Merge branch 'master' of https://github.com/GPUOpen-LibrariesAndSDKs/RadeonRays_SDK into HEAD

Author: yozhijk

CLW/CLW.lua

@@ -14,6 +14,10 @@ project "CLW"
 
     configuration {}
 
+	if _OPTIONS["allow_cpu_devices"] then
+        defines {"RR_ALLOW_CPU_DEVICES=1"}
+	end
+	
     -- we rely on RadeonRays to do the actual embedding for us
     if _OPTIONS["embed_kernels"] then
         defines {"RR_EMBED_KERNELS=1"}

CLW/CLWParallelPrimitives.cpp

@@ -30,7 +30,7 @@ THE SOFTWARE.
 
 #ifdef RR_EMBED_KERNELS
 #if USE_OPENCL
-#    include <CLW/kernelcache/clwkernels_cl.h>
+#include "CLW/kernelcache/clwkernels_cl.h"
 #endif
 #endif // RR_EMBED_KERNELS
 
@@ -952,4 +952,4 @@ CLWEvent CLWParallelPrimitives::Copy(unsigned int deviceIdx, CLWBuffer<cl_int> i
     copyKernel.SetArg(2, output);
 
     return context_.Launch1D(0, NUM_BLOCKS * WG_SIZE, WG_SIZE, copyKernel);
-}
+}

CLW/CLWPlatform.cpp

@@ -48,7 +48,12 @@ void CLWPlatform::CreateAllPlatforms(std::vector<CLWPlatform>& platforms)
     status = clGetPlatformIDs(numPlatforms, &platformIds[0], nullptr);
     ThrowIf(status != CL_SUCCESS, status, "clGetPlatformIDs failed");
 
-    cl_device_type type = CL_DEVICE_TYPE_ALL;
+
+#ifdef RR_ALLOW_CPU_DEVICES
+	cl_device_type type = CL_DEVICE_TYPE_ALL;
+#else
+	cl_device_type type = CL_DEVICE_TYPE_GPU;
+#endif
 
     // TODO: this is a workaround for nasty Apple's OpenCL runtime
     // which doesn't allow to have work group sizes > 1 on CPU devices
@@ -70,8 +75,8 @@ void CLWPlatform::CreateAllPlatforms(std::vector<CLWPlatform>& platforms)
 
         std::string versionstr(version.begin(), version.end());
 
-        if (versionstr.find("1.0") != std::string::npos ||
-            versionstr.find("1.1") != std::string::npos)
+        if (versionstr.find("OpenCL 1.0 ") != std::string::npos ||
+            versionstr.find("OpenCL 1.1") != std::string::npos)
         {
             continue;
         }

Calc/src/device_clw.cpp

@@ -459,9 +459,6 @@ namespace Calc
                 ""
 #endif
                 );
-#ifdef USE_SAFE_MATH
-            buildopts.append("-D USE_SAFE_MATH");
-#endif
 
             return new ExecutableClw(
                                      CLWProgram::CreateFromFile(filename, headernames, numheaders, buildopts.c_str(), m_context)

README.md

@@ -144,6 +144,9 @@ The app only supports loading of pure triangle .obj meshes. The list of supporte
 - dds (limited support)
 - tga
 
+## Run unit tests
+They need to be run from the <Radeon Rays_SDK path>/UnitTest path.
+Premake should be runned with the `--safe_math` option.
 
 # Hardware  support
 

RadeonRays/RadeonRays.lua

@@ -11,13 +11,17 @@ project "RadeonRays"
 
     if _OPTIONS["shared_calc"] then
         defines {"CALC_IMPORT_API"};
-	links {"dl"}
+        if os.is("windows") then
+            characterset ("MBCS")
+        else
+            links {"dl"}
+        end
     else
-	defines {"CALC_STATIC_LIBRARY"}
+    defines {"CALC_STATIC_LIBRARY"}
         links {"Calc"}
-	if _OPTIONS["use_opencl"] then
+    if _OPTIONS["use_opencl"] then
            links {"CLW"}
-	end
+    end
     end
 
     if _OPTIONS["enable_raymask"] then
@@ -108,8 +112,13 @@ project "RadeonRays"
     end
 
     if _OPTIONS["enable_raymask"] then
-       	configuration {}
-	defines {"RR_RAY_MASK"}
+        configuration {}
+        defines {"RR_RAY_MASK"}
+    end
+
+    if _OPTIONS["safe_math"] then
+        configuration {}
+        defines { "USE_SAFE_MATH" }
     end
 
     if _OPTIONS["use_vulkan"] then

RadeonRays/include/math/matrix.h

@@ -179,14 +179,13 @@ namespace RadeonRays
         return res*=c;
     }
 
-    inline float3 operator * (matrix const& m, float3 const& v)
+    inline float4 operator * (matrix const& m, float4 const& v)
     {
-        float3 res;
+        float4 res(0, 0, 0, 0);
 
