(original) (raw)
I think that you need two things:
1\. A direct-as-possible mapping in between the DAG and the stub backend instructions
2\. The features of the backend are all configurable (legal datatypes and instructions, what is promoted to what, etc.) so that we can test as much of the legalizer's parameter space as possible.
-Hal
Sent from my Verizon Wireless Droid
1\. A direct-as-possible mapping in between the DAG and the stub backend instructions
2\. The features of the backend are all configurable (legal datatypes and instructions, what is promoted to what, etc.) so that we can test as much of the legalizer's parameter space as possible.
-Hal
Sent from my Verizon Wireless Droid
-----Original message-----
From: Anton Korobeynikov>>> Are there any plans to this?
To: Jakob Stoklund Olesen
Cc: llvmdev@cs.uiuc.edu
Sent: Tue, Jul 10, 2012 22:09:01 GMT+00:00
Subject: Re: \[LLVMdev\] New backend
>>> I'd like to see a 'stub' backend in the llvm-tree (or somewhere else).
>> Well, I asked on the list previously, but noone ever responded….
> Don't targets like MSP430 and MBlaze serve as good stubs?
It's a good question... I think people are looking for something
minimal, so they already know that there are no "unnecessary" stuff...
--
With best regards, Anton Korobeynikov
Faculty of Mathematics and Mechanics, Saint Petersburg State University
\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_
LLVM Developers mailing list
LLVMdev@cs.uiuc.edu http://llvm.cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev