Design Doc: CSP in PaddlePaddle Fluid¶
Motivation¶
Concurrent programming is important for deep learning. Few example applications are:
- The main thread keeps reading the next mini-batch while another thread uses the GPU for computing.
- The main thread performs the computation while another thread uploads the local gradients from each trainer to the parameter server.
Most DL systems, including TensorFlow, Caffe2, and MxNet, can asynchronously execute operators in a graph. However, Fluid doesn’t have the concept of a graph at all, as the design goal of Fluid is that of a programming language.
Concurrent Programming Models¶
There were many concurrent programming models, implemented in various forms:
| concurrent programming model | implementation | |—–|—–| | mutex | types and functions in standard libraries | | semaphore | types and functions in standard libraries | | communicating sequential processes (CSP) | Go programming language | | actor model | Erlang programming language | | message passing | MPI | | bulk synchronous parallel (BSP) | Pregel distributed programming framework |
Since Fluid was designed to be a programming language, we would like to implement CSP in Fluid.
CSP v.s. Actor Model¶
A well-known implementation of Actor Model is the Erlang programming language. In Actor Model, processes could send messages to another process and receive messages from another process given the process IDs. We can find the three ingredients, process with ID, send, and recv, in MPI too. Indeed, we can rewrite Erlang programs in Python + MPI with possibly fewer lines of code. Our concern with Actor Model is that it doesn’t seem reasonable to implement process management in a programming language’s runtime library; instead, it should be the operating systems’ responsibility to manage processes and libraries like MPI for send/recv.
CSP in Fluid¶
Fluid has two fundamental control-flows: if-else and while. If we are to implement CSP, we need the following:
- a new data type: channel and operators send and recv,
- goroutine or thread, and
- a new control-flow: select.
We also need Python wrappers for the above components.
The type channel is conceptually the blocking queue. In Go, its implemented is a blocking circular queue, which supports send and recv.
The select operation has been in OS kernels long before Go language. All Unix kernels implement system calls poll and select. They monitor multiple file descriptors to see if I/O is possible on any of them. This takes O(N) time. Since Linux 2.6, a new system call, epoll, can do the same in O(1) time. In BSD systems, there is a similar system call kqueue. Go’s Linux implementation uses epoll.
It might be a good idea to implement Fluid’s select using epoll too. In this design doc, we start from the O(N) way, so we could focus on Python binding and the syntax.
Type Channel¶
Fluid supports many data types:
- Tensor,
- Row-sparse Tensor
- LoD Tensor,
- Tensor array, etc
Each data type is registered in the framework.proto as an enum value. To add a new type channel, we need to add a new type enum.
To expose a C++ type to Python, we need to edit the pybind.cc file. Here is an example how we expose C++ class LoDTensor.
Syntax Design¶
Create Channel¶
In Go, we create a channel by specifying the element type and buffer size:
ch := make(chan int) // a channel without buffer
ch1 := make(chan int, 100) // a channel that can buffer 100 ints.
In Fluid, we should be able to do the same:
ch = fluid.make_channel(dtype=INT)
ch1 = fluid.make_channel(dtype=INT, 100)
In addition to that, we want channels that can hold more complex element types, e.g., Tensors of float16:
ch = fluid.make_channel(dtype=Tensor, etype=float16)
or Tensors of Tensors of float16 etc.
The point here is that we need a consistent way to compose types, like in C++ we can have Tensor<Tensor<...<float16>...> >.
Send and Recv¶
In Go, we first create a channel as explained in the section above and then perform read and write operations on top of the channels.
ch1 := make(chan int)
ch2 := make(chan int, 100)
To write (or perform a Send operation) the value of a variable x, to channel ch1 above, we perform the following:
ch1 <- x
fmt.Println("Written to the channel")
Now to read (or perform a Recv operation) the value stored in ch2 into a variable y, we perform the following:
y <- ch2
fmt.Println("Received from channel")
In Fluid, we should be able to perform the above operations on the channel objects as well. As of now, we support two different kinds of channels : Buffered Channel and UnBuffered Channel
Send and Receive can be performed as following on a buffered channel:
import threading
def send_to_channel(channel, num_time=1):
for i in xrange(num_time):
channel.send(i)
# Create a buffered channel of capacity 10
buffer_size = 10;
ch = fluid.make_channel(dtype=INT, buffer_size)
# Now write three elements to the channel
thread = threading.Thread(target=send_to_channel, args=(ch, 3, ))
thread.daemon = True
thread.start()
# Read all the data from the channel
for i in xrange(3):
y = ch.recv()
# Done receiving , now close the channel
ch.close()
The send and receive operations will be similar for unbuffered channel as well, except for the fact that there is no buffer in an unbuffered channel, so the operations are completely synchronized. For example:
import threading
def send_to_channel(channel, data):
channel.send(data)
# Create an unbuffered channel
ch = fluid.make_channel(dtype=INT)
# Writes and Reads are synchronous otherwise the calls will block.
thread = threading.Thread(target=send_to_channel, args=(ch, 10, ))
thread.daemon = True
thread.start()
y = ch.recv()
# Done receiving , now close the channel
ch.close()