String Switch Lowering

[Steven Schveighoffer]

On 1/25/18 1:41 PM, H. S. Teoh wrote:
> On Thu, Jan 25, 2018 at 07:21:29PM +0100, Benjamin Thaut via Digitalmars-d wrote:

>> If we think this is a good idea, should we rewrite this particular
>> string switch to use a associative array instead to avoid the overly
>> long symbol name?
> [...]
> 
> I believe the original idea behind using a template for string switches
> was to allow the possibility for the implementation to be smarter about
> how to implement the switch (IIRC, string switches used to be
> implemented as a runtime function). Supposedly object.__switch could
> analyze the list of strings at compile-time and generate a perfect hash
> or something, to maximize runtime performance.

I believe that when the number of cases is small enough, the binary 
search of the strings is done recursively to allow full optimization. 
And these symbols should be relatively small.

But I think if the number of strings is large enough, it calls the same 
runtime function as before (essentially). But here we have a problem -- 
the wrapper for this function is essentially this giant symbol that 
generates the string array locally, and then calls the real function. 
It's a huge cost just for a simple inline-able wrapper to another call.

See the code here:
https://github.com/dlang/druntime/blob/master/src/object.d#L3873

The function that does the real work is __switchSearch, which is only 
templated on the char type. It's going to be very very infrequent that 
the exact same switch case list is used in multiple places, meaning you 
are paying a huge cost for essentially memoizing these string lists.

I think we need some way to mark a function as following a different 
mangling, or maybe even just avoid the whole memoization, do everything 
inline.

-Steve