hello,
i am building a c++ application using jolt as a shared library. i want
to add a JIT compiler to an existing smalltalk(-like) virtual machine.
we use jolt to compile code while the virtual machine is running. after
compilation we execute the code instead of interpreting it again and again.
i would like to profile my application with valgrind, gprof or a
similiar tool. unfortunately, valgrind seems to have problems with the
runtime code generation that is done in our application with jolt:
$ valgrind --smc-check=all --tool=callgrind <myApp and Parameters>
==8868== Callgrind, a call-graph generating cache profiler.
==8868== Copyright (C) 2002-2008, and GNU GPL'd, by Josef Weidendorfer
et al.
==8868== Using LibVEX rev 1884, a library for dynamic binary translation.
==8868== Copyright (C) 2004-2008, and GNU GPL'd, by OpenWorks LLP.
==8868== Using valgrind-3.4.1-Debian, a dynamic binary instrumentation
framework.
==8868== Copyright (C) 2000-2008, and GNU GPL'd, by Julian Seward et al.
==8868== For more details, rerun with: -v
==8868==
==8868== For interactive control, run 'callgrind_control -h'.
==8868==
==8868== Process terminating with default action of signal 11 (SIGSEGV)
==8868== Access not within mapped region at address 0xBEF4B000
==8868== at 0x421F3C2: GC_push_all_eager (in jolty.so)
==8868== by 0x421F407: GC_push_all_stack (in jolty.so)
==8868== by 0x4217B78: GC_push_all_stacks (in jolty.so)
==8868== by 0x4215512: GC_default_push_other_roots (in jolty.so)
==8868== by 0x422075B: GC_push_roots (in jolty.so)
==8868== by 0x421DF9F: GC_mark_some (in jolty.so)
==8868== by 0x421996D: GC_stopped_mark (in jolty.so)
==8868== by 0x421968D: GC_try_to_collect_inner (in jolty.so)
==8868== by 0x4213D9F: GC_init_inner (in jolty.so)
==8868== by 0x421394B: GC_init (in jolty.so)
==8868== by 0x4226122: _libid_init (libid.c:1048)
==8868== by 0x40EBCD6: GetJoltInitially (jolty.c:7)
==8868== If you believe this happened as a result of a stack overflow
in your
==8868== program's main thread (unlikely but possible), you can try to
increase
==8868== the size of the main thread stack using the --main-stacksize=
flag.
==8868== The main thread stack size used in this run was 8388608.
==8868==
==8868== Events : Ir
==8868== Collected : 4645259
==8868==
==8868== I refs: 4,645,259
Segmentation fault
the programs works fine when executed normally (without valgrind as a
profiler). jolty.so is the jolt shared library that we use. we use the
latest jolt revision available in the repository (r647).
i have tried using valgrinds "--smc-check=all" flag for self-modifying
code because the manual says:
"[...] Useful if you are debugging a JIT compiler or some other dynamic
code generation system. After this call, attempts to execute code in the
invalidated address range will cause Valgrind to make new translations
of that code, which is probably the semantics you want. [...]
Alternatively, for transparent self-modifying-code support,
use--smc-check=all, or run on ppc32/Linux or ppc64/Linux."
however, this seems to have no effect. the program is still terminated
with a segfault when running in valgrind.
is there another way to profile my vm application that uses jolt?
regards,
ingo
_______________________________________________
fonc mailing list
[email protected]
http://vpri.org/mailman/listinfo/fonc
