First, we experiment with HPJava on a simple Laplace Equation with red-black relaxation on the Sun Solaris 9 with 8 UltraSPARC III Cu 900MHz Processors and 16GB of main memory. Figure 7 shows the result of five different versions (HPJava with HPJOPT2 optimization, HPJava with PRE optimization, HPJava with naive translation, Java, and C) of red-black relaxation of the two dimensional Laplace equation. After applying HPJOPT2 for the naive translation, the speedup of HPJava is 177% on a single processor and 138% on 8 processors.
Second, The results of our benchmarks use an IBM SP3 running with four Power3 375MHz CPUs and 2GB of memory on each node. This machine uses AIX version 4.3 operating system and the IBM Developer Kit 1.3.1 (JIT) for the Java system. We are using the shared ``css0'' adapter with User Space (US) communication mode for MPI setting and -O compiler flag for Java. For comparison, we have also completed experiments for sequential Java, Fortran and HPF version of the HPJava programs. For the HPF version of program, we use IBM XL HPF version 1.4 with xlhpf95 compiler command and -O3 and -qhot flag. And XL Fortran for AIX with -O5 flag is used for Fortran version.
Figure 8 shows the results from four different versions (HPJava, sequential Java, HPF and Fortran) of red-black relaxation for the two dimensional Laplace equation with size of 512 by 512. In our runs parallel HPJava can out-perform sequential Java by up to 17 times. On 36 processors HPJava can get about 78% of the performance of HPF. This is satisfactory performance for the initial benchmark result. Scaling behavior of HPJava is slightly better than HPF. Probably this mainly reflects the low performance of a single Java node compared to Fortran6.