# Community Support **URL:** https://numba.discourse.group/c/numba/community-support/7.md [Latest](https://numba.discourse.group/latest.md) · [Categories](https://numba.discourse.group/categories.md) --- ## [About the Community Support category](https://numba.discourse.group/t/about-the-community-support-category/17)
**Author:** [@Siu](https://numba.discourse.group/u/Siu)\ **Replies:** 0
General questions and discussions about using Numba that do not fit other “Support …” categories --- ## [Feature request about supporting Arrow in Numba](https://numba.discourse.group/t/feature-request-about-supporting-arrow-in-numba/1668)
**Author:** [@Kang\_Ge](https://numba.discourse.group/u/Kang_Ge)\ **Replies:** 13\ **Last updated:** [March 30, 2026, 10:42am UTC](https://numba.discourse.group/t/feature-request-about-supporting-arrow-in-numba/1668 "2026-03-30T10:42:02Z")
Hi, guys. I guess we haven’t any specific plan to support arrow-related types natively for numba. At the same time, more and more projects, e.g., Pandas has ArrowStringArray, vaex has supported some arrow formats, Ray h… --- ## [Structured arrays type definition](https://numba.discourse.group/t/structured-arrays-type-definition/3223)
**Author:** [@pauljurczak](https://numba.discourse.group/u/pauljurczak)\ **Replies:** 0\ **Last updated:** [January 28, 2026, 7:03am UTC](https://numba.discourse.group/t/structured-arrays-type-definition/3223 "2026-01-28T07:03:04Z")
Is there a reason for structured array type having to be defined outside Numba function? This snippet: import numpy as np, numba as nb Point = np.dtype(\[('x', 'f4'), ('y', 'f4')\]) @nb.njit def f(): a = np.zeros((2,)… --- ## [Numba indexing faster with unsigned integers](https://numba.discourse.group/t/numba-indexing-faster-with-unsigned-integers/3222)
**Author:** [@ricardoV94](https://numba.discourse.group/u/ricardoV94)\ **Replies:** 1\ **Last updated:** [January 20, 2026, 4:07pm UTC](https://numba.discourse.group/t/numba-indexing-faster-with-unsigned-integers/3222 "2026-01-20T16:07:50Z")
This is probably known, but showed up when re-implementing sparse matrix algorithms. This helper is used in the first part of a sparse matmul. It’s considerably faster if we cast the indices to uint, as I suppose the com… --- ## [Avoid multiple copies of large numpy array in closure?](https://numba.discourse.group/t/avoid-multiple-copies-of-large-numpy-array-in-closure/2017)
**Author:** [@yxqd](https://numba.discourse.group/u/yxqd)\ **Replies:** 3\ **Last updated:** [December 18, 2025, 1:11pm UTC](https://numba.discourse.group/t/avoid-multiple-copies-of-large-numpy-array-in-closure/2017 "2025-12-18T13:11:53Z")
Hi Stan and Siu, Good morning! I am an author of mcvine, a Monte Carlo neutron ray tracing simulation package. We have been using numba to accelerate mcvine: mcvine.acc. We exchanged emails a couple years ago. Thanks … --- ## [Cannot activate SVML](https://numba.discourse.group/t/cannot-activate-svml/3093)
**Author:** [@ricardoV94](https://numba.discourse.group/u/ricardoV94)\ **Replies:** 1\ **Last updated:** [December 11, 2025, 3:15pm UTC](https://numba.discourse.group/t/cannot-activate-svml/3093 "2025-12-11T15:15:55Z")
$ mamba list | grep -e numba -e lib-rt intel-cmplr-lib-rt 2025.3.1 h04fd95a\_760 conda-forge numba 0.63.0 py314h8169c2f\_0 conda-forge $ numba -s | grep SVML \_\_SVML Inf… --- ## [How can I make better use of CPU to speed up np.linspace](https://numba.discourse.group/t/how-can-i-make-better-use-of-cpu-to-speed-up-np-linspace/3086)
