Subject: Re: [boost] [review] **NEXT WEEK** Review of Outcome (starts Fri-19-May)
From: Marc Glisse (marc.glisse_at_[hidden])
Date: 2017-05-14 08:33:22
On Sun, 14 May 2017, Niall Douglas via Boost wrote:
> There was a very interesting talk at ACCU (unfortunately in the
> non-videoed room) by a HFT guy explaining all the ways in which recent
> GCC versions have bugged __builtin_expect, like inverting the code path
> you specifically told the compiler was the hot path.
> GCC devs apparently don't care enough to fix it despite multiple
> persistent reports, and the feature is now useless on GCC >= 5. The
> speaker recommended clang, especially very recent clang which is
> apparently finally being competitive with GCC in generating very tight
> code. I'd back that up, clang 5.0 is generating much tighter code with
> Outcome than clang 3.x did, markedly so.
> But I can see a feature like __builtin_expect that going the way of the
> dodo as the compiler vendors really would prefer if you used profile
> guided optimisation instead. Passing that sort of micro-info from the
> parser to the backend is surely complex to get right.
I looked for bug reports about __builtin_expect in gcc's bugzilla, and the
only relevant one I found was
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=59521 about using it in a
switch. Could you back your statement with some links? Otherwise it sounds
like FUD (at least the part about ignoring multiple persistent reports).
-- Marc Glisse
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk