public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>
To: Ahmed Sayed Mousse via Gcc-patches <gcc-patches@gcc.gnu.org>
Cc: rep.dot.nop@gmail.com,
	Ahmed Sayed Mousse <ahmedsayedmousse@gmail.com>,
	jakub@redhat.com
Subject: Re: [patch] libgompd: Add thread handles
Date: Tue, 27 Sep 2022 08:40:53 +0200	[thread overview]
Message-ID: <20220927084053.78ea3f9b@nbbrfq> (raw)
In-Reply-To: <3835af74-e2f0-75eb-58a0-1ad04bc7d7f7@gmail.com>

On Tue, 27 Sep 2022 03:20:51 +0200
Ahmed Sayed Mousse via Gcc-patches <gcc-patches@gcc.gnu.org> wrote:

> diff --git a/libgomp/Makefile.am b/libgomp/Makefile.am
> index 6d913a93e7f..23f5bede1bf 100644
> --- a/libgomp/Makefile.am
> +++ b/libgomp/Makefile.am
> @@ -94,7 +94,7 @@ libgomp_la_SOURCES = alloc.c atomic.c barrier.c critical.c env.c error.c \
>  	priority_queue.c affinity-fmt.c teams.c allocator.c oacc-profiling.c \
>  	oacc-target.c ompd-support.c
>  
> -libgompd_la_SOURCES = ompd-init.c ompd-helper.c ompd-icv.c
> +libgompd_la_SOURCES = ompd-init.c ompd-helper.c ompd-icv.c ompd-threads.c
>  
>  include $(top_srcdir)/plugin/Makefrag.am
>  
> diff --git a/libgomp/Makefile.in b/libgomp/Makefile.in
> index 40f896b5f03..7acdcbf31d5 100644
> --- a/libgomp/Makefile.in
> +++ b/libgomp/Makefile.in
> @@ -233,7 +233,8 @@ am_libgomp_la_OBJECTS = alloc.lo atomic.lo barrier.lo critical.lo \
>  	affinity-fmt.lo teams.lo allocator.lo oacc-profiling.lo \
>  	oacc-target.lo ompd-support.lo $(am__objects_1)
>  libgomp_la_OBJECTS = $(am_libgomp_la_OBJECTS)
> -am_libgompd_la_OBJECTS = ompd-init.lo ompd-helper.lo ompd-icv.lo
> +am_libgompd_la_OBJECTS = ompd-init.lo ompd-helper.lo ompd-icv.lo \
> +	ompd-threads.lo
>  libgompd_la_OBJECTS = $(am_libgompd_la_OBJECTS)
>  AM_V_P = $(am__v_P_@AM_V@)
>  am__v_P_ = $(am__v_P_@AM_DEFAULT_V@)
> @@ -583,7 +584,7 @@ libgomp_la_SOURCES = alloc.c atomic.c barrier.c critical.c env.c \
>  	oacc-async.c oacc-plugin.c oacc-cuda.c priority_queue.c \
>  	affinity-fmt.c teams.c allocator.c oacc-profiling.c \
>  	oacc-target.c ompd-support.c $(am__append_7)
> -libgompd_la_SOURCES = ompd-init.c ompd-helper.c ompd-icv.c
> +libgompd_la_SOURCES = ompd-init.c ompd-helper.c ompd-icv.c ompd-threads.c
>  
>  # Nvidia PTX OpenACC plugin.
>  @PLUGIN_NVPTX_TRUE@libgomp_plugin_nvptx_version_info = -version-info $(libtool_VERSION)
> @@ -801,6 +802,7 @@ distclean-compile:
>  @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/ompd-icv.Plo@am__quote@
>  @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/ompd-init.Plo@am__quote@
>  @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/ompd-support.Plo@am__quote@
> +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/ompd-threads.Plo@am__quote@
>  @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/ordered.Plo@am__quote@
>  @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/parallel.Plo@am__quote@
>  @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/priority_queue.Plo@am__quote@
> diff --git a/libgomp/ompd-support.c b/libgomp/ompd-support.c
> index 27c5ad148e0..5b1afd37788 100644
> --- a/libgomp/ompd-support.c
> +++ b/libgomp/ompd-support.c
> @@ -33,6 +33,8 @@ const unsigned short gompd_sizeof_gomp_thread_handle
>    __attribute__ ((used)) OMPD_SECTION = 0;
>  #endif
>  
> +unsigned long gompd_thread_initial_tls_bias __attribute__ ((used));
> +
>  /* Get offset of the member m in struct t.  */
