Empty framework for a Tensor Core GEMM project without CMake.
makeOptional architecture override:
make CUDA_ARCH=sm80make run
# or
./bin/sgemm --m 4096 --n 4096 --k 4096 --kernel wmma --iters 100 --warmup 20include/: common structs and autotune config definitionssrc/: benchmark driver + kernel/autotune/cutlass placeholdersscripts/: helper scripts for benchmark/profiling/plottingresults/: CSV/NCU/figure outputsdocs/: design and methodology notes
This is intentionally a scaffold. Kernel bodies and autotune logic are stubs.