[Gmp-commit] /var/hg/www: 2 new changesets

[mercurial at gmplib.org]

details:   /var/hg/www/rev/a1005124c71a
changeset: 77:a1005124c71a
user:      Torbjorn Granlund <tege at gmplib.org>
date:      Mon May 27 18:55:24 2013 +0200
description:
Describe stalling problems according to current code status.

details:   /var/hg/www/rev/d4257ae6ab7e
changeset: 78:d4257ae6ab7e
user:      Torbjorn Granlund <tege at gmplib.org>
date:      Mon May 27 18:55:36 2013 +0200
description:
Mark an item as done.

diffstat:

 devel/GMPng.html |   4 ++--
 devel/sparc.html |  10 +++++-----
 2 files changed, 7 insertions(+), 7 deletions(-)

diffs (55 lines):

diff -r a28632dbef68 -r d4257ae6ab7e devel/GMPng.html
--- a/devel/GMPng.html	Mon May 27 18:54:10 2013 +0200
+++ b/devel/GMPng.html	Mon May 27 18:55:36 2013 +0200
@@ -340,7 +340,7 @@
   grained tuning, without trying to invent ever stranger configure CPU names.
   </font>
 
-  <font color="#a00000">
+  <font color="#00a000">
   <p> Make GMP_CPU_TYPE fat CPU selection standard for a fat build (but perhaps
   rename it to something more specific, GMP_FAT_CPU_TYPE_SELECT).  Motive:
   Testability.
@@ -407,7 +407,7 @@
 
 </div>
 
-<font size="-4">Last modified: 2013-05-13 </font>
+<font size="-4">Last modified: 2013-05-23 </font>
 <div id="footer">
 <table cellpadding=0 width="100%" bgcolor="#e8e8e8">
   <tr>
diff -r a28632dbef68 -r d4257ae6ab7e devel/sparc.html
--- a/devel/sparc.html	Mon May 27 18:54:10 2013 +0200
+++ b/devel/sparc.html	Mon May 27 18:55:36 2013 +0200
@@ -85,9 +85,8 @@
 <h3> T4-T5 projects </h3>
 <p>
 <ul>
-  <li> Schedule mpn/sparc64/ultrasparct3/aormul_2.asm to cut two cycles from
-    the loop.  This should be quite easy; just bubble mulx/umulxhi dependees
-    downwards a few lines.
+  <li> Several mul primitives run one or two cycles slower per iteration than
+    anticipated.  It does not seem to be directly latency scheduling related.
   </li>
   <li> Explore using the mpmul instruction for mpn_mul_basecase.  Since mpmul
     handles just same-size operands, a GMP {up,un} × {vp,vn} multiply (where
@@ -103,7 +102,8 @@
     result.
   </li>
   <li> Consider using the mpmul instruction for mpn_sqr_basecase, for
-    large-enough operands.
+    large-enough operands.  Unlike for mul_basecase, this will need just a
+    simple cutoff point to a discrete sqr_bascase loop.
   </li>
   <li> Implement dual-limb inverse "pi2" Euclidean and Hensel division
     primitives.  This will double small-divisor division performance, since the
@@ -139,7 +139,7 @@
 </div>
 
 <div id="footer">
-<font size="-4">Last modified: 2013-05-04 </font>
+<font size="-4">Last modified: 2013-05-22 </font>
 <table cellpadding=0 width="100%" bgcolor="#e8e8e8">
   <tr>
     <td align="center">