AccelerateHS / AccelerateHS/accelerate

New-pipeline: Make reference counting explicit in UniformSchedule

未关闭
#569 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看
主要语言
Haskell
星标
1k
派生
135
PR 合并指标
30 天内没有已合并 PR

描述

We are using reference counting for memory management on new-pipeline. The reference counting logic is currently hardcoded for the Native backend. To also use reference counting on the PTX backend, we should perform reference counting in the shared code instead of in the backend code. We can do that by making reference counting explicit in the last shared representation: UniformSchedule.

贡献指南

这个仓库没有索引到贡献指南

评估

这个 Issue 还没有评估数据。

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。