[SYCL] Add support for union types as kernel parameter #2285

smanna12 · 2020-08-07T22:49:55Z

This patch adds support for union types kernel arguments.
- Add 'VisitUnion' function.
- Add separate handler to implement a diagnostic on attempt to pass union with accessor/sampler/stream member
as a kernel argument.
- Add CodeGen/Sema/Intengreation header/runtime tests.

Signed-off-by: Soumi Manna soumi.manna@intel.com

Signed-off-by: Soumi Manna <soumi.manna@intel.com>

smanna12 · 2020-08-07T23:03:26Z

@erichkeane, I have not added any tests yet. I am working on it. This patch adds the changes from PR #2270 and adds a separate handler SyclKernelUnionBodyChecker for the diagnostic. Could you please review this change?

erichkeane

Looks right to me. When writing the union tests, make sure you use one in an array to make sure it is properly passed. Also, a struct-with-an-array-of-unions, and a array-of-struct-with-a-union.

Otherwise, LGTM once we get sufficient tests.

clang/lib/Sema/SemaSYCL.cpp

smanna12 · 2020-08-10T17:21:48Z

Thanks Erich for the review and comments.

The following test compiles fine. I am passing accessor array as a member inside union.
The test is supposed to fail since Accessor is not allowed inside union. Am i missing anything here?

@erichkeane any thought?

smanna@scsel-cfl-05:/export/iusers/smanna/myworkws/llvm/clang/test/SemaSYCL$ cat test.cpp
// RUN: %clang_cc1 -I %S/Inputs -fsycl -fsycl-is-device -fcxx-exceptions -verify -fsyntax-only %s

#include <sycl.hpp>

using namespace cl::sycl;

template <typename name, typename Func>
attribute((sycl_kernel)) void a_kernel(Func kernelFunc) {
kernelFunc();
}

int main() {

using Accessor =
accessor<int, 1, access::mode::read_write, access::target::global_buffer>;

union union_acc_t {
Accessor member_acc[2];
} union_acc;

a_kernel(
= {
union_acc.member_acc[2].use();
});
}

erichkeane · 2020-08-10T17:24:46Z

Thanks Erich for the review and comments.

The following test compiles fine. I am passing accessor array as a member inside union.
The test is supposed to fail since Accessor is not allowed inside union. Am i missing anything here?

@erichkeane any thought?

smanna@scsel-cfl-05:/export/iusers/smanna/myworkws/llvm/clang/test/SemaSYCL$ cat test.cpp
// RUN: %clang_cc1 -I %S/Inputs -fsycl -fsycl-is-device -fcxx-exceptions -verify -fsyntax-only %s

#include <sycl.hpp>

using namespace cl::sycl;

template <typename name, typename Func>
attribute((sycl_kernel)) void a_kernel(Func kernelFunc) {
kernelFunc();
}

int main() {

using Accessor =
accessor<int, 1, access::mode::read_write, access::target::global_buffer>;

union union_acc_t {
Accessor member_acc[2];
} union_acc;

a_kernel(
= {
union_acc.member_acc[2].use();
});
}

Hmm... do you perhaps also have to handle the array-element function? Or is that not properly visiting its types? @elizabethandrews : Do you know why the visitor wouldn't be visiting the elements of an array?

elizabethandrews · 2020-08-10T18:04:43Z

Thanks Erich for the review and comments.
The following test compiles fine. I am passing accessor array as a member inside union.
The test is supposed to fail since Accessor is not allowed inside union. Am i missing anything here?
@erichkeane any thought?
smanna@scsel-cfl-05:/export/iusers/smanna/myworkws/llvm/clang/test/SemaSYCL$ cat test.cpp
// RUN: %clang_cc1 -I %S/Inputs -fsycl -fsycl-is-device -fcxx-exceptions -verify -fsyntax-only %s
#include <sycl.hpp>
using namespace cl::sycl;
template <typename name, typename Func>
attribute((sycl_kernel)) void a_kernel(Func kernelFunc) {
kernelFunc();
}
int main() {
using Accessor =
accessor<int, 1, access::mode::read_write, access::target::global_buffer>;
union union_acc_t {
Accessor member_acc[2];
} union_acc;
a_kernel(
= {
union_acc.member_acc[2].use();
});
}

Hmm... do you perhaps also have to handle the array-element function? Or is that not properly visiting its types? @elizabethandrews : Do you know why the visitor wouldn't be visiting the elements of an array?

I don't know why it wouldn't visit the elements. I haven't had a chance to look at Erich's template changes properly. I need to actually step through it to understand what is happening. It is quite complicated.

Array elements are handled in VisitElement. Am I right in understanding that the call to 'handleSyclAccessorType' here should call the handler defined in SyclKernelUnionBodyChecker?

erichkeane · 2020-08-10T18:09:17Z

Thanks Erich for the review and comments.
The following test compiles fine. I am passing accessor array as a member inside union.
The test is supposed to fail since Accessor is not allowed inside union. Am i missing anything here?
@erichkeane any thought?
smanna@scsel-cfl-05:/export/iusers/smanna/myworkws/llvm/clang/test/SemaSYCL$ cat test.cpp
// RUN: %clang_cc1 -I %S/Inputs -fsycl -fsycl-is-device -fcxx-exceptions -verify -fsyntax-only %s
#include <sycl.hpp>
using namespace cl::sycl;
template <typename name, typename Func>
attribute((sycl_kernel)) void a_kernel(Func kernelFunc) {
kernelFunc();
}
int main() {
using Accessor =
accessor<int, 1, access::mode::read_write, access::target::global_buffer>;
union union_acc_t {
Accessor member_acc[2];
} union_acc;
a_kernel(
= {
union_acc.member_acc[2].use();
});
}

Hmm... do you perhaps also have to handle the array-element function? Or is that not properly visiting its types? @elizabethandrews : Do you know why the visitor wouldn't be visiting the elements of an array?

I don't know why it wouldn't visit the elements. I haven't had a chance to look at Erich's template changes properly. I need to actually step through it to understand what is happening. It is quite complicated.

Array elements are handled in VisitElement. Am I right in understanding that the call to 'handleSyclAccessorType' here should call the handler defined in SyclKernelUnionBodyChecker?

I don't think my union changes really matter here. I'd expect the visitor to visit the elements themselves (via VisitUnion?), then each of the sub-elements to it. VisitElement calls VisitUnion in this patch, so I'd expect that to work? Or did I screw up the visiting there or something? Perhaps we need a bit of time to be debugging.

elizabethandrews · 2020-08-10T18:25:09Z

Thanks Erich for the review and comments.
The following test compiles fine. I am passing accessor array as a member inside union.
The test is supposed to fail since Accessor is not allowed inside union. Am i missing anything here?
@erichkeane any thought?
smanna@scsel-cfl-05:/export/iusers/smanna/myworkws/llvm/clang/test/SemaSYCL$ cat test.cpp
// RUN: %clang_cc1 -I %S/Inputs -fsycl -fsycl-is-device -fcxx-exceptions -verify -fsyntax-only %s
#include <sycl.hpp>
using namespace cl::sycl;
template <typename name, typename Func>
attribute((sycl_kernel)) void a_kernel(Func kernelFunc) {
kernelFunc();
}
int main() {
using Accessor =
accessor<int, 1, access::mode::read_write, access::target::global_buffer>;
union union_acc_t {
Accessor member_acc[2];
} union_acc;
a_kernel(
= {
union_acc.member_acc[2].use();
});
}

Hmm... do you perhaps also have to handle the array-element function? Or is that not properly visiting its types? @elizabethandrews : Do you know why the visitor wouldn't be visiting the elements of an array?

I don't know why it wouldn't visit the elements. I haven't had a chance to look at Erich's template changes properly. I need to actually step through it to understand what is happening. It is quite complicated.
Array elements are handled in VisitElement. Am I right in understanding that the call to 'handleSyclAccessorType' here should call the handler defined in SyclKernelUnionBodyChecker?

I don't think my union changes really matter here. I'd expect the visitor to visit the elements themselves (via VisitUnion?), then each of the sub-elements to it. VisitElement calls VisitUnion in this patch, so I'd expect that to work? Or did I screw up the visiting there or something? Perhaps we need a bit of time to be debugging.

The handlers in VisitElement are to handle individual elements of an array. For example, add __init calls for each accessor element. I don't fully understand your patch yet. I need to step through it but I think VisitUnion here means you're handling an array of unions.

erichkeane · 2020-08-10T18:47:48Z

Ah, I might have it :) The union-checker is never added in ConstructOpenCLKernel as a new visitor! Add it after 'checker'.

smanna12 · 2020-08-10T18:48:53Z

Thanks Erich for the review and comments.
The following test compiles fine. I am passing accessor array as a member inside union.
The test is supposed to fail since Accessor is not allowed inside union. Am i missing anything here?
@erichkeane any thought?
smanna@scsel-cfl-05:/export/iusers/smanna/myworkws/llvm/clang/test/SemaSYCL$ cat test.cpp
// RUN: %clang_cc1 -I %S/Inputs -fsycl -fsycl-is-device -fcxx-exceptions -verify -fsyntax-only %s
#include <sycl.hpp>
using namespace cl::sycl;
template <typename name, typename Func>
attribute((sycl_kernel)) void a_kernel(Func kernelFunc) {
kernelFunc();
}
int main() {
using Accessor =
accessor<int, 1, access::mode::read_write, access::target::global_buffer>;
union union_acc_t {
Accessor member_acc[2];
} union_acc;
a_kernel(
= {
union_acc.member_acc[2].use();
});
}

Hmm... do you perhaps also have to handle the array-element function? Or is that not properly visiting its types? @elizabethandrews : Do you know why the visitor wouldn't be visiting the elements of an array?

I don't know why it wouldn't visit the elements. I haven't had a chance to look at Erich's template changes properly. I need to actually step through it to understand what is happening. It is quite complicated.
Array elements are handled in VisitElement. Am I right in understanding that the call to 'handleSyclAccessorType' here should call the handler defined in SyclKernelUnionBodyChecker?

I don't think my union changes really matter here. I'd expect the visitor to visit the elements themselves (via VisitUnion?), then each of the sub-elements to it. VisitElement calls VisitUnion in this patch, so I'd expect that to work? Or did I screw up the visiting there or something? Perhaps we need a bit of time to be debugging.

The handlers in VisitElement are to handle individual elements of an array. For example, add __init calls for each accessor element. I don't fully understand your patch yet. I need to step through it but I think VisitUnion here means you're handling an array of unions.

I see that we have added " SyclKernelFieldChecker" in two places:
ConstructOpenCLKernel
SyclKernelDeclCreator

I have not added "SyclKernelUnionBodyChecker " anywhere in the patch. This might cause the issue here.

erichkeane · 2020-08-10T18:54:01Z

You need it in the ConstructOpenCLKernel function (see my response right above yours :) )

smanna12 · 2020-08-10T18:55:57Z

You need it in the ConstructOpenCLKernel function (see my response right above yours :) )

Yes, i missed this. Thanks Erich.

Signed-off-by: Soumi Manna <soumi.manna@intel.com>

erichkeane

Additionally, this needs a test to show the diagnostics happening.

Yes, i am missing diagnostic tests. UnionBodyChecker seems like not working properly. I am debugging this issue. The following test expects to fail since we are trying to pass accessor array inside union. This test generates no diagnostic. Same is happening with sampler/stream.

// RUN: %clang_cc1 -I %S/Inputs -fsycl -fsycl-is-device -fcxx-exceptions -verify -fsyntax-only %s
#include <sycl.hpp>
using namespace cl::sycl;
template <typename name, typename Func>
attribute((sycl_kernel)) void a_kernel(Func kernelFunc) {
kernelFunc();
}
int main() {
using Accessor =
accessor<int, 1, access::mode::read_write, access::target::global_buffer>;
union union_acc_t {
Accessor member_acc[2];
} union_acc;

a_kernel(
= {
union_acc.member_acc[2].use();
});
}

clang/lib/Sema/SemaSYCL.cpp

erichkeane · 2020-08-11T13:52:49Z

clang/lib/Sema/SemaSYCL.cpp

 // A type to Create and own the FunctionDecl for the kernel.
 class SyclKernelDeclCreator : public SyclKernelFieldHandler {
  FunctionDecl *KernelDecl;
  llvm::SmallVector<ParmVarDecl *, 8> Params;
  SyclKernelFieldChecker &ArgChecker;
+  SyclKernelUnionBodyChecker &ArgChecker1;


I'd probably call this 'UnionChecker'. Additionally, note that #2289 makes this part unnecessary.

Ok. I will change that.

Actually, looking again, adding ArgChecker1 is completely unnecessary. ArgChecker itself is completely unnecessary here too (and likely should be removed). I remove it in #2289 (depending on when that gets in), but it seems to be a leftover from when it was necessary.

It seems like only place we can add this handler "SyclKernelUnionBodyChecker checker1(*this)" to ConstructOpenCLKernel.

Yep, I thought you would have noticed trying to debug why diagnostics weren't working right, but it needs to be added to these two lines instead:

Visitor.VisitRecordBases(KernelObj, checker, kernel_decl, kernel_body,
int_header);

Visitor.VisitRecordFields(KernelObj, checker, kernel_decl, kernel_body,
int_header);

erichkeane · 2020-08-11T13:54:35Z

clang/test/CodeGenSYCL/union-kernel-param.cpp

+// CHECK-NEXT: const kernel_param_desc_t kernel_signatures[] = {
+// CHECK-NEXT:   //--- _ZTSZZ5test0vENK3$_0clERN2cl4sycl7handlerEE8MyKernel
+// CHECK-NEXT:  { kernel_param_kind_t::kind_accessor, 4062, 0 },
+// CHECK-NEXT:  { kernel_param_kind_t::kind_std_layout, 12, 16 },


I don't think this is right, is it? 'S' is just the union, right? We should be capturing that as a single std_layout, not as however many these are.

erichkeane · 2020-08-11T13:55:15Z

clang/test/CodeGenSYCL/union-kernel-param1.cpp

+  }
+}
+
+// CHECK MyKernel parameters


@elizabethandrews can you take a look at this? The union should be a simple un-decomposed value.

From the Sema tests below, it looks like parameters are being passed individually as well? This in incorrect right?

Yes, the test seems incorrect. I will update the test. Thanks Erich and Elizabeth.

elizabethandrews · 2020-08-11T15:28:13Z

clang/test/SemaSYCL/union-kernel-param1.cpp

+// Check kernel_A parameters
+// CHECK: FunctionDecl {{.*}}kernel_A{{.*}} 'void (union union_acc_t, int, int)'
+// CHECK-NEXT: ParmVarDecl {{.*}} used _arg_ 'union union_acc_t':'union_acc_t'
+// CHECK-NEXT: ParmVarDecl {{.*}} used _arg_member_acc 'int'


It looks decomposition is happening ? member_acc is the union array. This should not be passed as individual parameters right? It looks like you are passing the 'not-decomposed' union as well as it's fields?

elizabethandrews · 2020-08-11T15:38:23Z

clang/test/SemaSYCL/union-kernel-param.cpp

+
+typedef float realw;
+
+typedef union dpct_type_54e08f {


Can you add more than one field in this union? Also why do we need the typedefs? A simple test with 2-3 non-array fields will be best to check basic functionality.

Yep! I also want to see an:

array of unions. (should be 1 union per element)

Array of structs-with-unions. (should be decomposed structs, but unions still only a single).

I will update tests.

erichkeane · 2020-08-11T15:42:57Z

clang/lib/Sema/SemaSYCL.cpp

+// A type to check the validity of passing union with accessor/sampler/stream
+// member as a kernel argument types.
+class SyclKernelUnionBodyChecker : public SyclKernelFieldHandler {
+  static constexpr const bool VisitUnionBody = true;


I think your problem is that SyclKernelFieldHandler needs this line as well, except =false instead.

I think that results in the union-filtering not actually doing anything.

think that results in the union-filtering not actually doing anything.
Yes, Thanks Erich.

I can't find where it is done, that might be the reason why everything is decomposed.

Yep, it should be in the base class, not just the other checker. WHich is why the tests are still wrong.

erichkeane · 2020-08-11T15:54:47Z

I NEED us to be able to catch that missed constexpr variable, so please change the visitor to be like this:

https://godbolt.org/z/Tcbv8a

Changes:
-Each of the VisitUnion calls now has the FilteredHandlers type FIRST instead of right at the end in their param lists.
-The call on line 34 has filtered_handlers... moved to be FIRST.
-The call on line 48 has filtered_handlers... AND cur_handler moved to the front.

This will prevent silently missing that the flag is missing.

EDIT: Disregard the printf statements I added in that :) Those were for my own debugging.

smanna12 · 2020-08-12T23:37:28Z

I tried this simple union test without array. The AST looks like this.

union MyUnion {
int x;
char y;
float cuda;
};

template <typename name, typename Func>
_attribute((sycl_kernel)) void a_kernel(Func kernelFunc) {
kernelFunc();
}

int main() {
MyUnion accel;

a_kernel(
= {
float local = accel.cuda;
});
}

@erichkeane, Are we expecting this AST syntax ?

FunctionDecl 0x55d691256088 <line:26:7, > col:7 _ZTSZ4mainE6kernel 'void (MyUnion, int, char, float)'
|-ParmVarDecl 0x55d6912564b0 <> used arg 'MyUnion'
|-ParmVarDecl 0x55d6912565e0 <> used _arg_x 'int'
|-ParmVarDecl 0x55d6912566d8 <> used _arg_y 'char'
|-ParmVarDecl 0x55d6912567d8 <> used _arg_cuda 'float'
|-CompoundStmt 0x55d691256990 <>
| |-DeclStmt 0x55d691256260 <>
| | -VarDecl 0x55d6912561f8 <<invalid sloc>, col:11> <invalid sloc> used '(lambda at /export/iusers/smanna/myworkws/llvm/clang/test/SemaSYCL/union-kernel-param4.cpp:26:7)' cinit | | -InitListExpr 0x55d691256878 <> '(lambda at /export/iusers/smanna/myworkws/llvm/clang/test/SemaSYCL/union-kernel-param4.cpp:26:7)'
| | |-CXXConstructExpr 0x55d691256550 <> 'MyUnion' 'void (const MyUnion &) noexcept'
| | | -ImplicitCastExpr 0x55d691256538 <<invalid sloc>> 'const MyUnion' lvalue <NoOp> | | | -DeclRefExpr 0x55d691256518 <> 'MyUnion' lvalue ParmVar 0x55d6912564b0 'arg' 'MyUnion'
| | |-ImplicitCastExpr 0x55d691256668 <> 'int'
| | | -DeclRefExpr 0x55d691256648 <<invalid sloc>> 'int' lvalue ParmVar 0x55d6912565e0 '_arg_x' 'int' | | |-ImplicitCastExpr 0x55d691256760 <<invalid sloc>> 'char' <LValueToRValue> | | | -DeclRefExpr 0x55d691256740 <> 'char' lvalue ParmVar 0x55d6912566d8 '_arg_y' 'char'
| | -ImplicitCastExpr 0x55d691256860 <<invalid sloc>> 'float' <LValueToRValue> | | -DeclRefExpr 0x55d691256840 <> 'float' lvalue ParmVar 0x55d6912567d8 '_arg_cuda' 'float'
| `-CompoundStmt 0x55d691256978 <>

erichkeane · 2020-08-12T23:39:23Z

No, arg_x, arg_y, and arg_cuda shouldn't end up in there. For some reason, this is still visiting places it shouldn't. Do you have the latest patch up that someone can take a look at?

I will check with Maria and Elizabeth tomorrow and I will upload latest patch.

Signed-off-by: Soumi Manna <soumi.manna@intel.com>

smanna12 · 2020-08-14T20:20:11Z

Sorry for the multiple submissions. Clang format took wrong patches. This is fixed now.

smanna12 · 2020-08-15T22:50:27Z

@elizabethandrews , @Fznamznon,, @premanandrao could you please review the support? Thanks.

Fznamznon

A couple of minors in tests, otherwise looks ok

Fznamznon · 2020-08-17T07:47:36Z

clang/test/SemaSYCL/union-kernel-param1.cpp

+#include <sycl.hpp>
+
+using namespace cl::sycl;


I don't see usage of this include. Let's remove it.

Fznamznon · 2020-08-17T07:47:58Z

clang/test/SemaSYCL/union-kernel-param2.cpp

+#include <sycl.hpp>
+
+using namespace cl::sycl;


Signed-off-by: Soumi Manna <soumi.manna@intel.com>

smanna12 · 2020-08-17T14:29:02Z

Runtime test was reviewed and approved in previous commit (db9f49d) by @againull on behalf of intel/llvm-reviewers-runtime . Latest commit is just couple of minor nits in FE tests, no change in runtime test.

@bader , would it be possible to merge this into github today? Thanks!

smanna12 · 2020-08-17T16:45:19Z

Thanks everyone for the reviews.

bader · 2020-08-17T16:54:18Z

@bader , would it be possible to merge this into github today? Thanks!

@smanna12, could you add more details to the PR description about this patch, which I can use as commit message?

smanna12 · 2020-08-17T17:06:52Z

@bader , would it be possible to merge this into github today? Thanks!

@smanna12, could you add more details to the PR description about this patch, which I can use as commit message?

@bader, I have added PR description. Sorry about this.

Preserve DIExpression in DIGlobalVariableExpression. Preserving DIExpressions is necessary because of this change in LLVM: commit 638a839 Author: Michael Buch michaelbuch12@gmail.com Date: Mon Nov 13 06:04:27 2023 +0000 Reland "[clang][DebugInfo] Emit global variable definitions for static data members with constant initializers" (#71780) Original commit: KhronosGroup/SPIRV-LLVM-Translator@9edf714

[SYCL] Add support for union

7be969a

Signed-off-by: Soumi Manna <soumi.manna@intel.com>

smanna12 changed the title ~~[SYCL] Add support for union~~ [[SYCL] Add support for union types as kernel parameter Aug 7, 2020

smanna12 changed the title ~~[[SYCL] Add support for union types as kernel parameter~~ [SYCL] Add support for union types as kernel parameter Aug 7, 2020

