1 files changed, 88 insertions, 48 deletions
diff --git a/src/video_core/renderer_opengl/gl_device.cpp b/src/video_core/renderer_opengl/gl_device.cpp
index b30d5be74..413d8546b 100644
--- a/src/video_core/renderer_opengl/gl_device.cpp
+++ b/src/video_core/renderer_opengl/gl_device.cpp
@@ -5,7 +5,9 @@
 #include <algorithm>
 #include <array>
 #include <cstddef>
+#include <optional>
 #include <vector>
+
 #include <glad/glad.h>
 
 #include "common/logging/log.h"
@@ -17,6 +19,30 @@ namespace OpenGL {
 
 namespace {
 
+// One uniform block is reserved for emulation purposes
+constexpr u32 ReservedUniformBlocks = 1;
+
+constexpr u32 NumStages = 5;
+
+constexpr std::array LimitUBOs = {GL_MAX_VERTEX_UNIFORM_BLOCKS, GL_MAX_TESS_CONTROL_UNIFORM_BLOCKS,
+                                  GL_MAX_TESS_EVALUATION_UNIFORM_BLOCKS,
+                                  GL_MAX_GEOMETRY_UNIFORM_BLOCKS, GL_MAX_FRAGMENT_UNIFORM_BLOCKS};
+
+constexpr std::array LimitSSBOs = {
+    GL_MAX_VERTEX_SHADER_STORAGE_BLOCKS, GL_MAX_TESS_CONTROL_SHADER_STORAGE_BLOCKS,
+    GL_MAX_TESS_EVALUATION_SHADER_STORAGE_BLOCKS, GL_MAX_GEOMETRY_SHADER_STORAGE_BLOCKS,
+    GL_MAX_FRAGMENT_SHADER_STORAGE_BLOCKS};
+
+constexpr std::array LimitSamplers = {
+    GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS, GL_MAX_TESS_CONTROL_TEXTURE_IMAGE_UNITS,
+    GL_MAX_TESS_EVALUATION_TEXTURE_IMAGE_UNITS, GL_MAX_GEOMETRY_TEXTURE_IMAGE_UNITS,
+    GL_MAX_TEXTURE_IMAGE_UNITS};
+
+constexpr std::array LimitImages = {GL_MAX_VERTEX_IMAGE_UNIFORMS,
+                                    GL_MAX_TESS_CONTROL_IMAGE_UNIFORMS,
+                                    GL_MAX_TESS_EVALUATION_IMAGE_UNIFORMS,
+                                    GL_MAX_GEOMETRY_IMAGE_UNIFORMS, GL_MAX_FRAGMENT_IMAGE_UNIFORMS};
+
 template <typename T>
 T GetInteger(GLenum pname) {
     GLint temporary;
@@ -48,13 +74,71 @@ bool HasExtension(const std::vector<std::string_view>& images, std::string_view
     return std::find(images.begin(), images.end(), extension) != images.end();
 }
 
+u32 Extract(u32& base, u32& num, u32 amount, std::optional<GLenum> limit = {}) {
+    ASSERT(num >= amount);
+    if (limit) {
+        amount = std::min(amount, GetInteger<u32>(*limit));
+    }
+    num -= amount;
+    return std::exchange(base, base + amount);
+}
+
+std::array<Device::BaseBindings, Tegra::Engines::MaxShaderTypes> BuildBaseBindings() noexcept {
+    std::array<Device::BaseBindings, Tegra::Engines::MaxShaderTypes> bindings;
+
+    static std::array<std::size_t, 5> stage_swizzle = {0, 1, 2, 3, 4};
+    const u32 total_ubos = GetInteger<u32>(GL_MAX_UNIFORM_BUFFER_BINDINGS);
+    const u32 total_ssbos = GetInteger<u32>(GL_MAX_SHADER_STORAGE_BUFFER_BINDINGS);
+    const u32 total_samplers = GetInteger<u32>(GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS);
+
+    u32 num_ubos = total_ubos - ReservedUniformBlocks;
+    u32 num_ssbos = total_ssbos;
+    u32 num_samplers = total_samplers;
+
+    u32 base_ubo = ReservedUniformBlocks;
+    u32 base_ssbo = 0;
+    u32 base_samplers = 0;
+
+    for (std::size_t i = 0; i < NumStages; ++i) {
+        const std::size_t stage = stage_swizzle[i];
+        bindings[stage] = {
+            Extract(base_ubo, num_ubos, total_ubos / NumStages, LimitUBOs[stage]),
+            Extract(base_ssbo, num_ssbos, total_ssbos / NumStages, LimitSSBOs[stage]),
+            Extract(base_samplers, num_samplers, total_samplers / NumStages, LimitSamplers[stage])};
+    }
+
+    u32 num_images = GetInteger<u32>(GL_MAX_IMAGE_UNITS);
+    u32 base_images = 0;
+
+    // Reserve more image bindings on fragment and vertex stages.
+    bindings[4].image =
+        Extract(base_images, num_images, num_images / NumStages + 2, LimitImages[4]);
+    bindings[0].image =
+        Extract(base_images, num_images, num_images / NumStages + 1, LimitImages[0]);
+
+    // Reserve the other image bindings.
+    const u32 total_extracted_images = num_images / (NumStages - 2);
+    for (std::size_t i = 2; i < NumStages; ++i) {
+        const std::size_t stage = stage_swizzle[i];
+        bindings[stage].image =
+            Extract(base_images, num_images, total_extracted_images, LimitImages[stage]);
+    }
+
+    // Compute doesn't care about any of this.
+    bindings[5] = {0, 0, 0, 0};
+
+    return bindings;
+}
+
 } // Anonymous namespace
 
