[QST] Question about UniversalFMA

[leven-comeon]

What is your question?

When executing the sgemm_70.cu program, I used the print_latex tool to observe the structure of mmaC, and its specific details are shown in the figure.

As far as I know, UniversalFMA uses CUDA Cores for computation. However, during the execution of the General Matrix Multiplication (GEMM) operation, a puzzling phenomenon occurred: it seems that a thread can access the values in the internal registers of other threads. For example, thread T17 can read the values in the registers of threads T1 and T16 to complete the computation.
I'm wondering if you could kindly explain to me the reason behind this behavior?

[thakkarV]

The A and V TV mappings only show the first thread. Multiple threads are each reading A and B in reality