From owner-freebsd-java Tue Sep 5 6:43:54 2000 Delivered-To: freebsd-java@freebsd.org Received: from cafe.muraoka.info.waseda.ac.jp (cafe.muraoka.info.waseda.ac.jp [133.9.68.49]) by hub.freebsd.org (Postfix) with ESMTP id 2BB7A37B43E for ; Tue, 5 Sep 2000 06:43:50 -0700 (PDT) Received: from muraoka.info.waseda.ac.jp (shudoh@localhost [127.0.0.1]) by cafe.muraoka.info.waseda.ac.jp (8.9.1a/3.7W) with ESMTP id WAA12926; Tue, 5 Sep 2000 22:43:28 +0900 Message-Id: <200009051343.WAA12926@cafe.muraoka.info.waseda.ac.jp> To: freebsd-java@freebsd.org Cc: shudoh@muraoka.info.waseda.ac.jp Subject: shuJIT 0.6.5 released Date: Tue, 05 Sep 2000 22:43:27 +0900 From: SHUDO Kazuyuki Sender: owner-freebsd-java@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org Hi all, Now I've released shuJIT 0.6.5. Once, a year ago, I've announced the release of shuJIT 0.3.2. | To: freebsd-java@FreeBSD.ORG | Subject: shuJIT 0.3.2 relased | Date: Sun, 01 Aug 1999 17:39:29 +0900 If you're interested in the JIT, please see http://www.shudo.net/jit/. It is a JIT compiler for Sun classic VM / FreeBSD and Linux / x86. I've made comparisons between OpenJIT 1.1.14, TYA 1.7v2 and shuJIT 0.6.5 in their peak performance. They are JIT compilers which can be used on FreeBSD, Linux and x86. The results were produced on the Linux box, but these JIT compilers can work with FreeBSD JDK 1.1.8 and 1.2.2. So I believe that these results are valuable for FreeBSD users. [SPEC JVM98] http://www.spec.org/osg/jvm98/ This has been used as the standard benchmark suite to evaluate JVMs and JITs. Benchmark programs in ths suite are not toy programs, rather real applications like javac, JavaCC and MPEG audio layer 3 (MP3) decoder. I adopted the following command line options because the machine has a little memory (64MB): % appletviewer -J-Xmx32m -J-Xms32m http://... These measurements were performed on K6-2 400 MHz and Blackdown JDK 1.2.2 FCS. * results (second: lower is better) OpenJIT shuJIT TYA interpreter _201_compress 114.529 143.152 153.195 596.893 _202_jess 44.584 57.459 56.785 141.738 _209_db 150.592 129.824 148.438 295.294 _213_javac 98.65 124.729 115.22 213.525 _222_mpegaudio 99.409 110.923 125.69 502.524 _227_mtrt 56.582 85.534 79.384 177.097 _228_jack 52.637 69.653 68.288 194.796 * results (score: higher is better) OpenJIT shuJIT TYA interpreter _227_mtrt 8.13 5.38 5.79 2.60 _202_jess 8.52 6.61 6.69 2.68 _201_compress 10.3 8.21 7.67 1.97 _209_db 3.35 3.89 3.39 1.71 _222_mpegaudio 11.1 9.92 8.75 2.19 _228_jack 8.64 6.53 6.66 2.34 _213_javac 4.31 3.41 3.69 1.99 Geometric Mean 7.18 5.91 5.79 2.19 [CaffeineMark 3.0] http://www.pendragon-software.com/pendragon/cm3/ Recently, small benchmarks in this suite is not used to measure performance because the results do not reflect a feeling in using JVMs. But, it can be useful for me in examining the result of the improvement of a JIT compiler. I used Pentium II 333 MHz and Blackdown JDK 1.1.8v1 for this benchmark. * results (score: higher is better) OpenJIT shuJIT TYA Sieve 3224 1975 1911 Loop 4647 3323 2683 Logic 6869 6864 4958 String 4265 2521 2305 Float 3139 2505 1765 Method 4681 1826 1823 Graphics 336 343 360 Image 534 526 533 Dialog 133 107 129 CM3 1749 1291 1186 [Linpack benchmark] http://www.netlib.org/benchmark/linpackjava/ This benchmark performs solution of linear equations. The method of solution is base on Gaussian elimination. This benchmark is small and runs in a short time. Therefore JIT compilers (including IBM's) which adopt adaptive compilation techniques cannot make a good score. But, JIT compilers evaluated this time do not use those techniques in their default settings. The followings were made on Pentium II 333 MHz and Blackdown JDK 1.2.2 FCS. OpenJIT 1.1.13 13.464 shuJIT 0.6.5 13.205 TYA 1.7v2 11.444 OpenJIT 1.1.14 10.899 Any comment, suggestion or report is appreciated. Kazuyuki SHUDO Happy Hacking! Muraoka Lab., Grad. School of Sci. & Eng., Waseda Univ. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-java" in the body of the message