* OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
* USE OR OTHER DEALINGS IN THE SOFTWARE.
*/
+#include "r600_pipe.h"
#include "util/u_surface.h"
#include "util/u_blitter.h"
#include "util/u_format.h"
-#include "r600_pipe.h"
enum r600_blitter_op /* bitmask */
{
- R600_SAVE_TEXTURES = 1,
- R600_SAVE_FRAMEBUFFER = 2,
- R600_DISABLE_RENDER_COND = 4,
+ R600_SAVE_FRAGMENT_STATE = 1,
+ R600_SAVE_TEXTURES = 2,
+ R600_SAVE_FRAMEBUFFER = 4,
+ R600_DISABLE_RENDER_COND = 8,
+
+ R600_CLEAR = R600_SAVE_FRAGMENT_STATE,
- R600_CLEAR = 0,
+ R600_CLEAR_SURFACE = R600_SAVE_FRAGMENT_STATE | R600_SAVE_FRAMEBUFFER,
- R600_CLEAR_SURFACE = R600_SAVE_FRAMEBUFFER,
+ R600_COPY_BUFFER = R600_DISABLE_RENDER_COND,
- R600_COPY = R600_SAVE_FRAMEBUFFER | R600_SAVE_TEXTURES |
+ R600_COPY_TEXTURE = R600_SAVE_FRAGMENT_STATE | R600_SAVE_FRAMEBUFFER | R600_SAVE_TEXTURES |
R600_DISABLE_RENDER_COND,
- R600_DECOMPRESS = R600_SAVE_FRAMEBUFFER | R600_DISABLE_RENDER_COND,
+ R600_BLIT = R600_SAVE_FRAGMENT_STATE | R600_SAVE_FRAMEBUFFER | R600_SAVE_TEXTURES |
+ R600_DISABLE_RENDER_COND,
+
+ R600_DECOMPRESS = R600_SAVE_FRAGMENT_STATE | R600_SAVE_FRAMEBUFFER | R600_DISABLE_RENDER_COND,
+
+ R600_COLOR_RESOLVE = R600_SAVE_FRAGMENT_STATE | R600_SAVE_FRAMEBUFFER | R600_DISABLE_RENDER_COND
};
static void r600_blitter_begin(struct pipe_context *ctx, enum r600_blitter_op op)
{
struct r600_context *rctx = (struct r600_context *)ctx;
- r600_context_queries_suspend(rctx);
+ r600_suspend_nontimer_queries(rctx);
- util_blitter_save_blend(rctx->blitter, rctx->states[R600_PIPE_STATE_BLEND]);
- util_blitter_save_depth_stencil_alpha(rctx->blitter, rctx->states[R600_PIPE_STATE_DSA]);
- if (rctx->states[R600_PIPE_STATE_STENCIL_REF]) {
- util_blitter_save_stencil_ref(rctx->blitter, &rctx->stencil_ref);
- }
- util_blitter_save_rasterizer(rctx->blitter, rctx->states[R600_PIPE_STATE_RASTERIZER]);
- util_blitter_save_fragment_shader(rctx->blitter, rctx->ps_shader);
+ util_blitter_save_vertex_buffer_slot(rctx->blitter, rctx->vertex_buffer_state.vb);
+ util_blitter_save_vertex_elements(rctx->blitter, rctx->vertex_fetch_shader.cso);
util_blitter_save_vertex_shader(rctx->blitter, rctx->vs_shader);
- util_blitter_save_vertex_elements(rctx->blitter, rctx->vertex_elements);
- if (rctx->states[R600_PIPE_STATE_VIEWPORT]) {
- util_blitter_save_viewport(rctx->blitter, &rctx->viewport);
+ util_blitter_save_so_targets(rctx->blitter, rctx->b.streamout.num_targets,
+ (struct pipe_stream_output_target**)rctx->b.streamout.targets);
+ util_blitter_save_rasterizer(rctx->blitter, rctx->rasterizer_state.cso);
+
+ if (op & R600_SAVE_FRAGMENT_STATE) {
+ util_blitter_save_viewport(rctx->blitter, &rctx->viewport.state);
+ util_blitter_save_scissor(rctx->blitter, &rctx->scissor.scissor);
+ util_blitter_save_fragment_shader(rctx->blitter, rctx->ps_shader);
+ util_blitter_save_blend(rctx->blitter, rctx->blend_state.cso);
+ util_blitter_save_depth_stencil_alpha(rctx->blitter, rctx->dsa_state.cso);
+ util_blitter_save_stencil_ref(rctx->blitter, &rctx->stencil_ref.pipe_state);
+ util_blitter_save_sample_mask(rctx->blitter, rctx->sample_mask.sample_mask);
}
- util_blitter_save_vertex_buffers(rctx->blitter,
- rctx->vbuf_mgr->nr_vertex_buffers,
- rctx->vbuf_mgr->vertex_buffer);
- util_blitter_save_so_targets(rctx->blitter, rctx->num_so_targets,
- (struct pipe_stream_output_target**)rctx->so_targets);
if (op & R600_SAVE_FRAMEBUFFER)
- util_blitter_save_framebuffer(rctx->blitter, &rctx->framebuffer);
+ util_blitter_save_framebuffer(rctx->blitter, &rctx->framebuffer.state);
if (op & R600_SAVE_TEXTURES) {
util_blitter_save_fragment_sampler_states(
- rctx->blitter, rctx->ps_samplers.n_samplers,
- (void**)rctx->ps_samplers.samplers);
+ rctx->blitter, util_last_bit(rctx->samplers[PIPE_SHADER_FRAGMENT].states.enabled_mask),
+ (void**)rctx->samplers[PIPE_SHADER_FRAGMENT].states.states);
util_blitter_save_fragment_sampler_views(
- rctx->blitter, rctx->ps_samplers.n_views,
- (struct pipe_sampler_view**)rctx->ps_samplers.views);
+ rctx->blitter, util_last_bit(rctx->samplers[PIPE_SHADER_FRAGMENT].views.enabled_mask),
+ (struct pipe_sampler_view**)rctx->samplers[PIPE_SHADER_FRAGMENT].views.views);
}
if ((op & R600_DISABLE_RENDER_COND) && rctx->current_render_cond) {
- rctx->saved_render_cond = rctx->current_render_cond;
- rctx->saved_render_cond_mode = rctx->current_render_cond_mode;
- rctx->context.render_condition(&rctx->context, NULL, 0);
- }
-
+ util_blitter_save_render_condition(rctx->blitter,
