Skip to content
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

Debug combined mode #69

Open
wants to merge 1,183 commits into
base: master
Choose a base branch
from
Open

Debug combined mode #69

wants to merge 1,183 commits into from

Conversation

ronlieb
Copy link
Contributor

@ronlieb ronlieb commented May 1, 2023

No description provided.

rakumarh and others added 30 commits November 10, 2020 10:38
Change-Id: I927a0e71c5b139b275f0a29437578a006851621a
…ed array member of derived type

Earlier change for this caused a hang while compilation, it is now fixed by moving a condition.
Change-Id: Ifdd0aa586c3cc0ed130d11f4707cf6f3df2d5835
…ttributed array member of derived type" into amd-staging
…tion is not recognized. This needs to be reviewed in case there are alternative instructions or methods for ppc64.
fixes to support build of flang on ppc64.  For example PEFETCH instru…
… soure changes

Change-Id: I4f01042c2468b0cb5af845c0a13c21a1ddc7a7c1
…inter derived type

Change-Id: I1c728e51de869fdc91ae7f8f6a280a28488e3e4d
Change-Id: Ic5d18b384f8e39af7c3ea5a92c0c461da5510e8c
Change-Id: Ic1e643d85ab717713537ad2edf1fe545a28299bc
Change-Id: I1f39a0920c9e9ba608c8798ce5e8104912d8a903
Subscripts specified for non-array variable isym

Change-Id: Iaa8080e1538eff7bb1196d5fc97c2aa0a679abb1
Lowering Error: symbol mpair is an inconsistent array descriptor

Change-Id: I58f16bd2405cf27454b26c22f88379e57cca9976
Change-Id: Ia00d2eb1acc84233ba7f8eb46764b8c249aca6dc
Change-Id: I8f3217b5462cf163b88d8dc3f6a32f40f9b9655c
Change-Id: Ib45bb3fd7c4dbf4756e99d7c97f11543c81bcbfc
Change-Id: I13be1de467f5346fc0ed370d08fecc08524b493f
Change-Id: I8b2dba646b5aced3d79bf364667428ccbef0e159
Change-Id: I927a0e71c5b139b275f0a29437578a006851621a
…ed array member of derived type

Earlier change for this caused a hang while compilation, it is now fixed by moving a condition.
Change-Id: Ifdd0aa586c3cc0ed130d11f4707cf6f3df2d5835
… soure changes

Change-Id: I4f01042c2468b0cb5af845c0a13c21a1ddc7a7c1
Change-Id: Ic5d18b384f8e39af7c3ea5a92c0c461da5510e8c
Change-Id: Ic1e643d85ab717713537ad2edf1fe545a28299bc
Change-Id: I1f39a0920c9e9ba608c8798ce5e8104912d8a903
Change-Id: I248c9633d704c55056883e0ccceea260ab434ba1
Change-Id: I4c71440bca2a63b109348748ce5aba8129ba5287
Change-Id: I248c9633d704c55056883e0ccceea260ab434ba1
Change-Id: I4c71440bca2a63b109348748ce5aba8129ba5287
ronlieb and others added 28 commits July 10, 2022 11:16
Centos-9 had an issue where this library was not
being found. We should instead use find_library
to allow certain names to be searched in case
the OS does not have a libquadmath.so symlink.
The build will now fail in the cmake step if
quadmath is not found.
Flang adds hardcoded attribute to kernel functions:
amdgpu-flat-work-group-size="256,256" . This attribute
can cause wrong program execution especially when hardcoded
work group size attribute does not correspond to work group size
used in OpenMP runtime.

Signed-off-by: Dominik Adamski <dominik.adamski@amd.com>

Signed-off-by: Dominik Adamski <dominik.adamski@amd.com>
This reverts commit 092afa4.
  519.clvleaf_t:
    FOPTIMIZE += -Mx,201,2 # workaround for map copy issue
…-patch

restore genasys gest patch, and add flag to allow 519 to
Flang uses pgmath by default. Unfortunately this library is not supported
on the GPU. That's why we observe undefined symbol if we want to use
this symbol on the GPU. This patch adds call to libomptarget equivalent
function (powif) if the code is generated for the AMD GPU.

Signed-off-by: Dominik Adamski <dominik.adamski@amd.com>

Signed-off-by: Dominik Adamski <dominik.adamski@amd.com>
… region (#65)

Eliminate calls to variadic modulo runtime functions. Variadic functions cause
linking errors for OpenMP target pragma if we compile code without
inline optimization.

OpenMP math library is used for C/C++ and Fortran code. That's why Fortran
runtime functions names should not collide with user C++ function names.
Adding prefix __ before Fortran runtime functions name eliminates potential
conflict with user C++ function names.

Co-Author: Shivarama Rao
Clang and Flang can generate the code where the host side
can invoke target offloading procedure with more arguments
than the number of device OpenMP kernel arguments.

If given host symbol is not mapped into the device symbol
then it should be not treated as valid param for offloading.
We should modify type of target param only if:
  * we call the function which launches offloaded kernel
and:
  * host variable is not mapped as kernel argument
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.