{
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);
- util_blitter_save_rasterizer(sctx->blitter, sctx->queued.named.rasterizer);
- util_blitter_save_fragment_shader(sctx->blitter, sctx->ps_shader);
- util_blitter_save_geometry_shader(sctx->blitter, sctx->gs_shader);
- util_blitter_save_vertex_shader(sctx->blitter, sctx->vs_shader);
- util_blitter_save_vertex_elements(sctx->blitter, sctx->vertex_elements);
- if (sctx->queued.named.sample_mask) {
- util_blitter_save_sample_mask(sctx->blitter,
- sctx->queued.named.sample_mask->sample_mask);
- }
- if (sctx->queued.named.viewport[0]) {
- util_blitter_save_viewport(sctx->blitter, &sctx->queued.named.viewport[0]->viewport);
- }
- if (sctx->queued.named.scissor[0]) {
- util_blitter_save_scissor(sctx->blitter, &sctx->queued.named.scissor[0]->scissor);
- }
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_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);
}
struct pipe_surface *zsurf, *cbsurf, surf_tmpl;
sctx->dbcb_copy_sample = sample;
- sctx->db_render_state.dirty = true;
+ si_mark_atom_dirty(sctx, &sctx->db_render_state);
surf_tmpl.format = texture->resource.b.b.format;
surf_tmpl.u.tex.level = level;
sctx->dbcb_depth_copy_enabled = false;
sctx->dbcb_stencil_copy_enabled = false;
- sctx->db_render_state.dirty = true;
+ si_mark_atom_dirty(sctx, &sctx->db_render_state);
}
static void si_blit_decompress_depth_in_place(struct si_context *sctx,
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;
-
- sctx->db_inplace_flush_enabled = true;
- sctx->db_render_state.dirty = true;
+ unsigned *dirty_level_mask;
+
+ if (is_stencil_sampler) {
+ sctx->db_flush_stencil_inplace = true;
+ dirty_level_mask = &texture->stencil_dirty_level_mask;
+ } else {
+ sctx->db_flush_depth_inplace = true;
+ dirty_level_mask = &texture->dirty_level_mask;
+ }
+ si_mark_atom_dirty(sctx, &sctx->db_render_state);
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 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);
}
}
- sctx->db_inplace_flush_enabled = false;
- sctx->db_render_state.dirty = true;
+ sctx->db_flush_depth_inplace = false;
+ sctx->db_flush_stencil_inplace = false;
+ si_mark_atom_dirty(sctx, &sctx->db_render_state);
}
void si_flush_depth_textures(struct si_context *sctx,
while (mask) {
struct pipe_sampler_view *view;
+ struct si_sampler_view *sview;
struct r600_texture *tex;
i = u_bit_scan(&mask);
view = textures->views.views[i];
assert(view);
+ sview = (struct si_sampler_view*)view;
tex = (struct r600_texture *)view->texture;
assert(tex->is_depth && !tex->is_flushing_texture);
si_blit_decompress_depth_in_place(sctx, tex,
+ sview->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));
}
assert(view);
tex = (struct r600_texture *)view->texture;
- assert(tex->cmask.size || tex->fmask.size);
+ assert(tex->cmask.size || tex->fmask.size || tex->dcc_buffer);
si_blit_decompress_color(&sctx->b.b, tex,
view->u.tex.first_level, view->u.tex.last_level,
zsbuf ? (struct r600_texture*)zsbuf->texture : NULL;
if (buffers & PIPE_CLEAR_COLOR) {
- evergreen_do_fast_color_clear(&sctx->b, fb, &sctx->framebuffer.atom,
- &buffers, color);
+ evergreen_do_fast_color_clear(&sctx->b, fb,
+ &sctx->framebuffer.atom, &buffers,
+ &sctx->framebuffer.dirty_cbufs,
+ color);
+ if (!buffers)
+ return; /* all buffers have been fast cleared */
}
if (buffers & PIPE_CLEAR_COLOR) {
}
}
- 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.atom.dirty = true; /* updates DB_DEPTH_CLEAR */
- sctx->db_depth_clear = true;
- sctx->db_render_state.dirty = true;
+ 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);
sctx->db_depth_clear = false;
sctx->db_depth_disable_expclear = false;
zstex->depth_cleared = true;
- sctx->db_render_state.dirty = 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);
}
}
struct r600_texture *rtex = (struct r600_texture*)tex;
if (rtex->is_depth && !rtex->is_flushing_texture) {
- si_blit_decompress_depth_in_place(sctx, rtex,
+ si_blit_decompress_depth_in_place(sctx, rtex, false,
level, level,
first_layer, last_layer);
- } else if (rtex->fmask.size || rtex->cmask.size) {
+ if (rtex->surface.flags & RADEON_SURF_SBUFFER)
+ 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) {
si_blit_decompress_color(ctx, rtex, level, level,
first_layer, last_layer);
}
unsigned npix0_y;
};
-static void si_compressed_to_blittable(struct pipe_resource *tex,
- unsigned level,
- struct texture_orig_info *orig)
-{
- struct r600_texture *rtex = (struct r600_texture*)tex;
- unsigned pixsize = util_format_get_blocksize(rtex->resource.b.b.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);
-
- /* By dividing the dimensions by 4, we effectively decrement
- * last_level by 2, therefore the last 2 mipmap levels disappear and
- * aren't blittable. Note that the last 3 mipmap levels (4x4, 2x2,
- * 1x1) have equal slice sizes, which is an important assumption
- * for this to work.
