Skip to content

Commit

Permalink
[SYCL][Doc] Pass a device to graph constructor
Browse files Browse the repository at this point in the history 
Based on feedback from the Working Group it is
clearer to represent the current single device characteristic
of a graph by passing a device to the graph constructor that all
nodes will target. Then remove the `device` parameter from the
explicit API, as it can be added as an overload in a future revision
due to experimental nature of the extension

A diff of these changes against the commit prior to when the [original
multi-device PR](#83) went in
can be seen
[here](28acfa8...reble:llvm:ewan/single_device)
  • Loading branch information
@EwanC
EwanC committed Mar 30, 2023
1 parent 59a3522 commit ad5df75
Showing 1 changed file with 30 additions and 62 deletions.
92 changes: 30 additions & 62 deletions sycl/doc/extensions/proposed/sycl_ext_oneapi_graph.asciidoc
View file
Original file line number Diff line number Diff line change
Expand Up @@ -140,11 +140,11 @@ Another reason for deferring a serialize/deserialize API (8) is that its scope
could extend from emitting the graph in a binary format, to emitting a
standardized IR format that enables further device specific graph optimizations.

Multi-device support (7) is something that has been designed for in this
extension, with the definition of a graph node being device specific. However,
the ability for a user to define a single graph with nodes targeting different
devices is still disallowed until such a time as this feature can be backed up
by implementation coverage.
Multi-device support (7) is something that we are considering introducing into
the extension in later revisions, which may result in API changes. It has been
planned for to the extent that the definition of a graph node is device
specific, however currently all nodes in a graph must target the same device
provided to the graph constructor.

=== Graph Building Mechanisms

Expand Down Expand Up @@ -328,7 +328,7 @@ class command_graph {};
template<>
class command_graph<graph_state::modifiable> {
public:
command_graph(const property_list& propList = {});
command_graph(const device& syclDevice, const property_list& propList = {});
command_graph<graph_state::executable>
finalize(const context& syclContext, const property_list& propList = {}) const;
Expand All @@ -343,7 +343,7 @@ public:
node add(const property_list& propList = {});
template<typename T>
node add(const device& syclDevice, T cgf, const property_list& propList = {});
node add(T cgf, const property_list& propList = {});
void make_edge(node src, node dest);
};
Expand Down Expand Up @@ -422,9 +422,8 @@ This extension adds a new `command_graph` object which follows the
{crs}[common reference semantics] of other SYCL runtime objects.

A `command_graph` represents a directed acyclic graph of nodes, where each node
represents a single command for a specific device or a sub-graph. A graph may be
made up of nodes targeting different devices. The execution of a graph completes
when all its nodes have completed.
represents a single command for a specific device or a sub-graph. The execution
of a graph completes when all its nodes have completed.

A `command_graph` is built up by either recording queue submissions or
explicitly adding nodes, then once the user is happy that the graph is complete,
Expand Down Expand Up @@ -480,11 +479,11 @@ Table {counter: tableNumber}. Constructor of the `command_graph` class.
[source,c++]
----
using namespace ext::oneapi::experimental;
command_graph(const property_list& propList = {});
command_graph(const device& syclDevice, const property_list& propList = {});
----
|Creates a SYCL `command_graph` object in the modifiable state.
Zero or more properties can be provided to the constructed SYCL `command_graph`
via an instance of `property_list`.
|Creates a SYCL `command_graph` object in the modifiable state for device
`syclDevice`. Zero or more properties can be provided to the constructed SYCL
`command_graph` via an instance of `property_list`.

Preconditions:

Expand All @@ -493,6 +492,9 @@ Preconditions:

Parameters:

* `syclDevice` - Device that all nodes added to the graph will target,
an immutable characteristic of the graph.

* `propList` - Optional parameter for passing properties. No `command_graph`
constructor properties are defined by this extension.

Expand Down Expand Up @@ -536,7 +538,7 @@ Exceptions:
----
using namespace ext::oneapi::experimental;
template<typename T>
node add(const device& syclDevice, T cgf, const property_list& propList = {});
node add(T cgf, const property_list& propList = {});
----
|This function adds a command group function object to a graph. The function
object statically contains a group of commands, of which a single command is
Expand All @@ -552,13 +554,6 @@ Preconditions:

Parameters:

* `syclDevice` - Device to process `cgf` with to create node command-group.
Commands are not required to execute on this device if the behaviour is
consistent with how they would not execute on the device associated with a
`sycl::queue` during regular queue submission. In particular host tasks,
sub-graphs with nodes targeting other devices, and memory copies may not
execute on `syclDevice`.

* `cgf` - Command group function object to be added as a node.

* `propList` - Zero or more properties can be provided to the constructed node
Expand All @@ -571,9 +566,6 @@ Exceptions:
* Throws synchronously with error code `invalid` if a queue is recording
commands to the graph.

* Throws synchronously with error code `invalid` if `syclDevice` is a different
device from the device targeted by the existing nodes of the graph.

|
[source,c++]
----
Expand Down Expand Up @@ -676,9 +668,8 @@ Exceptions:
already recording to a different graph.

