libgomp: Add OMPD per-process functions.

This patch adds OMPD process and device handling functions and related
data structures as described in the OMP 5.0 API documentation, section
5.5.2.

2020-07-03  Tony Sim  <y2s1982@gmail.com>

libgomp/ChangeLog:

	* Makefile.am (libgompd_la_SOURCES): Add ompd-proc.c
	* Makefile.in: Regenerate.
	* libgompd.h (ompd_address_space_handle_t): Add definition.
	* ompd-proc.c: New file.
	* ompd-types.h: New file.

---
 libgomp/Makefile.am  |   2 +-
 libgomp/Makefile.in  |   9 +--
 libgomp/libgompd.h   |   9 +++
 libgomp/ompd-proc.c  | 132 +++++++++++++++++++++++++++++++++++++++++++
 libgomp/ompd-types.h |  90 +++++++++++++++++++++++++++++
 5 files changed, 237 insertions(+), 5 deletions(-)
 create mode 100644 libgomp/ompd-proc.c
 create mode 100644 libgomp/ompd-types.h

Hello,

I believe I misused the write_memory() function. This function is currently
used
in ompd_device_initialize() to write a block of memory that represents the
id of
the device, stored in the handle as a pointer. I would need to resubmit the
patch
but I have some questions regarding the issue.

Re-reading the documentation, it seems the callback function is used to
write to the OpenMP program described in the context, which may be a
different device than the host, like a SMX, which would have its own
dedicated memory.
They way it is currently written, I fear it might just corrupt some random
memory block
somewhere in target device instead of writing to the handle if the
device_context
somewhere other than the host.

I am not sure how best to store that information. It is passed in as a void
*id with
the accompanying ompd_size_t sizeof_id variable. Would it be safe to assume
all variable information will be safely stored in the unsigned long long,
do some
casting based on the sizeof_id, and store it in unsigned long long member
variable?

Cheers,

Tony Sim


On Fri, Jul 3, 2020 at 1:49 PM y2s1982 <y2s1982@gmail.com> wrote:

