[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Bug in Double.java
From: |
Tom Tromey |
Subject: |
Re: Bug in Double.java |
Date: |
23 Jul 2001 14:35:21 -0600 |
>>>>> "Eric" == Eric Blake <address@hidden> writes:
Eric> I also spotted some optimizations to add. I made the equals,
Eric> compare, and isNaN work without needing to use
Eric> doubleToLongBits(), since a native call carries overhead.
I think we definitely need to implement some kind of preprocessing
step for Classpath and libgcj. The above reasoning is exactly
backwards for gcj.
Eric> I also updated hashCode to use a cache, to reduce the number of
Eric> native calls made by the same Double.
This one too. For libgcj this one probably makes things even less
efficient, since it increases the size of a Double.
In this case I guess we can just do as Mark suggests and put stuff
into Configuration. Then the compiler will remove the unused branch.
I think I'd prefer not to have one big `GCJ' configuration name
though. For this one, maybe `FAST_NATIVE_CALLS' or `CNI' is what we
should use. Comments?
Tom