It's tough (and pretty stupid, IMHO) to compare them side-by-side:
DRLVM has way more options to swizzle than RVM, and some of them are
low-level ones. But anyway:

DRLVM (206 options to swizzle):
9317 runs, 5716 failures -> 61% failure rate

RVM (89 options to swizzle):
13721 runs, 4584 failures -> 33% failure rate

One could say that RVM is more reliable, but again, that could be the delusion.

Thanks,
Aleksey.

On Mon, Nov 24, 2008 at 10:49 AM, Alexei Fedotov
<[EMAIL PROTECTED]> wrote:
> Aleksey,
> Nice work! How robust was DRLVM compared to Jikes?
>
> Thanks!
>
> On Mon, Nov 24, 2008 at 9:18 AM, Aleksey Shipilev
> <[EMAIL PROTECTED]> wrote:
>> Hi,
>>
>> On Mon, Nov 24, 2008 at 4:07 AM, Xiao-Feng Li <[EMAIL PROTECTED]> wrote:
>>> Interesting and useful work!  Surely the confs deserve careful 
>>> investigations.
>> Yep. What I don't know yet, is how exactly to tackle all that stuff...
>> What I know for sure is, the verification of issues have to be
>> automated.
>>
>> On Mon, Nov 24, 2008 at 9:03 AM, xiaoming gu <[EMAIL PROTECTED]> wrote:
>>> Aleksey, what is "GA"? Thanks. -Xiaoming
>> GA == Genetic Algorithm.
>>
>> Thanks,
>> Aleksey.
>>
>
>
>
> --
> С уважением,
> Алексей Федотов,
> ЗАО «Телеком Экспресс»
>

Reply via email to