# C Programma # bin/fluid --initial --width 101 --batch -jacobi --max 100 --iters100 On Desire ------------------------------------------------------------ # iterations # C Runtime with Gauss-Seidel (ms) # RMS error with Gauss-Seidel # C Runtime with Jacobi (ms) # RMS error with Jacobi 10 470 418485 230 506213 20 880 313370 400 411765 30 1290 256805 570 350927 40 1770 218158 730 309038 50 2110 189868 900 277747 100 4190 113576 1740 187573 Width 64 # iters, C-GS(ms), C-JC(ms), Repa-JC-N1, N2, N3, N4 10 200 110 20 350 170 480 430 410 470 30 510 230 40 680 300 870 790 750 880 50 840 370 Width 128 # iters, C-GS(ms), C-JC(ms), Repa-JC-N1, N2, N3, N4 10 780 400 20 1440 670 1190 880 790 860 30 2100 930 40 2760 1200 1960 1550 1370 1550 50 3460 1480 Width 256 # iters, C-GS(ms), C-JC(ms), Repa-JC-N1, N2, N3, N4 10 3250 2300 20 5970 3990 4040 2630 2030 2230 30 8790 5690 40 11420 7390 6530 4180 3420 3780 50 14180 9150 Width 512 # iters, C-GS(ms), C-JC(ms), Repa-JC-N1, N2, N3, N4 10 14200 18430 20 25290 33340 13670 8370 6710 6870 30 40 47830 62560 22280 13820 10870 50 On Tesla ------------------------------------------------------------ # --width 101, --max 100 # iters, C-GS, C-Jacobi, Repa-Jacobi-N1 N2 N4 N8 10 760 350 20 1410 590 30 2070 820 40 2720 1060 50 3380 1290 100 6650 2470 # --width 64 # iters, C-GS(ms), C-JC(ms), Repa-JC-N1, N2, N4, 8 10 20 570 260 780 760 790 770 30 40 1080 470 1380 1370 1830 1430 50 # --width 96 # iters, C-GS(ms), C-JC(ms), Repa-JC-N1, N2, N4, 8 10 20 1280 560 1630 1130 1070 1020 30 40 2460 1020 2860 1990 1830 1890 50 # --width 128 # iters, C-GS(ms), C-JC(ms), Repa-JC-N1, N2, N4, N8 10 1210 580 20 2270 980 30 3330 1390 40 4380 1800 4040 3370 2220 2300 50 5440 2210 # --width 192 # iters, C-GS(ms), C-JC(ms), Repa-JC-N1, N2, N4, N8 10 20 5190 3180 30 40 10010 5950 7000 5600 3710 4010 50 # --width 256 # High variance with more threads. # iters, C-GS(ms), C-JC(ms), Repa-JC-N1, N2, N4, N8 10 5040 3890 20 9360 6980 6300 4170 3250 2990 30 13680 10080 40 18010 13180 9630 6410 5590 5150 50 22320 16280 # --width 384 # High variance with more threads. # iters, C-GS(ms), C-JC(ms), Repa-JC-N1, N2, N4, N8 10 20 22540 30 40 36490 19570 13060 10360 9310 50 # --width 512 # iters, C-GS(ms), C-JC(ms), Repa-JC-N1, N2, N4, N8 10 20 41420 40640 21560 13160 9960 8600 30 40 78960 73480 34050 21270 16630 15040 50 # --width 1024 # iters, C-GS(ms), C-JC(ms), Repa-JC-N1, N2, N4, N8 10 20 225120 30 40 559290 145260 89560 50