trunk96 / tm-benchmark

GCC Transactional Memories benchmark to check against possible reordering of reads/writes in a transaction

STM Benchmark

This benchmark application uses TinySTM as TM library.

To compile and run the application run compile.sh. It compiles TinySTM library and the benchmark.

To execute the benchmark go in applications/stm_synch and run "app" executable

The usage for the benchmark is

./app [number_of_threads] [dummy_time] [number_of_wait_cycles] [number_of_transaction_per_thread] [shared_memory_size] [number_of_test_repetitions]

• number_of_threads is the number of thread the benchmark will spawn
• dummy_time is the time spent doing nothing (transaction padding time)
• number_of_wait_cycles is how many times a transaction has to wait dummy_time
• number_of_transactions_per_thread is how many transactions each thread executes
• shared_memory_size indicates the size in double words (8 bytes) of the shared memory block accessed by the transactions. It should be a multiple of 4, since each instruction of a transaction accesses randomly an memory area in a sub-block that is 1/4 the original block.
• number_of_test_repetition indicates the times the test will run (in order to have good statistical data on transactions time and aborts number)

To better visualize the result of the benchmark it is possible to copy and paste the output of the benchmark in a text file, save it and run python3 results.py [file_name] It would print the time plot and then the aborts plot.

To disable Intel Hyperthreading you can use applications/stm_synch/stop_ht.sh (and start_ht.sh to restart it), modifying the logical processors number for your architecture (it is possible to see which are the HT logical cores with grep -H . /sys/devices/system/cpu/cpu*/topology/thread_siblings_list, that are the second number in each printed line; more information here)