Re: `process-send-*` performance seems ... bad?

[Daniel Pittman <slippycheeze@google.com>]

[-- Attachment #1: Type: text/plain, Size: 1774 bytes --]

Ah, no, that is running locally – the call to make-process is not using
file handlers, and the purpose of setting that default directory was to
ensure that it didn't.

That said, I retested with `(default-directory (expand-file-name "~"))`
there, and performance is identical:

========== pass 1 ==========
write-with-temp-buffer: 5.8827s with 0 GCs taking 0.0000s: (5.882696 0 0.0)
write-with-two-calls: 2.4786s with 0 GCs taking 0.0000s: (2.478616 0 0.0)
write-with-one-call: 1.9073s with 0 GCs taking 0.0000s: (1.907256 0 0.0)
========== pass 2 ==========
write-with-temp-buffer: 5.6867s with 0 GCs taking 0.0000s: (5.686659 0 0.0)
write-with-two-calls: 2.2408s with 0 GCs taking 0.0000s: (2.240782 0 0.0)
write-with-one-call: 1.8802s with 0 GCs taking 0.0000s: (1.8801780000000001
0 0.0)
========== pass 3 ==========
write-with-temp-buffer: 5.7305s with 0 GCs taking 0.0000s:
(5.730538999999999 0 0.0)
write-with-two-calls: 2.3394s with 0 GCs taking 0.0000s: (2.339416 0 0.0)
write-with-one-call: 1.9350s with 0 GCs taking 0.0000s: (1.9350479999999999
0 0.0)


On Wed, Jun 19, 2019 at 10:58 AM Noam Postavsky <npostavs@gmail.com> wrote:

> On Tue, 18 Jun 2019 at 17:55, Daniel Pittman <slippycheeze@google.com>
> wrote:
>
> > ] cat obarray.data > in & wc < out & time fzf -f "hook mode" < in > out
> ; wait
>
> > The code behind those results is attached
>
>           (let ((default-directory
>
> "/gssh:slippycheeze.c.googlers.com:
> /google/src/cloud/slippycheeze/work/google3/production/storage/chronicle/")
>
> It looks you're running locally in the shell case, but over TRAMP in
> the elisp case. I imagine that could have a pretty large effect on
> timing. Is there a reason for binding default-directory like this?
>

[-- Attachment #2: Type: text/html, Size: 2279 bytes --]