diff options
author | Christopher Ferris <cferris@google.com> | 2019-08-26 17:51:19 -0700 |
---|---|---|
committer | Christopher Ferris <cferris@google.com> | 2019-08-30 08:53:08 -0700 |
commit | 48d0744499477f371e0589ece5ad92976af29133 (patch) | |
tree | 2e09f06be81b031f85953be8192c0eddeefca3b2 /memory_replay/traces | |
parent | abde21870e073997a766055ad660dfd5a8f1fa22 (diff) | |
download | extras-48d0744499477f371e0589ece5ad92976af29133.tar.gz |
Refactor code with single trace parser.
This is the first step towards adding a new way to benchmark the
traces. This new way would be a way to add all threads running at once.
Also to try and make it possible to do the faster runs that jmgao had
done a while ago.
Modified the code so that it's possible for memory_replay to play
the zip file, no need to unzip the trace files.
Add a single parser for the trace files.
Remove as many possible allocation points in the system. This allows
the traces to be run as pure as possible.
Added new unit tests to cover new code.
Bug: 137795072
Test: All unit tests pass.
Test: Ran select traces using the old version and the new version.
Test: Both get nearlyt he same RSS, but slightly different VA space
Test: due to the new version avoiding allocations at all costs.
Test: Ran the trace benchmarks and verified that the old version and the
Test: new version produce nearly the same numbers.
Change-Id: I8e8a28436f23b09eba57772c62bb0b866d874437
Diffstat (limited to 'memory_replay/traces')
-rw-r--r-- | memory_replay/traces/README | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/memory_replay/traces/README b/memory_replay/traces/README index d306b9a4..88b7b59b 100644 --- a/memory_replay/traces/README +++ b/memory_replay/traces/README @@ -11,7 +11,7 @@ Format of dumps: <tid> The pid_t value that is the gettid() value recorded during the run. -<action_name> +<action_name> One of: malloc - Allocate memory using the malloc function. calloc - Allocate memory using the calloc function. |