+ rctx->current_render_cond,
+ rctx->current_render_cond_cond,
+ rctx->current_render_cond_mode);
+ }
}
static void r600_blitter_end(struct pipe_context *ctx)
{
struct r600_context *rctx = (struct r600_context *)ctx;
- if (rctx->saved_render_cond) {
- rctx->context.render_condition(&rctx->context,
- rctx->saved_render_cond,
- rctx->saved_render_cond_mode);
- rctx->saved_render_cond = NULL;
- }
- r600_context_queries_resume(rctx);
+ r600_resume_nontimer_queries(rctx);
}
-static unsigned u_num_layers(struct pipe_resource *r, unsigned level)
+static unsigned u_max_sample(struct pipe_resource *r)
{
- switch (r->target) {
- case PIPE_TEXTURE_CUBE:
- return 6;
- case PIPE_TEXTURE_3D:
- return u_minify(r->depth0, level);
- case PIPE_TEXTURE_1D_ARRAY:
- return r->array_size;
- case PIPE_TEXTURE_2D_ARRAY:
- return r->array_size;
- default:
- return 1;
- }
+ return r->nr_samples ? r->nr_samples - 1 : 0;
}
-void r600_blit_uncompress_depth(struct pipe_context *ctx, struct r600_resource_texture *texture)
+void r600_blit_decompress_depth(struct pipe_context *ctx,
+ struct r600_texture *texture,
+ struct r600_texture *staging,
+ unsigned first_level, unsigned last_level,
+ unsigned first_layer, unsigned last_layer,
+ unsigned first_sample, unsigned last_sample)
{
struct r600_context *rctx = (struct r600_context *)ctx;
- unsigned layer, level;
- float depth = 1.0f;
+ unsigned layer, level, sample, checked_last_layer, max_layer, max_sample;
+ struct r600_texture *flushed_depth_texture = staging ?
+ staging : texture->flushed_depth_texture;
+ const struct util_format_description *desc =
+ util_format_description(texture->resource.b.b.format);
+ float depth;
+
+ if (!staging && !texture->dirty_level_mask)
+ return;
+
+ max_sample = u_max_sample(&texture->resource.b.b);
- if (!texture->dirty_db)
+ /* XXX Decompressing MSAA depth textures is broken on R6xx.
+ * There is also a hardlock if CMASK and FMASK are not present.
+ * Just skip this until we find out how to fix it. */
+ if (rctx->b.chip_class == R600 && max_sample > 0) {
+ texture->dirty_level_mask = 0;
return;
+ }
- if (rctx->family == CHIP_RV610 || rctx->family == CHIP_RV630 ||
- rctx->family == CHIP_RV620 || rctx->family == CHIP_RV635)
+ if (rctx->b.family == CHIP_RV610 || rctx->b.family == CHIP_RV630 ||
+ rctx->b.family == CHIP_RV620 || rctx->b.family == CHIP_RV635)
depth = 0.0f;
+ else
+ depth = 1.0f;
- for (level = 0; level <= texture->resource.b.b.b.last_level; level++) {
- unsigned num_layers = u_num_layers(&texture->resource.b.b.b, level);
+ /* Enable decompression in DB_RENDER_CONTROL */
+ rctx->db_misc_state.flush_depthstencil_through_cb = true;
+ rctx->db_misc_state.copy_depth = util_format_has_depth(desc);
+ rctx->db_misc_state.copy_stencil = util_format_has_stencil(desc);
+ rctx->db_misc_state.copy_sample = first_sample;
+ rctx->db_misc_state.atom.dirty = true;
- for (layer = 0; layer < num_layers; layer++) {
- struct pipe_surface *zsurf, *cbsurf, surf_tmpl;
+ for (level = first_level; level <= last_level; level++) {
+ if (!staging && !(texture->dirty_level_mask & (1 << level)))
+ continue;
- surf_tmpl.format = texture->real_format;
- surf_tmpl.u.tex.level = level;
+ /* The smaller the mipmap level, the less layers there are
+ * as far as 3D textures are concerned. */
+ max_layer = util_max_layer(&texture->resource.b.b, level);
+ checked_last_layer = last_layer < max_layer ? last_layer : max_layer;
+
+ for (layer = first_layer; layer <= checked_last_layer; layer++) {
+ for (sample = first_sample; sample <= last_sample; sample++) {
+ struct pipe_surface *zsurf, *cbsurf, surf_tmpl;
+
+ if (sample != rctx->db_misc_state.copy_sample) {
+ rctx->db_misc_state.copy_sample = sample;
+ rctx->db_misc_state.atom.dirty = true;
+ }
+
+ surf_tmpl.format = texture->resource.b.b.format;
+ surf_tmpl.u.tex.level = level;
+ surf_tmpl.u.tex.first_layer = layer;
+ surf_tmpl.u.tex.last_layer = layer;
+
+ zsurf = ctx->create_surface(ctx, &texture->resource.b.b, &surf_tmpl);
+
+ surf_tmpl.format = flushed_depth_texture->resource.b.b.format;
+ cbsurf = ctx->create_surface(ctx,
+ &flushed_depth_texture->resource.b.b, &surf_tmpl);
+
+ r600_blitter_begin(ctx, R600_DECOMPRESS);
+ util_blitter_custom_depth_stencil(rctx->blitter, zsurf, cbsurf, 1 << sample,
+ rctx->custom_dsa_flush, depth);
+ r600_blitter_end(ctx);
+
+ pipe_surface_reference(&zsurf, NULL);
+ pipe_surface_reference(&cbsurf, NULL);
+ }
+ }
+
+ /* The texture will always be dirty if some layers or samples aren't flushed.
+ * I don't think this case occurs often though. */
+ if (!staging &&
+ first_layer == 0 && last_layer == max_layer &&
+ first_sample == 0 && last_sample == max_sample) {
+ texture->dirty_level_mask &= ~(1 << level);
+ }
+ }
+
+ /* reenable compression in DB_RENDER_CONTROL */
+ rctx->db_misc_state.flush_depthstencil_through_cb = false;
+ rctx->db_misc_state.atom.dirty = true;
+}
+
+static void r600_blit_decompress_depth_in_place(struct r600_context *rctx,
+ struct r600_texture *texture,
+ unsigned first_level, unsigned last_level,
+ unsigned first_layer, unsigned last_layer)
+{
+ struct pipe_surface *zsurf, surf_tmpl = {{0}};
+ unsigned layer, max_layer, checked_last_layer, level;
+
+ /* Enable decompression in DB_RENDER_CONTROL */
+ rctx->db_misc_state.flush_depthstencil_in_place = true;
+ rctx->db_misc_state.atom.dirty = true;
+
+ surf_tmpl.format = texture->resource.b.b.format;
+
+ for (level = first_level; level <= last_level; level++) {
+ if (!(texture->dirty_level_mask & (1 << level)))
+ continue;
+
+ surf_tmpl.u.tex.level = level;
+
+ /* The smaller the mipmap level, the less layers there are
+ * as far as 3D textures are concerned. */
+ max_layer = util_max_layer(&texture->resource.b.b, level);
+ checked_last_layer = last_layer < max_layer ? last_layer : max_layer;
+
+ for (layer = first_layer; layer <= checked_last_layer; layer++) {
surf_tmpl.u.tex.first_layer = layer;
surf_tmpl.u.tex.last_layer = layer;
- surf_tmpl.usage = PIPE_BIND_DEPTH_STENCIL;
- zsurf = ctx->create_surface(ctx, &texture->resource.b.b.b, &surf_tmpl);
+ zsurf = rctx->b.b.create_surface(&rctx->b.b, &texture->resource.b.b, &surf_tmpl);
+
+ r600_blitter_begin(&rctx->b.b, R600_DECOMPRESS);
+ util_blitter_custom_depth_stencil(rctx->blitter, zsurf, NULL, ~0,
+ rctx->custom_dsa_flush, 1.0f);
+ r600_blitter_end(&rctx->b.b);
- surf_tmpl.format = texture->flushed_depth_texture->real_format;
- surf_tmpl.usage = PIPE_BIND_RENDER_TARGET;
- cbsurf = ctx->create_surface(ctx,
- (struct pipe_resource*)texture->flushed_depth_texture, &surf_tmpl);
+ pipe_surface_reference(&zsurf, NULL);
+ }
+
+ /* The texture will always be dirty if some layers or samples aren't flushed.
+ * I don't think this case occurs often though. */
+ if (first_layer == 0 && last_layer == max_layer) {
+ texture->dirty_level_mask &= ~(1 << level);
+ }
+ }
+
+ /* Disable decompression in DB_RENDER_CONTROL */
+ rctx->db_misc_state.flush_depthstencil_in_place = false;
+ rctx->db_misc_state.atom.dirty = true;
+}
+
+void r600_decompress_depth_textures(struct r600_context *rctx,
+ struct r600_samplerview_state *textures)
+{
+ unsigned i;
+ unsigned depth_texture_mask = textures->compressed_depthtex_mask;
+
+ while (depth_texture_mask) {
+ struct pipe_sampler_view *view;
+ struct r600_texture *tex;
+
+ i = u_bit_scan(&depth_texture_mask);
+
+ view = &textures->views[i]->base;
+ assert(view);
+
+ tex = (struct r600_texture *)view->texture;
+ assert(tex->is_depth && !tex->is_flushing_texture);
+
+ if (rctx->b.chip_class >= EVERGREEN ||
+ r600_can_read_depth(tex)) {
+ r600_blit_decompress_depth_in_place(rctx, tex,
+ view->u.tex.first_level, view->u.tex.last_level,
+ 0, util_max_layer(&tex->resource.b.b, view->u.tex.first_level));
+ } else {
+ r600_blit_decompress_depth(&rctx->b.b, tex, NULL,
+ view->u.tex.first_level, view->u.tex.last_level,
+ 0, util_max_layer(&tex->resource.b.b, view->u.tex.first_level),
+ 0, u_max_sample(&tex->resource.b.b));
+ }
+ }
+}
+
+static void r600_blit_decompress_color(struct pipe_context *ctx,
+ struct r600_texture *rtex,
+ unsigned first_level, unsigned last_level,
+ unsigned first_layer, unsigned last_layer)
+{
+ struct r600_context *rctx = (struct r600_context *)ctx;
+ unsigned layer, level, checked_last_layer, max_layer;
+
+ if (!rtex->dirty_level_mask)
+ return;
+
+ for (level = first_level; level <= last_level; level++) {
+ if (!(rtex->dirty_level_mask & (1 << level)))
+ continue;
+
+ /* 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);
+ checked_last_layer = last_layer < max_layer ? last_layer : max_layer;
+
+ for (layer = first_layer; layer <= checked_last_layer; layer++) {
+ struct pipe_surface *cbsurf, surf_tmpl;
+
+ surf_tmpl.format = rtex->resource.b.b.format;
+ surf_tmpl.u.tex.level = level;
+ surf_tmpl.u.tex.first_layer = layer;
+ surf_tmpl.u.tex.last_layer = layer;
+ cbsurf = ctx->create_surface(ctx, &rtex->resource.b.b, &surf_tmpl);
r600_blitter_begin(ctx, R600_DECOMPRESS);
- util_blitter_custom_depth_stencil(rctx->blitter, zsurf, cbsurf, rctx->custom_dsa_flush, depth);
+ util_blitter_custom_color(rctx->blitter, cbsurf, rctx->custom_blend_decompress);
r600_blitter_end(ctx);
- pipe_surface_reference(&zsurf, NULL);
pipe_surface_reference(&cbsurf, NULL);
}
+
+ /* The texture will always be dirty if some layers aren't flushed.
+ * I don't think this case occurs often though. */
+ if (first_layer == 0 && last_layer == max_layer) {
+ rtex->dirty_level_mask &= ~(1 << level);
+ }
}
+}
+
+void r600_decompress_color_textures(struct r600_context *rctx,
+ struct r600_samplerview_state *textures)
+{
+ unsigned i;
+ unsigned mask = textures->compressed_colortex_mask;
- texture->dirty_db = FALSE;
+ while (mask) {
+ struct pipe_sampler_view *view;
+ struct r600_texture *tex;
+
+ i = u_bit_scan(&mask);
+
+ view = &textures->views[i]->base;
+ assert(view);
+
+ tex = (struct r600_texture *)view->texture;
+ assert(tex->cmask_size && tex->fmask_size);
+
+ r600_blit_decompress_color(&rctx->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));
+ }
}
-void r600_flush_depth_textures(struct r600_context *rctx)
+/* Helper for decompressing a portion of a color or depth resource before
+ * blitting if any decompression is needed.
+ * The driver doesn't decompress resources automatically while u_blitter is
+ * rendering. */
+static bool r600_decompress_subresource(struct pipe_context *ctx,
+ struct pipe_resource *tex,
+ unsigned level,
+ unsigned first_layer, unsigned last_layer)
{
- unsigned int i;
+ struct r600_context *rctx = (struct r600_context *)ctx;
+ struct r600_texture *rtex = (struct r600_texture*)tex;
+
+ if (rtex->is_depth && !rtex->is_flushing_texture) {
+ if (rctx->b.chip_class >= EVERGREEN ||
+ r600_can_read_depth(rtex)) {
+ r600_blit_decompress_depth_in_place(rctx, rtex,
+ level, level,
+ first_layer, last_layer);
+ } else {
+ if (!r600_init_flushed_depth_texture(ctx, tex, NULL))
+ return false; /* error */
+
+ r600_blit_decompress_depth(ctx, rtex, NULL,
+ level, level,
+ first_layer, last_layer,
+ 0, u_max_sample(tex));
+ }
+ } else if (rtex->fmask_size && rtex->cmask_size) {
+ r600_blit_decompress_color(ctx, rtex, level, level,
+ first_layer, last_layer);
+ }
+ return true;
+}
- /* FIXME: This handles fragment shader textures only. */
+static boolean is_simple_msaa_resolve(const struct pipe_blit_info *info)
+{
+ unsigned dst_width = u_minify(info->dst.resource->width0, info->dst.level);
+ unsigned dst_height = u_minify(info->dst.resource->height0, info->dst.level);
+ struct r600_texture *dst = (struct r600_texture*)info->dst.resource;
+ unsigned dst_tile_mode = dst->surface.level[info->dst.level].mode;
+
+ return info->dst.resource->format == info->src.resource->format &&
+ info->dst.resource->format == info->dst.format &&
+ info->src.resource->format == info->src.format &&
+ !info->scissor_enable &&
+ info->mask == PIPE_MASK_RGBA &&
+ dst_width == info->src.resource->width0 &&
+ dst_height == info->src.resource->height0 &&
+ info->dst.box.x == 0 &&
+ info->dst.box.y == 0 &&
+ info->dst.box.width == dst_width &&
+ info->dst.box.height == dst_height &&
+ info->src.box.x == 0 &&
+ info->src.box.y == 0 &&
+ info->src.box.width == dst_width &&
+ info->src.box.height == dst_height &&
+ /* Dst must be tiled. If it's not, we have to use a temporary
+ * resource which is tiled. */
+ dst_tile_mode >= RADEON_SURF_MODE_1D;
+}
- for (i = 0; i < rctx->ps_samplers.n_views; ++i) {
- struct r600_pipe_sampler_view *view;
- struct r600_resource_texture *tex;
+static void r600_clear_buffer(struct pipe_context *ctx, struct pipe_resource *dst,
+ unsigned offset, unsigned size, unsigned char value);
- view = rctx->ps_samplers.views[i];
- if (!view) continue;
+static void evergreen_set_clear_color(struct pipe_surface *cbuf,
+ const union pipe_color_union *color)
+{
+ unsigned *clear_value = ((struct r600_texture *)cbuf->texture)->color_clear_value;
+ union util_color uc;
- tex = (struct r600_resource_texture *)view->base.texture;
- if (!tex->depth)
- continue;
+ memset(&uc, 0, sizeof(uc));
- if (tex->is_flushing_texture)
- continue;
+ if (util_format_is_pure_uint(cbuf->format)) {
+ util_format_write_4ui(cbuf->format, color->ui, 0, &uc, 0, 0, 0, 1, 1);
+ } else if (util_format_is_pure_sint(cbuf->format)) {
+ util_format_write_4i(cbuf->format, color->i, 0, &uc, 0, 0, 0, 1, 1);
+ } else {
+ util_pack_color(color->f, cbuf->format, &uc);
+ }
- r600_blit_uncompress_depth(&rctx->context, tex);
+ memcpy(clear_value, &uc, 2 * sizeof(uint32_t));
+}
+
+static bool can_fast_clear_color(struct pipe_context *ctx)
+{
+ struct r600_context *rctx = (struct r600_context *)ctx;
+ struct pipe_framebuffer_state *fb = &rctx->framebuffer.state;
+ int i;
+
+ if (rctx->b.chip_class < EVERGREEN) {
+ return false;
}
- /* also check CB here */
- for (i = 0; i < rctx->framebuffer.nr_cbufs; i++) {
- struct r600_resource_texture *tex;
- tex = (struct r600_resource_texture *)rctx->framebuffer.cbufs[i]->texture;
+ for (i = 0; i < fb->nr_cbufs; i++) {
+ struct r600_texture *tex = (struct r600_texture *)fb->cbufs[i]->texture;
