Commit 93d7e6e8 authored by Paal Kvamme's avatar Paal Kvamme
Browse files

Update readme file, mentioning the additional parallelization done.

parent 6ee2d36a
......@@ -31,6 +31,34 @@ html/index.html. If you have
installed when building there will also be a single-document pdf
versions of those two next to the .tgz files.
## <span style="color:blue">Performance</span>
The pure Python reference implementation is single threaded.
OpenZGY/C++ will in most cases make use of multiple threads.
Applications are allowed to issue multiple read requests in parallel
to the same open file. Even if they don't, parallelization done inside
OpenZGY ensures that the performance is still reasonable
Read requests to cloud storage will automatically consolidate neighbor
bricks. This makes cloud I/O faster for typical access patterns, since
reading a typical brick (1MB) at a time from cloud storage is very
inefficient. Note that the brick size isn't increased, it just looks
that way. So when you need smaller bricks because you have scattered
access then this will work as well.
Individual read requests will be parallelized inside OpenZGY for
copy-out, int8/int16 to float conversion, compression, linux local
file read, and cloud read. As applicable.
Write requests will be serialized with respect to float to int8/int16
conversion, compression, and upload to cloud where applicable.
Also the algorithms that decimate for low resolution data.
The last two bullets are implemented for local file access but some
are still work in progress for cloud I/O.
## <span style="color:blue">Building and testing the core parts</span>
### Building Linux core
......
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment