Prefix environment variables NCPUS, NCUDA, NGORDON, WORKERS_CPUID and WORKERS_GPUID by STARPU_

find . -type f -not -name "*svn*"|xargs sed -i s/"\bNCPUS\b"/STARPU_NCPUS/g
find . -type f -not -name "*svn*"|xargs sed -i s/"\bNCUDA\b"/STARPU_NCUDA/g
find . -type f -not -name "*svn*"|xargs sed -i s/"\bNGORDON\b"/STARPU_NGORDON/g
find . -type f -not -name "*svn*"|xargs sed -i s/"\bWORKERS_CPUID\b"/STARPU_WORKERS_CPUID/g
find . -type f -not -name "*svn*"|xargs sed -i s/"\bWORKERS_GPUID\b"/STARPU_WORKERS_GPUID/g

ChangeLog

@@ -23,7 +23,7 @@ The asynchronous heterogeneous multi-accelerator release
     - Model data transfer overhead
     - One model is created for each accelerator
   * Support for CUDA's driver API is deprecated
-  * The WORKERS_GPUID and WORKERS_CPUID env. variables make it possible to
+  * The STARPU_WORKERS_GPUID and STARPU_WORKERS_CPUID env. variables make it possible to
     specify where to bind the workers
   * Use the hwloc library to detect the actual number of cores
 

doc/starpu.texi

@@ -233,15 +233,15 @@ TODO, explicit configuration (passed to starpu_init) overrides env variables.
 @section Configuring workers
 
 @menu
-* NCPUS     :: Number of CPU workers
-* NCUDA     :: Number of CUDA workers
-* NGORDON   :: Number of SPU workers (Cell)
-* WORKERS_CPUID  :: Bind workers to specific CPUs
-* WORKERS_GPUID  :: Select specific CUDA devices
+* STARPU_NCPUS     :: Number of CPU workers
+* STARPU_NCUDA     :: Number of CUDA workers
+* STARPU_NGORDON   :: Number of SPU workers (Cell)
+* STARPU_WORKERS_CPUID  :: Bind workers to specific CPUs
+* STARPU_WORKERS_GPUID  :: Select specific CUDA devices
 @end menu
 
-@node NCPUS
-@subsection @code{NCPUS} -- Number of CPU workers
+@node STARPU_NCPUS
+@subsection @code{STARPU_NCPUS} -- Number of CPU workers
 @table @asis
 
 @item @emph{Description}:
@@ -249,8 +249,8 @@ TODO
 
 @end table
 
-@node NCUDA
-@subsection @code{NCUDA} -- Number of CUDA workers
+@node STARPU_NCUDA
+@subsection @code{STARPU_NCUDA} -- Number of CUDA workers
 @table @asis
 
 @item @emph{Description}:
@@ -258,8 +258,8 @@ TODO
 
 @end table
 
-@node NGORDON
-@subsection @code{NGORDON} -- Number of SPU workers (Cell)
+@node STARPU_NGORDON
+@subsection @code{STARPU_NGORDON} -- Number of SPU workers (Cell)
 @table @asis
 
 @item @emph{Description}:
@@ -268,8 +268,8 @@ TODO
 @end table
 
 
-@node WORKERS_CPUID
-@subsection @code{WORKERS_CPUID} -- Bind workers to specific CPUs
+@node STARPU_WORKERS_CPUID
+@subsection @code{STARPU_WORKERS_CPUID} -- Bind workers to specific CPUs
 @table @asis
 
 @item @emph{Description}:
@@ -277,8 +277,8 @@ TODO
 
 @end table
 
-@node WORKERS_GPUID
-@subsection @code{WORKERS_GPUID} -- Select specific CUDA devices
+@node STARPU_WORKERS_GPUID
+@subsection @code{STARPU_WORKERS_GPUID} -- Select specific CUDA devices
 @table @asis
 
 @item @emph{Description}:
@@ -422,15 +422,15 @@ This is the name of the scheduling policy. This can also be specified with the
 
 @item @code{ncpus} (default = -1):
 This is the maximum number of CPU cores that StarPU can use. This can also be
-specified with the @code{NCPUS} environment variable.
+specified with the @code{STARPU_NCPUS} environment variable.
 
 @item @code{ncuda} (default = -1):
 This is the maximum number of CUDA devices that StarPU can use. This can also be
-specified with the @code{NCUDA} environment variable.
+specified with the @code{STARPU_NCUDA} environment variable.
 
