Parfor lowering as kernels and dpnp ufunc compilation to kernels #957
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
added 2 commits
March 4, 2023 12:18
- DPEXRTQueue_CreateFromFilterString: to help create a
DPCTLSyclQueueRef object from a filter string.
- DpexrtQueue_SubmitRange: to submit a ranged kernel.
diptorupd
force-pushed
the
feature/dpnp_parfor_v2
branch
3 times, most recently
from
d107b63 to
671393d
Compare
added 5 commits
March 4, 2023 15:18
The code is based on the existing DpexLowerer in
core/passes/lowerer, but refactored and cleaned up a bit.
- Add the ParforLoweringPass
- Utility modules kernel_builder and kernel_launcher
- Dpnp expressions that rely on the __array_ufunc__ method
of DpnpNdArray do not infer the usm_type and device type
of the LHS of the expression.
The pass is an incomplete implementation of an algorithm
that traverses all the basic blocks and checks CFD compliance
and fixes the LHS of all parfors created from dpnp array
expressions and pranges.
- The new DpjitPipeline uses the backported parfor passes
with the new parfor lowerer and parfor compute follows data
legalization pass.
- dpjit decorator now used DpjitCompiler.
diptorupd
force-pushed
the
feature/dpnp_parfor_v2
branch
3 times, most recently
from
f0557a9 to
dc25803
Compare
Added numba_dpex/core/typing/dpnpdecl.py. Added numba_dpex/dpnp_iface/dpnp_ufunc_db.py. Added numba_dpex/dpnp_iface/dpnpimpl.py.
- The commit makes expressions such as A*2, where A is a
dpnp.ndarray work.
- Adds support for dpnp.true_divide.
diptorupd
force-pushed
the
feature/dpnp_parfor_v2
branch
from
548f014 to
2e087a1
Compare
github-actions bot
added a commit
that referenced
this pull request
Mar 6, 2023
Parfor lowering as kernels and dpnp ufunc compilation to kernels 5a2238b
This was referenced Mar 6, 2023
Closed
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The PR is the initial set of changes to support dpnp functions and vector-style expressions in numba-dpex.
A large subset of Numba's Numpy ufuncs that can be lowered to a SPIR-V supporting device and are also supported by dpnp now can be compiled using numba-dpex.
Dpnp offloading follows the compute follows data programming model.
New infrastructure for lowering parfors to kernels
Have you added a test, reproducer or referred to an issue with a reproducer?
Have you tested your changes locally for CPU and GPU devices?
Have you made sure that new changes do not introduce compiler warnings?
If this PR is a work in progress, are you filing the PR as a draft?