Route threads-per-wave through a single source of truth
(DeviceProperties.warp_size, from torch.cuda.get_device_properties) and
thread it through the config-generation helpers, replacing _NUM_THREADS_PER_WARP,
cc_warp_size, and scattered warp_size or 32 / torch.version.hip branches.
Behavior change: _num_warps halves num_warps only when warp_size == 64
(AMD CDNA/gfx9), not on every HIP device; AMD RDNA (wave32) now follows
the NVIDIA path.
Authored with: Claude
Pull Request resolved:
Community-Analysen & Experten-Meinungen 0
Verwandte Story-Cluster & Quellen (Vektor-KI)
Ähnliche Beiträge
Auch interessante Nachrichten trunk/521f9d784733cb2f4571804b0f6e4687fb7a91be: [inductor] Unify threads-per-wave (warp_size) extraction (#181112)
Thematisch verwandte Begriffe: trunk521f9d784733cb2f4571804b0f6e4687fb7a91be, inductor, Unify, threadsperwave · 6 Treffer
Apple’s $129 AirPods 5 bring noise cancellation to the cheap seats — and the $149 pair is the one to buy
Apple releases first iOS 27.2, iPadOS 27.2, watchOS 27.2, tvOS 27.2, and visionOS 27.2 betas
Videos werden geladen ...
Beiträge werden geladen ...
Videos werden geladen ...
Beiträge werden geladen ...
Videos werden geladen ...
Beiträge werden geladen ...
Videos werden geladen ...
Beiträge werden geladen ...
Videos werden geladen ...
SOCIAL SHARE CARD GENERATOR