I made a new kind of testing, and here are the results.
I made heavy load with the kernel-compiling (make -j), and
paralelly, I started 7 other tasks like this:
------------
#!/bin/sh
while /bin/true
do
/bin/echo -n "#"
sleep $1
done
------------
The parameter was 1 2 4 8 16 32 64
These programs echoes one symbol after every x seconds. These are
redirected to different files, and counted after finishing the
kernel-compilation has finished.
The reading of the test results:
First line: Kernel image name, user space, kernel space time, and
CPU% (like time), and the total compilation time in seconds.
Second line: The (number of #s per second * 10000 * program
parameter ) in the shell script outputs. If this number is lower, it
menas that the program had less time to put the # sign. In normal
case, every test result should be around 10000, because, the shell
script speed is 1/parameter hashmarks/sec.
I think this test can be a good test for the interactive speed.
Isn't it?
The results are a bit surprising:
I thought that the 2.2.0pre5 with arca mini patch is the best, but
these test shows a bit different results:
If you see the kernel-comiling times, the best is the 2.2.0pre6.
(the smallest 'time' int the first row of tests). If has reasonable
interactive speed also.
If you see the interactive-speed, the best is 2.2.0pre6 with
arcan-VM-15, but the kernel-comiling time is poor.
So I think the best swap-code was in the 2.2.0-pre6.
Here are the results:
p4arc10 or p4a10 means pre4 with arca-VM-20
p5-am means pre5 with Arcangeli's first mini patch.
2.0.35 : usr: 866.06s, sys: 7.00s, cpu:68%, total: 1440s
01: 2283 02: 3414 04: 4775 08: 5830 16: 6440 32: 7551 64: 8439
2.0.35 : usr: 869.29s, sys: 4.84s, cpu:67%, total: 1486s
01: 2199 02: 3350 04: 4386 08: 5489 16: 6243 32: 7104 64: 8180
2.1.131 : usr: 864.11s, sys: 4.36s, cpu:28%, total: 3964s
01: 2961 02: 3586 04: 3924 08: 4237 16: 4398 32: 2986 64: 2905
2.1.131 : usr: 852.14s, sys: 6.51s, cpu:31%, total: 3563s
01: 3031 02: 3300 04: 3951 08: 4063 16: 4266 32: 4849 64: 5747
2.2.0-p4 : usr: 843.17s, sys: 5.71s, cpu:84%, total: 1116s
01: 3582 02: 5123 04: 6771 08: 7523 16: 8311 32: 9171 64: 9171
2.2.0-p4 : usr: 847.65s, sys: 6.71s, cpu:87%, total: 1077s
01: 3646 02: 5270 04: 6977 08: 8016 16: 8907 32: 9204 64: 9501
2.2.0-p4arc10: usr: 862.64s, sys: 3.08s, cpu:72%, total: 1432s
01: 3293 02: 4647 04: 5694 08: 6587 16: 7815 32: 8932 64: 8932
2.2.0-p4arc10: usr: 842.06s, sys: 1.52s, cpu:73%, total: 1374s
01: 2998 02: 4381 04: 5501 08: 6287 16: 7452 32: 8150 64: 9315
2.2.0-p4arc7: usr: 854.62s, sys: 5.94s, cpu:72%, total: 1436s
01: 3354 02: 4287 04: 5651 08: 6570 16: 7572 32: 8241 64: 8909
2.2.0-p4arc7: usr: 855.51s, sys: 9.46s, cpu:70%, total: 1492s
01: 3150 02: 4222 04: 5334 08: 6165 16: 6862 32: 7720 64: 8578
2.2.0-p4arc8: usr: 838.73s, sys: 3.53s, cpu:88%, total: 1112s
01: 3407 02: 4872 04: 6329 08: 7336 16: 8199 32: 9206 64: 9781
2.2.0-p4arc8: usr: 858.85s, sys: 3.15s, cpu:84%, total: 1202s
01: 3184 02: 4689 04: 5886 08: 7050 16: 8248 32: 9046 64: 9046
2.2.0-p5 : usr: 839.53s, sys: 2.04s, cpu:87%, total: 1081s
