[Ur] TechEmpower Benchmarks

[Adam Chlipala]

On 11/29/2013 12:37 PM, escalier at riseup.net wrote:
>> Do you know what value the 'max_threads' parameter is being 
>> instantiated with in practice?
>
> Based on the logs, it looks like 8. But I would imagine it will vary 
> from platform to platform in the actual benchmark.

I wonder if it would be better to multiply that parameter by 2 to get 
the Ur/Web thread count, then.  I believe the benchmark server includes 
8 hardware threads, and I'm used to doubling the hardware thread count 
to get the software thread count, to exploit more opportunities for 
overlapping IO and computation.

This might not be worth fussing with now. *shrug*

>
>> List.mapM -> List.app
>
> (I briefly looked for such a function when initially writing the code. 
> Just for your information, in Haskell, `app` is known as `mapM_` 
> following a convention that a trailing underscore discards the result 
> of a monadic action.)

Interesting convention!  The Ur/Web version is named following the 
Standard ML convention, where the same function has the same name, 
modulo lack of an explicit monad.