Slashdot Mirror

← Back to Domains

Domain: wfmh.org.pl

Stories and comments across the archive that link to wfmh.org.pl.

Comments · 7

  1. Major speedup for FP-math intensive apps. by Thorgal · · Score: 2, Informative · on Java SE 6 Released
    What I don't see mentioned here is a significant performance improvement that is especially pronounced in FP-math intensive code. Take a look at some of the timings obtained with JatMark benchmark - typically it finishes in half the time (results in seconds):
    • 206 - Core 2 Duo E6600 (2.4GHz)/DDR2@667Mhz/Linux 2.6.18.3 #5 SMP/Sun JDK 1.6.0-rc-b104
    • 223 - Opteron 275 2.2GHz/Linux 2.6.11.12-grsec/Sun JDK 1.6.0-beta-b59g
    • 349 - Pentium M 730 (1.6Ghz)/DDR2@400Mhz/Linux 2.6.15-cK1/Sun JDK 1.6.0-beta-b59g
    • 401 - Core 2 Duo E6600 (2.4GHz)/DDR2@667Mhz/Linux 2.6.18.3 #5 SMP/Sun JDK 1.5.0_05-b05
    • 408 - Opteron 275 2.2GHz/Linux 2.6.11.12-grsec/Sun JDK 1.5.0_01-b08
    • 415 - Opteron 250 2.4GHz (dual)/Linux 2.6.8/Sun JDK 1.5.0-b63
    • 596 - Pentium M 730 (1.6Ghz)/DDR2@400Mhz/Linux 2.6.15-cK1/Sun JDK 1.5.0_06-b05
    Full results table.
  2. Major speedup for FP-math intensive apps. by Thorgal · · Score: 2, Informative · on Java SE 6 Released
    What I don't see mentioned here is a significant performance improvement that is especially pronounced in FP-math intensive code. Take a look at some of the timings obtained with JatMark benchmark - typically it finishes in half the time (results in seconds):
    • 206 - Core 2 Duo E6600 (2.4GHz)/DDR2@667Mhz/Linux 2.6.18.3 #5 SMP/Sun JDK 1.6.0-rc-b104
    • 223 - Opteron 275 2.2GHz/Linux 2.6.11.12-grsec/Sun JDK 1.6.0-beta-b59g
    • 349 - Pentium M 730 (1.6Ghz)/DDR2@400Mhz/Linux 2.6.15-cK1/Sun JDK 1.6.0-beta-b59g
    • 401 - Core 2 Duo E6600 (2.4GHz)/DDR2@667Mhz/Linux 2.6.18.3 #5 SMP/Sun JDK 1.5.0_05-b05
    • 408 - Opteron 275 2.2GHz/Linux 2.6.11.12-grsec/Sun JDK 1.5.0_01-b08
    • 415 - Opteron 250 2.4GHz (dual)/Linux 2.6.8/Sun JDK 1.5.0-b63
    • 596 - Pentium M 730 (1.6Ghz)/DDR2@400Mhz/Linux 2.6.15-cK1/Sun JDK 1.5.0_06-b05
    Full results table.
  3. Re:Java eh? by Thorgal · · Score: 1 · on Can You Spare A Few Trillion Cycles?
    This isn't exactly true. As the author of global illumination renderer I witnessed many situations where FP code in Java just sucks. This is especially the case for trigonometrical operations, where implementations in JDKs >= 1.4 return bit-for-bit identical results on all platforms. This is achieved with math library, thus wasting the potential of direct computation with FPU. This is also the reason why some half-baked benchmarks concluded that with JDK 1.4 Java become much slower, while in fact it is definitely faster save for these FP operations.

    BTW, this is a good time to plug my Java app benchmark based on the renderer I wrote: JatMark.

  4. Java scientific computing performance benchmark by Thorgal · · Score: 1 · on Performance Benchmarks of Nine Languages

    While we are at it, you may wish to check Java application benchmark I wrote some time ago. JatMark

  5. Re:Frosted glass by Thorgal · · Score: 1 · on AP reports on renewed "Browser War"

    here it goes

  6. Re:Freezing doesn't truly burst cells... by Thorgal · · Score: 1 · on Freeze Recovery Drug - Step Toward Suspended Animation?
    Mirror for Europeans: http://wfmh.org.pl/enginesofcreation/.

    Milek
    --

  7. Illusions by Thorgal · · Score: 1 · on Seeking a Ghost via Web Cam

    Check out the entry for pareidolia in "The Skeptic's Dictionary" (http://www.skepdic.com/pareidol.html ) to learn more about why people see ghosts in a bunch of blurry pixels. You may also wish to visit my page, which deals specifically with such images on the Moon (http://wfmh.org.pl/~thorgal/Moon/). Enjoy.