p0nce

Hello, I'm using the latest v0.10.6 tag and depending upon sdlang-d is now emitting the following on build: ``` Warning: Sub package unit-threaded:from, referenced by unit-threaded:mocks 0.7.55 must be referenced...

Doing nothing vs going pragma(inline, true) vs function literals

1

https://github.com/ldc-developers/ldc/issues/4388#issuecomment-2002054510 Need to makes a well balanced decision for this! Surprising development. Problem statement: - currently we'll have no inlining unless dub --combined is used, which is a major problem...

AVX2

2

~~179 TODO~~ ~~159 TODO~~ ~~142 TODO~~ ~~132 TODO~~ ~~112 TODO~~ ~~100 TODO~~ ~~87 TODO~~ ~~76 TODO~~ ~~62 TODO~~ 49 TODO

Enabling AVX vectors in DMD D_SIMD

1

Right now = always emulated This will regress a ton, with the usual DMD backend drama

Evaluate faster _mm_pow_ss and _mm_exp_ss

1

Using vector instruction like that may well pessimize the performance. Try to get positive benchmark results out of a fully scalar path. Agner Fog's tables sometime show a lower cost...

DMD 2.102 bug on CI

5

no repro locally on Windows

Needed for PS Vita which is armv7, it already works, but non-officially

Add D_SIMD and GDC paths for internals.d functions: - [ ] `cmpss` - [ ] `cmpsd` - [ ] `cmppd` - [ ] `cmpps` - [ ] `cmppd256` - [...

Else this shows with DMD or GDC with AVX emulated vectors are not aligned, and we want to perform SSE aligned operations on them. Note that it won't work when...