Skip to content
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

[SYCL] January'21 Release Notes #3055

Merged
merged 6 commits into from
Jan 20, 2021
Merged

[SYCL] January'21 Release Notes #3055

Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
6e2e468
[SYCL] January'21 Release Notes
Jan 14, 2021
8286889
Address review comments.
Jan 15, 2021
93c6b65
Merge branch 'sycl' into jan-2021-rn
Jan 19, 2021
da0dac2
Augment notes
Jan 19, 2021
515af48
Address review comments.
Jan 19, 2021
5d768c1
Fix typo
s-kanaev Jan 19, 2021
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
107 changes: 107 additions & 0 deletions sycl/ReleaseNotes.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,110 @@
# January'21 release notes

Release notes for commit range 5eebd1e4bfce..2ffafb95f887

## New features
### SYCL Compiler
- Added template parameter support for `work_group_size` attributes
[ca9d816234c6] [289230c0b52c]
- Introduced `auto` mode for device code split feature. This mode is the
default one [184d258b902a]
- Enabled support for multiple AOCX device binaries for FPGA [6ea38f0f1f7a]
### SYCL Library
- Implemented [`online_compiler`](doc/extensions/OnlineCompilation/OnlineCompilation.asciidoc)
feature [91122526e74d]
### Documentation
- Added specification for [set kernel cache configuration extension](doc/extensions/IntelGPU/IntelGPUKernelCache.md)
[db62fe7166ec]

## Improvements
### SYCL Compiler
- Updated ESIMD support for split barrier, slm load and dp4a [7f5013901874]
- Disabled swapping dimension values of `WorkGroupAttr` for semantic checks
[7ef84b6fb60c]
- Enabled `clang-offload-deps` tool usage for generating device dependencies
[08a1c00028e5, 237675ba5a16, 84d2eb53e774, c8455167c2fa]
- Switched to linking only needed symbols from fat static libraries on device
side [5c3e5385b35d]
- Made Intel attributes consistent with clang attributes [aeb4de7e4fa5]
### SYCL Library
- Added filtering of partition properties to exclude non-SYCL ones
[7f7006410c1d]
- Fixed memory leaks of global object in the Level-Zero plugins [f58c568]
- Fixed excessive memory allocation for device handles in the OpenCL plugin
[8cfcdb20b06e]
- Added support for
`syc::ext::oneapi::property::buffer::use_pinned_host_memory` buffery property
in LevelZero plugin [0b9a7493a2f1]
- Improved performance of kernels which do not use work-item free functions on
the host device [b83a1a8028a8]
- Switched to retaining events in LevelZero event wait list until command has
completed [c55eb20c2f8c]
- Improved group scan/reduce algorithms performance for CUDA [bd0edfee2908]
- Enabled sub-group loads and stores in CUDA [cfce9965c366]
### Documentation
- Added notice of the `-o` build configuration option in GetStartedGuide
[665e4062a5ff]
- Added requirement to reference E2E tests for the pull-request in the commit
message [965b63c38ca7]

## Bug fixes
### SYCL Compiler
- Eliminated emission of `llvm.used` global variables to output module(s) in
the post-link tool [9b07d114f5c2] [b96e647c82b2]
- Fixed crash on scalar `fptoui` LLVM instruction [d7ca5dafcb3b]
- Fixed detection of free function calls [6cfc3add16e3]
- Put constant initializer list data in non-generic address space
[0f2cf4ddc2fd]
### SYCL Library
- Fixed memory leak in Level-Zero plugin
[d98b61a7bd28] [051e1407a195] [45e1b143f0a9]
- (ESIMD only) Fixed build error "no member named `cl_int2`" [d7f1b53dec62]
- Disabled execution range rounding for ESIMD kernels [5500262254c3]
- Fixed argument printing when tracing plugin interface calls [30db15fac6f6]
- Fixed `as()` function for swizzle vector [70e003acbbfe]
- Fixed synchronization of sub-group and group functions on CUDA [2b6f2cd7ba0a]

## API/ABI breakages
- None

## Known issues
- GlobalWorkOffset is not supported by Level Zero backend [6f9e9a76]
- The code with function pointer is hanging on Level Zero [d384295e]
- User-defined functions with the same name and signature (exact match of
arguments, return type doesn't matter) as of an OpenCL C built-in
function, can lead to Undefined Behavior.
- A DPC++ system that has FPGAs installed does not support multi-process
execution. Creating a context opens the device associated with the context
and places a lock on it for that process. No other process may use that
device. Some queries about the device through device.get_info<>() also
open up the device and lock it to that process since the runtime needs
to query the actual device to obtain that information.
- On Windows, DPC++ compiler enforces using dynamic C++ runtime for
application linked with SYCL library by:
- linking with msvcrt[d].dll when `-fsycl` switch is used;
- emitting an error on attempts to compile a program with static C++ RT
using `-fsycl` and `/MT` or `/MTd`.
That protects you from complicated runtime errors caused by C++ objects
crossing sycl[d].dll boundary and not always handled properly by different
versions of C++ RT used on app and sycl[d].dll sides.
- The format of the object files produced by the compiler can change between
versions. The workaround is to rebuild the application.
- Using `cl::sycl::program` API to refer to a kernel defined in another
translation unit leads to undefined behavior
- Linkage errors with the following message:
`error LNK2005: "bool const std::_Is_integral<bool>" (??$_Is_integral@_N@std@@3_NB) already defined`
can happen when a SYCL application is built using MS Visual Studio 2019
version below 16.3.0 and user specifies `-std=c++14` or `/std:c++14`.
- Employing read sampler for image accessor may result in sporadic issues with
Level Zero plugin/backend [2c50c03]
- Printing internal defines isn't supported on Windows [50628db]
- Group algorithms for MUL/AND/OR/XOR cannot be enabled for group scope due to
SPIR-V limitations, and are not enabled for sub-group scope yet as the
SPIR-V version isn't automatically raised from 1.1 to 1.3 [96da39e]
- We cannot run Dead Argument Elimination for ESIMD since the pointers to SPIR
kernel functions are saved in `!genx.kernels metadata` [cf10351]


# December'20 release notes

Release notes for commit range 5d7e0925..5eebd1e4bfce
Expand Down