- if (!tex->depth)
- continue;
+ if (tex->cmask_size == 0) {
+ return false;
+ }
- if (tex->is_flushing_texture)
- continue;
+ /* 128-bit formats are unuspported */
+ if (util_format_get_blocksizebits(fb->cbufs[i]->format) > 64) {
+ return false;
+ }
- r600_blit_uncompress_depth(&rctx->context, tex);
+ /* the clear is allowed if all layers are bound */
+ if (fb->cbufs[i]->u.tex.first_layer != 0 ||
+ fb->cbufs[i]->u.tex.last_layer != util_max_layer(&tex->resource.b.b, 0)) {
+ return false;
+ }
}
+
+ return true;
}
static void r600_clear(struct pipe_context *ctx, unsigned buffers,
double depth, unsigned stencil)
{
struct r600_context *rctx = (struct r600_context *)ctx;
- struct pipe_framebuffer_state *fb = &rctx->framebuffer;
+ struct pipe_framebuffer_state *fb = &rctx->framebuffer.state;
+
+ /* fast color clear on AA framebuffers (EG+) */
+ if ((buffers & PIPE_CLEAR_COLOR) && can_fast_clear_color(ctx)) {
+ int i;
+
+ for (i = 0; i < fb->nr_cbufs; i++) {
+ struct r600_texture *tex = (struct r600_texture *)fb->cbufs[i]->texture;
+
+ evergreen_set_clear_color(fb->cbufs[i], color);
+ r600_clear_buffer(ctx, fb->cbufs[i]->texture,
+ tex->cmask_offset, tex->cmask_size, 0);
+ tex->dirty_level_mask |= 1 << fb->cbufs[i]->u.tex.level;
+ }
+
+ rctx->framebuffer.atom.dirty = true;
+
+ buffers &= ~PIPE_CLEAR_COLOR;
+ if (!buffers)
+ return;
+ }
+
+ /* if hyperz enabled just clear hyperz */
+ if (fb->zsbuf && (buffers & PIPE_CLEAR_DEPTH)) {
+ struct r600_texture *rtex;
+ unsigned level = fb->zsbuf->u.tex.level;
+
+ rtex = (struct r600_texture*)fb->zsbuf->texture;
+
+ /* We can't use hyperz fast clear if each slice of a texture
+ * array are clear to different value. To simplify code just
+ * disable fast clear for texture array.
+ */
+ /* Only use htile for first level */
+ if (rtex->htile && !level && rtex->surface.array_size == 1) {
+ if (rtex->depth_clear != depth) {
+ rtex->depth_clear = depth;
+ rctx->db_state.atom.dirty = true;
+ }
+ rctx->db_misc_state.htile_clear = true;
+ rctx->db_misc_state.atom.dirty = true;
+ }
+ }
r600_blitter_begin(ctx, R600_CLEAR);
util_blitter_clear(rctx->blitter, fb->width, fb->height,
- fb->nr_cbufs, buffers, fb->nr_cbufs ? fb->cbufs[0]->format : PIPE_FORMAT_NONE,
- color, depth, stencil);
+ buffers, color, depth, stencil);
r600_blitter_end(ctx);
+
+ /* disable fast clear */
+ if (rctx->db_misc_state.htile_clear) {
+ rctx->db_misc_state.htile_clear = false;
+ rctx->db_misc_state.atom.dirty = true;
+ }
}
static void r600_clear_render_target(struct pipe_context *ctx,
r600_blitter_end(ctx);
}
+void r600_copy_buffer(struct pipe_context *ctx, struct pipe_resource *dst, unsigned dstx,
+ struct pipe_resource *src, const struct pipe_box *src_box)
+{
+ struct r600_context *rctx = (struct r600_context*)ctx;
+ if (rctx->screen->has_cp_dma) {
+ r600_cp_dma_copy_buffer(rctx, dst, dstx, src, src_box->x, src_box->width);
+ }
+ else if (rctx->screen->has_streamout &&
+ /* Require 4-byte alignment. */
+ dstx % 4 == 0 && src_box->x % 4 == 0 && src_box->width % 4 == 0) {
+
+ /* Flush both resources. */
+ r600_flag_resource_cache_flush(rctx, src);
+ r600_flag_resource_cache_flush(rctx, dst);
+
+ r600_blitter_begin(ctx, R600_COPY_BUFFER);
+ util_blitter_copy_buffer(rctx->blitter, dst, dstx, src, src_box->x, src_box->width);
+ r600_blitter_end(ctx);
+
+ /* Flush the dst in case the 3D engine has been prefetching the resource. */
+ r600_flag_resource_cache_flush(rctx, dst);
+ } else {
+ util_resource_copy_region(ctx, dst, 0, dstx, 0, 0, src, 0, src_box);
+ }
+}
-/* Copy a block of pixels from one surface to another using HW. */
-static void r600_hw_copy_region(struct pipe_context *ctx,
- struct pipe_resource *dst,
- unsigned dst_level,
- unsigned dstx, unsigned dsty, unsigned dstz,
- struct pipe_resource *src,
- unsigned src_level,
- const struct pipe_box *src_box)
+static void r600_clear_buffer(struct pipe_context *ctx, struct pipe_resource *dst,
+ unsigned offset, unsigned size, unsigned char value)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct r600_context *rctx = (struct r600_context*)ctx;
+ uint32_t v = value;
- r600_blitter_begin(ctx, R600_COPY);
- util_blitter_copy_texture(rctx->blitter, dst, dst_level, dstx, dsty, dstz,
- src, src_level, src_box, TRUE);
- r600_blitter_end(ctx);
-}
+ if (rctx->screen->has_cp_dma &&
+ rctx->b.chip_class >= EVERGREEN &&
+ offset % 4 == 0 && size % 4 == 0) {
+ uint32_t clear_value = v | (v << 8) | (v << 16) | (v << 24);
-struct texture_orig_info {
- unsigned format;
- unsigned width0;
- unsigned height0;
- unsigned npix_x;
- unsigned npix_y;
- unsigned npix0_x;
- unsigned npix0_y;
-};
+ evergreen_cp_dma_clear_buffer(rctx, dst, offset, size, clear_value);
