DrSocalkwe3n/luajit
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Disable FMA by default. Use -Ofma or jit.opt.start("+fma") to enable.

Browse Source 
See the discussion in #918 for the rationale.
This commit is contained in:
Mike Pall
2022-12-07 18:38:22 +01:00
parent 7d5d4a1b1a
commit de2e1ca9d3
7 changed files with 35 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
doc/running.html
View File

@@ -220,6 +220,12 @@ mix the three forms, but note that setting an optimization level
overrides all earlier flags.
</p>
<p>
Note that <tt>-Ofma</tt> is not enabled by default at any level,
because it affects floating-point result accuracy. Only enable this,
if you fully understand the trade-offs of FMA for performance (higher),
determinism (lower) and numerical accuracy (higher).
</p>
<p>
Here are the available flags and at what optimization levels they
are enabled:
</p>
@@ -251,6 +257,8 @@ are enabled:
<td class="flag_name">sink</td><td class="flag_level">&nbsp;</td><td class="flag_level">&nbsp;</td><td class="flag_level">&bull;</td><td class="flag_desc">Allocation/Store Sinking</td></tr>
<tr class="even">
<td class="flag_name">fuse</td><td class="flag_level">&nbsp;</td><td class="flag_level">&nbsp;</td><td class="flag_level">&bull;</td><td class="flag_desc">Fusion of operands into instructions</td></tr>
<tr class="odd">
<td class="flag_name">fma </td><td class="flag_level">&nbsp;</td><td class="flag_level">&nbsp;</td><td class="flag_level">&nbsp;</td><td class="flag_desc">Fused multiply-add</td></tr>
</table>
<p>
Here are the parameters and their default settings: