Re: [Ocf-linux-users] No performances improvment with AMD Geode
Brought to you by:
david-m
Menu
▾
▴
Re: [Ocf-linux-users] No performances improvment with AMD Geode
|
From: Adam C. (Le_Vert) <ga...@le...> - 2008-07-10 07:46:56
|
David McCullough a écrit : > Jivin "Adam Cécile (Le_Vert)" lays it down ... > >> David McCullough a écrit : >> >>> Jivin "Adam Cécile (Le_Vert)" lays it down ... >>> >>> >>>> Daniel Mueller a écrit : >>>> >>>> >>>>> On Mon, 02 Jun 2008 17:25:23 +0200 Adam Cécile wrote: >>>>> >>>>> I do not own a AMD Geode but.. >>>>> >>>>> >>>>> >>>>> >>>>>> Loaded kernel modules: >>>>>> gandalf@alix:~$ lsmod | grep -e cry -e oc >>>>>> cryptodev 13988 3 >>>>>> crypto_null 2624 0 >>>>>> cryptosoft 12136 0 >>>>>> ocf 28984 2 cryptodev,cryptosoft >>>>>> >>>>>> >>>>>> >>>>> .. you need to load the geode-aes module as well. You can find it in >>>>> your kernel configuration. >>>>> >>>>> Cryptographic API ---> >>>>> [*] Hardware crypto devices ---> >>>>> <M> Support for the Geode LX AES engine >>>>> >>>>> Try to load the modules in the following order: >>>>> geode-aes.ko >>>>> ocf.ko >>>>> cryptosoft.ko >>>>> cryptodev.ko >>>>> >>>>> bye, >>>>> danm >>>>> >>>>> >>>>> >>>>> >>>> gandalf@alix:~$ cat /etc/modules >>>> # /etc/modules: kernel modules to load at boot time. >>>> # >>>> # This file contains the names of kernel modules that should be loaded >>>> # at boot time, one per line. Lines beginning with "#" are ignored. >>>> >>>> # Geode WatchDog >>>> geodewdt >>>> >>>> # Geode LX AES hardware crypto >>>> geode-aes >>>> geode-rng >>>> ocf >>>> cryptosoft >>>> cryptodev >>>> >>>> # Alix LEDs driver >>>> leds-alix >>>> >>>> # Sensors >>>> lm90 >>>> >>>> It's already loaded. Any other idea ? ;) >>>> >>>> >>> Did you check that with a "lsmod" to be sure ? >>> >>> Thanks, >>> Davidm >>> >>> >>> >> Hello, >> >> I've updated to debian's kernel 2.6.25-3-686 + latest ocf patch and I >> still don't see any performance improvement. >> > > > Have a look for emails by "Andreas Steinel" just recently , he seems to > have got it working, might be some clues. > More below. > > > >> gandalf@wrap:~$ lsmod | grep -e cry -e oc >> cryptodev 13572 3 >> crypto_null 3392 0 >> cryptosoft 11624 0 >> ocf 29176 2 cryptodev,cryptosoft >> crypto_blkcipher 18564 4 crypto_null,ecb,cbc,geode_aes >> dock 10320 1 libata >> > > Make sure you load crypto_blkcipher before you load ocf. > > >> gandalf@wrap:~$ ls -lah /dev/crypto >> crw-rw-rw- 1 root root 10, 70 Jan 1 2000 /dev/crypto >> >> gandalf@wrap:~$ openssl engine >> (cryptodev) BSD cryptodev engine >> (padlock) VIA PadLock (no-RNG, no-ACE) >> (dynamic) Dynamic engine loading support >> >> gandalf@wrap:~$ openssl speed -evp aes128 -elapsed >> You have chosen to measure elapsed time instead of user CPU time. >> To get the most accurate results, try to run this >> program when this computer is idle. >> Doing aes-128-cbc for 3s on 16 size blocks: 123525 aes-128-cbc's in 3.00s >> Doing aes-128-cbc for 3s on 64 size blocks: 119105 aes-128-cbc's in 3.00s >> Doing aes-128-cbc for 3s on 256 size blocks: 103091 aes-128-cbc's in 3.00s >> Doing aes-128-cbc for 3s on 1024 size blocks: 68212 aes-128-cbc's in 3.00s >> Doing aes-128-cbc for 3s on 2048 size blocks: 