dmd codegen improvements

[tsbockman via Digitalmars-d]

On Tuesday, 18 August 2015 at 10:45:49 UTC, Walter Bright wrote:
> Martin ran some benchmarks recently that showed that ddmd 
> compiled with dmd was about 30% slower than when compiled with 
> gdc/ldc. This seems to be fairly typical.
>
> I'm interested in ways to reduce that gap.
>
> There are 3 broad kinds of optimizations that compilers do:
>
> 1. source translations like rewriting x*2 into x<<1, and 
> function inlining
>
> 2. instruction selection patterns like should one generate:
>
>     SETC AL
>     MOVZ EAX,AL
>
> or:
>     SBB EAX
>     NEG EAX
>
> 3. data flow analysis optimizations like constant propagation, 
> dead code elimination, register allocation, loop invariants, 
> etc.
>
> Modern compilers (including dmd) do all three.
>
> So if you're comparing code generated by dmd/gdc/ldc, and 
> notice something that dmd could do better at (1, 2 or 3), 
> please let me know. Often this sort of thing is low hanging 
> fruit that is fairly easily inserted into the back end.
>
> For example, recently I improved the usage of the SETcc 
> instructions.
>
> https://github.com/D-Programming-Language/dmd/pull/4901
> https://github.com/D-Programming-Language/dmd/pull/4904
>
> A while back I improved usage of BT instructions, the way 
> switch statements were implemented, and fixed integer divide by 
> a constant with multiply by its reciprocal.

I lack the assembly language skills to determine the cause(s) 
myself, but my 
[CheckedInt](https://github.com/tsbockman/CheckedInt) benchmark 
runs about 10x slower when compiled with DMD rather than GDC. I'm 
sure there's some low-hanging fruit in there somewhere...

Note that while it's far from being a minimal test case, the 
runtime code is nowhere near as complicated as it might appear at 
first - the vast majority of the complexity in the code is 
compile-time logic. I could produce a similar example without 
most of the compile-time obfuscation, if requested.

Also note that the speed difference has nothing to do with the 
use of core.checkedint intrinsics, as it was there before those 
were implemented in GDC.