memspeed on Aspenite
Siddarth Gore
gores at marvell.com
Mon Aug 23 09:37:48 EDT 2010
Hello All,
Please disregard this message, it has been sent to this list by mistake.
I am very sorry for the inconvenience it may have caused.
-siddarth
> -----Original Message-----
> From: linux-arm-kernel-bounces at lists.infradead.org [mailto:linux-arm-kernel-
> bounces at lists.infradead.org] On Behalf Of Siddarth Gore
> Sent: Monday, August 23, 2010 6:23 PM
> To: linux-arm-kernel at lists.infradead.org
> Subject: memspeed on Aspenite
>
> Hello,
>
> The results for memspeed app on Aspenite DB for LSP 2.2.8.2.3 as well as
> mainline 2.6.35 kernel, are given below.
>
> * both kernels have L2 cache disabled and L1 cache enabled.
> * uncached memory writes are noticeably slower on mainline kernel,
> whereas the speeds are comparable when memory block is already present
> in L1 (data) cache.
> * the reason for this could be slow speed/contention while accessing
> main memory, so i repeated the tests with no caching atall (L1 and L2
> disabled). These numbers came out same for LSP and mainline kernel.
> * further, bandwidth while accessing cached data is also much greater in
> the LSP kernel.
>
> I am looking into what might be happening, any ideas/suggestions
> welcome.
>
> -siddarth
> PS: code used to compile memspeed also attached.
>
> ***************Mainline (2.6.35) kernel without L2 (BUFFER = 64k)
>
> ~ # ./memspeed -c
> *** Memory Write Throughput (in MB/s) ***
> method uncached write alloc cached cached + wa
> STRB 188.46 331.38 577.25 899.28
> STR 786.38 382.35 2642.50 2652.90
> STM4 1624.31 385.44 2909.16 3171.53
> STM8 2034.44 386.43 4442.98 3939.27
> STRD 1631.52 386.09 3236.05 3941.39
>
> *** Uncached Memory Read/Copy Throughput (in MB/s) ***
> method PLD=0 PLD=1 PLD=2 PLD=4 PLD=8 PLD=16 PLD=32 PLD=64
> LDRB 370.42 370.18 456.03 507.48 541.11 442.01 400.24 380.38
> LDR 510.22 505.19 562.18 578.19 599.29 605.76 574.74 535.86
> LDM 548.45 532.68 588.53 594.42 602.35 610.00 603.39 559.04
> LDRD 477.59 430.61 511.33 552.34 576.48 566.28 498.13 469.23
> CPY_B 156.81 156.80 163.67 149.09 152.55 151.11 152.10 154.19
> CPY_Bwa 151.67 164.81 162.03 144.64 141.02 147.12 146.84 144.61
> CPY_R 305.83 302.81 313.71 283.41 283.56 295.05 311.54 320.04
> CPY_Rwa 192.24 207.53 209.01 180.25 175.17 184.07 183.95 180.23
> CPY_M 378.40 322.67 417.79 403.85 398.06 396.56 409.38 416.31
> CPY_Mwa 198.26 210.24 209.62 182.47 178.10 186.53 187.39 182.89
> CPY_D 336.05 315.36 361.08 377.71 356.48 358.65 397.45 399.02
> CPY_Dwa 192.46 204.63 203.66 177.79 172.57 181.01 182.23 178.57
>
> *** Cached Memory Read/Copy Throughput (in MB/s) ***
> method PLD=0 PLD=1 PLD=2 PLD=4 PLD=8 PLD=16 PLD=32 PLD=64
> LDRB 922.34 871.74 862.09 858.70 846.03 845.91 849.66 846.75
> LDR 2900.89 2464.00 2369.34 2328.55 2245.00 2235.03 2255.32 2277.83
> LDM 4497.51 3942.53 3027.01 2952.48 2819.55 2821.08 2847.01 2876.93
