From 94af0a00f67c9f28fcaf170458e55b7a95de76bf Mon Sep 17 00:00:00 2001
From: ameerj <52414509+ameerj@users.noreply.github.com>
Date: Mon, 12 Jul 2021 02:03:25 -0400
Subject: glsl: Clamp shared mem size to GL_MAX_COMPUTE_SHARED_MEMORY_SIZE

---
 src/shader_recompiler/backend/glsl/emit_glsl.cpp   | 11 +++++++++--
 src/shader_recompiler/profile.h                    |  2 ++
 src/video_core/renderer_opengl/gl_shader_cache.cpp |  1 +
 3 files changed, 12 insertions(+), 2 deletions(-)

(limited to 'src')

diff --git a/src/shader_recompiler/backend/glsl/emit_glsl.cpp b/src/shader_recompiler/backend/glsl/emit_glsl.cpp
index ffdc6dbba..c5e819a0a 100644
--- a/src/shader_recompiler/backend/glsl/emit_glsl.cpp
+++ b/src/shader_recompiler/backend/glsl/emit_glsl.cpp
@@ -218,8 +218,15 @@ std::string EmitGLSL(const Profile& profile, const RuntimeInfo& runtime_info, IR
     const std::string version{fmt::format("#version 450{}\n", GlslVersionSpecifier(ctx))};
     ctx.header.insert(0, version);
     if (program.shared_memory_size > 0) {
-        ctx.header +=
-            fmt::format("shared uint smem[{}];", Common::DivCeil(program.shared_memory_size, 4U));
+        const auto requested_size{program.shared_memory_size};
+        const auto max_size{profile.gl_max_compute_smem_size};
+        const bool needs_clamp{requested_size > max_size};
+        if (needs_clamp) {
+            LOG_WARNING(Shader_GLSL, "Requested shared memory size ({}) exceeds device limit ({})",
+                        requested_size, max_size);
+        }
+        const auto smem_size{needs_clamp ? max_size : requested_size};
+        ctx.header += fmt::format("shared uint smem[{}];", Common::DivCeil(smem_size, 4U));
     }
     ctx.header += "void main(){\n";
     if (program.local_memory_size > 0) {
diff --git a/src/shader_recompiler/profile.h b/src/shader_recompiler/profile.h
index 501dcaf71..f0c3b3b17 100644
--- a/src/shader_recompiler/profile.h
+++ b/src/shader_recompiler/profile.h
@@ -67,6 +67,8 @@ struct Profile {
     bool has_gl_precise_bug{};
     /// Ignores SPIR-V ordered vs unordered using GLSL semantics
     bool ignore_nan_fp_comparisons{};
+
+    u32 gl_max_compute_smem_size{};
 };
 
 } // namespace Shader
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index 24f035c37..7ecafc862 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -211,6 +211,7 @@ ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindo
           .has_gl_component_indexing_bug = device.HasComponentIndexingBug(),
           .has_gl_precise_bug = device.HasPreciseBug(),
           .ignore_nan_fp_comparisons = true,
+          .gl_max_compute_smem_size = device.GetMaxComputeSharedMemorySize(),
       },
       host_info{
           .support_float16 = false,
-- 
cgit v1.2.3