# Latest

**URL:** https://numba.discourse.group/latest.md

[Latest](https://numba.discourse.group/latest.md) · [Categories](https://numba.discourse.group/categories.md)

---

## [Welcome to Discourse](https://numba.discourse.group/t/welcome-to-discourse/7)

<div class="topic-metadata">

**Author:** [@system](https://numba.discourse.group/u/system)\
**Replies:** 0\
**Last updated:** [June 10, 2020, 5:45am UTC](https://numba.discourse.group/t/welcome-to-discourse/7 "2020-06-10T05:45:57Z")

</div>

For discussion of usage and development of the Numba project.

---

## [How to use scipy.special.polygamma with numba?](https://numba.discourse.group/t/how-to-use-scipy-special-polygamma-with-numba/3271)

<div class="topic-metadata">

**Author:** [@jbdurand](https://numba.discourse.group/u/jbdurand)\
**Replies:** 2\
**Last updated:** [September 21, 2026, 1:50pm UTC](https://numba.discourse.group/t/how-to-use-scipy-special-polygamma-with-numba/3271 "2026-09-21T13:50:44Z")

</div>

Dear everyone, I am trying to use the polygamma and digamma functions in numba. Here is a simple example that fails with polygamma and works with digamma: utils\_numba.py: import numpy as np import numba as nb from …

---

## [Passing a Numba compileable function as a cfunc callback including a Tuple of args on the example of Scipy integrate](https://numba.discourse.group/t/passing-a-numba-compileable-function-as-a-cfunc-callback-including-a-tuple-of-args-on-the-example-of-scipy-integrate/3270)

<div class="topic-metadata">

**Author:** [@max9111](https://numba.discourse.group/u/max9111)\
**Replies:** 2\
**Last updated:** [September 16, 2026, 1:58pm UTC](https://numba.discourse.group/t/passing-a-numba-compileable-function-as-a-cfunc-callback-including-a-tuple-of-args-on-the-example-of-scipy-integrate/3270 "2026-09-16T13:58:24Z")

</div>

I thought about an old answer of mine on https://stackoverflow.com/a/58561573/4045774 where I showed a bit tedicious way how to pass args to scipy integrate using a Numba cfunc. Here I want to put a method for discussio…

---

## [\[ANN\] Numba 0.68.0rc1 and llvmlite 0.50.0rc3](https://numba.discourse.group/t/ann-numba-0-68-0rc1-and-llvmlite-0-50-0rc3/3269)

<div class="topic-metadata">

**Author:** [@esc](https://numba.discourse.group/u/esc)\
**Replies:** 0\
**Last updated:** [September 14, 2026, 9:25am UTC](https://numba.discourse.group/t/ann-numba-0-68-0rc1-and-llvmlite-0-50-0rc3/3269 "2026-09-14T09:25:16Z")

</div>

Hi all, Happy to announce release candidates: Numba 0.68.0rc1 and llvmlite 0.50.0rc1. Importantly, these release candidates bring wheels for Python 3.15. Numba 0.68.0rc1 PyPI: pip install --pre numba==0.68.0rc1 conda: …

---

## [Should we move linux-64 wheels to manylinux\_2\_28 in Numba 0.68?](https://numba.discourse.group/t/should-we-move-linux-64-wheels-to-manylinux-2-28-in-numba-0-68/3267)

<div class="topic-metadata">

**Author:** [@swap357](https://numba.discourse.group/u/swap357)\
**Replies:** 1\
**Last updated:** [August 18, 2026, 4:56pm UTC](https://numba.discourse.group/t/should-we-move-linux-64-wheels-to-manylinux-2-28-in-numba-0-68/3267 "2026-08-18T16:56:56Z")

</div>

I’m proposing to make changes to Numba CI and building all linux-64 PyPI wheels as manylinux\_2\_28 (glibc 2.28) instead of manylinux2014 (glibc 2.17), starting with Numba 0.68 / llvmlite 0.51. 0.67 would be the last manyl…

---

## [\[ANN\] Numba 0.67 and llvmlite 0.49 final --- NumPy 2.5 and Windows ARM64 support](https://numba.discourse.group/t/ann-numba-0-67-and-llvmlite-0-49-final-numpy-2-5-and-windows-arm64-support/3266)

<div class="topic-metadata">

**Author:** [@sklam](https://numba.discourse.group/u/sklam)\
**Replies:** 0\
**Last updated:** [August 12, 2026, 2:17pm UTC](https://numba.discourse.group/t/ann-numba-0-67-and-llvmlite-0-49-final-numpy-2-5-and-windows-arm64-support/3266 "2026-08-12T14:17:46Z")

</div>

Hi all, Numba 0.67.0 and llvmlite 0.49.0 are now released. Numba 0.67.0 Numba now adds support for NumPy 2.5. This release also ships Python 3.14 conda packages and wheels for Windows ARM64. Install from PyPI.org: pi…

---

## [Community discussion Aug 4th 2026: Numba-\>WASM target](https://numba.discourse.group/t/community-discussion-aug-4th-2026-numba-wasm-target/3265)

<div class="topic-metadata">

**Author:** [@sklam](https://numba.discourse.group/u/sklam)\
**Replies:** 0\
**Last updated:** [July 31, 2026, 4:49pm UTC](https://numba.discourse.group/t/community-discussion-aug-4th-2026-numba-wasm-target/3265 "2026-07-31T16:49:31Z")

</div>

In the upcoming community meeting, Anutosh and his colleagues will discuss their work on getting Numba/llvmlite to target WASM. If you are interested in this technical discussion, please join us on August 4th 15:30–16:20…

---

## [Why won't Numba let me define a variable?](https://numba.discourse.group/t/why-wont-numba-let-me-define-a-variable/3263)

<div class="topic-metadata">

**Author:** [@transversalist](https://numba.discourse.group/u/transversalist)\
**Replies:** 3\
**Last updated:** [July 27, 2026, 3:10pm UTC](https://numba.discourse.group/t/why-wont-numba-let-me-define-a-variable/3263 "2026-07-27T15:10:30Z")

</div>

I have this piece of code I want to run faster, but when It defines “num”, it gives an odd error message

---

## [\[ANN\] Numba 0.67rc1 and llvmlite 0.49rc1 --- NumPy 2.5 and Windows ARM64 support](https://numba.discourse.group/t/ann-numba-0-67rc1-and-llvmlite-0-49rc1-numpy-2-5-and-windows-arm64-support/3264)

<div class="topic-metadata">

**Author:** [@sklam](https://numba.discourse.group/u/sklam)\
**Replies:** 1\
**Last updated:** [July 24, 2026, 6:34pm UTC](https://numba.discourse.group/t/ann-numba-0-67rc1-and-llvmlite-0-49rc1-numpy-2-5-and-windows-arm64-support/3264 "2026-07-24T18:34:35Z")

</div>

Hi all, Happy to announce release candidates for Numba 0.67.0rc1 and llvmlite 0.49.0rc1. This is a shorter release cycle to bring NumPy 2.5 support and experimental Windows ARM64 support (Tier 1.5). Numba and llvmlite a…

---

## [Numba does not recognize anything](https://numba.discourse.group/t/numba-does-not-recognize-anything/3262)

<div class="topic-metadata">

**Author:** [@transversalist](https://numba.discourse.group/u/transversalist)\
**Replies:** 1\
**Last updated:** [July 24, 2026, 3:41pm UTC](https://numba.discourse.group/t/numba-does-not-recognize-anything/3262 "2026-07-24T15:41:50Z")

</div>

Whenever I run my code with @jit(nopython=True), no matter what is after it, it says I have incorrect syntax. How do I fix this?

---

## [Why does compile result's mangled LLVM symbol name contain execution-specific counter-generated ID?](https://numba.discourse.group/t/why-does-compile-results-mangled-llvm-symbol-name-contain-execution-specific-counter-generated-id/3260)

<div class="topic-metadata">

**Author:** [@milton](https://numba.discourse.group/u/milton)\
**Replies:** 8\
**Last updated:** [July 9, 2026, 9:20pm UTC](https://numba.discourse.group/t/why-does-compile-results-mangled-llvm-symbol-name-contain-execution-specific-counter-generated-id/3260 "2026-07-09T21:20:42Z")

</div>

Here are a couple of examples of what names numba generates for this or that compile result : from numba import float64, int64, njit from numba.core.registry import CPUDispatcher from numba.core.typing.templates import …