+ } else if (rctx->screen->has_streamout && offset % 4 == 0 && size % 4 == 0) {
+ union pipe_color_union clear_value;
-static void r600_compressed_to_blittable(struct pipe_resource *tex,
- unsigned level,
- struct texture_orig_info *orig)
+ clear_value.ui[0] = v | (v << 8) | (v << 16) | (v << 24);
+
+ r600_flag_resource_cache_flush(rctx, dst);
+
+ r600_blitter_begin(ctx, R600_DISABLE_RENDER_COND);
+ util_blitter_clear_buffer(rctx->blitter, dst, offset, size,
+ 1, &clear_value);
+ r600_blitter_end(ctx);
+
+ /* Flush again in case the 3D engine has been prefetching the resource. */
+ r600_flag_resource_cache_flush(rctx, dst);
+ } else {
+ char *map = r600_buffer_mmap_sync_with_rings(rctx, r600_resource(dst),
+ PIPE_TRANSFER_WRITE);
+ memset(map + offset, value, size);
+ }
+}
+
+void r600_screen_clear_buffer(struct r600_screen *rscreen, struct pipe_resource *dst,
+ unsigned offset, unsigned size, unsigned char value)
{
- struct r600_resource_texture *rtex = (struct r600_resource_texture*)tex;
- unsigned pixsize = util_format_get_blocksize(rtex->real_format);
- int new_format;
- int new_height, new_width;
-
- orig->format = tex->format;
- orig->width0 = tex->width0;
- orig->height0 = tex->height0;
- orig->npix0_x = rtex->surface.level[0].npix_x;
- orig->npix0_y = rtex->surface.level[0].npix_y;
- orig->npix_x = rtex->surface.level[level].npix_x;
- orig->npix_y = rtex->surface.level[level].npix_y;
-
- if (pixsize == 8)
- new_format = PIPE_FORMAT_R16G16B16A16_UINT; /* 64-bit block */
- else
- new_format = PIPE_FORMAT_R32G32B32A32_UINT; /* 128-bit block */
-
- new_width = util_format_get_nblocksx(tex->format, orig->width0);
- new_height = util_format_get_nblocksy(tex->format, orig->height0);
-
- tex->width0 = new_width;
- tex->height0 = new_height;
- tex->format = new_format;
- rtex->surface.level[0].npix_x = util_format_get_nblocksx(orig->format, orig->npix0_x);
- rtex->surface.level[0].npix_y = util_format_get_nblocksy(orig->format, orig->npix0_y);
- rtex->surface.level[level].npix_x = util_format_get_nblocksx(orig->format, orig->npix_x);
- rtex->surface.level[level].npix_y = util_format_get_nblocksy(orig->format, orig->npix_y);
+ pipe_mutex_lock(rscreen->aux_context_lock);
+ r600_clear_buffer(rscreen->aux_context, dst, offset, size, value);
+ rscreen->aux_context->flush(rscreen->aux_context, NULL, 0);
+ pipe_mutex_unlock(rscreen->aux_context_lock);
}
-static void r600_reset_blittable_to_compressed(struct pipe_resource *tex,
- unsigned level,
- struct texture_orig_info *orig)
+static bool util_format_is_subsampled_2x1_32bpp(enum pipe_format format)
{
- struct r600_resource_texture *rtex = (struct r600_resource_texture*)tex;
-
- tex->format = orig->format;
- tex->width0 = orig->width0;
- tex->height0 = orig->height0;
- rtex->surface.level[0].npix_x = orig->npix0_x;
- rtex->surface.level[0].npix_y = orig->npix0_y;
- rtex->surface.level[level].npix_x = orig->npix_x;
- rtex->surface.level[level].npix_y = orig->npix_y;
+ const struct util_format_description *desc = util_format_description(format);
+
+ return desc->layout == UTIL_FORMAT_LAYOUT_SUBSAMPLED &&
+ desc->block.width == 2 &&
+ desc->block.height == 1 &&
+ desc->block.bits == 32;
}
static void r600_resource_copy_region(struct pipe_context *ctx,
unsigned src_level,
const struct pipe_box *src_box)
{
- struct r600_resource_texture *rsrc = (struct r600_resource_texture*)src;
- struct texture_orig_info orig_info[2];
- struct pipe_box sbox;
- const struct pipe_box *psbox;
- boolean restore_orig[2];
-
- memset(orig_info, 0, sizeof(orig_info));
+ struct r600_context *rctx = (struct r600_context *)ctx;
+ struct pipe_surface *dst_view, dst_templ;
+ struct pipe_sampler_view src_templ, *src_view;
+ unsigned dst_width, dst_height, src_width0, src_height0, src_widthFL, src_heightFL;
+ struct pipe_box sbox, dstbox;
- /* Fallback for buffers. */
+ /* Handle buffers first. */
if (dst->target == PIPE_BUFFER && src->target == PIPE_BUFFER) {
- util_resource_copy_region(ctx, dst, dst_level, dstx, dsty, dstz,
- src, src_level, src_box);
+ r600_copy_buffer(ctx, dst, dstx, src, src_box);
return;
}
- if (rsrc->depth && !rsrc->is_flushing_texture)
- r600_texture_depth_flush(ctx, src, FALSE);
+ assert(u_max_sample(dst) == u_max_sample(src));
- restore_orig[0] = restore_orig[1] = FALSE;
+ /* The driver doesn't decompress resources automatically while
+ * u_blitter is rendering. */
+ if (!r600_decompress_subresource(ctx, src, src_level,
+ src_box->z, src_box->z + src_box->depth - 1)) {
+ return; /* error */
+ }
+
+ dst_width = u_minify(dst->width0, dst_level);
+ dst_height = u_minify(dst->height0, dst_level);
+ src_width0 = src->width0;
+ src_height0 = src->height0;
+ src_widthFL = u_minify(src->width0, src_level);
+ src_heightFL = u_minify(src->height0, src_level);
+
+ util_blitter_default_dst_texture(&dst_templ, dst, dst_level, dstz);
+ util_blitter_default_src_texture(&src_templ, src, src_level);
if (util_format_is_compressed(src->format)) {
- r600_compressed_to_blittable(src, src_level, &orig_info[0]);
- restore_orig[0] = TRUE;
- sbox.x = util_format_get_nblocksx(orig_info[0].format, src_box->x);
- sbox.y = util_format_get_nblocksy(orig_info[0].format, src_box->y);
+ unsigned blocksize = util_format_get_blocksize(src->format);
+
+ if (blocksize == 8)
+ src_templ.format = PIPE_FORMAT_R16G16B16A16_UINT; /* 64-bit block */
+ else
+ src_templ.format = PIPE_FORMAT_R32G32B32A32_UINT; /* 128-bit block */
+ dst_templ.format = src_templ.format;
+
+ dst_width = util_format_get_nblocksx(dst->format, dst_width);
+ dst_height = util_format_get_nblocksy(dst->format, dst_height);
+ src_width0 = util_format_get_nblocksx(src->format, src_width0);
+ src_height0 = util_format_get_nblocksy(src->format, src_height0);
+ src_widthFL = util_format_get_nblocksx(src->format, src_widthFL);
+ src_heightFL = util_format_get_nblocksy(src->format, src_heightFL);
+
+ dstx = util_format_get_nblocksx(dst->format, dstx);
+ dsty = util_format_get_nblocksy(dst->format, dsty);
+
+ sbox.x = util_format_get_nblocksx(src->format, src_box->x);
+ sbox.y = util_format_get_nblocksy(src->format, src_box->y);
sbox.z = src_box->z;
- sbox.width = util_format_get_nblocksx(orig_info[0].format, src_box->width);
- sbox.height = util_format_get_nblocksy(orig_info[0].format, src_box->height);
+ sbox.width = util_format_get_nblocksx(src->format, src_box->width);
+ sbox.height = util_format_get_nblocksy(src->format, src_box->height);
sbox.depth = src_box->depth;
- psbox=&sbox;
- } else
- psbox=src_box;
-
- if (util_format_is_compressed(dst->format)) {
- r600_compressed_to_blittable(dst, dst_level, &orig_info[1]);
- restore_orig[1] = TRUE;
- /* translate the dst box as well */
- dstx = util_format_get_nblocksx(orig_info[1].format, dstx);
- dsty = util_format_get_nblocksy(orig_info[1].format, dsty);
+ src_box = &sbox;
+ } else if (!util_blitter_is_copy_supported(rctx->blitter, dst, src,
+ PIPE_MASK_RGBAZS)) {
+ if (util_format_is_subsampled_2x1_32bpp(src->format)) {
+
+ src_templ.format = PIPE_FORMAT_R8G8B8A8_UINT;
+ dst_templ.format = PIPE_FORMAT_R8G8B8A8_UINT;
+
+ dst_width = util_format_get_nblocksx(dst->format, dst_width);
+ src_width0 = util_format_get_nblocksx(src->format, src_width0);
+ src_widthFL = util_format_get_nblocksx(src->format, src_widthFL);
+
+ dstx = util_format_get_nblocksx(dst->format, dstx);
+
+ sbox = *src_box;
+ sbox.x = util_format_get_nblocksx(src->format, src_box->x);
+ sbox.width = util_format_get_nblocksx(src->format, src_box->width);
+ src_box = &sbox;
+ } else {
+ unsigned blocksize = util_format_get_blocksize(src->format);
+
+ switch (blocksize) {
+ case 1:
+ dst_templ.format = PIPE_FORMAT_R8_UNORM;
+ src_templ.format = PIPE_FORMAT_R8_UNORM;
+ break;
+ case 2:
+ dst_templ.format = PIPE_FORMAT_R8G8_UNORM;
+ src_templ.format = PIPE_FORMAT_R8G8_UNORM;
+ break;
+ case 4:
+ dst_templ.format = PIPE_FORMAT_R8G8B8A8_UNORM;
+ src_templ.format = PIPE_FORMAT_R8G8B8A8_UNORM;
+ break;
+ case 8:
+ dst_templ.format = PIPE_FORMAT_R16G16B16A16_UINT;
+ src_templ.format = PIPE_FORMAT_R16G16B16A16_UINT;
+ break;
+ case 16:
+ dst_templ.format = PIPE_FORMAT_R32G32B32A32_UINT;
+ src_templ.format = PIPE_FORMAT_R32G32B32A32_UINT;
+ break;
+ default:
+ fprintf(stderr, "Unhandled format %s with blocksize %u\n",
+ util_format_short_name(src->format), blocksize);
+ assert(0);
+ }
+ }
+ }
+
+ dst_view = r600_create_surface_custom(ctx, dst, &dst_templ, dst_width, dst_height);
+
+ if (rctx->b.chip_class >= EVERGREEN) {
+ src_view = evergreen_create_sampler_view_custom(ctx, src, &src_templ,
+ src_width0, src_height0);
+ } else {
+ src_view = r600_create_sampler_view_custom(ctx, src, &src_templ,
+ src_widthFL, src_heightFL);
}
- r600_hw_copy_region(ctx, dst, dst_level, dstx, dsty, dstz,
- src, src_level, psbox);
+ u_box_3d(dstx, dsty, dstz, abs(src_box->width), abs(src_box->height),
+ abs(src_box->depth), &dstbox);
- if (restore_orig[0])
- r600_reset_blittable_to_compressed(src, src_level, &orig_info[0]);
+ /* Copy. */
+ r600_blitter_begin(ctx, R600_COPY_TEXTURE);
+ util_blitter_blit_generic(rctx->blitter, dst_view, &dstbox,
+ src_view, src_box, src_width0, src_height0,
+ PIPE_MASK_RGBAZS, PIPE_TEX_FILTER_NEAREST, NULL,
+ TRUE);
+ r600_blitter_end(ctx);
- if (restore_orig[1])
- r600_reset_blittable_to_compressed(dst, dst_level, &orig_info[1]);
+ pipe_surface_reference(&dst_view, NULL);
+ pipe_sampler_view_reference(&src_view, NULL);
}
-void r600_init_blit_functions(struct r600_context *rctx)
