Hi!
On 2015-07-08T17:03:02+0200, I wrote:
> On Wed, 18 Feb 2015 09:50:15 +0100, I wrote:
>> So far, we have concentrated only on the 64-bit x86_64 configuration;
>> 32-bit has several known issues to be resolved.
>> <https://gcc.gnu.org/PR65099> filed.
(This still holds, and is unlikely to ever get addressed.)
> I have committed the following patch in r225560. This gets us rid of the
> lots of "expected FAILs" in the 32-bit part of
> RUNTESTFLAGS='--target_board=unix\{-m64,-m32\}' testing, for example.
>
> commit fe265ad3c9624da88f43be349137696449148f4f
> Author: tschwinge <tschwinge@138bc75d-0d04-0410-961f-82ee72b054a4>
> Date: Wed Jul 8 14:59:59 2015 +0000
>
> [nvptx offloading] Only 64-bit configurations are currently supported
>
> PR libgomp/65099
> gcc/
> * config/nvptx/mkoffload.c (main): Create an offload image only in
> 64-bit configurations.
(That remains in place.)
> libgomp/
> * plugin/plugin-nvptx.c (nvptx_get_num_devices): Return 0 if not
> in a 64-bit configuration.
That, for reasons given in the commit log, I've just refined, pushed
"[nvptx libgomp plugin] Build only in supported configurations" to master
branch in commit 6106dfb9f73a33c87108ad5b2dcd4842bdd7828e, and
cherry-picked into releases/gcc-10 branch in commit
1e56a7c9a6631b217299b2ddcd5c4d497bb3445e, releases/gcc-9 branch in commit
0f1e1069a753e912b058f0d4bf599f0edde28408, releases/gcc-8 branch in commit
f9267925c648f2ccd9e4680b699e581003125bcf, see attached.
Grüße
Thomas
-----------------
Mentor Graphics (Deutschland) GmbH, Arnulfstraße 201, 80634 München / Germany
Registergericht München HRB 106955, Geschäftsführer: Thomas Heurung, Alexander
Walter
>From 6106dfb9f73a33c87108ad5b2dcd4842bdd7828e Mon Sep 17 00:00:00 2001
From: Thomas Schwinge <tho...@codesourcery.com>
Date: Mon, 30 Nov 2020 15:15:20 +0100
Subject: [PATCH] [nvptx libgomp plugin] Build only in supported configurations
As recently again discussed in <https://gcc.gnu.org/PR97436> "[nvptx] -m32
support", nvptx offloading other than for 64-bit host has never been
implemented, tested, supported. So we simply should buildn't the nvptx libgomp
plugin in this case.
This avoids build problems if, for example, in a (standard) bi-arch
x86_64-pc-linux-gnu '-m64'/'-m32' build, libcuda is available only in a 64-bit
variant but not in a 32-bit one, which, for example, is the case if you build
GCC against the CUDA toolkit's 'stubs/libcuda.so' (see
<https://stackoverflow.com/a/52784819>).
This amends PR65099 commit a92defdab79a1268f4b9dcf42b937e4002a4cf15 (r225560)
"[nvptx offloading] Only 64-bit configurations are currently supported" to
match the way we're doing this for the HSA/GCN plugins.
libgomp/
PR libgomp/65099
* plugin/configfrag.ac (PLUGIN_NVPTX): Restrict to supported
configurations.
* configure: Regenerate.
* plugin/plugin-nvptx.c (nvptx_get_num_devices): Remove 64-bit
check.
---
libgomp/configure | 86 +++++++++++++++++++-------------
libgomp/plugin/configfrag.ac | 92 ++++++++++++++++++++---------------
libgomp/plugin/plugin-nvptx.c | 9 ----
3 files changed, 105 insertions(+), 82 deletions(-)
diff --git a/libgomp/configure b/libgomp/configure
index 89c17c571b7..48bf8e4a72c 100755
--- a/libgomp/configure
+++ b/libgomp/configure
@@ -15272,21 +15272,30 @@ if test x"$enable_offload_targets" != x; then
tgt_plugin=intelmic
;;
nvptx*)
- tgt_plugin=nvptx
- PLUGIN_NVPTX=$tgt
- if test "x$CUDA_DRIVER_LIB" != xno \
- && test "x$CUDA_DRIVER_LIB" != xno; then
- PLUGIN_NVPTX_CPPFLAGS=$CUDA_DRIVER_CPPFLAGS
- PLUGIN_NVPTX_LDFLAGS=$CUDA_DRIVER_LDFLAGS
- PLUGIN_NVPTX_LIBS='-lcuda'
-
- PLUGIN_NVPTX_save_CPPFLAGS=$CPPFLAGS
- CPPFLAGS="$PLUGIN_NVPTX_CPPFLAGS $CPPFLAGS"
- PLUGIN_NVPTX_save_LDFLAGS=$LDFLAGS
- LDFLAGS="$PLUGIN_NVPTX_LDFLAGS $LDFLAGS"
- PLUGIN_NVPTX_save_LIBS=$LIBS
- LIBS="$PLUGIN_NVPTX_LIBS $LIBS"
- cat confdefs.h - <<_ACEOF >conftest.$ac_ext
+ case "${target}" in
+ aarch64*-*-* | powerpc64le-*-* | x86_64-*-*)
+ case " ${CC} ${CFLAGS} " in
+ *" -m32 "* | *" -mx32 "*)
+ # PR libgomp/65099: Currently, we only support offloading in
+ # 64-bit configurations.
