Rusticl vs. Intel Compute Runtime Performance For OpenCL On Battlemage
Interestingly with the cl-mem copy performance, Rusticl did come out slightly ahead of the Intel Compute Runtime.
The Intel Compute Runtime with Battlemage had a huge advantage on lower OpenCL kernel latency compared to Rusticl. The massive lead for the Intel Compute Runtime comes down to the Ultra Low Latency Scheduling (ULLS) implemented. See Benchmarks: OpenCL Kernel Latency ~76x Lower For Intel Lunar Lake With Updated Compute Runtime for the ULLS impact. That's why the Intel Compute Runtime does so well here but not Rusticl.
The Intel Compute Runtime maintained the lead in most of the other clpeak benchmarks.
Interestingly the clpeak's enqueue read and write buffer tests showed nice leads on Rusticl over the Intel Compute Runtime OpenCL driver.
