AI/rocm-systems
2
0
Garpu 0
Kode Masalah Tarik Permintaan Actions Packages Projects Rilis Wiki Kegiatan

SWDEV-422207 - Track commands for capture

Telusuri Sumber 
- Track all captured commands under a new AccumulateCommand
- Add begin() and end() methods to capture commands
- Explicit TS object now passed to certain methods because
profilingBegin() and profilingEnd() now happen separately and thus can
run into threading issues

Change-Id: I171106bdcad72b057836cb2f3fc398db3533119f
This commit is contained in:
Saleel Kudchadker
2023-10-26 20:06:18 +00:00
orang tua 1338ff37e8
melakukan 40f41f4d0b
8 mengubah file dengan 110 tambahan dan 56 penghapusan
Download Patch File Download Diff File Expand all files Collapse all files
rocclr/platform/command.hpp
+25 -6
Melihat File
@@ -269,7 +269,8 @@ class Command : public Event {
uint32_t commandWaitBits_;
//! Construct a new command of the given OpenCL type.
Command(HostQueue& queue, cl_command_type type, const EventWaitList& eventWaitList = nullWaitList,
Command(HostQueue& queue, cl_command_type type,
const EventWaitList& eventWaitList = nullWaitList,
uint32_t commandWaitBits = 0, const Event* waitingEvent = nullptr);
//! Construct a new command of the given OpenCL type.
@@ -857,7 +858,7 @@ class CopyMemoryCommand : public TwoMemoryArgsCommand {
: TwoMemoryArgsCommand(queue, cmdType, eventWaitList, srcMemory, dstMemory),
srcOrigin_(srcOrigin),
dstOrigin_(dstOrigin),
size_(size),
size_(size),
copyMetadata_(copyMetadata){
// Sanity checks
assert(size.c[0] > 0 && "invalid");
@@ -1222,7 +1223,8 @@ class ExternalSemaphoreCmd : public Command {
public:
ExternalSemaphoreCmd(HostQueue& queue, const void* sem_ptr, uint64_t fence,
ExternalSemaphoreCmdType cmd_type)
: Command::Command(queue, CL_COMMAND_USER), sem_ptr_(sem_ptr), fence_(fence), cmd_type_(cmd_type) {}
: Command::Command(queue, CL_COMMAND_USER), sem_ptr_(sem_ptr), fence_(fence),
cmd_type_(cmd_type) {}
virtual void submit(device::VirtualDevice& device) {
device.submitExternalSemaphoreCmd(*this);
@@ -1239,12 +1241,28 @@ class Marker : public Command {
//! Create a new Marker
Marker(HostQueue& queue, bool userVisible, const EventWaitList& eventWaitList = nullWaitList,
const Event* waitingEvent = nullptr, bool cpu_wait = false)
: Command(queue, userVisible ? CL_COMMAND_MARKER : 0, eventWaitList, 0, waitingEvent) { cpu_wait_ = cpu_wait; }
: Command(queue, userVisible ? CL_COMMAND_MARKER : 0, eventWaitList, 0, waitingEvent)
{
cpu_wait_ = cpu_wait;
}
//! The actual command implementation.
virtual void submit(device::VirtualDevice& device) { device.submitMarker(*this); }
};
class AccumulateCommand : public Command {
public:
//! Create a new Marker
AccumulateCommand(HostQueue& queue, const EventWaitList& eventWaitList = nullWaitList,
const Event* waitingEvent = nullptr)
: Command(queue, CL_COMMAND_TASK, eventWaitList, 0, waitingEvent) {}
//! The command implementation
virtual void submit(device::VirtualDevice& device) {
device.submitAccumulate(*this);
}
};
/*! \brief Maps CL objects created from external ones and syncs the contents (blocking).
*
*/
@@ -1516,8 +1534,9 @@ class SvmFreeMemoryCommand : public Command {
void* userData_; //!< Data passed to user-defined callback
public:
SvmFreeMemoryCommand(HostQueue& queue, const EventWaitList& eventWaitList, uint32_t numSvmPointers,
void** svmPointers, freeCallBack pfnFreeFunc, void* userData)
SvmFreeMemoryCommand(HostQueue& queue, const EventWaitList& eventWaitList,
uint32_t numSvmPointers, void** svmPointers,
freeCallBack pfnFreeFunc, void* userData)
: Command(queue, CL_COMMAND_SVM_FREE, eventWaitList),
//! We copy svmPointers since it can be reused/deallocated after
// command creation