{
SI_SAVE_TEXTURES = 1,
SI_SAVE_FRAMEBUFFER = 2,
- SI_DISABLE_RENDER_COND = 4,
+ SI_SAVE_FRAGMENT_STATE = 4,
+ SI_DISABLE_RENDER_COND = 8,
- SI_CLEAR = 0,
+ SI_CLEAR = SI_SAVE_FRAGMENT_STATE,
- SI_CLEAR_SURFACE = SI_SAVE_FRAMEBUFFER,
+ SI_CLEAR_SURFACE = SI_SAVE_FRAMEBUFFER | SI_SAVE_FRAGMENT_STATE,
SI_COPY = SI_SAVE_FRAMEBUFFER | SI_SAVE_TEXTURES |
- SI_DISABLE_RENDER_COND,
+ SI_SAVE_FRAGMENT_STATE | SI_DISABLE_RENDER_COND,
- SI_BLIT = SI_SAVE_FRAMEBUFFER | SI_SAVE_TEXTURES,
+ SI_BLIT = SI_SAVE_FRAMEBUFFER | SI_SAVE_TEXTURES |
+ SI_SAVE_FRAGMENT_STATE,
- SI_DECOMPRESS = SI_SAVE_FRAMEBUFFER | SI_DISABLE_RENDER_COND,
+ SI_DECOMPRESS = SI_SAVE_FRAMEBUFFER | SI_SAVE_FRAGMENT_STATE |
+ SI_DISABLE_RENDER_COND,
- SI_COLOR_RESOLVE = SI_SAVE_FRAMEBUFFER
+ SI_COLOR_RESOLVE = SI_SAVE_FRAMEBUFFER | SI_SAVE_FRAGMENT_STATE
};
static void si_blitter_begin(struct pipe_context *ctx, enum si_blitter_op op)
r600_suspend_nontimer_queries(&sctx->b);
- util_blitter_save_blend(sctx->blitter, sctx->queued.named.blend);
- util_blitter_save_depth_stencil_alpha(sctx->blitter, sctx->queued.named.dsa);
- util_blitter_save_stencil_ref(sctx->blitter, &sctx->stencil_ref.state);
- util_blitter_save_rasterizer(sctx->blitter, sctx->queued.named.rasterizer);
- util_blitter_save_fragment_shader(sctx->blitter, sctx->ps_shader.cso);
- util_blitter_save_geometry_shader(sctx->blitter, sctx->gs_shader.cso);
+ util_blitter_save_vertex_buffer_slot(sctx->blitter, sctx->vertex_buffer);
+ util_blitter_save_vertex_elements(sctx->blitter, sctx->vertex_elements);
+ util_blitter_save_vertex_shader(sctx->blitter, sctx->vs_shader.cso);
util_blitter_save_tessctrl_shader(sctx->blitter, sctx->tcs_shader.cso);
util_blitter_save_tesseval_shader(sctx->blitter, sctx->tes_shader.cso);
- util_blitter_save_vertex_shader(sctx->blitter, sctx->vs_shader.cso);
- util_blitter_save_vertex_elements(sctx->blitter, sctx->vertex_elements);
- util_blitter_save_sample_mask(sctx->blitter, sctx->sample_mask.sample_mask);
- util_blitter_save_viewport(sctx->blitter, &sctx->viewports.states[0]);
- util_blitter_save_scissor(sctx->blitter, &sctx->scissors.states[0]);
- util_blitter_save_vertex_buffer_slot(sctx->blitter, sctx->vertex_buffer);
+ util_blitter_save_geometry_shader(sctx->blitter, sctx->gs_shader.cso);
util_blitter_save_so_targets(sctx->blitter, sctx->b.streamout.num_targets,
(struct pipe_stream_output_target**)sctx->b.streamout.targets);
+ util_blitter_save_rasterizer(sctx->blitter, sctx->queued.named.rasterizer);
+
+ if (op & SI_SAVE_FRAGMENT_STATE) {
+ util_blitter_save_blend(sctx->blitter, sctx->queued.named.blend);
+ util_blitter_save_depth_stencil_alpha(sctx->blitter, sctx->queued.named.dsa);
+ util_blitter_save_stencil_ref(sctx->blitter, &sctx->stencil_ref.state);
+ util_blitter_save_fragment_shader(sctx->blitter, sctx->ps_shader.cso);
+ util_blitter_save_sample_mask(sctx->blitter, sctx->sample_mask.sample_mask);
+ util_blitter_save_viewport(sctx->blitter, &sctx->viewports.states[0]);
+ util_blitter_save_scissor(sctx->blitter, &sctx->scissors.states[0]);
+ }
if (op & SI_SAVE_FRAMEBUFFER)
util_blitter_save_framebuffer(sctx->blitter, &sctx->framebuffer.state);
if (op & SI_SAVE_TEXTURES) {
util_blitter_save_fragment_sampler_states(
sctx->blitter, 2,
- sctx->samplers[PIPE_SHADER_FRAGMENT].states.saved_states);
+ sctx->samplers[PIPE_SHADER_FRAGMENT].views.sampler_states);
util_blitter_save_fragment_sampler_views(sctx->blitter, 2,
sctx->samplers[PIPE_SHADER_FRAGMENT].views.views);
}
- if ((op & SI_DISABLE_RENDER_COND) && sctx->b.current_render_cond) {
- util_blitter_save_render_condition(sctx->blitter,
- sctx->b.current_render_cond,
- sctx->b.current_render_cond_cond,
- sctx->b.current_render_cond_mode);
- }
+ if (op & SI_DISABLE_RENDER_COND)
+ sctx->b.render_cond_force_off = true;
}
static void si_blitter_end(struct pipe_context *ctx)
{
struct si_context *sctx = (struct si_context *)ctx;
+
+ sctx->b.render_cond_force_off = false;
r600_resume_nontimer_queries(&sctx->b);
}
si_mark_atom_dirty(sctx, &sctx->db_render_state);
}
-void si_flush_depth_textures(struct si_context *sctx,
- struct si_textures_info *textures)
+static void
+si_flush_depth_textures(struct si_context *sctx,
+ struct si_textures_info *textures)
{
unsigned i;
unsigned mask = textures->depth_texture_mask;
static void si_blit_decompress_color(struct pipe_context *ctx,
struct r600_texture *rtex,
unsigned first_level, unsigned last_level,
- unsigned first_layer, unsigned last_layer)
+ unsigned first_layer, unsigned last_layer,
+ bool need_dcc_decompress)
{
struct si_context *sctx = (struct si_context *)ctx;
unsigned layer, level, checked_last_layer, max_layer;
- if (!rtex->dirty_level_mask)
+ if (!rtex->dirty_level_mask && !need_dcc_decompress)
return;
for (level = first_level; level <= last_level; level++) {
- if (!(rtex->dirty_level_mask & (1 << level)))
+ void* custom_blend;
+
+ if (!(rtex->dirty_level_mask & (1 << level)) && !need_dcc_decompress)
continue;
+ if (rtex->dcc_offset && need_dcc_decompress) {
+ custom_blend = sctx->custom_blend_dcc_decompress;
+ } else if (rtex->fmask.size) {
+ custom_blend = sctx->custom_blend_decompress;
+ } else {
+ custom_blend = sctx->custom_blend_fastclear;
+ }
+
/* The smaller the mipmap level, the less layers there are
* as far as 3D textures are concerned. */
max_layer = util_max_layer(&rtex->resource.b.b, level);
cbsurf = ctx->create_surface(ctx, &rtex->resource.b.b, &surf_tmpl);
si_blitter_begin(ctx, SI_DECOMPRESS);
- util_blitter_custom_color(sctx->blitter, cbsurf,
- rtex->fmask.size ? sctx->custom_blend_decompress :
- sctx->custom_blend_fastclear);
+ util_blitter_custom_color(sctx->blitter, cbsurf, custom_blend);
si_blitter_end(ctx);
pipe_surface_reference(&cbsurf, NULL);
}
}
-void si_decompress_color_textures(struct si_context *sctx,
- struct si_textures_info *textures)
+static void
+si_decompress_sampler_color_textures(struct si_context *sctx,
+ struct si_textures_info *textures)
{
unsigned i;
unsigned mask = textures->compressed_colortex_mask;
assert(view);
tex = (struct r600_texture *)view->texture;
- assert(tex->cmask.size || tex->fmask.size || tex->dcc_buffer);
+ assert(tex->cmask.size || tex->fmask.size || tex->dcc_offset);
si_blit_decompress_color(&sctx->b.b, tex,
view->u.tex.first_level, view->u.tex.last_level,
- 0, util_max_layer(&tex->resource.b.b, view->u.tex.first_level));
+ 0, util_max_layer(&tex->resource.b.b, view->u.tex.first_level),
+ false);
+ }
+}
+
+static void
+si_decompress_image_color_textures(struct si_context *sctx,
+ struct si_images_info *images)
+{
+ unsigned i;
+ unsigned mask = images->compressed_colortex_mask;
+
+ while (mask) {
+ const struct pipe_image_view *view;
+ struct r600_texture *tex;
+
+ i = u_bit_scan(&mask);
+
+ view = &images->views[i];
+ assert(view->resource->target != PIPE_BUFFER);
+
+ tex = (struct r600_texture *)view->resource;
+ if (!tex->cmask.size && !tex->fmask.size && !tex->dcc_offset)
+ continue;
+
+ si_blit_decompress_color(&sctx->b.b, tex,
+ view->u.tex.level, view->u.tex.level,
+ 0, util_max_layer(&tex->resource.b.b, view->u.tex.level),
+ false);
+ }
+}
+
