroma_width, chroma_height, + chroma_blocks_x, chroma_blocks_y, TGSI_SWIZZLE_Z, idct_matrix)) + return false; + + if(!vl_idct_init(&ctx->idct_cb, ctx->pipe, chroma_width, chroma_height, + chroma_blocks_x, chroma_blocks_y, TGSI_SWIZZLE_Y, idct_matrix)) + return false; + + if (!vl_mpeg12_mc_renderer_init(&ctx->mc_renderer, ctx->pipe, + buffer_width, buffer_height, chroma_format)) { + ctx->pipe->destroy(ctx->pipe); + FREE(ctx); + return NULL; + } + + ctx->buffer_map = util_new_keymap(sizeof(unsigned), -1, delete_buffer); + if (!ctx->buffer_map) { + vl_mpeg12_mc_renderer_cleanup(&ctx->mc_renderer); + ctx->pipe->destroy(ctx->pipe); + FREE(ctx); + return NULL; + } + + if (!vl_compositor_init(&ctx->compositor, ctx->pipe)) { + util_delete_keymap(ctx->buffer_map, ctx); + vl_mpeg12_mc_renderer_cleanup(&ctx->mc_renderer); + ctx->pipe->destroy(ctx->pipe); + FREE(ctx); + return NULL; + } + + if (!init_pipe_state(ctx)) { + vl_compositor_cleanup(&ctx->compositor); + util_delete_keymap(ctx->buffer_map, ctx); + vl_mpeg12_mc_renderer_cleanup(&ctx->mc_renderer); + ctx->pipe->destroy(ctx->pipe); + FREE(ctx); + return NULL; + } + + return &ctx->base; +} diff --cc src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c index 0e5a21c18b6,00000000000..d45b0642e3c mode 100644,000000..100644 --- a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c +++ b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c @@@ -1,690 -1,0 +1,690 @@@ +/************************************************************************** + * + * Copyright 2009 Younes Manton. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "vl_mpeg12_mc_renderer.h" +#include "vl_vertex_buffers.h" +#include "vl_defines.h" +#include "util/u_draw.h" +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +enum VS_OUTPUT +{ + VS_O_VPOS, + VS_O_LINE, + VS_O_TEX0, + VS_O_TEX1, + VS_O_TEX2, + VS_O_EB_0, + VS_O_EB_1, + VS_O_INFO, + VS_O_MV0, + VS_O_MV1, + VS_O_MV2, + VS_O_MV3 +}; + +static void * +create_vert_shader(struct vl_mpeg12_mc_renderer *r) +{ + struct ureg_program *shader; + struct ureg_src block_scale, mv_scale; + struct ureg_src vrect, vpos, eb[2][2], vmv[4]; + struct ureg_dst t_vpos, t_vtex, t_vmv; + struct ureg_dst o_vpos, o_line, o_vtex[3], o_eb[2], o_vmv[4], o_info; + unsigned i, label; + + shader = ureg_create(TGSI_PROCESSOR_VERTEX); + if (!shader) + return NULL; + + t_vpos = ureg_DECL_temporary(shader); + t_vtex = ureg_DECL_temporary(shader); + t_vmv = ureg_DECL_temporary(shader); + + vrect = ureg_DECL_vs_input(shader, VS_I_RECT); + vpos = ureg_DECL_vs_input(shader, VS_I_VPOS); + eb[0][0] = ureg_DECL_vs_input(shader, VS_I_EB_0_0); + eb[1][0] = ureg_DECL_vs_input(shader, VS_I_EB_1_0); + eb[0][1] = ureg_DECL_vs_input(shader, VS_I_EB_0_1); + eb[1][1] = ureg_DECL_vs_input(shader, VS_I_EB_1_1); + + o_vpos = ureg_DECL_output(shader, TGSI_SEMANTIC_POSITION, VS_O_VPOS); + o_line = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_LINE); + o_vtex[0] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_TEX0); + o_vtex[1] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_TEX1); + o_vtex[2] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_TEX2); + o_eb[0] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_EB_0); + o_eb[1] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_EB_1); + o_info = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_INFO); + + for (i = 0; i < 4; ++i) { + vmv[i] = ureg_DECL_vs_input(shader, VS_I_MV0 + i); + o_vmv[i] = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_MV0 + i); + } + + /* + * block_scale = (MACROBLOCK_WIDTH, MACROBLOCK_HEIGHT) / (dst.width, dst.height) + * mv_scale = 0.5 / (dst.width, dst.height); + * + * t_vpos = (vpos + vrect) * block_scale + * o_vpos.xy = t_vpos + * o_vpos.zw = vpos + * + * o_eb[0..1] = vrect.x ? eb[0..1][1] : eb[0..1][0] + * + * o_frame_pred = frame_pred + * o_info.x = ref_frames + * o_info.y = ref_frames > 0 + * o_info.z = bkwd_pred + * + * // Apply motion vectors + * o_vmv[0..count] = t_vpos + vmv[0..count] * mv_scale + * + * o_line.xy = vrect * 8 + * o_line.z = interlaced + * + * if(eb[0][0].w) { //interlaced + * t_vtex.x = vrect.x + * t_vtex.y = vrect.y * 0.5 + * t_vtex += vpos + * + * o_vtex[0].xy = t_vtex * block_scale + * + * t_vtex.y += 0.5 + * o_vtex[1].xy = t_vtex * block_scale + * } else { + * o_vtex[0..1].xy = t_vpos + * } + * o_vtex[2].xy = t_vpos + * + */ + block_scale = ureg_imm2f(shader, + (float)MACROBLOCK_WIDTH / r->buffer_width, + (float)MACROBLOCK_HEIGHT / r->buffer_height); + + mv_scale = ureg_imm2f(shader, + 0.5f / r->buffer_width, + 0.5f / r->buffer_height); + + 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), block_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); + + ureg_CMP(shader, ureg_writemask(o_eb[0], TGSI_WRITEMASK_XYZ), + ureg_negate(ureg_scalar(vrect, TGSI_SWIZZLE_X)), + eb[0][1], eb[0][0]); + ureg_CMP(shader, ureg_writemask(o_eb[1], TGSI_WRITEMASK_XYZ), + ureg_negate(ureg_scalar(vrect, TGSI_SWIZZLE_X)), + eb[1][1], eb[1][0]); + + ureg_MOV(shader, ureg_writemask(o_info, TGSI_WRITEMASK_X), + ureg_scalar(eb[1][1], TGSI_SWIZZLE_W)); + ureg_SGE(shader, ureg_writemask(o_info, TGSI_WRITEMASK_Y), + ureg_scalar(eb[1][1], TGSI_SWIZZLE_W), ureg_imm1f(shader, 0.0f)); + ureg_MOV(shader, ureg_writemask(o_info, TGSI_WRITEMASK_Z), + ureg_scalar(eb[1][0], TGSI_SWIZZLE_W)); + + ureg_MAD(shader, ureg_writemask(o_vmv[0], TGSI_WRITEMASK_XY), mv_scale, vmv[0], ureg_src(t_vpos)); + ureg_MAD(shader, ureg_writemask(o_vmv[2], TGSI_WRITEMASK_XY), mv_scale, vmv[2], ureg_src(t_vpos)); + + ureg_CMP(shader, ureg_writemask(t_vmv, TGSI_WRITEMASK_XY), + ureg_negate(ureg_scalar(eb[0][1], TGSI_SWIZZLE_W)), + vmv[0], vmv[1]); + ureg_MAD(shader, ureg_writemask(o_vmv[1], TGSI_WRITEMASK_XY), mv_scale, ureg_src(t_vmv), ureg_src(t_vpos)); + + ureg_CMP(shader, ureg_writemask(t_vmv, TGSI_WRITEMASK_XY), + ureg_negate(ureg_scalar(eb[0][1], TGSI_SWIZZLE_W)), + vmv[2], vmv[3]); + ureg_MAD(shader, ureg_writemask(o_vmv[3], TGSI_WRITEMASK_XY), mv_scale, ureg_src(t_vmv), ureg_src(t_vpos)); + + ureg_MOV(shader, ureg_writemask(o_vtex[0], TGSI_WRITEMASK_XY), ureg_src(t_vpos)); + ureg_MOV(shader, ureg_writemask(o_vtex[1], TGSI_WRITEMASK_XY), ureg_src(t_vpos)); + ureg_MOV(shader, ureg_writemask(o_vtex[2], TGSI_WRITEMASK_XY), ureg_src(t_vpos)); + + ureg_MOV(shader, ureg_writemask(o_line, TGSI_WRITEMASK_X), ureg_scalar(vrect, TGSI_SWIZZLE_Y)); + ureg_MUL(shader, ureg_writemask(o_line, TGSI_WRITEMASK_Y), + vrect, ureg_imm1f(shader, MACROBLOCK_HEIGHT / 2)); + + ureg_IF(shader, ureg_scalar(eb[0][0], TGSI_SWIZZLE_W), &label); + + ureg_MOV(shader, ureg_writemask(t_vtex, TGSI_WRITEMASK_X), vrect); + ureg_MUL(shader, ureg_writemask(t_vtex, TGSI_WRITEMASK_Y), vrect, ureg_imm1f(shader, 0.5f)); + ureg_ADD(shader, ureg_writemask(t_vtex, TGSI_WRITEMASK_XY), vpos, ureg_src(t_vtex)); + ureg_MUL(shader, ureg_writemask(o_vtex[0], TGSI_WRITEMASK_XY), ureg_src(t_vtex), block_scale); + ureg_ADD(shader, ureg_writemask(t_vtex, TGSI_WRITEMASK_Y), ureg_src(t_vtex), ureg_imm1f(shader, 0.5f)); + ureg_MUL(shader, ureg_writemask(o_vtex[1], TGSI_WRITEMASK_XY), ureg_src(t_vtex), block_scale); + + ureg_MUL(shader, ureg_writemask(o_line, TGSI_WRITEMASK_X), + ureg_scalar(vrect, TGSI_SWIZZLE_Y), + ureg_imm1f(shader, MACROBLOCK_HEIGHT / 2)); + + ureg_fixup_label(shader, label, ureg_get_instruction_number(shader)); + ureg_ENDIF(shader); + + ureg_release_temporary(shader, t_vtex); + ureg_release_temporary(shader, t_vpos); + ureg_release_temporary(shader, t_vmv); + + ureg_END(shader); + + return ureg_create_shader_and_destroy(shader, r->pipe); +} + +static struct ureg_dst +calc_field(struct ureg_program *shader) +{ + struct ureg_dst tmp; + struct ureg_src line; + + tmp = ureg_DECL_temporary(shader); + + line = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_LINE, TGSI_INTERPOLATE_LINEAR); + + /* + * line.x going from 0 to 1 if not interlaced + * line.x going from 0 to 8 in steps of 0.5 if interlaced + * line.y going from 0 to 8 in steps of 0.5 + * + * tmp.xy = fraction(line) + * tmp.xy = tmp.xy >= 0.5 ? 