From 72a171242c4efb37f5207624989729b235c3cce1 Mon Sep 17 00:00:00 2001
From: "Peter W. Draper" <p.w.draper@durham.ac.uk>
Date: Wed, 11 Mar 2020 12:57:43 +0000
Subject: [PATCH] Also report the tag and source of the failed receive
---
swiftmpiproxies.c | 24 +++++++++++++++---------
1 file changed, 15 insertions(+), 9 deletions(-)
diff --git a/swiftmpiproxies.c b/swiftmpiproxies.c
index 9d48671..5dcecdf 100644
--- a/swiftmpiproxies.c
+++ b/swiftmpiproxies.c
@@ -267,8 +267,8 @@ int main(int argc, char *argv[]) {
int size = log->size;
res = MPI_Isend(&size, 1, MPI_INT, log->otherrank,
- basetag + proxy_tag_count,
- MPI_COMM_WORLD, &req_send_counts[k]);
+ basetag + proxy_tag_count, MPI_COMM_WORLD,
+ &req_send_counts[k]);
if (res != MPI_SUCCESS) error("Counts MPI_Isend failed.");
/* Start Isend of pcells, filling the data with a pattern for checking
@@ -313,8 +313,8 @@ int main(int argc, char *argv[]) {
datacheck_fill(1, pcells_size[pid], pcells_in[pid]);
res = MPI_Irecv(pcells_in[pid], pcells_size[pid], MPI_BYTE,
- log->otherrank, basetag + proxy_tag_cells,
- MPI_COMM_WORLD, &req_pcells_in[pid]);
+ log->otherrank, basetag + proxy_tag_cells, MPI_COMM_WORLD,
+ &req_pcells_in[pid]);
if (res != MPI_SUCCESS) error("Pcell MPI_Irecv failed.");
}
message("All proxy cell counts have arrived, pcells irecvs are launched");
@@ -338,6 +338,14 @@ int main(int argc, char *argv[]) {
/* Check the data received is correct. */
if (!datacheck_test(0, pcells_size[pid], pcells_in[pid])) {
message("Received data is not correct");
+
+ /* Report the tag and source of the request. */
+ int expected_tag =
+ status.MPI_SOURCE * proxy_tag_shift + proxy_tag_cells;
+ message("sent from rank %d, with tag %d/%d and error code %d",
+ status.MPI_SOURCE, status.MPI_TAG, expected_tag,
+ status.MPI_ERROR);
+
datacheck_fulltest(pcells_size[pid], pcells_in[pid]);
if (datacheck_test(1, pcells_size[pid], pcells_in[pid])) {
error("Received data is not modified on receive");
@@ -345,8 +353,7 @@ int main(int argc, char *argv[]) {
error("Received data is corrupt");
}
} else {
- if (verbose)
- message("Received data is correct");
+ if (verbose) message("Received data is correct");
}
free(pcells_in[pid]);
pcells_in[pid] = NULL;
@@ -365,14 +372,13 @@ int main(int argc, char *argv[]) {
datacheck_fulltest(log->size, log->data);
error("Sent data has been corrupted");
} else {
- if (verbose)
- message("Sent data is correct");
+ if (verbose) message("Sent data is correct");
}
free(log->data);
log->data = NULL;
}
- }/* nloop */
+ } /* nloop */
/* Shutdown MPI. */
res = MPI_Finalize();
--
GitLab