Merge branch 'shawnz_bug_fix' into 'master'

Bug fix for 5280038, 5277193, 5281036 and 5294720

See merge request cuda-samples/cuda-samples!112

CHANGELOG.md

@@ -43,9 +43,12 @@
     * `4_CUDA_Libraries`
         * `jitLto`
     * `7_libNVVM`
+        * `cuda-c-linking`
         * `device-side-launch`
         * `simple`
         * `uvmlite`
+    * `8_Platform_Specific/Tegra`
+        * `EGLSync_CUDAEvent_Interop`
 * Updated the sample using CUDA API "cudaGraphAddNode"/"cudaStreamGetCaptureInfo" with adding "cudaGraphEdgeData" pointer parameter as they are updated to "cudaGraphAddNode_v2"/"cudaStreamGetCaptureInfo_v3" by default in CUDA 13.0:
     * `3_CUDA_Features`
         * `graphConditionalNodes`
@@ -57,6 +60,23 @@
 * Replaced "thrust::identity<uint>()" with "cuda::std::identity()" as it is deprecated in CUDA 13.0.
     * `2_Concepts_and_Techniques`
         * `segmentationTreeThrust`
+* Updated the the headers file and samples for CUFFT error codes update.
+    * Deprecated CUFFT errors:
+        * `CUFFT_INCOMPLETE_PARAMETER_LIST`
+        * `CUFFT_PARSE_ERROR`
+        * `CUFFT_LICENSE_ERROR`
+    * New added CUFFT errors:
+        * `CUFFT_MISSING_DEPENDENCY`
+        * `CUFFT_NVRTC_FAILURE`
+        * `CUFFT_NVJITLINK_FAILURE`
+        * `CUFFT_NVSHMEM_FAILURE`
+    * Header files and samples that are related with this change:
+        * `Common/helper_cuda.h`
+        * `4_CUDA_Libraries`
+            * `simpleCUFFT`
+            * `simpleCUFFT_2d_MGPU`
+            * `simpleCUFFT_MGPU`
+            * `simpleCUFFT_callback`
 
 ### CUDA 12.9
 * Updated toolchain for cross-compilation for Tegra Linux platforms.

Common/helper_cuda.h

@@ -147,6 +147,9 @@ static const char *_cudaGetErrorEnum(cufftResult error) {
     case CUFFT_NOT_IMPLEMENTED:
       return "CUFFT_NOT_IMPLEMENTED";
 
+    case CUFFT_NOT_SUPPORTED:
+      return "CUFFT_NOT_SUPPORTED";
+
     case CUFFT_MISSING_DEPENDENCY:
       return "CUFFT_MISSING_DEPENDENCY";
 
@@ -670,7 +673,10 @@ inline int _ConvertSMVer2Cores(int major, int minor) {
       {0x90, 128},
       {0xa0, 128},
       {0xa1, 128},
+      {0xa3, 128},
+      {0xb0, 128},
       {0xc0, 128},
+      {0xc1, 128},
       {-1, -1}};
 
   int index = 0;
@@ -722,7 +728,10 @@ inline const char* _ConvertSMVer2ArchName(int major, int minor) {
       {0x90, "Hopper"},
       {0xa0, "Blackwell"},
       {0xa1, "Blackwell"},
+      {0xa3, "Blackwell"},
+      {0xb0, "Blackwell"},
       {0xc0, "Blackwell"},
+      {0xc1, "Blackwell"},
       {-1, "Graphics Device"}};
 
   int index = 0;

Common/helper_cuda_drvapi.h

@@ -118,7 +118,10 @@ inline int _ConvertSMVer2CoresDRV(int major, int minor) {
       {0x90, 128},
       {0xa0, 128},
       {0xa1, 128},
+      {0xa3, 128},
+      {0xb0, 128},
       {0xc0, 128},
+      {0xc1, 128},
       {-1, -1}};
 
   int index = 0;

Samples/3_CUDA_Features/cdpAdvancedQuicksort/CMakeLists.txt

@@ -23,6 +23,7 @@ set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -Wno-deprecated-gpu-targets")
 
 if(ENABLE_CUDA_DEBUG)
     set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -G")        # enable cuda-gdb (may significantly affect performance on some targets)
+    set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -maxrregcount=64") # Limit register usage to 64 for the 'big_bitonicsort kernel
 else()
     set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -lineinfo") # add line information to all builds for debug tools (exclusive to -G option)
 endif()

Samples/4_CUDA_Libraries/simpleCUFFT_callback/simpleCUFFT_callback.cu

@@ -182,14 +182,6 @@ int runTest(int argc, char **argv)
     checkCudaErrors(cudaMemcpyFromSymbol(&hostCopyOfCallbackPtr, myOwnCallbackPtr, sizeof(hostCopyOfCallbackPtr)));
 
     // Now associate the load callback with the plan.
-    cufftResult status =
-        cufftXtSetCallback(cb_plan, (void **)&hostCopyOfCallbackPtr, CUFFT_CB_LD_COMPLEX, (void **)&d_params);
-    if (status == CUFFT_LICENSE_ERROR) {
-        printf("This sample requires a valid license file.\n");
-        printf("The file was either not found, out of date, or otherwise invalid.\n");
-        return EXIT_WAIVED;
-    }
-
     checkCudaErrors(
         cufftXtSetCallback(cb_plan, (void **)&hostCopyOfCallbackPtr, CUFFT_CB_LD_COMPLEX, (void **)&d_params));
 

Samples/7_libNVVM/cuda-c-linking/cuda-c-linking.cpp

@@ -248,7 +248,7 @@ int main(int argc, char **argv)
 
     // Create the CUDA context.
     CUcontext context;
-    checkCudaErrors(cuCtxCreate(&context, 0, device));
+    checkCudaErrors(cuCtxCreate(&context, NULL, 0, device));
 
     // Create a JIT linker and generate the result CUBIN.
     CUlinkState  linker;