[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Gomp-discuss] Line of action ..
From: |
Lars Segerlund |
Subject: |
[Gomp-discuss] Line of action .. |
Date: |
Mon, 10 Feb 2003 17:11:07 +0100 |
User-agent: |
Mozilla/5.0 (X11; U; Linux i586; en-US; rv:1.2.1) Gecko/20021226 Debian/1.2.1-9 |
Well, Steven agreed to the suggestion that we first implement explicit
parallism in gcc, and then move on to se what it becomes.
So as for now, what we have to do to get on with it is :
1. Recognize and parse the pragmas.
2. Figure out what we actually need as a minimal subset in the
tree-nodes.
3. Figure out a 'minimal' or 'primitive' subset of omp
4. Investigate how to interact with the code generation and
optimization, ( not to do certain optimizations across parallell
boundaries ).
My opinion is that if we can do the parallell stuff, why get all
worked up about parsing a few pragmas ? Same goes for the g95 frontend.
What we have to work out is when we can resolve these statements ( omp
), and I believe that we must have a parsed tree of the following block
( it has to be a block ) before we get started on actually resolving
variables and such.
Point 4 is the one that wories me most, as I have no knowledge of gcc
optimizations.
/ Lars Segerlund.
- [Gomp-discuss] Line of action ..,
Lars Segerlund <=