Evaluation

For the evaluation, we run your solution on four different servers, courtesy of the HPI data center:

• AMD (EPYC 7F72)
• ARM (Ampere Altra Max)
• IBM (Power8)
• Intel (Xeon E7-4880)

For the final evaluation, we use four additional servers:

• AMD (EPYC 7343)
• ARM (NVIDIA GH200 Grace Hopper)
• IBM (Power10)
• Intel (Xeon Platinum 8352Y)

For the final evaluation after the submission deadline, four additional undisclosed server will be included. See starter code for more information on the provided hardware. On each server, we measure the runtime of every single query execution, including the context creation. Then, we sum up all execution times per server as \( R_{Intel}, R_{AMD}, R_{ARM}, R_{IBM} \).

Our overall metric is the geometric mean over all server runtimes, \( R = \sqrt[4]{R_{Intel} * R_{AMD} * R_{ARM} * R_{IBM}} \).