* USE OR OTHER DEALINGS IN THE SOFTWARE.
*/
#include "r600_pipe.h"
+#include "compute_memory_pool.h"
+#include "evergreen_compute.h"
#include "util/u_surface.h"
-#include "util/u_blitter.h"
#include "util/u_format.h"
#include "evergreend.h"
R600_COPY_TEXTURE = R600_SAVE_FRAGMENT_STATE | R600_SAVE_FRAMEBUFFER | R600_SAVE_TEXTURES |
R600_DISABLE_RENDER_COND,
- R600_BLIT = R600_SAVE_FRAGMENT_STATE | R600_SAVE_FRAMEBUFFER | R600_SAVE_TEXTURES |
- R600_DISABLE_RENDER_COND,
+ R600_BLIT = R600_SAVE_FRAGMENT_STATE | R600_SAVE_FRAMEBUFFER | R600_SAVE_TEXTURES,
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
+ R600_COLOR_RESOLVE = R600_SAVE_FRAGMENT_STATE | R600_SAVE_FRAMEBUFFER
};
static void r600_blitter_begin(struct pipe_context *ctx, enum r600_blitter_op op)
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_geometry_shader(rctx->blitter, rctx->gs_shader);
+ util_blitter_save_tessctrl_shader(rctx->blitter, rctx->tcs_shader);
+ util_blitter_save_tesseval_shader(rctx->blitter, rctx->tes_shader);
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[0].state);
- util_blitter_save_scissor(rctx->blitter, &rctx->scissor[0].scissor);
+ util_blitter_save_viewport(rctx->blitter, &rctx->viewport.state[0]);
+ util_blitter_save_scissor(rctx->blitter, &rctx->scissor.scissor[0]);
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);
(struct pipe_sampler_view**)rctx->samplers[PIPE_SHADER_FRAGMENT].views.views);
}
- if ((op & R600_DISABLE_RENDER_COND) && rctx->b.current_render_cond) {
- util_blitter_save_render_condition(rctx->blitter,
- rctx->b.current_render_cond,
- rctx->b.current_render_cond_cond,
- rctx->b.current_render_cond_mode);
- }
+ if (op & R600_DISABLE_RENDER_COND)
+ rctx->b.render_cond_force_off = true;
}
static void r600_blitter_end(struct pipe_context *ctx)
{
struct r600_context *rctx = (struct r600_context *)ctx;
- r600_resume_nontimer_queries(&rctx->b);
+
+ rctx->b.render_cond_force_off = false;
+ r600_resume_nontimer_queries(&rctx->b);
}
static unsigned u_max_sample(struct pipe_resource *r)
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;
+ r600_mark_atom_dirty(rctx, &rctx->db_misc_state.atom);
for (level = first_level; level <= last_level; level++) {
if (!staging && !(texture->dirty_level_mask & (1 << level)))
if (sample != rctx->db_misc_state.copy_sample) {
rctx->db_misc_state.copy_sample = sample;
- rctx->db_misc_state.atom.dirty = true;
+ r600_mark_atom_dirty(rctx, &rctx->db_misc_state.atom);
}
surf_tmpl.format = texture->resource.b.b.format;
/* reenable compression in DB_RENDER_CONTROL */
rctx->db_misc_state.flush_depthstencil_through_cb = false;
- rctx->db_misc_state.atom.dirty = true;
+ r600_mark_atom_dirty(rctx, &rctx->db_misc_state.atom);
}
static void r600_blit_decompress_depth_in_place(struct r600_context *rctx,
struct r600_texture *texture,
+ bool is_stencil_sampler,
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;
+ unsigned *dirty_level_mask;
/* Enable decompression in DB_RENDER_CONTROL */
- rctx->db_misc_state.flush_depthstencil_in_place = true;
- rctx->db_misc_state.atom.dirty = true;
+ if (is_stencil_sampler) {
+ rctx->db_misc_state.flush_stencil_inplace = true;
+ dirty_level_mask = &texture->stencil_dirty_level_mask;
+ } else {
+ rctx->db_misc_state.flush_depth_inplace = true;
+ dirty_level_mask = &texture->dirty_level_mask;
+ }
+ r600_mark_atom_dirty(rctx, &rctx->db_misc_state.atom);
surf_tmpl.format = texture->resource.b.b.format;
for (level = first_level; level <= last_level; level++) {
