Cold-Air-Networks/mirror-llvm-project
2
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 7

[OMPT] Add callback for omp_target_memset calls (#194168)

Browse source 
OpenMP v6.0 added new enum values to `ompt_callback_target_data_op_t`
for communicating `memset` events. Add a corresponding callback
invokation to `omp_target_memset` calls.

Expand the tests to ensure correct behavior.

Signed-off-by: Jan André Reuter <j.reuter@fz-juelich.de>
This commit is contained in:
Jan André Reuter 2026-05-28 17:52:04 +02:00 committed by GitHub
parent ea4cd9228f
commit 2c8ca9679e
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
7 changed files with 279 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

1
clang/docs/OpenMPSupport.rst
View file

@ -621,6 +621,7 @@ implementation.
| OMPT: ompt_target_data_transfer(_async) | :part:`partial` | :good:`N/A` | Enum: https://github.com/llvm/llvm-project/pull/195829 |
+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
| OMPT: ompt_target_data_memset(_async) | :part:`partial` | :good:`N/A` | Enum: https://github.com/llvm/llvm-project/pull/195829 |
| | | | Callbacks: https://github.com/llvm/llvm-project/pull/194168 |
+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
| OMPT: workdistribute work callback enum | :part:`partial` | :good:`N/A` | Enum: https://github.com/llvm/llvm-project/pull/195829 |
+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

13
offload/include/OpenMP/OMPT/Interface.h
View file

@ -126,6 +126,14 @@ public:
void endTargetDisassociatePointer(int64_t DeviceId, void *HstPtrBegin,
void *TgtPtrBegin, size_t Size, void *Code);
/// Top-level function for invoking callback before target memset API
void beginTargetMemset(int64_t DeviceId, void *HostPtrBegin,
void *TgtPtrBegin, size_t Size, void *Code);
/// Top-level function for invoking callback after target memset API
void endTargetMemset(int64_t DeviceId, void *HostPtrBegin, void *TgtPtrBegin,
size_t Size, void *Code);
// Target kernel callbacks
/// Top-level function for invoking callback before target construct
@ -166,6 +174,11 @@ public:
std::mem_fn(&Interface::beginTargetDisassociatePointer),
std::mem_fn(&Interface::endTargetDisassociatePointer));
if constexpr (OpType == ompt_target_data_memset ||
OpType == ompt_target_data_memset_async)
return std::make_pair(std::mem_fn(&Interface::beginTargetMemset),
std::mem_fn(&Interface::endTargetMemset));
llvm_unreachable("Unhandled target data operation type!");
}

5
offload/libomptarget/OpenMP/API.cpp
View file

@ -477,6 +477,11 @@ EXTERN void *omp_target_memset(void *Ptr, int ByteVal, size_t NumBytes,
ODBG(ODT_Interface) << "filling memory on host via memset";
memset(Ptr, ByteVal, NumBytes); // ignore return value, memset() cannot fail
} else {
OMPT_IF_BUILT(InterfaceRAII TargetMemsetRAII(
RegionInterface.getCallbacks<ompt_target_data_memset>(), DeviceNum,
nullptr, const_cast<void *>(Ptr), NumBytes,
__builtin_return_address(0)));
// TODO: replace the omp_target_memset() slow path with the fast path.
// That will require the ability to execute a kernel from within
// libomptarget.so (which we do not have at the moment).

27
offload/libomptarget/OpenMP/OMPT/Callback.cpp
View file

@ -387,6 +387,33 @@ void Interface::endTargetDisassociatePointer(int64_t DeviceId,
}
}
void Interface::beginTargetMemset(int64_t DeviceId, void *HostPtrBegin,
void *TgtPtrBegin, size_t Size, void *Code) {
beginTargetDataOperation();
if (ompt_callback_target_data_op_emi_fn) {
ompt_callback_target_data_op_emi_fn(
ompt_scope_begin, TargetTaskData, &TargetData, &HostOpId,
ompt_target_data_memset, HostPtrBegin, omp_initial_device, TgtPtrBegin,
DeviceId, Size, Code);
} else if (ompt_callback_target_data_op_fn) {
HostOpId = createOpId();
ompt_callback_target_data_op_fn(
TargetData.value, HostOpId, ompt_target_data_memset, HostPtrBegin,
omp_initial_device, TgtPtrBegin, DeviceId, Size, Code);
}
}
void Interface::endTargetMemset(int64_t DeviceId, void *HostPtrBegin,
void *TgtPtrBegin, size_t Size, void *Code) {
if (ompt_callback_target_data_op_emi_fn) {
ompt_callback_target_data_op_emi_fn(
ompt_scope_end, TargetTaskData, &TargetData, &HostOpId,
ompt_target_data_memset, HostPtrBegin, omp_initial_device, TgtPtrBegin,
DeviceId, Size, Code);
}
endTargetDataOperation();
}
void Interface::beginTarget(int64_t DeviceId, void *Code) {
beginTargetRegion();
if (ompt_callback_target_emi_fn) {

76
offload/test/ompt/target_memset.c Normal file
View file

@ -0,0 +1,76 @@
// clang-format off
// RUN: %libomptarget-compile-run-and-check-generic
// REQUIRES: ompt
// REQUIRES: gpu
// clang-format on
/*
* Verify that for the target OpenMP APIs, the return address is non-null and
* distinct.
*/
#include <omp.h>
#include <stdlib.h>
#include "callbacks.h"
#include "register_non_emi.h"
int main() {
int d = omp_get_default_device();
int id = omp_get_initial_device();
int q[128], i;
void *p;
void *result;
if (d < 0 || d >= omp_get_num_devices())
d = id;
p = omp_target_alloc(130 * sizeof(int), d);
if (p == NULL)
return 0;
for (i = 0; i < 128; i++)
q[i] = i;
result = omp_target_memset(p, 0, 130 * sizeof(int), d);
if (result != p) {
abort();
}
int q2[128];
for (i = 0; i < 128; ++i)
q2[i] = i;
if (omp_target_memcpy_async(q2, p, 128 * sizeof(int), 0, sizeof(int), id, d,
0, NULL))
abort();
#pragma omp taskwait
for (i = 0; i < 128; ++i)
if (q2[i] != 0)
abort();
omp_target_free(p, d);
return 0;
}
// clang-format off
/// CHECK: Callback DataOp: target_id=[[TARGET_ID:[0-9]+]] host_op_id=[[HOST_OP_ID:[0-9]+]] optype=ompt_target_data_alloc
/// CHECK-SAME: src_device_num=[[HOST:-1]]
/// CHECK-SAME: dest_device_num=[[DEVICE:[0-9]+]]
/// CHECK-NOT: code=(nil)
/// CHECK: code=[[CODE1:0x[0-f]+]]
/// CHECK: Callback DataOp: target_id=[[TARGET_ID:[0-9]+]] host_op_id=[[HOST_OP_ID:[0-9]+]] optype=ompt_target_data_memset
/// CHECK-SAME: src_device_num=[[HOST]] {{.+}} dest_device_num=[[DEVICE]]
/// CHECK-NOT: code=(nil)
/// CHECK-NOT: code=[[CODE1]]
/// CHECK: code=[[CODE2:0x[0-f]+]]
/// CHECK: Callback DataOp: target_id=[[TARGET_ID:[0-9]+]] host_op_id=[[HOST_OP_ID:[0-9]+]] optype=ompt_target_data_transfer_from_device
/// CHECK-SAME: src_device_num=[[DEVICE]] {{.+}} dest_device_num=[[HOST]]
/// CHECK-NOT: code=(nil)
/// CHECK-NOT: code=[[CODE2]]
/// CHECK: code=[[CODE3:0x[0-f]+]]
/// CHECK: Callback DataOp: target_id=[[TARGET_ID:[0-9]+]] host_op_id=[[HOST_OP_ID:[0-9]+]] optype=ompt_target_data_delete
/// CHECK-NOT: code=(nil)
/// CHECK-NOT: code=[[CODE3]]

