maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/clang/test/SemaCUDA/call-host-fn-from-device.cu

141 lines
4.4 KiB
Plaintext
Raw Normal View History

[CUDA] Emit deferred diagnostics during Sema rather than during codegen. Summary: Emitting deferred diagnostics during codegen was a hack. It did work, but usability was poor, both for us as compiler devs and for users. We don't codegen if there are any sema errors, so for users this meant that they wouldn't see deferred errors if there were any non-deferred errors. For devs, this meant that we had to carefully split up our tests so that when we tested deferred errors, we didn't emit any non-deferred errors. This change moves checking for deferred errors into Sema. See the big comment in SemaCUDA.cpp for an overview of the idea. This checking adds overhead to compilation, because we have to maintain a partial call graph. As a result, this change makes deferred errors a CUDA-only concept (whereas before they were a general concept). If anyone else wants to use this framework for something other than CUDA, we can generalize at that time. This patch makes the minimal set of test changes -- after this lands, I'll go back through and do a cleanup of the tests that we no longer have to split up. Reviewers: rnk Subscribers: cfe-commits, rsmith, tra Differential Revision: https://reviews.llvm.org/D25541 llvm-svn: 284158
2016-10-14 04:52:12 +08:00
// RUN: %clang_cc1 %s --std=c++11 -triple nvptx-unknown-unknown -fcuda-is-device \
[CUDA] When we emit an error that might have been deferred, also print a callstack. Summary: Previously, when you did something not allowed in a host+device function and then caused it to be codegen'ed, we would print out an error telling you that you did something bad, but we wouldn't tell you how we decided that the function needed to be codegen'ed. This change causes us to print out a callstack when emitting deferred errors. This is immensely helpful when debugging highly-templated code, where it's often unclear how a function became known-emitted. We only print the callstack once per function, after we print the all deferred errors. This patch also switches all of our hashtables to using canonical FunctionDecls instead of regular FunctionDecls. This prevents a number of bugs, some of which are caught by tests added here, in which we assume that two FDs for the same function have the same pointer value. Reviewers: rnk Subscribers: cfe-commits, tra Differential Revision: https://reviews.llvm.org/D25704 llvm-svn: 284647
2016-10-20 05:15:01 +08:00
// RUN: -emit-llvm -o /dev/null -verify -verify-ignore-unexpected=note
[CUDA] Raise an error if a wrong-side call is codegen'ed. Summary: Some function calls in CUDA are allowed to appear in semantically-correct programs but are an error if they're ever codegen'ed. Specifically, a host+device function may call a host function, but it's an error if such a function is ever codegen'ed in device mode (and vice versa). Previously, clang made no attempt to catch these errors. For the most part, they would be caught by ptxas, and reported as "call to unknown function 'foo'". Now we catch these errors and report them the same as we report other illegal calls (e.g. a call from a host function to a device function). This has a small change in error-message behavior for calls that were previously disallowed (e.g. calls from a host to a device function). Previously, we'd catch disallowed calls fairly early, before doing additional semantic checking e.g. of the call's arguments. Now we catch these illegal calls at the very end of our semantic checks, so we'll only emit a "illegal CUDA call" error if the call is otherwise well-formed. Reviewers: tra, rnk Subscribers: cfe-commits Differential Revision: https://reviews.llvm.org/D23242 llvm-svn: 278759
2016-08-16 07:00:49 +08:00
// Note: This test won't work with -fsyntax-only, because some of these errors
// are emitted during codegen.
#include "Inputs/cuda.h"
extern "C" void host_fn() {}
[CUDA] Simplify some repeated diagnostic expectations in CUDA tests. Instead of repeating the diagnostic, use "expected-note N". Test-only change. llvm-svn: 284882
2016-10-22 04:50:47 +08:00
// expected-note@-1 7 {{'host_fn' declared here}}
[CUDA] Do a better job at detecting wrong-side calls. Summary: Move CheckCUDACall from ActOnCallExpr and BuildDeclRefExpr to DiagnoseUseOfDecl. This lets us catch some edge cases we were missing, specifically around class operators. This necessitates a few other changes: - Avoid emitting duplicate deferred diags in CheckCUDACall. Previously we'd carefully placed our call to CheckCUDACall such that it would only ever run once for a particular callsite. But now this isn't the case. - Emit deferred diagnostics from a template specialization/instantiation's primary template, in addition to from the specialization/instantiation itself. DiagnoseUseOfDecl ends up putting the deferred diagnostics on the template, rather than the specialization, so we need to check both. Reviewers: rsmith Subscribers: cfe-commits, tra Differential Revision: https://reviews.llvm.org/D24573 llvm-svn: 283637
2016-10-08 09:07:11 +08:00
struct Dummy {};
[CUDA] Raise an error if a wrong-side call is codegen'ed. Summary: Some function calls in CUDA are allowed to appear in semantically-correct programs but are an error if they're ever codegen'ed. Specifically, a host+device function may call a host function, but it's an error if such a function is ever codegen'ed in device mode (and vice versa). Previously, clang made no attempt to catch these errors. For the most part, they would be caught by ptxas, and reported as "call to unknown function 'foo'". Now we catch these errors and report them the same as we report other illegal calls (e.g. a call from a host function to a device function). This has a small change in error-message behavior for calls that were previously disallowed (e.g. calls from a host to a device function). Previously, we'd catch disallowed calls fairly early, before doing additional semantic checking e.g. of the call's arguments. Now we catch these illegal calls at the very end of our semantic checks, so we'll only emit a "illegal CUDA call" error if the call is otherwise well-formed. Reviewers: tra, rnk Subscribers: cfe-commits Differential Revision: https://reviews.llvm.org/D23242 llvm-svn: 278759
2016-08-16 07:00:49 +08:00
struct S {
S() {}
[CUDA] Simplify some repeated diagnostic expectations in CUDA tests. Instead of repeating the diagnostic, use "expected-note N". Test-only change. llvm-svn: 284882
2016-10-22 04:50:47 +08:00
// expected-note@-1 2 {{'S' declared here}}
[CUDA] Raise an error if a wrong-side call is codegen'ed. Summary: Some function calls in CUDA are allowed to appear in semantically-correct programs but are an error if they're ever codegen'ed. Specifically, a host+device function may call a host function, but it's an error if such a function is ever codegen'ed in device mode (and vice versa). Previously, clang made no attempt to catch these errors. For the most part, they would be caught by ptxas, and reported as "call to unknown function 'foo'". Now we catch these errors and report them the same as we report other illegal calls (e.g. a call from a host function to a device function). This has a small change in error-message behavior for calls that were previously disallowed (e.g. calls from a host to a device function). Previously, we'd catch disallowed calls fairly early, before doing additional semantic checking e.g. of the call's arguments. Now we catch these illegal calls at the very end of our semantic checks, so we'll only emit a "illegal CUDA call" error if the call is otherwise well-formed. Reviewers: tra, rnk Subscribers: cfe-commits Differential Revision: https://reviews.llvm.org/D23242 llvm-svn: 278759
2016-08-16 07:00:49 +08:00
~S() { host_fn(); }
[CUDA] Fix "declared here" note on deferred wrong-side errors. Previously we weren't deferring these "declared here" notes, which is obviously wrong. llvm-svn: 278767
2016-08-16 08:48:21 +08:00
// expected-note@-1 {{'~S' declared here}}
[CUDA] Raise an error if a wrong-side call is codegen'ed. Summary: Some function calls in CUDA are allowed to appear in semantically-correct programs but are an error if they're ever codegen'ed. Specifically, a host+device function may call a host function, but it's an error if such a function is ever codegen'ed in device mode (and vice versa). Previously, clang made no attempt to catch these errors. For the most part, they would be caught by ptxas, and reported as "call to unknown function 'foo'". Now we catch these errors and report them the same as we report other illegal calls (e.g. a call from a host function to a device function). This has a small change in error-message behavior for calls that were previously disallowed (e.g. calls from a host to a device function). Previously, we'd catch disallowed calls fairly early, before doing additional semantic checking e.g. of the call's arguments. Now we catch these illegal calls at the very end of our semantic checks, so we'll only emit a "illegal CUDA call" error if the call is otherwise well-formed. Reviewers: tra, rnk Subscribers: cfe-commits Differential Revision: https://reviews.llvm.org/D23242 llvm-svn: 278759
2016-08-16 07:00:49 +08:00
int x;
};
struct T {
__host__ __device__ void hd() { host_fn(); }
// expected-error@-1 {{reference to __host__ function 'host_fn' in __host__ __device__ function}}
// No error; this is (implicitly) inline and is never called, so isn't
// codegen'ed.
__host__ __device__ void hd2() { host_fn(); }
__host__ __device__ void hd3();
void h() {}
[CUDA] Fix "declared here" note on deferred wrong-side errors. Previously we weren't deferring these "declared here" notes, which is obviously wrong. llvm-svn: 278767
2016-08-16 08:48:21 +08:00
// expected-note@-1 {{'h' declared here}}
[CUDA] Do a better job at detecting wrong-side calls. Summary: Move CheckCUDACall from ActOnCallExpr and BuildDeclRefExpr to DiagnoseUseOfDecl. This lets us catch some edge cases we were missing, specifically around class operators. This necessitates a few other changes: - Avoid emitting duplicate deferred diags in CheckCUDACall. Previously we'd carefully placed our call to CheckCUDACall such that it would only ever run once for a particular callsite. But now this isn't the case. - Emit deferred diagnostics from a template specialization/instantiation's primary template, in addition to from the specialization/instantiation itself. DiagnoseUseOfDecl ends up putting the deferred diagnostics on the template, rather than the specialization, so we need to check both. Reviewers: rsmith Subscribers: cfe-commits, tra Differential Revision: https://reviews.llvm.org/D24573 llvm-svn: 283637
2016-10-08 09:07:11 +08:00
void operator+();
// expected-note@-1 {{'operator+' declared here}}
void operator-(const T&) {}
// expected-note@-1 {{'operator-' declared here}}
operator Dummy() { return Dummy(); }
// expected-note@-1 {{'operator Dummy' declared here}}
Aligned allocation versus CUDA: make deallocation function preference order match other CUDA preference orders, per discussion with jlebar. We now model this in an attempt to match overload resolution as closely as possible: - First, we throw out all non-callable (due to CUDA host/device mismatch) operator delete functions. - Then we apply sizedness / alignedness preferences based on whether the type is overaligned and whether the deallocation function is a member. - Finally, we use the CUDA callability preference as a tiebreaker. llvm-svn: 283830
2016-10-11 08:21:10 +08:00
[CUDA] Ignore uncallable functions when we check for usual deallocators. Previously clang considered function variants from both sides of compilation and that resulted in picking up wrong deallocation function. Differential Revision: https://reviews.llvm.org/D51808 llvm-svn: 342749
2018-09-22 01:29:33 +08:00
__host__ void operator delete(void *) { host_fn(); };
__device__ void operator delete(void*, __SIZE_TYPE__);
Aligned allocation versus CUDA: make deallocation function preference order match other CUDA preference orders, per discussion with jlebar. We now model this in an attempt to match overload resolution as closely as possible: - First, we throw out all non-callable (due to CUDA host/device mismatch) operator delete functions. - Then we apply sizedness / alignedness preferences based on whether the type is overaligned and whether the deallocation function is a member. - Finally, we use the CUDA callability preference as a tiebreaker. llvm-svn: 283830
2016-10-11 08:21:10 +08:00
};
struct U {
[CUDA] Ignore uncallable functions when we check for usual deallocators. Previously clang considered function variants from both sides of compilation and that resulted in picking up wrong deallocation function. Differential Revision: https://reviews.llvm.org/D51808 llvm-svn: 342749
2018-09-22 01:29:33 +08:00
__device__ void operator delete(void*, __SIZE_TYPE__) = delete;
Aligned allocation versus CUDA: make deallocation function preference order match other CUDA preference orders, per discussion with jlebar. We now model this in an attempt to match overload resolution as closely as possible: - First, we throw out all non-callable (due to CUDA host/device mismatch) operator delete functions. - Then we apply sizedness / alignedness preferences based on whether the type is overaligned and whether the deallocation function is a member. - Finally, we use the CUDA callability preference as a tiebreaker. llvm-svn: 283830
2016-10-11 08:21:10 +08:00
__host__ __device__ void operator delete(void*);
[CUDA] Raise an error if a wrong-side call is codegen'ed. Summary: Some function calls in CUDA are allowed to appear in semantically-correct programs but are an error if they're ever codegen'ed. Specifically, a host+device function may call a host function, but it's an error if such a function is ever codegen'ed in device mode (and vice versa). Previously, clang made no attempt to catch these errors. For the most part, they would be caught by ptxas, and reported as "call to unknown function 'foo'". Now we catch these errors and report them the same as we report other illegal calls (e.g. a call from a host function to a device function). This has a small change in error-message behavior for calls that were previously disallowed (e.g. calls from a host to a device function). Previously, we'd catch disallowed calls fairly early, before doing additional semantic checking e.g. of the call's arguments. Now we catch these illegal calls at the very end of our semantic checks, so we'll only emit a "illegal CUDA call" error if the call is otherwise well-formed. Reviewers: tra, rnk Subscribers: cfe-commits Differential Revision: https://reviews.llvm.org/D23242 llvm-svn: 278759
2016-08-16 07:00:49 +08:00
};
__host__ __device__ void T::hd3() {
host_fn();
// expected-error@-1 {{reference to __host__ function 'host_fn' in __host__ __device__ function}}
}
template <typename T> __host__ __device__ void hd2() { host_fn(); }
[CUDA][HIP][Sema] Fix template kernel with function as template parameter If a kernel template has a function as its template parameter, a device function should be allowed as template argument since a kernel can call a device function. However, currently if the kernel template is instantiated in a host function, clang will emit an error message saying the device function is an invalid candidate for the template parameter. This happens because clang checks the reference to the device function during parsing the template arguments. At this point, the template is not instantiated yet. Clang incorrectly assumes the device function is called by the host function and emits the error message. This patch fixes the issue by disabling checking of device function during parsing template arguments and deferring the check to the instantion of the template. At that point, the template decl is already available, therefore the check can be done against the instantiated function template decl. Differential Revision: https://reviews.llvm.org/D56411 llvm-svn: 355421
2019-03-06 02:19:35 +08:00
// expected-error@-1 2 {{reference to __host__ function 'host_fn' in __host__ __device__ function}}
[CUDA] Raise an error if a wrong-side call is codegen'ed. Summary: Some function calls in CUDA are allowed to appear in semantically-correct programs but are an error if they're ever codegen'ed. Specifically, a host+device function may call a host function, but it's an error if such a function is ever codegen'ed in device mode (and vice versa). Previously, clang made no attempt to catch these errors. For the most part, they would be caught by ptxas, and reported as "call to unknown function 'foo'". Now we catch these errors and report them the same as we report other illegal calls (e.g. a call from a host function to a device function). This has a small change in error-message behavior for calls that were previously disallowed (e.g. calls from a host to a device function). Previously, we'd catch disallowed calls fairly early, before doing additional semantic checking e.g. of the call's arguments. Now we catch these illegal calls at the very end of our semantic checks, so we'll only emit a "illegal CUDA call" error if the call is otherwise well-formed. Reviewers: tra, rnk Subscribers: cfe-commits Differential Revision: https://reviews.llvm.org/D23242 llvm-svn: 278759
2016-08-16 07:00:49 +08:00
__global__ void kernel() { hd2<int>(); }
__host__ __device__ void hd() { host_fn(); }
// expected-error@-1 {{reference to __host__ function 'host_fn' in __host__ __device__ function}}
template <typename T> __host__ __device__ void hd3() { host_fn(); }
[CUDA][HIP][Sema] Fix template kernel with function as template parameter If a kernel template has a function as its template parameter, a device function should be allowed as template argument since a kernel can call a device function. However, currently if the kernel template is instantiated in a host function, clang will emit an error message saying the device function is an invalid candidate for the template parameter. This happens because clang checks the reference to the device function during parsing the template arguments. At this point, the template is not instantiated yet. Clang incorrectly assumes the device function is called by the host function and emits the error message. This patch fixes the issue by disabling checking of device function during parsing template arguments and deferring the check to the instantion of the template. At that point, the template decl is already available, therefore the check can be done against the instantiated function template decl. Differential Revision: https://reviews.llvm.org/D56411 llvm-svn: 355421
2019-03-06 02:19:35 +08:00
// expected-error@-1 2 {{reference to __host__ function 'host_fn' in __host__ __device__ function}}
[CUDA] Raise an error if a wrong-side call is codegen'ed. Summary: Some function calls in CUDA are allowed to appear in semantically-correct programs but are an error if they're ever codegen'ed. Specifically, a host+device function may call a host function, but it's an error if such a function is ever codegen'ed in device mode (and vice versa). Previously, clang made no attempt to catch these errors. For the most part, they would be caught by ptxas, and reported as "call to unknown function 'foo'". Now we catch these errors and report them the same as we report other illegal calls (e.g. a call from a host function to a device function). This has a small change in error-message behavior for calls that were previously disallowed (e.g. calls from a host to a device function). Previously, we'd catch disallowed calls fairly early, before doing additional semantic checking e.g. of the call's arguments. Now we catch these illegal calls at the very end of our semantic checks, so we'll only emit a "illegal CUDA call" error if the call is otherwise well-formed. Reviewers: tra, rnk Subscribers: cfe-commits Differential Revision: https://reviews.llvm.org/D23242 llvm-svn: 278759
2016-08-16 07:00:49 +08:00
__device__ void device_fn() { hd3<int>(); }
// No error because this is never instantiated.
template <typename T> __host__ __device__ void hd4() { host_fn(); }
__host__ __device__ void local_var() {
S s;
// expected-error@-1 {{reference to __host__ function 'S' in __host__ __device__ function}}
}
__host__ __device__ void placement_new(char *ptr) {
::new(ptr) S();
// expected-error@-1 {{reference to __host__ function 'S' in __host__ __device__ function}}
}
__host__ __device__ void explicit_destructor(S *s) {
s->~S();
// expected-error@-1 {{reference to __host__ function '~S' in __host__ __device__ function}}
}
Aligned allocation versus CUDA: make deallocation function preference order match other CUDA preference orders, per discussion with jlebar. We now model this in an attempt to match overload resolution as closely as possible: - First, we throw out all non-callable (due to CUDA host/device mismatch) operator delete functions. - Then we apply sizedness / alignedness preferences based on whether the type is overaligned and whether the deallocation function is a member. - Finally, we use the CUDA callability preference as a tiebreaker. llvm-svn: 283830
2016-10-11 08:21:10 +08:00
__host__ __device__ void class_specific_delete(T *t, U *u) {
delete t; // ok, call sized device delete even though host has preferable non-sized version
delete u; // ok, call non-sized HD delete rather than sized D delete
}
[CUDA] Raise an error if a wrong-side call is codegen'ed. Summary: Some function calls in CUDA are allowed to appear in semantically-correct programs but are an error if they're ever codegen'ed. Specifically, a host+device function may call a host function, but it's an error if such a function is ever codegen'ed in device mode (and vice versa). Previously, clang made no attempt to catch these errors. For the most part, they would be caught by ptxas, and reported as "call to unknown function 'foo'". Now we catch these errors and report them the same as we report other illegal calls (e.g. a call from a host function to a device function). This has a small change in error-message behavior for calls that were previously disallowed (e.g. calls from a host to a device function). Previously, we'd catch disallowed calls fairly early, before doing additional semantic checking e.g. of the call's arguments. Now we catch these illegal calls at the very end of our semantic checks, so we'll only emit a "illegal CUDA call" error if the call is otherwise well-formed. Reviewers: tra, rnk Subscribers: cfe-commits Differential Revision: https://reviews.llvm.org/D23242 llvm-svn: 278759
2016-08-16 07:00:49 +08:00
__host__ __device__ void hd_member_fn() {
T t;
// Necessary to trigger an error on T::hd. It's (implicitly) inline, so
// isn't codegen'ed until we call it.
t.hd();
}
__host__ __device__ void h_member_fn() {
T t;
t.h();
// expected-error@-1 {{reference to __host__ function 'h' in __host__ __device__ function}}
}
__host__ __device__ void fn_ptr() {
auto* ptr = &host_fn;
// expected-error@-1 {{reference to __host__ function 'host_fn' in __host__ __device__ function}}
}
template <typename T>
__host__ __device__ void fn_ptr_template() {
auto* ptr = &host_fn; // Not an error because the template isn't instantiated.
}
[CUDA] Do a better job at detecting wrong-side calls. Summary: Move CheckCUDACall from ActOnCallExpr and BuildDeclRefExpr to DiagnoseUseOfDecl. This lets us catch some edge cases we were missing, specifically around class operators. This necessitates a few other changes: - Avoid emitting duplicate deferred diags in CheckCUDACall. Previously we'd carefully placed our call to CheckCUDACall such that it would only ever run once for a particular callsite. But now this isn't the case. - Emit deferred diagnostics from a template specialization/instantiation's primary template, in addition to from the specialization/instantiation itself. DiagnoseUseOfDecl ends up putting the deferred diagnostics on the template, rather than the specialization, so we need to check both. Reviewers: rsmith Subscribers: cfe-commits, tra Differential Revision: https://reviews.llvm.org/D24573 llvm-svn: 283637
2016-10-08 09:07:11 +08:00
__host__ __device__ void unaryOp() {
T t;
(void) +t; // expected-error {{reference to __host__ function 'operator+' in __host__ __device__ function}}
}
__host__ __device__ void binaryOp() {
T t;
(void) (t - t); // expected-error {{reference to __host__ function 'operator-' in __host__ __device__ function}}
}
__host__ __device__ void implicitConversion() {
T t;
Dummy d = t; // expected-error {{reference to __host__ function 'operator Dummy' in __host__ __device__ function}}
}
template <typename T>
struct TmplStruct {
template <typename U> __host__ __device__ void fn() {}
};
template <>
template <>
__host__ __device__ void TmplStruct<int>::fn<int>() { host_fn(); }
// expected-error@-1 {{reference to __host__ function 'host_fn' in __host__ __device__ function}}
__device__ void double_specialization() { TmplStruct<int>().fn<int>(); }