Modified pfb correlator thread to match PFB thread. Also added conditional...

Modified pfb correlator thread to match PFB thread. Also added conditional path setting to install script and CUDA path setting

install

@@ -9,7 +9,34 @@ else
 fi
 
 # Change this to match your destination directory
-prefix=~/hash
+prefix=$HOME/hash
+
+# Add prefix to PATH
+if [[ "$PATH" == ?(*:)"$prefix/bin"?(:*) ]]
+then
+    echo "PATH includes $prefix/bin"
+else
+    echo "PATH does not include $prefix/bin... Adding now"
+    export PATH=$prefix/bin:$PATH
+fi
+
+# Add prefix to LD_LIBRARY_PATH
+if [[ "$LD_LIBRARY_PATH" == ?(*:)"$prefix/lib"?(:*) ]]
+then
+    echo "LD_LIBRARY_PATH includes $prefix/lib"
+else
+    echo "LD_LIBRARY_PATH does not include $prefix/lib... Adding now"
+    export LD_LIBRARY_PATH=$prefix/lib:$LD_LIBRARY_PATH
+fi
+
+# Check for CUDA
+if [ -z "$CUDA" ]
+then
+    echo "CUDA is not set-- using $prefix/cuda"
+    CUDA=$prefix/cuda
+else
+    echo "CUDA is set to $CUDA"
+fi
 
 # Set library source directories
 HASHPIPE_SRC=$FLAG_DIR/lib/hashpipe/src
@@ -25,6 +52,7 @@ echo "Navigating to $HASHPIPE_SRC"
 cd $HASHPIPE_SRC
 autoreconf -is
 ./configure --prefix=$prefix --silent
+make clean
 make
 make install
 echo "========================================"
@@ -35,6 +63,7 @@ echo "Installing xGPU"
 echo "========================================"
 echo "Navigating to $XGPU_SRC"
 cd $XGPU_SRC
+make clean
 make
 make install prefix=$prefix
 echo "========================================"
@@ -45,6 +74,7 @@ echo "Installing flagbeamformer"
 echo "========================================"
 echo "Navigating to $BEAM_SRC"
 cd $BEAM_SRC
+make clean
 make
 make install prefix=$prefix
 echo "========================================"
@@ -55,6 +85,7 @@ echo "Installing flagPower"
 echo "========================================"
 echo "Navigating to $POWER_SRC"
 cd $POWER_SRC
+make clean
 make
 make install prefix=$prefix
 echo "========================================"
@@ -65,8 +96,9 @@ echo "Installing flagpfb"
 echo "========================================"
 echo "Navigating to $PFB_SRC"
 cd $PFB_SRC
-make
-make install prefix=$prefix
+make clean
+make CUDA_DIR=$CUDA
+make install prefix=$prefix CUDA_DIR=$CUDA
 echo "========================================"
 echo
 
@@ -77,10 +109,13 @@ echo "Navigating to $FLAG_DIR/src"
 cd $FLAG_DIR/src
 autoreconf -is
 ./configure --prefix=$prefix --with-hashpipe=$prefix --with-xgpu=$prefix --with-xgpufrb=$prefix --with-xgpupfb=$prefix --with-flagpfb=$prefix --with-flagbeamformer=$prefix --with-flagpower=$prefix --silent
+make clean
 make
 make install
 echo "========================================"
 echo 
 
+cd $FLAG_DIR
+
 echo "DONE!"
 echo "The FLAG plugins have been installed to $prefix"

src/Makefile.am

@@ -61,14 +61,14 @@ flag_x_frb_threads = flag_net_thread.c \
                      flag_frb_correlator_thread.c \
                      flag_frb_corsave_thread.c
 
-#flag_fx_threads = flag_net_thread.c \
-#                     flag_transpose_thread.c \
-#                     flag_pfb_thread.c \
-#                     flag_pfb_correlator_thread.c \
-#                     flag_pfb_corsave_thread.c
+flag_fx_threads = flag_net_thread.c \
+                     flag_transpose_thread.c \
+                     flag_pfb_thread.c \
+                     flag_pfb_correlator_thread.c \
+                     flag_pfb_corsave_thread.c
 
