Hi,
this patch adds some notes on LTO/IPA changes and some statistics
on bulid-time/memory use improvements I collected this weekend.
I also added some of FDO changes I rememebr. Martin, perhaps you can
do more.
Honza
Index: changes.html
===================================================================
RCS file: /cvs/gcc/wwwdocs/htdocs/gcc-9/changes.html,v
retrieving revision 1.59
diff -u -r1.59 changes.html
--- changes.html 9 Apr 2019 21:08:28 -0000 1.59
+++ changes.html 30 Apr 2019 17:19:05 -0000
@@ -195,6 +195,11 @@
metadata such as the inlining chain, and profile information (if
available).
</li>
+ <li>Inter-procedural propagation of stack alignment can now be controlled by
+ <a
href="https://gcc.gnu.org/onlinedocs/gcc/Developer-Options.html#index-fipa-stack-alignment";>-fipa-stack-alignment</code></a>.
+ <li>Propagation of addressability, readonly and writeonly flags on
+ static variables can now be controlled by
+ <a
href="https://gcc.gnu.org/onlinedocs/gcc/Developer-Options.html#index-fipa-reference-addressable";>-fipa-reference-addressable</code></a>.
</ul>
<p>The following built-in functions have been introduced.</p>
<ul>
@@ -246,6 +251,52 @@
can be transformed into <code>100 * how + 5</code> (for values defined
in the switch statement).
</li>
+<li>Inter-prodcedural optimization improvements:
+ <ul>
+ <li>Inliner defaults was tuned to better suits modern C++ codebases
+ especially when built with link time optimizations.
+ New parameters <code>max-inline-insns-small</code>,
+ <code>max-inline-insns-size</code>,
+ <code>uninlined-function-insns</code>,
+ <code>uninlined-function-time</code>,
<code>uninlined-thunk-insns</code>,
+ and <code>uninlined-thunk-time</code> was added.</li>
+ <li>Hot/cold partitioning is now more precise and agressive.</li>
+ <li>Improved scalability for very large translation units (especially
+ when link-time optimizing large programs).</li>
+ </ul>
+<li>Profile driven optimization improvements:
+ <ul>
+ <li><code>-fprofile-use</code> now enables
+ <code>-fversion-loops-for-strides</code>,
+ <code>-floop-interchange</code>,
+ <code>-funroll-and-jam</code>,
+ <code>-ftree-loop-distribution</code>.</li>
+ <li>Streaming of counter histograms was removed. This reduces
+ the size of profile files. Histogram is computed on the fly
+ with link-time optimization.
+ Parameter <code>hot-bb-count-ws-permille</code> was reduced
+ from 999 to 990 to account for more precise histograms.</li>
+ </ul>
+<li>Link-time optimization improvements:
+ <ul>
+ <li>Types are now simplified prior streaming resulting in significant
+ reductions of the LTO object files, link-time memory use, and
+ improvements of link-time parallelism.</li>
+ <li>Default number of partitions (<code>--param lto-partitions</code>) was
+ increased from 32 to 128 enabling effective use of CPUs with more than
+ 32 hyperthreads. <code>--param lto-max-streaming-parallelism</code>
+ can now be used to control number of streaming processes.</li>
+ <li>Warnings on C++ One Decl Rule violations (<code>-Wodr</code>) are
+ now more informative and produce fewer redundant results.</li>
+ </ul>
+ Overal compile time of Firefox and LibreOffice was reduced by about 5%
+ compared to GCC 8.3. Size of LTO object files is reduced by 7%.
+ LTO link-time improves by 11% on 8-core machine and scales significantly
better
+ for more parallel build environments. Serial stage of the link-time
+ optimization is 28% faster consuming 20% less memory.
+ Parallel stage now partitions to 128 partitions rather than 32 and
+ reduces memory use for every worker by 30%.
+ </li>
</ul>
<p>The following improvements to the <code>gcov</code> command-line utility
have been made.</p>
