u32 GLGSRender::get_zcull_stats(u32 type)
{
if (g_cfg.video.disable_zcull_queries)
return 0u;
if (zcull_task_queue.active_query &&
zcull_task_queue.active_query->active &&
current_zcull_stats.zpass_pixel_cnt == 0 &&
type == CELL_GCM_ZPASS_PIXEL_CNT)
{
//The zcull unit is still bound as the read is happening and there are no results ready
check_zcull_status(false, true); //close current query
check_zcull_status(false, false); //start new query since stat counting is still active
}
switch (type)
{
case CELL_GCM_ZPASS_PIXEL_CNT:
{
if (current_zcull_stats.zpass_pixel_cnt > 0)
return UINT16_MAX;
synchronize_zcull_stats(true);
return (current_zcull_stats.zpass_pixel_cnt > 0)? UINT16_MAX : 0;
}
case CELL_GCM_ZCULL_STATS:
case CELL_GCM_ZCULL_STATS1:
case CELL_GCM_ZCULL_STATS2:
//TODO
return UINT16_MAX;
case CELL_GCM_ZCULL_STATS3:
{
//Some kind of inverse value
if (current_zcull_stats.zpass_pixel_cnt > 0)
return 0;
synchronize_zcull_stats(true);
return (current_zcull_stats.zpass_pixel_cnt > 0) ? 0 : UINT16_MAX;
}
default:
LOG_ERROR(RSX, "Unknown zcull stat type %d", type);
return 0;
}
}
void GLGSRender::clear_zcull_stats(u32 type)
{
if (g_cfg.video.disable_zcull_queries)
return;
if (type == CELL_GCM_ZPASS_PIXEL_CNT)
{
if (zcull_task_queue.active_query &&
zcull_task_queue.active_query->active &&
zcull_task_queue.active_query->num_draws > 0)
{
//discard active query results
check_zcull_status(false, true);
zcull_task_queue.active_query->pending = false;
//re-enable cull stats if stats are enabled
check_zcull_status(false, false);
}
current_zcull_stats.clear();
}
}
u32 GLGSRender::get_zcull_stats(u32 type)
{
if (g_cfg.video.disable_zcull_queries)
return 0u;
if (zcull_task_queue.active_query &&
zcull_task_queue.active_query->active &&
current_zcull_stats.zpass_pixel_cnt == 0)
{
//The zcull unit is still bound as the read is happening and there are no results ready
check_zcull_status(false, true);
}
switch (type)
{
case CELL_GCM_ZPASS_PIXEL_CNT:
{
if (current_zcull_stats.zpass_pixel_cnt > 0)
return UINT32_MAX;
//If we have no results, we might as well synchronize here and wait for results to become available
synchronize_zcull_stats(true);
return (current_zcull_stats.zpass_pixel_cnt > 0)? UINT32_MAX: 0;
}
case CELL_GCM_ZCULL_STATS:
case CELL_GCM_ZCULL_STATS1:
case CELL_GCM_ZCULL_STATS2:
//TODO
return UINT32_MAX;
case CELL_GCM_ZCULL_STATS3:
{
//Some kind of inverse value
if (current_zcull_stats.zpass_pixel_cnt > 0)
return 0;
synchronize_zcull_stats(true);
return (current_zcull_stats.zpass_pixel_cnt > 0) ? 0 : UINT32_MAX;
}
default:
LOG_ERROR(RSX, "Unknown zcull stat type %d", type);
return 0;
}
}
void GLGSRender::init_buffers(bool skip_reading)
{
if (draw_fbo && !m_rtts_dirty)
{
set_viewport();
return;
}
//We are about to change buffers, flush any pending requests for the old buffers
synchronize_buffers();
m_rtts_dirty = false;
zcull_surface_active = false;
const u16 clip_horizontal = rsx::method_registers.surface_clip_width();
const u16 clip_vertical = rsx::method_registers.surface_clip_height();
if (clip_horizontal == 0 || clip_vertical == 0)
{
LOG_ERROR(RSX, "Invalid framebuffer setup, w=%d, h=%d", clip_horizontal, clip_vertical);
framebuffer_status_valid = false;
return;
}
const auto pitchs = get_pitchs();
const auto surface_format = rsx::method_registers.surface_color();
const auto depth_format = rsx::method_registers.surface_depth_fmt();
const auto surface_addresses = get_color_surface_addresses();
const auto depth_address = get_zeta_surface_address();
m_rtts.prepare_render_target(nullptr, surface_format, depth_format, clip_horizontal, clip_vertical,
rsx::method_registers.surface_color_target(),
surface_addresses, depth_address);
draw_fbo.recreate();
for (int i = 0; i < rsx::limits::color_buffers_count; ++i)
{
if (std::get<0>(m_rtts.m_bound_render_targets[i]))
{
__glcheck draw_fbo.color[i] = *std::get<1>(m_rtts.m_bound_render_targets[i]);
std::get<1>(m_rtts.m_bound_render_targets[i])->set_rsx_pitch(pitchs[i]);
surface_info[i] = { surface_addresses[i], pitchs[i], false, surface_format, depth_format, clip_horizontal, clip_vertical };
//Verify pitch given is correct if pitch <= 64 (especially 64)
if (pitchs[i] <= 64)
{
