| commit | 954403938f00645d92520efc4433c440a133c0b9 | |
| author | Luboš Luňák <l.lunak@collabora.com> | |
| Tue, 19 Jun 2018 10:16:38 +0000 (19 12:16 +0200) | ||
| committer | Michael Meeks <michael.meeks@collabora.com> | |
| Thu, 21 Jun 2018 08:31:23 +0000 (21 10:31 +0200) | ||
| tree | e7e83437f9a55fc846bb546296f472c3e41aa289 | treesnapshot (tar.gz zip) |
| parent | f46f871f3170eb0d5a6522b7425c2390b601df5a | commitdiff |
discard cached cell values if the cell changes
FormulaGroupContext is actually a cache of cell results for OpenCL,
but the cached values are not always properly discarded. Happens e.g.
with testFormulaDepTracking in sc_ucalc fails if OpenCL is forced for
it (i.e. with mnOpenCLMinimumFormulaGroupSize disabled), because
a SetString() call for a cell doesn't invalidate the cache.
This obviously reduces the cache hit rate a bit, but according to my
tests it's not that bad (in fact the cache doesn't seem to get used
that often, so I even wonder if it's worth it).
Change-Id: Ia7ef2214956861d26ca3a42b84f9fecbff8316d0
Reviewed-on: https://gerrit.libreoffice.org/56087
Tested-by: Jenkins
Reviewed-by: Michael Meeks <michael.meeks@collabora.com>
FormulaGroupContext is actually a cache of cell results for OpenCL,
but the cached values are not always properly discarded. Happens e.g.
with testFormulaDepTracking in sc_ucalc fails if OpenCL is forced for
it (i.e. with mnOpenCLMinimumFormulaGroupSize disabled), because
a SetString() call for a cell doesn't invalidate the cache.
This obviously reduces the cache hit rate a bit, but according to my
tests it's not that bad (in fact the cache doesn't seem to get used
that often, so I even wonder if it's worth it).
Change-Id: Ia7ef2214956861d26ca3a42b84f9fecbff8316d0
Reviewed-on: https://gerrit.libreoffice.org/56087
Tested-by: Jenkins
Reviewed-by: Michael Meeks <michael.meeks@collabora.com>