---

## [\[ANN\] Numba 0.66.0 and llvmlite 0.48.0](https://numba.discourse.group/t/ann-numba-0-66-0-and-llvmlite-0-48-0/3261)

<div class="topic-metadata">

**Author:** [@sklam](https://numba.discourse.group/u/sklam)\
**Replies:** 0\
**Last updated:** [July 2, 2026, 8:27pm UTC](https://numba.discourse.group/t/ann-numba-0-66-0-and-llvmlite-0-48-0/3261 "2026-07-02T20:27:18Z")

</div>

Numba 0.66.0 and llvmlite 0.48.0 are now released. Numba now upgrades to llvmlite 0.48 with LLVM 22 and restores sys.monitoring on Python 3.14.4 and later. This release also adds support for typed set containers and mul…

---

## [You can now use an RFFT in Numba](https://numba.discourse.group/t/you-can-now-use-an-rfft-in-numba/3253)

<div class="topic-metadata">

**Author:** [@falseywinchnet](https://numba.discourse.group/u/falseywinchnet)\
**Replies:** 0\
**Last updated:** [June 23, 2026, 5:28pm UTC](https://numba.discourse.group/t/you-can-now-use-an-rfft-in-numba/3253 "2026-06-23T17:28:06Z")

</div>

I’m here to present an efficient, lightweight, directly accessible RFFT library that you can use in numba without objmode. GitHub - falseywinchnet/bfft: Bruun Fast Fourier Transform · GitHub Caveats: Power of two ONLY…

---

## [\[ANN\] Numba 0.66.0rc1 / llvmlite 0.48.0rc1](https://numba.discourse.group/t/ann-numba-0-66-0rc1-llvmlite-0-48-0rc1/3250)

<div class="topic-metadata">

**Author:** [@esc](https://numba.discourse.group/u/esc)\
**Replies:** 8\
**Last updated:** [June 23, 2026, 5:18pm UTC](https://numba.discourse.group/t/ann-numba-0-66-0rc1-llvmlite-0-48-0rc1/3250 "2026-06-23T17:18:57Z")

</div>

Hi all, I’m happy to announce fresh release candidates of Numba and llvmlite. Numba now upgrades to llvmlite 0.48 with LLVM 22 and restores sys.monitoring on Python 3.14.4 and later. Also, adds support for typed set con…

---

## [\[ANN\] Numba 0.66.0rc2](https://numba.discourse.group/t/ann-numba-0-66-0rc2/3252)

<div class="topic-metadata">

**Author:** [@sklam](https://numba.discourse.group/u/sklam)\
**Replies:** 2\
**Last updated:** [June 20, 2026, 6:00am UTC](https://numba.discourse.group/t/ann-numba-0-66-0rc2/3252 "2026-06-20T06:00:24Z")

</div>

The following issues were addressed in the Numba 0.66.0rc2 release: np.mean numerical issue: \`np.mean\` on float32 arrays larger than 2^24 elements diverge from NumPy · Issue #10647 · numba/numba · GitHub slice performa…

---

## [Numba cache issue when running on docker](https://numba.discourse.group/t/numba-cache-issue-when-running-on-docker/3251)

<div class="topic-metadata">

**Author:** [@yangeorget](https://numba.discourse.group/u/yangeorget)\
**Replies:** 1\
**Last updated:** [June 18, 2026, 8:21pm UTC](https://numba.discourse.group/t/numba-cache-issue-when-running-on-docker/3251 "2026-06-18T20:21:27Z")

</div>

Hello, I create a docker image with GitHub Actions and the Dockerfile warms up the cache (the numba cache is stored in the docker image). When I run my docker container on AWS EC2, the cache is not used. This does not…

---

## [Curious performance of fusion of Numba and pyomp](https://numba.discourse.group/t/curious-performance-of-fusion-of-numba-and-pyomp/3249)

<div class="topic-metadata">

**Author:** [@pauljurczak](https://numba.discourse.group/u/pauljurczak)\
**Replies:** 0\
**Last updated:** [June 8, 2026, 3:34am UTC](https://numba.discourse.group/t/curious-performance-of-fusion-of-numba-and-pyomp/3249 "2026-06-08T03:34:15Z")

</div>

These two single-threaded functions have unexpected performance: f1 is faster than f0. 1.6x faster on x86-64 CPU, 2.4x faster on aarch64 CPU. mp.omp\_set\_num\_threads(1) @nb.njit(fastmath=True) def f0(img, w, gMin=16): …

