The project will compare the runtime performance of Java and C on Palm OS device. I like to perform series of numerical computation and memory

Transcription

1 š œž Ÿ Ÿ š 4 ª «š Ÿ m œ; ± ² ³ œ µ ¹ ; º» 4.0/214365*798;:;<9=;>5*7?1;@A7?BC<ED;:AFG1AFH5*BI8 JLKNMPO?QSR JTVUXWSR YZ[SZ O\JP]L^LZ_`]aMPRNObMNY*cZK?de_PO6QPWLY*MNY Z_P].0fg=;<9BC>Nhi<91;8414j;=;<9=;8AF JLKNMPO?QSR JTLk MLlNMmQPcnZ OoZYZlpJ\YqaQaJblN^!rak MslpM\tXuev*JLdY.EwID4=47yxXz{fg ~}; w 7o5 fg1;>- ƒ1;84 xm?=;>84=;@ f 1;> B4 ˆ -5*7o5*1I@~ 4=;847 Š;<Œ1;>-?3.0fg=41436:;>`= 58E=oŽ =47 :AF5*BI8 F5<9= 1;8 <9=;<ŒBI>ah0:43b14j;= #$ -!" The project will compare the runtime performance of Java and C on Palm OS device. I like to perform series of numerical computation and memory management benchmark test with Java and C in resource limited embedded system. If time allows I also like to run some of official Java benchmark test from such as SpecJVM09 and SciMark. I will concentrate on the execution speed and memory usage of Java and C programs.

2 . / /;: E^L3E_SUIJY SUP`HJIBA3IJKaP]QbA3?JcC*PRdOEe?JQJQJK C*fg?TSC*PRWOE h1ikjol\monqp*r9sutsgvv8wuxzy {zsgm\}k{g~ u}k s ƒ yt{@v ytts >=@?BA3?DC*Ê IO?OE^L SUPŠK*IO?JcWX?JWOHVQOPOFeIJcŒ -djk h1 ŽpMƒ x1v s_ y tap* kr }kr h1ikt}_ 9 ytr p {z~šsbwzs {z~ksu{@ h1 œ}zv Gp {^ žsgtr }@ ŸpMy { #$ -!" Java is already used widely for many real-world applications because its fast development cycle and easy integration. And PDA and Wireless devices are becoming popular, so Java should be considered for development on those devices. But I like to find if Java s performances are acceptable or not on those resource limited PDA devices.

3 . / /;: hjiekml>npo qr smtvuwtyxz {}t~- ƒ ' nzˆ ŠgŒy Š Ž p m V šm Ž œÿža - jÿ nª~ «n G ²± ³µ o n o ¹»º; z r tys l¼µm½ ~¾µ µc ~ÀzgsgÁmµˆ ÂÄÃÄÅÇÆÈaÆcÃÊÉ»Ë ÌŸÍ ÎjÌŸÍ Ï ÐÒÑIÓGÃÔ Õ Ö²ÖØ >Ù Ú ÃÛÙ»ÜIÈ ÝßÞIÚyÚ¾ÍÙGÚÛàTÙGÚ@ÑVáNâãÞIäIäVå*Í_æmÞ Ï Í_ÙÛÎ1ÜIÈeçaÈVåÙÛäIèjÈVÎ Ï #$ -!" Here are some of the popular Java Virtual Machine available for some of popular PDA and Wireless devices. Different embedded JVM has its own specification and gives different performance output on the PDA devices.

4 .0/ /98 : ;=<7>@? ACBEDGFIHJDKFIL7MEN O!PQNSRTPVUXWZY[WCW\HJ]_^Kà^@b9b c[d\egf D e-h W\HJi d ijd hcc W e M*FkDlHCW\H[]mW\H[W\npo[FIM*F qsrtvuwt9xvyz {j}~ti * xi} ta~ƒ xj tkˆ~ { ˆ ŠŒ{ZŠŒxV Ž ˆ ktk} { ti k ~ ti { ˆ ~ xix xv X q9 Vš ~ *Šœ I xku {7Š {7 X~ž}vtI 2yk{2Š t I{ŸyvŽ škˆ I *Š ~- u { ~TŽ k{ ˆ Xˆ ~ {Itk x \r tvuwt9xvyz { }'~ q9 Kˆ {œti g ªtvŽ * kˆ~ { tk Ÿx ZzgtvuwtZ«ša~ƒ S «* { }'~ x qs Ju x ± }² { tx~³ * j X{ x yzg{ }'~ * Ÿ t7 w{ ˆx ~ µti!{ #$ " %'& &*)+, -!" Dennis Sosonoski carried out series of experiment on comparing Java and C performance. His results showed that Java performance can be improved if Java program is carefully written such that use Java primitive types as much as you can, avoid using java.util.vector, avoid creating new object whenever you can.

5 .0/ /98 : ;=<7>@? AB;=C65ED FHGJILKNMPOKRQSMUTVQPWXZY\[^]L_P`bacMPK-O*deQPWfdeIZaSgPÙThQUTO*IZ_ i9jlknm o-pqup*rtsuuwvxpyuu{znm{r9zwo} ~p z qnq p o-pƒyfu znu{q r k m o-p* nm*p ˆzwo}pyUu iv zxš zœ E ^ zxž p*u k7 v y Š{s7 ^ snz q y ^k7uxo-p*rts s{ pšuw œ iv zxš z r knm o-pqup*rts7u{vxpyuu{znm zn ^ zxž pvlzn ~ žzxžjy zn ^ zxž{vlÿrvˆmye s Ep u q sw pšu{ n iv zxš z9yuk o} s yf ~r vv p ªzxŠ z zn ^ zxž2 Rk uwo}p*r9s s{ pšuw 0pvVq pvz «nms{q #$ " %'& &*)+, -!" Moreira et al. carried out series of matrix computation experiment for Java, C and FORTRAN. FORTAN and C clearly outperforms Java in the experiment. But Java performance can be improve if Java array runtime checking is disabled.

6 1FE1HG?JI 1A:2G 9LKM58NO7 3 PQE9SR>14TU5 VXW Y[Z*\]Y_^*Ùa6bScD`6bUdeYD`6fhg6djikikdUl4mDdUlonqprlstÙa6\ud pvaxwjù^s yfz{wu ~} Ud] [Z*\ud6 ƒ `] D` pvl< <ˆ V knš Z*f g6djiœikdulq BY p6ž gd`6 Z*f;ƒ `] D` *f mddulonqprl-sxùa6\ud praxwu ~}HˆM -f ZiL`6\u\udUmjikÙgU^*d mddulonqprl-sxùa6\uduˆ VXW YD`jiQZs mulqpd DdUs duaji f \uùa_g6dest`6ud ikp ƒ `] D` ikp lc[aegddjiœi-dul<praxwu ~}Hˆ #$ -!" Most of the existing works compared Java and C/C++ performance on Unix or Window machines. My experiment will expand on those existing works by carrying out the similar benchmark test in embedded system such as Palm OS device. Hopefully I can answer these questions by end of the project.

7 .0/213/546/5798:/<; = [1] Dennis M. Sosnoski. Java performance programming, Part 1: Smart object-management saves the day. Java World, November performance.html [2] Dennis M. Sosnoski. Java Performance Comparison with C/C++. Sosnoski Software Solution, Inc. This report was presented in 1999 JavaOne conference. [3] J. E. Moreira, S. P. Midkiff, and M. Gupta. A Comparison of Java, C/C++, and FORTRAN for Numerical Computing. IEEE Antennas and Propagation Magazine, Vol. 40, No. 5, October #$ -! "