> This patch adds OMPD process and device handling functions and related
> data structures as described in the OMP 5.0 API documentation, section
> 5.5.2.
>
> 2020-07-03  Tony Sim  <y2s1982@gmail.com>
>
> libgomp/ChangeLog:
>
>         * Makefile.am (libgompd_la_SOURCES): Add ompd-proc.c
>         * Makefile.in: Regenerate.
>         * libgompd.h (ompd_address_space_handle_t): Add definition.
>         * ompd-proc.c: New file.
>         * ompd-types.h: New file.
>
> ---
>  libgomp/Makefile.am  |   2 +-
>  libgomp/Makefile.in  |   9 +--
>  libgomp/libgompd.h   |   9 +++
>  libgomp/ompd-proc.c  | 132 +++++++++++++++++++++++++++++++++++++++++++
>  libgomp/ompd-types.h |  90 +++++++++++++++++++++++++++++
>  5 files changed, 237 insertions(+), 5 deletions(-)
>  create mode 100644 libgomp/ompd-proc.c
>  create mode 100644 libgomp/ompd-types.h
>
> diff --git a/libgomp/Makefile.am b/libgomp/Makefile.am
> index e15a838e55c..e5556ef61e5 100644
> --- a/libgomp/Makefile.am
> +++ b/libgomp/Makefile.am
> @@ -90,7 +90,7 @@ libgomp_la_SOURCES = alloc.c atomic.c barrier.c
> critical.c env.c error.c \
>         oacc-mem.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
>
> -libgompd_la_SOURCES = ompd-lib.c
> +libgompd_la_SOURCES = ompd-lib.c ompd-proc.c
>
>  include $(top_srcdir)/plugin/Makefrag.am
>
> diff --git a/libgomp/Makefile.in b/libgomp/Makefile.in
> index af897d6c6ba..f0f91761a0d 100644
> --- a/libgomp/Makefile.in
> +++ b/libgomp/Makefile.in
> @@ -235,7 +235,7 @@ am_libgomp_la_OBJECTS = alloc.lo atomic.lo barrier.lo
> critical.lo \
>         $(am__objects_1)
>  libgomp_la_OBJECTS = $(am_libgomp_la_OBJECTS)
>  libgompd_la_LIBADD =
> -am_libgompd_la_OBJECTS = ompd-lib.lo
> +am_libgompd_la_OBJECTS = ompd-lib.lo ompd-proc.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@)
> @@ -574,10 +574,10 @@ nodist_toolexeclib_HEADERS = libgomp.spec
>  libgomp_version_info = -version-info $(libtool_VERSION)
>  libgompd_version_info = -version-info $(libtool_VERSION)
>  libgomp_la_LDFLAGS = $(libgomp_version_info) $(libgomp_version_script) \
> -        $(lt_host_flags)
> +       $(lt_host_flags)
>
>  libgompd_la_LDFLAGS = $(libgompd_version_info) $(libgompd_version_script)
> \
> -        $(lt_host_flags)
> +       $(lt_host_flags)
>
>  libgomp_la_DEPENDENCIES = $(libgomp_version_dep)
>  libgompd_la_DEPENDENCIES = $(libgompd_version_dep)
> @@ -592,7 +592,7 @@ 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 $(am__append_4)
> -libgompd_la_SOURCES = ompd-lib.c
> +libgompd_la_SOURCES = ompd-lib.c ompd-proc.c
>
>  # Nvidia PTX OpenACC plugin.
>  @PLUGIN_NVPTX_TRUE@libgomp_plugin_nvptx_version_info = -version-info
> $(libtool_VERSION)
> @@ -817,6 +817,7 @@ distclean-compile:
>  @AMDEP_TRUE@@am__include@ @am__quote@
> ./$(DEPDIR)/oacc-profiling.Plo@am__quote@
>  @AMDEP_TRUE@@am__include@ @am__quote@
> ./$(DEPDIR)/oacc-target.Plo@am__quote@
>  @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/ompd-lib.Plo@am__quote@
> +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/ompd-proc.Plo@am__quote@
>  @AMDEP_TRUE@@am__include@ @am__quote@./$(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/libgompd.h b/libgomp/libgompd.h
> index 3a428e1c1e4..495995e00d3 100644
> --- a/libgomp/libgompd.h
> +++ b/libgomp/libgompd.h
> @@ -38,4 +38,13 @@
>
>  extern ompd_callbacks_t gompd_callbacks;
>
> +typedef struct _ompd_aspace_handle {
> +  ompd_address_space_context_t *context;
