llvm dev - Oct 2020 - TableGen enhancements

I just saw this post. I'm certainly looking for areas to improve. I will
investigate the example you provided.

Thanks Paul - Its awesome to have someone looking at tblgen build 
performance!

Something that I've looked at every so often is the poor choice of 
containers - e.g. passing a const std::vector& around when ArrayRef<> 
would be a lot cheaper (no expensive iterator bounds checks....).

Another is using map/set everywhere - they are great when data is being 
parsed/collected, but once its sorted transfering to an array/vector 
could be a lot cheaper if really we spend more time sequentially 
iterating across the data many times.

Simon.

On 20/10/2020 23:31, Paul C. Anagnostopoulos via llvm-dev
wrote:
> I just saw this post. I'm certainly looking for areas to improve. I
will investigate the example you provided.
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev

It will take me awhile to learn my way around the matcher table emitter, but it
sounds like a big piece of fun!

At 10/21/2020 05:48 AM, Simon Pilgrim wrote:
>Thanks Paul - Its awesome to have someone looking at tblgen build
performance!
>
>Something that I've looked at every so often is the poor choice of
containers - e.g. passing a const std::vector& around when ArrayRef<>
would be a lot cheaper (no expensive iterator bounds checks....).
>
>Another is using map/set everywhere - they are great when data is being
parsed/collected, but once its sorted transfering to an array/vector could be a
lot cheaper if really we spend more time sequentially iterating across the data
many times.
>
>Simon.

At 10/21/2020 05:48 AM, Simon Pilgrim wrote:
>Thanks Paul - Its awesome to have someone looking at tblgen build
performance!
I thought you would be interesting in the timing of the phases in the DAGISel
emitter. This is for the ARM target.

===-------------------------------------------------------------------------==  
TableGen phase timing
===-------------------------------------------------------------------------== 
Total Execution Time: 18.3457 seconds (18.3570 wall clock)

   ---User Time---   --System Time--   --User+System--   ---Wall Time---  ---
Name ---
   7.6128 ( 41.5%)   0.0000 (  0.0%)   7.6128 ( 41.5%)   7.6274 ( 41.6%) 
Convert to matchers
   5.1636 ( 28.2%)   0.0156 (100.0%)   5.1792 ( 28.2%)   5.1753 ( 28.2%)  Parse,
build records
   4.3680 ( 23.8%)   0.0000 (  0.0%)   4.3680 ( 23.8%)   4.3592 ( 23.7%)  Emit
matcher table
   0.8892 (  4.9%)   0.0000 (  0.0%)   0.8892 (  4.8%)   0.8891 (  4.8%)  Sort
patterns
   0.2964 (  1.6%)   0.0000 (  0.0%)   0.2964 (  1.6%)   0.3060 (  1.7%) 
Optimize matchers
  18.3301 (100.0%)   0.0156 (100.0%)  18.3457 (100.0%)  18.3570 (100.0%)  Total 

----------------------------------------------------------------
Windfall               Paul C. Anagnostopoulos 
      ----------------------------------------------------------
   Software            978 369-0839
                             www.windfall.com
----------------------------------------------------------------
My life has been filled with calamities,
some of which actually happened.
---Mark Twain 

Guga 'mzimba, sala 'nhliziyo