Mailman 3 December 2023 - NumPy-Discussion

Re: incomplete BLAS/CBLAS linking (Telling meson build which CBLAS/LAPACK (LAPACKE?) to use via pkgconfig module)
by Ralf Gommers 01 Jan '24

01 Jan '24

(re-sending to list) On Fri, Dec 29, 2023 at 11:34 AM Ralf Gommers <ralf.gommers(a)gmail.com> wrote: > > > On Fri, Dec 29, 2023 at 12:00 AM Dr. Thomas Orgis < > thomas.orgis(a)uni-hamburg.de> wrote: > >> Am Thu, 28 Dec 2023 20:51:27 +0100 >> schrieb Ralf Gommers <ralf.gommers(a)gmail.com>: >> >> > > libblas.so >> > > liblapack.so (NEEDing libblas.so) >> > > libcblas.so (NEEDing libblas.so) >> > > libpapacke.so (NEEDing liblapack.so, hence libblas.so) >> > > >> > > and their respective .pc files. This is the natural order that occus >> to >> > > me when building from netlib upstream. >> > >> > >> > This should work fine. It's auto-detected in NumPy already, and will be >> in >> > SciPy in the future. For now, using `-Dblas=blas -Dlapack=lapack` in the >> > SciPy build should work. >> >> I noticed that with -Dblas=blas, which is in pkgsrc now. The detection >> code sets cblas and finds libcblas by dark magic / defaults that happen >> to match. But what if my setup uses -Dblas=netlib_blas? Then the >> internal guesswork would fail. >> > > If the library name is libcblas.so it will still be found. If it's also a > nonstandard name, then yes it's going to fail. I'd say though that (a) this > isn't a real-world situation as far as we know, (b) just don't do this as a > packager, and (c) if you really must, you can still make it work by > providing a custom `cblas.pc` (see > http://scipy.github.io/devdocs/building/blas_lapack.html#using-pkg-config-t… > ). > > Please consider a mode where the user specifies separate names for all >> 4 components. For package builds, we do not want any guess work, >> including assuming that libblas.so is accompanied by libcblas.so with >> that exact name. >> >> So I'd like >> >> -Dblas=$BLAS_PACKAGE -Dcblas=$CBLAS_PACKAGE \ >> -Dlapack=$LAPACK_PACKAGE -Dlapacke=$LAPACKE_PACKAGE >> >> where the values may all be the same or not. If I fail to provide one >> of those, feel free to guess for the rest (for example, assuming/trying >> that all of those are openblas if I say -Dblas=openblas). > > > We don't use LAPACKE, so that one can be ignored. For CBLAS, I'd honestly > rather get a bug report than add new CLI flags for a situation that seems > to be purely hypothetical. Things work on all known distributions I > believe, and this design isn't new but was the same that numpy.distutils > uses. We can consider a new `-Dcblas` flag at any point, there is nothing > in the design preventing us from adding it later. But I'd rather only do so > if there's a real need. > > >> I also realized that including LAPACK in OpenBLAS is needed, but any >> new BLAS code could start out just replacing the netlib piece by piece. >> The partitioning is there and it is probably good for managing the >> complexity, limiting scope of the individual libraries. >> >> > > Telling the meson build that BLAS is libcblas works as long as >> actually >> > > CBLAS symbols are used. >> > >> > >> > Please never do this. The library is BLAS, so you should use >> `-Dblas=blas` >> > for NumPy. It will find `cblas` just fine that way. >> >> Oh. As I wrote before, we now have >> >> -Csetup-args=-Dblas=${CBLAS_PC} >> -Csetup-args=-Dlapack=${LAPACK_PC} >> >> for math/py-numpy. That's CBLAS_PC, not BLAS_PC. And this works. >> > > I assume that it also passes if you'd pass in BLAS_PC? > > >> >> > This is probably a bug in SciPy. >> >> Well, apparently its just a miscommunication between us two. Scipy is >> fine with >> > > Phew:) I also just confirmed by writing a new SciPy CI job for the split > Netlib BLAS situation, based on how OpenSUSE packages it. And that passes. > > >> >> -Csetup-args=-Dblas=${BLAS_PC} >> -Csetup-args=-Dlapack=${LAPACK_PC} >> >> locating licblas by inferring it from libblas, and finding cblas in >> openblas_foobar, apparently. It prints those lines: >> >> Run-time dependency blas found: YES 3.11.0 >> Run-time dependency cblas found: YES 3.11.0 >> Run-time dependency lapack found: YES 3.11.0 >> blas : blas >> lapack : lapack >> >> While the numpy build does this: >> >> Run-time dependency cblas found: YES 3.11.0 >> Message: BLAS symbol suffix: >> Run-time dependency lapack found: YES 3.11.0 >> blas : cblas >> lapack : lapack >> >> This looks similar to the case of openblas_openmp for -Dblas and -Dlapack: >> >> Run-time dependency openblas_openmp found: YES 0.3.24 >> Message: BLAS symbol suffix: >> Run-time dependency openblas_openmp found: YES 0.3.24 >> blas : openblas_openmp >> lapack : openblas_openmp >> >> So scipy locates cblas based on the name blas, but doesn't really use >> cblas. > > > It does in a few places, like SuperLU. > > >> Numpy is happy with libcblas bringing libblas in and calls it >> blas, but really uses the cblas interface. This looks a bit confusing. >> > > I may be able to add something to the docs, but there should be no > confusion. We need "BLAS with CBLAS symbols". CBLAS should simply not be > considered as a completely separate dependency (it's one library with two > interfaces). I can't think of a reason to do so, nor of a build system that > does it like that. There is no FindCBLAS in CMake for example, it handles > it transparently within FindBLAS. > > >> I guess it makes more sense to continue that discussion on the meson >> PRs for this functionality … as it transcends NumPy, anyway. I hope we >> can settle on something that works for autodetection and prescription >> of all parts. >> > > I'll note that the -D command-line switches are specific to NumPy and > SciPy. Meson only provides a uniform API and the detection methods used > under the hood. Nothing will change in Meson if we decide to add `-Dcblas` > for NumPy and SciPy. > > >> And I need to ponder if I leave it at -Dblas=$CBLAS_PC for pkgsrc now. >> It's somewhat wrong, but also more correct, as NumPy _really_ means to >> use CBLAS API, not BLAS. >> > > My opinion is that this is not right. It may work with pkg-config in your > case, but I don't think it'll be robust for the fallback detection that is > used when pkg-config is not installed. > > Cheers, > Ralf > > > >> Alrighty then, >> >> Thomas >> >> -- >> Dr. Thomas Orgis >> HPC @ Universität Hamburg >> >

1 0

Fwd: incomplete BLAS/CBLAS linking (Telling meson build which CBLAS/LAPACK (LAPACKE?) to use via pkgconfig module)
by Ralf Gommers 01 Jan '24

01 Jan '24

---------- Forwarded message --------- From: Dr. Thomas Orgis <thomas.orgis(a)uni-hamburg.de> Date: Fri, Dec 29, 2023 at 12:00 AM Subject: Re: incomplete BLAS/CBLAS linking (Telling meson build which CBLAS/LAPACK (LAPACKE?) to use via pkgconfig module) To: Ralf Gommers <ralf.gommers(a)gmail.com> Am Thu, 28 Dec 2023 20:51:27 +0100 schrieb Ralf Gommers <ralf.gommers(a)gmail.com>: > > libblas.so > > liblapack.so (NEEDing libblas.so) > > libcblas.so (NEEDing libblas.so) > > libpapacke.so (NEEDing liblapack.so, hence libblas.so) > > > > and their respective .pc files. This is the natural order that occus to > > me when building from netlib upstream. > > > This should work fine. It's auto-detected in NumPy already, and will be in > SciPy in the future. For now, using `-Dblas=blas -Dlapack=lapack` in the > SciPy build should work. I noticed that with -Dblas=blas, which is in pkgsrc now. The detection code sets cblas and finds libcblas by dark magic / defaults that happen to match. But what if my setup uses -Dblas=netlib_blas? Then the internal guesswork would fail. Please consider a mode where the user specifies separate names for all 4 components. For package builds, we do not want any guess work, including assuming that libblas.so is accompanied by libcblas.so with that exact name. So I'd like -Dblas=$BLAS_PACKAGE -Dcblas=$CBLAS_PACKAGE \ -Dlapack=$LAPACK_PACKAGE -Dlapacke=$LAPACKE_PACKAGE where the values may all be the same or not. If I fail to provide one of those, feel free to guess for the rest (for example, assuming/trying that all of those are openblas if I say -Dblas=openblas). I also realized that including LAPACK in OpenBLAS is needed, but any new BLAS code could start out just replacing the netlib piece by piece. The partitioning is there and it is probably good for managing the complexity, limiting scope of the individual libraries. > > Telling the meson build that BLAS is libcblas works as long as actually > > CBLAS symbols are used. > > > Please never do this. The library is BLAS, so you should use `-Dblas=blas` > for NumPy. It will find `cblas` just fine that way. Oh. As I wrote before, we now have -Csetup-args=-Dblas=${CBLAS_PC} -Csetup-args=-Dlapack=${LAPACK_PC} for math/py-numpy. That's CBLAS_PC, not BLAS_PC. And this works. > This is probably a bug in SciPy. Well, apparently its just a miscommunication between us two. Scipy is fine with -Csetup-args=-Dblas=${BLAS_PC} -Csetup-args=-Dlapack=${LAPACK_PC} locating licblas by inferring it from libblas, and finding cblas in openblas_foobar, apparently. It prints those lines: Run-time dependency blas found: YES 3.11.0 Run-time dependency cblas found: YES 3.11.0 Run-time dependency lapack found: YES 3.11.0 blas : blas lapack : lapack While the numpy build does this: Run-time dependency cblas found: YES 3.11.0 Message: BLAS symbol suffix: Run-time dependency lapack found: YES 3.11.0 blas : cblas lapack : lapack This looks similar to the case of openblas_openmp for -Dblas and -Dlapack: Run-time dependency openblas_openmp found: YES 0.3.24 Message: BLAS symbol suffix: Run-time dependency openblas_openmp found: YES 0.3.24 blas : openblas_openmp lapack : openblas_openmp So scipy locates cblas based on the name blas, but doesn't really use cblas. Numpy is happy with libcblas bringing libblas in and calls it blas, but really uses the cblas interface. This looks a bit confusing. I guess it makes more sense to continue that discussion on the meson PRs for this functionality … as it transcends NumPy, anyway. I hope we can settle on something that works for autodetection and prescription of all parts. And I need to ponder if I leave it at -Dblas=$CBLAS_PC for pkgsrc now. It's somewhat wrong, but also more correct, as NumPy _really_ means to use CBLAS API, not BLAS. Alrighty then, Thomas -- Dr. Thomas Orgis HPC @ Universität Hamburg