+ PLUGIN_NVPTX=0
+ ;;
+ *)
+ tgt_plugin=nvptx
+ PLUGIN_NVPTX=$tgt
+ if test "x$CUDA_DRIVER_LIB" != xno \
+ && test "x$CUDA_DRIVER_LIB" != xno; then
+ PLUGIN_NVPTX_CPPFLAGS=$CUDA_DRIVER_CPPFLAGS
+ PLUGIN_NVPTX_LDFLAGS=$CUDA_DRIVER_LDFLAGS
+ PLUGIN_NVPTX_LIBS='-lcuda'
+
+ PLUGIN_NVPTX_save_CPPFLAGS=$CPPFLAGS
+ CPPFLAGS="$PLUGIN_NVPTX_CPPFLAGS $CPPFLAGS"
+ PLUGIN_NVPTX_save_LDFLAGS=$LDFLAGS
+ LDFLAGS="$PLUGIN_NVPTX_LDFLAGS $LDFLAGS"
+ PLUGIN_NVPTX_save_LIBS=$LIBS
+ LIBS="$PLUGIN_NVPTX_LIBS $LIBS"
+ cat confdefs.h - <<_ACEOF >conftest.$ac_ext
/* end confdefs.h. */
#include "cuda.h"
int
@@ -15302,28 +15311,35 @@ if ac_fn_c_try_link "$LINENO"; then :
fi
rm -f core conftest.err conftest.$ac_objext \
conftest$ac_exeext conftest.$ac_ext
- CPPFLAGS=$PLUGIN_NVPTX_save_CPPFLAGS
- LDFLAGS=$PLUGIN_NVPTX_save_LDFLAGS
- LIBS=$PLUGIN_NVPTX_save_LIBS
- fi
- case $PLUGIN_NVPTX in
- nvptx*)
- if (test "x$CUDA_DRIVER_INCLUDE" = x \
- || test "x$CUDA_DRIVER_INCLUDE" = xno) \
- && (test "x$CUDA_DRIVER_LIB" = x \
- || test "x$CUDA_DRIVER_LIB" = xno); then
- PLUGIN_NVPTX=1
- PLUGIN_NVPTX_CPPFLAGS='-I$(srcdir)/plugin/cuda'
- PLUGIN_NVPTX_LIBS='-ldl'
- PLUGIN_NVPTX_DYNAMIC=1
- else
- PLUGIN_NVPTX=0
- as_fn_error $? "CUDA driver package required for nvptx support" "$LINENO" 5
- fi
- ;;
+ CPPFLAGS=$PLUGIN_NVPTX_save_CPPFLAGS
+ LDFLAGS=$PLUGIN_NVPTX_save_LDFLAGS
+ LIBS=$PLUGIN_NVPTX_save_LIBS
+ fi
+ case $PLUGIN_NVPTX in
+ nvptx*)
+ if (test "x$CUDA_DRIVER_INCLUDE" = x \
+ || test "x$CUDA_DRIVER_INCLUDE" = xno) \
+ && (test "x$CUDA_DRIVER_LIB" = x \
+ || test "x$CUDA_DRIVER_LIB" = xno); then
+ PLUGIN_NVPTX=1
+ PLUGIN_NVPTX_CPPFLAGS='-I$(srcdir)/plugin/cuda'
+ PLUGIN_NVPTX_LIBS='-ldl'
+ PLUGIN_NVPTX_DYNAMIC=1
+ else
+ PLUGIN_NVPTX=0
+ as_fn_error $? "CUDA driver package required for nvptx support" "$LINENO" 5
+ fi
+ ;;
+ esac
+ ;;
+ esac
+ ;;
+ *-*-*)
+ # Target architecture not supported.
+ PLUGIN_NVPTX=0
+ ;;
esac
;;
-
amdgcn*)
case "${target}" in
x86_64-*-*)
diff --git a/libgomp/plugin/configfrag.ac b/libgomp/plugin/configfrag.ac
index 2e086c017fe..88550982eab 100644
--- a/libgomp/plugin/configfrag.ac
+++ b/libgomp/plugin/configfrag.ac
@@ -158,47 +158,63 @@ if test x"$enable_offload_targets" != x; then
tgt_plugin=intelmic
;;
nvptx*)
- tgt_plugin=nvptx
- PLUGIN_NVPTX=$tgt
- if test "x$CUDA_DRIVER_LIB" != xno \
- && test "x$CUDA_DRIVER_LIB" != xno; then
- PLUGIN_NVPTX_CPPFLAGS=$CUDA_DRIVER_CPPFLAGS
- PLUGIN_NVPTX_LDFLAGS=$CUDA_DRIVER_LDFLAGS
- PLUGIN_NVPTX_LIBS='-lcuda'
+ case "${target}" in
+ aarch64*-*-* | powerpc64le-*-* | x86_64-*-*)
+ case " ${CC} ${CFLAGS} " in
+ *" -m32 "* | *" -mx32 "*)
+ # PR libgomp/65099: Currently, we only support offloading in
+ # 64-bit configurations.
+ PLUGIN_NVPTX=0
+ ;;
+ *)
+ tgt_plugin=nvptx
+ PLUGIN_NVPTX=$tgt
+ if test "x$CUDA_DRIVER_LIB" != xno \
+ && test "x$CUDA_DRIVER_LIB" != xno; then
+ PLUGIN_NVPTX_CPPFLAGS=$CUDA_DRIVER_CPPFLAGS
+ PLUGIN_NVPTX_LDFLAGS=$CUDA_DRIVER_LDFLAGS
+ PLUGIN_NVPTX_LIBS='-lcuda'
- PLUGIN_NVPTX_save_CPPFLAGS=$CPPFLAGS
- CPPFLAGS="$PLUGIN_NVPTX_CPPFLAGS $CPPFLAGS"
- PLUGIN_NVPTX_save_LDFLAGS=$LDFLAGS
- LDFLAGS="$PLUGIN_NVPTX_LDFLAGS $LDFLAGS"
- PLUGIN_NVPTX_save_LIBS=$LIBS
- LIBS="$PLUGIN_NVPTX_LIBS $LIBS"
- AC_LINK_IFELSE(
- [AC_LANG_PROGRAM(
- [#include "cuda.h"],
- [CUresult r = cuCtxPushCurrent (NULL);])],
- [PLUGIN_NVPTX=1])
- CPPFLAGS=$PLUGIN_NVPTX_save_CPPFLAGS
- LDFLAGS=$PLUGIN_NVPTX_save_LDFLAGS
- LIBS=$PLUGIN_NVPTX_save_LIBS
- fi
- case $PLUGIN_NVPTX in
- nvptx*)
- if (test "x$CUDA_DRIVER_INCLUDE" = x \
- || test "x$CUDA_DRIVER_INCLUDE" = xno) \
- && (test "x$CUDA_DRIVER_LIB" = x \
- || test "x$CUDA_DRIVER_LIB" = xno); then
- PLUGIN_NVPTX=1
- PLUGIN_NVPTX_CPPFLAGS='-I$(srcdir)/plugin/cuda'
- PLUGIN_NVPTX_LIBS='-ldl'
- PLUGIN_NVPTX_DYNAMIC=1
- else
- PLUGIN_NVPTX=0
- AC_MSG_ERROR([CUDA driver package required for nvptx support])
- fi
- ;;
+ PLUGIN_NVPTX_save_CPPFLAGS=$CPPFLAGS
+ CPPFLAGS="$PLUGIN_NVPTX_CPPFLAGS $CPPFLAGS"
+ PLUGIN_NVPTX_save_LDFLAGS=$LDFLAGS
+ LDFLAGS="$PLUGIN_NVPTX_LDFLAGS $LDFLAGS"
+ PLUGIN_NVPTX_save_LIBS=$LIBS
+ LIBS="$PLUGIN_NVPTX_LIBS $LIBS"
+ AC_LINK_IFELSE(
+ [AC_LANG_PROGRAM(
+ [#include "cuda.h"],
+ [CUresult r = cuCtxPushCurrent (NULL);])],
+ [PLUGIN_NVPTX=1])
+ CPPFLAGS=$PLUGIN_NVPTX_save_CPPFLAGS
+ LDFLAGS=$PLUGIN_NVPTX_save_LDFLAGS
+ LIBS=$PLUGIN_NVPTX_save_LIBS
+ fi
+ case $PLUGIN_NVPTX in
+ nvptx*)
+ if (test "x$CUDA_DRIVER_INCLUDE" = x \
+ || test "x$CUDA_DRIVER_INCLUDE" = xno) \
+ && (test "x$CUDA_DRIVER_LIB" = x \
+ || test "x$CUDA_DRIVER_LIB" = xno); then
+ PLUGIN_NVPTX=1
+ PLUGIN_NVPTX_CPPFLAGS='-I$(srcdir)/plugin/cuda'
+ PLUGIN_NVPTX_LIBS='-ldl'
+ PLUGIN_NVPTX_DYNAMIC=1
+ else
+ PLUGIN_NVPTX=0
+ AC_MSG_ERROR([CUDA driver package required for nvptx support])
+ fi
+ ;;
+ esac
+ ;;
+ esac
+ ;;
+ *-*-*)
+ # Target architecture not supported.
