/*
* Copyright 2010 Jerome Glisse <glisse@freedesktop.org>
+ * Copyright 2015 Advanced Micro Devices, Inc.
+ * 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"),
#include "si_pipe.h"
#include "si_compute.h"
-#include "util/u_format.h"
+#include "util/format/u_format.h"
#include "util/u_log.h"
#include "util/u_surface.h"
SI_COLOR_RESOLVE = SI_SAVE_FRAMEBUFFER | SI_SAVE_FRAGMENT_STATE
};
-void si_blitter_begin(struct pipe_context *ctx, enum si_blitter_op op)
+void si_blitter_begin(struct si_context *sctx, enum si_blitter_op op)
{
- struct si_context *sctx = (struct si_context *)ctx;
-
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_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_scissor(sctx->blitter, &sctx->scissors.states[0]);
+ util_blitter_save_sample_mask(sctx->blitter, sctx->sample_mask);
+ util_blitter_save_scissor(sctx->blitter, &sctx->scissors[0]);
+ util_blitter_save_window_rectangles(sctx->blitter,
+ sctx->window_rectangles_include,
+ sctx->num_window_rectangles,
+ sctx->window_rectangles);
}
if (op & SI_SAVE_FRAMEBUFFER)
}
if (op & SI_DISABLE_RENDER_COND)
- sctx->b.render_cond_force_off = true;
+ sctx->render_cond_force_off = true;
+
+ if (sctx->screen->dpbb_allowed) {
+ sctx->dpbb_force_off = true;
+ si_mark_atom_dirty(sctx, &sctx->atoms.s.dpbb_state);
+ }
}
-void si_blitter_end(struct pipe_context *ctx)
+void si_blitter_end(struct si_context *sctx)
{
- struct si_context *sctx = (struct si_context *)ctx;
+ if (sctx->screen->dpbb_allowed) {
+ sctx->dpbb_force_off = false;
+ si_mark_atom_dirty(sctx, &sctx->atoms.s.dpbb_state);
+ }
- sctx->b.render_cond_force_off = false;
+ sctx->render_cond_force_off = false;
/* Restore shader pointers because the VS blit shader changed all
* non-global VS user SGPRs. */
sctx->shader_pointers_dirty |= SI_DESCS_SHADER_MASK(VERTEX);
- sctx->vertex_buffer_pointer_dirty = true;
- si_mark_atom_dirty(sctx, &sctx->shader_pointers.atom);
+ sctx->vertex_buffer_pointer_dirty = sctx->vb_descriptors_buffer != NULL;
+ si_mark_atom_dirty(sctx, &sctx->atoms.s.shader_pointers);
}
static unsigned u_max_sample(struct pipe_resource *r)
static unsigned
si_blit_dbcb_copy(struct si_context *sctx,
- struct r600_texture *src,
- struct r600_texture *dst,
+ struct si_texture *src,
+ struct si_texture *dst,
unsigned planes, unsigned level_mask,
unsigned first_layer, unsigned last_layer,
unsigned first_sample, unsigned last_sample)
sctx->dbcb_depth_copy_enabled = true;
if (planes & PIPE_MASK_S)
sctx->dbcb_stencil_copy_enabled = true;
- si_mark_atom_dirty(sctx, &sctx->db_render_state);
+ si_mark_atom_dirty(sctx, &sctx->atoms.s.db_render_state);
assert(sctx->dbcb_depth_copy_enabled || sctx->dbcb_stencil_copy_enabled);
/* The smaller the mipmap level, the less layers there are
* as far as 3D textures are concerned. */
- max_layer = util_max_layer(&src->resource.b.b, level);
+ max_layer = util_max_layer(&src->buffer.b.b, level);
checked_last_layer = MIN2(last_layer, max_layer);
surf_tmpl.u.tex.level = level;
for (layer = first_layer; layer <= checked_last_layer; layer++) {
struct pipe_surface *zsurf, *cbsurf;
- surf_tmpl.format = src->resource.b.b.format;
+ surf_tmpl.format = src->buffer.b.b.format;
surf_tmpl.u.tex.first_layer = layer;
surf_tmpl.u.tex.last_layer = layer;
- zsurf = sctx->b.b.create_surface(&sctx->b.b, &src->resource.b.b, &surf_tmpl);
+ zsurf = sctx->b.create_surface(&sctx->b, &src->buffer.b.b, &surf_tmpl);
- surf_tmpl.format = dst->resource.b.b.format;
- cbsurf = sctx->b.b.create_surface(&sctx->b.b, &dst->resource.b.b, &surf_tmpl);
+ surf_tmpl.format = dst->buffer.b.b.format;
+ cbsurf = sctx->b.create_surface(&sctx->b, &dst->buffer.b.b, &surf_tmpl);
for (sample = first_sample; sample <= last_sample; sample++) {
if (sample != sctx->dbcb_copy_sample) {
sctx->dbcb_copy_sample = sample;
- si_mark_atom_dirty(sctx, &sctx->db_render_state);
+ si_mark_atom_dirty(sctx, &sctx->atoms.s.db_render_state);
}
- si_blitter_begin(&sctx->b.b, SI_DECOMPRESS);
+ si_blitter_begin(sctx, SI_DECOMPRESS);
util_blitter_custom_depth_stencil(sctx->blitter, zsurf, cbsurf, 1 << sample,
sctx->custom_dsa_flush, 1.0f);
- si_blitter_end(&sctx->b.b);
+ si_blitter_end(sctx);
}
pipe_surface_reference(&zsurf, NULL);
}
if (first_layer == 0 && last_layer >= max_layer &&
- first_sample == 0 && last_sample >= u_max_sample(&src->resource.b.b))
+ first_sample == 0 && last_sample >= u_max_sample(&src->buffer.b.b))
fully_copied_levels |= 1u << level;
}
sctx->decompression_enabled = false;
sctx->dbcb_depth_copy_enabled = false;
sctx->dbcb_stencil_copy_enabled = false;
- si_mark_atom_dirty(sctx, &sctx->db_render_state);
+ si_mark_atom_dirty(sctx, &sctx->atoms.s.db_render_state);
return fully_copied_levels;
}
-static void si_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)
-{
- const struct util_format_description *desc;
- unsigned planes = 0;
-
- assert(staging != NULL && "use si_blit_decompress_zs_in_place instead");
-
- desc = util_format_description(staging->resource.b.b.format);
-
- if (util_format_has_depth(desc))
- planes |= PIPE_MASK_Z;
- if (util_format_has_stencil(desc))
- planes |= PIPE_MASK_S;
-
- si_blit_dbcb_copy(
- (struct si_context *)ctx, texture, staging, planes,
- u_bit_consecutive(first_level, last_level - first_level + 1),
- first_layer, last_layer, first_sample, last_sample);
-}
-
/* Helper function for si_blit_decompress_zs_in_place.
