Path: csiph.com!optima2.xanadu-bbs.net!xanadu-bbs.net!usenet.blueworldhosting.com!feeder01.blueworldhosting.com!border2.nntp.dca1.giganews.com!nntp.giganews.com!news.iecc.com!.POSTED!nerds-end
From: Hans-Peter Diettrich <DrDiettrich1@netscape.net>
Newsgroups: comp.compilers
Subject: Re: Maximal Munch instruction selection: how to connect tiles?
Date: Tue, 29 Sep 2015 17:16:27 +0200
Organization: Compilers Central
Lines: 28
Sender: news@iecc.com
Approved: comp.compilers@iecc.com
Message-ID: <15-09-026@comp.compilers>
References: <15-09-024@comp.compilers>
NNTP-Posting-Host: news.iecc.com
Mime-Version: 1.0
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
X-Trace: miucha.iecc.com 1443541993 67504 2001:470:1f07:1126:0:676f:7373:6970 (29 Sep 2015 15:53:13 GMT)
X-Complaints-To: abuse@iecc.com
NNTP-Posting-Date: Tue, 29 Sep 2015 15:53:13 +0000 (UTC)
Keywords: code, optimize
Posted-Date: 29 Sep 2015 11:53:13 EDT
X-submission-address: compilers@iecc.com
X-moderator-address: compilers-request@iecc.com
X-FAQ-and-archives: http://compilers.iecc.com
Xref: csiph.com comp.compilers:1621

CC)sar schrieb:
> Given the expression a = b + c; my compiler currently produces the
> following IR-tree:
>
>     =
>   /     \
> a       + (_t1)
>        /    \
>      b      c
>  ...

> I am trying to use the maximal munch strategy to produce assembly code
> for this tree but I'm stuck with the following questions: ...

You can consider many more cases, like the x86 addressing modes with
base, index and scaling, which can be used for some arithmetic
expressions in general.

In case the '+' operator stores the result in memory, a '+=' opcode (add
to memory) could be used in above expression. All that can end up in
multiple possible different instruction sequences, which have to be
"weighted" for the final selection of the "best" sequence. The sequences
then can map to different tree structures, which are created from the
AST. That means that modifications should be applied to copies of the
AST, and it's unpredictable how big the different trees for more complex
expressions will become.

DoDi