+/* For MSAA integer resolving to work, we change the format to NORM using this function. */
+static enum pipe_format int_to_norm_format(enum pipe_format format)
+{
+ switch (format) {
+#define REPLACE_FORMAT_SIGN(format,sign) \
+ case PIPE_FORMAT_##format##_##sign##INT: \
+ return PIPE_FORMAT_##format##_##sign##NORM
+#define REPLACE_FORMAT(format) \
+ REPLACE_FORMAT_SIGN(format, U); \
+ REPLACE_FORMAT_SIGN(format, S)
+
+ REPLACE_FORMAT_SIGN(B10G10R10A2, U);
+ REPLACE_FORMAT(R8);
+ REPLACE_FORMAT(R8G8);
+ REPLACE_FORMAT(R8G8B8X8);
+ REPLACE_FORMAT(R8G8B8A8);
+ REPLACE_FORMAT(A8);
+ REPLACE_FORMAT(I8);
+ REPLACE_FORMAT(L8);
+ REPLACE_FORMAT(L8A8);
+ REPLACE_FORMAT(R16);
+ REPLACE_FORMAT(R16G16);
+ REPLACE_FORMAT(R16G16B16X16);
+ REPLACE_FORMAT(R16G16B16A16);
+ REPLACE_FORMAT(A16);
+ REPLACE_FORMAT(I16);
+ REPLACE_FORMAT(L16);
+ REPLACE_FORMAT(L16A16);
+
+#undef REPLACE_FORMAT
+#undef REPLACE_FORMAT_SIGN
+ default:
+ return format;
+ }
+}
+
+static void r600_msaa_color_resolve(struct pipe_context *ctx,
+ const struct pipe_blit_info *info)
+{
+ struct r600_context *rctx = (struct r600_context *)ctx;
+ struct pipe_screen *screen = ctx->screen;
+ struct pipe_resource *tmp, templ;
+ struct pipe_blit_info blit;
+ unsigned sample_mask =
+ rctx->b.chip_class == CAYMAN ? ~0 :
+ ((1ull << MAX2(1, info->src.resource->nr_samples)) - 1);
+
+ assert(info->src.level == 0);
+ assert(info->src.box.depth == 1);
+ assert(info->dst.box.depth == 1);
+
+ if (is_simple_msaa_resolve(info)) {
+ r600_blitter_begin(ctx, R600_COLOR_RESOLVE);
+ util_blitter_custom_resolve_color(rctx->blitter,
+ info->dst.resource, info->dst.level,
+ info->dst.box.z,
+ info->src.resource, info->src.box.z,
+ sample_mask, rctx->custom_blend_resolve,
+ int_to_norm_format(info->dst.format));
+ r600_blitter_end(ctx);
+ return;
+ }
+
+ /* resolve into a temporary texture, then blit */
+ templ.target = PIPE_TEXTURE_2D;
+ templ.format = info->src.resource->format;
+ templ.width0 = info->src.resource->width0;
+ templ.height0 = info->src.resource->height0;
+ templ.depth0 = 1;
+ templ.array_size = 1;
+ templ.last_level = 0;
+ templ.nr_samples = 0;
+ templ.usage = PIPE_USAGE_STATIC;
+ templ.bind = PIPE_BIND_RENDER_TARGET | PIPE_BIND_SAMPLER_VIEW;
+ templ.flags = R600_RESOURCE_FLAG_FORCE_TILING; /* dst must not have a linear layout */
+
+ tmp = screen->resource_create(screen, &templ);
+
+ /* resolve */
+ r600_blitter_begin(ctx, R600_COLOR_RESOLVE);
+ util_blitter_custom_resolve_color(rctx->blitter,
+ tmp, 0, 0,
+ info->src.resource, info->src.box.z,
+ sample_mask, rctx->custom_blend_resolve,
+ int_to_norm_format(tmp->format));
+ r600_blitter_end(ctx);
+
+ /* blit */
+ blit = *info;
+ blit.src.resource = tmp;
+ blit.src.box.z = 0;
+
+ r600_blitter_begin(ctx, R600_BLIT);
+ util_blitter_blit(rctx->blitter, &blit);
+ r600_blitter_end(ctx);
+
+ pipe_resource_reference(&tmp, NULL);
+}
+
+static void r600_blit(struct pipe_context *ctx,
+ const struct pipe_blit_info *info)
{
- rctx->context.clear = r600_clear;
- rctx->context.clear_render_target = r600_clear_render_target;
- rctx->context.clear_depth_stencil = r600_clear_depth_stencil;
- rctx->context.resource_copy_region = r600_resource_copy_region;
+ struct r600_context *rctx = (struct r600_context*)ctx;
+
+ assert(util_blitter_is_blit_supported(rctx->blitter, info));
+
+ if (info->src.resource->nr_samples > 1 &&
+ info->dst.resource->nr_samples <= 1 &&
+ !util_format_is_depth_or_stencil(info->src.resource->format) &&
+ !util_format_is_pure_integer(int_to_norm_format(info->src.resource->format))) {
+ r600_msaa_color_resolve(ctx, info);
+ return;
+ }
+
+ /* The driver doesn't decompress resources automatically while
+ * u_blitter is rendering. */
+ if (!r600_decompress_subresource(ctx, info->src.resource, info->src.level,
+ info->src.box.z,
+ info->src.box.z + info->src.box.depth - 1)) {
+ return; /* error */
+ }
+
+ r600_blitter_begin(ctx, R600_BLIT);
+ util_blitter_blit(rctx->blitter, info);
+ r600_blitter_end(ctx);
}
-void r600_blit_push_depth(struct pipe_context *ctx, struct r600_resource_texture *texture)
+void r600_init_blit_functions(struct r600_context *rctx)
{
- struct pipe_box sbox;
-
- sbox.x = sbox.y = sbox.z = 0;
- sbox.width = texture->resource.b.b.b.width0;
- sbox.height = texture->resource.b.b.b.height0;
- /* XXX that might be wrong */
- sbox.depth = 1;
-
- r600_hw_copy_region(ctx, (struct pipe_resource *)texture, 0,
- 0, 0, 0,
- (struct pipe_resource *)texture->flushed_depth_texture, 0,
- &sbox);
+ rctx->b.b.clear = r600_clear;
+ rctx->b.b.clear_render_target = r600_clear_render_target;
+ rctx->b.b.clear_depth_stencil = r600_clear_depth_stencil;
+ rctx->b.b.resource_copy_region = r600_resource_copy_region;
+ rctx->b.b.blit = r600_blit;
}