Hi,
I am wondering if it is possible to get the time-stamp and the processor ID where an instruction is exectued in the IR? Thanks for your help in advance:)
Best,
Zhi
Hi,
I am wondering if it is possible to get the time-stamp and the processor ID where an instruction is exectued in the IR? Thanks for your help in advance:)
Best,
Zhi
Hi Zhi,
There is no standard (architecture-independent) way to get the processor id. We do have an intrinsic to get the cycle counter:
http://llvm.org/docs/LangRef.html#llvm-readcyclecounter-intrinsic
-Hal
Hi Hal,
Thanks for the pointer. Is it possible to get the processor ID on X86 architecture? There is a library call in linux, sched_getcpu(), to the ID.
Also, is it possible to get the program counter in the IR?
Best,
Zhi
From: "zhi chen" <zchenhn@gmail.com>
To: "Hal Finkel" <hfinkel@anl.gov>
Cc: llvm-dev@lists.llvm.org
Sent: Thursday, November 19, 2015 3:48:02 PM
Subject: Re: [llvm-dev] Get timestamp and processor ID in the IRHi Hal,
Thanks for the pointer. Is it possible to get the processor ID on X86
architecture? There is a library call in linux, sched_getcpu(), to
the ID.
What exactly do you mean by "processor ID"?
-Hal
Say if we have 8 cores, I want to get which core is running the current program. Thanks.
From: "zhi chen" <zchenhn@gmail.com>
To: "Hal Finkel" <hfinkel@anl.gov>
Cc: llvm-dev@lists.llvm.org
Sent: Thursday, November 19, 2015 5:47:45 PM
Subject: Re: [llvm-dev] Get timestamp and processor ID in the IRSay if we have 8 cores, I want to get which core is running the
current program. Thanks.
What do you intend to do with that information?
One obvious complication is that the identity of the core running your application might change at any time (whenever the operating system performs a context switch). If you need to need to restrict to running only on one core, you'll likely need to adjust the process's core bindings (affinity mask) using some operating-system specific method.
-Hal
Hi Hal,
I just want to see which core the program is assigned to. I know we could set the CPUaffinity, but I don’t want to specify it. I want it to be scheduled by the operating systems itself, then I will check which core it is scheduled. Is it possible to get this information? Thanks.