[refactor] Remove redundant codegen of BitExtractStmt

[strongoier]

Issue: #6134

[netlify]

✅ Deploy Preview for docsite-preview canceled.

Name	Link
🔨 Latest commit	b32a8f5
🔍 Latest deploy log	https://app.netlify.com/sites/docsite-preview/deploys/632c2d324ac26100086af250

[bobcao3]

There is actually hardware support on both Nvidia and AMD for native bit extract (a lot faster than using lower level bit operations). Maybe we should keep them not demoted?

[strongoier]

There is actually hardware support on both Nvidia and AMD for native bit extract (a lot faster than using lower level bit operations). Maybe we should keep them not demoted?

Could you elaborate a bit on what should we do to make use of that? Regarding this PR, demote_operations is just an abstract layer of what current codegen of different backends does.

[bobcao3]

There is actually hardware support on both Nvidia and AMD for native bit extract (a lot faster than using lower level bit operations). Maybe we should keep them not demoted?

Could you elaborate a bit on what should we do to make use of that? Regarding this PR, demote_operations is just an abstract layer of what current codegen of different backends does.

We would want to be able to optionally demote BFE when backend says it's not supported, and if supported the codegen will want to take the BFE CHI-IR and generate the backend native instruction

[strongoier]

We would want to be able to optionally demote BFE when backend says it's not supported, and if supported the codegen will want to take the BFE CHI-IR and generate the backend native instruction

Sounds good. Then I think not performing demotion in CHI IR is a better choice. I'll now close this PR and submit another one in the future.