[llvm] r204542 - [TableGen] Don't assert, produce an error, when an instruction has too few operands

Duncan P. N. Exon Smith dexonsmith at apple.com
Sat Mar 22 19:48:11 PDT 2014


On 2014 Mar 22, at 18:27, Hal Finkel <hfinkel at anl.gov> wrote:

> ----- Original Message -----
>> From: "Duncan P. N. Exon Smith" <dexonsmith at apple.com>
>> To: "Hal Finkel" <hfinkel at anl.gov>
>> Cc: llvm-commits at cs.uiuc.edu
>> Sent: Saturday, March 22, 2014 6:55:38 PM
>> Subject: Re: [llvm] r204542 - [TableGen] Don't assert, produce an error, when an instruction has too few operands
>> 
>> testcase?
> 
> We currently gave a grand total of zero independent test cases for the encoder generator, and, I believe, zero test cases for TableGen error outputs. This is certainly worth improving, but if all goes well, the entire support for positional operands (including this functionality) will soon disappear,

Fair enough.

> so I did not want to hold up this improvement while I built up a good set of tests for the encoder in general.
> 
> The next step, however, is to make the named matching support complex operands, which will allow all targets to switch completely away from positional operand matching. As I do this, I do intend to introduce a reasonable set of independent regression tests.

Great!




More information about the llvm-commits mailing list