RE: [wwwdocs] gcc-17: Create porting_to.html and document the AVX10.2 scalar SAT CVT rename
"Hu, Lin1" <[email protected]>
| Newsgroups | gmane.comp.gcc.patches |
|---|---|
| Message-ID | <PH3PPF801A91A7DD6DF89DE2A161926020AA6DE2@PH3PPF801A91A7D.namprd11.prod.outlook.com> |
Kindly ping. > -----Original Message----- > From: Hu, Lin1 <[email protected]> > Sent: Monday, August 3, 2026 4:48 PM > To: [email protected] > Cc: [email protected]; [email protected]; [email protected]; > [email protected]; Liu, Hongtao <[email protected]>; [email protected] > Subject: [wwwdocs] gcc-17: Create porting_to.html and document the AVX10.2 > scalar SAT CVT rename > > Hi, > > This follows up on the PR126581 discussion [1]. > > r16-9456 renamed the AVX10.2 scalar saturating conversion intrinsics from the > _ep[i|u]{32,64} suffix to _[i|u]{32,64}. The old names are kept on the release > branches but are gone in GCC 17, so document what to use instead. > > Ok for wwwdocs? > > [1] https://gcc.gnu.org/pipermail/gcc-patches/2026-August/726127.html > > BRs, > Lin > > --- > htdocs/gcc-17/changes.html | 15 ++++-- > htdocs/gcc-17/porting_to.html | 98 +++++++++++++++++++++++++++++++++++ > 2 files changed, 110 insertions(+), 3 deletions(-) create mode 100644 > htdocs/gcc-17/porting_to.html > > diff --git a/htdocs/gcc-17/changes.html b/htdocs/gcc-17/changes.html index > 3c3a24bb..9ae53b86 100644 > --- a/htdocs/gcc-17/changes.html > +++ b/htdocs/gcc-17/changes.html > @@ -17,11 +17,9 @@ > <p> > This page is a "brief" summary of some of the huge number of improvements in > GCC 17. > -<!-- > You may also want to check out our > <a href="porting_to.html">Porting to GCC 17</a> page and the <a > href="../onlinedocs/index.html#current">full GCC documentation</a>. > ---> > </p> > > <p>Note: GCC 17 has not been released yet, so this document is @@ -186,7 > +184,18 @@ for general information.</p> > </li> > </ul> > > -<!-- <h3 id="x86">IA-32/x86-64</h3> --> > +<h3 id="x86">IA-32/x86-64</h3> > + > +<ul> > + <li>The AVX10.2 scalar saturating conversion intrinsics have been renamed > + from the <code>_ep[i|u]{32,64}</code> suffix, which denotes packed > + elements, to <code>_[i|u]{32,64}</code>, with <code>_si{32,64}</code> > + aliases for the signed variants. The old names were never accepted by > + other x86 compilers. See the > + <a href="porting_to.html#x86_avx10_2_scalar_satcvt">porting guide</a> > + for the list of affected intrinsics. > + </li> > +</ul> > > <!-- <h3 id="mips">MIPS</h3> --> > > diff --git a/htdocs/gcc-17/porting_to.html b/htdocs/gcc-17/porting_to.html new > file mode 100644 index 00000000..4a65fd52 > --- /dev/null > +++ b/htdocs/gcc-17/porting_to.html > @@ -0,0 +1,98 @@ > +<!DOCTYPE html> > +<html lang="en"> > + > +<head> > +<meta http-equiv="Content-Type" content="text/html; charset=utf-8"> > +<title>Porting to GCC 17</title> <link rel="stylesheet" type="text/css" > +href="https://gcc.gnu.org/gcc.css"> > +</head> > + > +<body> > +<h1>Porting to GCC 17</h1> > + > +<p> > +The GCC 17 release series differs from previous GCC releases in <a > +href="changes.html">a number of ways</a>. Some of these are a result of > +bug fixing, and some old behaviors have been intentionally changed to > +support new standards, or relaxed in standards-conforming ways to > +facilitate compilation or