1 : 0 + */ + ureg_FRC(shader, ureg_writemask(tmp, TGSI_WRITEMASK_XY), line); + ureg_SGE(shader, ureg_writemask(tmp, TGSI_WRITEMASK_XY), ureg_src(tmp), ureg_imm1f(shader, 0.5f)); + + return tmp; +} + +static struct ureg_dst +fetch_ycbcr(struct vl_mpeg12_mc_renderer *r, struct ureg_program *shader, struct ureg_dst field) +{ + struct ureg_src tc[3], sampler[3], eb[2]; + struct ureg_dst texel, t_tc, t_eb_info; + unsigned i, label; + + texel = ureg_DECL_temporary(shader); + t_tc = ureg_DECL_temporary(shader); + t_eb_info = ureg_DECL_temporary(shader); + + tc[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_TEX0, TGSI_INTERPOLATE_LINEAR); + tc[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_TEX1, TGSI_INTERPOLATE_LINEAR); + tc[2] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_TEX2, TGSI_INTERPOLATE_LINEAR); + + eb[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_EB_0, TGSI_INTERPOLATE_CONSTANT); + eb[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_EB_1, TGSI_INTERPOLATE_CONSTANT); + + for (i = 0; i < 3; ++i) { + sampler[i] = ureg_DECL_sampler(shader, i); + } + + /* + * texel.y = tex(field.y ? tc[1] : tc[0], sampler[0]) + * texel.cb = tex(tc[2], sampler[1]) + * texel.cr = tex(tc[2], sampler[2]) + */ + + ureg_CMP(shader, ureg_writemask(t_tc, TGSI_WRITEMASK_XY), + ureg_negate(ureg_scalar(ureg_src(field), TGSI_SWIZZLE_X)), + tc[1], tc[0]); + + ureg_CMP(shader, ureg_writemask(t_eb_info, TGSI_WRITEMASK_XYZ), + ureg_negate(ureg_scalar(ureg_src(field), TGSI_SWIZZLE_X)), + eb[1], eb[0]); + + /* r600g is ignoring TGSI_INTERPOLATE_CONSTANT, just workaround this */ + ureg_SLT(shader, ureg_writemask(t_eb_info, TGSI_WRITEMASK_XYZ), ureg_src(t_eb_info), ureg_imm1f(shader, 0.5f)); + + ureg_MOV(shader, ureg_writemask(texel, TGSI_WRITEMASK_XYZ), ureg_imm1f(shader, 0.0f)); + for (i = 0; i < 3; ++i) { + ureg_IF(shader, ureg_scalar(ureg_src(t_eb_info), TGSI_SWIZZLE_X + i), &label); + + /* Nouveau can't writemask tex dst regs (yet?), so this won't work anymore on nvidia hardware */ + if(i==0 || r->chroma_format == PIPE_VIDEO_CHROMA_FORMAT_444) { + ureg_TEX(shader, ureg_writemask(texel, TGSI_WRITEMASK_X << i), TGSI_TEXTURE_3D, ureg_src(t_tc), sampler[i]); + } else { + ureg_TEX(shader, ureg_writemask(texel, TGSI_WRITEMASK_X << i), TGSI_TEXTURE_3D, tc[2], sampler[i]); + } + + ureg_fixup_label(shader, label, ureg_get_instruction_number(shader)); + ureg_ENDIF(shader); + } + + ureg_release_temporary(shader, t_tc); + ureg_release_temporary(shader, t_eb_info); + + return texel; +} + +static struct ureg_dst +fetch_ref(struct ureg_program *shader, struct ureg_dst field) +{ + struct ureg_src info; + struct ureg_src tc[4], sampler[2]; + struct ureg_dst ref[2], result; + unsigned i, intra_label, bi_label, label; + + info = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_INFO, TGSI_INTERPOLATE_CONSTANT); + + for (i = 0; i < 4; ++i) + tc[i] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_MV0 + i, TGSI_INTERPOLATE_LINEAR); + + for (i = 0; i < 2; ++i) { + sampler[i] = ureg_DECL_sampler(shader, i + 3); + ref[i] = ureg_DECL_temporary(shader); + } + + result = ureg_DECL_temporary(shader); + + ureg_MOV(shader, ureg_writemask(result, TGSI_WRITEMASK_XYZ), ureg_imm1f(shader, 0.5f)); + + ureg_IF(shader, ureg_scalar(info, TGSI_SWIZZLE_Y), &intra_label); + ureg_CMP(shader, ureg_writemask(ref[0], TGSI_WRITEMASK_XY), + ureg_negate(ureg_scalar(ureg_src(field), TGSI_SWIZZLE_Y)), + tc[1], tc[0]); + + ureg_IF(shader, ureg_scalar(info, TGSI_SWIZZLE_X), &bi_label); + + /* + * result = tex(field.z ? tc[1] : tc[0], sampler[bkwd_pred ? 1 : 0]) + */ + ureg_IF(shader, ureg_scalar(info, TGSI_SWIZZLE_Z), &label); + ureg_TEX(shader, result, TGSI_TEXTURE_2D, ureg_src(ref[0]), sampler[1]); + ureg_fixup_label(shader, label, ureg_get_instruction_number(shader)); + ureg_ELSE(shader, &label); + ureg_TEX(shader, result, TGSI_TEXTURE_2D, ureg_src(ref[0]), sampler[0]); + ureg_fixup_label(shader, label, ureg_get_instruction_number(shader)); + ureg_ENDIF(shader); + + ureg_fixup_label(shader, bi_label, ureg_get_instruction_number(shader)); + ureg_ELSE(shader, &bi_label); + + /* + * if (field.z) + * ref[0..1] = tex(tc[0..1], sampler[0..1]) + * else + * ref[0..1] = tex(tc[2..3], sampler[0..1]) + */ + ureg_CMP(shader, ureg_writemask(ref[1], TGSI_WRITEMASK_XY), + ureg_negate(ureg_scalar(ureg_src(field), TGSI_SWIZZLE_Y)), + tc[3], tc[2]); + ureg_TEX(shader, ref[0], TGSI_TEXTURE_2D, ureg_src(ref[0]), sampler[0]); + ureg_TEX(shader, ref[1], TGSI_TEXTURE_2D, ureg_src(ref[1]), sampler[1]); + + ureg_LRP(shader, ureg_writemask(result, TGSI_WRITEMASK_XYZ), ureg_imm1f(shader, 0.5f), + ureg_src(ref[0]), ureg_src(ref[1])); + + ureg_fixup_label(shader, bi_label, ureg_get_instruction_number(shader)); + ureg_ENDIF(shader); + ureg_fixup_label(shader, intra_label, ureg_get_instruction_number(shader)); + ureg_ENDIF(shader); + + for (i = 0; i < 2; ++i) + ureg_release_temporary(shader, ref[i]); + + return result; +} + +static void * +create_frag_shader(struct vl_mpeg12_mc_renderer *r) +{ + struct ureg_program *shader; + struct ureg_dst result; + struct ureg_dst field, texel; + struct ureg_dst fragment; + + shader = ureg_create(TGSI_PROCESSOR_FRAGMENT); + if (!shader) + return NULL; + + fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0); + + field = calc_field(shader); + texel = fetch_ycbcr(r, shader, field); + + result = fetch_ref(shader, field); + + ureg_ADD(shader, ureg_writemask(fragment, TGSI_WRITEMASK_XYZ), ureg_src(texel), ureg_src(result)); + + ureg_release_temporary(shader, field); + ureg_release_temporary(shader, texel); + ureg_release_temporary(shader, result); + ureg_END(shader); + + return ureg_create_shader_and_destroy(shader, r->pipe); +} + +static bool +init_pipe_state(struct vl_mpeg12_mc_renderer *r) +{ + struct pipe_sampler_state sampler; + struct pipe_rasterizer_state rs_state; + unsigned filters[5]; + unsigned i; + + assert(r); + + r->viewport.scale[0] = r->buffer_width; + r->viewport.scale[1] = r->buffer_height; + r->viewport.scale[2] = 1; + r->viewport.scale[3] = 1; + r->viewport.translate[0] = 0; + r->viewport.translate[1] = 0; + r->viewport.translate[2] = 0; + r->viewport.translate[3] = 0; + + r->fb_state.width = r->buffer_width; + r->fb_state.height = r->buffer_height; + r->fb_state.nr_cbufs = 1; + r->fb_state.zsbuf = NULL; + + /* Luma filter */ + filters[0] = PIPE_TEX_FILTER_NEAREST; + /* Chroma filters */ + if (r->chroma_format == PIPE_VIDEO_CHROMA_FORMAT_444 || true) { //TODO + filters[1] = PIPE_TEX_FILTER_NEAREST; + filters[2] = PIPE_TEX_FILTER_NEAREST; + } + else { + filters[1] = PIPE_TEX_FILTER_LINEAR; + filters[2] = PIPE_TEX_FILTER_LINEAR; + } + /* Fwd, bkwd ref filters */ + filters[3] = PIPE_TEX_FILTER_LINEAR; + filters[4] = PIPE_TEX_FILTER_LINEAR; + + for (i = 0; i < 5; ++i) { + memset(&sampler, 0, sizeof(sampler)); + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_BORDER; + sampler.min_img_filter = filters[i]; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + sampler.mag_img_filter = filters[i]; + sampler.compare_mode = PIPE_TEX_COMPARE_NONE; + sampler.compare_func = PIPE_FUNC_ALWAYS; + sampler.normalized_coords = 1; + /*sampler.shadow_ambient = ; */ + /*sampler.lod_bias = ; */ + sampler.min_lod = 0; + /*sampler.max_lod = ; */ + sampler.border_color[0] = 0.0f; + sampler.border_color[1] = 0.0f; + sampler.border_color[2] = 0.0f; + sampler.border_color[3] = 0.0f; + /*sampler.max_anisotropy = ; */ + r->samplers.all[i] = r->pipe->create_sampler_state(r->pipe, &sampler); + } + + memset(&rs_state, 0, sizeof(rs_state)); + /*rs_state.sprite_coord_enable */ + rs_state.sprite_coord_mode = PIPE_SPRITE_COORD_UPPER_LEFT; + rs_state.point_quad_rasterization = true; + rs_state.point_size = BLOCK_WIDTH; + rs_state.gl_rasterization_rules = true; + r->rs_state = r->pipe->create_rasterizer_state(r->pipe, &rs_state); + + return true; +} + +static void +cleanup_pipe_state(struct vl_mpeg12_mc_renderer *r) +{ + unsigned i; + + assert(r); + + for (i = 0; i < 5; ++i) + r->pipe->delete_sampler_state(r->pipe, r->samplers.all[i]); + + r->pipe->delete_rasterizer_state(r->pipe, r->rs_state); +} + +static struct pipe_sampler_view +*find_or_create_sampler_view(struct vl_mpeg12_mc_renderer *r, struct pipe_surface *surface) +{ + struct pipe_sampler_view *sampler_view; + assert(r); + assert(surface); + + sampler_view = (struct pipe_sampler_view*)util_keymap_lookup(r->texview_map, &surface); + if (!sampler_view) { + struct pipe_sampler_view templat; + boolean added_to_map; + + u_sampler_view_default_template(&templat, surface->texture, + surface->texture->format); + sampler_view = r->pipe->create_sampler_view(r->pipe, surface->texture, + &templat); + if (!sampler_view) + return NULL; + + added_to_map = util_keymap_insert(r->texview_map, &surface, + sampler_view, r->pipe); + assert(added_to_map); + } + + return sampler_view; +} + +static void +texview_map_delete(const struct keymap *map, + const void *key, void *data, + void *user) +{ + struct pipe_sampler_view *sv = (struct pipe_sampler_view*)data; + + assert(map); + assert(key); + assert(data); + assert(user); + + pipe_sampler_view_reference(&sv, NULL); +} + +bool +vl_mpeg12_mc_renderer_init(struct vl_mpeg12_mc_renderer *renderer, + struct pipe_context *pipe, + unsigned buffer_width, + unsigned buffer_height, + enum pipe_video_chroma_format chroma_format) +{ + assert(renderer); + assert(pipe); + + memset(renderer, 0, sizeof(struct vl_mpeg12_mc_renderer)); + + renderer->pipe = pipe; + renderer->buffer_width = buffer_width; + renderer->buffer_height = buffer_height; + renderer->chroma_format = chroma_format; + + renderer->texview_map = util_new_keymap(sizeof(struct pipe_surface*), -1, + texview_map_delete); + if (!renderer->texview_map) + return false; + + if (!init_pipe_state(renderer)) + goto error_pipe_state; + + renderer->vs = create_vert_shader(renderer); + renderer->fs = create_frag_shader(renderer); + + if (renderer->vs == NULL || renderer->fs == NULL) + goto error_shaders; + + return true; + +error_shaders: + cleanup_pipe_state(renderer); + +error_pipe_state: + util_delete_keymap(renderer->texview_map, renderer->pipe); + return false; +} + +void +vl_mpeg12_mc_renderer_cleanup(struct vl_mpeg12_mc_renderer *renderer) +{ + assert(renderer); + + util_delete_keymap(renderer->texview_map, renderer->pipe); + cleanup_pipe_state(renderer); + + renderer->pipe->delete_vs_state(renderer->pipe, renderer->vs); + renderer->pipe->delete_fs_state(renderer->pipe, renderer->fs); +} + +bool +vl_mpeg12_mc_init_buffer(struct vl_mpeg12_mc_renderer *renderer, struct vl_mpeg12_mc_buffer *buffer, + struct pipe_resource *y, struct pipe_resource *cr, struct pipe_resource *cb) +{ + struct pipe_sampler_view sampler_view; + + unsigned i; + + assert(renderer && buffer); + + buffer->surface = NULL; + buffer->past = NULL; + buffer->future = NULL; + + pipe_resource_reference(&buffer->textures.individual.y, y); + pipe_resource_reference(&buffer->textures.individual.cr, cr); + pipe_resource_reference(&buffer->textures.individual.cb, cb); + + for (i = 0; i < 3; ++i) { + u_sampler_view_default_template(&sampler_view, + buffer->textures.all[i], + buffer->textures.all[i]->format); + sampler_view.swizzle_r = i == 0 ? PIPE_SWIZZLE_RED : PIPE_SWIZZLE_ZERO; + sampler_view.swizzle_g = i == 1 ? PIPE_SWIZZLE_RED : PIPE_SWIZZLE_ZERO; + sampler_view.swizzle_b = i == 2 ? PIPE_SWIZZLE_RED : PIPE_SWIZZLE_ZERO; + sampler_view.swizzle_a = PIPE_SWIZZLE_ONE; + buffer->sampler_views.all[i] = renderer->pipe->create_sampler_view( + renderer->pipe, buffer->textures.all[i], &sampler_view); + } + + return true; +} + +void +vl_mpeg12_mc_cleanup_buffer(struct vl_mpeg12_mc_renderer *renderer, struct vl_mpeg12_mc_buffer *buffer) +{ + unsigned i; + + assert(renderer && buffer); + + for (i = 0; i < 3; ++i) { + pipe_sampler_view_reference(&buffer->sampler_views.all[i], NULL); + pipe_resource_reference(&buffer->textures.all[i], NULL); + } + + pipe_surface_reference(&buffer->surface, NULL); + pipe_surface_reference(&buffer->past, NULL); + pipe_surface_reference(&buffer->future, NULL); +} + +void +vl_mpeg12_mc_set_surfaces(struct vl_mpeg12_mc_renderer *renderer, + struct vl_mpeg12_mc_buffer *buffer, + struct pipe_surface *surface, + struct pipe_surface *past, + struct pipe_surface *future, + struct pipe_fence_handle **fence) +{ + assert(renderer && buffer); + assert(surface); + + if (surface != buffer->surface) { + pipe_surface_reference(&buffer->surface, surface); + pipe_surface_reference(&buffer->past, past); + pipe_surface_reference(&buffer->future, future); + buffer->fence = fence; + } else { + /* If the surface we're rendering hasn't changed the ref frames shouldn't change. */ + assert(buffer->past == past); + assert(buffer->future == future); + } +} + +void +vl_mpeg12_mc_renderer_flush(struct vl_mpeg12_mc_renderer *renderer, struct vl_mpeg12_mc_buffer *buffer, + unsigned not_empty_start_instance, unsigned not_empty_num_instances, + unsigned empty_start_instance, unsigned empty_num_instances) +{ + assert(renderer && buffer); + + if (not_empty_num_instances == 0 && empty_num_instances == 0) + return; + + renderer->fb_state.cbufs[0] = buffer->surface; + renderer->pipe->bind_rasterizer_state(renderer->pipe, renderer->rs_state); + renderer->pipe->set_framebuffer_state(renderer->pipe, &renderer->fb_state); + renderer->pipe->set_viewport_state(renderer->pipe, &renderer->viewport); + + if (buffer->past) { + buffer->sampler_views.individual.ref[0] = find_or_create_sampler_view(renderer, buffer->past); + } else { + buffer->sampler_views.individual.ref[0] = find_or_create_sampler_view(renderer, buffer->surface); + } + + if (buffer->future) { + buffer->sampler_views.individual.ref[1] = find_or_create_sampler_view(renderer, buffer->future); + } else { + buffer->sampler_views.individual.ref[1] = find_or_create_sampler_view(renderer, buffer->surface); + } + + renderer->pipe->set_fragment_sampler_views(renderer->pipe, 5, buffer->sampler_views.all); + renderer->pipe->bind_fragment_sampler_states(renderer->pipe, 5, renderer->samplers.all); + + renderer->pipe->bind_vs_state(renderer->pipe, renderer->vs); + renderer->pipe->bind_fs_state(renderer->pipe, renderer->fs); + + if (not_empty_num_instances > 0) + util_draw_arrays_instanced(renderer->pipe, PIPE_PRIM_QUADS, 0, 4, + not_empty_start_instance, not_empty_num_instances); + + if (empty_num_instances > 0) + util_draw_arrays_instanced(renderer->pipe, PIPE_PRIM_QUADS, 0, 4, + empty_start_instance, empty_num_instances); + - renderer->pipe->flush(renderer->pipe, PIPE_FLUSH_RENDER_CACHE, buffer->fence); ++ renderer->pipe->flush(renderer->pipe, buffer->fence); + + /* Next time we get this surface it may have new ref frames */ + pipe_surface_reference(&buffer->surface, NULL); + pipe_surface_reference(&buffer->past, NULL); + pipe_surface_reference(&buffer->future, NULL); +} diff --cc src/gallium/drivers/r600/r600_asm.c index 996d9f74a60,240093f9b9d..45e614977e2 --- a/src/gallium/drivers/r600/r600_asm.c +++ b/src/gallium/drivers/r600/r600_asm.c @@@ -1304,16 -1250,24 +1307,34 @@@ int r600_bc_add_alu(struct r600_bc *bc return r600_bc_add_alu_type(bc, alu, BC_INST(bc, V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU)); } +static void r600_bc_remove_alu(struct r600_bc_cf *cf, struct r600_bc_alu *alu) +{ + if (alu->last && alu->list.prev != &cf->alu) { + PREV_ALU(alu)->last = 1; + } + LIST_DEL(&alu->list); + free(alu); + cf->ndw -= 2; +} + + static unsigned r600_bc_num_tex_and_vtx_instructions(const struct r600_bc *bc) + { + switch (bc->chiprev) { + case CHIPREV_R600: + return 8; + + case CHIPREV_R700: + return 16; + + case CHIPREV_EVERGREEN: + return 64; + + default: + R600_ERR("Unknown chiprev %d.\n", bc->chiprev); + return 8; + } + } + int r600_bc_add_vtx(struct r600_bc *bc, const struct r600_bc_vtx *vtx) { struct r600_bc_vtx *nvtx = r600_bc_vtx(); @@@ -1523,64 -1453,24 +1520,72 @@@ static int r600_bc_alu_build(struct r60 return 0; } + static void r600_bc_cf_vtx_build(uint32_t *bytecode, const struct r600_bc_cf *cf) + { + *bytecode++ = S_SQ_CF_WORD0_ADDR(cf->addr >> 1); + *bytecode++ = S_SQ_CF_WORD1_CF_INST(cf->inst) | + S_SQ_CF_WORD1_BARRIER(1) | + S_SQ_CF_WORD1_COUNT((cf->ndw / 4) - 1); + } + -/* common for r600/r700 - eg in eg_asm.c */ -static int r600_bc_cf_build(struct r600_bc *bc, struct r600_bc_cf *cf) +enum cf_class { - unsigned id = cf->id; + CF_CLASS_ALU, + CF_CLASS_TEXTURE, + CF_CLASS_VERTEX, + CF_CLASS_EXPORT, + CF_CLASS_OTHER +}; +static enum cf_class r600_bc_cf_class(struct r600_bc_cf *cf) +{ switch (cf->inst) { case (V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU << 3): - case (V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_PUSH_BEFORE << 3): case (V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_POP_AFTER << 3): case (V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_POP2_AFTER << 3): + case (V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_PUSH_BEFORE << 3): + return CF_CLASS_ALU; + + case V_SQ_CF_WORD1_SQ_CF_INST_TEX: + return CF_CLASS_TEXTURE; + + case V_SQ_CF_WORD1_SQ_CF_INST_VTX: + case V_SQ_CF_WORD1_SQ_CF_INST_VTX_TC: + return CF_CLASS_VERTEX; + + case V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT: + case V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT_DONE: + case EG_V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT: + case