- *
- * In order to make the last 2 mipmap levels blittable, we have to
- * add the slice size of the last mipmap level to the texture
- * address, so that even though the hw thinks it reads last_level-2,
- * it will actually read last_level-1, and if we add the slice size*2,
- * it will read last_level. That's how this workaround works.
- */
- if (level > rtex->resource.b.b.last_level-2)
- rtex->mipmap_shift = level - (rtex->resource.b.b.last_level-2);
-}
-
-static void si_change_format(struct pipe_resource *tex,
- unsigned level,
- struct texture_orig_info *orig,
- enum pipe_format format)
-{
- struct r600_texture *rtex = (struct r600_texture*)tex;
-
- 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;
-
- tex->format = format;
-}
-
-static void si_reset_blittable_to_orig(struct pipe_resource *tex,
- unsigned level,
- struct texture_orig_info *orig)
-{
- struct r600_texture *rtex = (struct r600_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;
- rtex->mipmap_shift = 0;
-}
-
void si_resource_copy_region(struct pipe_context *ctx,
struct pipe_resource *dst,
unsigned dst_level,
const struct pipe_box *src_box)
{
struct si_context *sctx = (struct si_context *)ctx;
- struct r600_texture *rdst = (struct r600_texture*)dst;
struct pipe_surface *dst_view, dst_templ;
struct pipe_sampler_view src_templ, *src_view;
- struct texture_orig_info orig_info[2];
+ unsigned dst_width, dst_height, src_width0, src_height0;
+ unsigned src_force_level = 0;
struct pipe_box sbox, dstbox;
- boolean restore_orig[2];
- /* Fallback for buffers. */
+ /* Handle buffers first. */
if (dst->target == PIPE_BUFFER && src->target == PIPE_BUFFER) {
si_copy_buffer(sctx, dst, src, dstx, src_box->x, src_box->width, false);
return;
}
- memset(orig_info, 0, sizeof(orig_info));
+ assert(u_max_sample(dst) == u_max_sample(src));
/* The driver doesn't decompress resources automatically while
* u_blitter is rendering. */
si_decompress_subresource(ctx, src, src_level,
src_box->z, src_box->z + src_box->depth - 1);
- restore_orig[0] = restore_orig[1] = FALSE;
+ dst_width = u_minify(dst->width0, dst_level);
+ dst_height = u_minify(dst->height0, dst_level);
+ src_width0 = src->width0;
+ src_height0 = src->height0;
+
+ 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)) {
- si_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);
+
+ 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;
src_box = &sbox;
- si_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);
- } else if (!util_blitter_is_copy_supported(sctx->blitter, dst, src)) {
+ src_force_level = src_level;
+ } else if (!util_blitter_is_copy_supported(sctx->blitter, dst, src) ||
+ /* also *8_SNORM has precision issues, use UNORM instead */
+ util_format_is_snorm8(src->format)) {
if (util_format_is_subsampled_422(src->format)) {
- /* XXX untested */
- si_change_format(src, src_level, &orig_info[0],
- PIPE_FORMAT_R8G8B8A8_UINT);
- si_change_format(dst, dst_level, &orig_info[1],
- PIPE_FORMAT_R8G8B8A8_UINT);
+ 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);
+
+ dstx = util_format_get_nblocksx(dst->format, dstx);
sbox = *src_box;
- sbox.x = util_format_get_nblocksx(orig_info[0].format, src_box->x);
- sbox.width = util_format_get_nblocksx(orig_info[0].format, src_box->width);
+ 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;
- dstx = util_format_get_nblocksx(orig_info[1].format, dstx);
-
- restore_orig[0] = TRUE;
- restore_orig[1] = TRUE;
} else {
unsigned blocksize = util_format_get_blocksize(src->format);
switch (blocksize) {
case 1:
- si_change_format(src, src_level, &orig_info[0],
- PIPE_FORMAT_R8_UNORM);
- si_change_format(dst, dst_level, &orig_info[1],
- PIPE_FORMAT_R8_UNORM);
+ dst_templ.format = PIPE_FORMAT_R8_UNORM;
+ src_templ.format = PIPE_FORMAT_R8_UNORM;
break;
case 2:
- si_change_format(src, src_level, &orig_info[0],
- PIPE_FORMAT_R8G8_UNORM);
- si_change_format(dst, dst_level, &orig_info[1],
- PIPE_FORMAT_R8G8_UNORM);
+ dst_templ.format = PIPE_FORMAT_R8G8_UNORM;
+ src_templ.format = PIPE_FORMAT_R8G8_UNORM;
break;
case 4:
- si_change_format(src, src_level, &orig_info[0],
- PIPE_FORMAT_R8G8B8A8_UNORM);
- si_change_format(dst, dst_level, &orig_info[1],
- PIPE_FORMAT_R8G8B8A8_UNORM);
+ dst_templ.format = PIPE_FORMAT_R8G8B8A8_UNORM;
+ src_templ.format = PIPE_FORMAT_R8G8B8A8_UNORM;
break;
case 8:
- si_change_format(src, src_level, &orig_info[0],
- PIPE_FORMAT_R16G16B16A16_UINT);
- si_change_format(dst, dst_level, &orig_info[1],
- PIPE_FORMAT_R16G16B16A16_UINT);
+ dst_templ.format = PIPE_FORMAT_R16G16B16A16_UINT;
+ src_templ.format = PIPE_FORMAT_R16G16B16A16_UINT;
break;
case 16:
- si_change_format(src, src_level, &orig_info[0],
- PIPE_FORMAT_R32G32B32A32_UINT);
- si_change_format(dst, dst_level, &orig_info[1],
- PIPE_FORMAT_R32G32B32A32_UINT);
+ 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);
}
- restore_orig[0] = TRUE;
- restore_orig[1] = TRUE;
}
}
/* Initialize the surface. */
- util_blitter_default_dst_texture(&dst_templ, dst, dst_level, dstz);
dst_view = r600_create_surface_custom(ctx, dst, &dst_templ,
- rdst->surface.level[dst_level].npix_x,
- rdst->surface.level[dst_level].npix_y);
+ dst_width, dst_height);
/* Initialize the sampler view. */
- util_blitter_default_src_texture(&src_templ, src, src_level);
- src_view = ctx->create_sampler_view(ctx, src, &src_templ);
+ src_view = si_create_sampler_view_custom(ctx, src, &src_templ,
+ src_width0, src_height0,
+ src_force_level);
u_box_3d(dstx, dsty, dstz, abs(src_box->width), abs(src_box->height),
abs(src_box->depth), &dstbox);
/* Copy. */
si_blitter_begin(ctx, SI_COPY);
util_blitter_blit_generic(sctx->blitter, dst_view, &dstbox,
- src_view, src_box, src->width0, src->height0,
- PIPE_MASK_RGBAZS, PIPE_TEX_FILTER_NEAREST, NULL);
+ src_view, src_box, src_width0, src_height0,
+ PIPE_MASK_RGBAZS, PIPE_TEX_FILTER_NEAREST, NULL,
+ FALSE);
si_blitter_end(ctx);
pipe_surface_reference(&dst_view, NULL);
pipe_sampler_view_reference(&src_view, NULL);
-
- if (restore_orig[0])
- si_reset_blittable_to_orig(src, src_level, &orig_info[0]);
-
- if (restore_orig[1])
- si_reset_blittable_to_orig(dst, dst_level, &orig_info[1]);
}
/* For MSAA integer resolving to work, we change the format to NORM using this function. */
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 &&
info->src.box.depth == 1 &&
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->cmask.size || !dst->dirty_level_mask) && /* dst cannot be fast-cleared */
+ !dst->dcc_buffer) {
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_buffer)) {
si_blit_decompress_color(ctx, rtex, 0, res->last_level,
0, util_max_layer(res, 0));
}
}
+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;
projects / mesa.git / blobdiff