better understand to make programs faster with SIMT

[I C everything]

Hi
which algos are suitable/not suitable to parallelism?
similar to SIMD suitable algos?
well you might cheat,put it on some server,that more probable has more cores,than computer
I have noticed on computer,that without your own program has multiple threads,using file system OS,runs its read/write in separate thread
,in case of disk cache it updates file written first after close
earlier I mostly used a separate thread to read/write big files,to avoid GUI lag

[I C everything]

I have now tried testing with create 4 threads and 16 threads from main,the displaying calculation on console is seemly delayed in beginning of main,right after createthread calls
checking cpu performance,no matter single main or many threads there is slight delay before reaching about 100% performance