 @item @code{nspus} (default = -1):
 This is the maximum number of Cell SPUs that StarPU can use. This can also be
-specified with the @code{NGORDON} environment variable.
+specified with the @code{STARPU_NGORDON} environment variable.
 
 @item @code{calibrate} (default = 0):
 If this flag is set, StarPU will calibrate the performance models when

mpi/examples/perf.sh

@@ -42,7 +42,7 @@ echo "*******************************************">> log
 cat log
 cat log >> log.all
 
-NCPUS=0 NCUDA=$ngpus SCHED="dmda" PREFETCH=1 mpirun -machinefile machinefile.${PBS_JOBID} -np $np ./mpi_lu/plu_example_float -p $p -q $q -nblocks $nblocks -size $(($nblocks * $BLOCKSIZE)) -numa > log.out 2> log.err
+STARPU_NCPUS=0 STARPU_NCUDA=$ngpus SCHED="dmda" PREFETCH=1 mpirun -machinefile machinefile.${PBS_JOBID} -np $np ./mpi_lu/plu_example_float -p $p -q $q -nblocks $nblocks -size $(($nblocks * $BLOCKSIZE)) -numa > log.out 2> log.err
 cat log.out > log
 cat log.err >> log
 cat log

src/core/topology.c

@@ -54,7 +54,7 @@ static void _starpu_initialize_workers_gpuid(struct machine_config_s *config)
 	/* conf->workers_bindid indicates the successive cpu identifier that
 	 * should be used to bind the workers. It should be either filled
 	 * according to the user's explicit parameters (from starpu_conf) or
-	 * according to the WORKERS_CPUID env. variable. Otherwise, a
+	 * according to the STARPU_WORKERS_CPUID env. variable. Otherwise, a
 	 * round-robin policy is used to distributed the workers over the
 	 * cpus. */
 
@@ -66,16 +66,16 @@ static void _starpu_initialize_workers_gpuid(struct machine_config_s *config)
 			config->user_conf->workers_gpuid,
 			STARPU_NMAXWORKERS*sizeof(unsigned));
 	}
-	else if ((strval = getenv("WORKERS_GPUID")))
+	else if ((strval = getenv("STARPU_WORKERS_GPUID")))
 	{
-		/* WORKERS_GPUID certainly contains less entries than
+		/* STARPU_WORKERS_GPUID certainly contains less entries than
 		 * STARPU_NMAXWORKERS, so we reuse its entries in a round robin
 		 * fashion: "1 2" is equivalent to "1 2 1 2 1 2 .... 1 2". */
 		unsigned wrap = 0;
 		unsigned number_of_entries = 0;
 
 		char *endptr;
-		/* we use the content of the WORKERS_GPUID env. variable */
+		/* we use the content of the STARPU_WORKERS_GPUID env. variable */
 		for (i = 0; i < STARPU_NMAXWORKERS; i++)
 		{
 			if (!wrap) {
@@ -188,7 +188,7 @@ static int _starpu_init_machine_config(struct machine_config_s *config,
 			explicitval = user_conf->ncuda;
 		}
 		else {
-			explicitval = starpu_get_env_number("NCUDA");
+			explicitval = starpu_get_env_number("STARPU_NCUDA");
 		}
 
 		if (explicitval < 0) {
@@ -227,7 +227,7 @@ static int _starpu_init_machine_config(struct machine_config_s *config,
 		explicitval = user_conf->ncuda;
 	}
 	else {
-		explicitval = starpu_get_env_number("NGORDON");
+		explicitval = starpu_get_env_number("STARPU_NGORDON");
 	}
 
 	if (explicitval < 0) {
@@ -263,7 +263,7 @@ static int _starpu_init_machine_config(struct machine_config_s *config,
 		explicitval = user_conf->ncpus;
 	}
 	else {
-		explicitval = starpu_get_env_number("NCPUS");
+		explicitval = starpu_get_env_number("STARPU_NCPUS");
 	}
 
 	if (explicitval < 0) {
@@ -314,7 +314,7 @@ static void _starpu_initialize_workers_bindid(struct machine_config_s *config)
 	/* conf->workers_bindid indicates the successive cpu identifier that
 	 * should be used to bind the workers. It should be either filled
 	 * according to the user's explicit parameters (from starpu_conf) or
-	 * according to the WORKERS_CPUID env. variable. Otherwise, a
+	 * according to the STARPU_WORKERS_CPUID env. variable. Otherwise, a
 	 * round-robin policy is used to distributed the workers over the
 	 * cpus. */
 