 # This lists all of the plugins that will be created
-lib_LTLIBRARIES = flag_x.la flag_b.la flag_f.la flag_x_frb.la flag_bx.la 
+lib_LTLIBRARIES = flag_x.la flag_b.la flag_f.la flag_x_frb.la flag_bx.la flag_fx.la 
 
 # flag_x.la sources and libraries
 flag_x_la_SOURCES  = $(flag_databuf) $(flag_x_threads) $(fifo_codes)
@@ -109,12 +109,12 @@ flag_x_frb_la_LDFLAGS += -L"@XGPU_FRB_LIBDIR@" -Wl,-rpath,"@XGPU_FRB_LIBDIR@"
 flag_x_frb_la_LDFLAGS += -L"@HASHPIPE_LIBDIR@" -Wl,-rpath,"@HASHPIPE_LIBDIR@"
 
 # flag_fx.la sources and libraries
-#flag_fx_la_SOURCES = $(flag_databuf) $(flag_fx_threads) $(fifo_codes)
-#flag_fx_la_LIBADD   = -lrt -lxgpupfb -lcublas -L/usr/local/cuda/lib64 -lcfitsio -lflagpfb -lcufft -lcudart
-#flag_fx_la_LDFLAGS  = -avoid-version -module -shared -export-dynamic --enable-shared
-#flag_fx_la_LDFLAGS += -L"@XGPU_PFB_LIBDIR@" -Wl,-rpath,"@XGPU_PFB_LIBDIR@"
-#flag_fx_la_LDFLAGS += -L"@HASHPIPE_LIBDIR@" -Wl,-rpath,"@HASHPIPE_LIBDIR@"
-#flag_fx_la_LDFLAGS += -L"@FLAGPFB_LIBDIR@" -Wl, -rpath,"@FLAGPFB_LIBDIR@"
+flag_fx_la_SOURCES = $(flag_databuf) $(flag_fx_threads) $(fifo_codes)
+flag_fx_la_LIBADD   = -lrt -lxgpupfb -lcublas -L/usr/local/cuda/lib64 -lcfitsio -lflagpfb -lcufft -lcudart
+flag_fx_la_LDFLAGS  = -avoid-version -module -shared -export-dynamic --enable-shared
+flag_fx_la_LDFLAGS += -L"@XGPU_PFB_LIBDIR@" -Wl,-rpath,"@XGPU_PFB_LIBDIR@"
+flag_fx_la_LDFLAGS += -L"@HASHPIPE_LIBDIR@" -Wl,-rpath,"@HASHPIPE_LIBDIR@"
+flag_fx_la_LDFLAGS += -L"@FLAGPFB_LIBDIR@" -Wl, -rpath,"@FLAGPFB_LIBDIR@"
 
 
 # Installed scripts

src/flag_pfb_correlator_thread.c

@@ -31,7 +31,7 @@ typedef enum {
 //         (2b) Print out some data in the block
 static void * run(hashpipe_thread_args_t * args) {
     // Local aliases to shorten access to args fields
-    flag_pfb_gpu_input_databuf_t * db_in = (flag_pfb_gpu_input_databuf_t *)args->ibuf;
+    flag_gpu_pfb_output_databuf_t * db_in = (flag_gpu_pfb_output_databuf_t *)args->ibuf;
     flag_pfb_gpu_correlator_output_databuf_t * db_out = (flag_pfb_gpu_correlator_output_databuf_t *)args->obuf;
     hashpipe_status_t st = args->st;
     const char * status_key = args->thread_desc->skey;
@@ -62,7 +62,7 @@ static void * run(hashpipe_thread_args_t * args) {
     context.array_h  = (ComplexInput *)db_in->block[0].data;
     context.matrix_h = (Complex *)db_out->block[0].data;
    
-    context.array_len = (db_in->header.n_block * sizeof(flag_pfb_gpu_input_block_t) - sizeof(flag_gpu_input_header_t))/sizeof(ComplexInput);
+    context.array_len = (db_in->header.n_block * sizeof(flag_gpu_pfb_output_block_t) - sizeof(flag_gpu_input_header_t))/sizeof(ComplexInput);
     context.matrix_len = (db_out->header.n_block * sizeof(flag_pfb_gpu_correlator_output_block_t) - sizeof(flag_gpu_output_header_t))/sizeof(Complex);
 
     int xgpu_error = xgpuInit(&context, gpu_dev);
@@ -75,7 +75,7 @@ static void * run(hashpipe_thread_args_t * args) {
         printf("XGPU CONTEXT\n");
         printf("############################################################\n");
         printf("header.n_block = %lld\n", (long long int)(db_in->header.n_block));
-        printf("sizeof(flag_pfb_gpu_input_block_t) = %lld\n", (long long int)(sizeof(flag_pfb_gpu_input_block_t)));
+        printf("sizeof(flag_gpu_pfb_output_block_t) = %lld\n", (long long int)(sizeof(flag_gpu_pfb_output_block_t)));
         printf("sizeof(flag_gpu_input_header_t) = %lld\n", (long long int)(sizeof(flag_gpu_input_header_t)));
         printf("sizeof(flag_pfb_gpu_correlator_output_block_t) = %d\n", (int)(sizeof(flag_pfb_gpu_correlator_output_block_t)));
         printf("sizeof(flag_gpu_output_header_t) = %lld\n", (long long int)(sizeof(flag_gpu_output_header_t)));
@@ -114,7 +114,7 @@ static void * run(hashpipe_thread_args_t * args) {
         if (cur_state == ACQUIRE) {
             next_state = ACQUIRE;
 	    // Wait for input buffer block to be filled
-            while ((rv=flag_pfb_gpu_input_databuf_wait_filled(db_in, curblock_in)) != HASHPIPE_OK) {
+            while ((rv=flag_gpu_pfb_output_databuf_wait_filled(db_in, curblock_in)) != HASHPIPE_OK) {
                 if (rv==HASHPIPE_TIMEOUT) {
                     int cleanb;
                     hashpipe_status_lock_safe(&st);
@@ -154,7 +154,7 @@ static void * run(hashpipe_thread_args_t * args) {
                 #if VERBOSE==1
                     fprintf(stderr, "COR: Correlator is off...\n");
                 #endif
-                flag_pfb_gpu_input_databuf_set_free(db_in, curblock_in);
+                flag_gpu_pfb_output_databuf_set_free(db_in, curblock_in);
                 curblock_in = (curblock_in + 1) % db_in->header.n_block;
                 good_data = 1;
                 continue;
@@ -179,7 +179,7 @@ static void * run(hashpipe_thread_args_t * args) {
 
                     // starting mcnt not yet reached
                     // free block and continue
-                    flag_pfb_gpu_input_databuf_set_free(db_in, curblock_in);
+                    flag_gpu_pfb_output_databuf_set_free(db_in, curblock_in);
                     curblock_in = (curblock_in + 1) % db_in->header.n_block;
                     continue;
                 }
@@ -221,7 +221,7 @@ static void * run(hashpipe_thread_args_t * args) {
 
             // If we get here, then integ_status == "on"
             // Setup for current chunk
-            context.input_offset  = curblock_in  * sizeof(flag_pfb_gpu_input_block_t) / sizeof(ComplexInput);
+            context.input_offset  = curblock_in  * sizeof(flag_gpu_pfb_output_block_t) / sizeof(ComplexInput);
             context.output_offset = curblock_out * sizeof(flag_pfb_gpu_correlator_output_block_t) / sizeof(Complex);
         
             int doDump = 0;
@@ -287,7 +287,7 @@ static void * run(hashpipe_thread_args_t * args) {
             #if VERBOSE==1
             printf("COR: Marking input block %d as free\n", curblock_in);
             #endif
-            flag_pfb_gpu_input_databuf_set_free(db_in, curblock_in);
+            flag_gpu_pfb_output_databuf_set_free(db_in, curblock_in);
             curblock_in = (curblock_in + 1) % db_in->header.n_block;
         }
         }
@@ -334,7 +334,7 @@ static hashpipe_thread_desc_t x_thread = {
     skey: "CORSTAT",
     init: NULL,
     run:  run,
-    ibuf_desc: {flag_pfb_gpu_input_databuf_create},
+    ibuf_desc: {flag_gpu_pfb_output_databuf_create},
     obuf_desc: {flag_pfb_gpu_correlator_output_databuf_create}
 };