Re: Shark on MacBook Air
Re: Shark on MacBook Air
- Subject: Re: Shark on MacBook Air
- From: Rick Altherr <email@hidden>
- Date: Wed, 20 Feb 2008 11:21:40 -0800
On Feb 20, 2008, at 10:45 AM, David Dunham wrote:
I started doing some performance analysis of a game that uses
OpenGL, and happened to be using a MacBook Air. I'm seeing results
very different than I'd expect:
0.0% 81.8% mach_kernel dtrace_get_cpu_int_stack_top
Most of this time is in
0.1% 79.8% com.apple.driver.AppleIntelGMAX3100
IOIntel965GLContext::clientMemoryForType(unsigned long, unsigned
long*, IOMemoryDescriptor**)
So maybe integrated video (as opposed to a separate card) is sucking
up all the CPU. But the part that seems surprising is that it's in
dtrace. 1) I thought that was supposed to be low overhead; 2) I've
never seen Shark report a lot of time spent gathering performance data
Is this normal?
David Dunham
Voice/Fax: 206 783 7404 http://www.pensee.com/dunham/
Imagination is more important than knowledge. -- Albert Einstein
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Xcode-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
@apple.com
This email sent to email@hidden
Without more of the tree, I can't say much as to what the root cause
it. However, it's likely that your kernel is stripped so Shark is
guessing as to which symbol contains the address in the call stack.
dtrace_get_cpu_int_stack_top was likely the closest symbol to the
address.
You might ask why we just don't report unknown. The symbol table for
the linker doesn't specify ranges for symbols, but only start
addresses. All the debug tools that deal with symbols end up getting
the start addresses for each symbol and then "expand" them to
determine what is a likely end address. DWARF actually provides (or
at least can provide) end addresses for symbols, so no guessing is
needed.
In this specific case, the debug information for the kernel has been
stripped, so we rely solely on the linker symbol table. Thus we guess
and on a stripped binary, it can be quite wrong.
--
Rick Altherr
Architecture and Performance Group
email@hidden
Attachment:
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Xcode-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden