5090 FP4 Gemm Flops cublaslt_gemm benchmark error

this is logs
root@ubuntu:/data/superbenchmark/superbench/benchmarks/micro_benchmarks/cublaslt_gemm# cmake . -- The CXX compiler identification is GNU 11.4.0 -- Detecting CXX compiler ABI info -- Detecting CXX compiler ABI info - done -- Check for working CXX compiler: /usr/bin/c++ - skipped -- Detecting CXX compile features -- Detecting CXX compile features - done -- Looking for C++ include pthread.h -- Looking for C++ include pthread.h - found -- Performing Test CMAKE_HAVE_LIBC_PTHREAD -- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Success -- Found Threads: TRUE -- The CUDA compiler identification is NVIDIA 12.8.61 -- Detecting CUDA compiler ABI info -- Detecting CUDA compiler ABI info - done -- Check for working CUDA compiler: /usr/local/cuda-12.8/bin/nvcc - skipped -- Detecting CUDA compile features -- Detecting CUDA compile features - done -- Configuring done -- Generating done -- Build files have been written to: /data/superbenchmark/superbench/benchmarks/micro_benchmarks/cublaslt_gemm root@ubuntu:/data/superbenchmark/superbench/benchmarks/micro_benchmarks/cublaslt_gemm# make [ 25%] Building CXX object CMakeFiles/cublaslt_utils.dir/cublaslt_utils.cc.o [ 50%] Linking CXX shared library libcublaslt_utils.so [ 50%] Built target cublaslt_utils [ 75%] Building CUDA object CMakeFiles/cublaslt_gemm.dir/cublaslt_gemm.cu.o [100%] Linking CUDA executable cublaslt_gemm [100%] Built target cublaslt_gemm root@ubuntu:/data/superbenchmark/superbench/benchmarks/micro_benchmarks/cublaslt_gemm# ls CMakeCache.txt cmake_install.cmake cublaslt_gemm cublaslt_utils.cc libcublaslt_utils.so CMakeFiles CMakeLists.txt cublaslt_gemm.cu cublaslt_utils.h Makefile root@ubuntu:/data/superbenchmark/superbench/benchmarks/micro_benchmarks/cublaslt_gemm# ./cublaslt_gemm -w 50 -i 200 -t fp4e2m1 -m 8192 -n 8192 -k 8192 cuBLAS call cublasLtMatmulAlgoGetHeuristic(handle_.get(), op_desc_.get(), a_desc_.get(), b_desc_.get(), c_desc_.get(), d_desc_.get(), preference_.get(), max_algorithm_count, results.data(), &found_algorithm_count) failed at /data/superbenchmark/superbench/benchmarks/micro_benchmarks/cublaslt_gemm/cublaslt_utils.cc:216 'the requested functionality is not supported' root@ubuntu:/data/superbenchmark/superbench/benchmarks/micro_benchmarks/cublaslt_gemm# ./cublaslt_gemm -w 50 -i 200 -t fp8e4m3 -m 8192 -n 8192 -k 8192 8192 8192 8192 0 2280.463379 482.114525

nvidia-smi

The benchmark is in benchmarks/micro_benchmarks/cublaslt_gemm, with the failure reported at cublaslt_utils.cc:216 while building cublaslt_gemm.cu through CMake. Start by reproducing the fp4e2m1 command and comparing it with the working fp8e4m3 run. Done means the expected FP4 benchmark behavior is established and the reported heuristic error is resolved or clearly diagnosed.