MarathonMan
Emulator Developer
- Thread Starter
- #861
@arbin: Yes, I'll drop by.
With LTO and optimizations disabled, the callgraph is probably not really to be trusted.
The assertions that I wrote more or less should probably just be taken out all together. The assert should actually be 0x3FFFFF, too, since it's a 4-byte array and not 1-byte array. Thanks for commenting, I'll fix tonight.
Don't know if its of any use to you, but I did some profiling with google's pprof tool, see the attachment. One caveat is that I turned off all inlining to better see whats going on.
EDIT: A small bug in debug mode exclusive (the irony): In rdp/Core.c there is
#define RREADIDX32(in) (assert(in <= 0x7FFFFC), bswap32(rdram[in]));
which is later called as
RREADIDX32(dp_current_al & 0x3fffff);
It fails the assert because of the macro and operator precedence doing bad teamwork![]()
With LTO and optimizations disabled, the callgraph is probably not really to be trusted.
The assertions that I wrote more or less should probably just be taken out all together. The assert should actually be 0x3FFFFF, too, since it's a 4-byte array and not 1-byte array. Thanks for commenting, I'll fix tonight.