Minutes_2018_07_26
Attendees: Ehsan, Todd, Siu, Stuart, Stan
- #3169: parfors lowering error in
build_gufunc_wrapper
ParallelAccelerator- Need a better error message
- #3164: how to use python / numba / @vectorize non-scalar input
- blocked on ufunc rewrite
- #3163: @jit for generic jitclasses
- Good idea, type inference a little tricky, will be rolled into future redesign of jitclasses
- #3161: Caching for ParallelAccelerator without Threading
- Siu will take a look
- #3159: np.eye() does not support dtype argument, which is not stated in the docs and does not return meaningful error code
- Need better error and updated docs
- #3158: Structured dtypes with multi-dim fields: numba either throws or crashes
- Need strict dtype checking and raising of errors
- Notes in docs about what dtypes
- #3155: parfors hoist set item fail ParallelAccelerator bug
- Related to implied accumulation in expression (
x = x * 2
rather thanx *= 2
) - Can temporarily skip test to get builds to pass again
- Need to fix this more properly either detect and error (use inplace operators for accumulators), or detect and do the right thing
- Related to implied accumulation in expression (
- #3149: calling cuda math functions (eg. sincospi)
- Will advise user
- Need to think about where in numba namespace this should go
- #3146: Storing ctypes c_void_p type result in numpy array: Cannot cast void* to uint64
- Siu will take a look
- #3144: [Question] Would it be possible to support larger step size in
prange
? ParallelAccelerator- On wishlist
- #3141: numba acceleration not working in python-sgp4
- Stuart will add links to likely related bugs
- #3140: Jupyter integration. Output compiler log, diagnostic info, etc..
- Good wishlist ideas, need to see how Jupyter feature ends up
- #3139: Parfors accumulator reuse bug
- PR opened to fix
- #3138: Access the global variable defined in other modules
- Need to fix, not sure effort level yet
-
3168 Py37 bytes output fix.
- Merged this morning
-
3167 In build script activate env before installing.
- Merged this morning
-
3166 [WIP] Objmode with-block
- Seeing lots of places that could benefit from refactoring
-
3165 Add FMA intrinsic support
- Need to review and test in build farm
- Is there CPU support for FMA we should add?
- 3162 Support constant dtype string in nopython mode in functions like numpy.empty.
-
3160 First attempt at parallel diagnostics
- Looking for feedback on utility of output, impl is a bit hacky
- 3153 Fix canonicalize_array_math typing for calls with kw args
-
3152 [WIP] Use cuda driver api to get best blocksize for best occupancy
- Once tested on Volta, ready for review
- 3151 Keep a queue of references to last N deserialized functions. Fixes #3026
- 3148 Remove dead array equal @infer code
- 3145 [WIP] support for np.fill_diagonal
- 3142 Issue3139
- 3137 Fix for issue3103
-
3134 [WIP] Cfunc x86 abi
- Contributor updated patch and tests, need rereview?
-
3132 Adds an ~5 minute guide to Numba.
- Stan posted comments
-
3128 WIP: Fix recipe for jetson tx2/ARM
- Will merge when ready
-
3127 Support for reductions on arrays.
- First pass review, waiting for changes
-
3124 Fix 3119, raise for 0d arrays in reductions
- _
-
3122 WIP: Add inliner to object mode pipeline
- Fix closure inlining in object mode
- Needs test and review
-
3093 [WIP] Singledispatch overload support for cuda array interface.
- Needs review
-
3046 Pairwise sum implementation.
-
3017 Add facility to support with-contexts
-
#2999 Support LowLevelCallable
-
#2983 [WIP] invert mapping b/w binop operators and the operator module
-
#2950 Fix dispatcher to only consider contiguous-ness.
-
#2942 Fix linkage nature (declspec(dllexport)) of some test functions
-
#2894: [WIP] Implement jitclass default constructor arguments.
-
#2817: [WIP] Emit LLVM optimization remarks
===========================
- Experimental python mode blocks
- Refactored threadpool interface
- AMD GPU backend
- Parallel diagnostics
- Usual collection of bug fixes