---

## [Significant speedup with pyOMP](https://numba.discourse.group/t/significant-speedup-with-pyomp/3245)

<div class="topic-metadata">

**Author:** [@pauljurczak](https://numba.discourse.group/u/pauljurczak)\
**Replies:** 6\
**Last updated:** [June 6, 2026, 12:55am UTC](https://numba.discourse.group/t/significant-speedup-with-pyomp/3245 "2026-06-06T00:55:57Z")

</div>

This simple function shows over 2x speedup (single threaded) when compiled with pyOMP versus plain Numba: @mp.njit(fastmath=True) def f0(nSteps): step = 1.0/nSteps sum = 0.0 with openmp("parallel for reduction(+:…

---

## [NuCS vs Choco: a pure-Python solver meets a JVM veteran](https://numba.discourse.group/t/nucs-vs-choco-a-pure-python-solver-meets-a-jvm-veteran/3248)

<div class="topic-metadata">

**Author:** [@yangeorget](https://numba.discourse.group/u/yangeorget)\
**Replies:** 0\
**Last updated:** [June 1, 2026, 9:25pm UTC](https://numba.discourse.group/t/nucs-vs-choco-a-pure-python-solver-meets-a-jvm-veteran/3248 "2026-06-01T21:25:29Z")

</div>

NuCS (https://github.com/yangeorget/nucs), a constraint solver written entirely in Python and optimized with Numba, is benchmarked against Choco (https://github.com/chocoteam/choco-solver), a mature Java solver with two …

---

## [Charex - Enhance Numba with NumPy's string operations](https://numba.discourse.group/t/charex-enhance-numba-with-numpys-string-operations/2419)

<div class="topic-metadata">

**Author:** [@nmehran](https://numba.discourse.group/u/nmehran)\
**Replies:** 3\
**Last updated:** [June 1, 2026, 3:11pm UTC](https://numba.discourse.group/t/charex-enhance-numba-with-numpys-string-operations/2419 "2026-06-01T15:11:00Z")

</div>

I present a project potentially useful for the Numba ecosystem: charex. This package is designed to extend Numba’s capabilities by integrating NumPy’s string operations within Numba-optimized functions. Overview charex …

---

## [Strange behavior in binding.parse\_assembly](https://numba.discourse.group/t/strange-behavior-in-binding-parse-assembly/3242)

<div class="topic-metadata">

**Author:** [@Dreamerxic](https://numba.discourse.group/u/Dreamerxic)\
**Replies:** 1\
**Last updated:** [May 19, 2026, 11:21am UTC](https://numba.discourse.group/t/strange-behavior-in-binding-parse-assembly/3242 "2026-05-19T11:21:42Z")

</div>

I have a python str irstr that is a valid LLVM IR module with a struct type definition. When I use m=binding.parse\_assembly(irstr)multiple times, then print(str(m)), the struct type name changes like %struct.PgHdr.0, %st…

---

## [\[ANN\] Numba 0.65.1](https://numba.discourse.group/t/ann-numba-0-65-1/3243)

<div class="topic-metadata">

**Author:** [@esc](https://numba.discourse.group/u/esc)\
**Replies:** 0\
**Last updated:** [April 24, 2026, 9:29am UTC](https://numba.discourse.group/t/ann-numba-0-65-1/3243 "2026-04-24T09:29:06Z")

</div>

Hi all, I’m happy to announce a new patch release of Numba: 0.65.1. This patch release disables JIT sys.monitoring integration on Python 3.14.4 and later (#10538) because it was no longer functioning correctly. The wo…

---

## [\[ANN\] Numba 0.65.0 / llvmlite 0.47.0](https://numba.discourse.group/t/ann-numba-0-65-0-llvmlite-0-47-0/3240)

<div class="topic-metadata">

**Author:** [@esc](https://numba.discourse.group/u/esc)\
**Replies:** 1\
**Last updated:** [April 2, 2026, 12:27pm UTC](https://numba.discourse.group/t/ann-numba-0-65-0-llvmlite-0-47-0/3240 "2026-04-02T12:27:48Z")

</div>

Hi all, I’m happy to announce the new stable release of Numba and llvmlite. The latest release adds support for Python 3.14t (FREE THREADING). Numba 0.65.0 Install from PyPI.org: pip install numba==0.65.0 Install f…

---

## [Accessing numpy fft ufuncs from numba](https://numba.discourse.group/t/accessing-numpy-fft-ufuncs-from-numba/3241)

<div class="topic-metadata">

**Author:** [@sjperkins](https://numba.discourse.group/u/sjperkins)\
**Replies:** 0\
**Last updated:** [April 2, 2026, 10:10am UTC](https://numba.discourse.group/t/accessing-numpy-fft-ufuncs-from-numba/3241 "2026-04-02T10:10:58Z")

