shader: compute texture handles read from storage buffers for each draw

Switch 2 shaders can read a texture handle from a storage buffer (Switch
Sports' scenery: a per-object entry of a table whose pointer is in the driver
constant buffer). The bindless handle search looked through the load into its
address and returned the table pointer's constant buffer word as the handle,
which named an arbitrary descriptor; the game repoints that one at transient
render targets every few seconds, so the island and menu lighting flipped
between correct, dark and red.

The search no longer looks into memory loads. A handle it cannot trace is now
built into a small program over constant buffer words, storage buffer words and
integer operations (Shader::Info::handle_programs), which the pipelines evaluate
when they bind the texture. Handles a single evaluation cannot give (one that
changes in a loop) keep the previous search.

6 files changed, 98 insertions, 0 deletions

diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index 5ed43fc903..b3ef88c5f1 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -235,6 +235,7 @@ add_library(video_core STATIC
     shader_cache.h
     shader_environment.cpp
     shader_environment.h
+    shader_handle_program.h
     shader_notify.cpp
     shader_notify.h
     smaa_area_tex.h
diff --git a/src/video_core/renderer_opengl/gl_compute_pipeline.cpp b/src/video_core/renderer_opengl/gl_compute_pipeline.cpp
index f0a7baf9aa..eb134e86b5 100644
--- a/src/video_core/renderer_opengl/gl_compute_pipeline.cpp
+++ b/src/video_core/renderer_opengl/gl_compute_pipeline.cpp
@@ -12,6 +12,7 @@
 #include "video_core/renderer_opengl/gl_compute_pipeline.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
 #include "video_core/renderer_opengl/gl_shader_util.h"
+#include "video_core/shader_handle_program.h"
 
 namespace OpenGL {
 
@@ -105,6 +106,20 @@ void ComputePipeline::Configure() {
     const auto& cbufs{qmd.const_buffer_config};
     const bool via_header_index{qmd.linked_tsc != 0};
     const auto read_handle{[&](const auto& desc, u32 index) {
+        if constexpr (std::is_same_v<decltype(desc), const Shader::TextureDescriptor&>) {
+            if (desc.handle_program != ~0U) {
+                // Read by the shader from a storage buffer: computed for this dispatch.
+                return TexturePair(VideoCommon::EvaluateTextureHandle(
+                                       info, desc.handle_program, *gpu_memory,
+                                       [&](u32 cbuf) -> GPUVAddr {
+                                           return cbuf < cbufs.size() &&
+                                                          ((qmd.const_buffer_enable_mask >> cbuf) & 1) != 0
+                                                      ? cbufs[cbuf].Address()
+                                                      : GPUVAddr{};
+                                       }),
+                                   via_header_index);
+            }
+        }
         ASSERT(((qmd.const_buffer_enable_mask >> desc.cbuf_index) & 1) != 0);
         const u32 index_offset{index << desc.size_shift};
         const u32 offset{desc.cbuf_offset + index_offset};
diff --git a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp
index ecd6544b10..81950279d5 100644
--- a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp
+++ b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp
@@ -17,6 +17,7 @@
 #include "video_core/renderer_opengl/gl_shader_manager.h"
 #include "video_core/renderer_opengl/gl_shader_util.h"
 #include "video_core/renderer_opengl/gl_state_tracker.h"
+#include "video_core/shader_handle_program.h"
 #include "video_core/shader_notify.h"
 #include "video_core/texture_cache/texture_cache.h"
 
