On 12-Apr-06, at 3:07 AM, Mark Wielaard wrote:
On Tue, 2006-04-11 at 15:07 -0400, Bryce McKinlay wrote:
With GCJ and a simple benchmark, I get around a 1.5-3x performance
improvement depending on the String length. No mauve regressions.
According to the autobuilder (and I just checked locally
> "Mark" == Mark Wielaard <[EMAIL PROTECTED]> writes:
Mark> FAIL: gnu.testlet.java.io.ObjectInputOutput.OutputTest: Serializable:
Mark> gnu.testlet.java.io.ObjectInputOutput.Test$Extern () (number 2)
Mark> FAIL: gnu.testlet.java.text.DecimalFormatSymbols.serial (number 1)
The new code wasn't
Hi Bryce,
On Tue, 2006-04-11 at 15:07 -0400, Bryce McKinlay wrote:
> With GCJ and a simple benchmark, I get around a 1.5-3x performance
> improvement depending on the String length. No mauve regressions.
According to the autobuilder (and I just checked locally against mauve)
this does trigger 2
writeUTF() is heavily used by serialization/RMI and appears fairly
prominently on profiles of applications that use these. This patch
improves its performance in several ways:
- Instead of allocating a new, variable length byte[] buffer every time
writeUTF() is called, now use a fixed-length b
