Skip to content
This repository was archived by the owner on Mar 28, 2023. It is now read-only.

[SYCL][XPTI] Enable XPTI and XPTI Framework E2E tests #458

Merged
vladimirlaz merged 14 commits into from
Dec 22, 2021
Merged

[SYCL][XPTI] Enable XPTI and XPTI Framework E2E tests #458

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
14 commits
Select commit Hold shift + click to select a range
d26781a
[SYCL][XPTI] Enable XPTI and XPTI Frameworks E2E tests
alexbatashev Sep 11, 2021
3684efb
Merge remote-tracking branch 'upstream/intel' into xpti_tests
alexbatashev Sep 11, 2021
3eedfa4
clang-format
alexbatashev Sep 12, 2021
6f9223e
Address feedback
alexbatashev Sep 18, 2021
1e5f8f3
Update SYCL/XPTI/basic_event_collection.cpp
alexbatashev Sep 21, 2021
c2d58b9
Merge branch 'intel' into xpti_tests
Dec 17, 2021
ab58469
fix
Dec 17, 2021
9b7519b
Merge remote-tracking branch 'upstream/intel' into xpti_tests
Dec 17, 2021
4f1fd32
update test
Dec 17, 2021
1c9938e
update codeowners
alexbatashev Dec 18, 2021
5bb745c
minor fixes
Dec 20, 2021
ba8dadd
Merge branch 'xpti_tests' of github.com:alexbatashev/llvm-test-suite …
Dec 20, 2021
7c018cb
make tests device agnostic?
Dec 21, 2021
2606dbe
windows?
alexbatashev Dec 22, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions .github/CODEOWNERS
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -75,3 +75,6 @@ SYCL/BFloat16 @AlexeySotkin @MrSidims

# Deprecated features
SYCL/DeprecatedFeatures @intel/llvm-reviewers-runtime

# XPTI and XPTI Framework
SYCL/XPTI @tovinkere @andykaylor @alexbatashev
138 changes: 138 additions & 0 deletions SYCL/XPTI/Inputs/test_collector.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,138 @@
#include "xpti/xpti_trace_framework.hpp"

#include <iostream>
#include <mutex>
#include <string_view>

std::mutex GMutex;

XPTI_CALLBACK_API void syclCallback(uint16_t, xpti::trace_event_data_t *,
xpti::trace_event_data_t *, uint64_t,
const void *);
XPTI_CALLBACK_API void syclPiCallback(uint16_t, xpti::trace_event_data_t *,
xpti::trace_event_data_t *, uint64_t,
const void *);

XPTI_CALLBACK_API void xptiTraceInit(unsigned int MajorVersion,
unsigned int MinorVersion,
const char *VersionStr,
const char *StreamName) {
std::cout << "xptiTraceInit: Stream Name = " << StreamName << "\n";
std::string_view NameView{StreamName};

if (NameView == "sycl.pi") {
uint8_t StreamID = xptiRegisterStream(StreamName);
xptiRegisterCallback(
StreamID,
static_cast<uint16_t>(xpti::trace_point_type_t::function_begin),
syclPiCallback);
xptiRegisterCallback(
StreamID,
static_cast<uint16_t>(xpti::trace_point_type_t::function_with_args_end),
syclPiCallback);
}
if (NameView == "sycl") {
uint8_t StreamID = xptiRegisterStream(StreamName);
xptiRegisterCallback(
StreamID, static_cast<uint16_t>(xpti::trace_point_type_t::graph_create),
syclCallback);
xptiRegisterCallback(
StreamID, static_cast<uint16_t>(xpti::trace_point_type_t::node_create),
syclCallback);
xptiRegisterCallback(
StreamID, static_cast<uint16_t>(xpti::trace_point_type_t::edge_create),
syclCallback);
xptiRegisterCallback(
StreamID, static_cast<uint16_t>(xpti::trace_point_type_t::task_begin),
syclCallback);
xptiRegisterCallback(
StreamID, static_cast<uint16_t>(xpti::trace_point_type_t::task_end),
syclCallback);
xptiRegisterCallback(
StreamID, static_cast<uint16_t>(xpti::trace_point_type_t::signal),
syclCallback);
xptiRegisterCallback(
StreamID,
static_cast<uint16_t>(xpti::trace_point_type_t::barrier_begin),
syclCallback);
xptiRegisterCallback(
StreamID, static_cast<uint16_t>(xpti::trace_point_type_t::barrier_end),
syclCallback);
xptiRegisterCallback(
StreamID, static_cast<uint16_t>(xpti::trace_point_type_t::wait_begin),
syclCallback);
xptiRegisterCallback(
StreamID, static_cast<uint16_t>(xpti::trace_point_type_t::wait_end),
syclCallback);
xptiRegisterCallback(
StreamID, static_cast<uint16_t>(xpti::trace_point_type_t::signal),
syclCallback);
}
}

