jpsdr
What do you think if I make 64-byte alignment for buffers in x64 build, and in x86 build it will be forced for specific filter only (it will do inefficient reallocation)?
I am afraid that x86 process may run out of memory merely because of the increased alignment.
Unfortunately there is no already implemented way to enforce only useful alignment, and this seems difficult.
edit: nevermind, have better idea.
Last edited by shekh; 5th April 2017 at 14:30.
|