EG_V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT_DONE: + return CF_CLASS_EXPORT; + + case V_SQ_CF_WORD1_SQ_CF_INST_JUMP: + case V_SQ_CF_WORD1_SQ_CF_INST_ELSE: + case V_SQ_CF_WORD1_SQ_CF_INST_POP: + case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_NO_AL: + case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_END: + case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_CONTINUE: + case V_SQ_CF_WORD1_SQ_CF_INST_LOOP_BREAK: + case V_SQ_CF_WORD1_SQ_CF_INST_CALL_FS: + case V_SQ_CF_WORD1_SQ_CF_INST_RETURN: + case V_SQ_CF_WORD1_SQ_CF_INST_NOP: + return CF_CLASS_OTHER; + + default: + R600_ERR("unsupported CF instruction (0x%X)\n", cf->inst); + return -EINVAL; + } +} + +/* common for r600/r700 - eg in eg_asm.c */ +static int r600_bc_cf_build(struct r600_bc *bc, struct r600_bc_cf *cf) +{ + unsigned id = cf->id; + unsigned end_of_program = bc->cf.prev == &cf->list; + + switch (r600_bc_cf_class(cf)) { + case CF_CLASS_ALU: + assert(!end_of_program); bc->bytecode[id++] = S_SQ_CF_ALU_WORD0_ADDR(cf->addr >> 1) | S_SQ_CF_ALU_WORD0_KCACHE_MODE0(cf->kcache[0].mode) | S_SQ_CF_ALU_WORD0_KCACHE_BANK0(cf->kcache[0].bank) | @@@ -1590,19 -1480,20 +1595,18 @@@ S_SQ_CF_ALU_WORD1_KCACHE_MODE1(cf->kcache[1].mode) | S_SQ_CF_ALU_WORD1_KCACHE_ADDR0(cf->kcache[0].addr) | S_SQ_CF_ALU_WORD1_KCACHE_ADDR1(cf->kcache[1].addr) | - S_SQ_CF_ALU_WORD1_BARRIER(1) | - S_SQ_CF_ALU_WORD1_USES_WATERFALL(bc->chiprev == CHIPREV_R600 ? cf->r6xx_uses_waterfall : 0) | - S_SQ_CF_ALU_WORD1_COUNT((cf->ndw / 2) - 1); + S_SQ_CF_ALU_WORD1_BARRIER(cf->barrier) | + S_SQ_CF_ALU_WORD1_USES_WATERFALL(bc->chiprev == CHIPREV_R600 ? cf->r6xx_uses_waterfall : 0) | + S_SQ_CF_ALU_WORD1_COUNT((cf->ndw / 2) - 1); break; - case V_SQ_CF_WORD1_SQ_CF_INST_TEX: - case V_SQ_CF_WORD1_SQ_CF_INST_VTX: - case V_SQ_CF_WORD1_SQ_CF_INST_VTX_TC: + case CF_CLASS_TEXTURE: + case CF_CLASS_VERTEX: - bc->bytecode[id++] = S_SQ_CF_WORD0_ADDR(cf->addr >> 1); - bc->bytecode[id++] = S_SQ_CF_WORD1_CF_INST(cf->inst) | - S_SQ_CF_WORD1_BARRIER(cf->barrier) | - S_SQ_CF_WORD1_COUNT((cf->ndw / 4) - 1) | - S_SQ_CF_WORD1_END_OF_PROGRAM(end_of_program); + if (bc->chiprev == CHIPREV_R700) + r700_bc_cf_vtx_build(&bc->bytecode[id], cf); + else + r600_bc_cf_vtx_build(&bc->bytecode[id], cf); break; - case V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT: - case V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT_DONE: + case CF_CLASS_EXPORT: bc->bytecode[id++] = S_SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR(cf->output.gpr) | S_SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE(cf->output.elem_size) | S_SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE(cf->output.array_base) | diff --cc src/gallium/drivers/r600/r600_texture.c index 16b53a0b2a5,dc351bfb62d..d079b571156 --- a/src/gallium/drivers/r600/r600_texture.c +++ b/src/gallium/drivers/r600/r600_texture.c @@@ -306,19 -306,10 +306,19 @@@ static boolean permit_hardware_blit(str res->format, res->target, res->nr_samples, - PIPE_BIND_SAMPLER_VIEW, 0)) + PIPE_BIND_SAMPLER_VIEW)) return FALSE; - return TRUE; + switch (res->usage) { + case PIPE_USAGE_STREAM: + case PIPE_USAGE_STAGING: + case PIPE_USAGE_STATIC: + case PIPE_USAGE_IMMUTABLE: + return FALSE; + + default: + return TRUE; + } } static boolean r600_texture_get_handle(struct pipe_screen* screen, diff --cc src/gallium/include/pipe/p_defines.h index 8cf738fa2c0,bac3300df3c..e9d47983e1b --- a/src/gallium/include/pipe/p_defines.h +++ b/src/gallium/include/pipe/p_defines.h @@@ -504,40 -488,7 +488,33 @@@ enum pipe_shader_ca PIPE_SHADER_CAP_SUBROUTINES, /* BGNSUB, ENDSUB, CAL, RET */ }; - /** - * Referenced query flags. - */ - - #define PIPE_UNREFERENCED 0 - #define PIPE_REFERENCED_FOR_READ (1 << 0) - #define PIPE_REFERENCED_FOR_WRITE (1 << 1) +enum pipe_video_codec +{ + PIPE_VIDEO_CODEC_UNKNOWN = 0, + PIPE_VIDEO_CODEC_MPEG12, /**< MPEG1, MPEG2 */ + PIPE_VIDEO_CODEC_MPEG4, /**< DIVX, XVID */ + PIPE_VIDEO_CODEC_VC1, /**< WMV */ + PIPE_VIDEO_CODEC_MPEG4_AVC /**< H.264 */ +}; + +enum pipe_video_profile +{ + PIPE_VIDEO_PROFILE_UNKNOWN, + PIPE_VIDEO_PROFILE_MPEG1, + PIPE_VIDEO_PROFILE_MPEG2_SIMPLE, + PIPE_VIDEO_PROFILE_MPEG2_MAIN, + PIPE_VIDEO_PROFILE_MPEG4_SIMPLE, + PIPE_VIDEO_PROFILE_MPEG4_ADVANCED_SIMPLE, + PIPE_VIDEO_PROFILE_VC1_SIMPLE, + PIPE_VIDEO_PROFILE_VC1_MAIN, + PIPE_VIDEO_PROFILE_VC1_ADVANCED, + PIPE_VIDEO_PROFILE_MPEG4_AVC_BASELINE, + PIPE_VIDEO_PROFILE_MPEG4_AVC_MAIN, + PIPE_VIDEO_PROFILE_MPEG4_AVC_HIGH +}; + + /** * Composite query types */ diff --cc src/gallium/include/pipe/p_format.h index e2cc32222de,d53af598e75..f3b1e5561cf --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@@ -196,28 -196,21 +196,33 @@@ enum pipe_format PIPE_FORMAT_R8G8B8X8_UNORM = 134, PIPE_FORMAT_B4G4R4X4_UNORM = 135, + PIPE_FORMAT_YV12 = 136, + PIPE_FORMAT_YV16 = 137, + PIPE_FORMAT_IYUV = 138, /**< aka I420 */ + PIPE_FORMAT_NV12 = 139, + PIPE_FORMAT_NV21 = 140, + PIPE_FORMAT_AYUV = PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_FORMAT_VUYA = PIPE_FORMAT_B8G8R8A8_UNORM, + PIPE_FORMAT_XYUV = PIPE_FORMAT_X8R8G8B8_UNORM, + PIPE_FORMAT_VUYX = PIPE_FORMAT_B8G8R8X8_UNORM, + PIPE_FORMAT_IA44 = 141, + PIPE_FORMAT_AI44 = 142, + /* some stencil samplers formats */ - PIPE_FORMAT_X24S8_USCALED = 136, - PIPE_FORMAT_S8X24_USCALED = 137, - PIPE_FORMAT_X32_S8X24_USCALED = 138, + PIPE_FORMAT_X24S8_USCALED = 143, + PIPE_FORMAT_S8X24_USCALED = 144, + PIPE_FORMAT_X32_S8X24_USCALED = 145, - PIPE_FORMAT_B2G3R3_UNORM = 139, - PIPE_FORMAT_L16A16_UNORM = 140, - PIPE_FORMAT_A16_UNORM = 141, - PIPE_FORMAT_I16_UNORM = 142, + PIPE_FORMAT_B2G3R3_UNORM = 146, + PIPE_FORMAT_L16A16_UNORM = 147, + PIPE_FORMAT_A16_UNORM = 148, + PIPE_FORMAT_I16_UNORM = 149, + PIPE_FORMAT_LATC1_UNORM = 143, + PIPE_FORMAT_LATC1_SNORM = 144, + PIPE_FORMAT_LATC2_UNORM = 145, + PIPE_FORMAT_LATC2_SNORM = 146, + PIPE_FORMAT_COUNT }; diff --cc src/gallium/include/pipe/p_video_context.h index 73f03d58ea6,00000000000..2d59741ec18 mode 100644,000000..100644 --- a/src/gallium/include/pipe/p_video_context.h +++ b/src/gallium/include/pipe/p_video_context.h @@@ -1,182 -1,0 +1,181 @@@ +/************************************************************************** + * + * Copyright 2009 Younes Manton. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef PIPE_VIDEO_CONTEXT_H +#define PIPE_VIDEO_CONTEXT_H + +#ifdef __cplusplus +extern "C" { +#endif + +#include + +/* XXX: Move to an appropriate place */ +#define PIPE_CAP_DECODE_TARGET_PREFERRED_FORMAT 256 + +struct pipe_screen; +struct pipe_buffer; +struct pipe_surface; +struct pipe_macroblock; +struct pipe_picture_desc; +struct pipe_fence_handle; + +/** + * Gallium video rendering context + */ +struct pipe_video_context +{ + struct pipe_screen *screen; + enum pipe_video_profile profile; + enum pipe_video_chroma_format chroma_format; + unsigned width; + unsigned height; + + void *priv; /**< context private data (for DRI for example) */ + + /** + * Query an integer-valued capability/parameter/limit + * \param param one of PIPE_CAP_x + */ + int (*get_param)(struct pipe_video_context *vpipe, int param); + + /** + * Check if the given pipe_format is supported as a texture or + * drawing surface. + */ + boolean (*is_format_supported)(struct pipe_video_context *vpipe, + enum pipe_format format, - unsigned usage, - unsigned geom); ++ unsigned usage); + + void (*destroy)(struct pipe_video_context *vpipe); + + struct pipe_surface *(*create_surface)(struct pipe_video_context *vpipe, + struct pipe_resource *resource, + const struct pipe_surface *templat); + + /** + * Picture decoding and displaying + */ + /*@{*/ + void (*decode_bitstream)(struct pipe_video_context *vpipe, + unsigned num_bufs, + struct pipe_buffer **bitstream_buf); + + void (*decode_macroblocks)(struct pipe_video_context *vpipe, + struct pipe_surface *past, + struct pipe_surface *future, + unsigned num_macroblocks, + struct pipe_macroblock *macroblocks, + struct pipe_fence_handle **fence); + + void (*render_picture)(struct pipe_video_context *vpipe, + struct pipe_surface *src_surface, + enum pipe_mpeg12_picture_type picture_type, + /*unsigned num_past_surfaces, + struct pipe_surface *past_surfaces, + unsigned num_future_surfaces, + struct pipe_surface *future_surfaces,*/ + struct pipe_video_rect *src_area, + struct pipe_surface *dst_surface, + struct pipe_video_rect *dst_area, + struct pipe_fence_handle **fence); + + void (*clear_render_target)(struct pipe_video_context *vpipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + const float *rgba, + unsigned width, unsigned height); + + void (*resource_copy_region)(struct pipe_video_context *vpipe, + struct pipe_resource *dst, + unsigned dstx, unsigned dsty, unsigned dstz, + struct pipe_resource *src, + unsigned srcx, unsigned srcy, unsigned srcz, + unsigned width, unsigned height); + + struct pipe_transfer *(*get_transfer)(struct pipe_video_context *vpipe, + struct pipe_resource *resource, + unsigned level, + unsigned usage, /* a combination of PIPE_TRANSFER_x */ + const struct pipe_box *box); + + void (*transfer_destroy)(struct pipe_video_context *vpipe, + struct pipe_transfer *transfer); + + void* (*transfer_map)(struct pipe_video_context *vpipe, + struct pipe_transfer *transfer); + + void (*transfer_flush_region)(struct pipe_video_context *vpipe, + struct pipe_transfer *transfer, + const struct pipe_box *box); + + void (*transfer_unmap)(struct pipe_video_context *vpipe, + struct pipe_transfer *transfer); + + void (*transfer_inline_write)(struct pipe_video_context *vpipe, + struct pipe_resource *resource, + unsigned level, + unsigned usage, /* a combination of PIPE_TRANSFER_x */ + const struct pipe_box *box, + const void *data, + unsigned stride, + unsigned slice_stride); + + /*@}*/ + + /** + * Parameter-like states (or properties) + */ + /*@{*/ + void (*set_picture_background)(struct pipe_video_context *vpipe, + struct pipe_surface *bg, + struct pipe_video_rect *bg_src_rect); + + void (*set_picture_layers)(struct pipe_video_context *vpipe, + struct pipe_surface *layers[], + struct pipe_video_rect *src_rects[], + struct pipe_video_rect *dst_rects[], + unsigned num_layers); + + void (*set_picture_desc)(struct pipe_video_context *vpipe, + const struct pipe_picture_desc *desc); + + void (*set_decode_target)(struct pipe_video_context *vpipe, + struct pipe_surface *dt); + + void (*set_csc_matrix)(struct pipe_video_context *vpipe, const float *mat); + + /* TODO: Interface for scaling modes, post-processing, etc. */ + /*@}*/ +}; + + +#ifdef __cplusplus +} +#endif + +#endif /* PIPE_VIDEO_CONTEXT_H */ diff --cc src/gallium/state_trackers/vdpau/decode.c index 8bc54d7fc7d,00000000000..048cfe819f0 mode 100644,000000..100644 --- a/src/gallium/state_trackers/vdpau/decode.c +++ b/src/gallium/state_trackers/vdpau/decode.c @@@ -1,303 -1,0 +1,294 @@@ +/************************************************************************** + * + * Copyright 2010 Thomas Balling Sørensen. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "vdpau_private.h" +#include "mpeg2_bitstream_parser.h" +#include +#include +#include +#include + +VdpStatus +vlVdpDecoderCreate(VdpDevice device, + VdpDecoderProfile profile, + uint32_t width, uint32_t height, + uint32_t max_references, + VdpDecoder *decoder) +{ + enum pipe_video_profile p_profile = PIPE_VIDEO_PROFILE_UNKNOWN; + VdpStatus ret = VDP_STATUS_OK; + vlVdpDecoder *vldecoder = NULL; + + debug_printf("[VDPAU] Creating decoder\n"); + + if (!decoder) + return VDP_STATUS_INVALID_POINTER; + + if (!(width && height)) + return VDP_STATUS_INVALID_VALUE; + + vlVdpDevice *dev = vlGetDataHTAB(device); + if (!dev) { + ret = VDP_STATUS_INVALID_HANDLE; + goto inv_device; + } + + vldecoder = CALLOC(1,sizeof(vlVdpDecoder)); + if (!vldecoder) { + ret = VDP_STATUS_RESOURCES; + goto no_decoder; + } + + p_profile = ProfileToPipe(profile); + if (p_profile == PIPE_VIDEO_PROFILE_UNKNOWN) { + ret = VDP_STATUS_INVALID_DECODER_PROFILE; + goto inv_profile; + } + + // TODO: Define max_references. Used mainly for H264 + + vldecoder->profile = p_profile; + vldecoder->height = height; + vldecoder->width = width; + vldecoder->device = dev; + vldecoder->vctx = NULL; + + *decoder = vlAddDataHTAB(vldecoder); + if (*decoder == 0) { + ret = VDP_STATUS_ERROR; + goto no_handle; + } + debug_printf("[VDPAU] Decoder created succesfully\n"); + + return VDP_STATUS_OK; + +no_handle: + FREE(vldecoder); + inv_profile: +no_screen: +no_decoder: +inv_device: + return ret; +} + +VdpStatus +vlVdpDecoderDestroy(VdpDecoder decoder) +{ + debug_printf("[VDPAU] Destroying decoder\n"); + vlVdpDecoder *vldecoder; + + vldecoder = (vlVdpDecoder *)vlGetDataHTAB(decoder); + if (!vldecoder) { + return VDP_STATUS_INVALID_HANDLE; + } + + if (vldecoder->vctx) { + if (vldecoder->vctx->vscreen) + vl_screen_destroy(vldecoder->vctx->vscreen); + } + + if (vldecoder->vctx) + vl_video_destroy(vldecoder->vctx); + + FREE(vldecoder); + + return VDP_STATUS_OK; +} + +VdpStatus +vlVdpCreateSurfaceTarget(vlVdpDecoder *vldecoder, vlVdpSurface *vlsurf) +{ + struct pipe_surface surf_template; + struct pipe_resource tmplt; + struct pipe_resource *surf_tex; + struct pipe_video_context *vctx; + + debug_printf("[VDPAU] Creating surface\n"); + + if(!(vldecoder && vlsurf)) + return VDP_STATUS_INVALID_POINTER; + + vctx = vldecoder->vctx->vpipe; + ++ if (!vctx->is_format_supported(vctx, tmplt.format, PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET)) ++ return VDP_STATUS_RESOURCES; ++ + memset(&tmplt, 0, sizeof(struct pipe_resource)); + tmplt.target = PIPE_TEXTURE_2D; + tmplt.format = vctx->get_param(vctx,PIPE_CAP_DECODE_TARGET_PREFERRED_FORMAT); + tmplt.last_level = 0; - - if (vctx->is_format_supported(vctx, tmplt.format, - PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET, - PIPE_TEXTURE_GEOM_NON_POWER_OF_TWO)) { - tmplt.width0 = vlsurf->width; - tmplt.height0 = vlsurf->height; - } else { - assert(vctx->is_format_supported(vctx, tmplt.format, - PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET, - PIPE_TEXTURE_GEOM_NON_SQUARE)); - tmplt.width0 = util_next_power_of_two(vlsurf->width); - tmplt.height0 = util_next_power_of_two(vlsurf->height); - } - ++ tmplt.width0 = vlsurf->width; ++ tmplt.height0 = vlsurf->height; + tmplt.depth0 = 1; + tmplt.usage = PIPE_USAGE_DEFAULT; + tmplt.bind = PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET; + tmplt.flags = 0; + + surf_tex = vctx->screen->resource_create(vctx->screen, &tmplt); + + memset(&surf_template, 0, sizeof(surf_template)); + surf_template.format = surf_tex->format; + surf_template.usage = PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET; + vlsurf->psurface = vctx->create_surface(vctx->screen, surf_tex, &surf_template); + + pipe_resource_reference(&surf_tex, NULL); + + if (!vlsurf->psurface) + return VDP_STATUS_RESOURCES; + debug_printf("[VDPAU] Done creating surface\n"); + + return VDP_STATUS_OK; +} + +VdpStatus +vlVdpDecoderRenderMpeg2(vlVdpDecoder *vldecoder, + vlVdpSurface *vlsurf, + VdpPictureInfoMPEG1Or2 *picture_info, + uint32_t bitstream_buffer_count, + VdpBitstreamBuffer const *bitstream_buffers) +{ + struct pipe_video_context *vpipe; + vlVdpSurface *t_vdp_surf; + vlVdpSurface *p_vdp_surf; + vlVdpSurface *f_vdp_surf; + struct pipe_surface *t_surf; + struct pipe_surface *p_surf; + struct pipe_surface *f_surf; + uint32_t num_macroblocks; + struct pipe_mpeg12_macroblock *pipe_macroblocks; + VdpStatus ret; + + debug_printf("[VDPAU] Decoding MPEG2\n"); + + t_vdp_surf = vlsurf; + + /* if surfaces equals VDP_STATUS_INVALID_HANDLE, they are not used */ + if (picture_info->backward_reference == VDP_INVALID_HANDLE) + p_vdp_surf = NULL; + else { + p_vdp_surf = (vlVdpSurface *)vlGetDataHTAB(picture_info->backward_reference); + if (!p_vdp_surf) + return VDP_STATUS_INVALID_HANDLE; + } + + if (picture_info->forward_reference == VDP_INVALID_HANDLE) + f_vdp_surf = NULL; + else { + f_vdp_surf = (vlVdpSurface *)vlGetDataHTAB(picture_info->forward_reference); + if (!f_vdp_surf) + return VDP_STATUS_INVALID_HANDLE; + } + + if (f_vdp_surf == VDP_INVALID_HANDLE) f_vdp_surf = NULL; + + ret = vlVdpCreateSurfaceTarget(vldecoder,t_vdp_surf); + + vpipe = vldecoder->vctx->vpipe; + + if (vlVdpMPEG2BitstreamToMacroblock(vpipe->screen, bitstream_buffers, bitstream_buffer_count, + &num_macroblocks, &pipe_macroblocks)) + { + debug_printf("[VDPAU] Error in frame-header. Skipping.\n"); + + ret = VDP_STATUS_OK; + goto skip_frame; + } + + vpipe->set_decode_target(vpipe,t_surf); + vpipe->decode_macroblocks(vpipe, p_surf, f_surf, num_macroblocks, + (struct pipe_macroblock *)pipe_macroblocks, NULL); + + skip_frame: + return ret; +} + +VdpStatus +vlVdpDecoderRender(VdpDecoder decoder, + VdpVideoSurface target, + VdpPictureInfo const *picture_info, + uint32_t bitstream_buffer_count, + VdpBitstreamBuffer const *bitstream_buffers) +{ + vlVdpDecoder *vldecoder; + vlVdpSurface *vlsurf; + struct vl_screen *vscreen; + VdpStatus ret; + + debug_printf("[VDPAU] Decoding\n"); + + if (!