Version 28.4 of sci-mathematics/gimps is available for download. This is another beta for the new version. This version contains optimizations and performance improvements for Intel Haswell processors. Very fast memory is required to fully exploit these optimizations. Small performance improvements are expected for Sandy Bridge and Ivy Bridge Intel CPU as well and it works (with no known performance regressions) on all CPUs supported by 27.x Changes since the latest version are: 1. Performance tweaks: It reduces the sin/cos constants stored in memory by computing the missing sin/cos constants at runtime. Thus, this version will in most cases be slightly slower running one worker and slightly faster running all workers. 2. Benchmarking code updated. It now starts at 1024K. Multi-threaded benchmarks are now run only on the most useful combinations. For example, a quad-core hyperthreaded benchmark runs on 1 cpu, 1 cpu hyperthreaded, 2 cpus, 3 cpus, 4 cpus, and 4 cpus hyperthreaded. The trial factoring benchmark is not run by default. 3. Benchmarking also runs a multiple-worker test to measure memory bandwidth related degradation. 4. Many new undoc.txt options for benchmarking.
You may also add the improved systemd unit file from bug #503714.
I went straight to 28.5 which was also on the server and seems to have a 32bit version again (curiously suffixed "linux" instead of "linux32" as before). New systemd file is included.