XPTI_CALLBACK_API void xptiTraceFinish(const char *streamName) {
std::cout << "xptiTraceFinish: Stream Name = " << streamName << "\n";
}

XPTI_CALLBACK_API void syclPiCallback(uint16_t TraceType,
xpti::trace_event_data_t *,
xpti::trace_event_data_t *, uint64_t,
const void *UserData) {
std::lock_guard Lock{GMutex};
auto Type = static_cast<xpti::trace_point_type_t>(TraceType);
const char *funcName = static_cast<const char *>(UserData);
if (Type == xpti::trace_point_type_t::function_begin) {
std::cout << "PI Call Begin : ";
} else if (Type == xpti::trace_point_type_t::function_end) {
std::cout << "PI Call End : ";
}
std::cout << funcName << "\n";
}

XPTI_CALLBACK_API void syclCallback(uint16_t TraceType,
xpti::trace_event_data_t *,
xpti::trace_event_data_t *Event, uint64_t,
const void *UserData) {
std::lock_guard Lock{GMutex};
auto Type = static_cast<xpti::trace_point_type_t>(TraceType);
switch (Type) {
case xpti::trace_point_type_t::graph_create:
std::cout << "Graph create\n";
break;
case xpti::trace_point_type_t::node_create:
std::cout << "Node create\n";
break;
case xpti::trace_point_type_t::edge_create:
std::cout << "Edge create\n";
break;
case xpti::trace_point_type_t::task_begin:
std::cout << "Task begin\n";
break;
case xpti::trace_point_type_t::task_end:
std::cout << "Task end\n";
break;
case xpti::trace_point_type_t::signal:
std::cout << "Signal\n";
break;
case xpti::trace_point_type_t::wait_begin:
std::cout << "Wait begin\n";
break;
case xpti::trace_point_type_t::wait_end:
std::cout << "Wait end\n";
break;
case xpti::trace_point_type_t::barrier_begin:
std::cout << "Barrier begin\n";
break;
case xpti::trace_point_type_t::barrier_end:
std::cout << "Barrier end\n";
break;
default:
std::cout << "Unknown tracepoint\n";
}

xpti::metadata_t *Metadata = xptiQueryMetadata(Event);
for (auto &Item : *Metadata) {
std::cout << " " << xptiLookupString(Item.first) << " : "
<< xptiLookupString(Item.second) << "\n";
}
}
175 changes: 175 additions & 0 deletions SYCL/XPTI/basic_event_collection.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,175 @@
// REQUIRES: xptifw
// RUN: %clangxx -fsycl %s -DXPTI_COLLECTOR -shared -fPIC -std=c++17 -o %t_collector.dll -lxptifw
// RUN: %clangxx -fsycl -fsycl-unnamed-lambda -fsycl-targets=%sycl_triple %s -o %t.out
// RUN: env XPTI_TRACE_ENABLED=1 env XPTI_FRAMEWORK_DISPATCHER=%xptifw_lib env XPTI_SUBSCRIBERS=%t_collector.dll %CPU_RUN_PLACEHOLDER %t %CPU_CHECK_PLACEHOLDER
Copy link

@vladimirlaz vladimirlaz Sep 16, 2021

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The test is running on CPU only. Is it a requirement? If so, cpu should be added to REQUIRES. If not we can remove CPU_RUN_PLACEHOLDERs.

