Hi lads,
Looking at rte_ring move_head functions I noticed that all of them
use slightly different approach to guarantee desired order of memory accesses:
1. rte_ring_generic_pvt.h:
=====================
pseudo-c-code // related armv8
instructions
--------------------
--------------------------------------
head.load() // ldr [head]
rte_smp_rmb() // dmb ishld
opposite_tail.load() // ldr [opposite_tail]
...
rte_atomic32_cmpset(head, ...) // ldrex[head];... stlex[head]
2. rte_ring_c11_pvt.h
=====================
pseudo-c-code // related armv8
instructions
--------------------
--------------------------------------
head.atomic_load(relaxed) // ldr[head]
atomic_thread_fence(acquire) // dmb ish
opposite_tail.atomic_load(acquire) // lda[opposite_tail]
...
head.atomic_cas(..., relaxed) // ldrex[haed]; ...
strex[head]
3. rte_ring_hts_elem_pvt.h
==========================
pseudo-c-code // related armv8
instructions
--------------------
--------------------------------------
head.atomic_load(acquire) // lda [head]
opposite_tail.load() // ldr [opposite_tail]
...
head.atomic_cas(..., acquire) // ldaex[head]; ...
strex[head]
The questions that arose from these observations:
a) are all 3 approaches equivalent in terms of functionality?
b) if yes, is there any difference in terms of performance between:
"ldr; dmb; ldr;" vs "lda; ldr;"
?
c) Comapring at 1) and 2) above, combination of
ldr [head]; dmb; lda [opposite_tail]:
looks like an overkill to me. Wouldn't just:
ldr [head]; dmb; ldr[opposite_tail];
be sufficient here?
I.E.- for reading tail value - we don't need to use load(acquire).
Or probably I do miss something obvious here?
Thanks
Konstantin
For convenience, I created a godbot page with all these variants:
https://godbolt.org/z/Yjj73b8xa
#1 - __rte_ring_headtail_move_head()
#2 - __rte_ring_headtail_move_head_c11_v1
#3 - __rte_ring_headtail_move_head_c11_v2
#2 with c) - __rte_ring_headtail_move_head_c11_v3