@@ -308,6 +309,19 @@ bool GraphicsPipeline::ConfigureImpl(bool is_indexed) {
         }
         const auto& cbufs{maxwell3d->state.shader_stages[stage].const_buffers};
         const auto read_handle{[&](const auto& desc, u32 index) {
+            if constexpr (std::is_same_v<decltype(desc), const Shader::TextureDescriptor&>) {
+                if (desc.handle_program != ~0U) {
+                    // Read by the shader from a storage buffer: computed for this draw.
+                    return TexturePair(VideoCommon::EvaluateTextureHandle(
+                                           info, desc.handle_program, *gpu_memory,
+                                           [&](u32 cbuf) -> GPUVAddr {
+                                               return cbuf < cbufs.size() && cbufs[cbuf].enabled
+                                                          ? cbufs[cbuf].address
+                                                          : GPUVAddr{};
+                                           }),
+                                       via_header_index);
+                }
+            }
             ASSERT(cbufs[desc.cbuf_index].enabled);
             const u32 index_offset{index << desc.size_shift};
             const u32 offset{desc.cbuf_offset + index_offset};
diff --git a/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp b/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp
index 29356cfcec..ee3d94899e 100644
--- a/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp
+++ b/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp
@@ -18,6 +18,7 @@
 #include "video_core/renderer_vulkan/vk_pipeline_cache.h"
 #include "video_core/renderer_vulkan/vk_scheduler.h"
 #include "video_core/renderer_vulkan/vk_update_descriptor.h"
+#include "video_core/shader_handle_program.h"
 #include "video_core/shader_notify.h"
 #include "video_core/vulkan_common/vulkan_device.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
@@ -172,6 +173,20 @@ bool ComputePipeline::Configure(Tegra::Engines::KeplerCompute& kepler_compute,
     const auto& cbufs{qmd.const_buffer_config};
     const bool via_header_index{qmd.linked_tsc != 0};
     const auto read_handle{[&](const auto& desc, u32 index) {
+        if constexpr (std::is_same_v<decltype(desc), const Shader::TextureDescriptor&>) {
+            if (desc.handle_program != ~0U) {
+                // Read by the shader from a storage buffer: computed for this dispatch.
+                return TexturePair(VideoCommon::EvaluateTextureHandle(
+                                       info, desc.handle_program, gpu_memory,
+                                       [&](u32 cbuf) -> GPUVAddr {
+                                           return cbuf < cbufs.size() &&
+                                                          ((qmd.const_buffer_enable_mask >> cbuf) & 1) != 0
+                                                      ? cbufs[cbuf].Address()
+                                                      : GPUVAddr{};
+                                       }),
+                                   via_header_index);
+            }
+        }
         ASSERT(((qmd.const_buffer_enable_mask >> desc.cbuf_index) & 1) != 0);
         const u32 index_offset{index << desc.size_shift};
         const u32 offset{desc.cbuf_offset + index_offset};
diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
index 94bace432c..3282e85484 100644
--- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
+++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
@@ -24,6 +24,7 @@
 #include "video_core/renderer_vulkan/vk_scheduler.h"
 #include "video_core/renderer_vulkan/vk_texture_cache.h"
 #include "video_core/renderer_vulkan/vk_update_descriptor.h"
+#include "video_core/shader_handle_program.h"
 #include "video_core/shader_notify.h"
 #include "video_core/texture_cache/texture_cache.h"
 #include "video_core/vulkan_common/vulkan_device.h"
@@ -382,6 +383,19 @@ bool GraphicsPipeline::ConfigureImpl(bool is_indexed) {
         }
         const auto& cbufs{maxwell3d->state.shader_stages[stage].const_buffers};
         const auto read_handle{[&](const auto& desc, u32 index) {
+            if constexpr (std::is_same_v<decltype(desc), const Shader::TextureDescriptor&>) {
+                if (desc.handle_program != ~0U) {
+                    // Read by the shader from a storage buffer: computed for this draw.
+                    return TexturePair(VideoCommon::EvaluateTextureHandle(
+                                           info, desc.handle_program, *gpu_memory,
+                                           [&](u32 cbuf) -> GPUVAddr {
+                                               return cbuf < cbufs.size() && cbufs[cbuf].enabled
+                                                          ? cbufs[cbuf].address
+                                                          : GPUVAddr{};
+                                           }),
+                                       via_header_index);
+                }
+            }
             ASSERT(cbufs[desc.cbuf_index].enabled);
             const u32 index_offset{index << desc.size_shift};
             const u32 offset{desc.cbuf_offset + index_offset};
diff --git a/src/video_core/shader_handle_program.h b/src/video_core/shader_handle_program.h
new file mode 100644
index 0000000000..3d35524cc8
--- /dev/null
+++ b/src/video_core/shader_handle_program.h
@@ -0,0 +1,39 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#pragma once
+
+#include "common/common_types.h"
+#include "shader_recompiler/shader_info.h"
+#include "video_core/memory_manager.h"
+
+namespace VideoCommon {
+
+/// The texture handle a shader reads from a storage buffer, computed from the current constant
+/// buffer and storage buffer words (Shader::Info::handle_programs). cbuf_address(index) is the
+/// bound constant buffer's address, 0 when none is bound.
+template <typename CbufAddress>
+u32 EvaluateTextureHandle(const Shader::Info& info, u32 program, Tegra::MemoryManager& gpu_memory,
+                          CbufAddress&& cbuf_address) {
+    return Shader::EvaluateHandleProgram(
+               info.handle_programs[program],
+               [&](u32 index, u32 offset) -> u32 {
+                   const GPUVAddr base{cbuf_address(index)};
+                   return base != 0 ? gpu_memory.Read<u32>(base + offset) : 0U;
+               },
+               [&](u32 binding, u32 offset) -> u32 {
+                   if (binding >= info.storage_buffers_descriptors.size()) {
+                       return 0U;
+                   }
+                   const auto& descriptor{info.storage_buffers_descriptors[binding]};
+                   const GPUVAddr cbuf{cbuf_address(descriptor.cbuf_index)};
+                   if (cbuf == 0) {
+                       return 0U;
+                   }
+                   const GPUVAddr base{gpu_memory.Read<u64>(cbuf + descriptor.cbuf_offset)};
+                   return base != 0 ? gpu_memory.Read<u32>(base + offset) : 0U;
+               })
+        .value_or(0U);
+}
+
+} // namespace VideoCommon