(picture_info && bitstream_buffers)) + return VDP_STATUS_INVALID_POINTER; + + vldecoder = (vlVdpDecoder *)vlGetDataHTAB(decoder); + if (!vldecoder) + return VDP_STATUS_INVALID_HANDLE; + + vlsurf = (vlVdpSurface *)vlGetDataHTAB(target); + if (!vlsurf) + return VDP_STATUS_INVALID_HANDLE; + + if (vlsurf->device != vldecoder->device) + return VDP_STATUS_HANDLE_DEVICE_MISMATCH; + + /* Test doesn't make sence */ + /*if (vlsurf->chroma_format != vldecoder->chroma_format) + return VDP_STATUS_INVALID_CHROMA_TYPE;*/ + + vscreen = vl_screen_create(vldecoder->device->display, vldecoder->device->screen); + if (!vscreen) + return VDP_STATUS_RESOURCES; + + vldecoder->vctx = vl_video_create(vscreen, vldecoder->profile, vlsurf->chroma_format, vldecoder->width, vldecoder->height); + if (!vldecoder->vctx) + return VDP_STATUS_RESOURCES; + + // TODO: Right now only mpeg2 is supported. + switch (vldecoder->vctx->vpipe->profile) { + case PIPE_VIDEO_PROFILE_MPEG2_SIMPLE: + case PIPE_VIDEO_PROFILE_MPEG2_MAIN: + ret = vlVdpDecoderRenderMpeg2(vldecoder,vlsurf,(VdpPictureInfoMPEG1Or2 *)picture_info, + bitstream_buffer_count,bitstream_buffers); + break; + default: + return VDP_STATUS_INVALID_DECODER_PROFILE; + } + assert(0); + + return ret; +} + +VdpStatus +vlVdpGenerateCSCMatrix(VdpProcamp *procamp, + VdpColorStandard standard, + VdpCSCMatrix *csc_matrix) +{ + debug_printf("[VDPAU] Generating CSCMatrix\n"); + if (!(csc_matrix && procamp)) + return VDP_STATUS_INVALID_POINTER; + + return VDP_STATUS_OK; +} diff --cc src/gallium/state_trackers/vdpau/query.c index 74a3bd925cf,00000000000..e971b6dc02e mode 100644,000000..100644 --- a/src/gallium/state_trackers/vdpau/query.c +++ b/src/gallium/state_trackers/vdpau/query.c @@@ -1,282 -1,0 +1,281 @@@ +/************************************************************************** + * + * Copyright 2010 Younes Manton. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "vdpau_private.h" +#include +#include +#include +#include +#include +#include + + +VdpStatus +vlVdpGetApiVersion(uint32_t *api_version) +{ + if (!api_version) + return VDP_STATUS_INVALID_POINTER; + + *api_version = 1; + return VDP_STATUS_OK; +} + +VdpStatus +vlVdpGetInformationString(char const **information_string) +{ + if (!information_string) + return VDP_STATUS_INVALID_POINTER; + + *information_string = INFORMATION_STRING; + return VDP_STATUS_OK; +} + +VdpStatus +vlVdpVideoSurfaceQueryCapabilities(VdpDevice device, VdpChromaType surface_chroma_type, + VdpBool *is_supported, uint32_t *max_width, uint32_t *max_height) +{ + struct vl_screen *vlscreen; + uint32_t max_2d_texture_level; + VdpStatus ret; + + debug_printf("[VDPAU] Querying video surfaces\n"); + + if (!(is_supported && max_width && max_height)) + return VDP_STATUS_INVALID_POINTER; + + vlVdpDevice *dev = vlGetDataHTAB(device); + if (!dev) + return VDP_STATUS_INVALID_HANDLE; + + vlscreen = vl_screen_create(dev->display, dev->screen); + if (!vlscreen) + return VDP_STATUS_RESOURCES; + + /* XXX: Current limits */ + *is_supported = true; + if (surface_chroma_type != VDP_CHROMA_TYPE_420) { + *is_supported = false; + goto no_sup; + } + + max_2d_texture_level = vlscreen->pscreen->get_param( vlscreen->pscreen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS ); + if (!max_2d_texture_level) { + ret = VDP_STATUS_RESOURCES; + goto no_sup; + } + + /* I am not quite sure if it is max_2d_texture_level-1 or just max_2d_texture_level */ + *max_width = *max_height = pow(2,max_2d_texture_level-1); + + vl_screen_destroy(vlscreen); + + return VDP_STATUS_OK; + no_sup: + return ret; +} + +VdpStatus +vlVdpVideoSurfaceQueryGetPutBitsYCbCrCapabilities(VdpDevice device, VdpChromaType surface_chroma_type, + VdpYCbCrFormat bits_ycbcr_format, + VdpBool *is_supported) +{ + struct vl_screen *vlscreen; + + debug_printf("[VDPAU] Querying get put video surfaces\n"); + + if (!is_supported) + return VDP_STATUS_INVALID_POINTER; + + vlVdpDevice *dev = vlGetDataHTAB(device); + if (!dev) + return VDP_STATUS_INVALID_HANDLE; + + vlscreen = vl_screen_create(dev->display, dev->screen); + if (!vlscreen) + return VDP_STATUS_RESOURCES; + + if (bits_ycbcr_format != VDP_YCBCR_FORMAT_Y8U8V8A8 && bits_ycbcr_format != VDP_YCBCR_FORMAT_V8U8Y8A8) + *is_supported = vlscreen->pscreen->is_format_supported(vlscreen->pscreen, + FormatToPipe(bits_ycbcr_format), + PIPE_TEXTURE_2D, + 1, - PIPE_BIND_RENDER_TARGET, - PIPE_TEXTURE_GEOM_NON_SQUARE); ++ PIPE_BIND_RENDER_TARGET); + + vl_screen_destroy(vlscreen); + + return VDP_STATUS_OK; +} + +VdpStatus +vlVdpDecoderQueryCapabilities(VdpDevice device, VdpDecoderProfile profile, + VdpBool *is_supported, uint32_t *max_level, uint32_t *max_macroblocks, + uint32_t *max_width, uint32_t *max_height) +{ + enum pipe_video_profile p_profile; + uint32_t max_decode_width; + uint32_t max_decode_height; + uint32_t max_2d_texture_level; + struct vl_screen *vlscreen; + + debug_printf("[VDPAU] Querying decoder\n"); + + if (!(is_supported && max_level && max_macroblocks && max_width && max_height)) + return VDP_STATUS_INVALID_POINTER; + + vlVdpDevice *dev = vlGetDataHTAB(device); + if (!dev) + return VDP_STATUS_INVALID_HANDLE; + + vlscreen = vl_screen_create(dev->display, dev->screen); + if (!vlscreen) + return VDP_STATUS_RESOURCES; + + p_profile = ProfileToPipe(profile); + if (p_profile == PIPE_VIDEO_PROFILE_UNKNOWN) { + *is_supported = false; + return VDP_STATUS_OK; + } + + if (p_profile != PIPE_VIDEO_PROFILE_MPEG2_SIMPLE && p_profile != PIPE_VIDEO_PROFILE_MPEG2_MAIN) { + *is_supported = false; + return VDP_STATUS_OK; + } + + /* XXX hack, need to implement something more sane when the decoders have been implemented */ + max_2d_texture_level = vlscreen->pscreen->get_param( vlscreen->pscreen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS ); + max_decode_width = max_decode_height = pow(2,max_2d_texture_level-2); + if (!(max_decode_width && max_decode_height)) + return VDP_STATUS_RESOURCES; + + *is_supported = true; + *max_width = max_decode_width; + *max_height = max_decode_height; + *max_level = 16; + *max_macroblocks = (max_decode_width/16) * (max_decode_height/16); + + vl_screen_destroy(vlscreen); + + return VDP_STATUS_OK; +} + +VdpStatus +vlVdpOutputSurfaceQueryCapabilities(VdpDevice device, VdpRGBAFormat surface_rgba_format, + VdpBool *is_supported, uint32_t *max_width, uint32_t *max_height) +{ + if (!(is_supported && max_width && max_height)) + return VDP_STATUS_INVALID_POINTER; + + debug_printf("[VDPAU] Querying ouput surfaces\n"); + + return VDP_STATUS_NO_IMPLEMENTATION; +} + +VdpStatus +vlVdpOutputSurfaceQueryGetPutBitsNativeCapabilities(VdpDevice device, VdpRGBAFormat surface_rgba_format, + VdpBool *is_supported) +{ + debug_printf("[VDPAU] Querying output surfaces get put native cap\n"); + + if (!is_supported) + return VDP_STATUS_INVALID_POINTER; + + return VDP_STATUS_NO_IMPLEMENTATION; +} + +VdpStatus +vlVdpOutputSurfaceQueryPutBitsYCbCrCapabilities(VdpDevice device, VdpRGBAFormat surface_rgba_format, + VdpYCbCrFormat bits_ycbcr_format, + VdpBool *is_supported) +{ + debug_printf("[VDPAU] Querying output surfaces put ycrcb cap\n"); + if (!is_supported) + return VDP_STATUS_INVALID_POINTER; + + return VDP_STATUS_NO_IMPLEMENTATION; +} + +VdpStatus +vlVdpBitmapSurfaceQueryCapabilities(VdpDevice device, VdpRGBAFormat surface_rgba_format, + VdpBool *is_supported, uint32_t *max_width, uint32_t *max_height) +{ + debug_printf("[VDPAU] Querying bitmap surfaces\n"); + if (!(is_supported && max_width && max_height)) + return VDP_STATUS_INVALID_POINTER; + + return VDP_STATUS_NO_IMPLEMENTATION; +} + +VdpStatus +vlVdpVideoMixerQueryFeatureSupport(VdpDevice device, VdpVideoMixerFeature feature, + VdpBool *is_supported) +{ + debug_printf("[VDPAU] Querying mixer feature support\n"); + if (!is_supported) + return VDP_STATUS_INVALID_POINTER; + + return VDP_STATUS_NO_IMPLEMENTATION; +} + +VdpStatus +vlVdpVideoMixerQueryParameterSupport(VdpDevice device, VdpVideoMixerParameter parameter, + VdpBool *is_supported) +{ + if (!is_supported) + return VDP_STATUS_INVALID_POINTER; + + return VDP_STATUS_NO_IMPLEMENTATION; +} + +VdpStatus +vlVdpVideoMixerQueryParameterValueRange(VdpDevice device, VdpVideoMixerParameter parameter, + void *min_value, void *max_value) +{ + if (!(min_value && max_value)) + return VDP_STATUS_INVALID_POINTER; + + return VDP_STATUS_NO_IMPLEMENTATION; +} + +VdpStatus +vlVdpVideoMixerQueryAttributeSupport(VdpDevice device, VdpVideoMixerAttribute attribute, + VdpBool *is_supported) +{ + if (!is_supported) + return VDP_STATUS_INVALID_POINTER; + + return VDP_STATUS_NO_IMPLEMENTATION; +} + +VdpStatus +vlVdpVideoMixerQueryAttributeValueRange(VdpDevice device, VdpVideoMixerAttribute attribute, + void *min_value, void *max_value) +{ + if (!(min_value && max_value)) + return VDP_STATUS_INVALID_POINTER; + + return VDP_STATUS_NO_IMPLEMENTATION; +} diff --cc src/gallium/state_trackers/xorg/xvmc/surface.c index c90ad409c10,00000000000..799dac2efc8 mode 100644,000000..100644 --- a/src/gallium/state_trackers/xorg/xvmc/surface.c +++ b/src/gallium/state_trackers/xorg/xvmc/surface.c @@@ -1,527 -1,0 +1,521 @@@ +/************************************************************************** + * + * Copyright 2009 Younes Manton. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "xvmc_private.h" + +static enum pipe_mpeg12_macroblock_type TypeToPipe(int xvmc_mb_type) +{ + if (xvmc_mb_type & XVMC_MB_TYPE_INTRA) + return PIPE_MPEG12_MACROBLOCK_TYPE_INTRA; + if ((xvmc_mb_type & (XVMC_MB_TYPE_MOTION_FORWARD | XVMC_MB_TYPE_MOTION_BACKWARD)) == XVMC_MB_TYPE_MOTION_FORWARD) + return PIPE_MPEG12_MACROBLOCK_TYPE_FWD; + if ((xvmc_mb_type & (XVMC_MB_TYPE_MOTION_FORWARD | XVMC_MB_TYPE_MOTION_BACKWARD)) == XVMC_MB_TYPE_MOTION_BACKWARD) + return PIPE_MPEG12_MACROBLOCK_TYPE_BKWD; + if ((xvmc_mb_type & (XVMC_MB_TYPE_MOTION_FORWARD | XVMC_MB_TYPE_MOTION_BACKWARD)) == (XVMC_MB_TYPE_MOTION_FORWARD | XVMC_MB_TYPE_MOTION_BACKWARD)) + return PIPE_MPEG12_MACROBLOCK_TYPE_BI; + + assert(0); + + XVMC_MSG(XVMC_ERR, "[XvMC] Unrecognized mb type 0x%08X.\n", xvmc_mb_type); + + return -1; +} + +static enum pipe_mpeg12_picture_type PictureToPipe(int xvmc_pic) +{ + switch (xvmc_pic) { + case XVMC_TOP_FIELD: + return PIPE_MPEG12_PICTURE_TYPE_FIELD_TOP; + case XVMC_BOTTOM_FIELD: + return PIPE_MPEG12_PICTURE_TYPE_FIELD_BOTTOM; + case XVMC_FRAME_PICTURE: + return PIPE_MPEG12_PICTURE_TYPE_FRAME; + default: + assert(0); + } + + XVMC_MSG(XVMC_ERR, "[XvMC] Unrecognized picture type 0x%08X.\n", xvmc_pic); + + return -1; +} + +static enum pipe_mpeg12_motion_type MotionToPipe(int xvmc_motion_type, unsigned int xvmc_picture_structure) +{ + switch (xvmc_motion_type) { + case XVMC_PREDICTION_FRAME: + if (xvmc_picture_structure == XVMC_FRAME_PICTURE) + return PIPE_MPEG12_MOTION_TYPE_FRAME; + else + return PIPE_MPEG12_MOTION_TYPE_16x8; + break; + case XVMC_PREDICTION_FIELD: + return PIPE_MPEG12_MOTION_TYPE_FIELD; + case XVMC_PREDICTION_DUAL_PRIME: + return PIPE_MPEG12_MOTION_TYPE_DUALPRIME; + default: + assert(0); + } + + XVMC_MSG(XVMC_ERR, "[XvMC] Unrecognized motion type 0x%08X (with picture structure 0x%08X).\n", xvmc_motion_type, xvmc_picture_structure); + + return -1; +} + +#if 0 +static bool +CreateOrResizeBackBuffer(struct vl_context *vctx, unsigned int width, unsigned int height, + struct pipe_surface **backbuffer) +{ + struct pipe_video_context *vpipe; + struct pipe_resource template; + struct pipe_resource *tex; + + assert(vctx); + + vpipe = vctx->vpipe; + + if (*backbuffer) { + if ((*backbuffer)->width != width || (*backbuffer)->height != height) + pipe_surface_reference(backbuffer, NULL); + else + return true; + } + + memset(&template, 0, sizeof(struct pipe_resource)); + template.target = PIPE_TEXTURE_2D; + template.format = vctx->vscreen->format; + template.last_level = 0; + template.width0 = width; + template.height0 = height; + template.depth0 = 1; + template.array_size = 1; + template.usage = PIPE_USAGE_DEFAULT; + template.bind = PIPE_BIND_RENDER_TARGET | PIPE_BIND_DISPLAY_TARGET | PIPE_BIND_BLIT_SOURCE; + template.flags = 0; + + tex = vpipe->screen->resource_create(vpipe->screen, &template); + if (!tex) + return false; + + *backbuffer = vpipe->screen->get_tex_surface(vpipe->screen, tex, 0, 0, 0, + template.bind); + pipe_resource_reference(&tex, NULL); + + if (!*backbuffer) + return false; + + /* Clear the backbuffer in case the video doesn't cover the whole window */ + /* FIXME: Need to clear every time a frame moves and leaves dirty rects */ + vpipe->surface_fill(vpipe, *backbuffer, 0, 0, width, height, 0); + + return true; +} +#endif + +static void +MacroBlocksToPipe(struct pipe_screen *screen, + unsigned int xvmc_picture_structure, + const XvMCMacroBlockArray *xvmc_macroblocks, + const XvMCBlockArray *xvmc_blocks, + unsigned int first_macroblock, + unsigned int num_macroblocks, + struct pipe_mpeg12_macroblock *pipe_macroblocks) +{ + unsigned int i, j, k, l; + XvMCMacroBlock *xvmc_mb; + + assert(xvmc_macroblocks); + assert(xvmc_blocks); + assert(pipe_macroblocks); + assert(num_macroblocks); + + xvmc_mb = xvmc_macroblocks->macro_blocks + first_macroblock; + + for (i = 0; i < num_macroblocks; ++i) { + pipe_macroblocks->base.codec = PIPE_VIDEO_CODEC_MPEG12; + pipe_macroblocks->mbx = xvmc_mb->x; + pipe_macroblocks->mby = xvmc_mb->y; + pipe_macroblocks->mb_type = TypeToPipe(xvmc_mb->macroblock_type); + if (pipe_macroblocks->mb_type != PIPE_MPEG12_MACROBLOCK_TYPE_INTRA) + pipe_macroblocks->mo_type = MotionToPipe(xvmc_mb->motion_type, xvmc_picture_structure); + /* Get rid of Valgrind 'undefined' warnings */ + else + pipe_macroblocks->mo_type = -1; + pipe_macroblocks->dct_type = xvmc_mb->dct_type == XVMC_DCT_TYPE_FIELD ? + PIPE_MPEG12_DCT_TYPE_FIELD : PIPE_MPEG12_DCT_TYPE_FRAME; + + for (j = 0; j < 2; ++j) + for (k = 0; k < 2; ++k) + for (l = 0; l < 2; ++l) + pipe_macroblocks->pmv[j][k][l] = xvmc_mb->PMV[j][k][l]; + + pipe_macroblocks->mvfs[0][0] = xvmc_mb->motion_vertical_field_select & XVMC_SELECT_FIRST_FORWARD; + pipe_macroblocks->mvfs[0][1] = xvmc_mb->motion_vertical_field_select & XVMC_SELECT_FIRST_BACKWARD; + pipe_macroblocks->mvfs[1][0] = xvmc_mb->motion_vertical_field_select & XVMC_SELECT_SECOND_FORWARD; + pipe_macroblocks->mvfs[1][1] = xvmc_mb->motion_vertical_field_select & XVMC_SELECT_SECOND_BACKWARD; + + pipe_macroblocks->cbp = xvmc_mb->coded_block_pattern; + pipe_macroblocks->blocks = xvmc_blocks->blocks + xvmc_mb->index * BLOCK_SIZE_SAMPLES; + + ++pipe_macroblocks; + ++xvmc_mb; + } +} + +PUBLIC +Status XvMCCreateSurface(Display *dpy, XvMCContext *context, XvMCSurface *surface) +{ + XvMCContextPrivate *context_priv; + struct pipe_video_context *vpipe; + XvMCSurfacePrivate *surface_priv; + struct pipe_resource template; + struct pipe_resource *vsfc_tex; + struct pipe_surface surf_template; + struct pipe_surface *vsfc; + + XVMC_MSG(XVMC_TRACE, "[XvMC] Creating surface %p.\n", surface); + + assert(dpy); + + if (!context) + return XvMCBadContext; + if (!surface) + return XvMCBadSurface; + + context_priv = context->privData; + vpipe = context_priv->vctx->vpipe; + + surface_priv = CALLOC(1, sizeof(XvMCSurfacePrivate)); + if (!surface_priv) + return BadAlloc; + + memset(&template, 0, sizeof(struct pipe_resource)); + template.target = PIPE_TEXTURE_2D; + template.format = (enum pipe_format)vpipe->get_param(vpipe, PIPE_CAP_DECODE_TARGET_PREFERRED_FORMAT); - template.last_level = 0; - if (vpipe->is_format_supported(vpipe, template.format, - PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET, - PIPE_TEXTURE_GEOM_NON_POWER_OF_TWO)) { - template.width0 = context->width; - template.height0 = context->height; - } - else { - assert(vpipe->is_format_supported(vpipe, template.format, - PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET, - PIPE_TEXTURE_GEOM_NON_SQUARE)); - template.width0 = util_next_power_of_two(context->width); - template.height0 = util_next_power_of_two(context->height); ++ if (!vpipe->is_format_supported(vpipe, template.format, ++ PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET)) { ++ FREE(surface_priv); ++ return BadAlloc; + } ++ template.last_level = 0; ++ template.width0 = util_next_power_of_two(context->width); ++ template.height0 = util_next_power_of_two(context->height); + template.depth0 = 1; + template.array_size = 1; + template.usage = PIPE_USAGE_DEFAULT; + template.bind = PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET; + template.flags = 0; + vsfc_tex = vpipe->screen->resource_create(vpipe->screen, &template); + if (!vsfc_tex) { + FREE(surface_priv); + return BadAlloc; + } + + memset(&surf_template, 0, sizeof(surf_template)); + surf_template.format = vsfc_tex->format; + surf_template.usage = PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET; + vsfc = vpipe->create_surface(vpipe, vsfc_tex, &surf_template); + pipe_resource_reference(&vsfc_tex, NULL); + if (!vsfc) { + FREE(surface_priv); + return BadAlloc; + } + + surface_priv->pipe_vsfc = vsfc; + surface_priv->context = context; + + surface->surface_id = XAllocID(dpy); + surface->context_id = context->context_id; + surface->surface_type_id = context->surface_type_id; + surface->width = context->width; + surface->height = context->height; + surface->privData = surface_priv; + + SyncHandle(); + + XVMC_MSG(XVMC_TRACE, "[XvMC] Surface %p created.