@@ -326,16 +326,16 @@ static void _starpu_initialize_workers_bindid(struct machine_config_s *config)
 			config->user_conf->workers_bindid,
 			STARPU_NMAXWORKERS*sizeof(unsigned));
 	}
-	else if ((strval = getenv("WORKERS_CPUID")))
+	else if ((strval = getenv("STARPU_WORKERS_CPUID")))
 	{
-		/* WORKERS_CPUID certainly contains less entries than
+		/* STARPU_WORKERS_CPUID certainly contains less entries than
 		 * STARPU_NMAXWORKERS, so we reuse its entries in a round robin
 		 * fashion: "1 2" is equivalent to "1 2 1 2 1 2 .... 1 2". */
 		unsigned wrap = 0;
 		unsigned number_of_entries = 0;
 
 		char *endptr;
-		/* we use the content of the WORKERS_GPUID env. variable */
+		/* we use the content of the STARPU_WORKERS_GPUID env. variable */
 		for (i = 0; i < STARPU_NMAXWORKERS; i++)
 		{
 			if (!wrap) {

tests/cholesky/sched-one-gpu.sh

@@ -20,7 +20,7 @@ maxiter=5
 ROOTDIR=../../
 TIMINGDIR=$PWD/timing/
 
-export WORKERS_GPUID="1"
+export STARPU_WORKERS_GPUID="1"
 
 trace_sched()
 {
@@ -57,7 +57,7 @@ trace_sched()
 
 schedlist='dm dm dm dm greedy dm'
 
-export NCUDA=1
+export STARPU_NCUDA=1
 export CALIBRATE=1
 
 mkdir -p $TIMINGDIR

tests/coverage/coverage-test.sh

@@ -67,11 +67,11 @@ save_cov "tag_example2";
 # save_cov "spmv";
 # 
 # echo "spmv.gpu"
-# timing=`NCPUS=0 $BUILDDIR/examples/spmv/dw_spmv 2> /dev/null`
+# timing=`STARPU_NCPUS=0 $BUILDDIR/examples/spmv/dw_spmv 2> /dev/null`
 # save_cov "spmv.gpu";
 # 
 # echo "spmv.cpu"
-# timing=`NCUDA=0 $BUILDDIR/examples/spmv/dw_spmv 2> /dev/null`
+# timing=`STARPU_NCUDA=0 $BUILDDIR/examples/spmv/dw_spmv 2> /dev/null`
 # save_cov "spmv.cpu";
 # 
 # echo "spmv.dm"
@@ -106,11 +106,11 @@ timing=`CALIBRATE=1 SCHED="dmda" $EXAMPLEDIR/dw_cholesky -pin 2> /dev/null`
 save_cov "chol.dmda";
 
 echo "chol.cpu"
-timing=`CALIBRATE=1 NCUDA=0 SCHED="dm" $EXAMPLEDIR/dw_cholesky -pin 2> /dev/null`
+timing=`CALIBRATE=1 STARPU_NCUDA=0 SCHED="dm" $EXAMPLEDIR/dw_cholesky -pin 2> /dev/null`
 save_cov "chol.cpu";
 
 echo "chol.gpu"
-timing=`CALIBRATE=1 NCPUS=0 SCHED="dm" $EXAMPLEDIR/dw_cholesky -pin 2> /dev/null`
+timing=`CALIBRATE=1 STARPU_NCPUS=0 SCHED="dm" $EXAMPLEDIR/dw_cholesky -pin 2> /dev/null`
 save_cov "chol.gpu";
 
 echo "chol"

tests/heat/dmda.sh

@@ -55,8 +55,8 @@ mkdir -p $SAMPLINGDIR
 #schedlist="ws no-prio greedy prio dm random"
 #schedlist="random random random random"
 
-export NCUDA=3
-export NCPUS=8 
+export STARPU_NCUDA=3
+export STARPU_NCPUS=8 
 
 #sizelist="2048 4096 6144 8192 10240 12288 14336 16384 24576 28672 30720"
 sizelist=`seq 2048 2048 30720`

tests/heat/granularity.sh

@@ -60,7 +60,7 @@ trace_granularity()
 		for iter in `seq 1 $maxiter`
 		do
 			echo "$iter / $maxiter"
-			 val=`NCPUS=8 NCUDA=3 SCHED="dmda" PREFETCH=1 CALIBRATE=1 $ROOTDIR/examples/heat/heat $OPTIONS 2> /dev/null`
+			 val=`STARPU_NCPUS=8 STARPU_NCUDA=3 SCHED="dmda" PREFETCH=1 CALIBRATE=1 $ROOTDIR/examples/heat/heat $OPTIONS 2> /dev/null`
 			 echo "$val"
 			 echo "$val" >> $filename
 		done
@@ -121,12 +121,12 @@ calibrate_grain()
 
 	OPTIONS="-pin -nblocks $blocks -size $size -v3"
 
-	NCUDA=3 NCPUS=8 CALIBRATE=1 SCHED="dm" $ROOTDIR/examples/heat/heat $OPTIONS 2> /dev/null 
-	NCUDA=3 NCPUS=8 CALIBRATE=1 PREFETCH=1 SCHED="dmda" $ROOTDIR/examples/heat/heat $OPTIONS 2> /dev/null
-	NCUDA=3 NCPUS=8 CALIBRATE=1 PREFETCH=1 SCHED="dmda" $ROOTDIR/examples/heat/heat $OPTIONS 2> /dev/null
-	NCUDA=3 NCPUS=8 CALIBRATE=1 PREFETCH=1 SCHED="dmda" $ROOTDIR/examples/heat/heat $OPTIONS 2> /dev/null
-	NCUDA=3 NCPUS=8 CALIBRATE=1 PREFETCH=1 SCHED="dmda" $ROOTDIR/examples/heat/heat $OPTIONS 2> /dev/null
-	NCUDA=3 NCPUS=8 CALIBRATE=1 PREFETCH=1 SCHED="dmda" $ROOTDIR/examples/heat/heat $OPTIONS 2> /dev/null
+	STARPU_NCUDA=3 STARPU_NCPUS=8 CALIBRATE=1 SCHED="dm" $ROOTDIR/examples/heat/heat $OPTIONS 2> /dev/null 
+	STARPU_NCUDA=3 STARPU_NCPUS=8 CALIBRATE=1 PREFETCH=1 SCHED="dmda" $ROOTDIR/examples/heat/heat $OPTIONS 2> /dev/null
+	STARPU_NCUDA=3 STARPU_NCPUS=8 CALIBRATE=1 PREFETCH=1 SCHED="dmda" $ROOTDIR/examples/heat/heat $OPTIONS 2> /dev/null
+	STARPU_NCUDA=3 STARPU_NCPUS=8 CALIBRATE=1 PREFETCH=1 SCHED="dmda" $ROOTDIR/examples/heat/heat $OPTIONS 2> /dev/null
+	STARPU_NCUDA=3 STARPU_NCPUS=8 CALIBRATE=1 PREFETCH=1 SCHED="dmda" $ROOTDIR/examples/heat/heat $OPTIONS 2> /dev/null
+	STARPU_NCUDA=3 STARPU_NCPUS=8 CALIBRATE=1 PREFETCH=1 SCHED="dmda" $ROOTDIR/examples/heat/heat $OPTIONS 2> /dev/null
 }
 
 mkdir -p $TIMINGDIR

tests/heat/heat.sh

@@ -44,8 +44,8 @@ do
 			blocks=$BLOCKS
 			filename=$TIMINGDIR/timing.$cpus.$cublas.$size.$blocks
 
-			export NCPUS=$cpus
-			export NCUDA=$cublas
+			export STARPU_NCPUS=$cpus
+			export STARPU_NCUDA=$cublas
 
 			echo "size $size cpus $cpus cublas $cublas blocks $blocks" 
 			$ROOTDIR/examples/heat -nthick 34 -ntheta $(($theta+2)) -nblocks $BLOCKS 2>/dev/null| tee $filename
@@ -60,8 +60,8 @@ do
 			blocks=$BLOCKS
 			filename=$TIMINGDIR/timing.$cpus.$cublas.$size.$blocks
 
-			export NCPUS=$cpus
-			export NCUDA=$cublas
+			export STARPU_NCPUS=$cpus
+			export STARPU_NCUDA=$cublas
 
 			echo "size $size cpus $cpus cublas $cublas blocks $blocks" 
 			$ROOTDIR/examples/heat -nthick 34 -ntheta $(($theta+2)) -nblocks $BLOCKS 2>/dev/null| tee $filename
@@ -86,15 +86,15 @@ do
 	do
 		size=$(($theta*32))
 
