Larry's test was exporting a ProRes 4444 project to ProRes 422. I just did a similar test on my M1 Ultra using a timeline with no effects and got the results in the attached graphic. I compared FCP 10.6.3, Resolve Studio 17.4.6 and Premiere Pro 22.5.0 on Monterey 12.4, 128GB Mac Studio M1 Ultra, 8TB internal drive, all media and projects on the internal drive. The graphs were produced by iStat Menus.
FCP and Resolve are very fast at transcoding between various ProRes formats. It's interesting that Resolve was about 2x faster than FCP, but both were fast. Premiere was much slower.
In my tests the poor performance of Premiere seemed superficially reflected in traditional CPU, GPU and I/O measurements but in general those are not reliable and can be misleading. E.g, there is no instrumentation I'm aware of in MacOS for the all-important video engines. You can theoretically have an encode/decode workflow that shows low CPU, low GPU but completes rapidly. That's because much of the work is happening out of sight on the encode/decode units.
M1 Ultra has eight total video engines -- 4 encode/decode units for ProRes and 4 for H264/HEVC. In theory all 8 could be active but there is no visibility to the end user of their activity. For a single-stream task such as exporting to a single file, it's unclear how well that could be parallelized. In theory an all-intra codec like ProRes could be segmented and each piece dispatched to a separate engine, then concatenated for final output. Doing that on a Long GOP format is more difficult, and some of those are "open GOPs" meaning each GOP cannot be processed independently.