79
offload/test/ompt/target_memset_async.c Normal file
View file

@ -0,0 +1,79 @@
// clang-format off
// RUN: %libomptarget-compile-run-and-check-generic
// REQUIRES: ompt
// REQUIRES: gpu
// clang-format on
/*
* Verify that for the target OpenMP APIs, the return address is non-null and
* distinct.
*/
#include <omp.h>
#include <stdlib.h>
#include "callbacks.h"
#include "register_non_emi.h"
int main() {
int d = omp_get_default_device();
int id = omp_get_initial_device();
int q[128], i;
void *p;
void *result;
if (d < 0 || d >= omp_get_num_devices())
d = id;
p = omp_target_alloc(130 * sizeof(int), d);
if (p == NULL)
return 0;
for (i = 0; i < 128; i++)
q[i] = i;
result = omp_target_memset_async(p, 0, 130 * sizeof(int), d, 0, NULL);
#pragma omp taskwait
if (result != p) {
abort();
}
int q2[128];
for (i = 0; i < 128; ++i)
q2[i] = i;
if (omp_target_memcpy_async(q2, p, 128 * sizeof(int), 0, sizeof(int), id, d,
0, NULL))
abort();
#pragma omp taskwait
for (i = 0; i < 128; ++i)
if (q2[i] != 0)
abort();
omp_target_free(p, d);
return 0;
}
// clang-format off
/// CHECK: Callback DataOp: target_id=[[TARGET_ID:[0-9]+]] host_op_id=[[HOST_OP_ID:[0-9]+]] optype=ompt_target_data_alloc
/// CHECK-SAME: src_device_num=[[HOST:-1]]
/// CHECK-SAME: dest_device_num=[[DEVICE:[0-9]+]]
/// CHECK-NOT: code=(nil)
/// CHECK: code=[[CODE1:0x[0-f]+]]
/// CHECK: Callback DataOp: target_id=[[TARGET_ID:[0-9]+]] host_op_id=[[HOST_OP_ID:[0-9]+]] optype=ompt_target_data_memset
/// CHECK-SAME: src_device_num=[[HOST]] {{.+}} dest_device_num=[[DEVICE]]
/// CHECK-NOT: code=(nil)
/// CHECK-NOT: code=[[CODE1]]
/// CHECK: code=[[CODE2:0x[0-f]+]]
/// CHECK: Callback DataOp: target_id=[[TARGET_ID:[0-9]+]] host_op_id=[[HOST_OP_ID:[0-9]+]] optype=ompt_target_data_transfer_from_device
/// CHECK-SAME: src_device_num=[[DEVICE]] {{.+}} dest_device_num=[[HOST]]
/// CHECK-NOT: code=(nil)
/// CHECK-NOT: code=[[CODE2]]
/// CHECK: code=[[CODE3:0x[0-f]+]]
/// CHECK: Callback DataOp: target_id=[[TARGET_ID:[0-9]+]] host_op_id=[[HOST_OP_ID:[0-9]+]] optype=ompt_target_data_delete
/// CHECK-NOT: code=(nil)
/// CHECK-NOT: code=[[CODE3]]

78
offload/test/ompt/target_memset_emi.c Normal file
View file

@ -0,0 +1,78 @@
// clang-format off
// RUN: %libomptarget-compile-run-and-check-generic
// REQUIRES: ompt
// REQUIRES: gpu
// clang-format on
/*
* Verify correct callback sequence for memset API call.
*/
#include <omp.h>
#include <stdio.h>
#include <stdlib.h>
#include "callbacks.h"
#include "register_emi.h"
int main() {
int d = omp_get_default_device();
int id = omp_get_initial_device();
int q[128], i;
void *p;
void *result;
if (d < 0 || d >= omp_get_num_devices())
d = id;
p = omp_target_alloc(130 * sizeof(int), d);
if (p == NULL)
return 0;
for (i = 0; i < 128; i++)
q[i] = i;
result = omp_target_memset(p, 0, 130 * sizeof(int), d);
if (result != p) {
abort();
}
int q2[128];
for (i = 0; i < 128; ++i)
q2[i] = i;
if (omp_target_memcpy_async(q2, p, 128 * sizeof(int), 0, sizeof(int), id, d,
0, NULL))
abort();
#pragma omp taskwait
for (i = 0; i < 128; ++i)
if (q2[i] != 0)
abort();
omp_target_free(p, d);
return 0;
}
// clang-format off
/// CHECK: Callback Init:
/// CHECK: Callback DataOp EMI: endpoint=ompt_scope_begin optype=ompt_target_data_alloc
/// CHECK-SAME: src_device_num=[[HOST:-1]]
/// CHECK-SAME: dest_device_num=[[DEVICE:[0-9]+]]
/// CHECK: Callback DataOp EMI: endpoint=ompt_scope_end optype=ompt_target_data_alloc {{.+}} src_device_num=[[HOST]] {{.+}} dest_device_num=[[DEVICE]]
/// CHECK: Callback DataOp EMI: endpoint=ompt_scope_begin optype=ompt_target_data_memset {{.+}} src_device_num=[[HOST]] {{.+}} dest_device_num=[[DEVICE]]
/// CHECK: Callback DataOp EMI: endpoint=ompt_scope_end optype=ompt_target_data_memset {{.+}} src_device_num=[[HOST]] {{.+}} dest_device_num=[[DEVICE]]
/// CHECK: Callback DataOp EMI: endpoint=ompt_scope_begin optype=ompt_target_data_transfer_from_device {{.+}} src_device_num=[[DEVICE]] {{.+}} dest_device_num=[[HOST]]
/// CHECK: Callback DataOp EMI: endpoint=ompt_scope_end optype=ompt_target_data_transfer_from_device {{.+}} src_device_num=[[DEVICE]] {{.+}} dest_device_num=[[HOST]]
/// CHECK: Callback DataOp EMI: endpoint=ompt_scope_begin optype=ompt_target_data_delete {{.+}} src_device_num=[[DEVICE]]
/// CHECK: Callback DataOp EMI: endpoint=ompt_scope_end optype=ompt_target_data_delete {{.+}} src_device_num=[[DEVICE]]
/// CHECK: Callback Fini:
// clang-format on