aboutsummaryrefslogtreecommitdiff
path: root/doc
diff options
context:
space:
mode:
authorMike Pall <mike>2022-12-07 18:38:22 +0100
committerMike Pall <mike>2022-12-07 18:38:22 +0100
commitde2e1ca9d3d87e74c0c20c1e4ad3c32b31a5875b (patch)
treec6dd3a9575b77c9f019c9d8627a814a955227acf /doc
parent7d5d4a1b1a690d9fc87253868ba967bf25f4df6e (diff)
downloadluajit-de2e1ca9d3d87e74c0c20c1e4ad3c32b31a5875b.tar.gz
luajit-de2e1ca9d3d87e74c0c20c1e4ad3c32b31a5875b.tar.bz2
luajit-de2e1ca9d3d87e74c0c20c1e4ad3c32b31a5875b.zip
Disable FMA by default. Use -Ofma or jit.opt.start("+fma") to enable.
See the discussion in #918 for the rationale.
Diffstat (limited to 'doc')
-rw-r--r--doc/running.html8
1 files changed, 8 insertions, 0 deletions
diff --git a/doc/running.html b/doc/running.html
index 9979d223..edc049fb 100644
--- a/doc/running.html
+++ b/doc/running.html
@@ -220,6 +220,12 @@ mix the three forms, but note that setting an optimization level
220overrides all earlier flags. 220overrides all earlier flags.
221</p> 221</p>
222<p> 222<p>
223Note that <tt>-Ofma</tt> is not enabled by default at any level,
224because it affects floating-point result accuracy. Only enable this,
225if you fully understand the trade-offs of FMA for performance (higher),
226determinism (lower) and numerical accuracy (higher).
227</p>
228<p>
223Here are the available flags and at what optimization levels they 229Here are the available flags and at what optimization levels they
224are enabled: 230are enabled:
225</p> 231</p>
@@ -251,6 +257,8 @@ are enabled:
251<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> 257<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>
252<tr class="even"> 258<tr class="even">
253<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> 259<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>
260<tr class="odd">
261<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>
254</table> 262</table>
255<p> 263<p>
256Here are the parameters and their default settings: 264Here are the parameters and their default settings: