[Openmp-dev] Offloading app build error with Trunk

[Doerfert, Johannes via Openmp-dev]

Hi Itaru,,

The real solution to the math stuff is prototyped in D71179.
Unfortunately the discussion, which continued on D71241, got derailed.
Once we finally get it unstuck I will get things in asap, hopefully before the LLVM 10 fork.

Sorry for the wait,
  Johannes



---------------------------------------
Johannes Doerfert
Researcher

Argonne National Laboratory
Lemont, IL 60439, USA

jdoerfert at anl.gov

________________________________________
From: Openmp-dev <openmp-dev-bounces at lists.llvm.org> on behalf of Itaru Kitayama via Openmp-dev <openmp-dev at lists.llvm.org>
Sent: Sunday, December 15, 2019 23:42
To: openmp-dev
Subject: Re: [Openmp-dev] Offloading app build error with Trunk

Following up on my post:

Applying below to current Trunk as of this writing:
$ git diff
diff --git a/clang/lib/Headers/__clang_cuda_cmath.h b/clang/lib/Headers/__clang_
index 834a2e3..fb17660 100644
--- a/clang/lib/Headers/__clang_cuda_cmath.h
+++ b/clang/lib/Headers/__clang_cuda_cmath.h
@@ -53,8 +53,8 @@ __DEVICE__ double abs(double __x) { return ::fabs(__x); }
 #endif
 // TODO: remove once variat is supported.
 #if defined(_OPENMP) && defined(__cplusplus)
-__DEVICE__ const float abs(const float __x) { return ::fabsf((float)__x); }
-__DEVICE__ const double abs(const double __x) { return ::fabs((double)__x); }
+//__DEVICE__ const float abs(const float __x) { return ::fabsf((float)__x); }
+//__DEVICE__ const double abs(const double __x) { return ::fabs((double)__x); }
 #endif
 __DEVICE__ float acos(float __x) { return ::acosf(__x); }
 __DEVICE__ float asin(float __x) { return ::asinf(__x); }
diff --git a/clang/lib/Headers/__clang_cuda_math_forward_declares.h b/clang/lib/
index 0afe4db..46b95db 100644
--- a/clang/lib/Headers/__clang_cuda_math_forward_declares.h
+++ b/clang/lib/Headers/__clang_cuda_math_forward_declares.h
@@ -47,8 +47,8 @@ __DEVICE__ float abs(float);
 // TODO: Once variant is supported the additional functions will have
 // to be removed.
 #if defined(_OPENMP) && defined(__cplusplus)
-__DEVICE__ const double abs(const double);
-__DEVICE__ const float abs(const float);
+//__DEVICE__ const double abs(const double);
+//__DEVICE__ const float abs(const float);
 #endif
 __DEVICE__ int abs(int) __NOEXCEPT;
 __DEVICE__ double acos(double);

now my app builds again on POWER8.

Experts,
If you guys can come up a work-around patch, would you merge it in trunk so
we don't need to make local changes?

Thanks,

On Mon, Dec 16, 2019 at 11:15 AM Itaru Kitayama <itaru.kitayama at gmail.com<mailto:itaru.kitayama at gmail.com>> wrote:
Hi,
I am getting a build error:

In file included from /gpfs/software/opt/gcc/7.2.0/lib/gcc/powerpc64le-unknown-linux-gnu/7.2.0/../../../../include/c++/7.2.0/cstdlib:77:
/gpfs/software/opt/gcc/7.2.0/lib/gcc/powerpc64le-unknown-linux-gnu/7.2.0/../../../../include/c++/7.2.0/bits/std_abs.h:70:3: error: declaration conflicts with target of using declaration already in scope
  abs(double __x)
  ^
/p/scratch/cpcp0/pcp0151/opt/clang/201912160138/lib/clang/10.0.0/include/__clang_cuda_cmath.h:57:25: note: target of using declaration
__DEVICE__ const double abs(const double __x) { return ::fabs((double)__x); }
                        ^
/gpfs/software/opt/gcc/7.2.0/lib/gcc/powerpc64le-unknown-linux-gnu/7.2.0/../../../../include/c++/7.2.0/bits/std_abs.h:52:11: note: using declaration
  using ::abs;
          ^
/gpfs/software/opt/gcc/7.2.0/lib/gcc/powerpc64le-unknown-linux-gnu/7.2.0/../../../../include/c++/7.2.0/bits/std_abs.h:74:3: error: declaration conflicts with target of using declaration already in scope
  abs(float __x)
  ^
/p/scratch/cpcp0/pcp0151/opt/clang/201912160138/lib/clang/10.0.0/include/__clang_cuda_cmath.h:56:24: note: target of using declaration
__DEVICE__ const float abs(const float __x) { return ::fabsf((float)__x); }
                       ^
/gpfs/software/opt/gcc/7.2.0/lib/gcc/powerpc64le-unknown-linux-gnu/7.2.0/../../../../include/c++/7.2.0/bits/std_abs.h:52:11: note: using declaration
  using ::abs;
          ^
/p/project/cpcp0/pcp0151/projects/nest-simulator/librandom/binomial_randomdev.cpp:233:34: warning: implicit conversion from 'long' to 'double' changes value from 9223372036854775807 to 9223372036854775808 [-Wimplicit-int-float-conversion]
    static_cast< long >( 0.998 * std::numeric_limits< long >::max() );
                               ~ ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1 warning and 2 errors generated.
make[2]: *** [librandom/CMakeFiles/random.dir/binomial_randomdev.cpp.o] Error 1
make[1]: *** [librandom/CMakeFiles/random.dir/all] Error 2
make: *** [all] Error 2

Although I have this hunk:

$ git diff
diff --git a/clang/lib/Headers/__clang_cuda_math_forward_declares.h b/clang/lib/Headers/__clang_cuda_math_f
index 0afe4db..46b95db 100644
--- a/clang/lib/Headers/__clang_cuda_math_forward_declares.h
+++ b/clang/lib/Headers/__clang_cuda_math_forward_declares.h
@@ -47,8 +47,8 @@ __DEVICE__ float abs(float);
 // TODO: Once variant is supported the additional functions will have
 // to be removed.
 #if defined(_OPENMP) && defined(__cplusplus)
-__DEVICE__ const double abs(const double);
-__DEVICE__ const float abs(const float);
+//__DEVICE__ const double abs(const double);
+//__DEVICE__ const float abs(const float);
 #endif
 __DEVICE__ int abs(int) __NOEXCEPT;
 __DEVICE__ double acos(double);

, as Alexey suggested.