| Age | Commit message (Collapse) | Author |
|---|
|
This effectively reverts 271a663d2f098f3f665cab6da2e13b265a7eab93 "rtl: support
start/stop threads around pre-init" again, now that
df6ba650469a6f2fda06ef1c2e107ccdd3570505 "Remove 'officially dead now' rtl_cache
slab allocator mechanism" removed the wsupdate thread.
(rtl_alloc_preInit is an internal-use-only C function, so changing its arguments
doesn't affect URE compatibility.)
Change-Id: Ie9bce86377f9520e2600e4111ac525dddace10f8
Reviewed-on: https://gerrit.libreoffice.org/58443
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
|
|
...after <https://gerrit.libreoffice.org/#/c/58263/> "the custom SAL allocator
is no longer used" got merged. According to my July 31, 2018 comment there:
"However, [...] this change is effectively a final coffin nail for the
'rtl_cache' mechanism. It could be argued that the alleged benefits of that
mechanism (if it were still working) might be real (at least on some platforms;
which would need measurements), and that it should be made working again (by
reverting the effects of both ce906b8096081dee15dc8cc96e570d5b0b587955 'skip
tricky allocators on G_SLICE=always-malloc' and
bc6a5d8e79e7d0e7d75ac107aa8e6aa275e434e9 'Disable custom allocator' on
rtl_cache_alloc/free again). But it could just as well be argued that the
'rtl_cache' mechanism is effectively gone for long enough now (since end of
November, 2017, with bc6a5d8e79e7d0e7d75ac107aa8e6aa275e434e9 'Disable custom
allocator') without any (apparent) negative consequences, so that it can be
removed for good with this change."
Change-Id: I8c1e45d494fc22555a9e675ab27be9e6e404abde
Reviewed-on: https://gerrit.libreoffice.org/58369
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
|
|
rtl_arena_create's quantum_cache_max parameter (when non-zero) would have
allowed an rtl_arena to internally use an rtl_cache for chunk allocation. (And
none of the LO-internal uses of rtl_arena_create makes use of that.)
However, with the combination of old ce906b8096081dee15dc8cc96e570d5b0b587955
"skip tricky allocators on G_SLICE=always-malloc" and recent
bc6a5d8e79e7d0e7d75ac107aa8e6aa275e434e9 "Disable custom allocator", rtl_cache
unconditionally just uses malloc/free now, so the rtl_arena_create
quantum_cache_max feature has effectively become irrelevant. As those changes
appear to have no negative consequences overall (and help remove quite a chunk
of no-longer used code), leave it at that and officially abandon the
quantum_cache_max feature.
Change-Id: I7d186a4a1589db6a73059c2be164aa81d81aef47
Reviewed-on: https://gerrit.libreoffice.org/54388
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
Tested-by: Stephan Bergmann <sbergman@redhat.com>
|
|
Change-Id: I90f72b6a618436d8cd2115be4e57a07bac15befc
Reviewed-on: https://gerrit.libreoffice.org/47161
Tested-by: Jenkins <ci@libreoffice.org>
Reviewed-by: Julien Nabet <serval2412@yahoo.fr>
|
|
This is necessary to avoid having extra threads
while forking. After forking, the second stage
of pre-init is started and so we start the stopped
rtl threads.
The comment for rtl_alloc_preInit_phase_t has
more details.
Change-Id: I1a3f7be74d4b04d0b2fc4a72b02124c2faa3c047
Reviewed-on: https://gerrit.libreoffice.org/47060
Tested-by: Jenkins <ci@libreoffice.org>
Reviewed-by: Ashod Nakashian <ashnakash@gmail.com>
|
|
This saves several megabytes of dirtied pages for each LOK
client of Online.
Change-Id: I425a2e7896879f0a64d71fcc0655e9e1fa1256aa
|
|
Change these back to consistently use the "..." form to include other UNO API
include files, for the benefit of external users of this API.
Change-Id: I9c9188e895eb3495e20a71ad44abfa2f6061fa94
|
|
Change-Id: I6074241dfcdcd4ca8a365ab6e03f01028d2c1a7b
Reviewed-on: https://gerrit.libreoffice.org/43344
Reviewed-by: Michael Stahl <mstahl@redhat.com>
Tested-by: Michael Stahl <mstahl@redhat.com>
|
|
Change-Id: Idb2e46fcaa080d6763d2e3ed963f7673a2353eb2
|
|
Change-Id: Id28046eb318cd3b2ed0b813fd266617547cf6ee2
|
|
...by adding some further SAL_DLLPUBLIC_RTTI type annotations (cf.
b4f6b26b5a1a78fecfa95ec2eb7ac8b80495d8aa "SAL_DLLPUBLIC_RTTI for proper RTTI
visibility for LLVM") and by making sure relevant function types do not use
incomplete types in their parameter and return types (which would make the RTTI
hidden).
Change-Id: Id7aadcbc0704b9759968ae36266fc9ce11a2e340
|
|
Change-Id: I677d973fbcf118111b5fc93b09143c66b0afb0d9
|
|
Change-Id: I248d002d2ed0e61f97a35ea0d329c64832252ad7
|
|
OpenCL devices require this else we would get a performance hit.
Change-Id: I6b1db6320fa84f933b6446022a0fd02ba267bf21
|
|
Change-Id: Ic90a365a237aa23846f97131146a5aa2c46b5fd2
|
|
Change-Id: If1ee11da444a7f96f2d8668b277540da0bb4dbe9
|
|
Change-Id: Ib48a12e902f2311c295b2007f08f44dee28f431d
Reviewed-on: https://gerrit.libreoffice.org/3499
Reviewed-by: David Tardon <dtardon@redhat.com>
Tested-by: David Tardon <dtardon@redhat.com>
|
