Performance of std::pow - cache misses???

Posted by Eamon Nerbonne on Stack Overflow See other posts from Stack Overflow or by Eamon Nerbonne
Published on 2010-03-20T19:06:09Z Indexed on 2010/03/20 19:11 UTC
Read the original article Hit count: 517

Filed under:

cache

|

cpu

|

profiling

|

c++

I've been trying to optimize a numeric program of mine, and have run into something of a mystery. I'm looping over code that performs thousands of floating point operations, and just 1 call to pow nevertheless, that call takes 5% of the time... That's not necessarily a critical issue, but it is odd, so I'd like to understand what's happening.

When I profiled for cache misses, VS.NET 2010RC's profiler reports that virtually all cache misses are occurring in std::pow... so... what's up with that? Is there a faster alternative? I tried powf, but that's only slightly faster; it's still responsible for an abnormal number of cache misses.

Why would a basic function like pow cause cache-misses?

© Stack Overflow or respective owner

Related posts about cache

How to remove a package entirely?

as seen on Ask Ubuntu - Search for 'Ask Ubuntu'
Hi I'm quite new to Linux, but before using it I was hearing that Windows programs, after uninstallation, leaves a lot of remains on the hard disc, and Linux removes all. I'm using Ubuntu 10.04. To uninstall packages I'm using sudo apt-get autoremove application_name or sudo aptitude purge application_name… >>> More
LVM / Device Mapper maps wrong device

as seen on Server Fault - Search for 'Server Fault'
Hi, I run a LVM setup on a raid1 created by mdadm. md2 is based on sda6 (major:minor 8:6) and sdb6 (8:22). md2 is partition 9:2. The VG on top of md2 has 4 LVs, var, home, usr, tmp. First the problem: While booting it seems as if the device mapper takes the wrong partition for the mapping! Immediately… >>> More
Application throws NotSerializableException when run on an jboss cluster

as seen on Stack Overflow - Search for 'Stack Overflow'
Environment: JBoss 5.1.0, JBoss Seam 2.2.0 While trying to get my application running in a clustered environment after login I am getting the following exception. Post login we try to store the currentUser in jboss seam session context. java.io.NotSerializableException: org.jboss.seam.util.AnnotatedBeanProperty How… >>> More
Windows Azure Evolution – Caching (Preview)

as seen on Geeks with Blogs - Search for 'Geeks with Blogs'
Caching is a popular topic when we are building a high performance and high scalable system not only on top of the cloud platform but the on-premise environment as well. On March 2011 the Windows Azure AppFabric Caching had been production launched. It provides an in-memory, distributed caching service… >>> More
Locking Cache Key without Locking the entire Cache

as seen on Stack Overflow - Search for 'Stack Overflow'
I have servlets that caches user information rather then retrieving it from the user store on every request (shared Ehcache). The issue I have is that if a client is multi-threaded and they make more then one simultaneous request, before they have been authenticated, then I get this in my log: Retrieving… >>> More

Related posts about cpu

Speedstep and Intel x5570?

as seen on Server Fault - Search for 'Server Fault'
Hi, My new server has 2 x X5570 CPUs. Now here is the output of grep -i hz /proc/cpuinfo model name : Intel(R) Xeon(R) CPU X5570 @ 2.93GHz cpu MHz : 1600.231 model name : Intel(R) Xeon(R) CPU X5570 @ 2.93GHz cpu MHz : 1600.231 model name : Intel(R) Xeon(R) CPU … >>> More
VMware ESXi - varying CPU time (CPU reservation)

as seen on Server Fault - Search for 'Server Fault'
Hello! I'm running FreeBSD 7.2 under VMware ESXi 3.5. Host has 2 physical CPUs and the BSD box is currently the only running VM. Only one virtual CPU is assigned to the VM. When measuring CPU time of a specific program, I get very different results from time to time. Processor usage is reported… >>> More
How to calculate CPU % based on raw CPU ticks in SNMP

as seen on Server Fault - Search for 'Server Fault'
According to http://net-snmp.sourceforge.net/docs/mibs/ucdavis.html#scalar_notcurrent ssCpuUser, ssCpuSystem, ssCpuIdle, etc are deprecated in favor of the raw variants (ssCpuRawUser, etc). The former values (which don't cover things like nice, wait, kernel, interrupt, etc) returned a percentage… >>> More
disparity between `top`'s given CPU % and process CPU usage total

as seen on Super User - Search for 'Super User'
I've noticed that there are sometimes (large) differences between the reported total CPU usage and a summation of the per-process CPU utilization given by apps like top and wmtop. As an example: I recently ran a git filter-branch --index-filter on a fairly large repo, with the index-filter command… >>> More
How to read oom-killer syslog messages?

as seen on Server Fault - Search for 'Server Fault'
I have a Ubuntu 12.04 server which sometimes dies completely - no SSH, no ping, nothing until it is physically rebooted. After the reboot, I see in syslog that the oom-killer killed, well, pretty much everything. There's a lot of detailed memory usage information in them. How do I read these logs… >>> More