Skip to content

Device global copy kernel implementation #269

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

Draft
wants to merge 19 commits into
base: main
Choose a base branch
from
Draft

Device global copy kernel implementation #269

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
19 commits
Select commit Hold shift + click to select a range
39556cb
Resolve conflict and improve code style
sherry-yuan Jan 21, 2022
796733c
change kernel arg size to sizeof(void*)
sherry-yuan Jan 28, 2022
f8c394a
temporarily remove release kernel
sherry-yuan Jan 28, 2022
c1e8825
avoid declaraing array with variable length as windows MSVC does not …
sherry-yuan Jan 28, 2022
c6d89c3
fix klocwork issues
sherry-yuan Jan 28, 2022
1cf2d0f
Test both read and write device global
sherry-yuan Feb 4, 2022
309a9d1
use null to indicate optional clean up callback data
sherry-yuan Feb 7, 2022
c047c76
Resolve conflict and improve code style
sherry-yuan Feb 11, 2022
0185cf4
Resolve conflict and improve code style
sherry-yuan Feb 12, 2022
d8e8786
Add clear functional documentation and line comments
sherry-yuan Mar 11, 2022
b8773ad
Fix klocwork and remove commented code.
sherry-yuan Apr 1, 2022
aad07f9
Resolve conflict and improve code style
intel-liudean Feb 3, 2023
f2d0533
Modify code to pass ninja and unittest
intel-liudean Feb 6, 2023
7dd6a34
Fix style and comment
intel-liudean Feb 7, 2023
953e9c0
Clang format my change
intel-liudean Feb 7, 2023
1d3e29b
Clang format again
intel-liudean Feb 8, 2023
625faa1
Improve style
intel-liudean Feb 8, 2023
3e72c5a
Add line break for clamg format
intel-liudean Feb 8, 2023
6837a26
Clang format
intel-liudean Feb 8, 2023
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
Prev Previous commit
Next Next commit
Resolve conflict and improve code style
  • Loading branch information
@sherry-yuan @intel-liudean
sherry-yuan authored and intel-liudean committed Feb 3, 2023
commit c047c764a7740d1d319772867dc65da686cda88a
2 changes: 2 additions & 0 deletions include/acl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -248,6 +248,8 @@ typedef struct {

bool streaming_control_info_available;
acl_streaming_kernel_control_info streaming_control_info;
unsigned int device_global_address; /* Address of kernel's device global*/
unsigned int device_global_size; /* Size of address space of device global used by this kernel*/
} acl_accel_def_t;

/* An ACL system definition.
Expand Down
6 changes: 4 additions & 2 deletions src/acl_mem.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -424,8 +424,10 @@ CL_API_ENTRY cl_int clEnqueueReadGlobalVariableINTEL(
}

// dev_addr_t dev_global_address =
// kernel->dev_bin->get_devdef().autodiscovery_def.?
uintptr_t dev_global_address = 0x4000000;
uintptr_t dev_global_address = kernel->accel_def->device_global_address;
assert(kernel->accel_def->device_global_address == 4096); // TODO: remove when merging
// uintptr_t dev_global_address = 0x4000000;
// TODO: add checks for whether the copy will be out of bound for device global
void *dev_global_ptr =
(void *)(dev_global_address + offset * 8); // 1 unit of offset is 8 bits
status = set_kernel_arg_mem_pointer_without_checks(kernel, 0, dev_global_ptr);
Expand Down
3 changes: 3 additions & 0 deletions test/acl_auto_configure_test.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -269,6 +269,9 @@ TEST(auto_configure, simple) {
CHECK_EQUAL(0,
(int)m_device_def.autodiscovery_def.accel[0].max_work_group_size);
CHECK_EQUAL(1, (int)m_device_def.autodiscovery_def.accel[0].is_sycl_compile);
CHECK_EQUAL(4096, (int)m_device_def.autodiscovery_def.accel[0].device_global_address);
CHECK_EQUAL(2048, (int)m_device_def.autodiscovery_def.accel[0].device_global_size);


// Checks for device global entry.
CHECK_EQUAL(2, m_device_def.autodiscovery_def.device_global_mem_defs.size());
Expand Down
33 changes: 19 additions & 14 deletions test/acl_globals_test.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -198,20 +198,25 @@ static std::vector<acl_accel_def_t> acltest_complex_system_device0_accel = {
{},
{32768, 0, 0},
1},
{14,
ACL_RANGE_FROM_ARRAY(acltest_devicelocal[11]),
acltest_kernels[14],
acltest_laspace_info,
{0, 0, 0},
0,
0,
1,
0,
32768,
3,
{},
{32768, 0, 0},
1},
{14, // id
ACL_RANGE_FROM_ARRAY(acltest_devicelocal[11]), // mem
acltest_kernels[14], // iface
acltest_laspace_info, // local_aspaces
{0, 0, 0}, // compile_work_group_size
0, // is_workgroup_invariant
0, // is_workitem_invariant
1, // num_vector_lanes
0, // profiling_words_to_readback
32768, // max_work_group_size
3, // max_global_work_dim
{}, // printf_format_info
{32768, 0, 0}, // max_work_group_size_arr
1, // uses_global_work_offset
0, // fast_launch_depth
1, // is_sycl_compile
4096, // device_global_address
2048, // device_global_size
},
{1,
ACL_RANGE_FROM_ARRAY(acltest_devicelocal[1]),
acltest_kernels[1],
Expand Down