Skip to content

add tracing for cl_khr_unified_svm #444

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 10 commits into from
Nov 21, 2025
Merged

add tracing for cl_khr_unified_svm #444

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
b35286f
initial support for cl_khr_unified_svm
bashbaug Jan 4, 2025
479f764
Merge branch 'main' into cl_khr_unified_svm
bashbaug Feb 3, 2025
b518350
Merge branch 'main' into cl_khr_unified_svm
bashbaug Mar 21, 2025
a24a0cc
Merge branch 'main' into cl_khr_unified_svm
bashbaug May 15, 2025
ec87bde
Merge branch 'main' into cl_khr_unified_svm
bashbaug Jul 24, 2025
41e8e85
add SVM pointer info tracing
bashbaug Aug 18, 2025
23201b3
add USM capability string logging
bashbaug Sep 23, 2025
6d81d58
leak checking for async SVM frees
bashbaug Sep 27, 2025
4dd50b5
Merge branch 'main' into cl_khr_unified_svm
bashbaug Nov 21, 2025
7681d65
update to the latest unified SVM spec v0.9.0
bashbaug Nov 21, 2025
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
86 changes: 86 additions & 0 deletions intercept/src/cli_ext.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -984,6 +984,92 @@ cl_int CL_API_CALL clGetKernelSuggestedLocalWorkSizeKHR(
#define CL_QUEUE_THROTTLE_MED_KHR (1<<1)
#define CL_QUEUE_THROTTLE_LOW_KHR (1<<2)

///////////////////////////////////////////////////////////////////////////////
// cl_khr_unified_svm

// Note: This implements the proposed extension version 0.9.0.

#define CL_PLATFORM_SVM_TYPE_CAPABILITIES_KHR 0x0909

#define CL_DEVICE_SVM_TYPE_CAPABILITIES_KHR 0x1077

typedef cl_bitfield cl_svm_capabilities_khr;

#define CL_SVM_CAPABILITY_SINGLE_ADDRESS_SPACE_KHR (1 << 0)
#define CL_SVM_CAPABILITY_SYSTEM_ALLOCATED_KHR (1 << 1)
#define CL_SVM_CAPABILITY_DEVICE_OWNED_KHR (1 << 2)
#define CL_SVM_CAPABILITY_DEVICE_UNASSOCIATED_KHR (1 << 3)
#define CL_SVM_CAPABILITY_CONTEXT_ACCESS_KHR (1 << 4)
#define CL_SVM_CAPABILITY_HOST_OWNED_KHR (1 << 5)
#define CL_SVM_CAPABILITY_HOST_READ_KHR (1 << 6)
#define CL_SVM_CAPABILITY_HOST_WRITE_KHR (1 << 7)
#define CL_SVM_CAPABILITY_HOST_MAP_KHR (1 << 8)
#define CL_SVM_CAPABILITY_DEVICE_READ_KHR (1 << 9)
#define CL_SVM_CAPABILITY_DEVICE_WRITE_KHR (1 << 10)
#define CL_SVM_CAPABILITY_DEVICE_ATOMIC_ACCESS_KHR (1 << 11)
#define CL_SVM_CAPABILITY_CONCURRENT_ACCESS_KHR (1 << 12)
#define CL_SVM_CAPABILITY_CONCURRENT_ATOMIC_ACCESS_KHR (1 << 13)
#define CL_SVM_CAPABILITY_INDIRECT_ACCESS_KHR (1 << 14)

typedef cl_properties cl_svm_alloc_properties_khr;
typedef cl_bitfield cl_svm_alloc_access_flags_khr;
typedef cl_properties cl_svm_free_properties_khr;
typedef cl_bitfield cl_svm_free_flags_khr;
typedef cl_uint cl_svm_pointer_info_khr;