>  #define gompd_get_offset(t, m) \
>    const unsigned short gompd_access_##t##_##m __attribute__ ((used)) \
> @@ -67,6 +69,11 @@ gompd_load (void)
>    gompd_state |= OMPD_ENABLED;
>    ompd_dll_locations = &ompd_dll_locations_array[0];
>    ompd_dll_locations_valid ();
> +
> +  #if defined(LIBGOMP_USE_PTHREADS) && !defined(GOMP_NEEDS_THREAD_HANDLE)
> +  gompd_thread_initial_tls_bias = (unsigned long) ((char *) &gomp_tls_data
> +						   - (char *) pthread_self ());
> +  #endif
>  }
>  
>  #ifndef __ELF__
> diff --git a/libgomp/ompd-threads.c b/libgomp/ompd-threads.c
> new file mode 100644
> index 00000000000..723ef740181
> --- /dev/null
> +++ b/libgomp/ompd-threads.c
> @@ -0,0 +1,222 @@
> +/* Copyright (C) The GNU Toolchain Authors.
> +   Contributed by Ahmed Sayed <ahmedsayedmousse@gmail.com>.
> +   This file is part of the GNU Offloading and Multi Processing Library
> +   (libgomp).
> +
> +   Libgomp is free software; you can redistribute it and/or modify it
> +   under the terms of the GNU General Public License as published by
> +   the Free Software Foundation; either version 3, or (at your option)
> +   any later version.
> +
> +   Libgomp is distributed in the hope that it will be useful, but WITHOUT ANY
> +   WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
> +   FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
> +   more details.
> +
> +   Under Section 7 of GPL version 3, you are granted additional
> +   permissions described in the GCC Runtime Library Exception, version
> +   3.1, as published by the Free Software Foundation.
> +
> +   You should have received a copy of the GNU General Public License and
> +   a copy of the GCC Runtime Library Exception along with this program;
> +   see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
> +   <http://www.gnu.org/licenses/>.  */
> +
> +/* This file contains the implementation of functions defined in
> +   Section 5.5 ThreadHandles. */
> +
> +
> +#include "ompd-helper.h"
> +
> +ompd_rc_t
> +ompd_get_thread_in_parallel (ompd_parallel_handle_t *parallel_handle,
> +			     int thread_num,
> +			     ompd_thread_handle_t **thread_handle)
> +{
> +
> +  if (parallel_handle == NULL)
> +    return ompd_rc_stale_handle;
> +  CHECK (parallel_handle->ah);
> +
> +  ompd_address_space_context_t *context = parallel_handle->ah->context;
> +  ompd_rc_t ret;
> +
> +  ompd_word_t team_size_var = 1;
> +  if (parallel_handle->th.address)
> +    gompd_get_team_size (parallel_handle, &team_size_var);
> +
> +  if (thread_num < 0 || thread_num >= team_size_var)
> +    return ompd_rc_bad_input;
> +
> +  ompd_word_t temp_offset;
> +  ompd_address_t temp_symbol_addr, symbol_addr = {OMPD_SEGMENT_UNSPECIFIED, 0};
> +  ompd_addr_t temp_addr;
> +
> +  ACCESS_VALUE (context, NULL, "gompd_access_gomp_thread_pool_threads",
> +		temp_offset, 1, ret, symbol_addr, temp_symbol_addr, temp_addr);
> +
> +  symbol_addr.address += thread_num * target_sizes.sizeof_pointer;
> +
> +  DEREFERENCE (context, NULL, symbol_addr, target_sizes.sizeof_pointer, 1,
> +	       temp_addr, ret, 1);
> +
> +  ret = callbacks->alloc_memory (sizeof (ompd_thread_handle_t),
> +				 (void **) thread_handle);
> +
> +  CHECK_RET (ret);
> +
> +  if (symbol_addr.address == 0)
> +    return ompd_rc_unsupported;

Does the above leak the allocated memory, i.e. do you have to move the
check to right before the DEREFERENCE?

