This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[DOC PATCH]: PR target/34702: Document limitations of recip sequences for x86
- From: "Uros Bizjak" <ubizjak at gmail dot com>
- To: "GCC Patches" <gcc-patches at gcc dot gnu dot org>
- Cc: "Dominique Dhumieres" <dominiq at lps dot ens dot fr>
- Date: Tue, 8 Jan 2008 11:00:31 +0100
- Subject: [DOC PATCH]: PR target/34702: Document limitations of recip sequences for x86
Hello!
This patch documents limitations of recip sequences for x86 targets.
OK for mainline?
2008-01-08 Uros Bizjak <ubizjak@gmail.com>
PR target/34702
* doc/invoke.texi (i386 and x86-64 Options) [mrecip]: Document
limitations of reciprocal sequences on x86 targets.
Uros.
Index: invoke.texi
===================================================================
--- invoke.texi (revision 131392)
+++ invoke.texi (working copy)
@@ -10647,9 +10647,14 @@
@item -mrecip
@opindex mrecip
This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
-vectorized variants RCPPS and RSQRTPS) instead of DIVSS and SQRTSS (and their
-vectorized variants). These instructions will be generated only when
-@option{-funsafe-math-optimizations} is enabled.
+vectorized variants RCPPS and RSQRTPS) with additional Newton-Rhapson step
+to increase precision instead of DIVSS and SQRTSS (and their vectorized
+variants) for single precision floating point arguments. These instructions
+are generated only when @option{-funsafe-math-optimizations} is enabled
+together with @option{-finite-math-only} and @option{-fno-trapping-math}.
+Note that while the throughput of the sequence is higher than the throughput
+of the non-reciprocal instruction, the precision of the sequence can be
+decreased up to 2 ulp.
@item -mveclibabi=@var{type}
@opindex mveclibabi