const u16 native_pitch = std::get<1>(m_rtts.m_bound_render_targets[i])->get_native_pitch();
if (native_pitch > pitchs[i])
{
LOG_TRACE(RSX, "Bad color surface pitch given: surface_width=%d, format=%d, pitch=%d, native_pitch=%d",
clip_horizontal, (u32)surface_format, pitchs[i], native_pitch);
//Will not transfer this surface between cell and rsx due to misalignment
//TODO: Verify correct behaviour
surface_info[i].pitch = 0;
}
}
}
else
surface_info[i] = {};
}
if (std::get<0>(m_rtts.m_bound_depth_stencil))
{
if (depth_format == rsx::surface_depth_format::z24s8)
__glcheck draw_fbo.depth_stencil = *std::get<1>(m_rtts.m_bound_depth_stencil);
else
__glcheck draw_fbo.depth = *std::get<1>(m_rtts.m_bound_depth_stencil);
const u32 depth_surface_pitch = rsx::method_registers.surface_z_pitch();
std::get<1>(m_rtts.m_bound_depth_stencil)->set_rsx_pitch(rsx::method_registers.surface_z_pitch());
depth_surface_info = { depth_address, depth_surface_pitch, true, surface_format, depth_format, clip_horizontal, clip_vertical };
//Verify pitch given is correct if pitch <= 64 (especially 64)
if (depth_surface_pitch <= 64)
{
const u16 native_pitch = std::get<1>(m_rtts.m_bound_depth_stencil)->get_native_pitch();
if (native_pitch > depth_surface_pitch)
{
LOG_TRACE(RSX, "Bad depth surface pitch given: surface_width=%d, format=%d, pitch=%d, native_pitch=%d",
clip_horizontal, (u32)depth_format, depth_surface_pitch, native_pitch);
//Will not transfer this surface between cell and rsx due to misalignment
//TODO: Verify correct behaviour
depth_surface_info.pitch = 0;
}
}
}
else
depth_surface_info = {};
framebuffer_status_valid = draw_fbo.check();
if (!framebuffer_status_valid) return;
check_zcull_status(true, false);
draw_fbo.bind();
set_viewport();
switch (rsx::method_registers.surface_color_target())
{
case rsx::surface_target::none: break;
case rsx::surface_target::surface_a:
__glcheck draw_fbo.draw_buffer(draw_fbo.color[0]);
__glcheck draw_fbo.read_buffer(draw_fbo.color[0]);
break;
case rsx::surface_target::surface_b:
__glcheck draw_fbo.draw_buffer(draw_fbo.color[1]);
__glcheck draw_fbo.read_buffer(draw_fbo.color[1]);
break;
case rsx::surface_target::surfaces_a_b:
__glcheck draw_fbo.draw_buffers({ draw_fbo.color[0], draw_fbo.color[1] });
__glcheck draw_fbo.read_buffer(draw_fbo.color[0]);
break;
case rsx::surface_target::surfaces_a_b_c:
__glcheck draw_fbo.draw_buffers({ draw_fbo.color[0], draw_fbo.color[1], draw_fbo.color[2] });
__glcheck draw_fbo.read_buffer(draw_fbo.color[0]);
break;
case rsx::surface_target::surfaces_a_b_c_d:
__glcheck draw_fbo.draw_buffers({ draw_fbo.color[0], draw_fbo.color[1], draw_fbo.color[2], draw_fbo.color[3] });
__glcheck draw_fbo.read_buffer(draw_fbo.color[0]);
break;
}
//Mark buffer regions as NO_ACCESS on Cell visible side
if (g_cfg.video.write_color_buffers)
{
auto color_format = rsx::internals::surface_color_format_to_gl(surface_format);
for (u8 i = 0; i < rsx::limits::color_buffers_count; ++i)
{
if (!surface_info[i].address || !surface_info[i].pitch) continue;
const u32 range = surface_info[i].pitch * surface_info[i].height;
m_gl_texture_cache.lock_rtt_region(surface_info[i].address, range, surface_info[i].width, surface_info[i].height, surface_info[i].pitch,
color_format.format, color_format.type, color_format.swap_bytes, *std::get<1>(m_rtts.m_bound_render_targets[i]));
}
}
if (g_cfg.video.write_depth_buffer)
{
if (depth_surface_info.address && depth_surface_info.pitch)
{
auto depth_format_gl = rsx::internals::surface_depth_format_to_gl(depth_format);
u32 pitch = depth_surface_info.width * 2;
if (depth_surface_info.depth_format != rsx::surface_depth_format::z16) pitch *= 2;
const u32 range = pitch * depth_surface_info.height;
//TODO: Verify that depth surface pitch variance affects results
if (pitch != depth_surface_info.pitch)
LOG_WARNING(RSX, "Depth surface pitch does not match computed pitch, %d vs %d", depth_surface_info.pitch, pitch);
m_gl_texture_cache.lock_rtt_region(depth_surface_info.address, range, depth_surface_info.width, depth_surface_info.height, pitch,
depth_format_gl.format, depth_format_gl.type, true, *std::get<1>(m_rtts.m_bound_depth_stencil));
}
}
}
void GLGSRender::notify_zcull_info_changed()
{
check_zcull_status(false, false);
}
