wrap lambda to support Windows #2286
Merged
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
Codecov Report
@@ Coverage Diff @@
## develop #2286 +/- ##
========================================
Coverage 91.45% 91.45%
========================================
Files 433 433
Lines 16175 16177 +2
========================================
+ Hits 14793 14795 +2
Misses 1382 1382 |
|
Does it work if we construct the execute = std::function<argument(context& ctx, const std::vector<argument>& args)>([=](context&, const std::vector<argument>& args) { ... });We probably want to add a typedef for the function to make it more readable: execute = execute_function{[=](context&, const std::vector<argument>& args) { ... }};
If the cause is because of a temporal lambda then you can assign it to a local variable first and then assign the class member: auto local_execute = [=](context&, const std::vector<argument>& args) { ... };
execute = local_execute; |
pfultz2
reviewed
Oct 4, 2023
| @@ -95,7 +95,7 @@ template <class Derived, class Primitive> | |||
| struct dnnl_op : auto_register_op<Derived> | |||
| { | |||
| std::vector<post_op> post_ops; | |||
| std::function<argument(context& ctx, const std::vector<argument>& args)> execute; | |||
There was a problem hiding this comment.
Dont remove the parameter names from the function.
pfultz2
reviewed
Oct 4, 2023
| @@ -284,7 +284,7 @@ struct dnnl_op : auto_register_op<Derived> | |||
|
|
|||
| std::ptrdiff_t output_alias(const std::vector<shape>& shapes) const | |||
| { | |||
| return shapes.size() - 1; | |||
| return static_cast<std::ptrdiff_t>(shapes.size() - 1); | |||
There was a problem hiding this comment.
This doesnt seem to make sense, if the cast is to avoid underflow when size() is zero then that wont prevent that. I think you would need to cast the size: static_cast<std::ptrdiff_t>(shapes.size()) - 1.
There was a problem hiding this comment.
MSVC does not perform implicit type conversion here. It is just simply failing compilation. I will recheck it since I made that change some time ago.
|
I tried all your suggestions, i.e., casting to |
|
So looking at the error closer I see this: Which looks like it fails because of execute = [=](context&, const std::vector<argument>& args) mutable { ... }I dont have access to windows, but if that doesnt fix it, we need to investigate why |
apwojcik
force-pushed
the
windows_bind_dnnl
branch
from
fb658cb to
6e3012c
Compare
apwojcik
changed the title
|
I replaced lambda with functor, and it worked. |
Check results before merge 🔆 |
🔴bert_base_cased_fp16: FAILED: MIGraphX is not within tolerance - check verbose output🔴bert_large_uncased_fp16: FAILED: MIGraphX is not within tolerance - check verbose output🔴distilgpt2_fp16: FAILED: MIGraphX is not within tolerance - check verbose output |
|
Can we wrap the lambda in a function object? template<class F>
struct execute_wrapper
{
F f;
argument operator()(context& ctx, const std::vector<argument>& args) const
{
return f(ctx, args);
}
};
template<class F>
execute_wrapper<F> make_execute_wrapper(F f)
{
return {std::move(f)};
}Then do: execute = make_execute_wrapper([=](context&, const std::vector<argument>& args) { ... }); |
|
I tried wrapping lambda in a function object before, which also failed for me. Here's the error message from the piece of code you asked me to try. |
|
So we are definatly getting close to what the problem is: I am not sure why its taking Either way, we dont use this parameter so we can removed it from the lambda(but not the function object): template<class F>
struct execute_wrapper
{
F f;
argument operator()(context&, const std::vector<argument>& args) const
{
return f(args);
}
};
template<class F>
execute_wrapper<F> make_execute_wrapper(F f)
{
return {std::move(f)};
}And then create the lambda as: execute = make_execute_wrapper([=](const std::vector<argument>& args) { ... }); |
|
It worked. We have the final solution now. Thanks. |
apwojcik
changed the title
pfultz2
approved these changes
Oct 8, 2023
| @@ -91,6 +91,19 @@ struct post_op : reflect_equality<post_op>, reflect_stream<post_op> | |||
| } | |||
| }; | |||
|
|
|||
| template <class F> | |||
There was a problem hiding this comment.
Can you add a comment about wrapping this lambda because an issue with MSVC and the context parameter?
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.
The following error prevents compilation on Windows with Clang/MSVC. It seems the standard library from MSFT does not support assigning temporal lambdas to class data members.