erichkeane reviewed Aug 7, 2020

View reviewed changes

clang/lib/Sema/SemaSYCL.cpp Show resolved Hide resolved

clang/lib/Sema/SemaSYCL.cpp Show resolved Hide resolved

clang/lib/Sema/SemaSYCL.cpp Outdated Show resolved Hide resolved

smanna12 added 4 commits August 11, 2020 01:49

add tests

6de6dff

Signed-off-by: Soumi Manna <soumi.manna@intel.com>

Fix Clang-format issue

60f02ac

Signed-off-by: Soumi Manna <soumi.manna@intel.com>

Fix Clang-format issue

fc3999b

Signed-off-by: Soumi Manna <soumi.manna@intel.com>

Fix Clang-format issue

e71672c

Signed-off-by: Soumi Manna <soumi.manna@intel.com>

erichkeane mentioned this pull request Aug 11, 2020

[SYCL]Move kernel diagnostics to the call, enable template trail on errors #2289

Merged

smanna12 requested a review from erichkeane August 11, 2020 13:50

erichkeane reviewed Aug 11, 2020

View reviewed changes

elizabethandrews reviewed Aug 11, 2020

View reviewed changes

erichkeane reviewed Aug 11, 2020

View reviewed changes

Merge remote-tracking branch 'intel_llvm/sycl' into UnionKernelArgument

c36ab3f

smanna12 added 2 commits August 12, 2020 22:31

update tests and code changes

99e8b2a

Signed-off-by: Soumi Manna <soumi.manna@intel.com>

Fix Clang format issue

2e7b74b

Signed-off-by: Soumi Manna <soumi.manna@intel.com>

smanna12 added 2 commits August 14, 2020 12:19

Fix sema codes

03460a1

Signed-off-by: Soumi Manna <soumi.manna@intel.com>

Fix clang-format issue

f4450a6

Signed-off-by: Soumi Manna <soumi.manna@intel.com>

erichkeane previously approved these changes Aug 14, 2020

View reviewed changes

Fix clang-format issue and update source codes

f772189

Signed-off-by: Soumi Manna <soumi.manna@intel.com>

smanna12 dismissed erichkeane’s stale review via f772189 August 14, 2020 20:04

Fix clang-format issues

db9f49d

Signed-off-by: Soumi Manna <soumi.manna@intel.com>

smanna12 requested review from erichkeane and againull August 14, 2020 20:21

againull previously approved these changes Aug 14, 2020

View reviewed changes

smanna12 requested review from erichkeane and removed request for erichkeane August 14, 2020 21:15

erichkeane previously approved these changes Aug 14, 2020

View reviewed changes

Fznamznon reviewed Aug 17, 2020

View reviewed changes

update test based on reiew

347b21c

Signed-off-by: Soumi Manna <soumi.manna@intel.com>

smanna12 dismissed stale reviews from erichkeane and againull via 347b21c August 17, 2020 12:09

smanna12 requested review from Fznamznon and againull August 17, 2020 12:18

erichkeane approved these changes Aug 17, 2020

View reviewed changes

Fznamznon approved these changes Aug 17, 2020

View reviewed changes

smanna12 requested a review from a team August 17, 2020 13:17

againull approved these changes Aug 17, 2020

View reviewed changes

bader merged commit 5adfd79 into intel:sycl Aug 17, 2020

[SYCL] Add support for union types as kernel parameter #2285

[SYCL] Add support for union types as kernel parameter #2285

Conversation

smanna12 commented Aug 7, 2020 • edited Loading

smanna12 commented Aug 7, 2020

erichkeane left a comment

Choose a reason for hiding this comment

smanna12 commented Aug 10, 2020

erichkeane commented Aug 10, 2020

elizabethandrews commented Aug 10, 2020

erichkeane commented Aug 10, 2020

elizabethandrews commented Aug 10, 2020

erichkeane commented Aug 10, 2020

smanna12 commented Aug 10, 2020

erichkeane commented Aug 10, 2020

smanna12 commented Aug 10, 2020

erichkeane left a comment • edited by smanna12 Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

erichkeane commented Aug 11, 2020 • edited Loading

smanna12 commented Aug 12, 2020 • edited Loading

erichkeane commented Aug 12, 2020 • edited by smanna12 Loading

smanna12 commented Aug 14, 2020

smanna12 commented Aug 15, 2020 • edited Loading

Fznamznon left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

smanna12 commented Aug 17, 2020 • edited Loading

smanna12 commented Aug 17, 2020

bader commented Aug 17, 2020

smanna12 commented Aug 17, 2020

smanna12 commented Aug 7, 2020 •

edited

Loading

erichkeane left a comment •

edited by smanna12

Loading

erichkeane commented Aug 11, 2020 •

edited

Loading

smanna12 commented Aug 12, 2020 •

edited

Loading

erichkeane commented Aug 12, 2020 •

edited by smanna12

Loading

smanna12 commented Aug 15, 2020 •

edited

Loading

smanna12 commented Aug 17, 2020 •

edited

Loading