01: 3003 02: 4344 04: 5804 08: 6876 16: 7837 32: 8577 64: 8872
2.2.0-p5 : usr: 837.87s, sys: 0.49s, cpu:87%, total: 1089s
01: 2963 02: 4422 04: 5835 08: 6899 16: 7780 32: 8514 64: 8807
2.2.0-p5-am : usr: 855.32s, sys: 4.73s, cpu:87%, total: 1113s
01: 2793 02: 4346 04: 5604 08: 6753 16: 7759 32: 8621 64: 9196
2.2.0-p5-am : usr: 849.57s, sys: 3.98s, cpu:84%, total: 1138s
01: 3013 02: 4322 04: 5622 08: 6887 16: 7730 32: 8152 64: 8995
2.2.0-p6 : usr: 849.75s, sys: 2.82s, cpu:88%, total: 1091s
01: 3406 02: 5110 04: 6520 08: 7473 16: 8792 32: 9085 64: 9378
2.2.0-p6 : usr: 847.58s, sys: 1.34s, cpu:89%, total: 1073s
01: 3521 02: 4732 04: 6595 08: 7974 16: 8347 32: 8645 64: 9539
2.2.0-p6a15 : usr: 852.20s, sys: 5.27s, cpu:74%, total: 1359s
01: 4052 02: 5413 04: 6708 08: 7590 16: 8355 32: 8708 64: 9414
2.2.0-p6a15 : usr: 840.44s, sys: 4.30s, cpu:74%, total: 1286s
01: 3932 02: 5207 04: 6715 08: 7585 16: 8331 32: 8953 64: 8953
2.2.0-p7 : usr: 863.34s, sys: 6.69s, cpu:51%, total: 1961s
01: 2497 02: 3251 04: 3975 08: 4811 16: 5056 32: 5871 64: 7176
2.2.0-p7 : usr: 838.86s, sys: 8.44s, cpu:45%, total: 2184s
01: 2352 02: 2892 04: 3570 08: 3991 16: 4687 32: 5419 64: 7030
2.2.0-p7a18 : usr: 845.23s, sys: 9.98s, cpu:86%, total: 1108s
01: 3049 02: 4492 04: 5954 08: 6928 16: 7938 32: 8660 64: 9815
2.2.0-p7a18 : usr: 841.68s, sys: 9.09s, cpu:85%, total: 1116s
01: 2965 02: 4354 04: 5734 08: 6880 16: 7741 32: 8601 64: 9174
2.2.0-p7a19 : usr: 861.49s, sys: 4.22s, cpu:88%, total: 1237s
01: 2650 02: 3700 04: 4977 08: 5882 16: 6723 32: 7498 64: 8274
2.2.0-p7a19 : usr: 854.15s, sys: 9.74s, cpu:85%, total: 1267s
01: 2375 02: 3488 04: 4766 08: 5556 16: 6440 32: 7323 64: 8586
2.2.0-p7ac191: usr: 861.70s, sys: 6.90s, cpu:86%, total: 1137s
01: 2998 02: 4290 04: 5802 08: 6752 16: 7736 32: 8440 64: 9002
2.2.0-p7ac191: usr: 862.46s, sys: 6.13s, cpu:86%, total: 1129s
01: 3082 02: 4481 04: 5810 08: 7014 16: 7936 32: 8786 64: 9069
2.2.0-p7ac20: usr: 882.20s, sys: 3.09s, cpu:61%, total: 2064s
01: 2349 02: 3119 04: 4010 08: 4688 16: 6044 32: 6974 64: 7749
2.2.0-p7ac20: usr: 850.52s, sys: 8.33s, cpu:74%, total: 1470s
01: 2550 02: 3549 04: 4624 08: 5549 16: 6854 32: 7398 64: 8269
2.2.0-p8 : usr: 853.58s, sys: 7.60s, cpu:84%, total: 1145s
01: 2905 02: 4258 04: 5654 08: 6701 16: 7958 32: 8656 64: 8935
2.2.0-p8 : usr: 866.10s, sys: 5.73s, cpu:86%, total: 1130s
01: 3061 02: 4299 04: 5697 08: 6865 16: 7785 32: 8493 64: 9060
2.2.0-p8a25 : usr: 864.80s, sys: 2.47s, cpu:90%, total: 1593s
01: 1963 02: 2848 04: 3688 08: 4517 16: 5420 32: 6624 64: 7227
2.2.0-p8a25 : usr: 868.69s, sys: 7.73s, cpu:90%, total: 1443s
01: 2168 02: 3076 04: 4129 08: 5044 16: 5875 32: 6873 64: 7538
2.2.0-p8a26 : usr: 854.74s, sys: 5.39s, cpu:90%, total: 1373s
01: 2271 02: 3231 04: 4512 08: 5299 16: 6056 32: 6754 64: 8385
2.2.0-p8a26 : usr: 869.20s, sys: 7.80s, cpu:91%, total: 1400s
01: 2306 02: 3312 04: 4398 08: 5312 16: 6511 32: 7539 64: 8225
Keep up your good work, guys!
dLux
--
Isten ovja a kepernyot! God save the screen!
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/