https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93226
Thomas Schwinge <tschwinge at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Resolution|FIXED |---
Status|RESOLVED |REOPENED
Target Milestone|--- |16.0
Assignee|unassigned at gcc dot gnu.org |burnus at gcc dot
gnu.org
CC|clm at codesourcery dot com |
Keywords| |testsuite-fail
--- Comment #5 from Thomas Schwinge <tschwinge at gcc dot gnu.org> ---
Thanks. (I'd intentionally left this one available to work on for a beginner
contributor -- but yeah, we've been waiting for many years for one to pick it
up...)
However, I'm reopening this, as there are some issue:
On trunk, for some (but not all) nvptx offloading systems, I see:
PASS: libgomp.oacc-c/../libgomp.oacc-c-c++-common/acc_memcpy_device-1.c
-DACC_DEVICE_TYPE_nvidia=1 -DACC_MEM_SHARED=0 -foffload=nvptx-none -O0 (test
for excess errors)
PASS: libgomp.oacc-c/../libgomp.oacc-c-c++-common/acc_memcpy_device-1.c
-DACC_DEVICE_TYPE_nvidia=1 -DACC_MEM_SHARED=0 -foffload=nvptx-none -O0
execution test
PASS: libgomp.oacc-c/../libgomp.oacc-c-c++-common/acc_memcpy_device-1.c
-DACC_DEVICE_TYPE_nvidia=1 -DACC_MEM_SHARED=0 -foffload=nvptx-none -O2 (test
for excess errors)
FAIL: libgomp.oacc-c/../libgomp.oacc-c-c++-common/acc_memcpy_device-1.c
-DACC_DEVICE_TYPE_nvidia=1 -DACC_MEM_SHARED=0 -foffload=nvptx-none -O2
execution test
PASS: libgomp.oacc-c++/../libgomp.oacc-c-c++-common/acc_memcpy_device-1.c
-DACC_DEVICE_TYPE_nvidia=1 -DACC_MEM_SHARED=0 -foffload=nvptx-none -O0 (test
for excess errors)
PASS: libgomp.oacc-c++/../libgomp.oacc-c-c++-common/acc_memcpy_device-1.c
-DACC_DEVICE_TYPE_nvidia=1 -DACC_MEM_SHARED=0 -foffload=nvptx-none -O0
execution test
PASS: libgomp.oacc-c++/../libgomp.oacc-c-c++-common/acc_memcpy_device-1.c
-DACC_DEVICE_TYPE_nvidia=1 -DACC_MEM_SHARED=0 -foffload=nvptx-none -O2 (test
for excess errors)
FAIL: libgomp.oacc-c++/../libgomp.oacc-c-c++-common/acc_memcpy_device-1.c
-DACC_DEVICE_TYPE_nvidia=1 -DACC_MEM_SHARED=0 -foffload=nvptx-none -O2
execution test
libgomp: cuStreamSynchronize error: unspecified launch failure (perhaps
abort was called)
I'll try to figure out what that is.
On OG15, for both nvptx and GCN offloading, I see:
PASS: libgomp.oacc-fortran/acc_memcpy_device-1.f90
-DACC_DEVICE_TYPE_nvidia=1 -DACC_MEM_SHARED=0 -foffload=nvptx-none -O (test
for excess errors)
FAIL: libgomp.oacc-fortran/acc_memcpy_device-1.f90
-DACC_DEVICE_TYPE_nvidia=1 -DACC_MEM_SHARED=0 -foffload=nvptx-none -O
execution test
PASS: libgomp.oacc-fortran/acc_memcpy_device-1.f90
-DACC_DEVICE_TYPE_radeon=1 -DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa -O
(test for excess errors)
FAIL: libgomp.oacc-fortran/acc_memcpy_device-1.f90
-DACC_DEVICE_TYPE_radeon=1 -DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa -O
execution test
libgomp: Trying to map into device [0x102701e0..0x102711e0) object when
[0x102701e0..0x102711e0) is already mapped
So, probably some OpenACC/Fortran mapping issue? Maybe you could have a look
at that one?
