Add support for tail calls #1090
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
There are a few tests (including this one) which are currently failing because the target function has undergone TailCallElim (as part of optimize_module), thus the refined calls get tail'd. Should we bail out from |
nunoplopes
reviewed
Sep 29, 2024
| @@ -5,7 +5,7 @@ target triple = "x86_64-unknown-unknown" | |||
| declare i32 @memcmp(ptr nocapture, ptr nocapture, i64) | |||
|
|
|||
| define i32 @src(ptr nocapture readonly %x, ptr nocapture readonly %y) { | |||
| %call = tail call i32 @memcmp(ptr %x, ptr %y, i64 2) | |||
There was a problem hiding this comment.
revert this change; it should be fine?
nunoplopes
reviewed
Sep 29, 2024
| ret void | ||
| } | ||
|
|
||
| define void @tgt() { | ||
| tail call void (ptr, ...) @g(ptr null) | ||
| call void (ptr, ...) @g(ptr null) |
nunoplopes
reviewed
Sep 29, 2024
| @@ -24,7 +24,7 @@ for.body: ; preds = %entry, %for.body | |||
|
|
|||
| for.end: ; preds = %for.body | |||
| %3 = load i32, ptr @b, align 4 | |||
| %call = tail call i32 (ptr, ...) @print(ptr nonnull dereferenceable(1) getelementptr inbounds ([15 x i8], ptr @.str, i64 0, i64 0), i32 %3) | |||
| %call = call i32 (ptr, ...) @print(ptr nonnull dereferenceable(1) getelementptr inbounds ([15 x i8], ptr @.str, i64 0, i64 0), i32 %3) | |||
nunoplopes
reviewed
Sep 29, 2024
| @@ -74,7 +74,7 @@ for.end: ; preds = %for.end_us_lcssa, % | |||
| store i32 1, ptr @d, align 4 | |||
| store i32 %inc.lcssa, ptr @b, align 4 | |||
| store i32 %conv.lcssa, ptr @f, align 4 | |||
| %call = tail call i32 (ptr, ...) @myprintf(ptr getelementptr inbounds ([4 x i8], ptr @.str, i64 0, i64 0), i32 %conv.lcssa) | |||
| %call = call i32 (ptr, ...) @myprintf(ptr getelementptr inbounds ([4 x i8], ptr @.str, i64 0, i64 0), i32 %conv.lcssa) | |||
nunoplopes
reviewed
Sep 29, 2024
| @@ -39,7 +39,7 @@ if.end: ; preds = %delete.end, %entry | |||
| define i32 @tgt(ptr %bp) { | |||
| entry: | |||
| %add.ptr = getelementptr inbounds i32, ptr %bp, i64 1 | |||
| tail call void @_ZdaPv(ptr %add.ptr) | |||
| call void @_ZdaPv(ptr %add.ptr) | |||
nunoplopes
reviewed
Sep 29, 2024
| public: | ||
| Memcmp(Type &type, std::string &&name, Value &ptr1, Value &ptr2, Value &num, | ||
| bool is_bcmp): MemInstr(type, std::move(name)), ptr1(&ptr1), | ||
| ptr2(&ptr2), num(&num), is_bcmp(is_bcmp) {} | ||
| bool is_bcmp, bool is_tailcall): MemInstr(type, std::move(name)), ptr1(&ptr1), |
nunoplopes
reviewed
Sep 29, 2024
| @@ -171,6 +171,9 @@ class State { | |||
| // The value of a 'returned' input | |||
| std::optional<StateValue> returned_input; | |||
|
|
|||
| // Used to keep track of the current instruction | |||
| const Value *current_value = nullptr; | |||
There was a problem hiding this comment.
this doesn't seem needed. Instructions can pass *this to check_tailcall.
nunoplopes
reviewed
Sep 29, 2024
|
|
||
| auto it = s.getFn().bbOf(*call).instrs().begin(); | ||
| bool found = false; | ||
| for (auto e = s.getFn().bbOf(*call).instrs().end(); it != e; ++it) { |
There was a problem hiding this comment.
pls don't call bbOf() twice. you can have a single loop to fetch the instruction after the call.
|
I think overall this is the right direction, thanks! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Introduce support for tail calls: any subsequent access in the caller is UB after a tail call-site (bool only approach).