WebKit's New JavaScript Performance

Platform: PPC, PowerMac G4 2x867MHz, 2GB RAM, Mac OS X 10.4.11, Safari 3 web browser.

Below are each build's SunSpider benchmark results -- built locally, but no special build flags.

This is revision 36535, 2008-09-16...

http://www2.webkit.org/perf/sunspider-0.9/sunspider-results.html?%7B%223d-cube%22:%5B3020,2115,1647,2118,2115%5D,%223d-morph%22:%5B1176,1149,1140,1155,1151%5D,%223d-raytrace%22:%5B1344,1045,1064,1326,1339%5D,%22access-binary-trees%22:%5B193,135,132,136,131%5D,%22access-fannkuch%22:%5B238,236,236,236,237%5D,%22access-nbody%22:%5B3078,2361,2494,3040,3178%5D,%22access-nsieve%22:%5B90,85,90,83,97%5D,%22bitops-3bit-bits-in-byte%22:%5B91,83,89,81,85%5D,%22bitops-bits-in-byte%22:%5B116,115,121,115,118%5D,%22bitops-bitwise-and%22:%5B139,138,137,139,144%5D,%22bitops-nsieve-bits%22:%5B504,417,425,483,522%5D,%22controlflow-recursive%22:%5B81,72,73,75,84%5D,%22crypto-aes%22:%5B124,118,116,115,115%5D,%22crypto-md5%22:%5B384,327,327,350,411%5D,%22crypto-sha1%22:%5B334,346,337,337,362%5D,%22date-format-tofte%22:%5B271,256,264,257,261%5D,%22date-format-xparb%22:%5B408,417,400,403,410%5D,%22math-cordic%22:%5B1418,1102,1126,1386,1561%5D,%22math-partial-sums%22:%5B1761,1372,1395,1722,2215%5D,%22math-spectral-norm%22:%5B826,668,687,834,864%5D,%22regexp-dna%22:%5B1173,1126,1130,1133,1194%5D,%22string-base64%22:%5B205,189,209,212,217%5D,%22string-fasta%22:%5B435,409,421,434,449%5D,%22string-tagcloud%22:%5B556,544,531,525,589%5D,%22string-unpack-code%22:%5B452,438,449,439,422%5D,%22string-validate-input%22:%5B453,413,442,438,442%5D%7D


This is revision 36738, 2008-09-21...

http://www2.webkit.org/perf/sunspider-0.9/sunspider-results.html?%7B%223d-cube%22:%5B401,403,400,399,397%5D,%223d-morph%22:%5B509,510,510,506,505%5D,%223d-raytrace%22:%5B324,366,349,356,333%5D,%22access-binary-trees%22:%5B110,111,124,113,113%5D,%22access-fannkuch%22:%5B246,237,235,236,238%5D,%22access-nbody%22:%5B405,452,478,456,452%5D,%22access-nsieve%22:%5B86,82,82,85,85%5D,%22bitops-3bit-bits-in-byte%22:%5B87,87,87,88,88%5D,%22bitops-bits-in-byte%22:%5B118,118,117,117,117%5D,%22bitops-bitwise-and%22:%5B137,137,140,136,140%5D,%22bitops-nsieve-bits%22:%5B190,196,206,195,207%5D,%22controlflow-recursive%22:%5B78,78,77,77,77%5D,%22crypto-aes%22:%5B128,129,124,124,124%5D,%22crypto-md5%22:%5B130,129,134,135,127%5D,%22crypto-sha1%22:%5B132,131,131,131,133%5D,%22date-format-tofte%22:%5B232,232,233,244,247%5D,%22date-format-xparb%22:%5B325,321,321,316,319%5D,%22math-cordic%22:%5B334,328,328,328,334%5D,%22math-partial-sums%22:%5B402,418,405,408,404%5D,%22math-spectral-norm%22:%5B209,211,223,212,211%5D,%22regexp-dna%22:%5B1092,1067,1078,1069,1066%5D,%22string-base64%22:%5B157,158,152,156,154%5D,%22string-fasta%22:%5B268,267,272,262,270%5D,%22string-tagcloud%22:%5B480,475,468,463,466%5D,%22string-unpack-code%22:%5B425,426,434,440,432%5D,%22string-validate-input%22:%5B327,322,328,328,330%5D%7D


TEST COMPARISON FROM TO DETAILS

=============================================================================

** TOTAL **: 2.34x as fast 17262.6ms +/- 11.7% 7381.6ms +/- 0.6% significant

=============================================================================

3d: 3.65x as fast 4580.8ms +/- 16.8% 1253.6ms +/- 1.9% significant
cube: 5.51x as fast 2203.0ms +/- 28.2% 400.0ms +/- 0.7% significant
morph: 2.27x as fast 1154.2ms +/- 1.4% 508.0ms +/- 0.6% significant
raytrace: 3.54x as fast 1223.6ms +/- 15.7% 345.6ms +/- 6.1% significant

access: 3.73x as fast 3301.2ms +/- 14.6% 885.2ms +/- 3.6% significant
binary-trees: 1.27x as fast 145.4ms +/- 22.8% 114.2ms +/- 6.1% significant
fannkuch: ?? 236.6ms +/- 0.5% 238.4ms +/- 2.3% not conclusive: might be *1.01x as slow*
nbody: 6.31x as fast 2830.2ms +/- 16.4% 448.6ms +/- 7.4% significant
nsieve: - 89.0ms +/- 7.6% 84.0ms +/- 2.8%

bitops: 1.50x as fast 812.4ms +/- 7.6% 541.6ms +/- 2.0% significant
3bit-bits-in-byte: ?? 85.8ms +/- 6.0% 87.4ms +/- 0.8% not conclusive: might be *1.02x as slow*
bits-in-byte: ?? 117.0ms +/- 2.7% 117.4ms +/- 0.6% not conclusive: might be *1.00x as slow*
bitwise-and: - 139.4ms +/- 2.4% 138.0ms +/- 1.7%
nsieve-bits: 2.37x as fast 470.2ms +/- 12.4% 198.8ms +/- 4.6% significant

controlflow: ?? 77.0ms +/- 8.5% 77.4ms +/- 0.9% not conclusive: might be *1.01x as slow*
recursive: ?? 77.0ms +/- 8.5% 77.4ms +/- 0.9% not conclusive: might be *1.01x as slow*

crypto: 2.11x as fast 820.6ms +/- 6.7% 388.4ms +/- 0.8% significant
aes: *1.07x as slow* 117.6ms +/- 4.0% 125.8ms +/- 2.5% significant
md5: 2.75x as fast 359.8ms +/- 12.8% 131.0ms +/- 3.2% significant
sha1: 2.61x as fast 343.2ms +/- 4.1% 131.6ms +/- 0.8% significant

date: 1.20x as fast 669.4ms +/- 1.4% 558.0ms +/- 1.2% significant
format-tofte: 1.10x as fast 261.8ms +/- 2.9% 237.6ms +/- 3.8% significant
format-xparb: 1.27x as fast 407.6ms +/- 2.0% 320.4ms +/- 1.3% significant

math: 3.98x as fast 3787.4ms +/- 20.4% 951.0ms +/- 0.7% significant
cordic: 3.99x as fast 1318.6ms +/- 18.7% 330.4ms +/- 1.2% significant
partial-sums: 4.16x as fast 1693.0ms +/- 25.2% 407.4ms +/- 1.9% significant
spectral-norm: 3.64x as fast 775.8ms +/- 14.6% 213.2ms +/- 3.3% significant

regexp: 1.07x as fast 1151.2ms +/- 3.3% 1074.4ms +/- 1.3% significant
dna: 1.07x as fast 1151.2ms +/- 3.3% 1074.4ms +/- 1.3% significant

string: 1.25x as fast 2062.6ms +/- 3.0% 1652.0ms +/- 0.3% significant
base64: 1.33x as fast 206.4ms +/- 6.4% 155.4ms +/- 1.9% significant
fasta: 1.60x as fast 429.6ms +/- 4.4% 267.8ms +/- 1.7% significant
tagcloud: 1.17x as fast 549.0ms +/- 5.7% 470.4ms +/- 1.8% significant
unpack-code: - 440.0ms +/- 3.3% 431.4ms +/- 1.8%
validate-input: 1.34x as fast 437.6ms +/- 4.2% 327.0ms +/- 1.1% significant

Worth A Try

Simson Garfinkel blogs about getting hard disk drives working by putting them in your kitchen freezer.