- if (!(texture->dirty_level_mask & (1 << level)))
+ if (!(*dirty_level_mask & (1 << level)))
continue;
surf_tmpl.u.tex.level = level;
/* 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);
+ *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;
+ rctx->db_misc_state.flush_depth_inplace = false;
+ rctx->db_misc_state.flush_stencil_inplace = false;
+ r600_mark_atom_dirty(rctx, &rctx->db_misc_state.atom);
}
void r600_decompress_depth_textures(struct r600_context *rctx,
while (depth_texture_mask) {
struct pipe_sampler_view *view;
+ struct r600_pipe_sampler_view *rview;
struct r600_texture *tex;
i = u_bit_scan(&depth_texture_mask);
view = &textures->views[i]->base;
assert(view);
+ rview = (struct r600_pipe_sampler_view*)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,
+ rview->is_stencil_sampler,
view->u.tex.first_level, view->u.tex.last_level,
0, util_max_layer(&tex->resource.b.b, view->u.tex.first_level));
} else {
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,
+ r600_blit_decompress_depth_in_place(rctx, rtex, false,
level, level,
first_layer, last_layer);
+ if (rtex->surface.flags & RADEON_SURF_SBUFFER) {
+ r600_blit_decompress_depth_in_place(rctx, rtex, true,
+ level, level,
+ first_layer, last_layer);
+ }
} else {
if (!r600_init_flushed_depth_texture(ctx, tex, NULL))
return false; /* error */
if (buffers & PIPE_CLEAR_COLOR && rctx->b.chip_class >= EVERGREEN) {
evergreen_do_fast_color_clear(&rctx->b, fb, &rctx->framebuffer.atom,
- &buffers, color);
+ &buffers, NULL, color);
+ if (!buffers)
+ return; /* all buffers have been fast cleared */
}
if (buffers & PIPE_CLEAR_COLOR) {
* disable fast clear for texture array.
*/
/* Only use htile for first level */
- if (rtex->htile_buffer && !level && rtex->surface.array_size == 1) {
+ if (rtex->htile_buffer && !level &&
+ fb->zsbuf->u.tex.first_layer == 0 &&
+ fb->zsbuf->u.tex.last_layer == util_max_layer(&rtex->resource.b.b, level)) {
if (rtex->depth_clear_value != depth) {
rtex->depth_clear_value = depth;
- rctx->db_state.atom.dirty = true;
+ r600_mark_atom_dirty(rctx, &rctx->db_state.atom);
}
rctx->db_misc_state.htile_clear = true;
- rctx->db_misc_state.atom.dirty = true;
+ r600_mark_atom_dirty(rctx, &rctx->db_misc_state.atom);
}
}
r600_blitter_begin(ctx, R600_CLEAR);
- util_blitter_clear(rctx->blitter, fb->width, fb->height, 1,
+ util_blitter_clear(rctx->blitter, fb->width, fb->height,
+ util_framebuffer_get_num_layers(fb),
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;
+ r600_mark_atom_dirty(rctx, &rctx->db_misc_state.atom);
}
}
} else {
util_resource_copy_region(ctx, dst, 0, dstx, 0, 0, src, 0, src_box);
}
+
+ /* The index buffer (VGT) doesn't seem to see the result of the copying.
+ * Can we somehow flush the index buffer cache? Starting a new IB seems
+ * to do the trick. */
+ if (rctx->b.chip_class <= R700)
+ rctx->b.gfx.flush(ctx, RADEON_FLUSH_ASYNC, NULL);
}
/**
* Global buffers are not really resources, they are are actually offsets
* into a single global resource (r600_screen::global_pool). The means
- * they don't have their own cs_buf handle, so they cannot be passed
+ * they don't have their own buf handle, so they cannot be passed
* to r600_copy_buffer() and must be handled separately.
- *
- * XXX: It should be possible to implement this function using
- * r600_copy_buffer() by passing the memory_pool resource as both src
- * and dst and updating dstx and src_box to point to the correct offsets.
- * This would likely perform better than the current implementation.
