-
Notifications
You must be signed in to change notification settings - Fork 14.3k
[Clang] Demote always_inline error to warning for mismatching SME attrs #100740
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Clang] Demote always_inline error to warning for mismatching SME attrs #100740
Conversation
PR llvm#77936 introduced a diagnostic to avoid calls being inlined into functions with a different streaming mode, because inlining those functions may result in different runtime behaviour. This was necessary because LLVM doesn't check whether inlining is possible and thus blindly inlines the function without checking feasibility. In practice however, this introduces an artificial restriction that the user may not be able to work around. Calling an `always_inline` function from some header file that is out of the control of the user would result in an error that the user cannot remedy. Therefore, this patch demotes the error into a warning (for calls from streaming[-compatible] -> non-streaming), but the proper fix would be to fix the AlwaysInliner in LLVM to avoid inlining when it has analyzed the callee and has determined that inlining is not possible. Calling an always_inline function for calls from non-streaming -> streaming will remain an error, because there is little pre-existing code for SME, so it is expected that the header file can be modified by the user (e.g. by using `__arm_streaming_compatible` if the code is claimed to be compatible).
@llvm/pr-subscribers-clang-codegen @llvm/pr-subscribers-clang Author: Sander de Smalen (sdesmalen-arm) ChangesPR #77936 introduced a diagnostic to avoid calls being inlined into functions with a different streaming mode, because inlining those functions may result in different runtime behaviour. This was necessary because LLVM doesn't check whether inlining is possible and thus blindly inlines the function without checking feasibility. In practice however, this introduces an artificial restriction that the user may not be able to work around. Calling an Therefore, this patch demotes the error into a warning (for calls from streaming[-compatible] -> non-streaming), but the proper fix would be to fix the AlwaysInliner in LLVM to avoid inlining when it has analyzed the callee and has determined that inlining is not possible. Calling an always_inline function for calls from non-streaming -> streaming will remain an error, because there is little pre-existing code for SME, so it is expected that the header file can be modified by the user (e.g. by using Full diff: https://github.com/llvm/llvm-project/pull/100740.diff 3 Files Affected:
diff --git a/clang/include/clang/Basic/DiagnosticFrontendKinds.td b/clang/include/clang/Basic/DiagnosticFrontendKinds.td
index 12a4617c64d87..04f87f5217067 100644
--- a/clang/include/clang/Basic/DiagnosticFrontendKinds.td
+++ b/clang/include/clang/Basic/DiagnosticFrontendKinds.td
@@ -288,6 +288,9 @@ def err_function_needs_feature : Error<
let CategoryName = "Codegen ABI Check" in {
def err_function_always_inline_attribute_mismatch : Error<
"always_inline function %1 and its caller %0 have mismatching %2 attributes">;
+def warn_function_always_inline_attribute_mismatch : Warning<
+ "always_inline function %1 and its caller %0 have mismatching %2 attributes, "
+ "inlining may result in different runtime behaviour">, InGroup<AArch64SMEAttributes>;
def err_function_always_inline_new_za : Error<
"always_inline function %0 has new za state">;
diff --git a/clang/lib/CodeGen/Targets/AArch64.cpp b/clang/lib/CodeGen/Targets/AArch64.cpp
index b9df54b0c67c4..1dec3cd40ebd2 100644
--- a/clang/lib/CodeGen/Targets/AArch64.cpp
+++ b/clang/lib/CodeGen/Targets/AArch64.cpp
@@ -883,8 +883,10 @@ void AArch64TargetCodeGenInfo::checkFunctionCallABIStreaming(
if (!CalleeIsStreamingCompatible &&
(CallerIsStreaming != CalleeIsStreaming || CallerIsStreamingCompatible))
- CGM.getDiags().Report(CallLoc,
- diag::err_function_always_inline_attribute_mismatch)
+ CGM.getDiags().Report(
+ CallLoc, CalleeIsStreaming
+ ? diag::err_function_always_inline_attribute_mismatch
+ : diag::warn_function_always_inline_attribute_mismatch)
<< Caller->getDeclName() << Callee->getDeclName() << "streaming";
if (auto *NewAttr = Callee->getAttr<ArmNewAttr>())
if (NewAttr->isNewZA())
diff --git a/clang/test/CodeGen/aarch64-sme-inline-streaming-attrs.c b/clang/test/CodeGen/aarch64-sme-inline-streaming-attrs.c
index 25aebeced9379..2c565f6225468 100644
--- a/clang/test/CodeGen/aarch64-sme-inline-streaming-attrs.c
+++ b/clang/test/CodeGen/aarch64-sme-inline-streaming-attrs.c
@@ -20,7 +20,7 @@ void caller(void) {
#ifdef TEST_COMPATIBLE
void caller_compatible(void) __arm_streaming_compatible {
- inlined_fn(); // expected-error {{always_inline function 'inlined_fn' and its caller 'caller_compatible' have mismatching streaming attributes}}
+ inlined_fn(); // expected-warning {{always_inline function 'inlined_fn' and its caller 'caller_compatible' have mismatching streaming attributes, inlining may result in different runtime behaviour}}
inlined_fn_streaming_compatible();
inlined_fn_streaming(); // expected-error {{always_inline function 'inlined_fn_streaming' and its caller 'caller_compatible' have mismatching streaming attributes}}
inlined_fn_local(); // expected-error {{always_inline function 'inlined_fn_local' and its caller 'caller_compatible' have mismatching streaming attributes}}
@@ -29,7 +29,7 @@ void caller_compatible(void) __arm_streaming_compatible {
#ifdef TEST_STREAMING
void caller_streaming(void) __arm_streaming {
- inlined_fn(); // expected-error {{always_inline function 'inlined_fn' and its caller 'caller_streaming' have mismatching streaming attributes}}
+ inlined_fn(); // expected-warning {{always_inline function 'inlined_fn' and its caller 'caller_streaming' have mismatching streaming attributes, inlining may result in different runtime behaviour}}
inlined_fn_streaming_compatible();
inlined_fn_streaming();
inlined_fn_local();
@@ -39,7 +39,7 @@ void caller_streaming(void) __arm_streaming {
#ifdef TEST_LOCALLY
__arm_locally_streaming
void caller_local(void) {
- inlined_fn(); // expected-error {{always_inline function 'inlined_fn' and its caller 'caller_local' have mismatching streaming attributes}}
+ inlined_fn(); // expected-warning {{always_inline function 'inlined_fn' and its caller 'caller_local' have mismatching streaming attributes, inlining may result in different runtime behaviour}}
inlined_fn_streaming_compatible();
inlined_fn_streaming();
inlined_fn_local();
|
At a glance, this seems like a good use case for a warning + |
I'd rather not make that change, because in practice it will be unlikely that inlining will result in changes to runtime behaviour. We'd rather not burden the users with explicit errors for something that may be fine, as it's more a limitation of Clang that it has to emit this message in the first place. Separate to that, it might not be directly clear to users that the error can be demoted into a warning. Thanks for the suggestion though! |
Fair enough, that makes sense |
4ddaab2
to
377b19c
Compare
/cherry-pick 5430f73 |
…rs (llvm#100740) PR llvm#77936 introduced a diagnostic to avoid calls being inlined into functions with a different streaming mode, because inlining those functions may result in different runtime behaviour. This was necessary because LLVM doesn't check whether inlining is possible and thus blindly inlines the function without checking feasibility. In practice however, this introduces an artificial restriction that the user may not be able to work around. Calling an `always_inline` function from some header file that is out of the control of the user would result in an error that the user cannot remedy. Therefore, this patch demotes the error into a warning (for calls from streaming[-compatible] -> non-streaming), but the proper fix would be to fix the AlwaysInliner in LLVM to avoid inlining when it has analyzed the callee and has determined that inlining is not possible. Calling an always_inline function for calls from non-streaming -> streaming will remain an error, because there is little pre-existing code for SME, so it is expected that the header file can be modified by the user (e.g. by using `__arm_streaming_compatible` if the code is claimed to be compatible). (cherry picked from commit 5430f73)
/pull-request #100987 |
PR #77936 introduced a diagnostic to avoid calls being inlined into functions with a different streaming mode, because inlining those functions may result in different runtime behaviour. This was necessary because LLVM doesn't check whether inlining is possible and thus blindly inlines the function without checking feasibility.
In practice however, this introduces an artificial restriction that the user may not be able to work around. Calling an
always_inline
function from some header file that is out of the control of the user would result in an error that the user cannot remedy.Therefore, this patch demotes the error into a warning (for calls from streaming[-compatible] -> non-streaming), but the proper fix would be to fix the AlwaysInliner in LLVM to avoid inlining when it has analyzed the callee and has determined that inlining is not possible.
Calling an always_inline function for calls from non-streaming -> streaming will remain an error, because there is little pre-existing code for SME, so it is expected that the header file can be modified by the user (e.g. by using
__arm_streaming_compatible
if the code is claimed to be compatible).