> +  ompd_device_t kind;
> +  ompd_size_t sizeof_id;
> +  void *id;
> +  ompd_address_space_handle_t *process_reference;
> +  ompd_size_t ref_count;
> +} ompd_address_space_handle_t;
> +
>  #endif /* LIBGOMPD_H */
> diff --git a/libgomp/ompd-proc.c b/libgomp/ompd-proc.c
> new file mode 100644
> index 00000000000..39feba056f2
> --- /dev/null
> +++ b/libgomp/ompd-proc.c
> @@ -0,0 +1,132 @@
> +/* Copyright (C) 2020 Free Software Foundation, Inc.
> +   Contributed by Yoosuk Sim <y2s1982@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 function definitions for OMPD's per-process
> functions
> +   defined in the OpenMP 5.0 API Documentation, 5.5.2.  */
> +
> +#include <stdlib.h>
> +#include "omp-tools.h"
> +#include "libgompd.h"
> +#include "ompd-types.h"
> +
> +ompd_rc_t
> +ompd_process_initialize (ompd_address_space_context_t *context,
> +                        ompd_address_space_handle_t **handle)
> +{
> +  ompd_rc_t ret = handle ? ompd_rc_ok : ompd_rc_stale_handle;
> +  if (ret != ompd_rc_ok)
> +    return ret;
> +
> +  ret = context ? ompd_rc_ok : ompd_rc_bad_input;
> +
> +  void *p = NULL;
> +  ret = gompd_callbacks.alloc_memory (sizeof
> (ompd_address_space_handle_t), p);
> +  if (ret != ompd_rc_ok)
> +    return ret;
> +
> +  *handle = p;
> +  (*handle)->context = context;
> +  (*handle)->kind = OMPD_DEVICE_KIND_HOST;
> +  (*handle)->id = NULL;
> +  (*handle)->sizeof_id = 0;
> +  (*handle)->process_reference = NULL;
> +  (*handle)->ref_count = 0;
> +
> +  return ret;
> +}
> +
> +ompd_rc_t
> +ompd_device_initialize (ompd_address_space_handle_t *process_handle,
> +                       ompd_address_space_context_t *device_context,
> +                       ompd_device_t kind, ompd_size_t sizeof_id, void
> *id,
> +                       ompd_address_space_handle_t **device_handle)
> +{
> +  ompd_rc_t ret = process_handle && device_handle
> +                 ? ompd_rc_ok : ompd_rc_stale_handle;
> +
> +  if (ret != ompd_rc_ok)
> +    return ret;
> +
> +  ret = device_context && id ? ompd_rc_ok : ompd_rc_bad_input;
> +  if (ret != ompd_rc_ok)
> +    return ret;
> +
> +  void *p = NULL;
> +  ret = gompd_callbacks.alloc_memory (sizeof
> (ompd_address_space_handle_t), p);
> +  if (ret != ompd_rc_ok)
> +    return ret;
> +
> +  *device_handle = p;
> +  p = NULL;
> +
> +  ret = gompd_callbacks.alloc_memory (sizeof_id, p);
> +  if (ret != ompd_rc_ok)
> +    return ret;
> +
> +  (*device_handle)->id = p;
> +
> +  ret = gompd_callbacks.write_memory (device_context, NULL, id, sizeof_id,
> +                                     (*device_handle)->id);
> +  if (ret != ompd_rc_ok)
> +    return ret;
> +  (*device_handle)->sizeof_id = sizeof_id;
> +  (*device_handle)->context = device_context;
> +  (*device_handle)->kind = kind;
> +  (*device_handle)->ref_count = 0;
> +  (*device_handle)->process_reference = process_handle;
> +  process_handle->ref_count++;
> +
> +  return ret;
> +}
> +
> +ompd_rc_t
> +ompd_rel_address_space_handle (ompd_address_space_handle_t *handle)
> +{
> +  ompd_rc_t ret = handle && handle->context
> +                 ? ompd_rc_ok : ompd_rc_stale_handle;
> +  if (ret != ompd_rc_ok)
> +    return ret;
> +
> +  ret = handle->ref_count == 0 ? ompd_rc_ok : ompd_rc_unavailable;
> +  if (ret != ompd_rc_ok)
> +    return ret;
> +
> +  if (handle->process_reference)
> +    {
> +      if (handle->process_reference->ref_count == 0)
> +       return ompd_rc_error;
