Christian Thalinger wrote:
done.On Jan 4, 2011, at 10:00 PM, Paul Hohensee wrote:These two rfes implement per-thread approximate memory allocation tracking. 6173675 also adds multi-thread-id versions of getThreadCpuTime and getThreadUserTime.6173675 M&M: approximate memory allocation rate/amount per thread 7003271 Hotspot should track cumulative Java heap bytes allocated on a per-thread basissrc/os/solaris/vm/thread_solaris.inline.hpp: 56 Thread *candidate = ThreadLocalStorage::_get_thread_cache[ix]; You could have fixed the * in that line too. done, plus a bit more. :)src/share/vm/opto/macro.cpp: 1242 new (C, 4) StorePNode( needgc_false, contended_phi_rawmem, eden_top_adr, 1243 TypeRawPtr::BOTTOM, new_eden_top ); 1249 new (C, 5) StorePConditionalNode( needgc_false, contended_phi_rawmem, eden_top_adr, 1250 new_eden_top, fast_oop/*old_eden_top*/ ); 1284 Node* new_alloc_bytes = new (C, 3) AddLNode( alloc_bytes, alloc_size ); Could you also remove the spaces before and after the parentheses when you're at it? I'd rather leave this alone, since both eden_allocate() and tlab_allocate() use the two-parameter paradigmMacroAssembler::incr_allocated_bytes: It seems we could use a RegisterOrConstant size_in_bytes argument instead of Register var_size_in_bytes, int con_size_in_bytes. and I'd have to change those also. Thanks for the review!Otherwise it looks good. paul -- Christian |
- Pls review 6173675/7003271 Paul Hohensee
- Re: Pls review 6173675/7003271 Christian Thalinger
- Re: Pls review 6173675/7003271 Paul Hohensee
- Re: Pls review 6173675/7003271 Mandy Chung