Neon: Implement incomplete function families pt. 2 #843
Conversation
Glitch18
force-pushed
the
func-fam-2
branch
5 times, most recently
from
2da8e3a
to
1ead33a
Compare
There was a problem hiding this comment.
Overall, excellent. I added a few notes, mostly minor stuff. There are a few coding style breaks which I'll clean up when I merge, no big deal.
I'm going to merge all except:
- cvt
- qshlu_n
- ld1_dup
- st2_lane (due to the CI failure)
Those have issues which I'd prefer to have fixed before they are merged, but none should be particularly troublesome. qshlu_n might be ok to merge as-is, but I want to confirm that it doesn't cause problems for some compilers first.
|
I also dropped |
Glitch18
force-pushed
the
func-fam-2
branch
7 times, most recently
from
2efcf6f
to
66a73ca
Compare
simde/arm/neon/ld1_dup.h
Outdated
| simde_float64x1_t | ||
| simde_vld1_dup_f64(simde_float64 const * ptr) { | ||
| #if defined(SIMDE_ARM_NEON_A32V7_NATIVE) | ||
| return vreinterpret_f64_s64(vld1_dup_s64(ptr)); |
There was a problem hiding this comment.
You're going to need a reienterpret_cast to convert to inst64_t const * or you end up with a -Wincompatible-pointer-types
Glitch18
force-pushed
the
func-fam-2
branch
2 times, most recently
from
b128196
to
dfe22b5
Compare
This PR completes the following function families (WIP) :