Stream: git-cranelift

Topic: cranelift / PR #1400 add a "raw" function definition inte...

GitHub (Feb 21 2020 at 17:15):

froydnj opened PR #1400 from raw-output to master:

• [X] This has been discussed in issue #1386.
• [X] A short description of what this does, why it is needed: for certain functions (lucet's stack probe code is the motivating example), generating them via cranelift is not necessarily feasible. Other compilers have recognized this and provided inline assembly for such cases. This interface can be thought of as a very primitive form of inline assembly (and a limited form; e.g. relocations are not currently supported). The overarching intent is to make cranelift-module's API rich enough that lucet can operate entirely within cranelift-module's API and not have to drop down to faerie or object. See #1386 for more details.
• [X] This PR does not contain test cases.
• [X] A reviewer from the core maintainer team has been assigned for this PR.

GitHub (Feb 21 2020 at 17:15):

froydnj requested iximeow for a review on PR #1400.

GitHub (Feb 21 2020 at 17:21):

bjorn3 submitted PR Review.

GitHub (Feb 21 2020 at 17:21):

bjorn3 submitted PR Review.

GitHub (Feb 21 2020 at 17:21):

bjorn3 created PR Review Comment:

Maybe extract this as a function to be used in define_function too.

GitHub (Feb 21 2020 at 17:21):

bjorn3 created PR Review Comment:

Nit: comments missing trailing dots.

GitHub (Feb 21 2020 at 17:21):

bjorn3 created PR Review Comment:

            relocs: vec![],

GitHub (Feb 21 2020 at 19:41):

pchickey submitted PR Review.

GitHub (Feb 21 2020 at 20:49):

froydnj updated PR #1400 from raw-output to master:

• [X] This has been discussed in issue #1386.
• [X] A short description of what this does, why it is needed: for certain functions (lucet's stack probe code is the motivating example), generating them via cranelift is not necessarily feasible. Other compilers have recognized this and provided inline assembly for such cases. This interface can be thought of as a very primitive form of inline assembly (and a limited form; e.g. relocations are not currently supported). The overarching intent is to make cranelift-module's API rich enough that lucet can operate entirely within cranelift-module's API and not have to drop down to faerie or object. See #1386 for more details.
• [X] This PR does not contain test cases.
• [X] A reviewer from the core maintainer team has been assigned for this PR.

GitHub (Feb 21 2020 at 20:52):

pchickey submitted PR Review.

GitHub (Feb 21 2020 at 20:53):

iximeow submitted PR Review.

GitHub (Feb 21 2020 at 22:09):

froydnj updated PR #1400 from raw-output to master:

• [X] This has been discussed in issue #1386.
• [X] A short description of what this does, why it is needed: for certain functions (lucet's stack probe code is the motivating example), generating them via cranelift is not necessarily feasible. Other compilers have recognized this and provided inline assembly for such cases. This interface can be thought of as a very primitive form of inline assembly (and a limited form; e.g. relocations are not currently supported). The overarching intent is to make cranelift-module's API rich enough that lucet can operate entirely within cranelift-module's API and not have to drop down to faerie or object. See #1386 for more details.
• [X] This PR does not contain test cases.
• [X] A reviewer from the core maintainer team has been assigned for this PR.

GitHub (Feb 21 2020 at 22:13):

bjorn3 submitted PR Review.

GitHub (Feb 21 2020 at 23:14):

pchickey merged PR #1400.

Last updated: Dec 06 2025 at 05:03 UTC