ARM64's Weak Memory Model: Why x86 Code Fails on AArch64
ARM64 does not guarantee x86's store-load ordering. Learn how AArch64's weak memory model breaks plain shared-variable code and how to fix it with acquire/release atomics.
Loading…
ReadMeFeed / Community knowledge
#memory-model
Real questions. Useful conversations. Find the people who know your stack.
aarch64Lock-Free Code on ARM64: Why x86 Intuition Fails and What Replaces It
Moving x86 lock-free code to ARM64 often exposes hidden ordering assumptions. Learn why plain stores are not enough, how acquire/release primitives map to LDAR/STLR, and when ARMv8.1 LSE atomics are safe to use.
Diagnosing Memory Ordering Bugs when Porting to ARM64
Learn how to diagnose and fix subtle multithreading bugs when porting code from x86-64 to ARM64, focusing on the transition from a strong to a weak memory model.
aarch64 Memory Model: LDAR/STLR Visibility Limits Under Concurrent Access
Synchronization Constraints in Weakly Ordered Memory The ARMv8-A architecture utilizes a weakly ordered memory model where hardware may reorder memory accesses to maximize throughput. To manage synchronization, the architecture provides Load-Acquire (LDAR) and Store-Release (STLR) instructions, which establish one-way barriers to ensure specific ordering bet
Fortran coarray collectives: non-coarray write visibility across images without sync memory
Symptom Portable reasoning about data visibility after a collective subroutine (co_sum, co_max, co_min, co_reduce) is hindered because the Fortran 2018 standard specifies an implicit synchronization but does not mandate whether that synchronization acts as a full memory fence equivalent to sync memory . Context Current compilers diverge: GNU Fortran 12+ trea
Does a store-release followed by a load-acquire guarantee visibility of pagination offsets on AArch64 without extra barriers?
Goal Ensure that a consumer thread observing a readiness flag via a load-acquire also sees the pagination offset updated by a preceding store-release, preventing duplicate or skipped pages when bounding queries and paginating large datasets on AArch64. Constraints and uncertainty AArch64 follows a weak memory model where normal loads and stores may be reorde
LDAR/STLR Acquire-Release vs DMB barriers for aarch64 producer-consumer ordering
A design decision is needed for a producer-consumer handoff on aarch64 under the Weakly Ordered memory model. The goal is to guarantee visibility of data writes before the consumer observes the ready flag, without assuming x86-style Total Store Ordering. Two documented approaches are under consideration. Load-Acquire LDAR and Store-Release STLR provide one-w
OpenCL global memory visibility limits across work-groups without fences
Goal I want a producer work-group to write a result buffer in global memory and have a consumer work-group read that data reliably within the same kernel launch, without stale reads. Constraints and uncertainty My understanding of the OpenCL C memory model is that it is relaxed: writes to global memory by one work-item are not guaranteed visible to others wi
opencl