From b3e0b0d491c55bfdc1c40d26a421910103c1b9f2 Mon Sep 17 00:00:00 2001
From: jsmall-nvidia <jsmall@nvidia.com>
Date: Tue, 28 Jan 2020 12:41:09 -0500
Subject: Synthesizing CUDA tests (#1183)

* When using setUniform clamp the amount of data written to the buffer size.

* CUDA implement StructuredBuffer/ByteAddressBuffer as pointer/count as is on CPU.
Allow bounds check to zero index.
Update docs.

* Synthesize tests.

* Fix bug in CUDA output.

* Fixing more tests to run on CUDA.

* Added BaseType for layout of Vector and Matrix - as they are held as int32_t vector array types.

* Enable unbound array support on CUDA.

* Added unsized array support for CUDA documentation.
---
 tools/render-test/cuda/cuda-compute-util.cpp | 12 +++++++-----
 tools/render-test/cuda/cuda-compute-util.h   |  5 +++++
 tools/render-test/render-test-main.cpp       |  8 ++++++++
 3 files changed, 20 insertions(+), 5 deletions(-)

(limited to 'tools/render-test')

diff --git a/tools/render-test/cuda/cuda-compute-util.cpp b/tools/render-test/cuda/cuda-compute-util.cpp
index aa82d8d70..a50295063 100644
--- a/tools/render-test/cuda/cuda-compute-util.cpp
+++ b/tools/render-test/cuda/cuda-compute-util.cpp
@@ -572,13 +572,15 @@ static SlangResult _compute(CUcontext context, CUmodule module, const ShaderComp
                         auto elementCount = int(typeLayout->getElementCount());
                         if (elementCount == 0)
                         {
-                            void** array = location.getUniform<void*>();
-                            // If set, we setup the data needed for array on CPU side
-                            if (value && array)
+                            CUDAComputeUtil::Array array = { nullptr, 0 };
+                            auto resource = CUDAResource::getCUDAResource(value);
+                            if (resource)
                             {
-                                // TODO(JS): For now we'll just assume a pointer...
-                                *array = CUDAResource::getCUDAData(value);
+                                array.data = resource->m_cudaMemory;
+                                array.count = value->m_elementCount;
                             }
+
+                            location.setUniform(&array, sizeof(array));
                         }
                         break;
                     }
diff --git a/tools/render-test/cuda/cuda-compute-util.h b/tools/render-test/cuda/cuda-compute-util.h
index ea58b6343..f739ade91 100644
--- a/tools/render-test/cuda/cuda-compute-util.h
+++ b/tools/render-test/cuda/cuda-compute-util.h
@@ -21,6 +21,11 @@ struct CUDAComputeUtil
         void* data;
         size_t count;
     };
+    struct Array
+    {
+        void* data;
+        size_t count;
+    };
 
     struct Context
     {
diff --git a/tools/render-test/render-test-main.cpp b/tools/render-test/render-test-main.cpp
index 050a6d2c8..16f2d78d1 100644
--- a/tools/render-test/render-test-main.cpp
+++ b/tools/render-test/render-test-main.cpp
@@ -600,9 +600,17 @@ SLANG_TEST_TOOL_API SlangResult innerMain(Slang::StdWriters* stdWriters, SlangSe
 
 #if RENDER_TEST_CUDA
 
+        const uint64_t startTicks = ProcessUtil::getClockTick();
+
         CUDAComputeUtil::Context context;
         SLANG_RETURN_ON_FAIL(CUDAComputeUtil::execute(compilationAndLayout, context));
 
+        if (gOptions.performanceProfile)
+        {
+            const uint64_t endTicks = ProcessUtil::getClockTick();
+            _outputProfileTime(startTicks, endTicks);
+        }
+
         if (gOptions.outputPath)
         {
             // Dump everything out that was written
-- 
cgit v1.2.3