[kaffe] mipsel and jit3
dichro at rcpt.to
Fri Oct 8 19:54:21 PDT 2004
On Sat, 2004-10-09 at 12:01, Mikolaj Habryn wrote:
> Hmm. Having noticed KAFFE_VMDEBUG, I'll go play with that for a while.
Poking at this a little more, and with a working gdb no less, one
<JIT: java/lang/String.<clinit>()V time 41ms (41ms) @ 0x10093c08 (0x1006a0b4)>
<JIT: java/lang/String$ICComp.<init>(Ljava/lang/String$1;)V time 3ms (44ms) @ 0x100939e8 (0x1006bb30)>
<JIT: java/lang/String$ICComp.<init>()V time 3ms (47ms) @ 0x10093dc8 (0x1006baf4)>
<JIT: java/lang/Object.<init>()V time 0ms (47ms) @ 0x2ab50a50 (0x1005e2a4)>
Illegal instruction (core dumped)
The curious thing is that the last JITted method is apparently put into
0x2ab50a50, which, according to gdb on the core file, is entirely bogus.
The previously JITted methods are all in 0x100.... and can be seen in
Core was generated by `./kaffe-bin -v -verbosejit -Xbootclasspath:./sparse-2.jar -cp parsesol.jar to.r'.
Program terminated with signal 4, Illegal instruction.
#0 0x1005e2b0 in ?? ()
Curioser yet is exactly how it came to executing instructions at
0x100532b0 which looks damn close to where the Java bytecode is for this
And, and... how exactly did that method get JITted in 0 ms? Something
very, very fishy here.
More information about the kaffe