docs/build.sh runs rst2html rather than rst2html.py
Explicitly documented that fact that ICC needs the -mmic flag to compile for KNC. Updated ISPC User Guide with details on ICC compiler options that impact FP performance in generated code.
This commit is contained in:
@@ -1,14 +1,14 @@
|
|||||||
#!/bin/bash
|
#!/bin/bash
|
||||||
|
|
||||||
for i in ispc perfguide faq; do
|
for i in ispc perfguide faq; do
|
||||||
rst2html.py --template=template.txt --link-stylesheet \
|
rst2html --template=template.txt --link-stylesheet \
|
||||||
--stylesheet-path=css/style.css $i.rst > $i.html
|
--stylesheet-path=css/style.css $i.rst > $i.html
|
||||||
done
|
done
|
||||||
|
|
||||||
rst2html.py --template=template-news.txt --link-stylesheet \
|
rst2html --template=template-news.txt --link-stylesheet \
|
||||||
--stylesheet-path=css/style.css news.rst > news.html
|
--stylesheet-path=css/style.css news.rst > news.html
|
||||||
|
|
||||||
rst2html.py --template=template-perf.txt --link-stylesheet \
|
rst2html --template=template-perf.txt --link-stylesheet \
|
||||||
--stylesheet-path=css/style.css perf.rst > perf.html
|
--stylesheet-path=css/style.css perf.rst > perf.html
|
||||||
|
|
||||||
#rst2latex --section-numbering --documentclass=article --documentoptions=DIV=9,10pt,letterpaper ispc.txt > ispc.tex
|
#rst2latex --section-numbering --documentclass=article --documentoptions=DIV=9,10pt,letterpaper ispc.txt > ispc.tex
|
||||||
|
|||||||
@@ -582,6 +582,10 @@ code generated by ``ispc``, setting the ``#include`` search path so that it
|
|||||||
can find the ``examples/intrinsics/knc.h`` header file in the ``ispc``
|
can find the ``examples/intrinsics/knc.h`` header file in the ``ispc``
|
||||||
distribution.
|
distribution.
|
||||||
|
|
||||||
|
::
|
||||||
|
|
||||||
|
icc -mmic -Iexamples/intrinsics/ foo.cpp -o foo.o
|
||||||
|
|
||||||
With the current beta implementation, complex ``ispc`` programs are able to
|
With the current beta implementation, complex ``ispc`` programs are able to
|
||||||
run on Xeon Phi, though there are a number of known limitations:
|
run on Xeon Phi, though there are a number of known limitations:
|
||||||
|
|
||||||
@@ -600,6 +604,15 @@ run on Xeon Phi, though there are a number of known limitations:
|
|||||||
where the memory address is actually aligned. This may unnecessarily
|
where the memory address is actually aligned. This may unnecessarily
|
||||||
impact performance.
|
impact performance.
|
||||||
|
|
||||||
|
* When requesting that ICC generate code with strict floating point
|
||||||
|
precision compliance (using ICC option ``-fp-model strict``) or
|
||||||
|
using of floating point exceptions (using ICC option ``-fp-model
|
||||||
|
except``) the compiler will generate code that uses the x87 unit
|
||||||
|
rather than KNC's vector unit. For similar reasons, the options
|
||||||
|
``–ansi`` and ``–fmath-errno`` may result in calls to math functions
|
||||||
|
that are implemented in x87 rather than KNC instructions. This will
|
||||||
|
have a significant performance impact.
|
||||||
|
|
||||||
All of these issues are currently actively being addressed and will be
|
All of these issues are currently actively being addressed and will be
|
||||||
fixed in future releases.
|
fixed in future releases.
|
||||||
|
|
||||||
|
|||||||
Reference in New Issue
Block a user