Skip to content

Support buffer_location for shared/host USM allocations #141

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 2 commits into from
Jun 9, 2022
Merged

Support buffer_location for shared/host USM allocations #141

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
49d8874
Add new buffer location memory property to aocl_mmd.h
bsyrowik Jun 1, 2022
9ad15f6
Pass buffer location property through OpenCL runtime for USM shared a…
bsyrowik Jun 1, 2022
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
11 changes: 10 additions & 1 deletion include/MMD/aocl_mmd.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -484,7 +484,16 @@ typedef enum {
* memory bank. It is invalid to specify this property without also specifying
* AOCL_MMD_GLOBAL_MEMORY_INTERFACE.
*/
AOCL_MMD_MEM_PROPERTIES_MEMORY_BANK
AOCL_MMD_MEM_PROPERTIES_MEMORY_BANK,
/**
* Specifies the buffer location used for USM API calls. The buffer location
* is an integer index that corresponds to a global memory location in the
* board_spec.xml file. Similar to AOCL_MMD_MEM_PROPERTIES_GLOBAL_MEMORY.
* The buffer location value may not have been validated by the OpenCL
* runtime, so it is the responsibility of the MMD layer to ensure the value
* is valid.
*/
AOCL_MMD_MEM_PROPERTIES_BUFFER_LOCATION
} aocl_mmd_mem_properties_t;

/**
Expand Down
3 changes: 2 additions & 1 deletion include/acl_hal.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -230,7 +230,8 @@ typedef struct {

/// Allocates USM host memory
void *(*host_alloc)(const std::vector<cl_device_id> devices, size_t size,
size_t alignment, int *error);
size_t alignment, mem_properties_t *properties,
int *error);

/// Frees allocated memory by the MMD
int (*free)(cl_context context, void *ptr);
Expand Down
13 changes: 8 additions & 5 deletions src/acl_hal_mmd.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -65,7 +65,8 @@ void *acl_hal_mmd_shared_alloc(cl_device_id device, size_t size,
size_t alignment, mem_properties_t *properties,
int *error);
void *acl_hal_mmd_host_alloc(const std::vector<cl_device_id> devices,
size_t size, size_t alignment, int *error);
size_t size, size_t alignment,
mem_properties_t *properties, int *error);
int acl_hal_mmd_free(cl_context context, void *mem);
int acl_hal_mmd_try_devices(cl_uint num_devices, const cl_device_id *devices,
cl_platform_id platform);
Expand Down Expand Up @@ -2519,7 +2520,7 @@ void *acl_hal_mmd_legacy_shared_alloc(cl_context context, size_t size,
int error = 0;
std::vector<cl_device_id> devices = std::vector<cl_device_id>(
context->device, context->device + context->num_devices);
void *mem = acl_hal_mmd_host_alloc(devices, size, 0, &error);
void *mem = acl_hal_mmd_host_alloc(devices, size, 0, nullptr, &error);
device_ptr_out = static_cast<unsigned long long *>(mem);
if (error) {
switch (error) {
Expand Down Expand Up @@ -2728,7 +2729,8 @@ void *acl_hal_mmd_shared_alloc(cl_device_id device, size_t size,
}

void *acl_hal_mmd_host_alloc(const std::vector<cl_device_id> devices,
size_t size, size_t alignment, int *error) {
size_t size, size_t alignment,
mem_properties_t *properties, int *error) {
// Note we do not support devices in the same context with different MMDs
// Safe to get the mmd handle from first device
void *result = NULL;
Expand All @@ -2749,8 +2751,9 @@ void *acl_hal_mmd_host_alloc(const std::vector<cl_device_id> devices,
handles[i] = device_info[physical_device_id].handle;
}

result = dispatch->aocl_mmd_host_alloc(handles, devices.size(), size,
alignment, nullptr, error);
result = dispatch->aocl_mmd_host_alloc(
handles, devices.size(), size, alignment,
(aocl_mmd_mem_properties_t *)properties, error);
acl_delete_arr(handles);

if (error) {
Expand Down
44 changes: 41 additions & 3 deletions src/acl_usm.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,9 +3,11 @@

// System headers.
#include <algorithm>
#include <array>
#include <cassert>
#include <cstdio>
#include <iostream>
#include <optional>
#include <sstream>
#include <string>
#include <unordered_set>
Expand All @@ -28,6 +30,8 @@
#include <acl_usm.h>
#include <acl_util.h>

#include <MMD/aocl_mmd.h>

#ifdef __GNUC__
#pragma GCC visibility push(protected)
#endif
Expand Down Expand Up @@ -99,11 +103,15 @@ CL_API_ENTRY void *CL_API_CALL clHostMemAllocINTEL(
// Iterate over properties.
// The end of the properties list is specified with a zero.
cl_mem_alloc_flags_intel alloc_flags = 0;
std::optional<cl_uint> mem_id;
while (properties != NULL && *properties != 0) {
switch (*properties) {
case CL_MEM_ALLOC_FLAGS_INTEL: {
alloc_flags = *(properties + 1);
} break;
case CL_MEM_ALLOC_BUFFER_LOCATION_INTEL: {
mem_id = (cl_uint) * (properties + 1);
} break;
default: {
UNLOCK_BAIL_INFO(CL_INVALID_PROPERTY, context, "Invalid properties");
}
Expand Down Expand Up @@ -138,6 +146,16 @@ CL_API_ENTRY void *CL_API_CALL clHostMemAllocINTEL(
}

if (acl_get_hal()->host_alloc) {
std::array<mem_properties_t, 3> mmd_properties;
{
auto mmd_properties_it = mmd_properties.begin();
if (mem_id) {
*mmd_properties_it++ = AOCL_MMD_MEM_PROPERTIES_BUFFER_LOCATION;
*mmd_properties_it++ = *mem_id;
}
*mmd_properties_it++ = 0;
}

acl_usm_allocation_t *usm_alloc =
(acl_usm_allocation_t *)acl_malloc(sizeof(acl_usm_allocation_t));

Expand All @@ -146,7 +164,8 @@ CL_API_ENTRY void *CL_API_CALL clHostMemAllocINTEL(
}

int error = 0;
void *mem = acl_get_hal()->host_alloc(devices, size, alignment, &error);
void *mem = acl_get_hal()->host_alloc(devices, size, alignment,
mmd_properties.data(), &error);
if (error) {
acl_free(usm_alloc);
switch (error) {
Expand Down Expand Up @@ -380,6 +399,7 @@ clSharedMemAllocINTEL(cl_context context, cl_device_id device,
// The end of the properties list is specified with a zero.
cl_mem_alloc_flags_intel alloc_flags = 0;
std::unordered_set<unsigned long long> seen_flags;
std::optional<cl_uint> mem_id;
while (properties != NULL && *properties != 0) {
switch (*properties) {
case CL_MEM_ALLOC_FLAGS_INTEL: {
Expand All @@ -396,6 +416,14 @@ clSharedMemAllocINTEL(cl_context context, cl_device_id device,
}
alloc_flags = *(properties + 1);
} break;
case CL_MEM_ALLOC_BUFFER_LOCATION_INTEL: {
if (seen_flags.insert(CL_MEM_ALLOC_BUFFER_LOCATION_INTEL).second ==
false) {
UNLOCK_BAIL_INFO(CL_INVALID_PROPERTY, context,
"Property specified multiple times");
}
mem_id = (cl_uint) * (properties + 1);
} break;
default: {
UNLOCK_BAIL_INFO(CL_INVALID_PROPERTY, context, "Invalid properties");
}
Expand All @@ -404,6 +432,16 @@ clSharedMemAllocINTEL(cl_context context, cl_device_id device,
}

if (acl_get_hal()->shared_alloc) {
std::array<mem_properties_t, 3> mmd_properties;
{
auto mmd_properties_it = mmd_properties.begin();
if (mem_id) {
*mmd_properties_it++ = AOCL_MMD_MEM_PROPERTIES_BUFFER_LOCATION;
*mmd_properties_it++ = *mem_id;
}
*mmd_properties_it++ = 0;
}

acl_usm_allocation_t *usm_alloc =
(acl_usm_allocation_t *)acl_malloc(sizeof(acl_usm_allocation_t));

Expand All @@ -412,8 +450,8 @@ clSharedMemAllocINTEL(cl_context context, cl_device_id device,
}

int error;
void *mem =
acl_get_hal()->shared_alloc(device, size, alignment, nullptr, &error);
void *mem = acl_get_hal()->shared_alloc(device, size, alignment,
mmd_properties.data(), &error);
if (mem == NULL) {
acl_free(usm_alloc);
switch (error) {
Expand Down
10 changes: 4 additions & 6 deletions test/acl_hal_test.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -91,7 +91,8 @@ void *acl_test_hal_shared_alloc(cl_device_id device, size_t size,
size_t alignment, mem_properties_t *properties,
int *error);
void *acl_test_hal_host_alloc(const std::vector<cl_device_id> devices,
size_t size, size_t alignment, int *error);
size_t size, size_t alignment,
mem_properties_t *properties, int *error);
int acl_test_hal_free(cl_context context, void *ptr);

static acl_event_update_callback acltest_hal_event_callback = NULL;
Expand Down Expand Up @@ -661,11 +662,8 @@ void *acl_test_hal_shared_alloc(cl_device_id device, size_t size,
return (void *)0xdeadbeefdeadbeef;
}

void *acl_test_hal_host_alloc(const std::vector<cl_device_id> devices,
size_t size, size_t alignment, int *error) {
size = size;
alignment = alignment;
error = error;
void *acl_test_hal_host_alloc(const std::vector<cl_device_id>, size_t, size_t,
mem_properties_t *, int *) {
return (void *)0xdeadbeefdeadbeef;
}

Expand Down