-        for (int i=0;i<3;++i)
-        {
-            res[i] = 0.f;
-            for (int j=0;j<3;++j)
+        for (int i = 0; i < 4; ++i) {
+            //res[i] = 0.f;
+            for (int j = 0; j < 4; ++j)
                 res[i] += m.m[i][j] * v[j];
         }
 

RadeonRays/include/radeon_rays.h

@@ -28,6 +28,7 @@ THE SOFTWARE.
 #include "math/matrix.h"
 #include "math/ray.h"
 #include "math/mathutils.h"
+#include <cstdint>
 
 #define RADEONRAYS_API_VERSION 2.0
 

RadeonRays/src/device/embree_intersection_device.cpp

@@ -116,7 +116,7 @@ namespace RadeonRays
         if (result != RTC_NO_ERROR)
             std::cout << "Failed to create embree rtcDevice: " << result << std::endl;
 
-        m_scene = rtcDeviceNewScene(m_device, RTC_SCENE_STATIC, RTC_INTERSECT1 | RTC_INTERSECT4 | RTC_INTERSECT8 | RTC_INTERSECT16 | RTC_INTERSECTN);
+        m_scene = rtcDeviceNewScene(m_device, RTC_SCENE_STATIC, RTC_INTERSECT1 | RTC_INTERSECT4 | RTC_INTERSECT8 | RTC_INTERSECT16 );
         result = rtcDeviceGetError(m_device);
         if (result != RTC_NO_ERROR)
             std::cout << "Failed to create embree scene: " << result << std::endl;
@@ -171,7 +171,7 @@ namespace RadeonRays
         }
         m_instances.clear();
         rtcDeleteScene(m_scene); CheckEmbreeError();
-        m_scene = rtcDeviceNewScene(m_device, RTC_SCENE_STATIC, RTC_INTERSECT1 | RTC_INTERSECT4 | RTC_INTERSECT8 | RTC_INTERSECT16 | RTC_INTERSECTN); CheckEmbreeError();
+        m_scene = rtcDeviceNewScene(m_device, RTC_SCENE_STATIC, RTC_INTERSECT1 | RTC_INTERSECT4 | RTC_INTERSECT8 | RTC_INTERSECT16 ); CheckEmbreeError();
 
         for (auto i : world.shapes_)
         {
@@ -500,7 +500,7 @@ namespace RadeonRays
     {
         if (m_meshes.count(mesh))
             return m_meshes[mesh].scene;
-        RTCScene result = rtcDeviceNewScene(m_device, RTC_SCENE_STATIC, RTC_INTERSECT1 | RTC_INTERSECT4 | RTC_INTERSECT8 | RTC_INTERSECT16 | RTC_INTERSECTN);
+        RTCScene result = rtcDeviceNewScene(m_device, RTC_SCENE_STATIC, RTC_INTERSECT1 | RTC_INTERSECT4 | RTC_INTERSECT8 | RTC_INTERSECT16 );
         CheckEmbreeError();
         ThrowIf(!mesh->puretriangle(), "Only triangle meshes supported by now.");
 

RadeonRays/src/intersector/intersector_2level.cpp

@@ -130,10 +130,14 @@ namespace RadeonRays
     {
         std::string buildopts =
 #ifdef RR_RAY_MASK
-            "-D RR_RAY_MASK";
+            "-D RR_RAY_MASK ";
 #else
             "";
 #endif
+        
+#ifdef USE_SAFE_MATH
+        buildopts.append("-D USE_SAFE_MATH ");
+#endif
 
 #ifndef RR_EMBED_KERNELS
         if ( device->GetPlatform() == Calc::Platform::kOpenCL )