On Thu, Apr 23, 2026 at 1:09 AM Kugan Vivekanandarajah < [email protected]> wrote:
> Hi, > > Building on our discussions from the last Cauldron, we propose creating a > native, simplified AutoFDO tool for GCC to replace our current reliance on > external Google tools which is not actively maintained. I'll follow up with > a detailed design document as soon as we have a consensus on the proposal. > Thank you for writting it. In my opinion the tool is really needed to make AutoFDO practically useful, so I hope it will happen this stage 1 :) > > Thanks, > Kugan > > Summary > ======= > > We propose a standalone, minimal tool for generating AutoFDO profiles > that can be consumed by the GCC AutoFDO toolchain, with the goal of > integrating it into the GCC repository. The tool would > support: (1) offline read existing perf.data (single-process or > system-wide) and produce a profile for a target binary; (2) direct > attach to a process via the PMU (LBR or BRBE / SPE), bypassing perf > record and building the profile from the live sample stream; (3) > I think this is quite important missing feature of the existing tool. With branch history, It is hard to balance the setting to keep profile good while avoiding too large perf.data file. Data used by GCC is much smaller. > - Current AutoFDO tools (e.g. from Google) are not widely used with > GCC. LLVM has a similar profile-creation tool integrated with the > compiler. A tightly coupled tool for GCC would allow for easy > development and upgradation. > I also wonder if there is a chance for collaboration/integration with perf tools itself. In particular, the current tool is picky about perf.data file format that is not completely stable. So perhaps the perf.data loading can be sharing a library with perftools. On the other hand hwoever minimizing dependencies is quite important. Honza > >