**Author:** [@he113-code](https://numba.discourse.group/u/he113-code)\ **Replies:** 1\ **Last updated:** [December 4, 2025, 2:32am UTC](https://numba.discourse.group/t/how-can-i-make-better-use-of-cpu-to-speed-up-np-linspace/3086 "2025-12-04T02:32:40Z")
from numba import njit, prange # from skimage.draw import line\_nd import numpy as np import random def \_round\_safe(coords): if len(coords) \> 1 and coords\[0\] % 1 == 0.5 and coords\[1\] - coords\[0\] == 1: \_round… --- ## [Is there any way to get round the overhead of numba prange with one thread?](https://numba.discourse.group/t/is-there-any-way-to-get-round-the-overhead-of-numba-prange-with-one-thread/3085)
**Author:** [@lesshaste](https://numba.discourse.group/u/lesshaste)\ **Replies:** 6\ **Last updated:** [December 3, 2025, 9:11pm UTC](https://numba.discourse.group/t/is-there-any-way-to-get-round-the-overhead-of-numba-prange-with-one-thread/3085 "2025-12-03T21:11:16Z")
I like to be about to set the number of threads for my code to 1 sometimes. Unfortunately, all the numba code that uses prange then runs more slowly than if I had used range. Is there a way round this? Would a GitHub is… --- ## [Templates in Numba Cuda](https://numba.discourse.group/t/templates-in-numba-cuda/3078)
**Author:** [@pnk](https://numba.discourse.group/u/pnk)\ **Replies:** 4\ **Last updated:** [November 26, 2025, 8:00pm UTC](https://numba.discourse.group/t/templates-in-numba-cuda/3078 "2025-11-26T20:00:42Z")
Some AI chatbots have mentioned a template feature in Numba Cuda, so one could write @cuda.jit(template=True) before the definition of a kernel or device function. Is this just an AI hallucination, or is it supported,… --- ## [Feedback on custom caching strategy in PyTensor](https://numba.discourse.group/t/feedback-on-custom-caching-strategy-in-pytensor/3077)
**Author:** [@ricardoV94](https://numba.discourse.group/u/ricardoV94)\ **Replies:** 5\ **Last updated:** [November 19, 2025, 7:33pm UTC](https://numba.discourse.group/t/feedback-on-custom-caching-strategy-in-pytensor/3077 "2025-11-19T19:33:07Z")
Hi devs! I’m a developer of PyTensor, a lazy computational graph library that can be transpiled into Numba. We’re doing a strong push with our numba backend, with the goal of making it our default backend (the default … --- ## [Numba/LLVM doesn't optimize \`asarray(x).item()\`](https://numba.discourse.group/t/numba-llvm-doesnt-optimize-asarray-x-item/3069)
**Author:** [@ricardoV94](https://numba.discourse.group/u/ricardoV94)\ **Replies:** 1\ **Last updated:** [October 8, 2025, 12:14pm UTC](https://numba.discourse.group/t/numba-llvm-doesnt-optimize-asarray-x-item/3069 "2025-10-08T12:14:20Z")
Here is a MWE: import numba import numpy as np @numba.njit def foo(x): for i in range(100): x = np.asarray(x).item() \* -1 return x @numba.njit def bar(x): for i in range(100): x = x \* -1 … --- ## [Caching redefined functions](https://numba.discourse.group/t/caching-redefined-functions/3057)
**Author:** [@ricardoV94](https://numba.discourse.group/u/ricardoV94)\ **Replies:** 6\ **Last updated:** [October 2, 2025, 11:12am UTC](https://numba.discourse.group/t/caching-redefined-functions/3057 "2025-10-02T11:12:59Z")
In the following example, the cache fails for the first function, but not the second: from numba import njit @njit(cache=True) def incr\_basic(x): return x + 1 def create\_foo(): @njit(cache=True) def incr\_b… --- ## [Re-assigning structref's member doesn't decrease ref count](https://numba.discourse.group/t/re-assigning-structrefs-member-doesnt-decrease-ref-count/3018)
