Dear Haiku developers,
I compiled an application written in C++ with OpenMP directives, embarrassingly parallel in Haiku. I did the same thing in other operating systems such as windows, linux, freebsd, solaris. I get almost perfect scalability in all of them even running under a virtual machine. With haiku, the runtimes of some of the parallel runs becomes 5-6 times slower when running in 2 threads or more. Is there anyone interested to investigate this issue? I repeat, that the laptop is using hyperformance mode, it is connected to the plug (not on battery) and in other OS parallel scalability observed is the expected one 2times faster in 2 cores, 3.85 times faster in 4 cores and 7 times faster in 8 cores. In Haiku the parallel speedup decreases as the number of cores increases.
Obviously you should specify the VM software.
It might be the same issue I observed in Java apps don't scale on Haiku's openjdk · Issue #5136 · haikuports/haikuports · GitHub. I’ll try to reproduce it with R1/Beta5 and add some screenshots or a screen recording.
VM is VMWare Player. But under VMWare player on FreeBSD I do not observe this problem. Neither on Linux. Something is not carefully designed. It is obvious.