restructure results

1 files changed, 35 insertions, 0 deletions

diff --git a/results/classifier/zero-shot-user-mode/syscall/1606 b/results/classifier/zero-shot-user-mode/syscall/1606
new file mode 100644
index 000000000..3e3ba29f3
--- /dev/null
+++ b/results/classifier/zero-shot-user-mode/syscall/1606
@@ -0,0 +1,35 @@
+syscall: 0.437
+instruction: 0.315
+runtime: 0.249
+
+
+
+riscv: fence.i is not functional
+Description of problem:
+The attached user-level test is designed to do the following (in iteration):
+
+  - Thread P0 on CPU0 changes some text/code, while
+
+  - Thread P1 on CPU1 checks/reads the code, fence.i, then executes the same code.
+
+Results (in stdout) indicates that CPU1 has read the new code (1:x5=a009) but executed the old one (1:x7=1) (against the specification).
+Steps to reproduce:
+1. echo 2 > /proc/sys/vm/nr_hugepages
+2. ./CoRF+fence.i
+Additional information:
+Example output:
+```[CoRF+fence.i.c](/uploads/c150ca0910783cc4bfc3886789b64c28/CoRF+fence.i.c)
+Test CoRF+fence.i Allowed
+Histogram (4 states)
+25784  :>1:x5=0xa009; 1:x7=2;
+24207  *>1:x5=0xa009; 1:x7=1;   <--  THIS LINE
+8      :>1:x5=0xa019; 1:x7=1;
+1      :>1:x5=0xa019; 1:x7=2;
+Ok
+Witnesses
+Positive: 24207 Negative 25793
+Condition exists (1:x5=0xa009 /\ 1:x7=1) is  validated
+Observation CoRF+fence.i Sometimes 24207 25793
+Time CoRF+fence.i 0.85
+Hash=
+```