<antocuni>
a very short summary is that we worked on two main things
<antocuni>
1. the HPyField branch
<antocuni>
ah, maybe 3
<antocuni>
2. ronan's hpy-refactor-exceptions pypy branch. We concluded that we can probably merge it because the performance penalty is very small and/or inexistent
<antocuni>
3. investigate pypy performance on the piconumpy benchmark. It turned out that pypy is very suboptimal w.r.t allocations
<antocuni>
the issue has been somewhat treacherously to arigato in the hope that he feels the need to solve it 😅
<antocuni>
*has been assigned
<mattip>
about the double allocation: isn't this similar to the cpyext strategy for bytes and unicode (where we allocate an extra bufffer for the data)?