*/
static void
si_blit_decompress_zs_planes_in_place(struct si_context *sctx,
- struct r600_texture *texture,
+ struct si_texture *texture,
unsigned planes, unsigned level_mask,
unsigned first_layer, unsigned last_layer)
{
sctx->db_flush_stencil_inplace = true;
if (planes & PIPE_MASK_Z)
sctx->db_flush_depth_inplace = true;
- si_mark_atom_dirty(sctx, &sctx->db_render_state);
+ si_mark_atom_dirty(sctx, &sctx->atoms.s.db_render_state);
- surf_tmpl.format = texture->resource.b.b.format;
+ surf_tmpl.format = texture->buffer.b.b.format;
sctx->decompression_enabled = true;
/* 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);
+ max_layer = util_max_layer(&texture->buffer.b.b, level);
checked_last_layer = MIN2(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;
- zsurf = sctx->b.b.create_surface(&sctx->b.b, &texture->resource.b.b, &surf_tmpl);
+ zsurf = sctx->b.create_surface(&sctx->b, &texture->buffer.b.b, &surf_tmpl);
- si_blitter_begin(&sctx->b.b, SI_DECOMPRESS);
+ si_blitter_begin(sctx, SI_DECOMPRESS);
util_blitter_custom_depth_stencil(sctx->blitter, zsurf, NULL, ~0,
sctx->custom_dsa_flush,
1.0f);
- si_blitter_end(&sctx->b.b);
+ si_blitter_end(sctx);
pipe_surface_reference(&zsurf, NULL);
}
sctx->decompression_enabled = false;
sctx->db_flush_depth_inplace = false;
sctx->db_flush_stencil_inplace = false;
- si_mark_atom_dirty(sctx, &sctx->db_render_state);
+ si_mark_atom_dirty(sctx, &sctx->atoms.s.db_render_state);
}
/* Helper function of si_flush_depth_texture: decompress the given levels
*/
static void
si_blit_decompress_zs_in_place(struct si_context *sctx,
- struct r600_texture *texture,
+ struct si_texture *texture,
unsigned levels_z, unsigned levels_s,
unsigned first_layer, unsigned last_layer)
{
static void
si_decompress_depth(struct si_context *sctx,
- struct r600_texture *tex,
+ struct si_texture *tex,
unsigned required_planes,
unsigned first_level, unsigned last_level,
unsigned first_layer, unsigned last_layer)
}
}
- if (unlikely(sctx->b.log))
- u_log_printf(sctx->b.log,
+ if (unlikely(sctx->log))
+ u_log_printf(sctx->log,
"\n------------------------------------------------\n"
"Decompress Depth (levels %u - %u, levels Z: 0x%x S: 0x%x)\n\n",
first_level, last_level, levels_z, levels_s);
*/
if (copy_planes &&
(tex->flushed_depth_texture ||
- si_init_flushed_depth_texture(&sctx->b.b, &tex->resource.b.b, NULL))) {
- struct r600_texture *dst = tex->flushed_depth_texture;
+ si_init_flushed_depth_texture(&sctx->b, &tex->buffer.b.b))) {
+ struct si_texture *dst = tex->flushed_depth_texture;
unsigned fully_copied_levels;
unsigned levels = 0;
assert(tex->flushed_depth_texture);
- if (util_format_is_depth_and_stencil(dst->resource.b.b.format))
+ if (util_format_is_depth_and_stencil(dst->buffer.b.b.format))
copy_planes = PIPE_MASK_Z | PIPE_MASK_S;
if (copy_planes & PIPE_MASK_Z) {
fully_copied_levels = si_blit_dbcb_copy(
sctx, tex, dst, copy_planes, levels,
first_layer, last_layer,
- 0, u_max_sample(&tex->resource.b.b));
+ 0, u_max_sample(&tex->buffer.b.b));
if (copy_planes & PIPE_MASK_Z)
tex->dirty_level_mask &= ~fully_copied_levels;
}
if (inplace_planes) {
- bool has_htile = si_htile_enabled(tex, first_level);
- bool tc_compat_htile = vi_tc_compat_htile_enabled(tex, first_level);
+ bool has_htile = si_htile_enabled(tex, first_level, inplace_planes);
+ bool tc_compat_htile = vi_tc_compat_htile_enabled(tex, first_level,
+ inplace_planes);
/* Don't decompress if there is no HTILE or when HTILE is
* TC-compatible. */
/* Only in-place decompression needs to flush DB caches, or
* when we don't decompress but TC-compatible planes are dirty.
*/
- si_make_DB_shader_coherent(sctx, tex->resource.b.b.nr_samples,
+ si_make_DB_shader_coherent(sctx, tex->buffer.b.b.nr_samples,
inplace_planes & PIPE_MASK_S,
tc_compat_htile);
}
/* set_framebuffer_state takes care of coherency for single-sample.
* The DB->CB copy uses CB for the final writes.
*/
- if (copy_planes && tex->resource.b.b.nr_samples > 1)
- si_make_CB_shader_coherent(sctx, tex->resource.b.b.nr_samples,
- false);
+ if (copy_planes && tex->buffer.b.b.nr_samples > 1)
+ si_make_CB_shader_coherent(sctx, tex->buffer.b.b.nr_samples,
+ false, true /* no DCC */);
}
static void
while (mask) {
struct pipe_sampler_view *view;
struct si_sampler_view *sview;
- struct r600_texture *tex;
+ struct si_texture *tex;
i = u_bit_scan(&mask);
assert(view);
sview = (struct si_sampler_view*)view;
- tex = (struct r600_texture *)view->texture;
+ tex = (struct si_texture *)view->texture;
assert(tex->db_compatible);
si_decompress_depth(sctx, tex,
sview->is_stencil_sampler ? PIPE_MASK_S : PIPE_MASK_Z,
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->buffer.b.b, view->u.tex.first_level));
}
}
-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,
- bool need_dcc_decompress)
+static void si_blit_decompress_color(struct si_context *sctx,
+ struct si_texture *tex,
+ unsigned first_level, unsigned last_level,
+ unsigned first_layer, unsigned last_layer,
+ bool need_dcc_decompress,
+ bool need_fmask_expand)
{
- struct si_context *sctx = (struct si_context *)ctx;
void* custom_blend;
unsigned layer, checked_last_layer, max_layer;
unsigned level_mask =
u_bit_consecutive(first_level, last_level - first_level + 1);
if (!need_dcc_decompress)
- level_mask &= rtex->dirty_level_mask;
+ level_mask &= tex->dirty_level_mask;
if (!level_mask)
return;
- if (unlikely(sctx->b.log))
- u_log_printf(sctx->b.log,
+ if (unlikely(sctx->log))
+ u_log_printf(sctx->log,
"\n------------------------------------------------\n"
"Decompress Color (levels %u - %u, mask 0x%x)\n\n",
first_level, last_level, level_mask);
if (need_dcc_decompress) {
custom_blend = sctx->custom_blend_dcc_decompress;
- assert(rtex->dcc_offset);
+ assert(tex->surface.dcc_offset);
/* disable levels without DCC */
for (int i = first_level; i <= last_level; i++) {
- if (!vi_dcc_enabled(rtex, i))
+ if (!vi_dcc_enabled(tex, i))
level_mask &= ~(1 << i);
}
- } else if (rtex->fmask.size) {
+ } else if (tex->surface.fmask_size) {
custom_blend = sctx->custom_blend_fmask_decompress;
} else {
custom_blend = sctx->custom_blend_eliminate_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);
+ max_layer = util_max_layer(&tex->buffer.b.b, level);
checked_last_layer = MIN2(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.format = tex->buffer.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);
+ cbsurf = sctx->b.create_surface(&sctx->b, &tex->buffer.b.b, &surf_tmpl);
/* Required before and after FMASK and DCC_DECOMPRESS. */
if (custom_blend == sctx->custom_blend_fmask_decompress ||
custom_blend == sctx->custom_blend_dcc_decompress)
- sctx->b.flags |= SI_CONTEXT_FLUSH_AND_INV_CB;
+ sctx->flags |= SI_CONTEXT_FLUSH_AND_INV_CB;
- si_blitter_begin(ctx, SI_DECOMPRESS);
+ si_blitter_begin(sctx, SI_DECOMPRESS);
util_blitter_custom_color(sctx->blitter, cbsurf, custom_blend);
- si_blitter_end(ctx);
+ si_blitter_end(sctx);
if (custom_blend == sctx->custom_blend_fmask_decompress ||
custom_blend == sctx->custom_blend_dcc_decompress)
- sctx->b.flags |= SI_CONTEXT_FLUSH_AND_INV_CB;
+ sctx->flags |= SI_CONTEXT_FLUSH_AND_INV_CB;
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);
+ tex->dirty_level_mask &= ~(1 << level);
}
}
sctx->decompression_enabled = false;
- si_make_CB_shader_coherent(sctx, rtex->resource.b.b.nr_samples,
- vi_dcc_enabled(rtex, first_level));
+ si_make_CB_shader_coherent(sctx, tex->buffer.b.b.nr_samples,
+ vi_dcc_enabled(tex, first_level),
+ tex->surface.u.gfx9.dcc.pipe_aligned);
+
+ if (need_fmask_expand && tex->surface.fmask_offset && tex->fmask_is_not_identity) {
+ si_compute_expand_fmask(&sctx->b, &tex->buffer.b.b);
+ tex->fmask_is_not_identity = false;
+ }
}
static void
-si_decompress_color_texture(struct si_context *sctx, struct r600_texture *tex,
- unsigned first_level, unsigned last_level)