1 0

Fwd: incomplete BLAS/CBLAS linking (Telling meson build which CBLAS/LAPACK (LAPACKE?) to use via pkgconfig module)
by Ralf Gommers 01 Jan '24

01 Jan '24

(I took this off-list unintentionally, so I'm forward each email to the list now) ---------- Forwarded message --------- From: Ralf Gommers <ralf.gommers(a)gmail.com> Date: Thu, Dec 28, 2023 at 8:51 PM Subject: Re: incomplete BLAS/CBLAS linking (Telling meson build which CBLAS/LAPACK (LAPACKE?) to use via pkgconfig module) To: Dr. Thomas Orgis <thomas.orgis(a)uni-hamburg.de> On Mon, Dec 25, 2023 at 10:37 AM Dr. Thomas Orgis < thomas.orgis(a)uni-hamburg.de> wrote: > Hapy holidays … but I have an issue still that hopefully can be > addressed with the meson blas detection you are upstreaming(?). > Happy holidays to you too. And yes, I hope so:) > > Am Wed, 6 Dec 2023 18:06:01 +0100 > schrieb Ralf Gommers <ralf.gommers(a)gmail.com>: > > > > Well, now is another day. Pkgsrc uses python -m build and I added > > > > > > -Csetup-args=-Dblas=${CBLAS_PC} > -Csetup-args=-Dlapack=${LAPACK_PC} > > > > > > which seems to work out fine using cblas.pc and lapack.pc in the case > > > of the netlib install. In fact, most linking is done only to libblas.so > > > instead of libcblas.so, as the linker is smart enough to throw away the > > > unused lib. > > > > > > > Great, thanks for confirming! > > This works for numpy and also installs scipy nicely, but this produces > a broken scipy install when using netlib reference libraries from > pkgsrc. These come as > > libblas.so > liblapack.so (NEEDing libblas.so) > libcblas.so (NEEDing libblas.so) > libpapacke.so (NEEDing liblapack.so, hence libblas.so) > > and their respective .pc files. This is the natural order that occus to > me when building from netlib upstream. This should work fine. It's auto-detected in NumPy already, and will be in SciPy in the future. For now, using `-Dblas=blas -Dlapack=lapack` in the SciPy build should work. > This also means that one could > just replace BLAS and put stock LAPACK on top, what optimized BLAS libs > usually start out with. This is indeed possible, if unusual. It's supported and one reason for why we have separate `blas` and `lapack` flags. I'd discourage distros from shipping something like that by default though, since it tends to lead to problems. Arch Linux used to do this, shipping an OpenBLAS without LAPACK symbols. Luckily they finally fixed that. Shipping non-default build configs like that is invariably a bad idea, and should only be done if there's a pressing need. Only that they tend to pack all symbols into > one common library, which then project builds like numpy rely on. > > Telling the meson build that BLAS is libcblas works as long as actually > CBLAS symbols are used. Please never do this. The library is BLAS, so you should use `-Dblas=blas` for NumPy. It will find `cblas` just fine that way. > If not — I presume now, as I didn't yet see the > actual build lines that are triggered via the python -m build and meson > indirections — the linker might discard the -lcblas and leave symbols > unresolved (--as-needed but no --no-undefined). > > This happens with scipy: > > $ LANG=C readelf -d > /data/pkg/lib/python3.11/site-packages/scipy/sparse/linalg/_dsolve/_superlu.so > |grep NEEDED > 0x0000000000000001 (NEEDED) Shared library: [libm.so.6] > 0x0000000000000001 (NEEDED) Shared library: [libc.so.6] > This is probably a bug in SciPy. The build target depends on both `blas` and `lapack`, and sets `DUSE_VENDOR_BLAS=1`. However, it looks like it should depend on `cblas`. If you add `cblas` to this line, I think it'll fix the issue: https://github.com/scipy/scipy/blob/6452a48c9611d16140b160091de6cf5299fadd9… . > It would link against libopenblas_openmp.so if that had been the CBLAS > (and LAPACK) choice and all would be fine, but here, it should link > with libcblas.so, or directly to libblas.so, just like our regular > install of superlu: > > $ LANG=C readelf -d /data/pkg/lib/libsuperlu.so|grep NEEDED > 0x0000000000000001 (NEEDED) Shared library: [libblas.so.3] > 0x0000000000000001 (NEEDED) Shared library: [libm.so.6] > 0x0000000000000001 (NEEDED) Shared library: [libc.so.6] > > Of course, just not vendoring superlu would be one solution for scipy, > but I think the deeper issue with the meson BLAS support should be > solved: The 4 parts of the BLAS canon (not talking about SCALAPACK etc. > yet) need to be handled explicitly. > > It is confusing, though, as meson prints this: > > Run-time dependency blas found: YES 3.11.0 > Run-time dependency cblas found: YES 3.11.0 > Run-time dependency lapack found: YES 3.11.0 > > It suggests that it looked for and found 3 libraries, but actually, it > only cared for -llapack and -lcblas. It needs to find -lblas directly, > too (or the cblas package separately, for that matter, not as component > of blas). > > Is that easily fixable from your side? (I'm assuming numpy, scipy and > the future stock BLAS support of meson are handled together.) Is this > just an oversight on the scipy side and they could link the vendored > superly with -lblas without changing the meson machinery? > Yes, I think it's an oversight indeed, and we didn't notice because we have very limited CI for BLAS/LAPACK configs still for SciPy (just a couple of ATLAS usages next to the OpenBLAS default, no jobs with Netlib/MKL/Accelerate). Cheers, Ralf

1 0

NEP 54 - SIMD infrastructure evolution to C++ and adopting Google Highway
by Ralf Gommers 31 Dec '23

