Re: Any instructions on optimizing LCC?
- From: "Ira Baxter" <idbaxter@xxxxxxxxxxxxxx>
- Date: Sat, 2 Jun 2007 14:03:05 -0500
<baikeina@xxxxxxxxxxxx> wrote in message
news:1179729721.064601.324290@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Hi everyone:
I am asked to work on optimizing LCC for our own DSP chip,now mainly
to add generation of SIMD instructions in LCC .but LCC parse each
statement and represent it in tree ,then DAG, but SIMD needs the
analyze of multiple statements ,and some data structures like data
dependence graph and basic blocks ,use-define chain ,...are absent in
LCC ,so anyone have some idea of SIMD optimization for LCC ?
do I need to add the implementation of data dependence graph .etc .in
LCC if I want SIMD?
Thanks
I can't speak for LCC, because I don't know completely what is in it.
But in general if you want to do sophisticated code generation, you need
various types of flow analyzers and pretty sophisticated optimization
mechanisms that cross statement boundaries.
You can learn a little about a SIMD prototype we built for C++
for Electronic Arts. We found that to do SIMD well, one
really needed a language in which array operations were explicit,
so we extended C++ accordingly, and then generated SIMD code.
http://dataparallel.googlegroups.com/web/Electronics%20Art%20CGO%20DP%202007.pdf?gda=232OMFAAAAC71WKfsr8UdI2FEj4jNOY77HNTeaJV1XcQwgEn4MdeGGG1qiJ7UbTIup-M2XPURDQ1PKyNLczZ7-1yw4mSQxbYJz1-zCGDhmRQkbuglwv3ag&hl=en
--
Ira Baxter, CTO
www.semanticdesigns.com
.
- Prev by Date: Re: Progress in data processing
- Next by Date: Porting LCC for 6502 ISA, anyone has experiences?
- Previous by thread: yacc trouble
- Next by thread: Porting LCC for 6502 ISA, anyone has experiences?
- Index(es):
Relevant Pages
|
|