> +
> +  (*thread_handle)->th = symbol_addr;
> +  (*thread_handle)->ah = parallel_handle->ah;
> +  return ret;
> +}
> +
> +/* The ompd_get_thread_handle function that maps a native thread to an
> +   OMPD thread handle.  */
> +
> +ompd_rc_t
> +ompd_get_thread_handle (ompd_address_space_handle_t *handle,
> +			ompd_thread_id_t kind, ompd_size_t sizeof_thread_id,
> +			const void *thread_id,
> +			ompd_thread_handle_t **thread_handle)
> +{
> +  CHECK (handle);
> +  if (kind != OMPD_THREAD_ID_PTHREAD)
> +    return ompd_rc_unsupported;
> +
> +  ompd_address_space_context_t *context = handle->context;
> +  ompd_thread_context_t *tcontext;
> +  ompd_rc_t ret;
> +
> +  ret = callbacks->get_thread_context_for_thread_id (context, kind,
> +						     sizeof_thread_id,
> +						     thread_id, &tcontext);
> +  CHECK_RET (ret);
> +
> +  ompd_size_t temp_symbol_size, symbol_size;
> +  ompd_address_t temp_symbol_addr, symbol_addr = {OMPD_SEGMENT_UNSPECIFIED, 0};
> +
> +  GET_VALUE (context, NULL, "gompd_sizeof_gomp_thread", symbol_size,
> +	     temp_symbol_size, target_sizes.sizeof_short, 1, ret,
> +	     temp_symbol_addr);
> +
> +  GET_VALUE (context, tcontext, "gomp_tls_data", symbol_addr.address,
> +	     temp_symbol_addr.address, symbol_size, 1, ret, symbol_addr);
> +
> +  ret = callbacks->alloc_memory (sizeof (ompd_thread_handle_t),
> +				 (void **) thread_handle);
> +
> +  CHECK_RET (ret);
> +
> +  (*thread_handle)->ah = handle;
> +  (*thread_handle)->th = symbol_addr;
> +  (*thread_handle)->thread_context = tcontext;
> +  return ret;
> +}
> +
> +
> +ompd_rc_t
> +ompd_rel_thread_handle (ompd_thread_handle_t *thread_handle)
> +{
> +  if (thread_handle == NULL)
> +    return ompd_rc_stale_handle;
> +
> +  ompd_rc_t ret;
> +  ret = callbacks->free_memory ((void *) thread_handle);
> +  if (ret != ompd_rc_ok)
> +  return ret;

You seem to usually use CHECK_RET for the above.

> +
> +  return ompd_rc_ok;
> +}
> +
> +
> +/* Return -1, 0 or 1 for thread_handle_1 <, == or > thread_handle_2.  */
> +ompd_rc_t
> +ompd_thread_handle_compare (ompd_thread_handle_t *thread_handle_1,
> +			    ompd_thread_handle_t *thread_handle_2,
> +			    int	*cmp_value)
> +{
> +
> +  if (thread_handle_1 == NULL || thread_handle_2 == NULL)
> +    return ompd_rc_stale_handle;
> +  if (cmp_value == NULL)
> +    return ompd_rc_bad_input;
> +  if (thread_handle_1->ah->kind != thread_handle_2->ah->kind)
> +    return ompd_rc_bad_input;
> +
> +  if (thread_handle_1->th.address < thread_handle_2->th.address)
> +    *cmp_value = -1;
> +  else if (thread_handle_1->th.address > thread_handle_2->th.address)
> +    *cmp_value = 1;
> +  else
> +    *cmp_value = 0;
> +
> +  return ompd_rc_ok;
> +}
> +
> +
> +ompd_rc_t
> +ompd_get_thread_id (ompd_thread_handle_t *thread_handle, ompd_thread_id_t kind,
> +		    ompd_size_t sizeof_thread_id, void *thread_id)
> +{
> +  if (kind != OMPD_THREAD_ID_PTHREAD)
> +    return ompd_rc_unsupported;
> +  if (thread_id == NULL)
> +    return ompd_rc_bad_input;
> +  if (thread_handle == NULL)
> +    return ompd_rc_stale_handle;
> +
> +  CHECK (thread_handle->ah);
> +  ompd_address_space_context_t *context = thread_handle->ah->context;
> +
> +  ompd_rc_t ret;
> +  ompd_address_t taddr = thread_handle->th;
> +  ompd_address_t temp_symbol_addr, symbol_addr = {OMPD_SEGMENT_UNSPECIFIED, 0};
> +  ompd_size_t temp_symbol_size, symbol_size;
> +  ompd_word_t temp_offset, offset;
> +
> +  GET_VALUE (context, NULL, "gompd_sizeof_gomp_thread_handle", symbol_size,
> +	     temp_symbol_size, target_sizes.sizeof_short, 1, ret, symbol_addr);
> +
> +  if (symbol_size == 0)
> +    {
> +      GET_VALUE (context, NULL, "gompd_thread_initial_tls_bias", offset,
> +		 temp_offset, target_sizes.sizeof_long, 1, ret, symbol_addr);
> +
> +      ret = callbacks->symbol_addr_lookup (context, NULL,"gomp_tls_data",
> +					   &symbol_addr, NULL);

The above can never fail, no?
thanks,

> +      ret = callbacks->device_to_host (context, &temp_symbol_addr.address,
> +				       target_sizes.sizeof_long_long, 1,
> +				       &symbol_addr.address);
> +      CHECK_RET (ret);
> +
> +      taddr.address = symbol_addr.address + offset;
> +      ret = callbacks->read_memory (context, NULL, &taddr,
> +				    target_sizes.sizeof_long_long, thread_id);
> +    }
> +  else
> +    {
> +      if (sizeof_thread_id != symbol_size)
> +        return ompd_rc_bad_input;
> +
> +      GET_VALUE (context, NULL, "gompd_access_gomp_thread_handle", offset,
> +		 temp_offset, target_sizes.sizeof_short, 1, ret, symbol_addr);
> +      taddr.address += offset;
> +
> +      ret = callbacks->read_memory (context, NULL, &taddr, symbol_size,
> +				    thread_id);
> +    }
> +  return ret;
> +}
> +
> +
> +/* OMPD doesn't support GPUs for now.  */
> +ompd_rc_t ompd_get_device_from_thread (ompd_thread_handle_t *thread_handle,
> +				       ompd_address_space_handle_t **device)
> +{
> +  if (thread_handle == NULL)
> +    return ompd_rc_stale_handle;
> +  return ompd_rc_unsupported;
> +}
> diff --git a/libgomp/team.c b/libgomp/team.c
> index d53246961b7..9a84dc18bdb 100644
> --- a/libgomp/team.c
> +++ b/libgomp/team.c
> @@ -77,6 +77,7 @@ gomp_thread_start (void *xdata)
>    void *local_data;
>  
>    ompd_bp_thread_begin ();
> +
>  #if defined HAVE_TLS || defined USE_EMUTLS
>    thr = &gomp_tls_data;
>  #else
> @@ -313,6 +314,8 @@ gomp_free_thread (void *arg __attribute__((unused)))
>        gomp_end_task ();
>        free (task);
>      }
> +
> +  ompd_bp_thread_end ();
>  }
>  
>  /* Launch a team.  */


  reply	other threads:[~2022-09-27  6:42 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-09-27  1:12 Ahmed Sayed Mousse
2022-09-27  1:20 ` Ahmed Sayed Mousse
2022-09-27  6:40   ` Bernhard Reutner-Fischer [this message]
  -- strict thread matches above, loose matches on Subject: below --
2022-07-04 20:34 Ahmed Sayed Mousse
2022-07-27 12:11 ` Mohamed Atef
2022-07-29 11:22 ` Jakub Jelinek
2022-07-01  0:00 Ahmed Sayed Mousse
2022-06-06 22:21 Ahmed Sayed Mousse
2022-06-07  9:56 ` Jakub Jelinek

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20220927084053.78ea3f9b@nbbrfq \
    --to=rep.dot.nop@gmail.com \
    --cc=ahmedsayedmousse@gmail.com \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=jakub@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).