[LLVM] Trim intrinsics

[jurahul]

Rework intrinsic ID enums to pack target index and intrinsic index into the enum.

Add support for disabling support for a subset of intrinsics based on enabled targets in LLVM. This helps reduce code/data bloat for intrinsics that may not be exercised.

[github-actions]

✅ With the latest revision this PR passed the C/C++ code formatter.

[jrtc27]

This seems wrong. Emitting LLVM IR from the Clang frontend should not require anything about the set of targets enabled in LLVM itself.

[jrtc27]

"[LLVM] Trim intrinsics" is not a helpful commit message. What exactly are you trying to achieve and why? Is there an RFC for this given it seems to be having significant consequences across the tree?

[jurahul]

"[LLVM] Trim intrinsics" is not a helpful commit message. What exactly are you trying to achieve and why? Is there an RFC for this given it seems to be having significant consequences across the tree?

Yes, please see here: https://discourse.llvm.org/t/rfc-compress-intrinsic-name-table/82412/23

This is not yet the final PR so I did not add any commit message. The intent is to gather initial feedback and then split this PR into multiple ones for commit.

[jrtc27]

"[LLVM] Trim intrinsics" is not a helpful commit message. What exactly are you trying to achieve and why? Is there an RFC for this given it seems to be having significant consequences across the tree?

Yes, please see here: https://discourse.llvm.org/t/rfc-compress-intrinsic-name-table/82412/23

This is not yet the final PR so I did not add any commit message. The intent is to gather initial feedback and then split this PR into multiple ones for commit.

Please write that, especially including the RFC link, in future so anyone who didn't happen to see the Discourse thread go past but gets notified about the PR knows what on earth is going on.

[jurahul]

"[LLVM] Trim intrinsics" is not a helpful commit message. What exactly are you trying to achieve and why? Is there an RFC for this given it seems to be having significant consequences across the tree?

Yes, please see here: https://discourse.llvm.org/t/rfc-compress-intrinsic-name-table/82412/23

This is not yet the final PR so I did not add any commit message. The intent is to gather initial feedback and then split this PR into multiple ones for commit.

Please write that, especially including the RFC link, in future so anyone who didn't happen to see the Discourse thread go past but gets notified about the PR knows what on earth is going on.

Yes, agreed. This PR is not ready for review, I mistakenly marked it so and the reverted to draft. The split up PRs will have all those details.