31 Dec '23

Hi all, We have just NEP 54, "SIMD infrastructure evolution: adopting Google Highway when moving to C++?", with Draft status after a long review at https://github.com/numpy/numpy/pull/24138. It looks like it wasn't sent to this list before. Please see https://numpy.org/neps/nep-0054-simd-cpp-highway.html for the rendered version (complete text below). This is a complex topic, and the NEP captures more a discussion on the pros and cons of moving to Highway, and in what form. Most folks active in working on SIMD code in NumPy have weighed in in one of several calls, in the community meeting and the 3-weekly meeting of the recently formed NumPy Optimization Team. I think we can summarize the current status as follows: - Google Highway is now included in the main repo as a git submodule - We are +1 on using Highway for high-level operations where possible given accuracy constraints, and are already doing so for sorting functionality. - We are -1 on using Highway's dynamic dispatch, we prefer to stay with the current dynamic dispatch support via build system support, which has worked well for us for ~4 years now. - We are +0 to +0.5 on using Highway's form of 'universal intrinsics', in preference of moving our own universal intrinsics from C to C++. Both would be a major improvement on the current state of our C implementation. - For that latter decision, there isn't complete consensus on it, and also Highway is missing a few things that NumPy does have that we'd like to see it gain. In particular, a way to prototype and test new SIMD intrinsics from Python (see https://numpy.org/neps/nep-0054-simd-cpp-highway.html#the-simd-unit-testing… ). Cheers, Ralf full text of the NEP: =================================================================================== NEP 54 — SIMD infrastructure evolution: adopting Google Highway when moving to C++? =================================================================================== :Author: Sayed Adel, Jan Wassenberg, Matti Picus, Ralf Gommers, Chris Sidebottom :Status: Draft :Type: Standards Track :Created: 2023-07-06 :Resolution: TODO Abstract -------- We are moving the SIMD intrinsic framework, Universal Intrinsics, from C to C++. We have also moved to Meson as the build system. The Google Highway intrinsics project is proposing we use Highway instead of our Universal Intrinsics as described in `NEP 38`_. This is a complex and multi-faceted decision - this NEP is an attempt to describe the trade-offs involved and what would need to be done. Motivation and Scope -------------------- We want to refactor the C-based Universal Intrinsics (see :ref:`NEP 38 <NEP38>`) to C++. This work was ongoing for some time, and Google's Highway was suggested as an alternative, which was already written in C++ and had support for scalable SVE and other reusable components (such as VQSort). The move from C to C++ is motivated by (a) code readability and ease of development, (b) the need to add support for sizeless SIMD instructions (e.g., ARM's SVE, RISC-V's RVV). As an example of the readability improvement, here is a typical line of C code from our current C universal intrinsics framework: .. code:: // The @name@ is the numpy-specific templating in .c.src files npyv_@sfx@ a5 = npyv_load_@sfx@(src1 + npyv_nlanes_@sfx@ * 4); This will change (as implemented in PR `gh-21057`_) to: .. code:: C++ auto a5 = Load(src1 + nlanes * 4); If the above C++ code were to use Highway under the hood it would look quite similar, it uses similarly understandable names as ``Load`` for individual portable intrinsics. The ``@sfx`` in the C version above is the template variable for type identifiers, e.g.: ``#sfx = u8, s8, u16, s16, u32, s32, u64, s64, f32, f64#``. Explicit use of bitsize-encoded types like this won't work for sizeless SIMD instruction sets. With C++ this is easier to handle; PR `gh-21057`_ shows how and contains more complete examples of what the C++ code will look like. The scope of this NEP includes discussing most relevant aspects of adopting Google Highway to replace our current Universal Intrinsics framework, including but not limited to: - Maintainability, domain expertise availability, ease of onboarding new contributor, and other social aspects, - Key technical differences and constraints that may impact NumPy's internal design or performance, - Build system related aspects, - Release timing related aspects. Out of scope (at least for now) is revisiting other aspects of our current SIMD support strategy: - accuracy vs. performance trade-offs when adding SIMD support to a function - use of SVML and x86-simd-sort (and possibly its equivalents for aarch64) - pulling in individual bits or algorithms of Highway (as in `gh-24018`_) or SLEEF (as discussed in that same PR) Usage and Impact ---------------- N/A - there will be no significant user-visible changes. Backward compatibility ---------------------- There will be no changes in user-facing Python or C APIs: all the methods to control compilation and runtime CPU feature selection should remain, although there may be some changes due to moving to C++ without regards to the Highway/Universal Intrinsics choice. The naming of the CPU features in Highway is different from that of the Universal Intrinsics (see "Supported features/targets" below) On Windows, MSVC may have to be avoided, as a result of Highway's use of pragmas which are less well supported by MSVC. This means that we likely have to build our wheels with clang-cl or Mingw-w64. Both of those should work - we merged clang-cl support a while back (see `gh-20866`_), and SciPy builds with Mingw-w64. It may however impact other redistributors or end users who build from source on Windows. In response to the earlier dicussions around this NEP, Highway is now dual-licensed as Apache 2 / BSD-3. High-level considerations ------------------------- .. note:: Currently this section attempts to cover each topic separately, and comparing the future use of a NumPy-specific C++ implementation vs. use of Google Highway with our own numerical routines on top of that. It does not (yet) assume a decision or proposed decision is made. Hence this NEP is not "this is proposed" with another option in the Alternatives section, but rather a side-by-side comparison. Development effort and long-term maintainability ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Moving to Highway is likely to be a significant development effort. Longer-term, this will hopefully be offset by Highway itself having more maintainer bandwidth to deal with ongoing issues in compiler support and adding new platforms. Highway being used by other projects, like Chromium and `JPEG XL`_ (see `this more complete list < https://google.github.io/highway/en/master/README.html#examples>`__ in the Highway documentation), does imply that there is likely to be a benefit of a wider range of testing and bug reporting/fixing. One concern is that new instructions may have to be added, and that that is often best done as part of the process of developing the numerical kernel that needs the instruction. This will be a little more clumsy if the instruction lives in Highway which is a git submodule inside the NumPy repo - there will be a need to implement a temporary/generic version first, and then update the submodule after upstreaming the new intrinsic. Documentation-wise, Highway would be a clear win. NumPy's `CPU/SIMD Optimizations`_ docs are fairly sparse compared to `the Highway docs`_. Migration strategy - can it be gradual? ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ This is a story of two halves. Moving to Highway's statically dispatched intrinsics could be done gradually, as already seen in PR `gh-24018`_. However, adopting Highway's way of performing runtime dispatching has to be done in one go - we can't (or shouldn't) have two ways of doing that. Highway policies for compiler and platform support ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ When adding new instructions, Highway has a policy that they must be implemented in a way that fairly balances across CPU architectures. Regarding the support status and whether all currently-supported architectures will remain supported, Jan stated that Highway can commit to the following: 1. If it cross-compiles with Clang and can be tested via standard QEMU, it can go into Highway's CI. 2. If it cross-compiles via clang/gcc and can be tested with a new QEMU (possibly with extra flags), then it can be support via manual testing before each Highway release. 3. Existing targets will remain supported as long as they compile/run in QEMU. Highway is not subject to Google's "no longer supported" strategy (or, as written in its README, *This is not an officially supported Google product*). That is not a bad thing; it means that it is less likely to go unsupported due to a Google business decision about the project. Quite a few well-known open source projects under the ``google`` GitHub org state this, e.g. `JAX`_ and `tcmalloc`_. Supported features/targets ~~~~~~~~~~~~~~~~~~~~~~~~~~ Both frameworks support a large set of platforms and SIMD instruction sets, as well as generic scalar/fallback versions. The main differences right now are: - NumPy supports IBM Z-system (s390x, VX/VXE/VXE2) while Highway supports Z14, Z15. - Highway supports ARM SVE/SVE2 and RISC-V RVV (sizeless instructions), while NumPy does not. - The groundwork for sizeless SIMD support in NumPy has been done in `gh-21057`_, however SVE/SVE2 and RISC-V are not yet implemented there. There is also a difference in the granularity of instruction set groups: NumPy supports a more granular set of architectures than Highway. See the list of targets for Highway `here <https://github.com/google/highway/#targets>`__ (it's roughly per CPU family) and for NumPy `here < https://numpy.org/doc/1.25/reference/simd/build-options.html#supported-feat… >`__ (roughly per SIMD instruction set). Hence with Highway we'd lose some granularity - but that is probably fine, we don't really need this level of granularity, and there isn't much evidence that users explicitly play with this to squeeze out the last bit of performance for their own CPU. Compilation strategy for multiple targets and runtime dispatching ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Highway compiles once while using preprocessing tricks to generate multiple stanzas for each CPU feature within the same compilation unit (see the ``foreach_target.h`` usage and dynamic dispatch docs for how that is done). Universal Intrinsics generate multiple compilation units, one for each CPU feature group, and compiles multiple times, linking them all together (with different names) for runtime dispatch. The Highway technique may not work reliably on MSVC, the Universal Intrinsic technique does work on MSVC. Which one is more robust? The experts disagree. Jan thinks that the Highway approach is more robust and in particular avoids the linker pulling in functions with too-new instructions into the final binary. Sayed thinks that the current NumPy approach (also used by OpenCV) is more robust, and in particular is less likely to run into compiler-specific bugs or catch them earlier. Both agree the meson build system allows specifying object link order, which produces more consistent builds. However that does tie NumPy to meson. Matti and Ralf think the current build strategy is working well for NumPy and the advantages of changing the build and runtime dispatch, with possible unknown instabilities outweighs the advantages that adopting Highway's dynamic dispatch may bring. Our experience of the past four years says that bugs with "invalid instruction" type crashes are invariably due to issues with feature detection - most often because users are running under emulation, and sometimes because there are actual issues with our CPU feature detection code. There is little evidence we're aware of of the linker pulling in a function which is compiled multiple times for different architectures and picking the one with unsupported instructions. To ensure to avoid the issue, it's advisable to keep numerical kernels inside the source code and refrain from defining non-inlined functions within cache-able objects. C++ refactoring considerations ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We want to move from C to C++, which will naturally involve a significant amount of refactoring, for two main reasons: - get rid of the NumPy-specific templating language for more expressive C++ - this would make using sizeless intrinsics (like for SVE) easier. In addition, we see the following considerations: - If we use Highway, we would need to switch the C++ wrappers from universal intrinsics to Highway. On the other hand, the work to move to C++ is not complete. - If we use Highway, we'd need to rewrite existing kernels using Highway intrinsics. But again, moving to C++ requires touching all those kernels anyway. - One concern regarding Highway was whether it is possible to obtain a function pointer for an architecture-specific function instead of calling that function directly. This so that we can be sure that calling 1-D inner loop many times for a single Python API invocation does not incur the dispatching overhead many times. This was investigated: this can be done with Highway too. - A second concern was whether it's possible with Highway to allow the user at runtime to select or disable dispatching to certain instruction sets. This is possible. - Use of tags in Highway's C++ implementation reduces code duplication but the added templating makes C-level testing and tracing more complicated. The ``_simd`` unit testing module ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Rewriting the ``_simd testing`` module to use C++ was done very recently in PR `gh-24069`_. It depends on the main PR for the move to C++, `gh-21057`_. It allows one to access the C++ intrinsics with almost the same signature, but from Python. This is a great way not only for testing, but also for designing new SIMD kernels. It may be possible to add a similar testing and prototyping feature to Highway (which uses plain ``googletest``), however currently the NumPy way is quite a bit nicer. Math routines ~~~~~~~~~~~~~ Math or numerical routines are written at a higher level of abstraction than the universal intrinsics that are the main focus of this NEP. Highway has only a limited number of math routines, and they are not precise enough for NumPy's needs. So either way, NumPy's existing routines (which use universal intrinsics) will stay, and if we go the Highway route they'll simply have to use Highway primitives internally. We could still use Highway sorting routines. If we do accept lower-precision routines (via a user-supplied choice, i.e. extending ``errstate`` to allow a precision option), we could use Highway-native routines. There may be other libraries that have numerical routines that can be reused in NumPy (e.g., from SLEEF, or perhaps from JPEG XL or some other Highway-using libraries). There may be a small benefit here, but likely it doesn't matter too much. Supported and missing intrinsics ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Some specific intrinsics that NumPy needs may be missing from Highway. Similarly, some intrinsics that NumPy needs to implement routines are already implemented in Highway and are missing from NumPy. Highway has more instructions that NumPy's universal intrinsics, so it's possible that some future needs for NumPy kernels may already be met there. Either way, we will always have to implement intrinsics in either solution. Related Work ------------ - `Google Highway`_ - `Xsimd`_ - OpenCV's SIMD framework (`API reference < https://docs.opencv.org/4.x/df/d91/group__core__hal__intrin.html>`__, `docs <https://github.com/opencv/opencv/wiki/CPU-optimizations-build-options>`__) - `std::experimental::simd < https://en.cppreference.com/w/cpp/experimental/simd/simd>`__ - See the Related Work section in :ref:`NEP38` for more related work (as of 2019) Implementation -------------- TODO Alternatives ------------ Use Google Highway for dynamic dispatch. Other alternatives include: do nothing and stay with C universal intrinsics, use `Xsimd`_ as the SIMD framework (less comprehensive than Highway - no SVE or PowerPC support for example), or use/vendor `SLEEF`_ (a good library, but inconsistently maintained). Neither of these alternatives seems appealing. Discussion ---------- References and Footnotes ------------------------ .. [1] Each NEP must either be explicitly labeled as placed in the public domain (see this NEP as an example) or licensed under the `Open Publication License`_. .. _Open Publication License: https://www.opencontent.org/openpub/ .. _`NEP 38`: https://numpy.org/neps/nep-0038-SIMD-optimizations.html .. _`gh-20866`: https://github.com/numpy/numpy/pull/20866 .. _`gh-21057`: https://github.com/numpy/numpy/pull/21057 .. _`gh-23096`: https://github.com/numpy/numpy/pull/23096 .. _`gh-24018`: https://github.com/numpy/numpy/pull/24018 .. _`gh-24069`: https://github.com/numpy/numpy/pull/24069 .. _JPEG XL: https://github.com/libjxl/libjxl .. _CPU/SIMD Optimizations: https://numpy.org/doc/1.25/reference/simd/ .. _the Highway docs: https://google.github.io/highway/ .. _Google Highway: https://github.com/google/highway/ .. _Xsimd: https://github.com/xtensor-stack/xsimd .. _SLEEF: https://sleef.org/ .. _tcmalloc: https://github.com/google/tcmalloc .. _JAX: https://github.com/google/jax Copyright --------- This document has been placed in the public domain. [1]_