+ PLUGIN_NVPTX=0
+ ;;
esac
;;
-
amdgcn*)
case "${target}" in
x86_64-*-*)
diff --git a/libgomp/plugin/plugin-nvptx.c b/libgomp/plugin/plugin-nvptx.c
index 681c344b9c2..1215212d501 100644
--- a/libgomp/plugin/plugin-nvptx.c
+++ b/libgomp/plugin/plugin-nvptx.c
@@ -572,15 +572,6 @@ nvptx_get_num_devices (void)
{
int n;
- /* PR libgomp/65099: Currently, we only support offloading in 64-bit
- configurations. */
- if (sizeof (void *) != 8)
- {
- GOMP_PLUGIN_debug (0, "Disabling nvptx offloading;"
- " only 64-bit configurations are supported\n");
- return 0;
- }
-
/* This function will be called before the plugin has been initialized in
order to enumerate available devices, but CUDA API routines can't be used
until cuInit has been called. Just call it now (but don't yet do any
--
2.17.1
>From 1e56a7c9a6631b217299b2ddcd5c4d497bb3445e Mon Sep 17 00:00:00 2001
From: Thomas Schwinge <tho...@codesourcery.com>
Date: Mon, 30 Nov 2020 15:15:20 +0100
Subject: [PATCH] [nvptx libgomp plugin] Build only in supported configurations
As recently again discussed in <https://gcc.gnu.org/PR97436> "[nvptx] -m32
support", nvptx offloading other than for 64-bit host has never been
implemented, tested, supported. So we simply should buildn't the nvptx libgomp
plugin in this case.
This avoids build problems if, for example, in a (standard) bi-arch
x86_64-pc-linux-gnu '-m64'/'-m32' build, libcuda is available only in a 64-bit
variant but not in a 32-bit one, which, for example, is the case if you build
GCC against the CUDA toolkit's 'stubs/libcuda.so' (see
<https://stackoverflow.com/a/52784819>).
This amends PR65099 commit a92defdab79a1268f4b9dcf42b937e4002a4cf15 (r225560)
"[nvptx offloading] Only 64-bit configurations are currently supported" to
match the way we're doing this for the HSA/GCN plugins.
libgomp/
PR libgomp/65099
* plugin/configfrag.ac (PLUGIN_NVPTX): Restrict to supported
configurations.
* configure: Regenerate.
* plugin/plugin-nvptx.c (nvptx_get_num_devices): Remove 64-bit
check.
(cherry picked from commit 6106dfb9f73a33c87108ad5b2dcd4842bdd7828e)
---
libgomp/configure | 85 +++++++++++++++++++-------------
libgomp/plugin/configfrag.ac | 91 +++++++++++++++++++++--------------
libgomp/plugin/plugin-nvptx.c | 9 ----
3 files changed, 105 insertions(+), 80 deletions(-)
diff --git a/libgomp/configure b/libgomp/configure
index 69f57e31521..73f4a309f55 100755
--- a/libgomp/configure
+++ b/libgomp/configure
@@ -15294,21 +15294,30 @@ if test x"$enable_offload_targets" != x; then
tgt_plugin=intelmic
;;
nvptx*)
- tgt_plugin=nvptx
- PLUGIN_NVPTX=$tgt
- if test "x$CUDA_DRIVER_LIB" != xno \
- && test "x$CUDA_DRIVER_LIB" != xno; then
- PLUGIN_NVPTX_CPPFLAGS=$CUDA_DRIVER_CPPFLAGS
- PLUGIN_NVPTX_LDFLAGS=$CUDA_DRIVER_LDFLAGS
- PLUGIN_NVPTX_LIBS='-lcuda'
-
- PLUGIN_NVPTX_save_CPPFLAGS=$CPPFLAGS
- CPPFLAGS="$PLUGIN_NVPTX_CPPFLAGS $CPPFLAGS"
- PLUGIN_NVPTX_save_LDFLAGS=$LDFLAGS
- LDFLAGS="$PLUGIN_NVPTX_LDFLAGS $LDFLAGS"
- PLUGIN_NVPTX_save_LIBS=$LIBS
- LIBS="$PLUGIN_NVPTX_LIBS $LIBS"
- cat confdefs.h - <<_ACEOF >conftest.$ac_ext
+ case "${target}" in
+ aarch64*-*-* | powerpc64le-*-* | x86_64-*-*)
+ case " ${CC} ${CFLAGS} " in
+ *" -m32 "* | *" -mx32 "*)
+ # PR libgomp/65099: Currently, we only support offloading in
+ # 64-bit configurations.