*/
static void r600_copy_global_buffer(struct pipe_context *ctx,
struct pipe_resource *dst, unsigned
dstx, struct pipe_resource *src,
const struct pipe_box *src_box)
{
- struct pipe_box dst_box; struct pipe_transfer *src_pxfer,
- *dst_pxfer;
-
- u_box_1d(dstx, src_box->width, &dst_box);
- void *src_ptr = ctx->transfer_map(ctx, src, 0, PIPE_TRANSFER_READ,
- src_box, &src_pxfer);
- void *dst_ptr = ctx->transfer_map(ctx, dst, 0, PIPE_TRANSFER_WRITE,
- &dst_box, &dst_pxfer);
- memcpy(dst_ptr, src_ptr, src_box->width);
-
- ctx->transfer_unmap(ctx, src_pxfer);
- ctx->transfer_unmap(ctx, dst_pxfer);
+ struct r600_context *rctx = (struct r600_context*)ctx;
+ struct compute_memory_pool *pool = rctx->screen->global_pool;
+ struct pipe_box new_src_box = *src_box;
+
+ if (src->bind & PIPE_BIND_GLOBAL) {
+ struct r600_resource_global *rsrc =
+ (struct r600_resource_global *)src;
+ struct compute_memory_item *item = rsrc->chunk;
+
+ if (is_item_in_pool(item)) {
+ new_src_box.x += 4 * item->start_in_dw;
+ src = (struct pipe_resource *)pool->bo;
+ } else {
+ if (item->real_buffer == NULL) {
+ item->real_buffer =
+ r600_compute_buffer_alloc_vram(pool->screen,
+ item->size_in_dw * 4);
+ }
+ src = (struct pipe_resource*)item->real_buffer;
+ }
+ }
+ if (dst->bind & PIPE_BIND_GLOBAL) {
+ struct r600_resource_global *rdst =
+ (struct r600_resource_global *)dst;
+ struct compute_memory_item *item = rdst->chunk;
+
+ if (is_item_in_pool(item)) {
+ dstx += 4 * item->start_in_dw;
+ dst = (struct pipe_resource *)pool->bo;
+ } else {
+ if (item->real_buffer == NULL) {
+ item->real_buffer =
+ r600_compute_buffer_alloc_vram(pool->screen,
+ item->size_in_dw * 4);
+ }
+ dst = (struct pipe_resource*)item->real_buffer;
+ }
+ }
+
+ r600_copy_buffer(ctx, dst, dstx, src, &new_src_box);
}
static void r600_clear_buffer(struct pipe_context *ctx, struct pipe_resource *dst,
- unsigned offset, unsigned size, unsigned value)
+ unsigned offset, unsigned size, unsigned value,
+ bool is_framebuffer)
{
struct r600_context *rctx = (struct r600_context*)ctx;
} else {
uint32_t *map = r600_buffer_map_sync_with_rings(&rctx->b, r600_resource(dst),
PIPE_TRANSFER_WRITE);
+ map += offset / 4;
size /= 4;
for (unsigned i = 0; i < size; i++)
*map++ = value;
}
}
-static bool util_format_is_subsampled_2x1_32bpp(enum pipe_format format)
-{
- 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,
- 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)
+void r600_resource_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)
{
struct r600_context *rctx = (struct r600_context *)ctx;
struct pipe_surface *dst_view, dst_templ;
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)) {
+ if (util_format_is_compressed(src->format) ||
+ util_format_is_compressed(dst->format)) {
unsigned blocksize = util_format_get_blocksize(src->format);
if (blocksize == 8)
src_force_level = src_level;
} else if (!util_blitter_is_copy_supported(rctx->blitter, dst, src)) {
- if (util_format_is_subsampled_2x1_32bpp(src->format)) {
+ if (util_format_is_subsampled_422(src->format)) {
src_templ.format = PIPE_FORMAT_R8G8B8A8_UINT;
dst_templ.format = PIPE_FORMAT_R8G8B8A8_UINT;
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);
+ PIPE_MASK_RGBAZS, PIPE_TEX_FILTER_NEAREST, NULL,
+ FALSE);
r600_blitter_end(ctx);
pipe_surface_reference(&dst_view, NULL);
info->src.box.width == dst_width &&
info->src.box.height == dst_height &&
info->src.box.depth == 1 &&
- dst->surface.level[info->dst.level].mode >= RADEON_SURF_MODE_1D) {
- r600_blitter_begin(ctx, R600_COLOR_RESOLVE);
+ dst->surface.level[info->dst.level].mode >= RADEON_SURF_MODE_1D &&
+ (!dst->cmask.size || !dst->dirty_level_mask) /* dst cannot be fast-cleared */) {
+ r600_blitter_begin(ctx, R600_COLOR_RESOLVE |
+ (info->render_condition_enable ? 0 : R600_DISABLE_RENDER_COND));
util_blitter_custom_resolve_color(rctx->blitter,
info->dst.resource, info->dst.level,
info->dst.box.z,
return; /* error */
}
- r600_blitter_begin(ctx, R600_BLIT);
+ if (rctx->screen->b.debug_flags & DBG_FORCE_DMA &&
+ util_try_blit_via_copy_region(ctx, info))
+ return;
+
+ r600_blitter_begin(ctx, R600_BLIT |
+ (info->render_condition_enable ? 0 : R600_DISABLE_RENDER_COND));
util_blitter_blit(rctx->blitter, info);
r600_blitter_end(ctx);
}
if (!rtex->is_depth && rtex->cmask.size) {
r600_blit_decompress_color(ctx, rtex, 0, res->last_level,
- 0, res->array_size - 1);
+ 0, util_max_layer(res, 0));
}
}
projects / mesa.git / blobdiff