-Device::Device() {
+Device::Device() : base_bindings{BuildBaseBindings()} {
     const std::string_view vendor = reinterpret_cast<const char*>(glGetString(GL_VENDOR));
     const std::vector extensions = GetExtensions();
 
     const bool is_nvidia = vendor == "NVIDIA Corporation";
+    const bool is_amd = vendor == "ATI Technologies Inc.";
+    const bool is_intel = vendor == "Intel";
 
     uniform_buffer_alignment = GetInteger<std::size_t>(GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT);
     shader_storage_alignment = GetInteger<std::size_t>(GL_SHADER_STORAGE_BUFFER_OFFSET_ALIGNMENT);
@@ -66,8 +150,9 @@ Device::Device() {
     has_vertex_viewport_layer = GLAD_GL_ARB_shader_viewport_layer_array;
     has_image_load_formatted = HasExtension(extensions, "GL_EXT_shader_image_load_formatted");
     has_variable_aoffi = TestVariableAoffi();
-    has_component_indexing_bug = TestComponentIndexingBug();
+    has_component_indexing_bug = is_amd;
     has_precise_bug = TestPreciseBug();
+    has_broken_compute = is_intel;
     has_fast_buffer_sub_data = is_nvidia;
 
     LOG_INFO(Render_OpenGL, "Renderer_VariableAOFFI: {}", has_variable_aoffi);
@@ -85,6 +170,7 @@ Device::Device(std::nullptr_t) {
     has_image_load_formatted = true;
     has_variable_aoffi = true;
     has_component_indexing_bug = false;
+    has_broken_compute = false;
     has_precise_bug = false;
 }
 
@@ -99,52 +185,6 @@ void main() {
 })");
 }
 
-bool Device::TestComponentIndexingBug() {
-    const GLchar* COMPONENT_TEST = R"(#version 430 core
-layout (std430, binding = 0) buffer OutputBuffer {
-    uint output_value;
-};
-layout (std140, binding = 0) uniform InputBuffer {
-    uvec4 input_value[4096];
-};
-layout (location = 0) uniform uint idx;
-void main() {
-    output_value = input_value[idx >> 2][idx & 3];
-})";
-    const GLuint shader{glCreateShaderProgramv(GL_VERTEX_SHADER, 1, &COMPONENT_TEST)};
-    SCOPE_EXIT({ glDeleteProgram(shader); });
-    glUseProgram(shader);
-
-    OGLVertexArray vao;
-    vao.Create();
-    glBindVertexArray(vao.handle);
-
-    constexpr std::array<GLuint, 8> values{0, 0, 0, 0, 0x1236327, 0x985482, 0x872753, 0x2378432};
-    OGLBuffer ubo;
-    ubo.Create();
-    glNamedBufferData(ubo.handle, sizeof(values), values.data(), GL_STATIC_DRAW);
-    glBindBufferBase(GL_UNIFORM_BUFFER, 0, ubo.handle);
-
-    OGLBuffer ssbo;
-    ssbo.Create();
-    glNamedBufferStorage(ssbo.handle, sizeof(GLuint), nullptr, GL_CLIENT_STORAGE_BIT);
-
-    for (GLuint index = 4; index < 8; ++index) {
-        glInvalidateBufferData(ssbo.handle);
-        glBindBufferBase(GL_SHADER_STORAGE_BUFFER, 0, ssbo.handle);
-
-        glProgramUniform1ui(shader, 0, index);
-        glDrawArrays(GL_POINTS, 0, 1);
-
-        GLuint result;
-        glGetNamedBufferSubData(ssbo.handle, 0, sizeof(result), &result);
-        if (result != values.at(index)) {
-            return true;
-        }
-    }
-    return false;
-}
-
 bool Device::TestPreciseBug() {
     return !TestProgram(R"(#version 430 core
 in vec3 coords;