\n", surface); + + return Success; +} + +PUBLIC +Status XvMCRenderSurface(Display *dpy, XvMCContext *context, unsigned int picture_structure, + XvMCSurface *target_surface, XvMCSurface *past_surface, XvMCSurface *future_surface, + unsigned int flags, unsigned int num_macroblocks, unsigned int first_macroblock, + XvMCMacroBlockArray *macroblocks, XvMCBlockArray *blocks +) +{ + struct pipe_video_context *vpipe; + struct pipe_surface *t_vsfc; + struct pipe_surface *p_vsfc; + struct pipe_surface *f_vsfc; + XvMCContextPrivate *context_priv; + XvMCSurfacePrivate *target_surface_priv; + XvMCSurfacePrivate *past_surface_priv; + XvMCSurfacePrivate *future_surface_priv; + struct pipe_mpeg12_macroblock pipe_macroblocks[num_macroblocks]; + + XVMC_MSG(XVMC_TRACE, "[XvMC] Rendering to surface %p.\n", target_surface); + + assert(dpy); + + if (!context || !context->privData) + return XvMCBadContext; + if (!target_surface || !target_surface->privData) + return XvMCBadSurface; + + if (picture_structure != XVMC_TOP_FIELD && + picture_structure != XVMC_BOTTOM_FIELD && + picture_structure != XVMC_FRAME_PICTURE) + return BadValue; + /* Bkwd pred equivalent to fwd (past && !future) */ + if (future_surface && !past_surface) + return BadMatch; + + assert(context->context_id == target_surface->context_id); + assert(!past_surface || context->context_id == past_surface->context_id); + assert(!future_surface || context->context_id == future_surface->context_id); + + assert(macroblocks); + assert(blocks); + + assert(macroblocks->context_id == context->context_id); + assert(blocks->context_id == context->context_id); + + assert(flags == 0 || flags == XVMC_SECOND_FIELD); + + target_surface_priv = target_surface->privData; + past_surface_priv = past_surface ? past_surface->privData : NULL; + future_surface_priv = future_surface ? future_surface->privData : NULL; + + assert(target_surface_priv->context == context); + assert(!past_surface || past_surface_priv->context == context); + assert(!future_surface || future_surface_priv->context == context); + + context_priv = context->privData; + vpipe = context_priv->vctx->vpipe; + + t_vsfc = target_surface_priv->pipe_vsfc; + p_vsfc = past_surface ? past_surface_priv->pipe_vsfc : NULL; + f_vsfc = future_surface ? future_surface_priv->pipe_vsfc : NULL; + + MacroBlocksToPipe(vpipe->screen, picture_structure, macroblocks, blocks, first_macroblock, + num_macroblocks, pipe_macroblocks); + + vpipe->set_decode_target(vpipe, t_vsfc); + vpipe->decode_macroblocks(vpipe, p_vsfc, f_vsfc, num_macroblocks, + &pipe_macroblocks->base, &target_surface_priv->render_fence); + + XVMC_MSG(XVMC_TRACE, "[XvMC] Submitted surface %p for rendering.\n", target_surface); + + return Success; +} + +PUBLIC +Status XvMCFlushSurface(Display *dpy, XvMCSurface *surface) +{ + assert(dpy); + + if (!surface) + return XvMCBadSurface; + + return Success; +} + +PUBLIC +Status XvMCSyncSurface(Display *dpy, XvMCSurface *surface) +{ + assert(dpy); + + if (!surface) + return XvMCBadSurface; + + return Success; +} + +PUBLIC +Status XvMCPutSurface(Display *dpy, XvMCSurface *surface, Drawable drawable, + short srcx, short srcy, unsigned short srcw, unsigned short srch, + short destx, short desty, unsigned short destw, unsigned short desth, + int flags) +{ + static int dump_window = -1; + + struct pipe_video_context *vpipe; + XvMCSurfacePrivate *surface_priv; + XvMCContextPrivate *context_priv; + XvMCSubpicturePrivate *subpicture_priv; + XvMCContext *context; + struct pipe_video_rect src_rect = {srcx, srcy, srcw, srch}; + struct pipe_video_rect dst_rect = {destx, desty, destw, desth}; + struct pipe_surface *drawable_surface; + + XVMC_MSG(XVMC_TRACE, "[XvMC] Displaying surface %p.\n", surface); + + assert(dpy); + + if (!surface || !surface->privData) + return XvMCBadSurface; + + surface_priv = surface->privData; + context = surface_priv->context; + context_priv = context->privData; + + drawable_surface = vl_drawable_surface_get(context_priv->vctx, drawable); + if (!drawable_surface) + return BadDrawable; + + assert(flags == XVMC_TOP_FIELD || flags == XVMC_BOTTOM_FIELD || flags == XVMC_FRAME_PICTURE); + assert(srcx + srcw - 1 < surface->width); + assert(srcy + srch - 1 < surface->height); + /* + * Some apps (mplayer) hit these asserts because they call + * this function after the window has been resized by the WM + * but before they've handled the corresponding XEvent and + * know about the new dimensions. The output should be clipped + * until the app updates destw and desth. + */ + /* + assert(destx + destw - 1 < drawable_surface->width); + assert(desty + desth - 1 < drawable_surface->height); + */ + + subpicture_priv = surface_priv->subpicture ? surface_priv->subpicture->privData : NULL; + vpipe = context_priv->vctx->vpipe; + +#if 0 + if (!CreateOrResizeBackBuffer(context_priv->vctx, width, height, &context_priv->backbuffer)) + return BadAlloc; +#endif + + if (subpicture_priv) { + struct pipe_video_rect src_rect = {surface_priv->subx, surface_priv->suby, surface_priv->subw, surface_priv->subh}; + struct pipe_video_rect dst_rect = {surface_priv->surfx, surface_priv->surfy, surface_priv->surfw, surface_priv->surfh}; + struct pipe_video_rect *src_rects[1] = {&src_rect}; + struct pipe_video_rect *dst_rects[1] = {&dst_rect}; + + XVMC_MSG(XVMC_TRACE, "[XvMC] Surface %p has subpicture %p.\n", surface, surface_priv->subpicture); + + assert(subpicture_priv->surface == surface); + vpipe->set_picture_layers(vpipe, &subpicture_priv->sfc, src_rects, dst_rects, 1); + + surface_priv->subpicture = NULL; + subpicture_priv->surface = NULL; + } + else + vpipe->set_picture_layers(vpipe, NULL, NULL, NULL, 0); + + vpipe->render_picture(vpipe, surface_priv->pipe_vsfc, PictureToPipe(flags), &src_rect, + drawable_surface, &dst_rect, &surface_priv->disp_fence); + + XVMC_MSG(XVMC_TRACE, "[XvMC] Submitted surface %p for display. Pushing to front buffer.\n", surface); + + vpipe->screen->flush_frontbuffer + ( + vpipe->screen, + drawable_surface->texture, + 0, 0, + vl_contextprivate_get(context_priv->vctx, drawable_surface) + ); + + pipe_surface_reference(&drawable_surface, NULL); + + if(dump_window == -1) { + dump_window = debug_get_num_option("XVMC_DUMP", 0); + } + + if(dump_window) { + static unsigned int framenum = 0; + char cmd[256]; + sprintf(cmd, "xwd -id %d -out xvmc_frame_%08d.xwd", (int)drawable, ++framenum); + system(cmd); + } + + XVMC_MSG(XVMC_TRACE, "[XvMC] Pushed surface %p to front buffer.\n", surface); + + return Success; +} + +PUBLIC +Status XvMCGetSurfaceStatus(Display *dpy, XvMCSurface *surface, int *status) +{ + assert(dpy); + + if (!surface) + return XvMCBadSurface; + + assert(status); + + *status = 0; + + return Success; +} + +PUBLIC +Status XvMCDestroySurface(Display *dpy, XvMCSurface *surface) +{ + XvMCSurfacePrivate *surface_priv; + + XVMC_MSG(XVMC_TRACE, "[XvMC] Destroying surface %p.\n", surface); + + assert(dpy); + + if (!surface || !surface->privData) + return XvMCBadSurface; + + surface_priv = surface->privData; + pipe_surface_reference(&surface_priv->pipe_vsfc, NULL); + FREE(surface_priv); + surface->privData = NULL; + + XVMC_MSG(XVMC_TRACE, "[XvMC] Surface %p destroyed.\n", surface); + + return Success; +} + +PUBLIC +Status XvMCHideSurface(Display *dpy, XvMCSurface *surface) +{ + assert(dpy); + + if (!surface || !surface->privData) + return XvMCBadSurface; + + /* No op, only for overlaid rendering */ + + return Success; +} diff --cc src/mesa/state_tracker/st_format.c index 22a1450cfe7,03655ee0483..3d8637ab37e --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@@ -397,6 -407,15 +407,15 @@@ st_pipe_format_to_mesa_format(enum pipe case PIPE_FORMAT_RGTC2_SNORM: return MESA_FORMAT_SIGNED_RG_RGTC2; + case PIPE_FORMAT_LATC1_UNORM: + return MESA_FORMAT_L_LATC1; + case PIPE_FORMAT_LATC1_SNORM: + return MESA_FORMAT_SIGNED_L_LATC1; + case PIPE_FORMAT_LATC2_UNORM: + return MESA_FORMAT_LA_LATC2; - case PIPE_FORMAT_LATC2_SNORM: - return MESA_FORMAT_SIGNED_LA_LATC2; ++ //case PIPE_FORMAT_LATC2_SNORM: ++ // return MESA_FORMAT_SIGNED_LA_LATC2; + default: assert(0); return MESA_FORMAT_NONE;