+void si_decompress_textures(struct si_context *sctx)
+{
+ unsigned compressed_colortex_counter;
+
+ if (sctx->blitter->running)
+ return;
+
+ /* Update the compressed_colortex_mask if necessary. */
+ compressed_colortex_counter = p_atomic_read(&sctx->screen->b.compressed_colortex_counter);
+ if (compressed_colortex_counter != sctx->b.last_compressed_colortex_counter) {
+ sctx->b.last_compressed_colortex_counter = compressed_colortex_counter;
+ si_update_compressed_colortex_masks(sctx);
+ }
+
+ /* Flush depth textures which need to be flushed. */
+ for (int i = 0; i < SI_NUM_SHADERS; i++) {
+ if (sctx->samplers[i].depth_texture_mask) {
+ si_flush_depth_textures(sctx, &sctx->samplers[i]);
+ }
+ if (sctx->samplers[i].compressed_colortex_mask) {
+ si_decompress_sampler_color_textures(sctx, &sctx->samplers[i]);
+ }
+ if (sctx->images[i].compressed_colortex_mask) {
+ si_decompress_image_color_textures(sctx, &sctx->images[i]);
+ }
}
}
}
}
- if (buffers & PIPE_CLEAR_DEPTH &&
- zstex && zstex->htile_buffer &&
+ if (zstex && zstex->htile_buffer &&
zsbuf->u.tex.level == 0 &&
zsbuf->u.tex.first_layer == 0 &&
zsbuf->u.tex.last_layer == util_max_layer(&zstex->resource.b.b, 0)) {
- /* Need to disable EXPCLEAR temporarily if clearing
- * to a new value. */
- if (zstex->depth_cleared && zstex->depth_clear_value != depth) {
- sctx->db_depth_disable_expclear = true;
+ if (buffers & PIPE_CLEAR_DEPTH) {
+ /* Need to disable EXPCLEAR temporarily if clearing
+ * to a new value. */
+ if (zstex->depth_cleared && zstex->depth_clear_value != depth) {
+ sctx->db_depth_disable_expclear = true;
+ }
+
+ zstex->depth_clear_value = depth;
+ sctx->framebuffer.dirty_zsbuf = true;
+ si_mark_atom_dirty(sctx, &sctx->framebuffer.atom); /* updates DB_DEPTH_CLEAR */
+ sctx->db_depth_clear = true;
+ si_mark_atom_dirty(sctx, &sctx->db_render_state);
}
- zstex->depth_clear_value = depth;
- sctx->framebuffer.dirty_zsbuf = true;
- si_mark_atom_dirty(sctx, &sctx->framebuffer.atom); /* updates DB_DEPTH_CLEAR */
- sctx->db_depth_clear = true;
- si_mark_atom_dirty(sctx, &sctx->db_render_state);
+ if (buffers & PIPE_CLEAR_STENCIL) {
+ stencil &= 0xff;
+
+ /* Need to disable EXPCLEAR temporarily if clearing
+ * to a new value. */
+ if (zstex->stencil_cleared && zstex->stencil_clear_value != stencil) {
+ sctx->db_stencil_disable_expclear = true;
+ }
+
+ zstex->stencil_clear_value = stencil;
+ sctx->framebuffer.dirty_zsbuf = true;
+ si_mark_atom_dirty(sctx, &sctx->framebuffer.atom); /* updates DB_STENCIL_CLEAR */
+ sctx->db_stencil_clear = true;
+ si_mark_atom_dirty(sctx, &sctx->db_render_state);
+ }
}
si_blitter_begin(ctx, SI_CLEAR);
zstex->depth_cleared = true;
si_mark_atom_dirty(sctx, &sctx->db_render_state);
}
+
+ if (sctx->db_stencil_clear) {
+ sctx->db_stencil_clear = false;
+ sctx->db_stencil_disable_expclear = false;
+ zstex->stencil_cleared = true;
+ si_mark_atom_dirty(sctx, &sctx->db_render_state);
+ }
}
static void si_clear_render_target(struct pipe_context *ctx,
si_blit_decompress_depth_in_place(sctx, rtex, true,
level, level,
first_layer, last_layer);
- } else if (rtex->fmask.size || rtex->cmask.size || rtex->dcc_buffer) {
+ } else if (rtex->fmask.size || rtex->cmask.size || rtex->dcc_offset) {
si_blit_decompress_color(ctx, rtex, level, level,
- first_layer, last_layer);
+ first_layer, last_layer, false);
}
}
enum pipe_format format = int_to_norm_format(info->dst.format);
unsigned sample_mask = ~0;
+ /* Hardware MSAA resolve doesn't work if SPI format = NORM16_ABGR and
+ * the format is R16G16. Use R16A16, which does work.
+ */
+ if (format == PIPE_FORMAT_R16G16_UNORM)
+ format = PIPE_FORMAT_R16A16_UNORM;
+ if (format == PIPE_FORMAT_R16G16_SNORM)
+ format = PIPE_FORMAT_R16A16_SNORM;
+
if (info->src.resource->nr_samples > 1 &&
info->dst.resource->nr_samples <= 1 &&
util_max_layer(info->src.resource, 0) == 0 &&
dst->surface.level[info->dst.level].mode >= RADEON_SURF_MODE_1D &&
!(dst->surface.flags & RADEON_SURF_SCANOUT) &&
(!dst->cmask.size || !dst->dirty_level_mask) && /* dst cannot be fast-cleared */
- !dst->dcc_buffer) {
+ !dst->dcc_offset) {
si_blitter_begin(ctx, SI_COLOR_RESOLVE |
(info->render_condition_enable ? 0 : SI_DISABLE_RENDER_COND));
util_blitter_custom_resolve_color(sctx->blitter,
assert(res->target != PIPE_BUFFER);
- if (!rtex->is_depth && rtex->cmask.size) {
+ if (!rtex->is_depth && (rtex->cmask.size || rtex->dcc_offset)) {
si_blit_decompress_color(ctx, rtex, 0, res->last_level,
- 0, util_max_layer(res, 0));
+ 0, util_max_layer(res, 0), false);
+ }
+}
+
+static void si_decompress_dcc(struct pipe_context *ctx,
+ struct r600_texture *rtex)
+{
+ if (!rtex->dcc_offset)
+ return;
+
+ si_blit_decompress_color(ctx, rtex, 0, rtex->resource.b.b.last_level,
+ 0, util_max_layer(&rtex->resource.b.b, 0),
+ true);
+}
+
+static void si_pipe_clear_buffer(struct pipe_context *ctx,
+ struct pipe_resource *dst,
+ unsigned offset, unsigned size,
+ const void *clear_value_ptr,
+ int clear_value_size)
+{
+ struct si_context *sctx = (struct si_context*)ctx;
+ uint32_t dword_value;
+ unsigned i;
+
+ assert(offset % clear_value_size == 0);
+ assert(size % clear_value_size == 0);
+
+ if (clear_value_size > 4) {
+ const uint32_t *u32 = clear_value_ptr;
+ bool clear_dword_duplicated = true;
+
+ /* See if we can lower large fills to dword fills. */
+ for (i = 1; i < clear_value_size / 4; i++)
+ if (u32[0] != u32[i]) {
+ clear_dword_duplicated = false;
+ break;
+ }
+
+ if (!clear_dword_duplicated) {
+ /* Use transform feedback for 64-bit, 96-bit, and
+ * 128-bit fills.
+ */
+ union pipe_color_union clear_value;
+
+ memcpy(&clear_value, clear_value_ptr, clear_value_size);
+ si_blitter_begin(ctx, SI_DISABLE_RENDER_COND);
+ util_blitter_clear_buffer(sctx->blitter, dst, offset,
+ size, clear_value_size / 4,
+ &clear_value);
+ si_blitter_end(ctx);
+ return;
+ }
}
+
+ /* Expand the clear value to a dword. */
+ switch (clear_value_size) {
+ case 1:
+ dword_value = *(uint8_t*)clear_value_ptr;
+ dword_value |= (dword_value << 8) |
+ (dword_value << 16) |
+ (dword_value << 24);
+ break;
+ case 2:
+ dword_value = *(uint16_t*)clear_value_ptr;
+ dword_value |= dword_value << 16;
+ break;
+ default:
+ dword_value = *(uint32_t*)clear_value_ptr;
+ }
+
+ sctx->b.clear_buffer(ctx, dst, offset, size, dword_value, false);
}
void si_init_blit_functions(struct si_context *sctx)
{
sctx->b.b.clear = si_clear;
+ sctx->b.b.clear_buffer = si_pipe_clear_buffer;
sctx->b.b.clear_render_target = si_clear_render_target;
sctx->b.b.clear_depth_stencil = si_clear_depth_stencil;
sctx->b.b.resource_copy_region = si_resource_copy_region;
sctx->b.b.blit = si_blit;
sctx->b.b.flush_resource = si_flush_resource;
sctx->b.blit_decompress_depth = si_blit_decompress_depth;
+ sctx->b.decompress_dcc = si_decompress_dcc;
}