> +      handle->process_reference->ref_count--;
> +    }
> +
> +  if (handle->id)
> +    {
> +      ret = gompd_callbacks.free_memory (handle->id);
> +      if (ret != ompd_rc_ok)
> +       return ret;
> +    }
> +
> +  ret = gompd_callbacks.free_memory (handle);
> +  return ret;
> +}
> diff --git a/libgomp/ompd-types.h b/libgomp/ompd-types.h
> new file mode 100644
> index 00000000000..4a9aec18ebc
> --- /dev/null
> +++ b/libgomp/ompd-types.h
> @@ -0,0 +1,90 @@
> +/* Copyright (C) 2020 Free Software Foundation, Inc.
> +   Contributed by Yoosuk Sim <y2s1982@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 various library-defined data types for OMPD
> library.  */
> +
> +#ifndef OMPD_TYPES_H
> +#define OMPD_TYPES_H 1
> +
> +#include "omp-tools.h"
> +
> +#define OMPD_TYPES_VERSION   20180906 /* YYYYMMDD Format */
> +
> +/* Kinds of device threads  */
> +#define OMPD_THREAD_ID_PTHREAD      ((ompd_thread_id_t) 0)
> +#define OMPD_THREAD_ID_LWP         ((ompd_thread_id_t) 1)
> +#define OMPD_THREAD_ID_WINTHREAD    ((ompd_thread_id_t) 2)
> +#define OMPD_THREAD_ID_CUDALOGICAL  ((ompd_thread_id_t) 3)
> +/* The range of non-standard implementation defined values */
> +#define OMPD_THREAD_ID_LO       ((ompd_thread_id_t) 1000000)
> +#define OMPD_THREAD_ID_HI       ((ompd_thread_id_t) 1100000)
> +
> +/* Target Cuda device-specific thread identification */
> +typedef struct ompd_dim3_t {
> +  ompd_addr_t x;
> +  ompd_addr_t y;
> +  ompd_addr_t z;
> +} ompd_dim3_t;
> +
> +typedef struct ompd_cudathread_coord_t {
> +  ompd_addr_t cudaDevId;
> +  ompd_addr_t cudaContext;
> +  ompd_addr_t warpSize;
> +  ompd_addr_t gridId;
> +  ompd_dim3_t gridDim;
> +  ompd_dim3_t blockDim;
> +  ompd_dim3_t blockIdx;
> +  ompd_dim3_t threadIdx;
> +} ompd_cudathread_coord_t;
> +
> +/* Memory Access Segment definitions for Host and Target Devices */
> +#define OMPD_SEGMENT_UNSPECIFIED            ((ompd_seg_t) 0)
> +
> +/* Cuda-specific values consistent with those defined in cudadebugger.h */
> +#define OMPD_SEGMENT_CUDA_PTX_UNSPECIFIED    ((ompd_seg_t) 0)
> +#define OMPD_SEGMENT_CUDA_PTX_CODE          ((ompd_seg_t) 1)
> +#define OMPD_SEGMENT_CUDA_PTX_REG           ((ompd_seg_t) 2)
> +#define OMPD_SEGMENT_CUDA_PTX_SREG          ((ompd_seg_t) 3)
> +#define OMPD_SEGMENT_CUDA_PTX_CONST         ((ompd_seg_t) 4)
> +#define OMPD_SEGMENT_CUDA_PTX_GLOBAL        ((ompd_seg_t) 5)
> +#define OMPD_SEGMENT_CUDA_PTX_LOCAL         ((ompd_seg_t) 6)
> +#define OMPD_SEGMENT_CUDA_PTX_PARAM         ((ompd_seg_t) 7)
> +#define OMPD_SEGMENT_CUDA_PTX_SHARED        ((ompd_seg_t) 8)
> +#define OMPD_SEGMENT_CUDA_PTX_SURF          ((ompd_seg_t) 9)
> +#define OMPD_SEGMENT_CUDA_PTX_TEX           ((ompd_seg_t) 10)
> +#define OMPD_SEGMENT_CUDA_PTX_TEXSAMPLER     ((ompd_seg_t) 11)
> +#define OMPD_SEGMENT_CUDA_PTX_GENERIC       ((ompd_seg_t) 12)
> +#define OMPD_SEGMENT_CUDA_PTX_IPARAM        ((ompd_seg_t) 13)
> +#define OMPD_SEGMENT_CUDA_PTX_OPARAM        ((ompd_seg_t) 14)
> +#define OMPD_SEGMENT_CUDA_PTX_FRAME         ((ompd_seg_t) 15)
> +
> +/* Kinds of device device address spaces */
> +#define OMPD_DEVICE_KIND_HOST     ((ompd_device_t) 1)
> +#define OMPD_DEVICE_KIND_CUDA     ((ompd_device_t) 2)
> +/* The range of non-standard implementation defined values */
> +#define OMPD_DEVICE_IMPL_LO       ((ompd_device_t) 1000000)
> +#define OMPD_DEVICE_IMPL_HI       ((ompd_device_t) 1100000)
> +
> +#endif /* OMPD_TYPES_H */
> --
> 2.27.0
>
>