1 0

How is "round to N decimal places" defined for binary floating point numbers?
by Stefano Miccoli 29 Dec '23

29 Dec '23

I have always been puzzled about how to correctly define the python built-in `round(number, ndigits)` when `number` is a binary float and `ndigits` is greater than zero. Apparently CPython and numpy disagree: >>> round(2.765, 2) 2.77 >>> np.round(2.765, 2) 2.76 My question for the numpy devs are: - Is there an authoritative source that explains what `round(number, ndigits)` means when the digits are counted in a base different from the one used in the floating point representation? - Which was the first programming language to implement an intrinsic function `round(number, ndigits)` where ndgits are always decimal, irrespective of the representation of the floating point number? (I’m not interested in algorithms for printing a decimal representation, but in languages that allow to store and perform computations with the rounded value.) - Is `round(number, ndigits)` a useful function that deserves a rigorous definition, or is its use limited to fuzzy situations, where accuracy can be safely traded for speed? Personally I cannot think of sensible uses of `round(number, ndigits)` for binary floats: whenever you positively need `round(number, ndigits)`, you should use a decimal floating point representation. Stefano

3 3

next NumPy Newcomers' Hour - December 28th, 2023 at 12 PM UTC
by Ganesh Kathiresan 28 Dec '23

28 Dec '23

Apologies for the late notice! Our next Newcomers' Hour will be held this Thursday, 28th, 2023 at 12 PM UTC, Stop by to ask questions, share your progress, celebrate success, or just to say hi. To add to the meeting agenda the topics you’d like to discuss, follow the link: https://hackmd.io/3f3otyyuTte3FU9y3QzsLg?both. Join the meeting via Zoom: https://us06web.zoom.us/j/82563808729?pwd=ZFU3Z2dMcXBGb05YemRsaGE1OW5nQT09. Thanks, Ganesh

1 0

Telling meson build which CBLAS/LAPACK (LAPACKE?) to use via pkgconfig module
by Dr. Thomas Orgis 25 Dec '23

25 Dec '23

Hi, I'm involved in packaging NumPy for http://pkgsrc.org/. We install a set of possible BLAS/CBLAS/LAPACK/LAPACKE packages side-by-side in the same prefix. This includes multiple variants of OpenBLAS with regard to multithreading (and indexing). For this purpose, we point software to use the build-time chosen BLAS implementation via BLAS_LIBS and similar variables, or, as seems to be appropriate for the new meson build of NumPy, via .pc files. A package depends on the generic BLAS library family and central user configuration chooses which one the packages should use during build. What would be the correct way to force the NumPy build to just use our BLAS choice, avoiding any automatisms that might surprise us? How agnostic is NumPy regarding the offered BLAS? Does it have to know that it is using OpenMP-parallelized OpenBLAS vs. the serial one (I'd imagine just setting OMP_NUM_THREADS handles paralellism) or MKL, Netlib, … ? It doesn't scale to have to tell it "openblas" or "netlib", as there is no universal vocabulary to name the variants, and NumPy doesn't even know openblas_openmp from serial openblas or openblas_pthread (right?). Basically, I want to do meson setup -Dcblas_pc=$CBLAS_PC with CBLAS_PC being the module name of one of $prefix/lib/pkgconfig/cblas.pc $prefix/lib/pkgconfig/openblas_pthread.pc $prefix/lib/pkgconfig/openblas_openmp.pc $prefix/lib/pkgconfig/openblas64_openmp.pc so that pkg-config does its thing without the NumPy build guessing around. Is that feasible already? Is it easily supportable with some changes to the build? I dream of a world where package build scripts don't have to add dozens of idiosyncratic lines to detect these libs. I'd like things to work like for CMake's FindBLAS with -DBLA_PREFER_PKGCONFIG and -DBLA_PKGCONFIG_BLAS=$BLAS_PC (see https://cmake.org/cmake/help/latest/module/FindBLAS.html since version 3.25). Can we have that? And: NumPy needs CBLAS … does it also need LAPACKE instead of LAPACK? These are differing libraries, possibly coming in differing binaries, even if your OpenBLAS builds also combine them. So I guess it should be -Dcblas_pc and -Dlapacke_pc, both being possibly identical. A build of the reference Netlib implementation provides four distinct libraries and .pc files: $prefix/lib/pkgconfig/cblas.pc $prefix/lib/pkgconfig/blas.pc $prefix/lib/pkgconfig/lapacke.pc $prefix/lib/pkgconfig/lapack.pc We do support installing openblas64 and friends alongside the others and I imagine just setting an ILP64 option and repective symbol suffix (none as of yet, as it's not a settled thing upstream) for the NumPy build if a 64 variant is chosen by the user. I wonder a bit if there are possible pitfalls combining other libraries with Python and indirectly some incompatible BLAS variant via NumPy … but one point of our user choice is that they could ensure that all packages really use the same BLAS. Alrighty then, Thomas -- Dr. Thomas Orgis HPC @ Universität Hamburg

2 7

ENH: ignore_nan parameter in np.average
by alexwest.mays＠gmail.com 24 Dec '23

24 Dec '23

Hi there, I would love some input on this issue, https://github.com/numpy/numpy/issues/21375. I have created a draft PR which I believe resolves it, https://github.com/numpy/numpy/pull/25474. Decided to go with adding ignore_nan to np.average rather than adding weights to np.nanmean as it seemed logical and was the path of least resistance. Please let me know what you think! -awestm

2 1

updated 2.0-dev nightlies, including macOS Accelerate wheels
by Ralf Gommers 23 Dec '23

23 Dec '23

Hi all, We had some issues with nightlies, the macOS, Linux aarch64 and PyPy ones were about a month out of date. That is fixed now, new nightlies for all supported platforms are up on https://anaconda.org/scientific-python-nightly-wheels/numpy. Note that a lot changed in `main` over the last month, so if you see lots of failures in pre-release CI jobs of downstream packages, getting all of those changes at the same time will be why. The 2.0 migration guide should be mostly up to date, and covers the most important changes: https://numpy.org/devdocs/numpy_2_0_migration_guide.html. In addition, we have new wheels for arm64 macOS >=14.0 which use the new Apple Accelerate rather than OpenBLAS. They're ~3x smaller and linalg operations are significantly faster as a result. Cheers, Ralf

1 0

ANN: SciPy 1.12.0rc1 -- please test
by Tyler Reddy 21 Dec '23

21 Dec '23

Hi all, On behalf of the SciPy development team I'm pleased to announce the release candidate SciPy 1.12.0rc1. Please help us test this pre-release. Sources and binary wheels can be found at: https://pypi.org/project/SciPy/ and at: https://github.com/scipy/scipy/releases/tag/v1.12.0rc1 One of a few ways install the release candidate with pip: pip install scipy==1.12.0rc1 ========================== SciPy 1.12.0 Release Notes ========================== Note: SciPy 1.12.0 is not released yet! SciPy 1.12.0 is the culmination of 6 months of hard work. It contains many new features, numerous bug-fixes, improved test coverage and better documentation. There have been a number of deprecations and API changes in this release, which are documented below. All users are encouraged to upgrade to this release, as there are a large number of bug-fixes and optimizations. Before upgrading, we recommend that users check that their own code does not use deprecated SciPy functionality (to do so, run your code with ``python -Wd`` and check for ``DeprecationWarning`` s). Our development attention will now shift to bug-fix releases on the 1.12.x branch, and on adding new features on the main branch. This release requires Python 3.9+ and NumPy 1.22.4 or greater. For running on PyPy, PyPy3 6.0+ is required. ************************** Highlights of this release ************************** - Experimental support for the array API standard has been added to part of `scipy.special`, and to all of `scipy.fft` and `scipy.cluster`. There are likely to be bugs and early feedback for usage with CuPy arrays, PyTorch tensors, and other array API compatible libraries is appreciated. Use the ``SCIPY_ARRAY_API`` environment variable for testing. - A new class, ``ShortTimeFFT``, provides a more versatile implementation of the short-time Fourier transform (STFT), its inverse (ISTFT) as well as the (cross-) spectrogram. It utilizes an improved algorithm for calculating the ISTFT. - Several new constructors have been added for sparse arrays, and many operations now additionally support sparse arrays, further facilitating the migration from sparse matrices. - A large portion of the `scipy.stats` API now has improved support for handling ``NaN`` values, masked arrays, and more fine-grained shape-handling. The accuracy and performance of a number of ``stats`` methods have been improved, and a number of new statistical tests and distributions have been added. ************ New features ************ `scipy.cluster` improvements ============================ - Experimental support added for the array API standard; PyTorch tensors, CuPy arrays and array API compatible array libraries are now accepted (GPU support is limited to functions with pure Python implementations). CPU arrays which can be converted to and from NumPy are supported module-wide and returned arrays will match the input type. This behaviour is enabled by setting the ``SCIPY_ARRAY_API`` environment variable before importing ``scipy``. This experimental support is still under development and likely to contain bugs - testing is very welcome. `scipy.fft` improvements ======================== - Experimental support added for the array API standard; functions which are part of the ``fft`` array API standard extension module, as well as the Fast Hankel Transforms and the basic FFTs which are not in the extension module, now accept PyTorch tensors, CuPy arrays and array API compatible array libraries. CPU arrays which can be converted to and from NumPy arrays are supported module-wide and returned arrays will match the input type. This behaviour is enabled by setting the ``SCIPY_ARRAY_API`` environment variable before importing ``scipy``. This experimental support is still under development and likely to contain bugs - testing is very welcome. `scipy.integrate` improvements ============================== - Added `scipy.integrate.cumulative_simpson` for cumulative quadrature from sampled data using Simpson's 1/3 rule. `scipy.interpolate` improvements ================================ - New class ``NdBSpline`` represents tensor-product splines in N dimensions. This class only knows how to evaluate a tensor product given coefficients and knot vectors. This way it generalizes ``BSpline`` for 1D data to N-D, and parallels ``NdPPoly`` (which represents N-D tensor product polynomials). Evaluations exploit the localized nature of b-splines. - ``NearestNDInterpolator.__call__`` accepts ``**query_options``, which are passed through to the ``KDTree.query`` call to find nearest neighbors. This allows, for instance, to limit the neighbor search distance and parallelize the query using the ``workers`` keyword. - ``BarycentricInterpolator`` now allows computing the derivatives. - It is now possible to change interpolation values in an existing ``CloughTocher2DInterpolator`` instance, while also saving the barycentric coordinates of interpolation points. `scipy.linalg` improvements =========================== - Access to new low-level LAPACK functions is provided via ``dtgsyl`` and ``stgsyl``. `scipy.ndimage` improvements ============================ `scipy.optimize` improvements ============================= - `scipy.optimize.nnls` is rewritten in Python and now implements the so-called fnnls or fast nnls. - The result object of `scipy.optimize.root` and `scipy.optimize.root_scalar` now reports the method used. - The ``callback`` method of `scipy.optimize.differential_evolution` can now be passed more detailed information via the ``intermediate_results`` keyword parameter. Also, the evolution ``strategy`` now accepts a callable for additional customization. The performance of ``differential_evolution`` has also been improved. - ``minimize`` method ``Newton-CG`` has been made slightly more efficient. - ``minimize`` method ``BFGS`` now accepts an initial estimate for the inverse of the Hessian, which allows for more efficient workflows in some circumstances. The new parameter is ``hess_inv0``. - ``minimize`` methods ``CG``, ``Newton-CG``, and ``BFGS`` now accept parameters ``c1`` and ``c2``, allowing specification of the Armijo and curvature rule parameters, respectively. - ``curve_fit`` performance has improved due to more efficient memoization of the callable function. - ``isotonic_regression`` has been added to allow nonparametric isotonic regression. `scipy.signal` improvements =========================== - ``freqz``, ``freqz_zpk``, and ``group_delay`` are now more accurate when ``fs`` has a default value. - The new class ``ShortTimeFFT`` provides a more versatile implementation of the short-time Fourier transform (STFT), its inverse (ISTFT) as well as the (cross-) spectrogram. It utilizes an improved algorithm for calculating the ISTFT based on dual windows and provides more fine-grained control of the parametrization especially in regard to scaling and phase-shift. Functionality was implemented to ease working with signal and STFT chunks. A section has been added to the "SciPy User Guide" providing algorithmic details. The functions ``stft``, ``istft`` and ``spectrogram`` have been marked as legacy. `scipy.sparse` improvements =========================== - ``sparse.linalg`` iterative solvers ``sparse.linalg.cg``, ``sparse.linalg.cgs``, ``sparse.linalg.bicg``, ``sparse.linalg.bicgstab``, ``sparse.linalg.gmres``, and ``sparse.linalg.qmr`` are rewritten in Python. - Updated vendored SuperLU version to ``6.0.1``, along with a few additional fixes. - Sparse arrays have gained additional constructors: ``eye_array``, ``random_array``, ``block_array``, and ``identity``. ``kron`` and ``kronsum`` have been adjusted to additionally support operation on sparse arrays. - Sparse matrices now support a transpose with ``axes=(1, 0)``, to mirror the ``.T`` method. - ``LaplacianNd`` now allows selection of the largest subset of eigenvalues, and additionally now supports retrieval of the corresponding eigenvectors. The performance of ``LaplacianNd`` has also been improved. - The performance of ``dok_matrix`` and ``dok_array`` has been improved, and their inheritance behavior should be more robust. - ``hstack``, ``vstack``, and ``block_diag`` now work with sparse arrays, and preserve the input sparse type. - A new function, `scipy.sparse.linalg.matrix_power`, has been added, allowing for exponentiation of sparse arrays. `scipy.spatial` improvements ============================ - Two new methods were implemented for ``spatial.transform.Rotation``: ``__pow__`` to raise a rotation to integer or fractional power and ``approx_equal`` to check if two rotations are approximately equal. - The method ``Rotation.align_vectors`` was extended to solve a constrained alignment problem where two vectors are required to be aligned precisely. Also when given a single pair of vectors, the algorithm now returns the rotation with minimal magnitude, which can be considered as a minor backward incompatible change. - A new representation for ``spatial.transform.Rotation`` called Davenport angles is available through ``from_davenport`` and ``as_davenport`` methods. - Performance improvements have been added to ``distance.hamming`` and ``distance.correlation``. - Improved performance of ``SphericalVoronoi`` ``sort_vertices_of_regions`` and two dimensional area calculations. `scipy.special` improvements ============================ - Added `scipy.special.stirling2` for computation of Stirling numbers of the second kind. Both exact calculation and an asymptotic approximation (the default) are supported via ``exact=True`` and ``exact=False`` (the default) respectively. - Added `scipy.special.betaincc` for computation of the complementary incomplete Beta function and `scipy.special.betainccinv` for computation of its inverse. - Improved precision of `scipy.special.betainc` and `scipy.special.betaincinv` - Experimental support added for alternative backends: functions `scipy.special.log_ndtr`, `scipy.special.ndtr`, `scipy.special.ndtri`, `scipy.special.erf`, `scipy.special.erfc`, `scipy.special.i0`, `scipy.special.i0e`, `scipy.special.i1`, `scipy.special.i1e`, `scipy.special.gammaln`, `scipy.special.gammainc`, `scipy.special.gammaincc`, `scipy.special.logit`, and `scipy.special.expit` now accept PyTorch tensors and CuPy arrays. These features are still under development and likely to contain bugs, so they are disabled by default; enable them by setting a ``SCIPY_ARRAY_API`` environment variable to ``1`` before importing ``scipy``. Testing is