+si_decompress_color_texture(struct si_context *sctx, struct si_texture *tex,
+ unsigned first_level, unsigned last_level,
+ bool need_fmask_expand)
{
/* CMASK or DCC can be discarded and we can still end up here. */
- if (!tex->cmask.size && !tex->fmask.size && !tex->dcc_offset)
+ if (!tex->cmask_buffer && !tex->surface.fmask_size && !tex->surface.dcc_offset)
return;
- si_blit_decompress_color(&sctx->b.b, tex, first_level, last_level, 0,
- util_max_layer(&tex->resource.b.b, first_level),
- false);
+ si_blit_decompress_color(sctx, tex, first_level, last_level, 0,
+ util_max_layer(&tex->buffer.b.b, first_level),
+ false, need_fmask_expand);
}
static void
while (mask) {
struct pipe_sampler_view *view;
- struct r600_texture *tex;
+ struct si_texture *tex;
i = u_bit_scan(&mask);
view = textures->views[i];
assert(view);
- tex = (struct r600_texture *)view->texture;
+ tex = (struct si_texture *)view->texture;
si_decompress_color_texture(sctx, tex, view->u.tex.first_level,
- view->u.tex.last_level);
+ view->u.tex.last_level, false);
}
}
while (mask) {
const struct pipe_image_view *view;
- struct r600_texture *tex;
+ struct si_texture *tex;
i = u_bit_scan(&mask);
view = &images->views[i];
assert(view->resource->target != PIPE_BUFFER);
- tex = (struct r600_texture *)view->resource;
+ tex = (struct si_texture *)view->resource;
si_decompress_color_texture(sctx, tex, view->u.tex.level,
- view->u.tex.level);
+ view->u.tex.level,
+ view->access & PIPE_IMAGE_ACCESS_WRITE);
}
}
static void si_check_render_feedback_texture(struct si_context *sctx,
- struct r600_texture *tex,
+ struct si_texture *tex,
unsigned first_level,
unsigned last_level,
unsigned first_layer,
{
bool render_feedback = false;
- if (!tex->dcc_offset)
+ if (!tex->surface.dcc_offset)
return;
for (unsigned j = 0; j < sctx->framebuffer.state.nr_cbufs; ++j) {
- struct r600_surface * surf;
+ struct si_surface * surf;
if (!sctx->framebuffer.state.cbufs[j])
continue;
- surf = (struct r600_surface*)sctx->framebuffer.state.cbufs[j];
+ surf = (struct si_surface*)sctx->framebuffer.state.cbufs[j];
- if (tex == (struct r600_texture *)surf->base.texture &&
+ if (tex == (struct si_texture *)surf->base.texture &&
surf->base.u.tex.level >= first_level &&
surf->base.u.tex.level <= last_level &&
surf->base.u.tex.first_layer <= last_layer &&
}
if (render_feedback)
- si_texture_disable_dcc(&sctx->b, tex);
+ si_texture_disable_dcc(sctx, tex);
}
static void si_check_render_feedback_textures(struct si_context *sctx,
while (mask) {
const struct pipe_sampler_view *view;
- struct r600_texture *tex;
+ struct si_texture *tex;
unsigned i = u_bit_scan(&mask);
if(view->texture->target == PIPE_BUFFER)
continue;
- tex = (struct r600_texture *)view->texture;
+ tex = (struct si_texture *)view->texture;
si_check_render_feedback_texture(sctx, tex,
view->u.tex.first_level,
while (mask) {
const struct pipe_image_view *view;
- struct r600_texture *tex;
+ struct si_texture *tex;
unsigned i = u_bit_scan(&mask);
if (view->resource->target == PIPE_BUFFER)
continue;
- tex = (struct r600_texture *)view->resource;
+ tex = (struct si_texture *)view->resource;
si_check_render_feedback_texture(sctx, tex,
view->u.tex.level,
util_dynarray_foreach(&sctx->resident_tex_handles,
struct si_texture_handle *, tex_handle) {
struct pipe_sampler_view *view;
- struct r600_texture *tex;
+ struct si_texture *tex;
view = (*tex_handle)->view;
if (view->texture->target == PIPE_BUFFER)
continue;
- tex = (struct r600_texture *)view->texture;
+ tex = (struct si_texture *)view->texture;
si_check_render_feedback_texture(sctx, tex,
view->u.tex.first_level,
util_dynarray_foreach(&sctx->resident_img_handles,
struct si_image_handle *, img_handle) {
struct pipe_image_view *view;
- struct r600_texture *tex;
+ struct si_texture *tex;
view = &(*img_handle)->view;
if (view->resource->target == PIPE_BUFFER)
continue;
- tex = (struct r600_texture *)view->resource;
+ tex = (struct si_texture *)view->resource;
si_check_render_feedback_texture(sctx, tex,
view->u.tex.level,
static void si_check_render_feedback(struct si_context *sctx)
{
-
if (!sctx->need_check_render_feedback)
return;
+ /* There is no render feedback if color writes are disabled.
+ * (e.g. a pixel shader with image stores)
+ */
+ if (!si_get_total_colormask(sctx))
+ return;
+
for (int i = 0; i < SI_NUM_SHADERS; ++i) {
si_check_render_feedback_images(sctx, &sctx->images[i]);
si_check_render_feedback_textures(sctx, &sctx->samplers[i]);
util_dynarray_foreach(&sctx->resident_tex_needs_color_decompress,
struct si_texture_handle *, tex_handle) {
struct pipe_sampler_view *view = (*tex_handle)->view;
- struct r600_texture *tex = (struct r600_texture *)view->texture;
+ struct si_texture *tex = (struct si_texture *)view->texture;
si_decompress_color_texture(sctx, tex, view->u.tex.first_level,
- view->u.tex.last_level);
+ view->u.tex.last_level, false);
}
util_dynarray_foreach(&sctx->resident_tex_needs_depth_decompress,
struct si_texture_handle *, tex_handle) {
struct pipe_sampler_view *view = (*tex_handle)->view;
struct si_sampler_view *sview = (struct si_sampler_view *)view;
- struct r600_texture *tex = (struct r600_texture *)view->texture;
+ struct si_texture *tex = (struct si_texture *)view->texture;
si_decompress_depth(sctx, tex,
sview->is_stencil_sampler ? PIPE_MASK_S : PIPE_MASK_Z,
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->buffer.b.b, view->u.tex.first_level));
}
}
util_dynarray_foreach(&sctx->resident_img_needs_color_decompress,
struct si_image_handle *, img_handle) {
struct pipe_image_view *view = &(*img_handle)->view;
- struct r600_texture *tex = (struct r600_texture *)view->resource;
+ struct si_texture *tex = (struct si_texture *)view->resource;
si_decompress_color_texture(sctx, tex, view->u.tex.level,
- view->u.tex.level);
+ view->u.tex.level,
+ view->access & PIPE_IMAGE_ACCESS_WRITE);
}
}
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;
+ compressed_colortex_counter = p_atomic_read(&sctx->screen->compressed_colortex_counter);
+ if (compressed_colortex_counter != sctx->last_compressed_colortex_counter) {
+ sctx->last_compressed_colortex_counter = compressed_colortex_counter;
si_update_needs_color_decompress_masks(sctx);
}
si_decompress_resident_textures(sctx);
if (sctx->uses_bindless_images)
si_decompress_resident_images(sctx);
+
+ if (sctx->ps_uses_fbfetch) {
+ struct pipe_surface *cb0 = sctx->framebuffer.state.cbufs[0];
+ si_decompress_color_texture(sctx,
+ (struct si_texture*)cb0->texture,
+ cb0->u.tex.first_layer,
+ cb0->u.tex.last_layer, false);
+ }
+
+ si_check_render_feedback(sctx);
} else if (shader_mask & (1 << PIPE_SHADER_COMPUTE)) {
- if (sctx->cs_shader_state.program->uses_bindless_samplers)
+ if (sctx->cs_shader_state.program->sel.info.uses_bindless_samplers)
si_decompress_resident_textures(sctx);
- if (sctx->cs_shader_state.program->uses_bindless_images)
+ if (sctx->cs_shader_state.program->sel.info.uses_bindless_images)
si_decompress_resident_images(sctx);
}
-
- si_check_render_feedback(sctx);
}
/* Helper for decompressing a portion of a color or depth resource before
unsigned first_layer, unsigned last_layer)
{
struct si_context *sctx = (struct si_context *)ctx;
- struct r600_texture *rtex = (struct r600_texture*)tex;
+ struct si_texture *stex = (struct si_texture*)tex;
- if (rtex->db_compatible) {
+ if (stex->db_compatible) {
planes &= PIPE_MASK_Z | PIPE_MASK_S;
- if (!rtex->surface.has_stencil)
+ if (!stex->surface.has_stencil)
planes &= ~PIPE_MASK_S;
/* If we've rendered into the framebuffer and it's a blitting
sctx->framebuffer.state.zsbuf->texture == tex)
si_update_fb_dirtiness_after_rendering(sctx);
- si_decompress_depth(sctx, rtex, planes,
+ si_decompress_depth(sctx, stex, planes,
level, level,
first_layer, last_layer);
- } else if (rtex->fmask.size || rtex->cmask.size || rtex->dcc_offset) {
+ } else if (stex->surface.fmask_size || stex->cmask_buffer || stex->surface.dcc_offset) {
/* If we've rendered into the framebuffer and it's a blitting
* source, make sure the decompression pass is invoked
* by dirtying the framebuffer.
}
}
- si_blit_decompress_color(ctx, rtex, level, level,
- first_layer, last_layer, false);
+ si_blit_decompress_color(sctx, stex, level, level,
+ first_layer, last_layer, false, false);
}
}
const struct pipe_box *src_box)
{
struct si_context *sctx = (struct si_context *)ctx;
- struct r600_texture *rsrc = (struct r600_texture*)src;
+ struct si_texture *ssrc = (struct si_texture*)src;
+ struct si_texture *sdst = (struct si_texture*)dst;
struct pipe_surface *dst_view, dst_templ;
struct pipe_sampler_view src_templ, *src_view;
unsigned dst_width, dst_height, src_width0, src_height0;
/* 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, 0);
+ si_copy_buffer(sctx, dst, src, dstx, src_box->x, src_box->width);
+ return;
+ }
+
+ if (!util_format_is_compressed(src->format) &&
+ !util_format_is_compressed(dst->format) &&
+ !util_format_is_depth_or_stencil(src->format) &&
+ src->nr_samples <= 1 &&
+ !sdst->surface.dcc_offset &&
+ !(dst->target != src->target &&
+ (src->target == PIPE_TEXTURE_1D_ARRAY || dst->target == PIPE_TEXTURE_1D_ARRAY))) {
+ si_compute_copy_image(sctx, dst, dst_level, src, src_level, dstx, dsty, dstz, src_box);
return;
}
if (util_format_is_compressed(src->format) ||
util_format_is_compressed(dst->format)) {
- unsigned blocksize = rsrc->surface.bpe;
+ unsigned blocksize = ssrc->surface.bpe;
if (blocksize == 8)
src_templ.format = PIPE_FORMAT_R16G16B16A16_UINT; /* 64-bit block */
sbox.width = util_format_get_nblocksx(src->format, src_box->width);
src_box = &sbox;
} else {
- unsigned blocksize = rsrc->surface.bpe;
+ unsigned blocksize = ssrc->surface.bpe;
switch (blocksize) {
case 1:
* Note that some chips avoid this issue by using SDMA.