I've worked this trick since the mid-1990s. Sometimes it works, sometimes it doesn't; but the price is right.

I always place the drive in a "zip-lock" bag — get as much air out as possible. Any moister in the air will freeze, of course.

Hard drives are designed to passively move heat. I have always felt that an hour or two was long enough. Then, I try mounting the drive while it is still freezer cold, but I also try as (and after) it has warmed back to room temperatures.

When a hard drive is obsolete or beyond repair they are always good for a few rare-earth magnets!

Please Sir, May I Have Some More?

No one likes a whiner; therefore I haven't written to my blog lately.

Here are a few quick list of things happening right now:

1. After nine years working independently, I'm looking for a full-time salary job, wish me luck!
2. I've been doing volunteer work for WebKit. WebKit is the web engine behind Apple's Safari web browser.
3. PT Wetter 3 is actively remaking his to-do system, Frictionless. I will be trying this out tonight.
4. I have been fixing-up my 1993 Volvo 850 GLT sedan. It's got 179,000 miles and going strong.
5. Now that it is past 15 May it's Summer in this part of Arizona.

Optimize Mail.app Database

I found Hawk Wing's blog entry very notable this morning, since I use Apple's Mail.app all the time. This tip optimizes the Mail.app "Envelope" database.

From my Terminal.app:
% ls -lah ~/Library/Mail/Envelope\ Index
-rw-r--r-- 1 user admin 32M Mar 1 17:23 /Users/user/Library/Mail/Envelope Index
% sqlite3 ~/Library/Mail/Envelope\ Index vacuum;
% ls -lah ~/Library/Mail/Envelope\ Index
-rw-r--r-- 1 user admin 26M Mar 2 08:07 /Users/user/Library/Mail/Envelope Index
% exit

From http://www.sqlite.org/lang_vacuum.html:

SQL As Understood By SQLite

VACUUM

sql-statement ::= VACUUM [index-or-table-name]

The VACUUM command is an SQLite extension modeled after a similar command found in PostgreSQL. If VACUUM is invoked with the name of a table or index then it is suppose to clean up the named table or index. In version 1.0 of SQLite, the VACUUM command would invoke gdbm_reorganize() to clean up the backend database file.

VACUUM became a no-op when the GDBM backend was removed from SQLITE in version 2.0.0. VACUUM was reimplemented in version 2.8.1. The index or table name argument is now ignored.

When an object (table, index, or trigger) is dropped from the database, it leaves behind empty space. This makes the database file larger than it needs to be, but can speed up inserts. In time inserts and deletes can leave the database file structure fragmented, which slows down disk access to the database contents. The VACUUM command cleans the main database by copying its contents to a temporary database file and reloading the original database file from the copy. This eliminates free pages, aligns table data to be contiguous, and otherwise cleans up the database file structure. It is not possible to perform the same process on an attached database file.

This command will fail if there is an active transaction. This command has no effect on an in-memory database.

As of SQLite version 3.1, an alternative to using the VACUUM command is auto-vacuum mode, enabled using the auto_vacuum pragma.

This page last modified on 2007/02/13 02:03:25

Before You Trash It

Before disposing of an Apple Performa 6400 or Power Mac 6400 or Power Mac 6500 you will want to pull the really unique speaker (subwoofer) from the case. Collect two of them and you might really have something.

Getting into the case is not straight forward. They don't build 'em like they used to, which is good in this instance. Skipping one of the following steps will not speed things up!

1. Remove the front façade, starting at the bottom. Release the two bottom tabs and pull very hard.
2. Pry-off the small front façade, staring at its side-seams.
3. Remove the screw from the top-back; slide the top façade off by pushing to the back.
4. Lightly pry-up the 2 centimeter tab molded into the plastic panel's front edge on the computer's right side façade. Now slide this panel off  by pushing to the back.
5. Remove the nine Philips-head screws from the steel side panel.
6. Remove the side panel by sliding it to the rear about 1 centimeter and then lift up.
7. Remove the four T-10 screws from the bottom. These have rubber bushings that are worth saving too.
8. Release the red and black speaker lead wires. It is attached, by just friction, to the small printed circuit board (PCB) near the top-rear of the computer.
9. Remove the speaker and fish out the lead wires.
10. Done.