diff options
| author | Ben Gamari <ben@smart-cactus.org> | 2019-09-27 18:49:06 +0000 |
|---|---|---|
| committer | Ben Gamari <ben@smart-cactus.org> | 2020-10-24 21:01:18 -0400 |
| commit | edb4b92b218cee5b309866f3d236da30c5621567 (patch) | |
| tree | 13db4eade35c006e066f838f48de57f1d626355a /includes | |
| parent | bf1b0bc78da7dbe5f6fbda54b37a9cb165ff857f (diff) | |
| download | haskell-edb4b92b218cee5b309866f3d236da30c5621567.tar.gz | |
rts/WSDeque: Rewrite with proper atomicswip/tsan/wsdeque
After a few attempts at shoring up the previous implementation, I ended
up turning to the literature and now use the proven implementation,
> N.M. Lê, A. Pop, A.Cohen, and F.Z. Nardelli. "Correct and Efficient
> Work-Stealing for Weak Memory Models". PPoPP'13, February 2013,
> ACM 978-1-4503-1922/13/02.
Note only is this approach formally proven correct under C11 semantics
but it is also proved to be a bit faster in practice.
Diffstat (limited to 'includes')
| -rw-r--r-- | includes/stg/SMP.h | 15 |
1 files changed, 15 insertions, 0 deletions
diff --git a/includes/stg/SMP.h b/includes/stg/SMP.h index fa52a913c4..4679d7986e 100644 --- a/includes/stg/SMP.h +++ b/includes/stg/SMP.h @@ -453,6 +453,14 @@ load_load_barrier(void) { // Non-atomic addition for "approximate" counters that can be lossy #define NONATOMIC_ADD(ptr,val) RELAXED_STORE(ptr, RELAXED_LOAD(ptr) + val) +// Explicit fences +// +// These are typically necessary only in very specific cases (e.g. WSDeque) +// where the ordered operations aren't expressive enough to capture the desired +// ordering. +#define RELEASE_FENCE() __atomic_thread_fence(__ATOMIC_RELEASE) +#define SEQ_CST_FENCE() __atomic_thread_fence(__ATOMIC_SEQ_CST) + /* ---------------------------------------------------------------------- */ #else /* !THREADED_RTS */ @@ -479,6 +487,10 @@ EXTERN_INLINE void load_load_barrier () {} /* nothing */ // Non-atomic addition for "approximate" counters that can be lossy #define NONATOMIC_ADD(ptr,val) *ptr += val +// Fences +#define RELEASE_FENCE() +#define SEQ_CST_FENCE() + #if !IN_STG_CODE || IN_STGCRUN INLINE_HEADER StgWord xchg(StgPtr p, StgWord w) @@ -527,6 +539,9 @@ atomic_dec(StgVolatilePtr p) } #endif +/* An alias for the C11 declspec */ +#define ATOMIC + #define VOLATILE_LOAD(p) ((StgWord)*((StgWord*)(p))) #endif /* !THREADED_RTS */ |