*/
if (util_format_is_snorm8(dst_templ.format)) {
- switch (dst_templ.format) {
- case PIPE_FORMAT_R8_SNORM:
- dst_templ.format = src_templ.format = PIPE_FORMAT_R8_SINT;
- break;
- case PIPE_FORMAT_R8G8_SNORM:
- dst_templ.format = src_templ.format = PIPE_FORMAT_R8G8_SINT;
- break;
- case PIPE_FORMAT_R8G8B8X8_SNORM:
- dst_templ.format = src_templ.format = PIPE_FORMAT_R8G8B8X8_SINT;
- break;
- case PIPE_FORMAT_R8G8B8A8_SNORM:
- /* There are no SINT variants for ABGR and XBGR, so we have to use RGBA. */
- case PIPE_FORMAT_A8B8G8R8_SNORM:
- case PIPE_FORMAT_X8B8G8R8_SNORM:
- dst_templ.format = src_templ.format = PIPE_FORMAT_R8G8B8A8_SINT;
- break;
- case PIPE_FORMAT_A8_SNORM:
- dst_templ.format = src_templ.format = PIPE_FORMAT_A8_SINT;
- break;
- case PIPE_FORMAT_L8_SNORM:
- dst_templ.format = src_templ.format = PIPE_FORMAT_L8_SINT;
- break;
- case PIPE_FORMAT_L8A8_SNORM:
- dst_templ.format = src_templ.format = PIPE_FORMAT_L8A8_SINT;
- break;
- case PIPE_FORMAT_I8_SNORM:
- dst_templ.format = src_templ.format = PIPE_FORMAT_I8_SINT;
- break;
- default:; /* fall through */
- }
+ dst_templ.format = src_templ.format =
+ util_format_snorm8_to_sint8(dst_templ.format);
}
- vi_disable_dcc_if_incompatible_format(&sctx->b, dst, dst_level,
+ vi_disable_dcc_if_incompatible_format(sctx, dst, dst_level,
dst_templ.format);
- vi_disable_dcc_if_incompatible_format(&sctx->b, src, src_level,
+ vi_disable_dcc_if_incompatible_format(sctx, src, src_level,
src_templ.format);
/* Initialize the surface. */
abs(src_box->depth), &dstbox);
/* Copy. */
- si_blitter_begin(ctx, SI_COPY);
+ si_blitter_begin(sctx, 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,
false);
- si_blitter_end(ctx);
+ si_blitter_end(sctx);
pipe_surface_reference(&dst_view, NULL);
pipe_sampler_view_reference(&src_view, NULL);
enum pipe_format format)
{
/* Required before and after CB_RESOLVE. */
- sctx->b.flags |= SI_CONTEXT_FLUSH_AND_INV_CB;
+ sctx->flags |= SI_CONTEXT_FLUSH_AND_INV_CB;
- si_blitter_begin(&sctx->b.b, SI_COLOR_RESOLVE |
+ si_blitter_begin(sctx, SI_COLOR_RESOLVE |
(info->render_condition_enable ? 0 : SI_DISABLE_RENDER_COND));
util_blitter_custom_resolve_color(sctx->blitter, dst, dst_level, dst_z,
info->src.resource, info->src.box.z,
~0, sctx->custom_blend_resolve,
format);
- si_blitter_end(&sctx->b.b);
+ si_blitter_end(sctx);
/* Flush caches for possible texturing. */
- si_make_CB_shader_coherent(sctx, 1, false);
+ si_make_CB_shader_coherent(sctx, 1, false, true /* no DCC */);
}
static bool do_hardware_msaa_resolve(struct pipe_context *ctx,
const struct pipe_blit_info *info)
{
struct si_context *sctx = (struct si_context*)ctx;
- struct r600_texture *src = (struct r600_texture*)info->src.resource;
- struct r600_texture *dst = (struct r600_texture*)info->dst.resource;
- MAYBE_UNUSED struct r600_texture *rtmp;
+ struct si_texture *src = (struct si_texture*)info->src.resource;
+ struct si_texture *dst = (struct si_texture*)info->dst.resource;
+ ASSERTED struct si_texture *stmp;
unsigned dst_width = u_minify(info->dst.resource->width0, info->dst.level);
unsigned dst_height = u_minify(info->dst.resource->height0, info->dst.level);
enum pipe_format format = info->src.format;
info->src.box.height == dst_height &&
info->src.box.depth == 1 &&
!dst->surface.is_linear &&
- (!dst->cmask.size || !dst->dirty_level_mask)) { /* dst cannot be fast-cleared */
+ (!dst->cmask_buffer || !dst->dirty_level_mask)) { /* dst cannot be fast-cleared */
/* Check the last constraint. */
if (src->surface.micro_tile_mode != dst->surface.micro_tile_mode) {
/* The next fast clear will switch to this mode to
* get direct hw resolve next time if the mode is
* different now.
+ *
+ * TODO-GFX10: This does not work in GFX10 because MSAA
+ * is restricted to 64KB_R_X and 64KB_Z_X swizzle modes.
+ * In some cases we could change the swizzle of the
+ * destination texture instead, but the more general
+ * solution is to implement compute shader resolve.
*/
src->last_msaa_resolve_target_micro_mode =
dst->surface.micro_tile_mode;
* This is still the fastest codepath even with this clear.
*/
if (vi_dcc_enabled(dst, info->dst.level)) {
- /* TODO: Implement per-level DCC clears for GFX9. */
- if (sctx->b.chip_class >= GFX9 &&
- info->dst.resource->last_level != 0)
+ if (!vi_dcc_clear_level(sctx, dst, info->dst.level,
+ DCC_UNCOMPRESSED))
goto resolve_to_temp;
- vi_dcc_clear_level(sctx, dst, info->dst.level,
- 0xFFFFFFFF);
dst->dirty_level_mask &= ~(1 << info->dst.level);
}
templ.depth0 = 1;
templ.array_size = 1;
templ.usage = PIPE_USAGE_DEFAULT;
- templ.flags = R600_RESOURCE_FLAG_FORCE_TILING |
- R600_RESOURCE_FLAG_DISABLE_DCC;
+ templ.flags = SI_RESOURCE_FLAG_FORCE_MSAA_TILING |
+ SI_RESOURCE_FLAG_DISABLE_DCC;
/* The src and dst microtile modes must be the same. */
if (src->surface.micro_tile_mode == RADEON_MICRO_MODE_DISPLAY)
tmp = ctx->screen->resource_create(ctx->screen, &templ);
if (!tmp)
return false;
- rtmp = (struct r600_texture*)tmp;
+ stmp = (struct si_texture*)tmp;
- assert(!rtmp->surface.is_linear);
- assert(src->surface.micro_tile_mode == rtmp->surface.micro_tile_mode);
+ assert(!stmp->surface.is_linear);
+ assert(src->surface.micro_tile_mode == stmp->surface.micro_tile_mode);
/* resolve */
si_do_CB_resolve(sctx, info, tmp, 0, 0, format);
blit.src.resource = tmp;
blit.src.box.z = 0;
- si_blitter_begin(ctx, SI_BLIT |
+ si_blitter_begin(sctx, SI_BLIT |
(info->render_condition_enable ? 0 : SI_DISABLE_RENDER_COND));
util_blitter_blit(sctx->blitter, &blit);
- si_blitter_end(ctx);
+ si_blitter_end(sctx);
pipe_resource_reference(&tmp, NULL);
return true;
const struct pipe_blit_info *info)
{
struct si_context *sctx = (struct si_context*)ctx;
- struct r600_texture *rdst = (struct r600_texture *)info->dst.resource;
+ struct si_texture *dst = (struct si_texture *)info->dst.resource;
if (do_hardware_msaa_resolve(ctx, info)) {
return;
* resource_copy_region can't do this yet, because dma_copy calls it
* on failure (recursion).
*/
- if (rdst->surface.is_linear &&
- sctx->b.dma_copy &&
+ if (dst->surface.is_linear &&
+ sctx->dma_copy &&
util_can_blit_via_copy_region(info, false)) {
- sctx->b.dma_copy(ctx, info->dst.resource, info->dst.level,
+ sctx->dma_copy(ctx, info->dst.resource, info->dst.level,
info->dst.box.x, info->dst.box.y,
info->dst.box.z,
info->src.resource, info->src.level,
/* The driver doesn't decompress resources automatically while
* u_blitter is rendering. */
- vi_disable_dcc_if_incompatible_format(&sctx->b, info->src.resource,
+ vi_disable_dcc_if_incompatible_format(sctx, info->src.resource,
info->src.level,
info->src.format);
- vi_disable_dcc_if_incompatible_format(&sctx->b, info->dst.resource,
+ vi_disable_dcc_if_incompatible_format(sctx, info->dst.resource,
info->dst.level,
info->dst.format);
- si_decompress_subresource(ctx, info->src.resource, info->mask,
+ si_decompress_subresource(ctx, info->src.resource, PIPE_MASK_RGBAZS,
info->src.level,
info->src.box.z,
info->src.box.z + info->src.box.depth - 1);
- if (sctx->screen->b.debug_flags & DBG(FORCE_DMA) &&
+ if (sctx->screen->debug_flags & DBG(FORCE_DMA) &&
util_try_blit_via_copy_region(ctx, info))
return;
- si_blitter_begin(ctx, SI_BLIT |
+ si_blitter_begin(sctx, SI_BLIT |
(info->render_condition_enable ? 0 : SI_DISABLE_RENDER_COND));
util_blitter_blit(sctx->blitter, info);
- si_blitter_end(ctx);
+ si_blitter_end(sctx);
}
-static boolean si_generate_mipmap(struct pipe_context *ctx,
- struct pipe_resource *tex,
- enum pipe_format format,
- unsigned base_level, unsigned last_level,
- unsigned first_layer, unsigned last_layer)
+static bool si_generate_mipmap(struct pipe_context *ctx,
+ struct pipe_resource *tex,
+ enum pipe_format format,
+ unsigned base_level, unsigned last_level,
+ unsigned first_layer, unsigned last_layer)
{
struct si_context *sctx = (struct si_context*)ctx;
- struct r600_texture *rtex = (struct r600_texture *)tex;
+ struct si_texture *stex = (struct si_texture *)tex;
if (!util_blitter_is_copy_supported(sctx->blitter, tex, tex))
return false;
