?

Log in

No account? Create an account
x86 on IA-64 - brad's life — LiveJournal [entries|archive|friends|userinfo]
Brad Fitzpatrick

[ website | bradfitz.com ]
[ userinfo | livejournal userinfo ]
[ archive | journal archive ]

x86 on IA-64 [Dec. 27th, 2004|10:54 am]
Brad Fitzpatrick
I'm running a CPU-heavy app on an Itanium2 box now. For a while I couldn't get it to compile IA-64 so I was using the x86 binary.... 100% CPU. Got it to rebuild and it's now at 80% CPU and running 4x faster.
LinkReply

Comments:
[User Picture]From: gaal
2004-12-27 08:36 pm (UTC)
Wouldn't you prefer it to be 100% CPU and 5x faster?
(Reply) (Thread)
[User Picture]From: brad
2004-12-27 08:40 pm (UTC)
It's IO-bound now.
(Reply) (Parent) (Thread)
[User Picture]From: agreg
2004-12-27 09:34 pm (UTC)
What is it? :)
(Reply) (Parent) (Thread)
[User Picture]From: brad
2004-12-27 10:08 pm (UTC)
InnoDB checksum validator... pulling 16k at a time off a 340 GB file and verifying the checksum of that database pages matches what it thinks it is.
(Reply) (Parent) (Thread)
From: jamesd
2004-12-27 10:44 pm (UTC)

InnoDB checksum validator

Please say more about the InnoDB checksum validator. Something other than check table?
(Reply) (Parent) (Thread)
[User Picture]From: brad
2004-12-27 11:59 pm (UTC)

Re: InnoDB checksum validator

Yeah, it does the same per-16k-page integrity checks that InnoDB does itself. marksmith basically gutted the InnoDB source code and found the checksumming bits and made an independent tool:

http://www.livejournal.com/~xb95/260706.html

We gave them permission to own the code and include it in the next version of MySQL ... we'll see if they do.
(Reply) (Parent) (Thread)
From: jamesd
2004-12-28 12:59 am (UTC)

Re: InnoDB checksum validator

Thanks. I want that.:) Arjen was in IRC and I told him as much, told him that I was telling you that you could say Wikipedia wants it as well if you like... He mentioned that if you want it in the next version it's really up to you to watch over it and make sure it's in there.

I like the sound of compaction as well. Also defragmentation. Journal post coming up about that...
(Reply) (Parent) (Thread)
[User Picture]From: supersat
2004-12-27 10:08 pm (UTC)
Intel really dropped the ball on the IA-64. Sure, the x86 has a ton of problems, but the existing software base is so large that it makes sense to make it as backward-compatible as possible. I think they had a similar problem with the Pentium Pro, which wouldn't run 16-bit software (like parts of Windows 9x) without major slowdowns.
(Reply) (Thread)
[User Picture]From: brad
2004-12-27 11:57 pm (UTC)
IA-64 works great. I don't give a fuck about binary compatability. MySQL, Linux, etc run awesome.

The bug compiling this latest app was a real bug... fixed it and it compiled fine.
(Reply) (Parent) (Thread)
From: e_carter
2004-12-28 05:19 pm (UTC)

Question:

Did you get your answer from F5 yet?

Did it help?

Let me know...

(I'm on the inside :-)
(Reply) (Thread)
[User Picture]From: brad
2004-12-28 05:58 pm (UTC)

Re: Question:

I haven't been touching them. Last I heard the latest "upgrade" totally killed one of them.
(Reply) (Parent) (Thread)