United StatesChange Country, Oracle Worldwide Web Sites Communities I am a... I want to...
Bug ID: 6313320 Performance regression of String.toUpperCase() and toLowerCase()
6313320 : Performance regression of String.toUpperCase() and toLowerCase()

Details
Type:
Bug
Submit Date:
2005-08-19
Status:
Resolved
Updated Date:
2010-12-03
Project Name:
JDK
Resolved Date:
2005-10-10
Component:
core-libs
OS:
solaris
Sub-Component:
java.lang
CPU:
x86
Priority:
P3
Resolution:
Fixed
Affected Versions:
6
Fixed Versions:
6

Related Reports
Backport:
JDK-2130788 - Performance regression of String.toUpperCase() and toLowerCase()
Relates:
JDK-4125644 - RFE: Improve locale handling in String.toLowerCase/toUpperCase
Relates:
JDK-6265809 - Huge String.toLowerCase() performance regression

Sub Tasks

Description
Even after the fix for 6265809, String.toUpperCase() and toLowerCase() in Mustang workspace is slower than the same methods in 1.5.0_X and much slower than 1.4.2_X. This should be fixed.

Comments
EVALUATION

Added a separate fast path for non-supplementary character case conversions.  Here are the numbers which are in seconds to perform 1000000 case conversions, so lower is better.

On Solaris 10 (Sunblade 2cpu):

               LC(NON-S)  LC(SUPPL) UC(NON-S) UC(SUPPL) 
1.4.2_08:       0.900      N/A       0.954     N/A
5.0u6:          1.513      2.667     1.427     2.601
Mustang w/fix:  0.445      0.856     0.500     1.028

On Windows XP (Pentium 4 2.5GHz, 1GB memory):

               LC(NON-S)  LC(SUPPL) UC(NON-S) UC(SUPPL) 
1.4.2_09:       0.401      N/A       0.440     N/A
5.0u4:          0.831      1.609     0.821     1.623
Mustang w/fix:  0.450      0.711     0.441     0.701 

Test cases are attached to this bug.
2005-09-22


Hardware and Software, Engineered to Work Together