> LDRD 1616.92 1469.47 1458.91 1467.13 1444.78 1447.80 1461.38 1461.31
> CPY_B 491.15 486.40 466.49 467.21 463.10 461.04 461.72 465.06
> CPY_Bwa 477.82 477.00 462.64 458.03 457.13 454.09 454.84 456.76
> CPY_R 1781.87 1685.79 1501.97 1475.36 1451.06 1443.40 1451.77 1459.49
> CPY_Rwa 1605.59 1603.00 1429.02 1414.78 1385.34 1379.93 1394.19 1393.15
> CPY_M 2666.13 2462.52 2073.39 2037.13 1966.12 1971.33 1991.05 1995.38
> CPY_Mwa 2465.72 2275.77 1944.52 1919.53 1861.64 1857.50 1879.41 1873.19
> CPY_D 1405.17 1348.93 1267.71 1257.24 1224.85 1225.88 1232.08 1239.05
> CPY_Dwa 1342.51 1241.43 1213.40 1207.56 1180.84 1180.08 1189.68 1190.25
>
>
> ***************LSP (2.2.8.2.3) kernel without L2 (BUFFER = 64k)
>
> ~ # ./memspeed -c
> *** Memory Write Throughput (in MB/s) ***
> method uncached write alloc cached cached + wa
> STRB 922.09 335.67 920.63 893.59
> STR 1917.30 388.38 2883.57 2648.31
> STM4 1630.89 391.49 3501.35 3160.13
> STM8 2031.07 392.65 4463.82 3924.40
> STRD 2029.12 390.60 4470.54 3931.82
>
> *** Uncached Memory Read/Copy Throughput (in MB/s) ***
> method PLD=0 PLD=1 PLD=2 PLD=4 PLD=8 PLD=16 PLD=32 PLD=64
> LDRB 369.06 369.25 452.36 504.15 539.96 437.91 399.73 380.75
> LDR 509.63 505.75 559.96 572.86 594.95 601.14 570.57 533.24
> LDM 544.69 528.20 585.61 590.51 601.14 606.34 601.24 559.95
> LDRD 477.41 429.09 509.10 547.05 571.84 562.20 498.04 469.70
> CPY_B 220.17 220.59 276.18 309.26 320.63 290.12 273.31 263.76
> CPY_Bwa 145.52 153.85 156.61 139.90 135.80 142.68 142.82 141.49
> CPY_R 341.04 338.97 383.76 374.03 373.13 390.59 416.70 412.97
> CPY_Rwa 183.95 192.84 199.53 173.22 167.64 177.95 178.17 175.70
> CPY_M 347.05 294.75 399.88 378.63 372.31 371.14 397.91 411.09
> CPY_Mwa 188.47 195.73 202.40 175.96 171.07 180.76 181.11 178.87
> CPY_D 336.53 331.34 357.69 365.06 369.98 379.19 406.97 398.67
> CPY_Dwa 184.39 189.68 196.50 171.02 165.10 175.72 175.80 173.78
>
> *** Cached Memory Read/Copy Throughput (in MB/s) ***
> method PLD=0 PLD=1 PLD=2 PLD=4 PLD=8 PLD=16 PLD=32 PLD=64
> LDRB 920.36 871.56 859.23 853.41 842.45 842.32 845.01 846.46
> LDR 2884.00 2451.59 2356.95 2313.08 2234.44 2233.52 2252.71 2262.72
> LDM 4464.70 3924.43 3013.70 2941.92 2816.51 2814.44 2845.41 2860.78
> LDRD 1603.57 1460.50 1458.29 1457.61 1441.90 1439.00 1448.91 1454.05
> CPY_B 489.71 482.38 465.19 463.44 460.18 460.17 460.94 461.37
> CPY_Bwa 475.51 475.38 458.62 456.87 453.79 453.70 454.53 454.88
> CPY_R 1776.31 1684.34 1490.75 1473.20 1440.56 1440.44 1448.34 1452.47
> CPY_Rwa 1602.42 1601.08 1425.43 1409.13 1379.68 1379.20 1386.52 1390.25
> CPY_M 2645.24 2445.68 2058.14 2024.33 1964.26 1963.14 1978.23 1985.58
> CPY_Mwa 2445.75 2274.42 1935.22 1905.66 1852.03 1851.33 1864.35 1871.35
> CPY_D 1395.71 1338.04 1259.41 1246.73 1224.40 1222.21 1229.33 1232.92
> CPY_Dwa 1338.12 1236.86 1212.41 1200.58 1179.88 1177.74 1184.57 1187.82
More information about the linux-arm-kernel
mailing list