-		export NCPUS=4
-		export NCUDA=0
+		export STARPU_NCPUS=4
+		export STARPU_NCUDA=0
 
 		echo "size $size cpus 4 cublas 0 blocks $blocks"
 		filename=$TIMINGDIR/timing.4.0.$size.$blocks
 		$ROOTDIR/examples/heat -nthick 34 -ntheta $(($theta+2)) -nblocks $blocks 2>/dev/null| tee $filename
 
-		export NCPUS=3
-		export NCUDA=1
+		export STARPU_NCPUS=3
+		export STARPU_NCUDA=1
 
 		echo "size $size cpus 3 cublas 1 blocks $blocks"
 		filename=$TIMINGDIR/timing.3.1.$size.$blocks

tests/heat/sched.sh

@@ -97,8 +97,8 @@ mkdir -p $SAMPLINGDIR
 #schedlist="ws no-prio greedy prio dm random"
 #schedlist="random random random random"
 
-export NCUDA=3
-export NCPUS=8 
+export STARPU_NCUDA=3
+export STARPU_NCPUS=8 
 
 trace_sched "greedy" 0 0;
 trace_sched "dm" 0 1;

tests/heat/speedup.sh

@@ -37,7 +37,7 @@ do
 
 	for cpus in 1 2 4 6 8 10 12 14 16
 	do
-		export NCPUS=$cpus
+		export STARPU_NCPUS=$cpus
 
 		echo "ncpus $cpus size $size"
 

tests/memory/memstress.sh

@@ -31,8 +31,8 @@ trace_stress()
 {
 	memstress=$1
 
-	export NCPUS=0
-	export NCUDA=1
+	export STARPU_NCPUS=0
+	export STARPU_NCUDA=1
 	export STRESS_MEM=$memstress
 
 	line="$memstress"

tests/mult/sched.sh

@@ -37,7 +37,7 @@ trace_sched()
 		do
 			echo "$iter / $maxiter"
 			#echo "$ROOTDIR/examples/mult/sgemm $OPTIONS 2> /dev/null"
-			NCUDA=1 CALIBRATE=1 SCHED="$sched" $ROOTDIR/examples/mult/sgemm $OPTIONS 2> /dev/null >> $filename
+			STARPU_NCUDA=1 CALIBRATE=1 SCHED="$sched" $ROOTDIR/examples/mult/sgemm $OPTIONS 2> /dev/null >> $filename
 			tail -1 $filename
 		done
 	done

tests/mult/speedup.sh

@@ -35,7 +35,7 @@ do
 
 	for cpus in 1 2 4 6 8 10 12 14 16
 	do
-		export NCPUS=$cpus
+		export STARPU_NCPUS=$cpus
 
 		echo "ncpus $cpus size $size"
 

tests/perf-models/bench_sgemm.sh

@@ -40,7 +40,7 @@ gpu_output=$DIR/output.gpu
 rm -f $cpu_output
 rm -f $gpu_output
 
-export WORKERS_CPUID="2"
+export STARPU_WORKERS_CPUID="2"
 export CALIBRATE=1
 export SCHED="dm"
 
@@ -56,7 +56,7 @@ do
 	fi
 
 	echo "GotoBLAS -> size $size niter $niter"
-	timing=`NCPUS=1 NCUDA=0 $INSTALLDIR/lib/starpu/examples/dw_mult_no_filters -x $size -y $size -z $size -nblocks 1 -iter $niter 2> /dev/null`
+	timing=`STARPU_NCPUS=1 STARPU_NCUDA=0 $INSTALLDIR/lib/starpu/examples/dw_mult_no_filters -x $size -y $size -z $size -nblocks 1 -iter $niter 2> /dev/null`
 	echo "$size	$timing	$niter" >> $cpu_output
 done
 
@@ -73,7 +73,7 @@ do
 	fi
 
 	echo "CUBLAS -> size $size niter $niter"
-	timing=`NCPUS=0 NCUDA=1 $INSTALLDIR/lib/starpu/examples/dw_mult_no_filters -x $size -y $size -z $size -nblocks 1 -iter $niter 2 -pin 2> /dev/null`
+	timing=`STARPU_NCPUS=0 STARPU_NCUDA=1 $INSTALLDIR/lib/starpu/examples/dw_mult_no_filters -x $size -y $size -z $size -nblocks 1 -iter $niter 2 -pin 2> /dev/null`
 	echo "$size	$timing	$niter" >> $gpu_output
 done