Re: RFR(L): 8028468: Add inlining information into ciReplay

[Vladimir Kozlov]

Thank you, Roland

On 1/8/14 12:41 AM, Roland Westrelin wrote:

Thanks for the new webrev and explanations.

ciReplay.hpp
59 // Replay data file replay_pid%p.log is also created when VM crushes

when VM crashes

Fixed.

Thanks,
Vladimir

That looks good to me.

Roland.

On Jan 8, 2014, at 5:02 AM, Vladimir Kozlov <vladimir.koz...@oracle.com> wrote:

Thank you, Roland

On 1/7/14 2:39 AM, Roland Westrelin wrote:

http://cr.openjdk.java.net/~kvn/8028468/webrev/

New webrev:

http://cr.openjdk.java.net/~kvn/8028468/webrev.01/

Should the agent/doc/cireplay.html be updated? Is there another doc on how to 
use the replay support somewhere (wiki)?

cireplay.html describes how to use SA to extract compilation replay data from 
core files. Nothing changed there with my changes.
I don't think we have any wiki for compilation replay.

I added all replay command descriptions and examples into ciReplay.hpp.

ciReplay.cpp
typos:
// Use pointer because we may need to retirn inline records

// Replay Inlinig

vmError.cpp
typo:
// Do not overwite file during replay

Typos are fixed.

compile.hpp
Shouldn’t _replay_inline_data be private with an accessor?

1183   // Dump inlining replay data to the stream.
1184   void dump_inline_data(outputStream* out);
1185   void* _replay_inline_data;

Done.

ciReplay.cpp
Why was this changed? Why was it there in the first place?
1052   // Make sure we don't run with background compilation
1053   //  BackgroundCompilation = false;

I forgot to uncomment it. It is done only when compilation (not inlining) is 
replayed. It is special case. Such replaying is done by putting compilation 
task on compile queue immediately after VM initialization. So we should wait 
(-Xbatch, -XX:-BackgroundCompilation) until it is finished because we don't 
need to execute java code (there is no program to execute). And after 
compilation we exit VM:

void ciReplay::replay(TRAPS) {
  int exit_code = replay_impl(THREAD);

  Threads::destroy_vm();

  vm_exit(exit_code);
}

Existing fields in class CompileReplay that don’t start with _ make the code 
confusing:

   char* bufptr;
   char* buffer;
   int   buffer_length;
   int   buffer_end;
   int   line_no;

etc. Maybe it would be worthwhile to fix that as part of this change?

I added '_' to all fields in this file.

Can buffer be a growableArray? If not factor this code and other similar code 
in a method?
  437       if (pos + 1 >= buffer_length) {
  438         int newl = buffer_length * 2;
  439         char* newb = NEW_RESOURCE_ARRAY(char, newl);
  440         memcpy(newb, buffer, pos);
  441         buffer = newb;
  442         buffer_length = newl;
  443       }

Same for:
  445         // null terminate it, reset the pointer and process the line
  446         buffer[pos] = '\0';
  447         buffer_end = pos++;
  448         bufptr = buffer;

Done. That code is moved to new method get_line().

Shouldn’t the fields below be private?
  422   ciKlass* _iklass;
  423   Method* _imethod;
  424   int _entry_bci;
  425   int _comp_level;

Done.

I think
ciInlineRecord* find_ciInlineRecord(Method* method, int bci, int inline_depth)
should be implemented by calling:
static ciInlineRecord* find_ciInlineRecord(GrowableArray<ciInlineRecord*>*  
records, Method* method, int bci, int inline_depth)
to avoid duplication of code.

Done.

I also missed _ci_inline_records field initialization so I added it to 
CompileReplay constructor.

Thanks,
Vladimir

Roland.