**Author:** [@milton](https://numba.discourse.group/u/milton)\ **Replies:** 5\ **Last updated:** [June 28, 2025, 2:38pm UTC](https://numba.discourse.group/t/re-assigning-structrefs-member-doesnt-decrease-ref-count/3018 "2025-06-28T14:38:54Z")
I have defined StructRef type S with a single member x being 1d array of integers. As we know, in the NRT context, arrays have their own MemInfo structure associated with it that counts references to the memory region m… --- ## [Understanding Initial Execution Delay in Numba Cached Functions](https://numba.discourse.group/t/understanding-initial-execution-delay-in-numba-cached-functions/2783)
**Author:** [@yafare](https://numba.discourse.group/u/yafare)\ **Replies:** 9\ **Last updated:** [June 25, 2025, 10:23am UTC](https://numba.discourse.group/t/understanding-initial-execution-delay-in-numba-cached-functions/2783 "2025-06-25T10:23:19Z")
I have a question about the cache loading mechanism in Numba. Here’s a piece of test code I wrote: import numba as nb @nb.njit(cache=True) def test\_jit(): a = 0 for i in range(1000000): a += i retu… --- ## [Parallel Computing with Numba across Nodes](https://numba.discourse.group/t/parallel-computing-with-numba-across-nodes/3002)
**Author:** [@ydx22](https://numba.discourse.group/u/ydx22)\ **Replies:** 2\ **Last updated:** [June 9, 2025, 1:01am UTC](https://numba.discourse.group/t/parallel-computing-with-numba-across-nodes/3002 "2025-06-09T01:01:49Z")
I’m currently using Numba version 0.60.0 on a supercomputer running the CentOS operating system. Each node has 56 cores. I want Numba to utilize more cores for parallel computing with prange. So, I set the environment v… --- ## [Can a "nogil" configged in the outter-most function applys to all the routines called inside?](https://numba.discourse.group/t/can-a-nogil-configged-in-the-outter-most-function-applys-to-all-the-routines-called-inside/2785)
**Author:** [@victor](https://numba.discourse.group/u/victor)\ **Replies:** 1\ **Last updated:** [May 29, 2025, 2:01pm UTC](https://numba.discourse.group/t/can-a-nogil-configged-in-the-outter-most-function-applys-to-all-the-routines-called-inside/2785 "2025-05-29T14:01:37Z")
When a function is jitted with option nogil=True, will all the routines it calls automatically release the gil? from numba import njit from numba.experimental import jitclass @njit def inner(...) ... # not explicitly … --- ## [JIT within multiprocessing or threading: what happens?](https://numba.discourse.group/t/jit-within-multiprocessing-or-threading-what-happens/2996)
**Author:** [@mostsquares](https://numba.discourse.group/u/mostsquares)\ **Replies:** 1\ **Last updated:** [May 18, 2025, 4:04am UTC](https://numba.discourse.group/t/jit-within-multiprocessing-or-threading-what-happens/2996 "2025-05-18T04:04:11Z")
I’m curious about what happens when @jit/@njit functions are called from multiple processes or threads. Does each process JIT the function separately? Is the answer the same in threading? Further, if cache=True, is the c… --- ## [How to set type (precision) of operations inside a function?](https://numba.discourse.group/t/how-to-set-type-precision-of-operations-inside-a-function/2989)
**Author:** [@pauljurczak](https://numba.discourse.group/u/pauljurczak)\ **Replies:** 4\ **Last updated:** [April 20, 2025, 11:24pm UTC](https://numba.discourse.group/t/how-to-set-type-precision-of-operations-inside-a-function/2989 "2025-04-20T23:24:49Z")
This function: @nb.njit('(f4)(f4)', fastmath=True, inline='always') def f(a): return 42\*a uses float64 for internal operations and casts it back to the requested float32: def f(a): # --- LINE 8 --- # $const4… --- ## [Different behavior of code with np.nan when jitted](https://numba.discourse.group/t/different-behavior-of-code-with-np-nan-when-jitted/2988)
**Author:** [@pauljurczak](https://numba.discourse.group/u/pauljurczak)\ **Replies:** 3\ **Last updated:** [April 19, 2025, 11:44pm UTC](https://numba.discourse.group/t/different-behavior-of-code-with-np-nan-when-jitted/2988 "2025-04-19T23:44:30Z")
This script: import numpy as np, numba as nb @nb.njit(fastmath=True) def f(): return np.nan #@nb.njit(fastmath=True) def g(): print(f(), np.isnan(f())) g() produces different results: nan False when uncommen… --- ## [How to find my statement that lead to: NumbaTypeSafetyWarning: unsafe cast from int64 to int8. Precision may be lost.?](https://numba.discourse.group/t/how-to-find-my-statement-that-lead-to-numbatypesafetywarning-unsafe-cast-from-int64-to-int8-precision-may-be-lost/1041)
**Author:** [@alatif-alatif](https://numba.discourse.group/u/alatif-alatif)\ **Replies:** 5\ **Last updated:** [April 7, 2025, 4:53pm UTC](https://numba.discourse.group/t/how-to-find-my-statement-that-lead-to-numbatypesafetywarning-unsafe-cast-from-int64-to-int8-precision-may-be-lost/1041 "2025-04-07T16:53:21Z")
Hi, when executing my own code, I get the following warning from numba (numba 0.54.1): .....lib/python3.8/site-packages/numba/typed/typeddict.py:33: NumbaTypeSafetyWarning: unsafe cast from int64 to int8. Precision may… --- ## [Statically linking in dependencies resulting in large memory needs](https://numba.discourse.group/t/statically-linking-in-dependencies-resulting-in-large-memory-needs/2940)
**Author:** [@milton](https://numba.discourse.group/u/milton)\ **Replies:** 10\ **Last updated:** [April 1, 2025, 6:27pm UTC](https://numba.discourse.group/t/statically-linking-in-dependencies-resulting-in-large-memory-needs/2940 "2025-04-01T18:27:01Z")
This is going to be a bit lengthy. In our project, we have a large number of JIT’ed functions performing laborious calculations, and it’s a common occurrence for one JIT’ed function to be calling a number of other JIT’e… --- ## [Numba simple multiplication giving incorrect result](https://numba.discourse.group/t/numba-simple-multiplication-giving-incorrect-result/2965)
**Author:** [@woot](https://numba.discourse.group/u/woot)\ **Replies:** 3\ **Last updated:** [March 26, 2025, 6:09pm UTC](https://numba.discourse.group/t/numba-simple-multiplication-giving-incorrect-result/2965 "2025-03-26T18:09:45Z")
Hello all! I am new to Numba but decided the best way to learn was to start using it on my existing codebase. I am doing scientific computing with version 0.61.0. I am currently trying to understand why Numba’s jit(nopy… --- ## [Dijkstra on grid](https://numba.discourse.group/t/dijkstra-on-grid/1483)
**Author:** [@etienne87](https://numba.discourse.group/u/etienne87)\ **Replies:** 7\ **Last updated:** [March 22, 2025, 5:34pm UTC](https://numba.discourse.group/t/dijkstra-on-grid/1483 "2025-03-22T17:34:52Z")
Hi, i am using numba to accelerate a dijkstra graph search on 2d grid. The code is pretty slow (~1s for 1000x1000 image). Is there something obvious i could do to speed this up? import heapq import numpy as np from num… --- ## [How to call into LLVM-IR code that has multiple functions and global variables](https://numba.discourse.group/t/how-to-call-into-llvm-ir-code-that-has-multiple-functions-and-global-variables/2947)
**Author:** [@max9111](https://numba.discourse.group/u/max9111)\ **Replies:** 1\ **Last updated:** [March 10, 2025, 3:28pm UTC](https://numba.discourse.group/t/how-to-call-into-llvm-ir-code-that-has-multiple-functions-and-global-variables/2947 "2025-03-10T15:28:57Z")
I just wanted to try to call multipe functions from LLMIR, which interact through a global variable (in this case the function pointer of this scipy.special function which has to be initiated at runtime). This is just a… --- ## [Support for free-threaded python](https://numba.discourse.group/t/support-for-free-threaded-python/2916)
