diff mbox series

nvptx, libgfortran: Switch out of "minimal" mode

Message ID 87r0damfjn.fsf@euler.schwinge.ddns.net
State New
Headers show
Series nvptx, libgfortran: Switch out of "minimal" mode | expand

Commit Message

Thomas Schwinge June 6, 2024, 12:17 p.m. UTC
Hi!

On 2023-01-20T22:16:00+0100, Hi wrote:
> On 2023-01-20T22:04:02+0100, I wrote:
>> We've been (t)asked to enable (portions of) GCC/Fortran I/O for nvptx
>> offloading, which means building a normal (non-'LIBGFOR_MINIMAL')
>> configuration of libgfortran.
>
> This is achieved by 'nvptx, libgfortran: Switch out of "minimal" mode',
> see attached, again based on WIP work by Andrew Stubbs.

I've recently slightly revised this, in particular:

> The OpenACC XFAILs: "[...] overflows the stack [...]"

... I now avoid by use of commit 0d25989d60d15866ef4737d66e02432f50717255
"nvptx offloading: 'GOMP_NVPTX_NATIVE_GPU_THREAD_STACK_SIZE' environment variable [PR97384, PR105274]".

The underlying issue remains...

> [...] unresolved at this point; see the discussion around
> "Handling of large stack objects in GPU code generation -- maybe transform into heap allocation?",
> and my "nvptx: '-mframe-malloc-threshold', '-Wframe-malloc-threshold'"
> experimenting.  (The latter works to some extent, but also has other
> issues that I shall detail at some later point in time.)

(No progress.)


Pushed to trunk branch commit 3a4775d4403f2e88b589e88a9937cc1fd45a0e87
'nvptx, libgfortran: Switch out of "minimal" mode', see attached.

This, unsurprisingly, also greatly improves GCC/Fortran test results for
nvptx target.


Grüße
 Thomas
diff mbox series

Patch

From 3a4775d4403f2e88b589e88a9937cc1fd45a0e87 Mon Sep 17 00:00:00 2001
From: Thomas Schwinge <tschwinge@baylibre.com>
Date: Wed, 5 Jun 2024 13:13:24 +0200
Subject: [PATCH] nvptx, libgfortran: Switch out of "minimal" mode

..., in order to enable (portions of) Fortran I/O, for example.

	libgfortran/
	* configure.ac: No longer set 'LIBGFOR_MINIMAL' for nvptx.
	* configure: Regenerate.
	libgomp/
	* libgomp.texi (nvptx): Update.
	* testsuite/libgomp.fortran/target-print-1-nvptx.f90: Remove.
	* testsuite/libgomp.fortran/target-print-1.f90: Adjust.
	* testsuite/libgomp.oacc-fortran/error_stop-2-nvptx.f: New.
	* testsuite/libgomp.oacc-fortran/error_stop-2.f: Adjust.
	* testsuite/libgomp.oacc-fortran/print-1-nvptx.f90: Adjust.
	* testsuite/libgomp.oacc-fortran/print-1.f90: Adjust.
	* testsuite/libgomp.oacc-fortran/stop-2-nvptx.f: New.
	* testsuite/libgomp.oacc-fortran/stop-2.f: Adjust.

Co-authored-by: Andrew Stubbs <ams@gcc.gnu.org>
---
 libgfortran/configure                         | 21 ++++------
 libgfortran/configure.ac                      | 17 +++-----
 libgomp/libgomp.texi                          | 10 +++--
 .../libgomp.fortran/target-print-1-nvptx.f90  | 11 -----
 .../libgomp.fortran/target-print-1.f90        |  3 --
 .../libgomp.oacc-fortran/error_stop-2-nvptx.f | 39 ++++++++++++++++++
 .../libgomp.oacc-fortran/error_stop-2.f       |  3 +-
 .../libgomp.oacc-fortran/print-1-nvptx.f90    | 40 +++++++++++++++----
 .../libgomp.oacc-fortran/print-1.f90          |  4 +-
 .../libgomp.oacc-fortran/stop-2-nvptx.f       | 36 +++++++++++++++++
 .../testsuite/libgomp.oacc-fortran/stop-2.f   |  3 +-
 11 files changed, 134 insertions(+), 53 deletions(-)
 delete mode 100644 libgomp/testsuite/libgomp.fortran/target-print-1-nvptx.f90
 create mode 100644 libgomp/testsuite/libgomp.oacc-fortran/error_stop-2-nvptx.f
 create mode 100644 libgomp/testsuite/libgomp.oacc-fortran/stop-2-nvptx.f

diff --git a/libgfortran/configure b/libgfortran/configure
index 774dd52fc95..11a1bc5f070 100755
--- a/libgfortran/configure
+++ b/libgfortran/configure
@@ -6207,17 +6207,12 @@  else
 fi
 
 
-# For GPU offloading, not everything in libfortran can be supported.
-# Currently, the only target that has this problem is nvptx.  The
-# following is a (partial) list of features that are unsupportable on
-# this particular target:
-# * Constructors
-# * alloca
-# * C library support for I/O, with printf as the one notable exception
-# * C library support for other features such as signal, environment
-#   variables, time functions
-
- if test "x${target_cpu}" = xnvptx; then
+# "Minimal" mode is for targets that cannot (yet) support all features of
+# libgfortran.  It avoids the need for working constructors, alloca, and C
+# library support for I/O, signals, environment variables, time functions, etc.
+# At present there are no targets that require this mode.
+
+ if false; then
   LIBGFOR_MINIMAL_TRUE=
   LIBGFOR_MINIMAL_FALSE='#'
 else
@@ -12852,7 +12847,7 @@  else
   lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<_LT_EOF
-#line 12855 "configure"
+#line 12850 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -12958,7 +12953,7 @@  else
   lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<_LT_EOF
-#line 12961 "configure"
+#line 12956 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
diff --git a/libgfortran/configure.ac b/libgfortran/configure.ac
index 46585a3ee14..cca1ea0ea97 100644
--- a/libgfortran/configure.ac
+++ b/libgfortran/configure.ac
@@ -209,17 +209,12 @@  AM_CONDITIONAL(LIBGFOR_USE_SYMVER, [test "x$gfortran_use_symver" != xno])
 AM_CONDITIONAL(LIBGFOR_USE_SYMVER_GNU, [test "x$gfortran_use_symver" = xgnu])
 AM_CONDITIONAL(LIBGFOR_USE_SYMVER_SUN, [test "x$gfortran_use_symver" = xsun])
 
-# For GPU offloading, not everything in libfortran can be supported.
-# Currently, the only target that has this problem is nvptx.  The
-# following is a (partial) list of features that are unsupportable on
-# this particular target:
-# * Constructors
-# * alloca
-# * C library support for I/O, with printf as the one notable exception
-# * C library support for other features such as signal, environment
-#   variables, time functions
-
-AM_CONDITIONAL(LIBGFOR_MINIMAL, [test "x${target_cpu}" = xnvptx])
+# "Minimal" mode is for targets that cannot (yet) support all features of
+# libgfortran.  It avoids the need for working constructors, alloca, and C
+# library support for I/O, signals, environment variables, time functions, etc.
+# At present there are no targets that require this mode.
+
+AM_CONDITIONAL(LIBGFOR_MINIMAL, false)
 
 # Some compiler target support may have limited support for integer
 # or floating point numbers – or may want to reduce the libgfortran size
diff --git a/libgomp/libgomp.texi b/libgomp/libgomp.texi
index c52bb2672c6..eb608915938 100644
--- a/libgomp/libgomp.texi
+++ b/libgomp/libgomp.texi
@@ -6354,7 +6354,7 @@  The used sizes are
 
 The implementation remark:
 @itemize
-@item I/O within OpenMP target regions and OpenACC parallel/kernels is supported
+@item I/O within OpenMP target regions and OpenACC compute regions is supported
       using the C library @code{printf} functions and the Fortran
       @code{print}/@code{write} statements.
 @item Reverse offload regions (i.e. @code{target} regions with
@@ -6429,9 +6429,11 @@  CUDA version and hardware.
 
