From 1be510f9529cb082f802408b472a77d074b394c0 Mon Sep 17 00:00:00 2001 From: Nicolas Vigier Date: Sun, 14 Apr 2013 13:46:12 +0000 Subject: Add zarb MLs html archives --- zarb-ml/mageia-dev/2013-March/023906.html | 103 ++++++++++++++++++++++++++++++ 1 file changed, 103 insertions(+) create mode 100644 zarb-ml/mageia-dev/2013-March/023906.html (limited to 'zarb-ml/mageia-dev/2013-March/023906.html') diff --git a/zarb-ml/mageia-dev/2013-March/023906.html b/zarb-ml/mageia-dev/2013-March/023906.html new file mode 100644 index 000000000..00ef33c7c --- /dev/null +++ b/zarb-ml/mageia-dev/2013-March/023906.html @@ -0,0 +1,103 @@ + + + + [Mageia-dev] Packages not rebuilt since Mageia 1 + + + + + + + + + +

[Mageia-dev] Packages not rebuilt since Mageia 1

+ eatdirt + dirteat at gmail.com +
+ Wed Mar 27 23:38:15 CET 2013 +

+
+ +
On 26/03/13 15:24, Pascal Terjan wrote:
+> There are 26 packages.
+>
+> atlas
+
+I think, and unless someone complains, this one should be dropped. Atlas 
+is supposed to provice fast routines already provided by blas and lapack 
+(that we have), but:
+
+http://math-atlas.sourceforge.net/atlas_install/node27.html
+
+says:
+
+"Many users ask how ATLAS can be used to build libraries that will run 
+on all x86 platforms. In general, this is a bad idea: ATLAS gets its 
+speed by specializing for particular platforms, so the more generic a 
+library is the less performance it will achieve!"
+
+and this is what we are doing. We even compile it with sseX options 
+which is soon worse than doing nothing:
+
+"
+Never use these libraries unless this portability is absolutely 
+required. They must use portable settings for blocking, for instance, 
+which will mean that on many platforms they will use only a fraction of 
+the actual cache, causing large performance drops. Even worse, peak 
+performance may by reduced by as much as factor of 8, due to not using 
+the proper ISA extension. The most portable ISA uses only the x87 unit, 
+which has a much lower peak rate on most modern machines (eg., an Intel 
+sandy bridge can do 16 flops/cycle using AVX in single precision, but 
+only 8 flops/cycle if using SSE, and only 2 flops/cycle using the x87 unit).
+"
+
+
+Cheers,
+Chris,
+
+
+ + + + + + + + + + + + + +
+

+ +
+More information about the Mageia-dev +mailing list
+ -- cgit v1.2.1