**Author:** [@khan](https://numba.discourse.group/u/khan)\ **Replies:** 1\ **Last updated:** [February 18, 2025, 4:13pm UTC](https://numba.discourse.group/t/support-for-free-threaded-python/2916 "2025-02-18T16:13:58Z")
Hello, I was wondering whether there are any plans for releasing free-threaded wheels for numba & llvmlite. I saw this comment but couldn’t find any other related issue. It would be nice to have an issue to track suppor… --- ## [Understanding Sliding Dot Product Performance](https://numba.discourse.group/t/understanding-sliding-dot-product-performance/2912)
**Author:** [@seanlaw](https://numba.discourse.group/u/seanlaw)\ **Replies:** 6\ **Last updated:** [February 12, 2025, 9:36pm UTC](https://numba.discourse.group/t/understanding-sliding-dot-product-performance/2912 "2025-02-12T21:36:03Z")
When computing a sliding dot product between two 1D arrays, Q and T, one can naively do: def naive\_sliding\_dot\_product(Q, T): m = len(Q) l = len(T) - m + 1 out = np.empty(l) for i in range(l): ou… --- ## [Are nested structarrays valid in numba?](https://numba.discourse.group/t/are-nested-structarrays-valid-in-numba/2897)
**Author:** [@adrian.gralak95](https://numba.discourse.group/u/adrian.gralak95)\ **Replies:** 2\ **Last updated:** [February 6, 2025, 7:56am UTC](https://numba.discourse.group/t/are-nested-structarrays-valid-in-numba/2897 "2025-02-06T07:56:42Z")
Hello friendly people :wink: The documentation seems to be a bit unclear about this, so thought I’ll double check here: base\_dtype = np.dtype(\[(f'x', np.uint32), (f'y', np.uint16), (f'z', np.uint8)\]) n\_types = 5 comple… --- ## [Numba for microcontroller such as Cortex-M](https://numba.discourse.group/t/numba-for-microcontroller-such-as-cortex-m/2837)
**Author:** [@guillaume-pais](https://numba.discourse.group/u/guillaume-pais)\ **Replies:** 6\ **Last updated:** [January 24, 2025, 10:41am UTC](https://numba.discourse.group/t/numba-for-microcontroller-such-as-cortex-m/2837 "2025-01-24T10:41:32Z")
Has anyone already tried to use Numba to generate llvm-code for microcontroller? Using a cfunc callback function, it is possible. We were able to integrate the generated LLVM-IR code in a C++ project in a Linux environm… --- ## [Performance is comparable between two different numba cuda approaches](https://numba.discourse.group/t/performance-is-comparable-between-two-different-numba-cuda-approaches/2848)
**Author:** [@myociss](https://numba.discourse.group/u/myociss)\ **Replies:** 6\ **Last updated:** [December 31, 2024, 1:30pm UTC](https://numba.discourse.group/t/performance-is-comparable-between-two-different-numba-cuda-approaches/2848 "2024-12-31T13:30:53Z")
I am implementing a ray tracer using numba cuda and trying to optimize the performance since I have tons of frames to trace. Two approaches are taking about the same amount of time even though it seems that one should be… --- ## [Mangler memory characteristics](https://numba.discourse.group/t/mangler-memory-characteristics/2826)
**Author:** [@nelson2005](https://numba.discourse.group/u/nelson2005)\ **Replies:** 2\ **Last updated:** [December 5, 2024, 6:20pm UTC](https://numba.discourse.group/t/mangler-memory-characteristics/2826 "2024-12-05T18:20:56Z")
Windows 10 Numba 0.56.4 I’ve observed significant memory and disk usage that may be related to mangling. We see mangled type strings that are ~500kB in size and they seem to appear at least ten times in the llvm code,… [Next page](https://numba.discourse.group/c/numba/community-support/7.md?page=1)