appreciated! `scipy.stats` improvements ========================== - Added `scipy.stats.quantile_test`, a nonparametric test of whether a hypothesized value is the quantile associated with a specified probability. The ``confidence_interval`` method of the result object gives a confidence interval of the quantile. - `scipy.stats.wasserstein_distance` now computes the Wasserstein distance in the multidimensional case. - `scipy.stats.sampling.FastGeneratorInversion` provides a convenient interface to fast random sampling via numerical inversion of distribution CDFs. - `scipy.stats.geometric_discrepancy` adds geometric/topological discrepancy metrics for random samples. - `scipy.stats.multivariate_normal` now has a ``fit`` method for fitting distribution parameters to data via maximum likelihood estimation. - `scipy.stats.bws_test` performs the Baumgartner-Weiss-Schindler test of whether two-samples were drawn from the same distribution. - `scipy.stats.jf_skew_t` implements the Jones and Faddy skew-t distribution. - `scipy.stats.anderson_ksamp` now supports a permutation version of the test using the ``method`` parameter. - The ``fit`` methods of `scipy.stats.halfcauchy`, `scipy.stats.halflogistic`, and `scipy.stats.halfnorm` are faster and more accurate. - `scipy.stats.beta` ``entropy`` accuracy has been improved for extreme values of distribution parameters. - The accuracy of ``sf`` and/or ``isf`` methods have been improved for several distributions: `scipy.stats.burr`, `scipy.stats.hypsecant`, `scipy.stats.kappa3`, `scipy.stats.loglaplace`, `scipy.stats.lognorm`, `scipy.stats.lomax`, `scipy.stats.pearson3`, `scipy.stats.rdist`, and `scipy.stats.pareto`. - The following functions now support parameters ``axis``, ``nan_policy``, and ``keep_dims``: `scipy.stats.entropy`, `scipy.stats.differential_entropy`, `scipy.stats.variation`, `scipy.stats.ansari`, `scipy.stats.bartlett`, `scipy.stats.levene`, `scipy.stats.fligner`, `scipy.stats.cirmean, `scipy.stats.circvar`, `scipy.stats.circstd`, `scipy.stats.tmean`, `scipy.stats.tvar`, `scipy.stats.tstd`, `scipy.stats.tmin`, `scipy.stats.tmax`, and `scipy.stats.tsem`. - The ``logpdf`` and ``fit`` methods of `scipy.stats.skewnorm` have been improved. - The beta negative binomial distribution is implemented as `scipy.stats.betanbinom`. - The speed of `scipy.stats.invwishart` ``rvs`` and ``logpdf`` have been improved. - A source of intermediate overflow in `scipy.stats.boxcox_normmax` with ``method='mle'`` has been eliminated, and the returned value of ``lmbda`` is constrained such that the transformed data will not overflow. - `scipy.stats.nakagami` ``stats`` is more accurate and reliable. - A source of intermediate overflow in `scipy.norminvgauss.pdf` has been eliminated. - Added support for masked arrays to ``stats.circmean``, ``stats.circvar``, ``stats.circstd``, and ``stats.entropy``. - ``dirichlet`` has gained a new covariance (``cov``) method. - Improved accuracy of ``multivariate_t`` entropy with large degrees of freedom. - ``loggamma`` has an improved ``entropy`` method. ******************* Deprecated features ******************* - Error messages have been made clearer for objects that don't exist in the public namespace and warnings sharpened for private attributes that are not supposed to be imported at all. - `scipy.signal.cmplx_sort` has been deprecated and will be removed in SciPy 1.14. A replacement you can use is provided in the deprecation message. - Values the the argument ``initial`` of `scipy.integrate.cumulative_trapezoid` other than ``0`` and ``None`` are now deprecated. - `scipy.stats.rvs_ratio_uniforms` is deprecated in favour of `scipy.stats.sampling.RatioUniforms` - `scipy.integrate.quadrature` and `scipy.integrate.romberg` have been deprecated due to accuracy issues and interface shortcomings. They will be removed in SciPy 1.14. Please use `scipy.integrate.quad` instead. - Coinciding with upcoming changes to function signatures (e.g. removal of a deprecated keyword), we are deprecating positional use of keyword arguments for the affected functions, which will raise an error starting with SciPy 1.14. In some cases, this has delayed the originally announced removal date, to give time to respond to the second part of the deprecation. Affected functions are: - ``linalg.{eigh, eigvalsh, pinv}`` - ``integrate.simpson`` - ``signal.{firls, firwin, firwin2, remez}`` - ``sparse.linalg.{bicg, bicgstab, cg, cgs, gcrotmk, gmres, lgmres, minres, qmr, tfqmr}`` - ``special.comb`` - ``stats.kendalltau`` - All wavelet functions have been deprecated, as PyWavelets provides suitable implementations; affected functions are: ``signal.{daub, qmf, cascade, morlet, morlet2, ricker, cwt}`` ********************* Expired Deprecations ********************* There is an ongoing effort to follow through on long-standing deprecations. The following previously deprecated features are affected: - The ``centered`` keyword of `stats.qmc.LatinHypercube` has been removed. Use ``scrambled=False`` instead of ``centered=True``. ****************************** Backwards incompatible changes ****************************** ************* Other changes ************* - The arguments used to compile and link SciPy are now available via ``show_config``. ******* Authors ******* * Name (commits) * endolith (1) * h-vetinari (29) * Tom Adamczewski (3) + * Anudeep Adiraju (1) + * akeemlh (1) * Alex Amadori (2) + * Raja Yashwanth Avantsa (2) + * Seth Axen (1) + * Ross Barnowski (1) * Dan Barzilay (1) + * Ashish Bastola (1) + * Christoph Baumgarten (2) * Ben Beasley (3) + * Doron Behar (1) * Peter Bell (1) * Sebastian Berg (1) * Ben Boeckel (1) + * David Boetius (1) + * Jake Bowhay (102) * Larry Bradley (1) + * Dietrich Brunn (5) * Evgeni Burovski (101) * Matthias Bussonnier (18) * CJ Carey (6) * Colin Carroll (1) + * Aadya Chinubhai (1) + * Luca Citi (1) * Lucas Colley (140) + * com3dian (1) + * Anirudh Dagar (4) * Danni (1) + * Dieter Werthmüller (1) * John Doe (2) + * Philippe DONNAT (2) + * drestebon (1) + * Thomas Duvernay (1) * elbarso (1) + * emilfrost (2) + * Paul Estano (8) + * Evandro (2) * Franz Király (1) + * Nikita Furin (1) + * gabrielthomsen (1) + * Lukas Geiger (9) + * Artem Glebov (22) + * Caden Gobat (1) * Ralf Gommers (125) * Alexander Goscinski (2) + * Rohit Goswami (2) + * Olivier Grisel (1) * Matt Haberland (243) * Charles Harris (1) * harshilkamdar (1) + * Alon Hovav (2) + * Gert-Ludwig Ingold (1) * Romain Jacob (1) + * jcwhitehead (1) + * Julien Jerphanion (13) * He Jia (1) * JohnWT (1) + * jokasimr (1) + * Evan W Jones (1) * Karen Róbertsdóttir (1) + * Ganesh Kathiresan (1) * Robert Kern (11) * Andrew Knyazev (4) * Uwe L. Korn (1) + * Rishi Kulkarni (1) * Kale Kundert (3) + * Jozsef Kutas (2) * Kyle0 (2) + * Robert Langefeld (1) + * Jeffrey Larson (1) + * Jessy Lauer (1) + * lciti (1) + * Hoang Le (1) + * Antony Lee (5) * Thilo Leitzbach (4) + * LemonBoy (2) + * Ellie Litwack (8) + * Thomas Loke (4) + * Malte Londschien (1) + * Christian Lorentzen (6) * Adam Lugowski (9) + * lutefiskhotdish (1) * mainak33 (1) + * Ben Mares (11) + * mart-mihkel (2) + * Mateusz Sokół (24) + * Nikolay Mayorov (4) * Nicholas McKibben (1) * Melissa Weber Mendonça (7) * Kat Mistberg (2) + * mkiffer (1) + * mocquin (1) + * Nicolas Mokus (2) + * Sturla Molden (1) * Roberto Pastor Muela (3) + * Bijay Nayak (1) + * Andrew Nelson (105) * Praveer Nidamaluri (2) + * Lysandros Nikolaou (2) * Dimitri Papadopoulos Orfanos (7) * Pablo Rodríguez Pérez (1) + * Dimitri Papadopoulos (2) * Tirth Patel (14) * Kyle Paterson (1) + * Paul (4) + * Yann Pellegrini (2) + * Matti Picus (4) * Ilhan Polat (36) * Pranav (1) + * Bharat Raghunathan (1) * Chris Rapson (1) + * Matteo Raso (4) * Tyler Reddy (165) * Martin Reinecke (1) * Tilo Reneau-Cardoso (1) + * resting-dove (2) + * Simon Segerblom Rex (4) * Lucas Roberts (2) * Pamphile Roy (31) * Feras Saad (3) + * Atsushi Sakai (3) * Masahiro Sakai (2) + * Omar Salman (14) * Andrej Savikin (1) + * Daniel Schmitz (52) * Dan Schult (19) * Scott Shambaugh (9) * Sheila-nk (2) + * Mauro Silberberg (3) + * Maciej Skorski (1) + * Laurent Sorber (1) + * Albert Steppi (28) * Kai Striega (1) * Saswat Susmoy (1) + * Alex Szatmary (1) + * Søren Fuglede Jørgensen (3) * othmane tamri (3) + * Ewout ter Hoeven (1) * Will Tirone (1) * TLeitzbach (1) + * Kevin Topolski (1) + * Edgar Andrés Margffoy Tuay (1) * Dipansh Uikey (1) + * Matus Valo (3) * Christian Veenhuis (2) * Nicolas Vetsch (1) + * Isaac Virshup (7) * Hielke Walinga (2) + * Stefan van der Walt (2) * Warren Weckesser (7) * Bernhard M. Wiedemann (4) * Levi John Wolf (1) * Xuefeng Xu (4) + * Rory Yorke (2) * YoussefAli1 (1) + * Irwin Zaid (4) + * Jinzhe Zeng (1) + * JIMMY ZHAO (1) + A total of 161 people contributed to this release. People with a "+" by their names contributed a patch for the first time. This list of names is automatically generated, and may not be fully complete. ************************ Issues closed for 1.12.0 ************************ * `#2725 <https://github.com/scipy/scipy/issues/2725>`__: Barycentric interpolation should allow evaluation of derivatives * `#4244 <https://github.com/scipy/scipy/issues/4244>`__: betaincinv accuracy * `#4677 <https://github.com/scipy/scipy/issues/4677>`__: Unexpected behavior from scipy.special.btdtri * `#4819 <https://github.com/scipy/scipy/issues/4819>`__: Comparison operator overloading doesn't work with sparse matrices * `#5022 <https://github.com/scipy/scipy/issues/5022>`__: bicg returns last iterate, not the smallest-residue vector * `#6198 <https://github.com/scipy/scipy/issues/6198>`__: callback for Krylov methods * `#7241 <https://github.com/scipy/scipy/issues/7241>`__: ENH: Implement Chandrupatla's algorithm for root finding (simpler/faster... * `#9527 <https://github.com/scipy/scipy/issues/9527>`__: Anderson-Darling ksamples can not estimate p-values beyond given... * `#11516 <https://github.com/scipy/scipy/issues/11516>`__: Recommend ccache for benchmarks in contributor documentation * `#12017 <https://github.com/scipy/scipy/issues/12017>`__: Expose FACTOR parameter instead of using magic number in scipy.optimize.nnls * `#12748 <https://github.com/scipy/scipy/issues/12748>`__: Convergence issue of GMRES * `#12796 <https://github.com/scipy/scipy/issues/12796>`__: BUG: nonmonotonicity in betaincinv, btdtri, stats.beta.ppf * `#13306 <https://github.com/scipy/scipy/issues/13306>`__: griddata with multiple data values * `#13437 <https://github.com/scipy/scipy/issues/13437>`__: DOC: Add example as to how to use solve_ivp to solve complex... * `#14037 <https://github.com/scipy/scipy/issues/14037>`__: ENH: other quality metrics for random sampling * `#14480 <https://github.com/scipy/scipy/issues/14480>`__: LSODA implementation of dense output yields incorrect result * `#15533 <https://github.com/scipy/scipy/issues/15533>`__: BUG: test failure with MKL in presence of AVX512-capable processor * `#15676 <https://github.com/scipy/scipy/issues/15676>`__: ENH: Implement \`multivariate_normal.fit\` * `#15738 <https://github.com/scipy/scipy/issues/15738>`__: DEP: change default of atol in \`scipy.sparse.linalg.