--- Begin Message ---
Subject: |
[PATCH] guix: parallelize building the manual-database |
Date: |
Fri, 12 Jul 2019 23:42:45 +0200 |
* guix/profiles.scm (manual-database): par-map over the entries. This
distributes the load roughly equally over all cores and avoids blocking on
I/O. The order of the entries stays the same since write-mandb-database sorts
them.
---
guix/profiles.scm | 20 ++++++++++++--------
1 file changed, 12 insertions(+), 8 deletions(-)
diff --git a/guix/profiles.scm b/guix/profiles.scm
index f5c863945c..374f0f8a90 100644
--- a/guix/profiles.scm
+++ b/guix/profiles.scm
@@ -1312,15 +1312,11 @@ the entries in MANIFEST."
#~(begin
(use-modules (guix man-db)
(guix build utils)
+ (ice-9 threads)
(srfi srfi-1)
(srfi srfi-19))
- (define (compute-entries)
- ;; This is the most expensive part (I/O and CPU, due to
- ;; decompression), so report progress as we traverse INPUTS.
- (let* ((inputs '#$(manifest-inputs manifest))
- (total (length inputs)))
- (append-map (lambda (directory count)
+ (define (compute-entry directory count total)
(format #t "\r[~3d/~3d] building list of \
man-db entries..."
count total)
@@ -1330,8 +1326,16 @@ man-db entries..."
(if (directory-exists? man)
(mandb-entries man)
'())))
- inputs
- (iota total 1))))
+
+ (define (compute-entries)
+ ;; This is the most expensive part (I/O and CPU, due to
+ ;; decompression), so report progress as we traverse INPUTS.
+ (let* ((inputs '#$(manifest-inputs manifest))
+ (total (length inputs)))
+ (apply append (par-map compute-entry
+ inputs
+ (iota total 1)
+ (make-list total total)))))
(define man-directory
(string-append #$output "/share/man"))
--
2.22.0
--- End Message ---
--- Begin Message ---
Subject: |
Re: [bug#36630] [PATCH] guix: parallelize building the manual-database |
Date: |
Tue, 31 Mar 2020 15:02:47 +0200 |
User-agent: |
Gnus/5.13 (Gnus v5.13) Emacs/26.3 (gnu/linux) |
Hi Arne,
Arne Babenhauserheide <address@hidden> skribis:
> I now reduced the thread count to exactly 2 (to avoid running into
> resource troubles; I hope that two should be safe) and added a mutex for
> status messages to ensure that writes don’t overlap.
It’s been 9 months but I finally committed a slightly modified variant
as ef4b5f2fed3ca13a0e15a821ba7e561cd4395aa6. It turns out that the
mutex was unnecessary as ports are thread-safe.
As noted in the log, I see a 36% speedup on my SSD laptop with 4 cores
(slightly less with 2 cores). It’s not great, but still an improvement!
Thanks,
Ludo’.
--- End Message ---