+ PLUGIN_NVPTX=0
+ ;;
+ *)
+ tgt_plugin=nvptx
+ PLUGIN_NVPTX=$tgt
+ if test "x$CUDA_DRIVER_LIB" != xno \
+ && test "x$CUDA_DRIVER_LIB" != xno; then
+ PLUGIN_NVPTX_CPPFLAGS=$CUDA_DRIVER_CPPFLAGS
+ PLUGIN_NVPTX_LDFLAGS=$CUDA_DRIVER_LDFLAGS
+ PLUGIN_NVPTX_LIBS='-lcuda'
+
+ PLUGIN_NVPTX_save_CPPFLAGS=$CPPFLAGS
+ CPPFLAGS="$PLUGIN_NVPTX_CPPFLAGS $CPPFLAGS"
+ PLUGIN_NVPTX_save_LDFLAGS=$LDFLAGS
+ LDFLAGS="$PLUGIN_NVPTX_LDFLAGS $LDFLAGS"
+ PLUGIN_NVPTX_save_LIBS=$LIBS
+ LIBS="$PLUGIN_NVPTX_LIBS $LIBS"
+ cat confdefs.h - <<_ACEOF >conftest.$ac_ext
/* end confdefs.h. */
#include "cuda.h"
int
@@ -15324,25 +15333,33 @@ if ac_fn_c_try_link "$LINENO"; then :
fi
rm -f core conftest.err conftest.$ac_objext \
conftest$ac_exeext conftest.$ac_ext
- CPPFLAGS=$PLUGIN_NVPTX_save_CPPFLAGS
- LDFLAGS=$PLUGIN_NVPTX_save_LDFLAGS
- LIBS=$PLUGIN_NVPTX_save_LIBS
- fi
- case $PLUGIN_NVPTX in
- nvptx*)
- if (test "x$CUDA_DRIVER_INCLUDE" = x \
- || test "x$CUDA_DRIVER_INCLUDE" = xno) \
- && (test "x$CUDA_DRIVER_LIB" = x \
- || test "x$CUDA_DRIVER_LIB" = xno); then
- PLUGIN_NVPTX=1
- PLUGIN_NVPTX_CPPFLAGS='-I$(srcdir)/plugin/cuda'
- PLUGIN_NVPTX_LIBS='-ldl'
- PLUGIN_NVPTX_DYNAMIC=1
- else
- PLUGIN_NVPTX=0
- as_fn_error $? "CUDA driver package required for nvptx support" "$LINENO" 5
- fi
- ;;
+ CPPFLAGS=$PLUGIN_NVPTX_save_CPPFLAGS
+ LDFLAGS=$PLUGIN_NVPTX_save_LDFLAGS
+ LIBS=$PLUGIN_NVPTX_save_LIBS
+ fi
+ case $PLUGIN_NVPTX in
+ nvptx*)
+ if (test "x$CUDA_DRIVER_INCLUDE" = x \
+ || test "x$CUDA_DRIVER_INCLUDE" = xno) \
+ && (test "x$CUDA_DRIVER_LIB" = x \
+ || test "x$CUDA_DRIVER_LIB" = xno); then
+ PLUGIN_NVPTX=1
+ PLUGIN_NVPTX_CPPFLAGS='-I$(srcdir)/plugin/cuda'
+ PLUGIN_NVPTX_LIBS='-ldl'
+ PLUGIN_NVPTX_DYNAMIC=1
+ else
+ PLUGIN_NVPTX=0
+ as_fn_error $? "CUDA driver package required for nvptx support" "$LINENO" 5
+ fi
+ ;;
+ esac
+ ;;
+ esac
+ ;;
+ *-*-*)
+ # Target architecture not supported.
+ PLUGIN_NVPTX=0
+ ;;
esac
;;
hsa*)
diff --git a/libgomp/plugin/configfrag.ac b/libgomp/plugin/configfrag.ac
index fc91702a434..7eb137472c2 100644
--- a/libgomp/plugin/configfrag.ac
+++ b/libgomp/plugin/configfrag.ac
@@ -167,44 +167,61 @@ if test x"$enable_offload_targets" != x; then
tgt_plugin=intelmic
;;
nvptx*)
- tgt_plugin=nvptx
- PLUGIN_NVPTX=$tgt
- if test "x$CUDA_DRIVER_LIB" != xno \
- && test "x$CUDA_DRIVER_LIB" != xno; then
- PLUGIN_NVPTX_CPPFLAGS=$CUDA_DRIVER_CPPFLAGS
- PLUGIN_NVPTX_LDFLAGS=$CUDA_DRIVER_LDFLAGS
- PLUGIN_NVPTX_LIBS='-lcuda'
+ case "${target}" in
+ aarch64*-*-* | powerpc64le-*-* | x86_64-*-*)
+ case " ${CC} ${CFLAGS} " in
+ *" -m32 "* | *" -mx32 "*)
+ # PR libgomp/65099: Currently, we only support offloading in
+ # 64-bit configurations.
+ PLUGIN_NVPTX=0
+ ;;
+ *)
+ tgt_plugin=nvptx
+ PLUGIN_NVPTX=$tgt
+ if test "x$CUDA_DRIVER_LIB" != xno \
+ && test "x$CUDA_DRIVER_LIB" != xno; then
+ PLUGIN_NVPTX_CPPFLAGS=$CUDA_DRIVER_CPPFLAGS
+ PLUGIN_NVPTX_LDFLAGS=$CUDA_DRIVER_LDFLAGS
+ PLUGIN_NVPTX_LIBS='-lcuda'
- PLUGIN_NVPTX_save_CPPFLAGS=$CPPFLAGS
- CPPFLAGS="$PLUGIN_NVPTX_CPPFLAGS $CPPFLAGS"
- PLUGIN_NVPTX_save_LDFLAGS=$LDFLAGS
- LDFLAGS="$PLUGIN_NVPTX_LDFLAGS $LDFLAGS"
- PLUGIN_NVPTX_save_LIBS=$LIBS
- LIBS="$PLUGIN_NVPTX_LIBS $LIBS"
- AC_LINK_IFELSE(
- [AC_LANG_PROGRAM(
- [#include "cuda.h"],
- [CUresult r = cuCtxPushCurrent (NULL);])],
- [PLUGIN_NVPTX=1])
- CPPFLAGS=$PLUGIN_NVPTX_save_CPPFLAGS
- LDFLAGS=$PLUGIN_NVPTX_save_LDFLAGS
- LIBS=$PLUGIN_NVPTX_save_LIBS
- fi
- case $PLUGIN_NVPTX in
- nvptx*)
- if (test "x$CUDA_DRIVER_INCLUDE" = x \
- || test "x$CUDA_DRIVER_INCLUDE" = xno) \
- && (test "x$CUDA_DRIVER_LIB" = x \
- || test "x$CUDA_DRIVER_LIB" = xno); then
- PLUGIN_NVPTX=1
- PLUGIN_NVPTX_CPPFLAGS='-I$(srcdir)/plugin/cuda'
- PLUGIN_NVPTX_LIBS='-ldl'
- PLUGIN_NVPTX_DYNAMIC=1
- else
- PLUGIN_NVPTX=0
- AC_MSG_ERROR([CUDA driver package required for nvptx support])
- fi
- ;;
+ PLUGIN_NVPTX_save_CPPFLAGS=$CPPFLAGS
+ CPPFLAGS="$PLUGIN_NVPTX_CPPFLAGS $CPPFLAGS"
+ PLUGIN_NVPTX_save_LDFLAGS=$LDFLAGS
+ LDFLAGS="$PLUGIN_NVPTX_LDFLAGS $LDFLAGS"
+ PLUGIN_NVPTX_save_LIBS=$LIBS
+ LIBS="$PLUGIN_NVPTX_LIBS $LIBS"
+ AC_LINK_IFELSE(
+ [AC_LANG_PROGRAM(
+ [#include "cuda.h"],
+ [CUresult r = cuCtxPushCurrent (NULL);])],
+ [PLUGIN_NVPTX=1])
+ CPPFLAGS=$PLUGIN_NVPTX_save_CPPFLAGS
+ LDFLAGS=$PLUGIN_NVPTX_save_LDFLAGS
+ LIBS=$PLUGIN_NVPTX_save_LIBS
+ fi
+ case $PLUGIN_NVPTX in
+ nvptx*)
+ if (test "x$CUDA_DRIVER_INCLUDE" = x \
+ || test "x$CUDA_DRIVER_INCLUDE" = xno) \
+ && (test "x$CUDA_DRIVER_LIB" = x \
+ || test "x$CUDA_DRIVER_LIB" = xno); then
+ PLUGIN_NVPTX=1
+ PLUGIN_NVPTX_CPPFLAGS='-I$(srcdir)/plugin/cuda'
+ PLUGIN_NVPTX_LIBS='-ldl'
+ PLUGIN_NVPTX_DYNAMIC=1
+ else
+ PLUGIN_NVPTX=0
+ AC_MSG_ERROR([CUDA driver package required for nvptx support])
+ fi
+ ;;
+ esac
+ ;;
+ esac
+ ;;
+ *-*-*)
+ # Target architecture not supported.
