With real-time profiling and observability becoming more important for all kinds of workloads, we should enable frame pointers for all arches to maximize the usability of such tooling for Linux workload and application development.
Of note, several Linux distributions have done this change:
Shipping the distribution with frame pointers is a major competitive benefit because it makes SUSE distributions more attractive for developers, operators, and analysts who want to make performant workloads. A full description of the benefits of doing so can be read in the Fedora and Arch Linux documents for this.
I've also proposed it for openSUSE Factory:
factory@
packaging@
Hello Neal,
until we have clarified situation on Factory I'd not touch Leap 16.0 Please be aware that even if we enable this, packages inherited from SLFO, might be built with a different set of options.
Lubos
Yes, well, doing a mass build to apply it to everything isn't hard.
Metadata Update from @lkocman: - Custom field SUSE Jira - SUSE Linux Enterprise adjusted to https://jira.suse.com/browse/PED-11594
Proposed the same for SLFO / SLES 16 https://jira.suse.com/browse/PED-11594
It does seem that it would result into noticeable performance decline. Seems like RHEL 10 will skip it too. I'll double check with team before closing.
There shouldn't be a significant decline because the major one was an older Python version that neither RHEL 10 nor SLE 16 would offer. Not to mention, even if we do it by default for everything, I would add a way for specific packages to opt-out of it.
RHEL 10 is skipping it not because of performance reasons but because the RHEL toolchain team themselves dislike frame pointers out of principle.
Log in to comment on this ticket.