Revised calcHTranspose operator (mmbell#68)

* Initial version with revised H^t operator * Next step in Memory reduction. * More explicit definitions of ints. * Code cleanup for new CSR format for H^t matrix * Verified that this works for the hurricane_4panel case. * More bug fixes for new-H^t operator. * Reduced int length for JH array * Change integer size * Cleanup calculation of nnz * Removed remaining old H^t code. * Minor cleanup of the code. * Cleanup * Clarification on the size of int used in the code. * Fixed OpenACC directive --------- Co-authored-by: Jian Sun <[email protected]>
sjsprecious · Jul 15, 2024 · 6e2fd8e · 6e2fd8e
1 parent 12e3fc1
commit 6e2fd8e
Show file tree

Hide file tree

Showing 6 changed files with 122 additions and 109 deletions.
diff --git a/ncar_scripts/casper_h100_submit.sh b/ncar_scripts/casper_h100_submit.sh
@@ -27,8 +27,7 @@ cd ncar_scripts
 # Run a case #
 ##############
 suffix="casper_gpu"
-for i in beltrami supercell hurricane typhoonChanthu2020  # hurricane_4panel
-
+for i in beltrami supercell hurricane typhoonChanthu2020 hurricane_4panel
 do
   ./ncar_run.sh $SAMURAI_ROOT/ncar_scripts/TDRP/${i}.tdrp >& log_${i}_$suffix.$ID
   if [ ! -d  ${i}_${suffix} ]; then

diff --git a/ncar_scripts/derecho_a100_submit.sh b/ncar_scripts/derecho_a100_submit.sh
@@ -1,7 +1,7 @@
 #!/bin/bash -l
 #PBS -N SAMURAI
 #PBS -A NEOL0013
-#PBS -l select=1:ncpus=64:ompthreads=1:mem=100GB:ngpus=1
+#PBS -l select=1:ncpus=64:ompthreads=1:mem=300GB:ngpus=1
 #PBS -q main
 #PBS -l walltime=02:30:00
 #PBS -j oe
@@ -12,6 +12,10 @@ cd ..
 export SAMURAI_ROOT=$(pwd)
 
 ID=`date '+%Y%m%d%H%M'`
+
+sed -i 's/cc70/cc80/g' CMakeLists.txt
+sed -i 's/cc90/cc80/g' CMakeLists.txt
+
 ##################
 # Build the code #
 ##################
@@ -25,7 +29,7 @@ cd ncar_scripts
 # Run a case #
 ##############
 suffix="derecho_gpu"
-for i in  beltrami supercell hurricane typhoonChanthu2020 # hurricane_4panel
+for i in beltrami supercell hurricane typhoonChanthu2020 # hurricane_4panel
 do
 
   ./ncar_run.sh $SAMURAI_ROOT/ncar_scripts/TDRP/${i}.tdrp >& log_${i}_$suffix.$ID

diff --git a/src/CostFunction.h b/src/CostFunction.h
@@ -28,7 +28,7 @@ class CostFunction
 
 protected:
 	int ls_cnt;
-	int64_t mObs;
+	uint64_t mObs;
 	int nState;
 	real* currState;
 	real* currGradient;