 The implementation remark:
 @itemize
-@item I/O within OpenMP target regions and OpenACC parallel/kernels is supported
-      using the C library @code{printf} functions. Note that the Fortran
-      @code{print}/@code{write} statements are not supported, yet.
+@item I/O within OpenMP target regions and OpenACC compute regions is supported
+      using the C library @code{printf} functions.
+      Additionally, the Fortran @code{print}/@code{write} statements are
+      supported within OpenMP target regions, but not yet OpenACC compute
+      regions.  @c The latter needs 'GOMP_NVPTX_NATIVE_GPU_THREAD_STACK_SIZE'.
 @item Compilation OpenMP code that contains @code{requires reverse_offload}
       requires at least @code{-march=sm_35}, compiling for @code{-march=sm_30}
       is not supported.
diff --git a/libgomp/testsuite/libgomp.fortran/target-print-1-nvptx.f90 b/libgomp/testsuite/libgomp.fortran/target-print-1-nvptx.f90
deleted file mode 100644
index a89c9c33484..00000000000
--- a/libgomp/testsuite/libgomp.fortran/target-print-1-nvptx.f90
+++ /dev/null
@@ -1,11 +0,0 @@ 
-! Ensure that write on the offload device works, nvptx offloading variant.
-
-! This doesn't compile: for nvptx offloading we're using a minimal libgfortran
-! configuration.
-! { dg-do link } ! ..., but still apply 'dg-do run' options.
-! { dg-xfail-if "minimal libgfortran" { offload_target_nvptx } }
-
-! Skip duplicated testing.
-! { dg-skip-if "separate file" { ! offload_target_nvptx } }
-
-include 'target-print-1.f90'
diff --git a/libgomp/testsuite/libgomp.fortran/target-print-1.f90 b/libgomp/testsuite/libgomp.fortran/target-print-1.f90
index 327bb22cb6d..9ac70e5a85f 100644
--- a/libgomp/testsuite/libgomp.fortran/target-print-1.f90
+++ b/libgomp/testsuite/libgomp.fortran/target-print-1.f90
@@ -3,9 +3,6 @@ 
 ! { dg-do run }
 ! { dg-output "The answer is 42(\n|\r\n|\r)+" }
 
-! Separate file 'target-print-1-nvptx.f90' for nvptx offloading.
-! { dg-skip-if "separate file" { offload_target_nvptx } }
-
 program main
   implicit none
   integer :: var = 42
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/error_stop-2-nvptx.f b/libgomp/testsuite/libgomp.oacc-fortran/error_stop-2-nvptx.f
new file mode 100644
index 00000000000..ef93e80d9d4
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/error_stop-2-nvptx.f
@@ -0,0 +1,39 @@ 
+! 'error_stop-2.f' nvptx offloading variant
+
+! { dg-do run { target openacc_nvidia_accel_selected } }
+!
+! The PTX JIT doesn't understand the 'ERROR STOP' call graph, and therefore:
+!     warning : Stack size for entry function 'main$_omp_fn$0' cannot be statically determined
+! Running with default 1024-bytes GPU thread stack size overflows the stack,
+! so raise it to an arbitrarily higher value:
+! { dg-set-target-env-var GOMP_NVPTX_NATIVE_GPU_THREAD_STACK_SIZE 3333 }
+
+      PROGRAM MAIN
+      IMPLICIT NONE
+
+! Initialize before the checkpoint, in case this produces any output.
+!$ACC PARALLEL
+!$ACC END PARALLEL
+
+      PRINT *, "CheCKpOInT"
+!$ACC PARALLEL
+      ERROR STOP 35
+!$ACC END PARALLEL
+      PRINT *, "WrONg WAy"
+
+      END PROGRAM MAIN
+
+! { dg-output "CheCKpOInT(\n|\r\n|\r)+" }
+
+! { dg-output "ERROR STOP 35(\n|\r\n|\r)+" }
+!
+! In gfortran's main program, libfortran's set_options is called - which sets
+! compiler_options.backtrace = 1 by default.  For an offload libgfortran, this
+! is never called and, hence, "Error termination." is never printed.  Thus:
+! { dg-output "Error termination.*" { target { ! { openacc_nvidia_accel_selected || openacc_radeon_accel_selected } } } }
+
+! PR85463.  The 'exit' implementation used with nvptx
+! offloading is a little bit different.
+! { dg-output "libgomp: cuStreamSynchronize error.*" { target openacc_nvidia_accel_selected } }
+
+! { dg-shouldfail "" }
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/error_stop-2.f b/libgomp/testsuite/libgomp.oacc-fortran/error_stop-2.f
index 5951e8cbe64..c0762835faa 100644
--- a/libgomp/testsuite/libgomp.oacc-fortran/error_stop-2.f
+++ b/libgomp/testsuite/libgomp.oacc-fortran/error_stop-2.f
@@ -17,7 +17,8 @@ 
 
 ! { dg-output "CheCKpOInT(\n|\r\n|\r)+" }
 
-! { dg-output "ERROR STOP 35(\n|\r\n|\r)+" }
+! See 'error_stop-2-nvptx.f' regarding the nvptx offloading XFAIL.
+! { dg-output "ERROR STOP 35(\n|\r\n|\r)+" { xfail openacc_nvidia_accel_selected } }
 !
 ! In gfortran's main program, libfortran's set_options is called - which sets
 ! compiler_options.backtrace = 1 by default.  For an offload libgfortran, this
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/print-1-nvptx.f90 b/libgomp/testsuite/libgomp.oacc-fortran/print-1-nvptx.f90
index 866c8654355..e8080e6c8ac 100644
--- a/libgomp/testsuite/libgomp.oacc-fortran/print-1-nvptx.f90
+++ b/libgomp/testsuite/libgomp.oacc-fortran/print-1-nvptx.f90
@@ -1,11 +1,37 @@ 
 ! Ensure that write on the offload device works, nvptx offloading variant.
 
-! This doesn't compile: for nvptx offloading we're using a minimal libgfortran
-! configuration.
-! { dg-do link } ! ..., but still apply 'dg-do run' options.
-! { dg-xfail-if "minimal libgfortran" { offload_target_nvptx } }
+! { dg-do run { target openacc_nvidia_accel_selected } }
+! { dg-output "The answer is 42(\n|\r\n|\r)+" }
+!
+! The PTX JIT doesn't understand the 'write' call graph, and therefore:
+!     warning : Stack size for entry function 'main$_omp_fn$0' cannot be statically determined
+! Running with default 1024-bytes GPU thread stack size overflows the stack,
+! so raise it to an arbitrarily higher value:
+! { dg-set-target-env-var GOMP_NVPTX_NATIVE_GPU_THREAD_STACK_SIZE 3333 }
 
-! Skip duplicated testing.
-! { dg-skip-if "separate file" { ! offload_target_nvptx } }
+! { dg-additional-options "-fopt-info-note-omp" }
+! { dg-additional-options "-foffload=-fopt-info-note-omp" }
 
-include 'print-1.f90'
+! { dg-additional-options "--param=openacc-privatization=noisy" }
+! { dg-additional-options "-foffload=--param=openacc-privatization=noisy" }
+! Prune a few: uninteresting, and potentially varying depending on GCC configuration (data types):
+! { dg-prune-output {note: variable 'D\.[0-9]+' declared in block isn't candidate for adjusting OpenACC privatization level: not addressable} } */
+
+! It's only with Tcl 8.5 (released in 2007) that "the variable 'varName'
+! passed to 'incr' may be unset, and in that case, it will be set to [...]",
+! so to maintain compatibility with earlier Tcl releases, we manually
+! initialize counter variables:
+! { dg-line l_dummy[variable c_compute 0] }
+! { dg-message dummy {} { target iN-VAl-Id } l_dummy } to avoid
+! "WARNING: dg-line var l_dummy defined, but not used".
+
+program main
+  implicit none
+  integer :: var = 42
+
+!$acc parallel ! { dg-line l_compute[incr c_compute] }
+  ! { dg-note {variable 'dt_parm\.[0-9]+' declared in block isn't candidate for adjusting OpenACC privatization level: artificial} {} { target *-*-* } l_compute$c_compute }
+  write (0, '("The answer is ", I2)') var
+!$acc end parallel
+
+end program main
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/print-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/print-1.f90
index d2f89d915f8..00f5c634a20 100644
--- a/libgomp/testsuite/libgomp.oacc-fortran/print-1.f90
+++ b/libgomp/testsuite/libgomp.oacc-fortran/print-1.f90
@@ -3,8 +3,8 @@ 
 ! { dg-do run }
 ! { dg-output "The answer is 42(\n|\r\n|\r)+" }
 
-! Separate file 'print-1-nvptx.f90' for nvptx offloading.
-! { dg-skip-if "separate file" { offload_target_nvptx } }
+! See 'print-1-nvptx.f90' regarding the nvptx offloading XFAIL.
+! { dg-xfail-run-if TODO { openacc_nvidia_accel_selected } }
 
 ! { dg-additional-options "-fopt-info-note-omp" }
 ! { dg-additional-options "-foffload=-fopt-info-note-omp" }
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/stop-2-nvptx.f b/libgomp/testsuite/libgomp.oacc-fortran/stop-2-nvptx.f
new file mode 100644
index 00000000000..3c8b45c102f
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/stop-2-nvptx.f
@@ -0,0 +1,36 @@ 
+! 'stop-2.f' nvptx offloading variant
+
+! { dg-do run { target openacc_nvidia_accel_selected } }
+!
+! The PTX JIT doesn't understand the 'STOP' call graph, and therefore:
+!     warning : Stack size for entry function 'main$_omp_fn$0' cannot be statically determined
+! Running with default 1024-bytes GPU thread stack size overflows the stack,
+! so raise it to an arbitrarily higher value:
+! { dg-set-target-env-var GOMP_NVPTX_NATIVE_GPU_THREAD_STACK_SIZE 3333 }
+
+      PROGRAM MAIN
+      IMPLICIT NONE
+
+! Initialize before the checkpoint, in case this produces any output.
+!$ACC PARALLEL
+!$ACC END PARALLEL
+
+      PRINT *, "CheCKpOInT"
+!$ACC PARALLEL
+      STOP 35
+!$ACC END PARALLEL
+      PRINT *, "WrONg WAy"
+
+      END PROGRAM MAIN
+
+! { dg-output "CheCKpOInT(\n|\r\n|\r)+" }
+
+! { dg-output "STOP 35(\n|\r\n|\r)+" }
+!
+! PR85463.  The 'exit' implementation used with nvptx
+! offloading is a little bit different.
+! { dg-output "libgomp: cuStreamSynchronize error.*" { target openacc_nvidia_accel_selected } }
+
+! { dg-output "$" }
+
+! { dg-shouldfail "" }
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/stop-2.f b/libgomp/testsuite/libgomp.oacc-fortran/stop-2.f
index fe7ee37813a..e08297eda5d 100644
--- a/libgomp/testsuite/libgomp.oacc-fortran/stop-2.f
+++ b/libgomp/testsuite/libgomp.oacc-fortran/stop-2.f
@@ -17,7 +17,8 @@ 
 
 ! { dg-output "CheCKpOInT(\n|\r\n|\r)+" }
 
-! { dg-output "STOP 35(\n|\r\n|\r)+" }
+! See 'stop-2-nvptx.f' regarding the nvptx offloading XFAIL.
+! { dg-output "STOP 35(\n|\r\n|\r)+" { xfail openacc_nvidia_accel_selected } }
 !
 ! PR85463.  The 'exit' implementation used with nvptx
 ! offloading is a little bit different.
-- 
2.34.1