get_number_of_processors does not take into account hyper-threading

[m-ober]

Currently, the procedure get_number_of_processors is reading /proc/cpuinfo and returns the number of logical cores. On a 6-core CPU with hyper-threading, this results in 12 threads being launched for synthesis - when enabling out of context mode. This is probably slower than launching only 6 threads due to HT overhead. It also happens that this number of threads exhausts even large amounts of RAM (32 GiB in my case), so it could be required to lower the number of threads even below the number of available (physical) cores.

My suggestion is:

• Use nproc (output can be used directly) instead of (probably error prone) parsing of /proc/cpuinfo
• Maybe introduce an environment variable like TAPASCO_MAX_THREADS to set a maximum number of threads or TAPASCO_NUM_THREADS to use a fixed number of threads

[cahz]

Selecting the right number of threads is a problem. Relying on the number of cores is not a good metric as memory usage can explode, depending on which FPGA is targeted.

• nproc is simpler to use, but still includes hyper-threading
• there is already a command line parameter --maxThreads and --maxTasks to set the maximum number of threads / tasks