OProfile is a powerful system-wide profiler for Linux. Read more at http://oprofile.sf.net OProfile 0.7 has been released. OProfile is still in alpha, but has been proven stable for many users. Release notes ------------- For 2.2 kernels, the module must be compiled as the same user that owns the kernel source tree. nosmp is not supported in kernels before 2.4.10 (bug #463087). The pre-emptable kernel option is not supported in 2.4 (bug #478516). Power management on laptops can be incompatible with OProfile in 2.4 (bug #554927). The sample file format has changed again from previous releases. You will not be able to read older sample files with OProfile 0.7, or vice versa. You'll find a small utilities at http://perso.wanadoo.fr/phil.el/ to change old format to new format. The default event count for performance-counter CPUs has been changed to a set value, instead of scaling on CPU frequency. This means that the granulity of sampling is now constant, though faster CPUs will have a higher interrupt load. New features ------------ If a binary could not be read or found, the samples for it are no longer discarded in the opreport output: instead they appear as if the binary had no symbols with the "(no symbols)" marker. By default, opreport no longer shows the VMA address of each symbol. You can re-enable this with the --show-address / -w option. Per processor and per thread samples separation has been implemented Bug fixes --------- Compilation with Qt 2.3.1 has been fixed. A race which could lead to empty sample files, triggering an assertion in the post-profiling tools, has been fixed. --global-percent now applies to assembly instruction percentages in --details output. A symbolic link passed to the post-profiling tools is now correctly resolved to the target binary. The limitation of 64 kernel modules has finally been removed.