Updated notes section.

git-svn-id: svn://svn.valgrind.org/valgrind/trunk@8797

diff --git a/drd/scripts/run-splash2 b/drd/scripts/run-splash2
index 916836aae615153e9944ca39839b9a295b2ea281..0db3b79d7aec12750f2d79db65921b516b608986 100755 (executable)
--- a/drd/scripts/run-splash2
+++ b/drd/scripts/run-splash2
@@ -116,7 +116,14 @@ fi
 # 20. Execution time ratio for Intel Thread Checker -p4 versus -p4.
 # 21. Execution time ratio for Intel Thread Checker -p4 versus -p4.
 #
-# Note: Helgrind uses a granularity of four bytes for data race detection, and DRD a granularity of one byte.
+# Notes:
+# - Both Helgrind and DRD use a granularity of one byte for data race detection.
+# - Helgrind does detect data races on stack variables. DRD only detects
+#   data races on stack variables with --check-stack-var=yes.
+# - The ITC tests have been run on a 4-way 2.5 GHz Pentium 4 workstation.
+#   Not yet clear to me: have these tests been run on a 32-bit or on a 64-bit
+#   OS ? Which OS ? Which granularity does ITC use ? And which m4 macro's
+#   have been used by ITC as implementation of the synchronization primitives ?
 #
 #     1                    2     3      4      5      6     7     8   9    10   11   12  13   14  15   16    17   18    19   20   21
 ####################################################################################################################################
@@ -210,8 +217,10 @@ run_test ${SPLASH2}/codes/kernels/lu/contiguous_blocks/LU -n512
 run_test ${SPLASH2}/codes/apps/ocean/contiguous_partitions/OCEAN -n258
 #run_test ${SPLASH2}/codes/apps/ocean/non_contiguous_partitions/OCEAN -n258
 
-# Radiosity.
+# Radiosity. Runs fine on a 32-bit OS, but deadlocks on a 64-bit OS. Not clear to me why.
+if [ $(uname -p) = "i686" ]; then
 psep=' ' run_test ${SPLASH2}/codes/apps/radiosity/RADIOSITY -batch -room -ae 5000.0 -en 0.050 -bf 0.10
+fi
 
 # Radix
 run_test ${SPLASH2}/codes/kernels/radix/RADIX -n$((2**20)) -r1024