Friday, May 22, 2015

Log 052215

Implemented a general compare function on the AST pointers. Did some memory profiling. Valgrind is my best friend. For future reference:

valgrind --tool=massif --pages-as-heap=yes --massif-out-file=massif.out.

I really have to get used to how much memory is allocated these days. Printing an AST:

-rw-rw-r--. 1 marco marco 2596096 May 22 20:26 test0.hi

Right,source code of 2,596,096 bytes becomes 279,101,440 bytes heap usage; 2.6 seconds to print. Ah well, at least I don't expect Moore's law to end soon when it comes to memory.

Plotted (x-axis is sample number over time, the 2.6 seconds), it becomes:

Right. Two spikes I don't get. Possibly it copies the entire vector of declarations while rewriting it. Nothing I want to solve at the moment.

The steady climb is a bit worrying, but valgrind reports no memory leaks. Moreover, I am pretty sure I do everything right, it copies pointers a lot while visiting the AST, a lot of refcounting but no 'real' pressure on the garbage collector. Probably due to that the AST remains in memory until the program exits, or C's memory allocator 'gives up' upto the next garbage collection point (yes, C does garbage collection too); yeah well, RAII...

No comments:

Post a Comment