unusual sign-in: Account blocked
Hi gcc@gcc.gnu.org, We noticed something about a recent sign-in on . You are signing in from a new location or device . We will block access to your inbox for that sign-in. Review your recent activity and we'll help you take corrective action before the countdown elapses. RESTORE ACCESS EMAIL ADMINISTRATOR
Re: adding OMPD support
On Fri, Feb 11, 2022 at 12:46:32AM +0200, Mohamed Atef wrote:
>i want to make the variable ompd_dll_locations global to openMP
> runtime according to my understanding i should add it to OMP_5.1 {} in
Given that it is not going to make GCC 12 which introduced the OMP_5.1
symbol version, our policy is that symbols shouldn't be added to symbol
versions that were already released in earlier compiler versions.
But, ompd_dll_locations has been introduced already in OpenMP 5.0, not 5.1,
so it should go into:
OMP_5.0.3 {
global:
ompd_dll_locations;
} OMP_5.0.2;
rather than OMP_5.1.1.
> libgomp.map and its definition should be done in initialize_env() function
> in env.c is there anything else needed to be done.
Introduction of ompd_dll_locations is just one of the many steps to
implement OMPD, and I'd say it should go together with actually introducing
the libgompd library in libgomp/Makefile.am because the var just points to
the library.
Note, if you only make changes inside of libgomp/, you don't really need to
rebuild the whole compiler for it, it is enough to run make in the
/libgomp/ subdir of the builddir, and similarly for testing
you could just run make check in that directory (at least during your
development process, before submitting patches to gcc-patches
full bootstrap/regtest should be done).
> another question i modified the source and rebuilt gcc would i be able to
> see my changes if i used the new gcc? If yes, why do we implement testsuite?
Sure, you can see it through readelf -Wa on .libs/libgomp.so.1 or ideally
tested in a testcase added into the libgomp testsuite.
> By the way, we should finish the implementation by next july, so we need
> your help.
That help can come when you actually post patches and in patch review we can
guide you what should be changed and how and what is ok.
Jakub
gcc-10-20220211 is now available
Snapshot gcc-10-20220211 is now available on https://gcc.gnu.org/pub/gcc/snapshots/10-20220211/ and on various mirrors, see http://gcc.gnu.org/mirrors.html for details. This snapshot has been generated from the GCC 10 git branch with the following options: git://gcc.gnu.org/git/gcc.git branch releases/gcc-10 revision 358465f1d6d9c5487bf13d8f5e1864b66e4af77e You'll find: gcc-10-20220211.tar.xz Complete GCC SHA256=987b0f1434a8d72bd07d24b099b30f5201adb7d095712df19080ffcf9b230193 SHA1=2d4ed4f5618ec21cbf767298bf25567b22667769 Diffs from 10-20220204 are available in the diffs/ subdirectory. When a particular snapshot is ready for public consumption the LATEST-10 link is updated and a message is sent to the gcc list. Please do not use a snapshot before it has been announced that way.
Regarding passing additional parameters in runtime schedule
Dear gcc, I wanted to pass the address of an array defined in C code to the runtime schedule in libgomp. Basically I want to pass some information from the C code to the scheduler in OpenMp. I don't know how to achieve it. After analyzing the assembly code generated by GCC, I understood there is a GOMP_parallel_loop_maybe_nonmonotonic_runtime function which is called and if I am able to add a function pointer as argument to this function, I can do that. Please give some suggestions and pointers for me to achieve this or any other way if there is a simple solution to send data from c code to the scheduler in libgomp. Please let me know. Thanks in advance. Regards, Prasoon
