Gabi
2010-Apr-14 05:57 UTC
[LLVMdev] Most efficient way to count # of intrcutions in a module ?
I need to count the total number of instructions in a given module. The only way I found is using the obvious double iteration as seen below. Is there any more efficient way to achieve this ? I don't like the fact that simple calculation like that has performance of o(F*B) (# of functions * # of blocks per function) unsigned int calcSize(Module* mod) { unsigned int size = 0; for (Module::iterator f = mod->begin(), fe = mod->end(); f != fe; ++f) for (Function::iterator b = f->begin(), be = f->end(); b != be; ++b) { BasicBlock* bb = (BasicBlock*) b; size+=bb->getInstList().size(); } return size; } -- Regards, Gabi http://bugspy.net
Nick Lewycky
2010-Apr-14 06:13 UTC
[LLVMdev] Most efficient way to count # of intrcutions in a module ?
Gabi wrote:> I need to count the total number of instructions in a given module. > The only way I found is using the obvious double iteration as seen > below. Is there any more efficient way to achieve this ? I don't like > the fact that simple calculation like that has performance of o(F*B) > (# of functions * # of blocks per function)If you just want to know given a .bc file, you can run 'opt -instcount -stats' which is a pass that does pretty much what you wrote, except that it does a per-instruction breakdown which means that it will be a little slower.> unsigned int calcSize(Module* mod) > { > unsigned int size = 0; > for (Module::iterator f = mod->begin(), fe = mod->end(); f != fe; ++f) > for (Function::iterator b = f->begin(), be = f->end(); b != be; ++b) > { > BasicBlock* bb = (BasicBlock*) b; > size+=bb->getInstList().size(); > } > return size; > }Your implementation isn't actually O(n^2) if that's what you're worried about. It's just O(n) where n is the # instructions in the whole program. You're doing it as efficiently as possible. Also, you should be able to write 'size += b->getInstList().size()' without the explicit cast. Nick
Maybe Matching Threads
- [LLVMdev] JIT : Does it cache the compiled code of executed functions upon runFunction(..)?
- [LLVMdev] JIT : Does it cache the compiled code of executed functions upon runFunction(..)?
- [LLVMdev] How to avoid memory leaks
- [LLVMdev] JIT : Does it cache the compiled code of executed functions upon runFunction(..)?
- [LLVMdev] JIT : Does it cache the compiled code of executed functions upon runFunction(..)?