\*\` * `#16729 <https://github.com/scipy/scipy/issues/16729>`__: _fitpack / dfitpack duplication * `#16880 <https://github.com/scipy/scipy/issues/16880>`__: ENH: Add Rotation.align_vector * `#17290 <https://github.com/scipy/scipy/issues/17290>`__: ENH: multi dimensional wasserstein/earth mover distance in Scipy * `#17398 <https://github.com/scipy/scipy/issues/17398>`__: BUG: Documentation for \`scipy.optimize.differential_evolution\`... * `#17462 <https://github.com/scipy/scipy/issues/17462>`__: ENH: Create Rotations by aligning a primary axis and best-fitting... * `#17493 <https://github.com/scipy/scipy/issues/17493>`__: BUG: stats: Occasional failures of some tests of \`levy_stable\` * `#17572 <https://github.com/scipy/scipy/issues/17572>`__: BUG: Deprecation warning says to use non-existent symbols * `#17706 <https://github.com/scipy/scipy/issues/17706>`__: ENH: add isotonic regression * `#17734 <https://github.com/scipy/scipy/issues/17734>`__: BUG: dijkstra algorithm is returning different results in v1.10... * `#17744 <https://github.com/scipy/scipy/issues/17744>`__: BUG: test_maxiter_worsening[lgmres] failed on riscv * `#17756 <https://github.com/scipy/scipy/issues/17756>`__: DOC: dimension clarification in \`directed_hausdorff \` * `#17771 <https://github.com/scipy/scipy/issues/17771>`__: BUG: cannot import ODEintWarning from scipy.integrate * `#17864 <https://github.com/scipy/scipy/issues/17864>`__: ENH: feature request for initial hessian estimate in scipy bfgs... * `#17890 <https://github.com/scipy/scipy/issues/17890>`__: ENH: Stirling Numbers of the second Kind * `#18093 <https://github.com/scipy/scipy/issues/18093>`__: BUG: entropy calculations failing for large values * `#18279 <https://github.com/scipy/scipy/issues/18279>`__: BUG: deprecation warnings for private API are misleading * `#18316 <https://github.com/scipy/scipy/issues/18316>`__: DOC: update scipy.stats.truncnorm.rvs example * `#18389 <https://github.com/scipy/scipy/issues/18389>`__: BUG: Yeo-Johnson Power Transformer gives Numpy warning * `#18404 <https://github.com/scipy/scipy/issues/18404>`__: ENH: Add wrapper for LAPACK functions stgsyl and dtgsyl * `#18432 <https://github.com/scipy/scipy/issues/18432>`__: BUG: levy_stable.pdf does not use pdf_default_method_name attribute * `#18452 <https://github.com/scipy/scipy/issues/18452>`__: BUG: DST 2 and 3 with \`norm="ortho"\` are not orthogonal * `#18457 <https://github.com/scipy/scipy/issues/18457>`__: DOC: clarify that \`prewitt\` filter does not return the magnitude... * `#18506 <https://github.com/scipy/scipy/issues/18506>`__: BUG: Strange behavior of scipy.stats.hypergeom.sf function with... * `#18511 <https://github.com/scipy/scipy/issues/18511>`__: BUG: Incorrect hypergeometric mean calculation * `#18564 <https://github.com/scipy/scipy/issues/18564>`__: BUG: \`rv_discrete\` fails when support is unbounded below * `#18568 <https://github.com/scipy/scipy/issues/18568>`__: BUG: circe-ci eigenvalue-LOBPCG benchmarks do not check accuracy... * `#18577 <https://github.com/scipy/scipy/issues/18577>`__: DEP: deprecate positional arguments in \`sparse.gmres\` * `#18578 <https://github.com/scipy/scipy/issues/18578>`__: DEP: deprecate positional arguments in \`stats.kendalltau\` * `#18579 <https://github.com/scipy/scipy/issues/18579>`__: DEP: deprecate positional arguments in \`firwin\*\` * `#18580 <https://github.com/scipy/scipy/issues/18580>`__: DEP: removal of \`scale='raw'\` in \`iqr\` * `#18581 <https://github.com/scipy/scipy/issues/18581>`__: DEP: removal of \`stats.binom_test\` * `#18582 <https://github.com/scipy/scipy/issues/18582>`__: DEP: removal of parameter \`centered\` of \`stats.qmc.LatinHypercube\` * `#18592 <https://github.com/scipy/scipy/issues/18592>`__: Semantics of sparse array creation functions * `#18637 <https://github.com/scipy/scipy/issues/18637>`__: BUG: Anaconda environment creation with python 3.11 * `#18639 <https://github.com/scipy/scipy/issues/18639>`__: ENH: \`multivariate_normal.rvs\` extremely slow * `#18643 <https://github.com/scipy/scipy/issues/18643>`__: DOC: Problem with the randint description * `#18647 <https://github.com/scipy/scipy/issues/18647>`__: BUG: dgbmv gives "(len(x)>offx+(trans==0?m-1:n-1)\*abs(incx))... * `#18651 <https://github.com/scipy/scipy/issues/18651>`__: DOC: Missing equal sign in equation of Lomb-Scargle periodogram * `#18669 <https://github.com/scipy/scipy/issues/18669>`__: DOC: sparse docs "matrix" and "array" mixup * `#18680 <https://github.com/scipy/scipy/issues/18680>`__: stats: XSLOW test failures in TestFit * `#18702 <https://github.com/scipy/scipy/issues/18702>`__: DEP: deprecate aliased integration methods that are (close to)... * `#18721 <https://github.com/scipy/scipy/issues/18721>`__: TST: Cause of \`test_linsolve.py\` test failure * `#18748 <https://github.com/scipy/scipy/issues/18748>`__: [DOC] Incorrect docstring for \`boxcox_normmax\` argument \`method="mle"\`,... * `#18836 <https://github.com/scipy/scipy/issues/18836>`__: DOC: API reference doesn't point users to replacement for interp1d * `#18838 <https://github.com/scipy/scipy/issues/18838>`__: BUG: stats.burr12: distribution returns invalid moments * `#18839 <https://github.com/scipy/scipy/issues/18839>`__: ENH: concatenating sparse arrays should return sparse arrays * `#18849 <https://github.com/scipy/scipy/issues/18849>`__: BUG: \`python dev.py bench\` fails * `#18860 <https://github.com/scipy/scipy/issues/18860>`__: BUG/BENCH: clough_tocher benchmark fails * `#18864 <https://github.com/scipy/scipy/issues/18864>`__: BUG: curve_fit memoization causes significant overhead in function... * `#18868 <https://github.com/scipy/scipy/issues/18868>`__: CI/BENCH: Fail on error * `#18876 <https://github.com/scipy/scipy/issues/18876>`__: DOC: integrate: clarify that event and jac must have the same... * `#18881 <https://github.com/scipy/scipy/issues/18881>`__: DOC: lobpcg examples improvement * `#18903 <https://github.com/scipy/scipy/issues/18903>`__: DOC: links in hacking.rst go back to hacking.rst instead of their... * `#18939 <https://github.com/scipy/scipy/issues/18939>`__: ENH: add new representations for Sakurai and Mikota matrices... * `#18944 <https://github.com/scipy/scipy/issues/18944>`__: How to change the default setup for using res.plot() * `#18953 <https://github.com/scipy/scipy/issues/18953>`__: BUG: \`\`scipy.optimize._differentialevolution\`\` incorrectly... * `#18966 <https://github.com/scipy/scipy/issues/18966>`__: BUG: \`PyArray_MAX\` no longer available in \`numpy/arrayobject.h\` * `#18981 <https://github.com/scipy/scipy/issues/18981>`__: ENH: Publish musllinux wheels to PyPI * `#18984 <https://github.com/scipy/scipy/issues/18984>`__: DOC: scipy.stats.lognorm — SciPy v1.12.0.dev Manual * `#18987 <https://github.com/scipy/scipy/issues/18987>`__: BUG: reproducible builds problem in _ stats_pythran.cpython-311-x86_64-linux-gnu.so * `#19008 <https://github.com/scipy/scipy/issues/19008>`__: DOC: error in scipy.integrate.DOP853 * `#19009 <https://github.com/scipy/scipy/issues/19009>`__: DOC: Maybe a typo on the parameter 'alternative' in the Mann-Whitney... * `#19024 <https://github.com/scipy/scipy/issues/19024>`__: BUG: #18481 made BFGS minimization less accurate * `#19039 <https://github.com/scipy/scipy/issues/19039>`__: BUG: stats.zscore returns all NaN if given masked array with... * `#19049 <https://github.com/scipy/scipy/issues/19049>`__: TST: add test for gh-17918 * `#19056 <https://github.com/scipy/scipy/issues/19056>`__: ENH: add computing a selected number \`m\` of extreme eigenpairs... * `#19063 <https://github.com/scipy/scipy/issues/19063>`__: DOC: typo in scipy.linalg.pinv.html * `#19075 <https://github.com/scipy/scipy/issues/19075>`__: ENH: Allow to choose line-search parameters in related optimization... * `#19082 <https://github.com/scipy/scipy/issues/19082>`__: MAINT, REL: Programming Language directive * `#19090 <https://github.com/scipy/scipy/issues/19090>`__: BUG: root_scalar (newton method), inconsistent objective function... * `#19092 <https://github.com/scipy/scipy/issues/19092>`__: ENH: update vendored SuperLU version * `#19098 <https://github.com/scipy/scipy/issues/19098>`__: ENH: add method in RootResults of root_scalar and root * `#19102 <https://github.com/scipy/scipy/issues/19102>`__: BUG: \`least_squares\` with \`method='trf'\` with initial params... * `#19117 <https://github.com/scipy/scipy/issues/19117>`__: TST/MAINT: \`AssertionError\`s with fresh build on M1 macOS * `#19118 <https://github.com/scipy/scipy/issues/19118>`__: BUG: \`np\`-coercible array-likes are not accepted with array... * `#19140 <https://github.com/scipy/scipy/issues/19140>`__: DOC: stats: alternative loc-scale handling in levy_stable * `#19147 <https://github.com/scipy/scipy/issues/19147>`__: BUG: asfarray is removed from numpy * `#19150 <https://github.com/scipy/scipy/issues/19150>`__: build warnings from pythran (?) * `#19158 <https://github.com/scipy/scipy/issues/19158>`__: BUG: Source archives have bogus timestamps * `#19161 <https://github.com/scipy/scipy/issues/19161>`__: ENH: allow \`sparse_matrix.transpose(axes=(1, 0))\` * `#19167 <https://github.com/scipy/scipy/issues/19167>`__: BUG: build graph non-determinism * `#19177 <https://github.com/scipy/scipy/issues/19177>`__: TST: SphericalVoronoi tests not sensitive to regions internal... * `#19185 <https://github.com/scipy/scipy/issues/19185>`__: BUG: minimize_scalar not checking entire area specified by bounds * `#19188 <https://github.com/scipy/scipy/issues/19188>`__: DOC: 1D arrays with KDTree * `#19190 <https://github.com/scipy/scipy/issues/19190>`__: TST: skip reason incorrect in some cases, related to array API... * `#19193 <https://github.com/scipy/scipy/issues/19193>`__: DOC: typo in scipy.stats.Covariance * `#19241 <https://github.com/scipy/scipy/issues/19241>`__: BUG: utils from scipy.sparse.sputils are not available from scipy.sparse... * `#19273 <https://github.com/scipy/scipy/issues/19273>`__: Calculation of frequencies w is