Merge pull request #7586 from raffenet/4.3.x-fixes

[4.3.x] misc fixes for 4.3.2rc1

Approved-by: Hui Zhou <[email protected]>

Author: raffenet

CHANGES

@@ -10,10 +10,19 @@
 
 # Update libfabric usage to silence deprecation warnings
 
+# Update yaksa for improved reproducibility in code generation
+
 # Update embedded UCX to v1.19.0
 
+# Update embedded libfabric to fix build issue with GCC 15
+
 # Add compatibility with CUDA 13
 
+# Fix missing const in nondestructive request test and status query
+  functions
+
+# Fix HCOLL support
+
 # Fix crash with GPU-aware build when running on systems with no GPUs
 
 # Fix singleton init with Hydra

maint/release.pl

@@ -259,7 +259,7 @@ sub run_cmd
 print("===> Removing unnecessary files in the main codebase... ");
 chdir($expdir);
 run_cmd("rm -rf README.vin maint/config.log maint/config.status unusederr.txt");
-run_cmd("find . -name autom4te.cache | xargs rm -rf");
+run_cmd("find . -name autom4te.cache -o -name __pycache__ | xargs rm -rf");
 print("done\n");
 
 # Get docs

modules/libfabric

@@ -1805,18 +1805,18 @@ MPI_Request_get_status:
     status: STATUS, direction=out, [status object if flag is true]
 MPI_Request_get_status_all:
     count: ARRAY_LENGTH_NNI, lists length
-    array_of_requests: REQUEST, direction=in, length=count, [array of requests]
+    array_of_requests: REQUEST, constant=True, direction=in, length=count, [array of requests]
     flag: LOGICAL, direction=out
     array_of_statuses: STATUS, direction=out, length=*, pointer=False, [array of status objects]
 MPI_Request_get_status_any:
     count: ARRAY_LENGTH_NNI, list length
-    array_of_requests: REQUEST, direction=in, length=count, [array of requests]
+    array_of_requests: REQUEST, constant=True, direction=in, length=count, [array of requests]
     index: INDEX, direction=out, [index of operation that completed or MPI_UNDEFINED if none completed]
     flag: LOGICAL, direction=out, [true if one of the operations is complete]
     status: STATUS, direction=out
 MPI_Request_get_status_some:
     incount: ARRAY_LENGTH_NNI, [length of array_of_requests]
-    array_of_requests: REQUEST, direction=in, length=incount, [array of requests]
+    array_of_requests: REQUEST, constant=True, direction=in, length=incount, [array of requests]
     outcount: ARRAY_LENGTH, direction=out, [number of completed requests]
     array_of_indices: INDEX, direction=out, length=*, [array of indices of operations that completed]
     array_of_statuses: STATUS, direction=out, length=*, pointer=False, [array of status objects for operations that completed]
@@ -2070,13 +2070,13 @@ MPI_Status_f2f08:
     f_status: F90_STATUS, pointer=True, constant=True, [status object declared as array]
     f08_status: F08_STATUS, direction=out, [status object declared as named type]
 MPI_Status_get_error:
-    status: STATUS, direction=in, [status from which to retrieve source rank]
+    status: STATUS, constant=True, direction=in, [status from which to retrieve source rank]
     error: ERROR_CODE, direction=out, [error set in the MPI_ERROR field]
 MPI_Status_get_source:
-    status: STATUS, direction=in, [status from which to retrieve source rank]
+    status: STATUS, constant=True, direction=in, [status from which to retrieve source rank]
     source: RANK, direction=out, [rank set in the MPI_SOURCE field]
 MPI_Status_get_tag:
-    status: STATUS, direction=in, [status from which to retrieve source rank]
+    status: STATUS, constant=True, direction=in, [status from which to retrieve source rank]
     tag: TAG, direction=out, [tag set in the MPI_TAG field]
 MPI_Status_set_error:
     status: STATUS, direction=inout, [status with which to associate error]

modules/yaksa

@@ -15,7 +15,9 @@ AM_CPPFLAGS = -I$(top_builddir)/include -I$(top_srcdir)/include
 
 # This is right for many platforms, but not all.  The right fix involves a
 # configure test, but this version is no worse than the simplemake version was.
-AM_FFLAGS = -I.
+# Add -w to suppress all warnings. Compilers such as gfortrans warns even with
+# -fallow-argument-mismatch.
+AM_FFLAGS = -I. -w
 
 LDADD = $(top_builddir)/util/libmtest_f77.la
 mtest_c_objects = $(top_builddir)/util/libmtest_single.la

src/binding/mpi_standard_api.txt

@@ -63,12 +63,12 @@ check_copy() {
 }
 
 generate_benchmarks() {
-    MYDEF_BOOT=$PWD/../../modules/mydef_boot
-    if test -d $MYDEF_BOOT ; then
+    MYDEF_BOOT="$PWD/../../modules/mydef_boot"
+    if test -d "$MYDEF_BOOT/bin" ; then
         echo "Generating benchmark tests"
-        export PATH=$MYDEF_BOOT/bin:$PATH
-        export PERL5LIB=$MYDEF_BOOT/lib/perl5
-        export MYDEFLIB=$MYDEF_BOOT/lib/MyDef
+        export PATH="$MYDEF_BOOT/bin:$PATH"
+        export PERL5LIB="$MYDEF_BOOT/lib/perl5"
+        export MYDEFLIB="$MYDEF_BOOT/lib/MyDef"
         (cd bench && ./autogen.sh)
     fi
 }

test/mpi/Makefile_f77.mtest

@@ -67,7 +67,7 @@ int main(int argc, char **argv)
         if (rank == 0) {
             printf("nprocs=%d NVARS=%d LEN=%d\n", nprocs, NVARS, LEN);
             printf("Expecting file size=%lld bytes (%.1f MB, %.1f GB)\n",
-                   fsize, (float) fsize / 1048576, (float) fsize / 1073741824);
+                   (long long) fsize, (float) fsize / 1048576, (float) fsize / 1073741824);
             printf("Each global variable is of size %d bytes (%.1f MB)\n",
                    gsize[0] * gsize[1], (float) gsize[0] * gsize[1] / 1048576);
             printf("Each process writes %zd bytes (%.1f MB, %.1f GB)\n",

test/mpi/autogen.sh

@@ -156,11 +156,12 @@ int main(int argc, char **argv)
             printf("Output file name = %s\n", filename);
             printf("nprocs=%d nvars=%d len=%d\n", nprocs, nvars, len);
             printf("Expecting file size=%lld bytes (%.1f MB, %.1f GB)\n",
-                   fsize * 2, (float) fsize * 2 / 1048576, (float) fsize * 2 / 1073741824);
-            printf("Each global variable is of size %d bytes (%.1f MB)\n",
-                   gsize[0] * gsize[1], (float) gsize[0] * gsize[1] / 1048576);
-            printf("Each process writes %zd bytes (%.1f MB, %.1f GB)\n",
-                   buf_len, (float) buf_len / 1048576, (float) buf_len / 1073741824);
+                   (long long) fsize * 2, (float) fsize * 2 / 1048576,
+                   (float) fsize * 2 / 1073741824);
+            printf("Each global variable is of size %d bytes (%.1f MB)\n", gsize[0] * gsize[1],
+                   (float) gsize[0] * gsize[1] / 1048576);
+            printf("Each process writes %zd bytes (%.1f MB, %.1f GB)\n", buf_len,
+                   (float) buf_len / 1048576, (float) buf_len / 1073741824);
             printf("** For nonblocking I/O test, the amount is twice\n");
             printf("-------------------------------------------------------\n");
         }