run-time performance. > +</p> > + > +<p> > +Some of these changes are user visible and can cause grief when porting > +to GCC 17. This document is an effort to identify common issues > +and provide solutions. Let us know if you have suggestions for improvements! > +</p> > + > +<h2 id="x86">IA-32/x86-64 issues</h2> > + > +<h3 id="x86_avx10_2_scalar_satcvt">Renamed AVX10.2 scalar saturating > +conversion intrinsics</h3> > + > +<p>The AVX10.2 scalar saturating conversion intrinsics were originally > +named with an <code>_ep[i|u]{32,64}</code> suffix. That suffix denotes > +packed elements, whereas these intrinsics convert a single scalar value > +and return a scalar integer, so the names did not follow the convention > +used by scalar conversion intrinsics since AVX-512F. The old names > +were never accepted by other x86 compilers. GCC 17 renames them to use > +the <code>_[i|u]{32,64}</code> suffix, with <code>_si{32,64}</code> > +aliases for the signed variants:</p> > + > +<table> > + <tr> > + <th>Removed in GCC 17</th> <th>Use instead</th> > + </tr> > + <tr> > + <td><code>_mm_cvtts_sd_epi32</code></td> > +<td><code>_mm_cvtts_sd_i32</code>, <code>_mm_cvtts_sd_si32</code></td> > + </tr> > + <tr> > + <td><code>_mm_cvtts_sd_epu32</code></td> > +<td><code>_mm_cvtts_sd_u32</code></td> > + </tr> > + <tr> > + <td><code>_mm_cvtts_ss_epi32</code></td> > +<td><code>_mm_cvtts_ss_i32</code>, <code>_mm_cvtts_ss_si32</code></td> > + </tr> > + <tr> > + <td><code>_mm_cvtts_ss_epu32</code></td> > +<td><code>_mm_cvtts_ss_u32</code></td> > + </tr> > + <tr> > + <td><code>_mm_cvtts_roundsd_epi32</code></td> > +<td><code>_mm_cvtts_roundsd_i32</code>, > +<code>_mm_cvtts_roundsd_si32</code></td> > + </tr> > + <tr> > + <td><code>_mm_cvtts_roundsd_epu32</code></td> > +<td><code>_mm_cvtts_roundsd_u32</code></td> > + </tr> > + <tr> > + <td><code>_mm_cvtts_roundss_epi32</code></td> > +<td><code>_mm_cvtts_roundss_i32</code>, > +<code>_mm_cvtts_roundss_si32</code></td> > + </tr> > + <tr> > + <td><code>_mm_cvtts_roundss_epu32</code></td> > +<td><code>_mm_cvtts_roundss_u32</code></td> > + </tr> > + <tr> > + <td><code>_mm_cvtts_sd_epi64</code></td> > +<td><code>_mm_cvtts_sd_i64</code>, <code>_mm_cvtts_sd_si64</code></td> > + </tr> > + <tr> > + <td><code>_mm_cvtts_sd_epu64</code></td> > +<td><code>_mm_cvtts_sd_u64</code></td> > + </tr> > + <tr> > + <td><code>_mm_cvtts_ss_epi64</code></td> > +<td><code>_mm_cvtts_ss_i64</code>, <code>_mm_cvtts_ss_si64</code></td> > + </tr> > + <tr> > + <td><code>_mm_cvtts_ss_epu64</code></td> > +<td><code>_mm_cvtts_ss_u64</code></td> > + </tr> > + <tr> > + <td><code>_mm_cvtts_roundsd_epi64</code></td> > +<td><code>_mm_cvtts_roundsd_i64</code>, > +<code>_mm_cvtts_roundsd_si64</code></td> > + </tr> > + <tr> > + <td><code>_mm_cvtts_roundsd_epu64</code></td> > +<td><code>_mm_cvtts_roundsd_u64</code></td> > + </tr> > + <tr> > + <td><code>_mm_cvtts_roundss_epi64</code></td> > +<td><code>_mm_cvtts_roundss_i64</code>, > +<code>_mm_cvtts_roundss_si64</code></td> > + </tr> > + <tr> > + <td><code>_mm_cvtts_roundss_epu64</code></td> > +<td><code>_mm_cvtts_roundss_u64</code></td> > + </tr> > +</table> > + > +<p>The old names are still accepted on the GCC 15 and GCC 16 release > +branches, so code can be updated before switching to GCC 17.</p> > + > +</body> > +</html> > -- > 2.31.1