* Throws synchronously with error code `invalid` if `recordingQueue` is
associated with a device which is different from the device already targeted
by the nodes of the graph.

associated with a device that is different from the device used on creation
of the graph.
|
[source, c++]
----
Expand Down Expand Up @@ -1015,10 +1006,7 @@ result in a synchronous exception.
When a queue is in recording mode asynchronous exceptions will not be
generated, as no device execution is occurring. Synchronous errors specified as
being thrown in the default queue executing state, will still be thrown when a
queue is in the recording state. A synchronous error with error code `invalid`
will also be thrown if a queue in recording mode tries to record a command to a
graph which already has nodes that target a device different from the device
associated with the recording queue.
queue is in the recording state.

The `command_graph::begin_recording` and `command_graph::end_recording`
entry-points return a `bool` value informing the user whether a related queue
Expand Down Expand Up @@ -1098,7 +1086,7 @@ evaluated as normal during command graph execution.
[source,c++]
----
using namespace ext::oneapi::experimental;
auto node = graph.add(device, [&](sycl::handler& cgh){
auto node = graph.add([&](sycl::handler& cgh){
// Host code here is evaluated during the call to add()
cgh.host_task([=](){
// Code here is evaluated as part of executing the command graph node
Expand Down Expand Up @@ -1163,23 +1151,6 @@ problem this extension currently aims to solve, it is the responsibility of the
user to decide the device each command will be processed for, not the SYCL
runtime.

A drawback of this design is that it is less convenient for the use-case where
a user would like to run the same graph on N devices. Rather than finalizing a
single modifiable graph N times for N devices in this scenario, the user now has
to record N modifiable graphs and then as finalize each of them. If this use
case does become a usability issue, we could provide a specific API to support
it. For example, an update API for the modifiable `command_graph`, similar to
parameters, so that the same modifiable graph can be updated to new devices then
re-finalized. There may need to be limitations on what devices a graph can be
updated to however, as device specific processing might not be possible to roll
back or defer. As such, it may be only possible to update to identical physical
devices, rather than re-target a GPU constructed graph to a CPU.

The queue an executable graph is submitted to needs to have the same context as
the executable graph, however execution of a multi-device graph is not limited
to the device associated to the queue. Instead, the queue is used as the
mechanism to submit the graph and express execution dependencies.

== Examples

[NOTE]
Expand Down Expand Up @@ -1209,9 +1180,7 @@ int main() {
float gamma = 3.0f;
sycl::queue q;
sycl::device device = q.get_device();
sycl_ext::command_graph g;
sycl_ext::command_graph g(q.get_device());
float *dotp = sycl::malloc_shared<float>(1, q);
float *x = sycl::malloc_device<float>(n, q);
Expand All @@ -1227,7 +1196,7 @@ int main() {
// c
/* init data on the device */
auto node_i = g.add(device, [&](sycl::handler& h) {
auto node_i = g.add([&](sycl::handler& h) {
h.parallel_for(n, [=](sycl::id<1> it){
const size_t i = it[0];
x[i] = 1.0f;
Expand All @@ -1236,21 +1205,21 @@ int main() {
});
});
auto node_a = g.add(device, [&](sycl::handler& h) {
auto node_a = g.add([&](sycl::handler& h) {
h.parallel_for(sycl::range<1>{n}, [=](sycl::id<1> it) {
const size_t i = it[0];
x[i] = alpha * x[i] + beta * y[i];
});
}, { sycl_ext::property::node::depends_on(node_i)});
auto node_b = g.add(device, [&](sycl::handler& h) {
auto node_b = g.add([&](sycl::handler& h) {
h.parallel_for(sycl::range<1>{n}, [=](sycl::id<1> it) {
const size_t i = it[0];
z[i] = gamma * z[i] + beta * y[i];
});
}, { sycl_ext::property::node::depends_on(node_i)});
auto node_c = g.add(device,
auto node_c = g.add(
[&](sycl::handler& h) {
h.parallel_for(sycl::range<1>{n},
sycl::reduction(dotp, 0.0f, std::plus()),
Expand Down Expand Up @@ -1295,7 +1264,7 @@ submitted in its entirety for execution via
queue q{default_selector{}};
// New object representing graph of command-groups
ext::oneapi::experimental::command_graph graph;
ext::oneapi::experimental::command_graph graph(q.get_device());
{
buffer<T> bufferA{dataA.data(), range<1>{elements}};
buffer<T> bufferB{dataB.data(), range<1>{elements}};
Expand Down Expand Up @@ -1368,10 +1337,9 @@ submitted in its entirety for execution via

Allow an executable graph to contain nodes targeting different devices.

**Outcome:** This feature has been designed for with the definition of a graph
node being device specific. However, the ability for a user to define a single
graph with nodes targeting different devices is still disallowed until such a
time as this feature can be backed up by implementation coverage.
**Outcome:** This feature is something that we are considering introducing into
the extension in later revisions. It has been planned for to the extent that the
definition of a graph node is device specific.

=== Memory Allocation API

Expand Down

0 comments on commit ad5df75

Please sign in to comment.