[PATCH] vm: open zipfiles in the classpath only once

This fixes a performance issue with zip files that appeared because we used to open the zip files for _every_ class that we wanted to load. Signed-off-by: Vegard Nossum vegard.nos...@gmail.com --- vm/classloader.c | 186 +++--- 1 files changed,

[PATCH 2/3] vm: implement bytecode_offset_to_line_no()

Signed-off-by: Tomek Grabiec tgrab...@gmail.com --- include/vm/method.h |2 ++ jit/bc-offset-mapping.c | 18 +- test/arch-x86/Makefile |1 + vm/method.c | 11 ++- 4 files changed, 22 insertions(+), 10 deletions(-) diff --git

[PATCH 2/3] vm: implement native_vmruntime_maplibraryname()

Signed-off-by: Tomek Grabiec tgrab...@gmail.com --- vm/jato.c | 61 +++-- 1 files changed, 59 insertions(+), 2 deletions(-) diff --git a/vm/jato.c b/vm/jato.c index 865deda..fb7ecb0 100644 --- a/vm/jato.c +++ b/vm/jato.c @@ -130,8

[PATCH 3/3] vm: preliminary support for JNI

Support for loading objects with native code, invoking JNI native methods. Few JNI native interface functions implemented. Signed-off-by: Tomek Grabiec tgrab...@gmail.com --- Makefile |4 +- arch/x86/emit-code.c | 24 +++ arch/x86/insn-selector_32.brg |3

Re: [PATCH] vm: implement java/io/VMFile.isDirectory()

2009/7/5 Tomek Grabiec tgrab...@gmail.com: Signed-off-by: Tomek Grabiec tgrab...@gmail.com ---  vm/jato.c |   24  1 files changed, 24 insertions(+), 0 deletions(-) diff --git a/vm/jato.c b/vm/jato.c index dfdd60b..9086f1d 100644 --- a/vm/jato.c +++ b/vm/jato.c

[PATCH 2/2] vm: fix native_vmobject_getclass()

Add missing return after throw_from_native() and put object-class check in assertion Reported-by: Vegard Nossum vegard.nos...@gmail.com Signed-off-by: Tomek Grabiec tgrab...@gmail.com --- vm/jato.c |5 - 1 files changed, 4 insertions(+), 1 deletions(-) diff --git a/vm/jato.c

[PATCH 1/7] regression: compile and run the test cases in alphabetic order

This makes it somewhat easier to find tests if you're looking them up by name. It looks better too, in my opinion. Signed-off-by: Vegard Nossum vegard.nos...@gmail.com --- Makefile| 50 +++--- regression/run-suite.sh | 46

[PATCH 5/7] vm: add vmtype - bytecode type conversion

I suspect that we should just make object allocation take a vm type instead (and convert the callers that need it to send a vm type instead of a bytecode type). But that's for a later cleanup, we need this now. Signed-off-by: Vegard Nossum vegard.nos...@gmail.com --- include/vm/types.h |1 +

[PATCH 7/7] regression: add CloneTest

Signed-off-by: Vegard Nossum vegard.nos...@gmail.com --- Makefile |1 + regression/jvm/CloneTest.java | 64 + regression/run-suite.sh |1 + 3 files changed, 66 insertions(+), 0 deletions(-) create mode 100644

[PATCH 6/7] vm: implement [VM]Object.clone()

Signed-off-by: Vegard Nossum vegard.nos...@gmail.com --- include/vm/object.h|2 + test/vm/preload-stub.c |1 + vm/class.c |5 ++- vm/jato.c |9 - vm/object.c| 78 5 files changed, 91

[PATCH 2/2] x86: fix bug in emit_jni_trampoline()

We cannot overwrite a register which is not saved (ESI) because this will lead to a corruption in JIT code. The proper solution is to copy the call arguments. Signed-off-by: Tomek Grabiec tgrab...@gmail.com --- arch/x86/emit-code.c| 37 +++--