epsilon-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Jitter v0.9.249: better profiling [Was: Re: [Jitter v0.9.248] Jitter now


From: Luca Saiu
Subject: Jitter v0.9.249: better profiling [Was: Re: [Jitter v0.9.248] Jitter now has sample-based profiling as well; updated poke submodule reference]
Date: Mon, 28 Dec 2020 02:47:14 +0100
User-agent: Gnus (Gnus v5.13), GNU Emacs 27.0.50, x86_64-pc-linux-gnu

I became very suspicious after seeing a VM remainder instruction taking,
according to sample-profiling, just a few hundreds of picoseconds on my
machine instead of the typical ten nanoseconds.

I have now just made profiling more accurate, forcing instrumentation
code to be generated *before* user code.  Allowing GCC to freely reorder
it could start the measurement at any point of the VM instruction
execution, including *after* the work is done.

The current solution is still not perfect because the measurement takes
place (deterministically) *after* residual materialisation.  But in its
current state profiling will be useful, particularly for slow VM
instructions.  For very fast VM instructions this kind of profiling will
never be reliable, since the instrumentation code can easily be slower
than the instruction itself.

No API changes.

New tag v0.9.249.  Updated submodule in poke.

-- 
Luca Saiu
* My personal web site:  http://ageinghacker.net
* GNU epsilon:           http://www.gnu.org/software/epsilon
* Jitter:                http://ageinghacker.net/projects/jitter

I support everyone's freedom of mocking any opinion or belief, no
matter how deeply held, with open disrespect and the same unrelented
enthusiasm of a toddler who has just learned the word "poo".

Attachment: signature.asc
Description: PGP signature


reply via email to

[Prev in Thread] Current Thread [Next in Thread]