</div>

I’ve just released a small package on pypi called numba-ufunc-hooks: GitHub - ratt-ru/numba-ufunc-hooks: Numba hooks for NumPy ufuncs · GitHub It may be of interest as it contains a strategy for accessing internal num…

---

## [Feature request about supporting Arrow in Numba](https://numba.discourse.group/t/feature-request-about-supporting-arrow-in-numba/1668)

<div class="topic-metadata">

**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")

</div>

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…

---

## [\[ANN\] Numba 0.65.0rc1 / llvmlite 0.47.0rc1](https://numba.discourse.group/t/ann-numba-0-65-0rc1-llvmlite-0-47-0rc1/3239)

<div class="topic-metadata">

**Author:** [@esc](https://numba.discourse.group/u/esc)\
**Replies:** 3\
**Last updated:** [March 22, 2026, 3:25pm UTC](https://numba.discourse.group/t/ann-numba-0-65-0rc1-llvmlite-0-47-0rc1/3239 "2026-03-22T15:25:39Z")

</div>

Hi all, I’m happy to announce the new stable release of Numba and llvmlite. The latest release adds support for Python 3.14t (FREE THREADING). Numba 0.65.0rc1 Install from PyPI.org: pip install --pre numba==0.65.0rc1…

---

## [Slowdown when omitting optional keyword arguments](https://numba.discourse.group/t/slowdown-when-omitting-optional-keyword-arguments/3235)

<div class="topic-metadata">

**Author:** [@cchapmanbird](https://numba.discourse.group/u/cchapmanbird)\
**Replies:** 1\
**Last updated:** [March 6, 2026, 11:59am UTC](https://numba.discourse.group/t/slowdown-when-omitting-optional-keyword-arguments/3235 "2026-03-06T11:59:04Z")

</div>

I am finding that function call cost increases significantly when omitting an optional keywrd argument in a function. For example, consider foo = jit(lambda a, b=2: a + b) If this is called without passing the kwarg, i…

---

## [How can fused multiply-add (fma) be enforced in jit-compiled code?](https://numba.discourse.group/t/how-can-fused-multiply-add-fma-be-enforced-in-jit-compiled-code/3232)

<div class="topic-metadata">

**Author:** [@MothNik](https://numba.discourse.group/u/MothNik)\
**Replies:** 5\
**Last updated:** [March 5, 2026, 4:40pm UTC](https://numba.discourse.group/t/how-can-fused-multiply-add-fma-be-enforced-in-jit-compiled-code/3232 "2026-03-05T16:40:19Z")

</div>

This question is about Fused Multiply-Add (fma). It is related to Numba-Issue 10071, but has a slightly different scope. Rather than supporting Python’s relatively new builtin math.fma, which computes a \* b + c in a fus…

---

## [TypeError running numba code](https://numba.discourse.group/t/typeerror-running-numba-code/3234)

<div class="topic-metadata">

**Author:** [@ps\_nithin](https://numba.discourse.group/u/ps_nithin)\
**Replies:** 3\
**Last updated:** [March 4, 2026, 1:50pm UTC](https://numba.discourse.group/t/typeerror-running-numba-code/3234 "2026-03-04T13:50:21Z")

</div>

Hi, Could someone help me fix this bug im getting running numba code. I have shared the code in reply. The error is TypeError: Signature mismatch: 2 argument types given, but function takes 3 arguments But the functio…

---

## [Numba very slow with default arguments depending on which arguments are provided](https://numba.discourse.group/t/numba-very-slow-with-default-arguments-depending-on-which-arguments-are-provided/2309)

<div class="topic-metadata">

**Author:** [@Nin17](https://numba.discourse.group/u/Nin17)\
**Replies:** 1\
**Last updated:** [February 23, 2026, 7:57pm UTC](https://numba.discourse.group/t/numba-very-slow-with-default-arguments-depending-on-which-arguments-are-provided/2309 "2026-02-23T19:57:03Z")

</div>

Crossposting my stackoverflow question, I’m implementing a function for which some arguments are compulsory and some are set as default in numba, however depending on the values set for the default arguments, which order…

[Next page](https://numba.discourse.group/latest.md?page=1)
