[javascript] [cuda] Only spawn memory allocator daemon when CUDA is enabled (JS 8/n)

[AmesingFlank]

Related issue = #3781

The daemon is only used for allocation memory in CUDA. (Actually, do we even use that in CUDA anymore, we have switched completely to pre-allocation right?)

[vercel]

This pull request is being automatically deployed with Vercel (learn more).
To see the status of your deployment, click below or on the icon next to each commit.

🔍 Inspect: https://vercel.com/taichi-api-docs-preview/taichi/CBUrXq9UziwubVpZhYZ68LhsZK7u
✅ Preview: Canceled

[Deployment for 46d300e canceled]

[netlify]

✔️ Deploy Preview for docsite-preview ready!

🔨 Explore the source changes: 46d300e

🔍 Inspect the deploy log: https://app.netlify.com/sites/docsite-preview/deploys/61f0b900c6cbc200070ae79f

😎 Browse the preview: https://deploy-preview-4115--docsite-preview.netlify.app

[qiao-bo]

Memory allocation in CUDA/CPU has been switched to using preallocated. The memory pool is not used IIUC. ;/

[AmesingFlank]

Memory allocation in CUDA/CPU has been switched to using preallocated. The memory pool is not used IIUC. ;/

Right! Thanks for confirming.

Actually I think this memory pool is still used for allocating things like host-side result buffer. But we should consider removing the daemon entirely.

Created #4135 to track this.

[ailzhang]

LGTM!

[AmesingFlank]

CI on macOS keeps failing.. interesting. Not sure why this change would affect macOS. Will investigate.

[feisuzhu]

Closing ancient PR