As for my understanding, it should work for any BE so it is better to removed dependencies on specific device type.


#ifdef XPTI_COLLECTOR

#include "Inputs/test_collector.cpp"

#else

#include <sycl/sycl.hpp>

int main() {
sycl::queue Q{sycl::default_selector{}};

auto Ptr = sycl::malloc_device<int>(1, Q);

auto Evt1 = Q.single_task([=]() { Ptr[0] = 1; });

auto Evt2 = Q.submit([&](sycl::handler &CGH) {
CGH.depends_on(Evt1);
CGH.single_task([=]() { Ptr[0]++; });
});

Evt2.wait();

int Res = 0;
Q.memcpy(&Res, Ptr, 1);
Q.wait();

assert(Res == 2);

return 0;
}

#endif

// CHECK: xptiTraceInit: Stream Name = sycl
// CHECK-NEXT: Graph create
// CHECK-NEXT: xptiTraceInit: Stream Name = sycl.pi
// CHECK-NEXT: xptiTraceInit: Stream Name = sycl.pi.debug
// CHECK-NEXT: PI Call Begin : piPlatformsGet
// CHECK-NEXT: PI Call Begin : piPlatformsGet
// CHECK-NEXT: PI Call Begin : piDevicesGet
// CHECK-NEXT: PI Call Begin : piDevicesGet
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceRetain
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piPlatformGetInfo
// CHECK-NEXT: PI Call Begin : piPlatformGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceRelease
// CHECK-NEXT: PI Call Begin : piDevicesGet
// CHECK-NEXT: PI Call Begin : piDevicesGet
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDevicesGet
// CHECK-NEXT: PI Call Begin : piDevicesGet
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDevicesGet
// CHECK-NEXT: PI Call Begin : piDevicesGet
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceRetain
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDevicesGet
// CHECK-NEXT: PI Call Begin : piDevicesGet
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piContextCreate
// CHECK-NEXT: PI Call Begin : piQueueCreate
// CHECK-NEXT: PI Call Begin : piextUSMDeviceAlloc
// CHECK-NEXT: PI Call Begin : piextDeviceSelectBinary
// CHECK-NEXT: Node create
// CHECK-NEXT: from_source : false
// CHECK-NEXT: kernel_name :
// CHECK-NEXT: sycl_device : CPU
// CHECK-NEXT: Task begin
// CHECK-NEXT: from_source : false
// CHECK-NEXT: kernel_name :
// CHECK-NEXT: sycl_device : CPU
// CHECK-NEXT: PI Call Begin : piextDeviceSelectBinary
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piPlatformGetInfo
// CHECK-NEXT: PI Call Begin : piPlatformGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piPlatformGetInfo
// CHECK-NEXT: PI Call Begin : piPlatformGetInfo
// CHECK-NEXT: PI Call Begin : piProgramCreate
// CHECK-NEXT: PI Call Begin : piProgramBuild
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piDeviceGetInfo
// CHECK-NEXT: PI Call Begin : piPlatformGetInfo
// CHECK-NEXT: PI Call Begin : piPlatformGetInfo
// CHECK-NEXT: PI Call Begin : piKernelCreate
// CHECK-NEXT: PI Call Begin : piKernelSetExecInfo
// CHECK-NEXT: PI Call Begin : piextKernelSetArgPointer
// CHECK-NEXT: PI Call Begin : piKernelGetGroupInfo
// CHECK-NEXT: PI Call Begin : piEnqueueKernelLaunch
// CHECK-NEXT: Signal
// CHECK-NEXT: from_source : false
// CHECK-NEXT: kernel_name :
// CHECK-NEXT: sycl_device : CPU
// CHECK-NEXT: Task end
// CHECK-NEXT: from_source : false
// CHECK-NEXT: kernel_name :
// CHECK-NEXT: sycl_device : CPU
// CHECK-NEXT: PI Call Begin : piextDeviceSelectBinary
// CHECK-NEXT: Node create
// CHECK-NEXT: from_source : false
// CHECK-NEXT: kernel_name :
// CHECK-NEXT: sycl_device : CPU
// CHECK-NEXT: Node create
// CHECK-NEXT: kernel_name : virtual_node
// CHECK-NEXT: Edge create
// CHECK-NEXT: event : Event
// CHECK-NEXT: kernel_name : virtual_node
// CHECK-NEXT: Task begin
// CHECK-NEXT: from_source : false
// CHECK-NEXT: kernel_name :
// CHECK-NEXT: sycl_device : CPU
// CHECK-NEXT: PI Call Begin : piKernelCreate
// CHECK-NEXT: PI Call Begin : piKernelSetExecInfo
// CHECK-NEXT: PI Call Begin : piextKernelSetArgPointer
// CHECK-NEXT: PI Call Begin : piKernelGetGroupInfo
// CHECK-NEXT: PI Call Begin : piEnqueueKernelLaunch
// CHECK-NEXT: Signal
// CHECK-NEXT: from_source : false
// CHECK-NEXT: kernel_name :
// CHECK-NEXT: sycl_device : CPU
// CHECK-NEXT: Task end
// CHECK-NEXT: from_source : false
// CHECK-NEXT: kernel_name :
// CHECK-NEXT: sycl_device : CPU
// CHECK-NEXT: Wait begin
// CHECK-NEXT: PI Call Begin : piEventsWait
// CHECK-NEXT: Wait end
// CHECK-NEXT: PI Call Begin : piextUSMEnqueueMemcpy
// CHECK-NEXT: PI Call Begin : piEventRelease
// CHECK-NEXT: Wait begin
// CHECK-NEXT: sym_line_no : 34
// CHECK-NEXT: sym_function_name : main
// CHECK-NEXT: sycl_device : CPU
// CHECK-NEXT: PI Call Begin : piQueueFinish
// CHECK-NEXT: Wait end
// CHECK-NEXT: sym_line_no : 34
// CHECK-NEXT: sym_function_name : main
// CHECK-NEXT: sycl_device : CPU
// CHECK-NEXT: PI Call Begin : piEventRelease
// CHECK-NEXT: PI Call Begin : piEventRelease
// CHECK-NEXT: PI Call Begin : piQueueRelease
// CHECK-NEXT: PI Call Begin : piContextRelease
// CHECK-NEXT: PI Call Begin : piKernelRelease
// CHECK-NEXT: PI Call Begin : piKernelRelease
// CHECK-NEXT: PI Call Begin : piProgramRelease
// CHECK-NEXT: PI Call Begin : piDeviceRelease
// CHECK-NEXT: PI Call Begin : piTearDown
// CHECK-NEXT: xptiTraceFinish: Stream Name = sycl.pi
// CHECK-NEXT: xptiTraceFinish: Stream Name = sycl.pi.debug
// CHECK-NEXT: xptiTraceFinish: Stream Name = sycl
13 changes: 13 additions & 0 deletions SYCL/lit.cfg.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -313,6 +313,19 @@
if find_executable('sycl-ls'):
config.available_features.add('sycl-ls')

# TODO properly set XPTIFW include and runtime dirs
if find_executable('sycl-ls'):
sycl_ls_full_path = find_executable('sycl-ls')
sycl_ls_loc = os.path.dirname(sycl_ls_full_path)
xptifw_lib = ""
if platform.system() == "Linux":
xptifw_lib = os.path.join(sycl_ls_loc, '..', 'lib', 'libxptifw.so')
elif platform.system() == "Windows":
xptifw_lib = os.path.join(sycl_ls_loc, 'xptifw.dll')
if (os.path.exists(xptifw_lib)):
config.available_features.add('xptifw')
config.substitutions.append(('%xptifw_lib', xptifw_lib))

llvm_tools = ["llvm-spirv", "llvm-link"]
for llvm_tool in llvm_tools:
llvm_tool_path = find_executable(llvm_tool)
Expand Down