+ PLUGIN_NVPTX=0
+ ;;
esac
;;
hsa*)
diff --git a/libgomp/plugin/plugin-nvptx.c b/libgomp/plugin/plugin-nvptx.c
index 390804ad1fa..ee4a3ed2264 100644
--- a/libgomp/plugin/plugin-nvptx.c
+++ b/libgomp/plugin/plugin-nvptx.c
@@ -546,15 +546,6 @@ nvptx_get_num_devices (void)
{
int n;
- /* PR libgomp/65099: Currently, we only support offloading in 64-bit
- configurations. */
- if (sizeof (void *) != 8)
- {
- GOMP_PLUGIN_debug (0, "Disabling nvptx offloading;"
- " only 64-bit configurations are supported\n");
- return 0;
- }
-
/* This function will be called before the plugin has been initialized in
order to enumerate available devices, but CUDA API routines can't be used
until cuInit has been called. Just call it now (but don't yet do any
--
2.17.1
>From 0f1e1069a753e912b058f0d4bf599f0edde28408 Mon Sep 17 00:00:00 2001
From: Thomas Schwinge <tho...@codesourcery.com>
Date: Mon, 30 Nov 2020 15:15:20 +0100
Subject: [PATCH] [nvptx libgomp plugin] Build only in supported configurations
As recently again discussed in <https://gcc.gnu.org/PR97436> "[nvptx] -m32
support", nvptx offloading other than for 64-bit host has never been
implemented, tested, supported. So we simply should buildn't the nvptx libgomp
plugin in this case.
This avoids build problems if, for example, in a (standard) bi-arch
x86_64-pc-linux-gnu '-m64'/'-m32' build, libcuda is available only in a 64-bit
variant but not in a 32-bit one, which, for example, is the case if you build
GCC against the CUDA toolkit's 'stubs/libcuda.so' (see
<https://stackoverflow.com/a/52784819>).
This amends PR65099 commit a92defdab79a1268f4b9dcf42b937e4002a4cf15 (r225560)
"[nvptx offloading] Only 64-bit configurations are currently supported" to
match the way we're doing this for the HSA/GCN plugins.
libgomp/
PR libgomp/65099
* plugin/configfrag.ac (PLUGIN_NVPTX): Restrict to supported
configurations.
* configure: Regenerate.
* plugin/plugin-nvptx.c (nvptx_get_num_devices): Remove 64-bit
check.
(cherry picked from commit 6106dfb9f73a33c87108ad5b2dcd4842bdd7828e)
---
libgomp/configure | 85 +++++++++++++++++++-------------
libgomp/plugin/configfrag.ac | 91 +++++++++++++++++++++--------------
libgomp/plugin/plugin-nvptx.c | 9 ----
3 files changed, 105 insertions(+), 80 deletions(-)
diff --git a/libgomp/configure b/libgomp/configure
index b4bc4f43628..de31f97c2c6 100755
--- a/libgomp/configure
+++ b/libgomp/configure
@@ -15641,21 +15641,30 @@ if test x"$enable_offload_targets" != x; then
tgt_plugin=intelmic
;;
nvptx*)
- tgt_plugin=nvptx
- PLUGIN_NVPTX=$tgt
- if test "x$CUDA_DRIVER_LIB" != xno \
- && test "x$CUDA_DRIVER_LIB" != xno; then
- PLUGIN_NVPTX_CPPFLAGS=$CUDA_DRIVER_CPPFLAGS
- PLUGIN_NVPTX_LDFLAGS=$CUDA_DRIVER_LDFLAGS
- PLUGIN_NVPTX_LIBS='-lcuda'
-
- PLUGIN_NVPTX_save_CPPFLAGS=$CPPFLAGS
- CPPFLAGS="$PLUGIN_NVPTX_CPPFLAGS $CPPFLAGS"
- PLUGIN_NVPTX_save_LDFLAGS=$LDFLAGS
- LDFLAGS="$PLUGIN_NVPTX_LDFLAGS $LDFLAGS"
- PLUGIN_NVPTX_save_LIBS=$LIBS
- LIBS="$PLUGIN_NVPTX_LIBS $LIBS"
- cat confdefs.h - <<_ACEOF >conftest.$ac_ext
+ case "${target}" in
+ aarch64*-*-* | powerpc64le-*-* | x86_64-*-*)
+ case " ${CC} ${CFLAGS} " in
+ *" -m32 "* | *" -mx32 "*)
+ # PR libgomp/65099: Currently, we only support offloading in
+ # 64-bit configurations.
