[Issue]: [CMS] Race condition in 256x256x64NN/TT BF16 case

[Alex-Vasile]

Problem Description

The last LRA1 issued at 66/67 (codepath 0/1) is only guaranteed by the SWaitCnt at 84, rather than the one at 74. (The message on the SWaitCnt is also wrong).

This is a problem since the PackA1s start at 75, the pack at index 77 uses the results for the last LRA1.

Operating System

Linux

CPU

AMD

GPU

AMD

ROCm Version

7.1

ROCm Component

No response

Steps to Reproduce

No response

(Optional for Linux users) Output of /opt/rocm/bin/rocminfo --support

No response

Additional Information

No response