#define CL_SVM_ALLOC_ASSOCIATED_DEVICE_HANDLE_KHR 0x2078
#define CL_SVM_ALLOC_ACCESS_FLAGS_KHR 0x2079
#define CL_SVM_ALLOC_ALIGNMENT_KHR 0x207A

#define CL_SVM_ALLOC_ACCESS_HOST_NOREAD_KHR (1 << 0)
#define CL_SVM_ALLOC_ACCESS_HOST_NOWRITE_KHR (1 << 1)
#define CL_SVM_ALLOC_ACCESS_DEVICE_NOREAD_KHR (1 << 8)
#define CL_SVM_ALLOC_ACCESS_DEVICE_NOWRITE_KHR (1 << 9)

#define CL_SVM_INFO_TYPE_INDEX_KHR 0x2088
#define CL_SVM_INFO_CAPABILITIES_KHR 0x2089
#define CL_SVM_INFO_PROPERTIES_KHR 0x208A
#define CL_SVM_INFO_ACCESS_FLAGS_KHR 0x208B
#define CL_SVM_INFO_BASE_PTR_KHR 0x419B
#define CL_SVM_INFO_SIZE_KHR 0x419C
#define CL_SVM_INFO_ASSOCIATED_DEVICE_HANDLE_KHR 0x419D

#define CL_KERNEL_EXEC_INFO_SVM_INDIRECT_ACCESS_KHR 0x11BB

extern CL_API_ENTRY
void* CL_API_CALL clSVMAllocWithPropertiesKHR(
cl_context context,
const cl_svm_alloc_properties_khr* properties,
cl_uint svm_type_index,
size_t size,
cl_int* errcode_ret);

extern CL_API_ENTRY
cl_int CL_API_CALL clSVMFreeWithPropertiesKHR(
cl_context context,
const cl_svm_free_properties_khr* properties,
cl_svm_free_flags_khr flags,
void* ptr);

extern CL_API_ENTRY
cl_int CL_API_CALL clGetSVMPointerInfoKHR(
cl_context context,
cl_device_id device,
const void* ptr,
cl_svm_pointer_info_khr param_name,
size_t param_value_size,
void* param_value,
size_t* param_value_size_ret);

extern CL_API_ENTRY
cl_int CL_API_CALL clGetSVMSuggestedTypeIndexKHR(
cl_context context,
cl_svm_capabilities_khr required_capabilities,
cl_svm_capabilities_khr desired_capabilities,
const cl_svm_alloc_properties_khr* properties,
size_t size,
cl_uint* suggested_svm_type_index);

///////////////////////////////////////////////////////////////////////////////
// cl_ext_atomic_counters

Expand Down
222 changes: 211 additions & 11 deletions intercept/src/dispatch.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -57,7 +57,8 @@ CL_API_ENTRY cl_int CL_API_CALL CLIRN(clGetPlatformIDs)(

if( pIntercept && pIntercept->dispatch().clGetPlatformIDs )
{
LOG_CLINFO();
CACHE_PLATFORM_INFO();
LOG_CL_INFO();

GET_ENQUEUE_COUNTER();
CALL_LOGGING_ENTER();
Expand Down Expand Up @@ -5382,10 +5383,8 @@ CL_API_ENTRY void* CL_API_CALL CLIRN(clGetExtensionFunctionAddress)(

return retVal;
}
else
{
return NULL;
}

return NULL;
}

///////////////////////////////////////////////////////////////////////////////
Expand Down Expand Up @@ -5438,10 +5437,8 @@ CL_API_ENTRY void* CL_API_CALL CLIRN(clGetExtensionFunctionAddressForPlatform)(

return retVal;
}
else
{
return NULL;
}

return NULL;
}

///////////////////////////////////////////////////////////////////////////////
Expand Down Expand Up @@ -6448,8 +6445,10 @@ CL_API_ENTRY cl_int CL_API_CALL CLIRN(clEnqueueSVMFree) (

HOST_PERFORMANCE_TIMING_END();
DEVICE_PERFORMANCE_TIMING_END( command_queue, event );
// TODO: REMOVE_SVM_ALLOCATIONS?
CHECK_ERROR( retVal );
ADD_OBJECT_ALLOCATION( event ? event[0] : NULL );
ADD_POINTER_FREES( pfn_free_func, num_svm_pointers, svm_pointers );
CALL_LOGGING_EXIT_EVENT( retVal, event );
ADD_EVENT( event ? event[0] : NULL );
}
Expand Down Expand Up @@ -8969,6 +8968,207 @@ CL_API_ENTRY cl_int CL_API_CALL clSetKernelArgDevicePointerEXT(
NULL_FUNCTION_POINTER_RETURN_ERROR(CL_INVALID_KERNEL);
}

///////////////////////////////////////////////////////////////////////////////
//
// cl_khr_unified_svm
CL_API_ENTRY void* CL_API_CALL clSVMAllocWithPropertiesKHR(
cl_context context,
const cl_svm_alloc_properties_khr* properties,
cl_uint svm_type_index,
size_t size,
cl_int* errcode_ret)
{
CLIntercept* pIntercept = GetIntercept();

if( pIntercept )
{
const auto& dispatchX = pIntercept->dispatchX(context);
if( dispatchX.clSVMAllocWithPropertiesKHR )
{
GET_ENQUEUE_COUNTER();

std::string propsStr;
std::string typeStr;
if( pIntercept->config().CallLogging )
{
pIntercept->getSVMAllocPropertiesString(
properties,
propsStr );
pIntercept->getSVMTypeIndexCapabilitiesString(
context,
svm_type_index,
typeStr );
}
CALL_LOGGING_ENTER( "context = %p, properties = [ %s ], svm_type_index = %s (%u), size = %zu",
context,
propsStr.c_str(),
typeStr.c_str(),
svm_type_index,
size );
CHECK_ERROR_INIT( errcode_ret );
HOST_PERFORMANCE_TIMING_START();

void* retVal = dispatchX.clSVMAllocWithPropertiesKHR(
context,
properties,
svm_type_index,
size,
errcode_ret );

HOST_PERFORMANCE_TIMING_END();
ADD_SVM_ALLOCATION( retVal, size ); // TODO: Should this be SVM or USM?
CHECK_ERROR( errcode_ret[0] );
ADD_POINTER_ALLOCATION( retVal );
CALL_LOGGING_EXIT( errcode_ret[0], "returned %p", retVal );

return retVal;
}
}

NULL_FUNCTION_POINTER_SET_ERROR_RETURN_NULL(errcode_ret, CL_INVALID_CONTEXT);
}

///////////////////////////////////////////////////////////////////////////////
//
// cl_khr_unified_svm
CL_API_ENTRY cl_int CL_API_CALL clSVMFreeWithPropertiesKHR(
cl_context context,
const cl_svm_free_properties_khr* properties,
cl_svm_free_flags_khr flags,
void* ptr)
{
CLIntercept* pIntercept = GetIntercept();

if( pIntercept )
{
const auto& dispatchX = pIntercept->dispatchX(context);
if( dispatchX.clSVMFreeWithPropertiesKHR )
{
GET_ENQUEUE_COUNTER();
CALL_LOGGING_ENTER( "context = %p, ptr = %p",
context,
ptr );
HOST_PERFORMANCE_TIMING_START();

cl_int retVal = dispatchX.clSVMFreeWithPropertiesKHR(
context,
properties,
flags,
ptr );

HOST_PERFORMANCE_TIMING_END();
REMOVE_SVM_ALLOCATION( ptr ); // TODO: Should this be SVM or USM?
CHECK_ERROR( retVal );
ADD_POINTER_FREE( ptr );
CALL_LOGGING_EXIT( retVal );

return retVal;
}
}

NULL_FUNCTION_POINTER_RETURN_ERROR(CL_INVALID_CONTEXT);
}

///////////////////////////////////////////////////////////////////////////////
//
// cl_khr_unified_svm
CL_API_ENTRY cl_int CL_API_CALL clGetSVMPointerInfoKHR(
cl_context context,
cl_device_id device,
const void* ptr,
cl_svm_pointer_info_khr param_name,
size_t param_value_size,
void* param_value,
size_t* param_value_size_ret)
{
CLIntercept* pIntercept = GetIntercept();

if( pIntercept )
{
const auto& dispatchX = pIntercept->dispatchX(context);
if( dispatchX.clGetSVMPointerInfoKHR )
{
GET_ENQUEUE_COUNTER();
CALL_LOGGING_ENTER( "context = %p, ptr = %p, param_name = %s (%08X)",
context,
ptr,
pIntercept->enumName().name( param_name ).c_str(),
param_name );
HOST_PERFORMANCE_TIMING_START();

cl_int retVal = dispatchX.clGetSVMPointerInfoKHR(
context,
device,
ptr,
param_name,
param_value_size,
param_value,
param_value_size_ret );

HOST_PERFORMANCE_TIMING_END();
CHECK_ERROR( retVal );
CALL_LOGGING_EXIT( retVal );

return retVal;
}
}

NULL_FUNCTION_POINTER_RETURN_ERROR(CL_INVALID_CONTEXT);
}

///////////////////////////////////////////////////////////////////////////////
//
// cl_khr_unified_svm
CL_API_ENTRY cl_int CL_API_CALL clGetSVMSuggestedTypeIndexKHR(
cl_context context,
cl_svm_capabilities_khr required_capabilities,
cl_svm_capabilities_khr desired_capabilities,
const cl_svm_alloc_properties_khr* properties,
size_t size,
cl_uint* suggested_svm_type_index)
{
CLIntercept* pIntercept = GetIntercept();

if( pIntercept )
{
const auto& dispatchX = pIntercept->dispatchX(context);
if( dispatchX.clGetSVMSuggestedTypeIndexKHR )
{
GET_ENQUEUE_COUNTER();
std::string propsStr;
if( pIntercept->config().CallLogging )
{
pIntercept->getSVMAllocPropertiesString(
properties,
propsStr );
}
CALL_LOGGING_ENTER( "context = %p, required_capabilities = %llX, desired_capabilities = %llX, properties = [ %s ], size = %zu",
context,
required_capabilities,
desired_capabilities,
propsStr.c_str(),
size );
HOST_PERFORMANCE_TIMING_START();

cl_int retVal = dispatchX.clGetSVMSuggestedTypeIndexKHR(
context,
required_capabilities,
desired_capabilities,
properties,
size,
suggested_svm_type_index );

HOST_PERFORMANCE_TIMING_END();
CHECK_ERROR( retVal );
CALL_LOGGING_EXIT( retVal );

return retVal;
}
}

NULL_FUNCTION_POINTER_RETURN_ERROR(CL_INVALID_CONTEXT);
}

///////////////////////////////////////////////////////////////////////////////
//
// cl_ext_image_requirements_info
Expand Down Expand Up @@ -9806,7 +10006,7 @@ CL_API_ENTRY void* CL_API_CALL clHostMemAllocINTEL(
}
}

return NULL;
NULL_FUNCTION_POINTER_SET_ERROR_RETURN_NULL(errcode_ret, CL_INVALID_CONTEXT);
}

///////////////////////////////////////////////////////////////////////////////
Expand Down Expand Up @@ -9883,7 +10083,7 @@ CL_API_ENTRY void* CL_API_CALL clDeviceMemAllocINTEL(
}
}

return NULL;
NULL_FUNCTION_POINTER_SET_ERROR_RETURN_NULL(errcode_ret, CL_INVALID_CONTEXT);
}

///////////////////////////////////////////////////////////////////////////////
Expand Down
Loading