Files
test/source/blender/gpu/intern/gpu_debug.cc
Christoph Neuhauser ad4adccdeb GPU: Added image format validation for shader image bindings
The OpenGL specs require that the storage image qualifier in shaders
(e.g., "rgba32f") needs to be compatible with the format of a bound image
(see https://registry.khronos.org/OpenGL/specs/gl/glspec46.core.pdf#page=318).

We know that Blender currently does not handle this correctly in
multiple places. AMD and NVIDIA seem to silently ignore a mismatch and
just seem to use the format of the bound image. However, for the
Intel Windows drivers, this seems to lead to visual corruptions
(#141436, #141173). While a more graceful handling of a mismatch may
be nice, this is in line with the OpenGL specs.

This PR adds code for validating image formats for bindings.

Pull Request: https://projects.blender.org/blender/blender/pulls/143791
2025-08-27 15:43:09 +02:00

203 lines
5.2 KiB
C++

/* SPDX-FileCopyrightText: 2020 Blender Authors
*
* SPDX-License-Identifier: GPL-2.0-or-later */
/** \file
* \ingroup gpu
*
* Debug features of OpenGL.
*/
#include "BKE_global.hh"
#include "BLI_string.h"
#include "gpu_context_private.hh"
#include "GPU_debug.hh"
using namespace blender;
using namespace blender::gpu;
void GPU_debug_group_begin(const char *name)
{
if (!(G.debug & G_DEBUG_GPU) && !G.profile_gpu) {
return;
}
Context *ctx = Context::get();
DebugStack &stack = ctx->debug_stack;
stack.append(StringRef(name));
ctx->debug_group_begin(name, stack.size());
}
void GPU_debug_group_end()
{
if (!(G.debug & G_DEBUG_GPU) && !G.profile_gpu) {
return;
}
Context *ctx = Context::get();
ctx->debug_stack.pop_last();
ctx->debug_group_end();
}
void GPU_debug_get_groups_names(int name_buf_len, char *r_name_buf)
{
Context *ctx = Context::get();
if (ctx == nullptr) {
return;
}
DebugStack &stack = ctx->debug_stack;
if (stack.is_empty()) {
r_name_buf[0] = '\0';
return;
}
size_t len = 0;
for (StringRef &name : stack) {
len += BLI_snprintf_rlen(r_name_buf + len, name_buf_len - len, "%s > ", name.data());
}
r_name_buf[len - 3] = '\0';
}
bool GPU_debug_group_match(const char *ref)
{
/* Otherwise there will be no names. */
BLI_assert(G.debug & G_DEBUG_GPU);
Context *ctx = Context::get();
if (ctx == nullptr) {
return false;
}
const DebugStack &stack = ctx->debug_stack;
for (const StringRef &name : stack) {
if (name == ref) {
return true;
}
}
return false;
}
void GPU_debug_capture_begin(const char *title)
{
/* GPU Frame capture is only enabled when --debug-gpu is specified. */
if (!(G.debug & G_DEBUG_GPU)) {
return;
}
Context *ctx = Context::get();
if (ctx && !ctx->debug_is_capturing) {
ctx->debug_is_capturing = ctx->debug_capture_begin(title);
/* Call GPU_finish to ensure all desired GPU commands occur within the capture boundary. */
GPU_finish();
}
}
void GPU_debug_capture_end()
{
/* GPU Frame capture is only enabled when --debug-gpu is specified. */
if (!(G.debug & G_DEBUG_GPU)) {
return;
}
Context *ctx = Context::get();
if (ctx && ctx->debug_is_capturing) {
/* Call GPU_finish to ensure all desired GPU commands occur within the capture boundary. */
GPU_finish();
ctx->debug_capture_end();
ctx->debug_is_capturing = false;
}
}
void *GPU_debug_capture_scope_create(const char *name)
{
/* GPU Frame capture is only enabled when --debug-gpu is specified. */
if (!(G.debug & G_DEBUG_GPU)) {
return nullptr;
}
Context *ctx = Context::get();
if (!ctx) {
return nullptr;
}
return ctx->debug_capture_scope_create(name);
}
bool GPU_debug_capture_scope_begin(void *scope)
{
/* Early exit if scope does not exist or not in debug mode. */
if (!(G.debug & G_DEBUG_GPU) || !scope) {
return false;
}
Context *ctx = Context::get();
if (!ctx) {
return false;
}
/* Declare beginning of capture scope region. */
bool scope_capturing = ctx->debug_capture_scope_begin(scope);
if (scope_capturing && !ctx->debug_is_capturing) {
/* Call GPU_finish to ensure all desired GPU commands occur within the capture boundary. */
GPU_finish();
ctx->debug_is_capturing = true;
}
return ctx->debug_is_capturing;
}
void GPU_debug_capture_scope_end(void *scope)
{
/* Early exit if scope does not exist or not in debug mode. */
if (!(G.debug & G_DEBUG_GPU) || !scope) {
return;
}
Context *ctx = Context::get();
if (!ctx) {
return;
}
/* If capturing, call GPU_finish to ensure all desired GPU commands occur within the capture
* boundary. */
if (ctx->debug_is_capturing) {
GPU_finish();
ctx->debug_is_capturing = false;
}
/* Declare end of capture scope region. */
ctx->debug_capture_scope_end(scope);
}
namespace blender::gpu {
void debug_validate_binding_image_format()
{
if (!(G.debug & G_DEBUG_GPU)) {
return;
}
const auto &texture_formats_state = Context::get()->state_manager->image_formats;
const auto &texture_formats_shader = Context::get()->shader->interface->image_formats_;
for (int image_unit = 0; image_unit < GPU_MAX_IMAGE; image_unit++) {
TextureWriteFormat format_state = texture_formats_state[image_unit];
TextureWriteFormat format_shader = texture_formats_shader[image_unit];
if (format_state != TextureWriteFormat::Invalid &&
format_shader == TextureWriteFormat::Invalid)
{
/* It is allowed for an image to be bound in the state manager but to be unused in the
* shader. */
continue;
}
if (UNLIKELY(texture_formats_shader[image_unit] != texture_formats_state[image_unit])) {
fprintf(
stderr,
"Error in GPU_debug_validate_binding_image_format: Image format mismatch detected for "
"shader '%s' at binding %d (shader format '%s' vs. bound texture format '%s').\n",
Context::get()->shader->name_get().c_str(),
image_unit,
GPU_texture_format_name(to_texture_format(texture_formats_shader[image_unit])),
GPU_texture_format_name(to_texture_format(texture_formats_state[image_unit])));
BLI_assert_unreachable();
}
}
}
} // namespace blender::gpu