46401 aes-128-cbc's in 3.00s >> OpenSSL 0.9.8g 19 Oct 2007 >> built on: Mon Jun 2 14:52:51 UTC 2008 >> options:bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long) >> aes(partial) blowfish(idx) >> compiler: gcc -fPIC -DOPENSSL_PIC -DZLIB -DOPENSSL_THREADS -D_REENTRANT >> -DDSO_DLFCN -DHAVE_DLFCN_H -DHAVE_CRYPTODEV -DL_ENDIAN -DTERMIO -O3 >> -march=i586 -Wa,--noexecstack -g -Wall -DOPENSSL_BN_ASM_PART_WORDS >> -DOPENSSL_IA32_SSE2 -DSHA1_ASM -DMD5_ASM -DRMD160_ASM -DAES_ASM >> available timing options: TIMES TIMEB HZ=100 [sysconf value] >> timing function used: ftime >> The 'numbers' are in 1000s of bytes per second processed. >> type 16 bytes 64 bytes 256 bytes 1024 bytes 2048 >> bytes >> aes-128-cbc 658.80k 2540.91k 8797.10k 23283.03k >> 31676.42k >> > > > Ok, this looks like you are using OCF for crypto, it is easy to tell > because the small packets do not get the same total throughput as the > large packets. > > If you "rmmod cryptodev" then run the test again it should confirm this. > > > >> gandalf@wrap:~$ openssl speed -evp aes128 -elapsed -engine cryptodev >> engine "cryptodev" set. >> You have chosen to measure elapsed time instead of user CPU time. >> To get the most accurate results, try to run this >> program when this computer is idle. >> Doing aes-128-cbc for 3s on 16 size blocks: 123212 aes-128-cbc's in 3.00s >> Doing aes-128-cbc for 3s on 64 size blocks: 119413 aes-128-cbc's in 3.00s >> Doing aes-128-cbc for 3s on 256 size blocks: 103493 aes-128-cbc's in 3.00s >> Doing aes-128-cbc for 3s on 1024 size blocks: 67906 aes-128-cbc's in 3.00s >> Doing aes-128-cbc for 3s on 2048 size blocks: 46018 aes-128-cbc's in 3.00s >> OpenSSL 0.9.8g 19 Oct 2007 >> built on: Mon Jun 2 14:52:51 UTC 2008 >> options:bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long) >> aes(partial) blowfish(idx) >> compiler: gcc -fPIC -DOPENSSL_PIC -DZLIB -DOPENSSL_THREADS -D_REENTRANT >> -DDSO_DLFCN -DHAVE_DLFCN_H -DHAVE_CRYPTODEV -DL_ENDIAN -DTERMIO -O3 >> -march=i586 -Wa,--noexecstack -g -Wall -DOPENSSL_BN_ASM_PART_WORDS >> -DOPENSSL_IA32_SSE2 -DSHA1_ASM -DMD5_ASM -DRMD160_ASM -DAES_ASM >> available timing options: TIMES TIMEB HZ=100 [sysconf value] >> timing function used: ftime >> The 'numbers' are in 1000s of bytes per second processed. >> type 16 bytes 64 bytes 256 bytes 1024 bytes 2048 >> bytes >> aes-128-cbc 657.13k 2546.63k 8831.40k 23178.58k >> 31414.95k >> > > These numbers at the same as above. The reason for that is openssl will > use ocf by default if it is enabled. The best way to check is to remove > the cryptodev driver and see what the performance is. > > Either way, you are getting 31414.95k bytes of AES throughput. Thats > around 250Mbits/s using 2k packets. Are you sure that is not fast enough ? > > Check your numbers with cryptodev unloaded to know for sure. > > >> Only OpenSSL hasn't been rebuilt with latest OCF patch but I'll do asap. >> Any pointers would be greatly appreciated (maybe my bench commands are >> just wrong ?). >> > > No need to do that, the patch hasn't changed since the last release (at > least I don't recall any changes ;-) > > Cheers, > Davidm > > gandalf@wrap:~$ cat /etc/modules # /etc/modules: kernel modules to load at boot time. # # This file contains the names of kernel modules that should be loaded # at boot time, one per line. Lines beginning with "#" are ignored. # Geode WatchDog geodewdt # Geode LX AES hardware crypto geode-aes geode-rng crypto_blkcipher ocf cryptosoft cryptodev # Alix LEDs driver leds-alix # Sensors lm90 wrap:/home/gandalf# openssl speed -evp aes128 -elapsed You have chosen to measure elapsed time instead of user CPU time. To get the most accurate results, try to run this program when this computer is idle. Doing aes-128-cbc for 3s on 16 size blocks: 104083 aes-128-cbc's in 3.00s Doing aes-128-cbc for 3s on 64 size blocks: 101925 aes-128-cbc's in 3.00s Doing aes-128-cbc for 3s on 256 size blocks: 89677 aes-128-cbc's in 3.00s Doing aes-128-cbc for 3s on 1024 size blocks: 61529 aes-128-cbc's in 3.00s Doing aes-128-cbc for 3s on 2048 size blocks: 42862 aes-128-cbc's in 3.00s OpenSSL 0.9.8g 19 Oct 2007 built on: Wed Jul 9 18:31:21 UTC 2008 options:bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long) aes(partial) blowfish(idx) compiler: gcc -fPIC -DOPENSSL_PIC -DZLIB -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -DHAVE_CRYPTODEV -DL_ENDIAN -DTERMIO -O3 -march=i586 -Wa,--noexecstack -g -Wall -DOPENSSL_BN_ASM_PART_WORDS -DOPENSSL_IA32_SSE2 -DSHA1_ASM -DMD5_ASM -DRMD160_ASM -DAES_ASM available timing options: TIMES TIMEB HZ=100 [sysconf value] timing function used: ftime The 'numbers' are in 1000s of bytes per second processed. type 16 bytes 64 bytes 256 bytes 1024 bytes 2048 bytes aes-128-cbc 555.11k 2173.68k 7652.44k 21001.90k 29260.46k wrap:/home/gandalf# lsmod | grep cry crypto_blkcipher 18564 1 geode_aes wrap:/home/gandalf# lsmod | grep ocf wrap:/home/gandalf# openssl speed -evp aes128 -elapsed You have chosen to measure elapsed time instead of user CPU time. To get the most accurate results, try to run this program when this computer is idle. Doing aes-128-cbc for 3s on 16 size blocks: 895899 aes-128-cbc's in 3.00s Doing aes-128-cbc for 3s on 64 size blocks: 324390 aes-128-cbc's in 3.00s Doing aes-128-cbc for 3s on 256 size blocks: 92243 aes-128-cbc's in 3.00s Doing aes-128-cbc for 3s on 1024 size blocks: 23823 aes-128-cbc's in 3.00s Doing aes-128-cbc for 3s on 2048 size blocks: 11998 aes-128-cbc's in 3.00s OpenSSL 0.9.8g 19 Oct 2007 built on: Wed Jul 9 18:31:21 UTC 2008 options:bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long) aes(partial) blowfish(idx) compiler: gcc -fPIC -DOPENSSL_PIC -DZLIB -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -DHAVE_CRYPTODEV -DL_ENDIAN -DTERMIO -O3 -march=i586 -Wa,--noexecstack -g -Wall -DOPENSSL_BN_ASM_PART_WORDS -DOPENSSL_IA32_SSE2 -DSHA1_ASM -DMD5_ASM -DRMD160_ASM -DAES_ASM available timing options: TIMES TIMEB HZ=100 [sysconf value] timing function used: ftime The 'numbers' are in 1000s of bytes per second processed. type 16 bytes 64 bytes 256 bytes 1024 bytes 2048 bytes aes-128-cbc 4778.13k 6920.32k 7871.40k 8131.58k 8190.63k Okay there's performance difference when I unload all ocf/cryptodev kernel modules. But are you sure it comes from hardware ? It may just come from the fact that ocf software perform better than the standard linux crypto stack. What's your mind about this ? Thanks in advance, Regards, Adam. |
