@@ -4,6 +4,72 @@ All notable changes to this project will be documented in this file.
4
4
The format is based on [ Keep a Changelog] ( https://keepachangelog.com/en/1.0.0/ ) ,
5
5
and this project adheres to [ Semantic Versioning] ( https://semver.org/spec/v2.0.0.html ) .
6
6
7
+ ## Unreleased
8
+
9
+ ### Added
10
+ * New dpjit decorator supporting dpnp compilation (#887 )
11
+ * Boxing and unboxing functionality for dpnp.ndarray to numba_dpex (#902 )
12
+ * New DpexTarget and dispatcher for compiling dpnp using numba-dpex (#887 )
13
+ * Overload implementation for dpnp.empty (#902 )
14
+ * Overload implementation for dpnp.empty_like, dpnp.zeros_like and
15
+ dpnp.ones_like inside dpjit (#928 )
16
+ * Overload implementation for dpnp.zeros and dpnp.ones inside dpjit (#923 )
17
+ * Compilation and offload support for dpnp vector style expressions using Numba
18
+ parfors (#957 )
19
+ * Compilation of over 70 ufuncs for dpnp inside dpjit (#957 )
20
+ * Backported the split parfor pass from upstream Numba. (#949 )
21
+ * Numba type aliases to numba_dpex. (#851 )
22
+ * Numba prange alias inside numba_dpex. (#957 )
23
+ * New LRU cache for kernels (#804 ) and funcs (#877 )
24
+ * New Range and NdRange classes for kernel submission that follow sycl's range
25
+ and ndrange syntax. (#888 )
26
+ * Monkey pacthes to Numba 0.56.4 to support dpnp ufuncs, allocating dpnp
27
+ arrays (#954 )
28
+ * New config flag (NUMBA_DPEX_DUMP_KERNEL_LLVM) to dump a kernel's
29
+ LLVM IR (#924 )
30
+ * A badge to our gitter chatroom (#919 )
31
+ * A small script to update copyright headers (#917 )
32
+ * A new dpexrt_python extension to support USM allocators for Numba
33
+ NRT_MemInfo (#902 )
34
+ * Updated examples for kernel API demonstrating compute-follows-data programming
35
+ model. (#826 )
36
+
37
+ ### Changed
38
+ * ` CLK_GLOBAL_MEM_FENCE ` and ` CLK_LOCAL_MEM_FENCE ` flags renamed to
39
+ ` GLOBAL_MEM_FENCE ` and ` LOCAL_MEM_FENCE ` . (#844 )
40
+ * Switched from Ubuntu-latest to Ubuntu-20.04 for conda package build (#836 )
41
+ * Rename USMNdArrayType to USMNdArray (#851 )
42
+ * Changes to the Numba type to represent dpnp ndarray typess now renamed to
43
+ DpnpNdarray (#880 )
44
+ * Improved exceptions and user errors (#804 )
45
+ * Updated internal API for kernel interface with improved support for
46
+ ` __sycl_usm_array_interface__ ` protocol (#804 )
47
+ * Pin generated spirv version for kernels to 1.1 (#885 )
48
+ * Rename DpexContext and DpexTypingContext to DpexKernelTarget and
49
+ DpexKernelTypingContext (#887 )
50
+ * Renamed existing dpnp overloads that used stubs to dpnp_stubs_impl.py (#953 )
51
+ * Dpctl version requirement mismatch is now a warning and not an
52
+ ImportError (#925 )
53
+ * Update to versioneer 0.28 (#827 )
54
+ * Update to dpctl 0.14 (#858 )
55
+ * Update linters: black to 23.1.0, isort to 5.12.0 (#900 )
56
+ * License in setup.py to match actual project licensing (#904 )
57
+ ### Fixed
58
+ * Kernel specialization, compute follows data programming model for
59
+ kernels (#804 )
60
+ * Dispatcher/caching rewrite to address performance regression (#912 , #896 )
61
+ * func decorator qualname ambiguation fix (#905 )
62
+
63
+ ### Removed
64
+ * Removes the numpy_usm_shared module from numba_dpex. (#841 )
65
+ * Removes the usage of llvmlite.llvmpy (#932 )
66
+
67
+ ### Deprecated
68
+ * Support for NumPy arrays as kernel arguments (#804 )
69
+ * Kernel argument access specifiers (#804 )
70
+ * Support for dpctl.device_context to launch kernels and njit offloading (#804 )
71
+ * Dpnp overloads using stubs. (#953 )
72
+
7
73
## [ 0.19.0] - 2022-11-21
8
74
9
75
### Added
0 commit comments