diff --git a/src/video_core/renderer_opengl/gl_arb_decompiler.cpp b/src/video_core/renderer_opengl/gl_arb_decompiler.cpp
index eb5158407d..31dd08fd22 100644
--- a/src/video_core/renderer_opengl/gl_arb_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_arb_decompiler.cpp
@@ -892,11 +892,19 @@ void ARBDecompiler::DeclareCompute() {
const ComputeInfo& info = registry.GetComputeInfo();
AddLine("GROUP_SIZE {} {} {};", info.workgroup_size[0], info.workgroup_size[1],
info.workgroup_size[2]);
- if (info.shared_memory_size_in_words > 0) {
- const u32 size_in_bytes = info.shared_memory_size_in_words * 4;
- AddLine("SHARED_MEMORY {};", size_in_bytes);
- AddLine("SHARED shared_mem[] = {{program.sharedmem}};");
+ if (info.shared_memory_size_in_words == 0) {
+ return;
}
+ const u32 limit = device.GetMaxComputeSharedMemorySize();
+ u32 size_in_bytes = info.shared_memory_size_in_words * 4;
+ if (size_in_bytes > limit) {
+ LOG_ERROR(Render_OpenGL, "Shared memory size {} is clamped to host's limit {}",
+ size_in_bytes, limit);
+ size_in_bytes = limit;
+ }
+
+ AddLine("SHARED_MEMORY {};", size_in_bytes);
+ AddLine("SHARED shared_mem[] = {{program.sharedmem}};");
}
void ARBDecompiler::DeclareInputAttributes() {
diff --git a/src/video_core/renderer_opengl/gl_device.cpp b/src/video_core/renderer_opengl/gl_device.cpp
index c1f20f0aba..b97c9da15d 100644
--- a/src/video_core/renderer_opengl/gl_device.cpp
+++ b/src/video_core/renderer_opengl/gl_device.cpp
@@ -212,6 +212,7 @@ Device::Device()
shader_storage_alignment = GetInteger<std::size_t>(GL_SHADER_STORAGE_BUFFER_OFFSET_ALIGNMENT);
max_vertex_attributes = GetInteger<u32>(GL_MAX_VERTEX_ATTRIBS);
max_varyings = GetInteger<u32>(GL_MAX_VARYING_VECTORS);
+ max_compute_shared_memory_size = GetInteger<u32>(GL_MAX_COMPUTE_SHARED_MEMORY_SIZE);
has_warp_intrinsics = GLAD_GL_NV_gpu_shader5 && GLAD_GL_NV_shader_thread_group &&
GLAD_GL_NV_shader_thread_shuffle;
has_shader_ballot = GLAD_GL_ARB_shader_ballot;
@@ -248,6 +249,7 @@ Device::Device(std::nullptr_t) {
shader_storage_alignment = 4;
max_vertex_attributes = 16;
max_varyings = 15;
+ max_compute_shared_memory_size = 0x10000;
has_warp_intrinsics = true;
has_shader_ballot = true;
has_vertex_viewport_layer = true;
diff --git a/src/video_core/renderer_opengl/gl_device.h b/src/video_core/renderer_opengl/gl_device.h
index e1d8119666..e713449783 100644
--- a/src/video_core/renderer_opengl/gl_device.h
+++ b/src/video_core/renderer_opengl/gl_device.h
@@ -52,6 +52,10 @@ public:
return max_varyings;
}
+ u32 GetMaxComputeSharedMemorySize() const {
+ return max_compute_shared_memory_size;
+ }
+
bool HasWarpIntrinsics() const {
return has_warp_intrinsics;
}
@@ -114,6 +118,7 @@ private:
std::size_t shader_storage_alignment{};
u32 max_vertex_attributes{};
u32 max_varyings{};
+ u32 max_compute_shared_memory_size{};
bool has_warp_intrinsics{};
bool has_shader_ballot{};
bool has_vertex_viewport_layer{};
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 2c49aeaacc..6a9602ff80 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -602,8 +602,15 @@ private:
return;
}
const auto& info = registry.GetComputeInfo();
- if (const u32 size = info.shared_memory_size_in_words; size > 0) {
- code.AddLine("shared uint smem[{}];", size);
+ if (u32 size = info.shared_memory_size_in_words * 4; size > 0) {
+ const u32 limit = device.GetMaxComputeSharedMemorySize();
+ if (size > limit) {
+ LOG_ERROR(Render_OpenGL, "Shared memory size {} is clamped to host's limit {}",
+ size, limit);
+ size = limit;
+ }
+
+ code.AddLine("shared uint smem[{}];", size / 4);
code.AddNewLine();
}
code.AddLine("layout (local_size_x = {}, local_size_y = {}, local_size_z = {}) in;",
diff --git a/src/video_core/renderer_vulkan/vk_device.h b/src/video_core/renderer_vulkan/vk_device.h
index ae5c21baa1..529744f2d1 100644
--- a/src/video_core/renderer_vulkan/vk_device.h
+++ b/src/video_core/renderer_vulkan/vk_device.h
@@ -122,6 +122,11 @@ public:
return properties.limits.maxPushConstantsSize;
}
+ /// Returns the maximum size for shared memory.
+ u32 GetMaxComputeSharedMemorySize() const {
+ return properties.limits.maxComputeSharedMemorySize;
+ }
+
/// Returns true if ASTC is natively supported.
bool IsOptimalAstcSupported() const {
return is_optimal_astc_supported;
diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
index 97429cc59f..694452fd83 100644
--- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
+++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
@@ -685,13 +685,19 @@ private:
}
t_smem_uint = TypePointer(spv::StorageClass::Workgroup, t_uint);
- const u32 smem_size = specialization.shared_memory_size;
+ u32 smem_size = specialization.shared_memory_size * 4;
if (smem_size == 0) {
// Avoid declaring an empty array.
return;
}
- const auto element_count = static_cast<u32>(Common::AlignUp(smem_size, 4) / 4);
- const Id type_array = TypeArray(t_uint, Constant(t_uint, element_count));
+ const u32 limit = device.GetMaxComputeSharedMemorySize();
+ if (smem_size > limit) {
+ LOG_ERROR(Render_Vulkan, "Shared memory size {} is clamped to host's limit {}",
+ smem_size, limit);
+ smem_size = limit;
+ }
+
+ const Id type_array = TypeArray(t_uint, Constant(t_uint, smem_size / 4));
const Id type_pointer = TypePointer(spv::StorageClass::Workgroup, type_array);
Name(type_pointer, "SharedMemory");