+ PLUGIN_NVPTX=0
+ ;;
+ *)
+ tgt_plugin=nvptx
+ PLUGIN_NVPTX=$tgt
+ if test "x$CUDA_DRIVER_LIB" != xno \
+ && test "x$CUDA_DRIVER_LIB" != xno; then
+ PLUGIN_NVPTX_CPPFLAGS=$CUDA_DRIVER_CPPFLAGS
+ PLUGIN_NVPTX_LDFLAGS=$CUDA_DRIVER_LDFLAGS
+ PLUGIN_NVPTX_LIBS='-lcuda'
+
+ PLUGIN_NVPTX_save_CPPFLAGS=$CPPFLAGS
+ CPPFLAGS="$PLUGIN_NVPTX_CPPFLAGS $CPPFLAGS"
+ PLUGIN_NVPTX_save_LDFLAGS=$LDFLAGS
+ LDFLAGS="$PLUGIN_NVPTX_LDFLAGS $LDFLAGS"
+ PLUGIN_NVPTX_save_LIBS=$LIBS
+ LIBS="$PLUGIN_NVPTX_LIBS $LIBS"
+ cat confdefs.h - <<_ACEOF >conftest.$ac_ext
/* end confdefs.h. */
#include "cuda.h"
int
@@ -15671,25 +15680,33 @@ if ac_fn_c_try_link "$LINENO"; then :
fi
rm -f core conftest.err conftest.$ac_objext \
conftest$ac_exeext conftest.$ac_ext
- CPPFLAGS=$PLUGIN_NVPTX_save_CPPFLAGS
- LDFLAGS=$PLUGIN_NVPTX_save_LDFLAGS
- LIBS=$PLUGIN_NVPTX_save_LIBS
- fi
- case $PLUGIN_NVPTX in
- nvptx*)
- if (test "x$CUDA_DRIVER_INCLUDE" = x \
- || test "x$CUDA_DRIVER_INCLUDE" = xno) \
- && (test "x$CUDA_DRIVER_LIB" = x \
- || test "x$CUDA_DRIVER_LIB" = xno); then
- PLUGIN_NVPTX=1
- PLUGIN_NVPTX_CPPFLAGS='-I$(srcdir)/plugin/cuda'
- PLUGIN_NVPTX_LIBS='-ldl'
- PLUGIN_NVPTX_DYNAMIC=1
- else
- PLUGIN_NVPTX=0
- as_fn_error $? "CUDA driver package required for nvptx support" "$LINENO" 5
- fi
- ;;
+ CPPFLAGS=$PLUGIN_NVPTX_save_CPPFLAGS
+ LDFLAGS=$PLUGIN_NVPTX_save_LDFLAGS
+ LIBS=$PLUGIN_NVPTX_save_LIBS
+ fi
+ case $PLUGIN_NVPTX in
+ nvptx*)
+ if (test "x$CUDA_DRIVER_INCLUDE" = x \
+ || test "x$CUDA_DRIVER_INCLUDE" = xno) \
+ && (test "x$CUDA_DRIVER_LIB" = x \
+ || test "x$CUDA_DRIVER_LIB" = xno); then
+ PLUGIN_NVPTX=1
+ PLUGIN_NVPTX_CPPFLAGS='-I$(srcdir)/plugin/cuda'
+ PLUGIN_NVPTX_LIBS='-ldl'
+ PLUGIN_NVPTX_DYNAMIC=1
+ else
+ PLUGIN_NVPTX=0
+ as_fn_error $? "CUDA driver package required for nvptx support" "$LINENO" 5
+ fi
+ ;;
+ esac
+ ;;
+ esac
+ ;;
+ *-*-*)
+ # Target architecture not supported.
+ PLUGIN_NVPTX=0
+ ;;
esac
;;
hsa*)
diff --git a/libgomp/plugin/configfrag.ac b/libgomp/plugin/configfrag.ac
index 9718ac752e2..77e1cda1a73 100644
--- a/libgomp/plugin/configfrag.ac
+++ b/libgomp/plugin/configfrag.ac
@@ -158,44 +158,61 @@ if test x"$enable_offload_targets" != x; then
tgt_plugin=intelmic
;;
nvptx*)
- tgt_plugin=nvptx
- PLUGIN_NVPTX=$tgt
- if test "x$CUDA_DRIVER_LIB" != xno \
- && test "x$CUDA_DRIVER_LIB" != xno; then
- PLUGIN_NVPTX_CPPFLAGS=$CUDA_DRIVER_CPPFLAGS
- PLUGIN_NVPTX_LDFLAGS=$CUDA_DRIVER_LDFLAGS
- PLUGIN_NVPTX_LIBS='-lcuda'
+ case "${target}" in
+ aarch64*-*-* | powerpc64le-*-* | x86_64-*-*)
+ case " ${CC} ${CFLAGS} " in
+ *" -m32 "* | *" -mx32 "*)
+ # PR libgomp/65099: Currently, we only support offloading in
+ # 64-bit configurations.
+ PLUGIN_NVPTX=0
+ ;;
+ *)
+ tgt_plugin=nvptx
+ PLUGIN_NVPTX=$tgt
+ if test "x$CUDA_DRIVER_LIB" != xno \
+ && test "x$CUDA_DRIVER_LIB" != xno; then
+ PLUGIN_NVPTX_CPPFLAGS=$CUDA_DRIVER_CPPFLAGS
+ PLUGIN_NVPTX_LDFLAGS=$CUDA_DRIVER_LDFLAGS
+ PLUGIN_NVPTX_LIBS='-lcuda'
- PLUGIN_NVPTX_save_CPPFLAGS=$CPPFLAGS
- CPPFLAGS="$PLUGIN_NVPTX_CPPFLAGS $CPPFLAGS"
- PLUGIN_NVPTX_save_LDFLAGS=$LDFLAGS
- LDFLAGS="$PLUGIN_NVPTX_LDFLAGS $LDFLAGS"
- PLUGIN_NVPTX_save_LIBS=$LIBS
- LIBS="$PLUGIN_NVPTX_LIBS $LIBS"
- AC_LINK_IFELSE(
- [AC_LANG_PROGRAM(
- [#include "cuda.h"],
- [CUresult r = cuCtxPushCurrent (NULL);])],
- [PLUGIN_NVPTX=1])
- CPPFLAGS=$PLUGIN_NVPTX_save_CPPFLAGS
- LDFLAGS=$PLUGIN_NVPTX_save_LDFLAGS
- LIBS=$PLUGIN_NVPTX_save_LIBS
- fi
- case $PLUGIN_NVPTX in
- nvptx*)
- if (test "x$CUDA_DRIVER_INCLUDE" = x \
- || test "x$CUDA_DRIVER_INCLUDE" = xno) \
- && (test "x$CUDA_DRIVER_LIB" = x \
- || test "x$CUDA_DRIVER_LIB" = xno); then
- PLUGIN_NVPTX=1
- PLUGIN_NVPTX_CPPFLAGS='-I$(srcdir)/plugin/cuda'
- PLUGIN_NVPTX_LIBS='-ldl'
- PLUGIN_NVPTX_DYNAMIC=1
- else
- PLUGIN_NVPTX=0
- AC_MSG_ERROR([CUDA driver package required for nvptx support])
- fi
- ;;
+ PLUGIN_NVPTX_save_CPPFLAGS=$CPPFLAGS
+ CPPFLAGS="$PLUGIN_NVPTX_CPPFLAGS $CPPFLAGS"
+ PLUGIN_NVPTX_save_LDFLAGS=$LDFLAGS
+ LDFLAGS="$PLUGIN_NVPTX_LDFLAGS $LDFLAGS"
+ PLUGIN_NVPTX_save_LIBS=$LIBS
+ LIBS="$PLUGIN_NVPTX_LIBS $LIBS"
+ AC_LINK_IFELSE(
+ [AC_LANG_PROGRAM(
+ [#include "cuda.h"],
+ [CUresult r = cuCtxPushCurrent (NULL);])],
+ [PLUGIN_NVPTX=1])
+ CPPFLAGS=$PLUGIN_NVPTX_save_CPPFLAGS
+ LDFLAGS=$PLUGIN_NVPTX_save_LDFLAGS
+ LIBS=$PLUGIN_NVPTX_save_LIBS
+ fi
+ case $PLUGIN_NVPTX in
+ nvptx*)
+ if (test "x$CUDA_DRIVER_INCLUDE" = x \
+ || test "x$CUDA_DRIVER_INCLUDE" = xno) \
+ && (test "x$CUDA_DRIVER_LIB" = x \
+ || test "x$CUDA_DRIVER_LIB" = xno); then
+ PLUGIN_NVPTX=1
+ PLUGIN_NVPTX_CPPFLAGS='-I$(srcdir)/plugin/cuda'
+ PLUGIN_NVPTX_LIBS='-ldl'
+ PLUGIN_NVPTX_DYNAMIC=1
+ else
+ PLUGIN_NVPTX=0
+ AC_MSG_ERROR([CUDA driver package required for nvptx support])
+ fi
+ ;;
+ esac
+ ;;
+ esac
+ ;;
+ *-*-*)
+ # Target architecture not supported.
+ PLUGIN_NVPTX=0
+ ;;
esac
;;
hsa*)
diff --git a/libgomp/plugin/plugin-nvptx.c b/libgomp/plugin/plugin-nvptx.c
index 387e7cc6dd3..eaa8a956573 100644
--- a/libgomp/plugin/plugin-nvptx.c
+++ b/libgomp/plugin/plugin-nvptx.c
@@ -876,15 +876,6 @@ nvptx_get_num_devices (void)
{
int n;
- /* PR libgomp/65099: Currently, we only support offloading in 64-bit
- configurations. */
- if (sizeof (void *) != 8)
- {
- GOMP_PLUGIN_debug (0, "Disabling nvptx offloading;"
- " only 64-bit configurations are supported\n");
- return 0;
- }
-
/* This function will be called before the plugin has been initialized in
order to enumerate available devices, but CUDA API routines can't be used
until cuInit has been called. Just call it now (but don't yet do any
--
2.17.1
>From f9267925c648f2ccd9e4680b699e581003125bcf Mon Sep 17 00:00:00 2001
From: Thomas Schwinge <tho...@codesourcery.com>
Date: Mon, 30 Nov 2020 15:15:20 +0100
Subject: [PATCH] [nvptx libgomp plugin] Build only in supported configurations
As recently again discussed in <https://gcc.gnu.org/PR97436> "[nvptx] -m32
support", nvptx offloading other than for 64-bit host has never been
implemented, tested, supported. So we simply should buildn't the nvptx libgomp
plugin in this case.
This avoids build problems if, for example, in a (standard) bi-arch
x86_64-pc-linux-gnu '-m64'/'-m32' build, libcuda is available only in a 64-bit
variant but not in a 32-bit one, which, for example, is the case if you build
GCC against the CUDA toolkit's 'stubs/libcuda.so' (see
<https://stackoverflow.com/a/52784819>).
This amends PR65099 commit a92defdab79a1268f4b9dcf42b937e4002a4cf15 (r225560)
"[nvptx offloading] Only 64-bit configurations are currently supported" to
match the way we're doing this for the HSA/GCN plugins.
libgomp/
PR libgomp/65099
* plugin/configfrag.ac (PLUGIN_NVPTX): Restrict to supported
configurations.
* configure: Regenerate.
* plugin/plugin-nvptx.c (nvptx_get_num_devices): Remove 64-bit
check.
(cherry picked from commit 6106dfb9f73a33c87108ad5b2dcd4842bdd7828e)
---
libgomp/configure | 75 +++++++++++++++++++-------------
libgomp/plugin/configfrag.ac | 81 +++++++++++++++++++++--------------
libgomp/plugin/plugin-nvptx.c | 9 ----
3 files changed, 95 insertions(+), 70 deletions(-)
diff --git a/libgomp/configure b/libgomp/configure
index ced7606b355..2529a8e0603 100755
--- a/libgomp/configure
+++ b/libgomp/configure
@@ -15398,19 +15398,28 @@ if test x"$enable_offload_targets" != x; then
tgt_name=intelmic
;;
nvptx*)
- tgt_name=nvptx
- PLUGIN_NVPTX=$tgt
- PLUGIN_NVPTX_CPPFLAGS=$CUDA_DRIVER_CPPFLAGS
- PLUGIN_NVPTX_LDFLAGS=$CUDA_DRIVER_LDFLAGS
- PLUGIN_NVPTX_LIBS='-lcuda'
-
- PLUGIN_NVPTX_save_CPPFLAGS=$CPPFLAGS
- CPPFLAGS="$PLUGIN_NVPTX_CPPFLAGS $CPPFLAGS"
- PLUGIN_NVPTX_save_LDFLAGS=$LDFLAGS
- LDFLAGS="$PLUGIN_NVPTX_LDFLAGS $LDFLAGS"
- PLUGIN_NVPTX_save_LIBS=$LIBS
- LIBS="$PLUGIN_NVPTX_LIBS $LIBS"
- cat confdefs.h - <<_ACEOF >conftest.$ac_ext
+ case "${target}" in
+ aarch64*-*-* | powerpc64le-*-* | x86_64-*-*)
+ case " ${CC} ${CFLAGS} " in
+ *" -m32 "* | *" -mx32 "*)
+ # PR libgomp/65099: Currently, we only support offloading in
+ # 64-bit configurations.
+ PLUGIN_NVPTX=0
+ ;;
+ *)
+ tgt_name=nvptx
+ PLUGIN_NVPTX=$tgt
+ PLUGIN_NVPTX_CPPFLAGS=$CUDA_DRIVER_CPPFLAGS
+ PLUGIN_NVPTX_LDFLAGS=$CUDA_DRIVER_LDFLAGS
+ PLUGIN_NVPTX_LIBS='-lcuda'
+
+ PLUGIN_NVPTX_save_CPPFLAGS=$CPPFLAGS
+ CPPFLAGS="$PLUGIN_NVPTX_CPPFLAGS $CPPFLAGS"
+ PLUGIN_NVPTX_save_LDFLAGS=$LDFLAGS
+ LDFLAGS="$PLUGIN_NVPTX_LDFLAGS $LDFLAGS"
+ PLUGIN_NVPTX_save_LIBS=$LIBS
+ LIBS="$PLUGIN_NVPTX_LIBS $LIBS"
+ cat confdefs.h - <<_ACEOF >conftest.$ac_ext
/* end confdefs.h. */
#include "cuda.h"
int
@@ -15426,22 +15435,30 @@ if ac_fn_c_try_link "$LINENO"; then :
fi
rm -f core conftest.err conftest.$ac_objext \
conftest$ac_exeext conftest.$ac_ext
- CPPFLAGS=$PLUGIN_NVPTX_save_CPPFLAGS
- LDFLAGS=$PLUGIN_NVPTX_save_LDFLAGS
- LIBS=$PLUGIN_NVPTX_save_LIBS
- case $PLUGIN_NVPTX in
- nvptx*)
- if test "x$CUDA_DRIVER_INCLUDE" = x \
- && test "x$CUDA_DRIVER_LIB" = x; then
- PLUGIN_NVPTX=1
- PLUGIN_NVPTX_CPPFLAGS='-I$(srcdir)/plugin/cuda'
- PLUGIN_NVPTX_LIBS='-ldl'
- PLUGIN_NVPTX_DYNAMIC=1
- else
- PLUGIN_NVPTX=0
- as_fn_error "CUDA driver package required for nvptx support" "$LINENO" 5
- fi
- ;;
+ CPPFLAGS=$PLUGIN_NVPTX_save_CPPFLAGS
+ LDFLAGS=$PLUGIN_NVPTX_save_LDFLAGS
+ LIBS=$PLUGIN_NVPTX_save_LIBS
+ case $PLUGIN_NVPTX in
+ nvptx*)
+ if test "x$CUDA_DRIVER_INCLUDE" = x \
+ && test "x$CUDA_DRIVER_LIB" = x; then
+ PLUGIN_NVPTX=1
+ PLUGIN_NVPTX_CPPFLAGS='-I$(srcdir)/plugin/cuda'
+ PLUGIN_NVPTX_LIBS='-ldl'
+ PLUGIN_NVPTX_DYNAMIC=1
+ else
+ PLUGIN_NVPTX=0
+ as_fn_error "CUDA driver package required for nvptx support" "$LINENO" 5
+ fi
+ ;;
+ esac
+ ;;
+ esac
+ ;;
+ *-*-*)
+ # Target architecture not supported.
+ PLUGIN_NVPTX=0
+ ;;
esac
;;
hsa*)
diff --git a/libgomp/plugin/configfrag.ac b/libgomp/plugin/configfrag.ac
index 864817d44d1..d3470f82f8c 100644
--- a/libgomp/plugin/configfrag.ac
+++ b/libgomp/plugin/configfrag.ac
@@ -148,39 +148,56 @@ if test x"$enable_offload_targets" != x; then
tgt_name=intelmic
;;
nvptx*)
- tgt_name=nvptx
- PLUGIN_NVPTX=$tgt
- PLUGIN_NVPTX_CPPFLAGS=$CUDA_DRIVER_CPPFLAGS
- PLUGIN_NVPTX_LDFLAGS=$CUDA_DRIVER_LDFLAGS
- PLUGIN_NVPTX_LIBS='-lcuda'
+ case "${target}" in
+ aarch64*-*-* | powerpc64le-*-* | x86_64-*-*)
+ case " ${CC} ${CFLAGS} " in
+ *" -m32 "* | *" -mx32 "*)
+ # PR libgomp/65099: Currently, we only support offloading in
+ # 64-bit configurations.
+ PLUGIN_NVPTX=0
+ ;;
+ *)
+ tgt_name=nvptx
+ PLUGIN_NVPTX=$tgt
+ PLUGIN_NVPTX_CPPFLAGS=$CUDA_DRIVER_CPPFLAGS
+ PLUGIN_NVPTX_LDFLAGS=$CUDA_DRIVER_LDFLAGS
+ PLUGIN_NVPTX_LIBS='-lcuda'
- PLUGIN_NVPTX_save_CPPFLAGS=$CPPFLAGS
- CPPFLAGS="$PLUGIN_NVPTX_CPPFLAGS $CPPFLAGS"
- PLUGIN_NVPTX_save_LDFLAGS=$LDFLAGS
- LDFLAGS="$PLUGIN_NVPTX_LDFLAGS $LDFLAGS"
- PLUGIN_NVPTX_save_LIBS=$LIBS
- LIBS="$PLUGIN_NVPTX_LIBS $LIBS"
- AC_LINK_IFELSE(
- [AC_LANG_PROGRAM(
- [#include "cuda.h"],
- [CUresult r = cuCtxPushCurrent (NULL);])],
- [PLUGIN_NVPTX=1])
- CPPFLAGS=$PLUGIN_NVPTX_save_CPPFLAGS
- LDFLAGS=$PLUGIN_NVPTX_save_LDFLAGS
- LIBS=$PLUGIN_NVPTX_save_LIBS
- case $PLUGIN_NVPTX in
- nvptx*)
- if test "x$CUDA_DRIVER_INCLUDE" = x \
- && test "x$CUDA_DRIVER_LIB" = x; then
- PLUGIN_NVPTX=1
- PLUGIN_NVPTX_CPPFLAGS='-I$(srcdir)/plugin/cuda'
- PLUGIN_NVPTX_LIBS='-ldl'
- PLUGIN_NVPTX_DYNAMIC=1
- else
- PLUGIN_NVPTX=0
- AC_MSG_ERROR([CUDA driver package required for nvptx support])
- fi
- ;;
+ PLUGIN_NVPTX_save_CPPFLAGS=$CPPFLAGS
+ CPPFLAGS="$PLUGIN_NVPTX_CPPFLAGS $CPPFLAGS"
+ PLUGIN_NVPTX_save_LDFLAGS=$LDFLAGS
+ LDFLAGS="$PLUGIN_NVPTX_LDFLAGS $LDFLAGS"
+ PLUGIN_NVPTX_save_LIBS=$LIBS
+ LIBS="$PLUGIN_NVPTX_LIBS $LIBS"
+ AC_LINK_IFELSE(
+ [AC_LANG_PROGRAM(
+ [#include "cuda.h"],
+ [CUresult r = cuCtxPushCurrent (NULL);])],
+ [PLUGIN_NVPTX=1])
+ CPPFLAGS=$PLUGIN_NVPTX_save_CPPFLAGS
+ LDFLAGS=$PLUGIN_NVPTX_save_LDFLAGS
+ LIBS=$PLUGIN_NVPTX_save_LIBS
+ case $PLUGIN_NVPTX in
+ nvptx*)
+ if test "x$CUDA_DRIVER_INCLUDE" = x \
+ && test "x$CUDA_DRIVER_LIB" = x; then
+ PLUGIN_NVPTX=1
+ PLUGIN_NVPTX_CPPFLAGS='-I$(srcdir)/plugin/cuda'
+ PLUGIN_NVPTX_LIBS='-ldl'
+ PLUGIN_NVPTX_DYNAMIC=1
+ else
+ PLUGIN_NVPTX=0
+ AC_MSG_ERROR([CUDA driver package required for nvptx support])
+ fi
+ ;;
+ esac
+ ;;
+ esac
+ ;;
+ *-*-*)
+ # Target architecture not supported.
+ PLUGIN_NVPTX=0
+ ;;
esac
;;
hsa*)
diff --git a/libgomp/plugin/plugin-nvptx.c b/libgomp/plugin/plugin-nvptx.c
index 9ae60953a9a..0a4f4f410bb 100644
--- a/libgomp/plugin/plugin-nvptx.c
+++ b/libgomp/plugin/plugin-nvptx.c
@@ -835,15 +835,6 @@ nvptx_get_num_devices (void)
{
int n;
- /* PR libgomp/65099: Currently, we only support offloading in 64-bit
- configurations. */
- if (sizeof (void *) != 8)
- {
- GOMP_PLUGIN_debug (0, "Disabling nvptx offloading;"
- " only 64-bit configurations are supported\n");
- return 0;
- }
-
/* This function will be called before the plugin has been initialized in
order to enumerate available devices, but CUDA API routines can't be used
until cuInit has been called. Just call it now (but don't yet do any
--
2.17.1
