fix reduction test for gfx1201 (#374)

* fix reduction for gfx942 and 1201

match the synchronizaation of internal_putmem_wg and internal_getmem_wg
to their non-internal counterparts. the internal_putmem_wg is used in
the ipc reduction

* move specialization to internal_putmem

scripts/functional_tests/driver.sh

@@ -178,7 +178,6 @@ ExecTest() {
   fi
 
   CMD+=" >> $LOG_DIR/$TEST_LOG_NAME.log 2>&1"
-
   # Run Test
   if [ $NUM_GPUS -ge $NUM_RANKS ] || [[ "" != "$HOSTFILE" ]]; then
     echo $TEST_LOG_NAME
@@ -452,7 +451,6 @@ TestColl() {
 
   ExecTest  "teambroadcast"    2       1            64        32768
 
-  ExecTest  "fcollect"         2       1            64        512
   ExecTest  "fcollect"         2       1            64        32768
 
   ExecTest  "teamreduction"    2       1            64        32768
@@ -641,7 +639,6 @@ TestGDA() {
 
   ExecTest  "teambroadcast"    2       1            1         32768
 
-  ExecTest  "fcollect"         2       1            1         512
   ExecTest  "fcollect"         2       1            1         32768
 
 # deadlock on gda, size 8KB