[BUG] Mixed Input H100 Kernel Hangs

FE4M3 x BF16 Kernel Hangs when run with beta=1

Please compile the kernel cutlass3x_sm90_tensorop_s64x128x16gemm_e4m3_bf16_f32_bf16_bf16_cvt_64x128x128_8x1x1_0_tnt_align16_warpspecialized_pingpong_epi_tma in profiler and run it with beta=0 and beta=1. It is not just the profiler, we are using this kernel for some of our shapes where it came to be the winner; however, it hangs when we run it with beta = 1, and our runs are not with profiler. So this might be a bug in the kernel and not just a profiler issue. We used CUDA 12.4.99.

beta = 1 hangs

./tools/profiler/cutlass_profiler --kernels=cutlass3x_sm90_tensorop_s64x128x16gemm_e4m3_bf16_f32_bf16_bf16_cvt_64x128x128_8x1x1_0_tnt_align16_warpspecialized_pingpong_epi_tma --m=128 --n=256 --k=512 --beta=1

beta = 0 works

./tools/profiler/cutlass_profiler --kernels=cutlass3x_sm90_tensorop_s64x128x16gemm_e4m3_bf16_f32_bf16_bf16_cvt_64x128x128_8x1x1_0_tnt_align16_warpspecialized_pingpong_epi_tma --m=128 --n=25
6 --k=512



=============================
  Problem ID: 1

        Provider: CUTLASS
   OperationKind: gemm
       Operation: cutlass3x_sm90_tensorop_s64x128x16gemm_e4m3_bf16_f32_bf16_bf16_cvt_64x128x128_8x1x1_0_tnt_align16_warpspecialized_pingpong_epi_tma

          Status: Success
    Verification: ON
     Disposition: Not verified

reference_device: Not run
          cuBLAS: Not run
           cuDNN: Not run

            Math: 4002.22 GFLOP/s
=============================

Start with tools/profiler/cutlass_profiler and reproduce the named CUTLASS kernel using the provided m=128, n=256, k=512 commands under CUDA 12.4.99, comparing beta=0 with beta=1. Determine whether the hang is in the kernel or profiler, and verify that the reported behavior is resolved or clearly characterized.