/* The driver doesn't decompress resources automatically while
* u_blitter is rendering. */
- vi_disable_dcc_if_incompatible_format(&sctx->b, tex, base_level,
+ vi_disable_dcc_if_incompatible_format(sctx, tex, base_level,
format);
si_decompress_subresource(ctx, tex, PIPE_MASK_RGBAZS,
base_level, first_layer, last_layer);
/* Clear dirty_level_mask for the levels that will be overwritten. */
assert(base_level < last_level);
- rtex->dirty_level_mask &= ~u_bit_consecutive(base_level + 1,
+ stex->dirty_level_mask &= ~u_bit_consecutive(base_level + 1,
last_level - base_level);
- sctx->generate_mipmap_for_depth = rtex->is_depth;
+ sctx->generate_mipmap_for_depth = stex->is_depth;
- si_blitter_begin(ctx, SI_BLIT | SI_DISABLE_RENDER_COND);
+ si_blitter_begin(sctx, SI_BLIT | SI_DISABLE_RENDER_COND);
util_blitter_generate_mipmap(sctx->blitter, tex, format,
base_level, last_level,
first_layer, last_layer);
- si_blitter_end(ctx);
+ si_blitter_end(sctx);
sctx->generate_mipmap_for_depth = false;
return true;
static void si_flush_resource(struct pipe_context *ctx,
struct pipe_resource *res)
{
- struct r600_texture *rtex = (struct r600_texture*)res;
+ struct si_context *sctx = (struct si_context*)ctx;
+ struct si_texture *tex = (struct si_texture*)res;
assert(res->target != PIPE_BUFFER);
- assert(!rtex->dcc_separate_buffer || rtex->dcc_gather_statistics);
+ assert(!tex->dcc_separate_buffer || tex->dcc_gather_statistics);
/* st/dri calls flush twice per frame (not a bug), this prevents double
* decompression. */
- if (rtex->dcc_separate_buffer && !rtex->separate_dcc_dirty)
+ if (tex->dcc_separate_buffer && !tex->separate_dcc_dirty)
return;
- if (!rtex->is_depth && (rtex->cmask.size || rtex->dcc_offset)) {
- si_blit_decompress_color(ctx, rtex, 0, res->last_level,
+ if (!tex->is_depth && (tex->cmask_buffer || tex->surface.dcc_offset)) {
+ si_blit_decompress_color(sctx, tex, 0, res->last_level,
0, util_max_layer(res, 0),
- rtex->dcc_separate_buffer != NULL);
+ tex->dcc_separate_buffer != NULL, false);
+
+ if (tex->surface.display_dcc_offset)
+ si_retile_dcc(sctx, tex);
}
/* Always do the analysis even if DCC is disabled at the moment. */
- if (rtex->dcc_gather_statistics && rtex->separate_dcc_dirty) {
- rtex->separate_dcc_dirty = false;
- vi_separate_dcc_process_and_reset_stats(ctx, rtex);
+ if (tex->dcc_gather_statistics) {
+ bool separate_dcc_dirty = tex->separate_dcc_dirty;
+
+ /* If the color buffer hasn't been unbound and fast clear hasn't
+ * been used, separate_dcc_dirty is false, but there may have been
+ * new rendering. Check if the color buffer is bound and assume
+ * it's dirty.
+ *
+ * Note that DRI2 never unbinds window colorbuffers, which means
+ * the DCC pipeline statistics query would never be re-set and would
+ * keep adding new results until all free memory is exhausted if we
+ * didn't do this.
+ */
+ if (!separate_dcc_dirty) {
+ for (unsigned i = 0; i < sctx->framebuffer.state.nr_cbufs; i++) {
+ if (sctx->framebuffer.state.cbufs[i] &&
+ sctx->framebuffer.state.cbufs[i]->texture == res) {
+ separate_dcc_dirty = true;
+ break;
+ }
+ }
+ }
+
+ if (separate_dcc_dirty) {
+ tex->separate_dcc_dirty = false;
+ vi_separate_dcc_process_and_reset_stats(ctx, tex);
+ }
}
}
-static void si_decompress_dcc(struct pipe_context *ctx,
- struct r600_texture *rtex)
+void si_decompress_dcc(struct si_context *sctx, struct si_texture *tex)
{
- if (!rtex->dcc_offset)
+ /* If graphics is disabled, we can't decompress DCC, but it shouldn't
+ * be compressed either. The caller should simply discard it.
+ */
+ if (!tex->surface.dcc_offset || !sctx->has_graphics)
return;
- si_blit_decompress_color(ctx, rtex, 0, rtex->resource.b.b.last_level,
- 0, util_max_layer(&rtex->resource.b.b, 0),
- true);
+ si_blit_decompress_color(sctx, tex, 0, tex->buffer.b.b.last_level,
+ 0, util_max_layer(&tex->buffer.b.b, 0),
+ true, false);
}
void si_init_blit_functions(struct si_context *sctx)
{
- 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.b.generate_mipmap = si_generate_mipmap;
- sctx->b.blit_decompress_depth = si_blit_decompress_depth;
- sctx->b.decompress_dcc = si_decompress_dcc;
+ sctx->b.resource_copy_region = si_resource_copy_region;
+
+ if (sctx->has_graphics) {
+ sctx->b.blit = si_blit;
+ sctx->b.flush_resource = si_flush_resource;
+ sctx->b.generate_mipmap = si_generate_mipmap;
+ }
}