From 190689fd0106774731c5053c2aa5455039eec4d4 Mon Sep 17 00:00:00 2001
From: Mathieu Faverge <mathieu.faverge@inria.fr>
Date: Fri, 2 Oct 2015 23:37:29 +0000
Subject: [PATCH] At the user levl, we wait for a sequence. We need to add a
MORSE_Barrier function that waits for all sequences currently running
---
example/lapack_to_morse/step2.c | 4 ++--
example/lapack_to_morse/step3.c | 4 ++--
example/lapack_to_morse/step4.c | 7 ++++---
example/lapack_to_morse/step5.c | 7 ++++---
example/lapack_to_morse/step6.c | 7 ++++---
5 files changed, 16 insertions(+), 13 deletions(-)
diff --git a/example/lapack_to_morse/step2.c b/example/lapack_to_morse/step2.c
index 87fca9420..d3cb31436 100644
--- a/example/lapack_to_morse/step2.c
+++ b/example/lapack_to_morse/step2.c
@@ -38,8 +38,8 @@ int main(int argc, char *argv[]) {
size_t i, j;
size_t N; // matrix order
- size_t NB; // number of rows and columns in tiles
- size_t NRHS; // number of RHS vectors
+ int NB; // number of rows and columns in tiles
+ int NRHS; // number of RHS vectors
int NCPU; // number of cores to use
int NGPU; // number of gpus (cuda devices) to use
int UPLO = MorseUpper; // where is stored L
diff --git a/example/lapack_to_morse/step3.c b/example/lapack_to_morse/step3.c
index 5330666ed..3a9f5ae7e 100644
--- a/example/lapack_to_morse/step3.c
+++ b/example/lapack_to_morse/step3.c
@@ -36,8 +36,8 @@ int main(int argc, char *argv[]) {
size_t i, j;
size_t N; // matrix order
- size_t NB; // number of rows and columns in tiles
- size_t NRHS; // number of RHS vectors
+ int NB; // number of rows and columns in tiles
+ int NRHS; // number of RHS vectors
int NCPU; // number of cores to use
int NGPU; // number of gpus (cuda devices) to use
int UPLO = MorseUpper; // where is stored L
diff --git a/example/lapack_to_morse/step4.c b/example/lapack_to_morse/step4.c
index 7e585ac10..9a2ec2158 100644
--- a/example/lapack_to_morse/step4.c
+++ b/example/lapack_to_morse/step4.c
@@ -40,8 +40,8 @@ int main(int argc, char *argv[]) {
size_t i, j;
size_t N; // matrix order
- size_t NB; // number of rows and columns in tiles
- size_t NRHS; // number of RHS vectors
+ int NB; // number of rows and columns in tiles
+ int NRHS; // number of RHS vectors
int NCPU; // number of cores to use
int NGPU; // number of gpus (cuda devices) to use
int UPLO = MorseUpper; // where is stored L
@@ -144,7 +144,8 @@ int main(int argc, char *argv[]) {
/* Synchronization barrier (the runtime ensures that all submitted tasks
* have been terminated */
- RUNTIME_barrier(morse);
+ MORSE_Sequence_Wait(sequence);
+
/* Ensure that all data processed on the gpus we are depending on are back
* in main memory */
RUNTIME_desc_getoncpu(descA);
diff --git a/example/lapack_to_morse/step5.c b/example/lapack_to_morse/step5.c
index a49a5ff82..1267532e3 100644
--- a/example/lapack_to_morse/step5.c
+++ b/example/lapack_to_morse/step5.c
@@ -40,8 +40,8 @@ int main(int argc, char *argv[]) {
size_t i, j;
size_t N; // matrix order
- size_t NB; // number of rows and columns in tiles
- size_t NRHS; // number of RHS vectors
+ int NB; // number of rows and columns in tiles
+ int NRHS; // number of RHS vectors
int NCPU; // number of cores to use
int NGPU; // number of gpus (cuda devices) to use
int UPLO = MorseUpper; // where is stored L
@@ -148,7 +148,8 @@ int main(int argc, char *argv[]) {
/* Synchronization barrier (the runtime ensures that all submitted tasks
* have been terminated */
- RUNTIME_barrier(morse);
+ MORSE_Sequence_Wait(sequence);
+
/* Ensure that all data processed on the gpus we are depending on are back
* in main memory */
RUNTIME_desc_getoncpu(descA);
diff --git a/example/lapack_to_morse/step6.c b/example/lapack_to_morse/step6.c
index 70e27d311..ee7cc0775 100644
--- a/example/lapack_to_morse/step6.c
+++ b/example/lapack_to_morse/step6.c
@@ -38,8 +38,8 @@ int main(int argc, char *argv[]) {
size_t i, j;
size_t N; // matrix order
- size_t NB; // number of rows and columns in tiles
- size_t NRHS; // number of RHS vectors
+ int NB; // number of rows and columns in tiles
+ int NRHS; // number of RHS vectors
int NCPU; // number of cores to use
int NGPU; // number of gpus (cuda devices) to use
int GRID_P; // parameter of the 2D block cyclic distribution
@@ -170,7 +170,8 @@ int main(int argc, char *argv[]) {
/* Synchronization barrier (the runtime ensures that all submitted tasks
* have been terminated */
- RUNTIME_barrier(morse);
+ MORSE_Sequence_Wait(sequence);
+
/* Ensure that all data processed on the gpus we are depending on are back
* in main memory */
RUNTIME_desc_getoncpu(descA);
--
GitLab