not accurate when fs is default... * `#19276 <https://github.com/scipy/scipy/issues/19276>`__: MAINT: array types: make \`compliance_scipy\` more strict * `#19332 <https://github.com/scipy/scipy/issues/19332>`__: ENH: help \`skewnorm.fit\` find global minimum * `#19352 <https://github.com/scipy/scipy/issues/19352>`__: CI, BUILD: SciPy build failure with Cython 3.0.3 (bisected) * `#19363 <https://github.com/scipy/scipy/issues/19363>`__: DOC: inconsistent terminology at scipy.interpolate.griddata.html * `#19373 <https://github.com/scipy/scipy/issues/19373>`__: DOC: interpolate: add note to BarycentricInterpolator.add_xi... * `#19396 <https://github.com/scipy/scipy/issues/19396>`__: BUG: L-BFGS-B has surprisingly high memory consumption * `#19398 <https://github.com/scipy/scipy/issues/19398>`__: BUG: Missing assertion in test_filter_design.py * `#19406 <https://github.com/scipy/scipy/issues/19406>`__: CI, MAINT: known CI issue with NumPy main/latest * `#19442 <https://github.com/scipy/scipy/issues/19442>`__: BUG: Error collecting tests due to inconsistent parameterization... * `#19448 <https://github.com/scipy/scipy/issues/19448>`__: DOC:fft: next_fast_len signature is empty in docs * `#19490 <https://github.com/scipy/scipy/issues/19490>`__: MAINT: lint: fail inventory * `#19544 <https://github.com/scipy/scipy/issues/19544>`__: DOC/MAINT: refguide-check errors * `#19565 <https://github.com/scipy/scipy/issues/19565>`__: DOC/DX: \`meson-python\` missing from 'required build dependencies' * `#19568 <https://github.com/scipy/scipy/issues/19568>`__: DOC/DX: \`cd scipy\` missing from 'Building from source' * `#19575 <https://github.com/scipy/scipy/issues/19575>`__: BUG: scipy.ndimage.watershed_ift cost data type is too small... * `#19577 <https://github.com/scipy/scipy/issues/19577>`__: Windows build problems related to fast matrix market parts * `#19599 <https://github.com/scipy/scipy/issues/19599>`__: ENH: Returning NotImplemented when multiplying sparse arrays... * `#19612 <https://github.com/scipy/scipy/issues/19612>`__: ENH: add best cost function evaluation to differential evolution... * `#19616 <https://github.com/scipy/scipy/issues/19616>`__: BUG: Normal Inverse Gaussian numerically unstable * `#19620 <https://github.com/scipy/scipy/issues/19620>`__: _lib: Test error in test_warning_calls_filters because of a Python... * `#19636 <https://github.com/scipy/scipy/issues/19636>`__: DOC: issue in documentation for the callback argument in scipy.optimize.minimize * `#19640 <https://github.com/scipy/scipy/issues/19640>`__: CI, MAINT: pre-release job failures with scipy-openblas32 ************************ Pull requests for 1.12.0 ************************ * `#12680 <https://github.com/scipy/scipy/pull/12680>`__: ENH: stats: add nonparametric one-sample quantile test and CI * `#14103 <https://github.com/scipy/scipy/pull/14103>`__: DOC: integrate: Add complex matrix DE solution to examples * `#14552 <https://github.com/scipy/scipy/pull/14552>`__: BUG: Fix LSODA interpolation scheme * `#17408 <https://github.com/scipy/scipy/pull/17408>`__: ENH: Short-Time Fourier Transform (STFT) Enhancements * `#17452 <https://github.com/scipy/scipy/pull/17452>`__: ENH: Add the __pow__ operator for Rotations * `#17460 <https://github.com/scipy/scipy/pull/17460>`__: ENH: New Rotation method approx_equal() * `#17473 <https://github.com/scipy/scipy/pull/17473>`__: ENH: stats: multivariate Wasserstein distance/EMD * `#17495 <https://github.com/scipy/scipy/pull/17495>`__: DOC: Optimize: update DE documentation * `#17542 <https://github.com/scipy/scipy/pull/17542>`__: ENH: Extend Rotation.align_vectors() to allow an infinite weight,... * `#17697 <https://github.com/scipy/scipy/pull/17697>`__: ENH: special: Improvements for the incomplete beta functions. * `#17719 <https://github.com/scipy/scipy/pull/17719>`__: ENH: Add Chandrupatla's algorithm to optimize._zeros_py.py to... * `#17722 <https://github.com/scipy/scipy/pull/17722>`__: ENH add pava and isotonic_regression * `#17728 <https://github.com/scipy/scipy/pull/17728>`__: ENH: Implemented Rotation.from_davenport and Rotation.as_davenport * `#17757 <https://github.com/scipy/scipy/pull/17757>`__: DOC: clarify input dimensions for \`directed_hausdorff\` method * `#17955 <https://github.com/scipy/scipy/pull/17955>`__: ENH: add simplified fast numerical inversion to stats.sampling * `#18103 <https://github.com/scipy/scipy/pull/18103>`__: ENH: add stirling2 function to \`scipy.special\` * `#18133 <https://github.com/scipy/scipy/pull/18133>`__: BUG: Fix fails of some tests and bad behaviour for x=zeta in... * `#18151 <https://github.com/scipy/scipy/pull/18151>`__: ENH: add cumulative_simpson integration to scipy.integrate * `#18156 <https://github.com/scipy/scipy/pull/18156>`__: ENH: Added loggamma entropy method * `#18197 <https://github.com/scipy/scipy/pull/18197>`__: ENH: Add derivatives for BarycentricInterpolator * `#18219 <https://github.com/scipy/scipy/pull/18219>`__: DEP: integrate: Deprecate \`initial\` values other than None... * `#18348 <https://github.com/scipy/scipy/pull/18348>`__: ENH: add private function to bracket root of monotonic function * `#18361 <https://github.com/scipy/scipy/pull/18361>`__: ENH: add \`fit\` method to \`multivariate_normal\` * `#18363 <https://github.com/scipy/scipy/pull/18363>`__: ENH: stats: add Baumgartner-Weiss-Schindler test * `#18376 <https://github.com/scipy/scipy/pull/18376>`__: ENH: CloughTocher2DInterpolator multiple times with different... * `#18465 <https://github.com/scipy/scipy/pull/18465>`__: ENH: asymptotic expansion for multivariate t entropy * `#18470 <https://github.com/scipy/scipy/pull/18470>`__: ENH: stats.anderson_ksamp: re-add permutation version of test * `#18477 <https://github.com/scipy/scipy/pull/18477>`__: DOC: stats.truncnorm: add example about truncation points * `#18481 <https://github.com/scipy/scipy/pull/18481>`__: MAINT: differentiable fns respect float width. Closes #15602 * `#18488 <https://github.com/scipy/scipy/pull/18488>`__: MAINT:ENH:sparse.linalg: Rewrite iterative solvers in Python,... * `#18492 <https://github.com/scipy/scipy/pull/18492>`__: Add NdBSpline: n-dim tensor product b-spline object * `#18496 <https://github.com/scipy/scipy/pull/18496>`__: ENH: Faster _select_samples in _differential_evolution.py * `#18499 <https://github.com/scipy/scipy/pull/18499>`__: ENH: asymptotic expansion for beta entropy for large a and b * `#18544 <https://github.com/scipy/scipy/pull/18544>`__: ENH: sparse.linalg: Implement matrix_power() * `#18552 <https://github.com/scipy/scipy/pull/18552>`__: DOC: stats.laplace_asymmetric: note relationship between scale... * `#18570 <https://github.com/scipy/scipy/pull/18570>`__: ENH:optimize: Rewrite nnls in Python * `#18571 <https://github.com/scipy/scipy/pull/18571>`__: ENH: linalg: Add wrapper for \`?tgsyl\` * `#18575 <https://github.com/scipy/scipy/pull/18575>`__: REL: set version to 1.12.0.dev0 * `#18585 <https://github.com/scipy/scipy/pull/18585>`__: DOC: stats.rv_discrete: note that default methods are not compatible... * `#18586 <https://github.com/scipy/scipy/pull/18586>`__: ENH: override sf for rdist distribution * `#18587 <https://github.com/scipy/scipy/pull/18587>`__: DEP: signal: deprecate cmplx_sort * `#18589 <https://github.com/scipy/scipy/pull/18589>`__: DEP: remove parameter centered of stats.qmc.LatinHypercube * `#18594 <https://github.com/scipy/scipy/pull/18594>`__: DOC: more explicit example usage of scipy.linalg.lu_factor * `#18602 <https://github.com/scipy/scipy/pull/18602>`__: MAINT: stats.hypergeom.mean: correct for large args * `#18606 <https://github.com/scipy/scipy/pull/18606>`__: ENH: override sf for Pearson3 distribution * `#18609 <https://github.com/scipy/scipy/pull/18609>`__: TST: update nonlin tests for sparse arrays * `#18610 <https://github.com/scipy/scipy/pull/18610>`__: DEP: stats: remove deprecated binom_test * `#18612 <https://github.com/scipy/scipy/pull/18612>`__: DEP: stats.iqr: remove deprecated \`scale='raw'\` * `#18613 <https://github.com/scipy/scipy/pull/18613>`__: BUG: fix ortho mode in DST type 2/3 * `#18614 <https://github.com/scipy/scipy/pull/18614>`__: MAINT: don't override sf for loguniform/reciprocal distribution * `#18616 <https://github.com/scipy/scipy/pull/18616>`__: ENH: override sf in loglaplace distribution * `#18617 <https://github.com/scipy/scipy/pull/18617>`__: MAINT: interpolate: delete duplicated FITPACK bisplev interface * `#18620 <https://github.com/scipy/scipy/pull/18620>`__: MAINT: signal: avoid eval/exec in hilbert2 * `#18622 <https://github.com/scipy/scipy/pull/18622>`__: MAINT: Move rvs_ratio_uniforms to sampling * `#18624 <https://github.com/scipy/scipy/pull/18624>`__: DEP: adjust deprecation of positional arguments * `#18631 <https://github.com/scipy/scipy/pull/18631>`__: ENH: Add \`fast_matrix_market\` to scipy.io * `#18633 <https://github.com/scipy/scipy/pull/18633>`__: BUG: Fix issue levy_stable.pdf * `#18640 <https://github.com/scipy/scipy/pull/18640>`__: MAINT: Fix lint warnings in \`_traversal.pyx\` * `#18641 <https://github.com/scipy/scipy/pull/18641>`__: ENH: makes \`_covariance\` a cached property * `#18649 <https://github.com/scipy/scipy/pull/18649>`__: DEP: scipy.stats.morestats: clarify deprecation warnings * `#18652 <https://github.com/scipy/scipy/pull/18652>`__: BUG: Fixes how the length of argument x of linalg.blas.?gbmv... * `#18653 <https://github.com/scipy/scipy/pull/18653>`__: DOC: Add missing equal sign to Lomb-Scargle periodogram equation * `#18661 <https://github.com/scipy/scipy/pull/18661>`__: CI: update \`SCIPY_NIGHTLY_UPLOAD_TOKEN\` on CirrusCI. * `#18664 <https://github.com/scipy/scipy/pull/18664>`__: ENH: stats.dirichlet: add covariance method * `#18666 <https://github.com/scipy/scipy/pull/18666>`__: DEV: update environment.yml, removing setuptools * `#18667 <https://github.com/scipy/scipy/pull/18667>`__: DEV: Add Windows specific instructions to environment.yml * `#18668 <https://github.com/scipy/scipy/pull/18668>`__: ENH: add machinery to support Array API * `#18670 <https://github.com/scipy/scipy/pull/18670>`__: MAINT: signal: deduplicate \*ord functions * `#18677 <https://github.com/scipy/scipy/pull/18677>`__: TST: linalg: improve seeding of some tests that are a problem... * `#18679 <https://github.com/scipy/scipy/pull/18679>`__: MAINT:stats:Cythonize and remove Fortran statlib code * `#18682 <https://github.com/scipy/scipy/pull/18682>`__: TST: stats: fix xslow test failures * `#18686 <https://github.com/scipy/scipy/pull/18686>`__: DEV: Add ability to run memory benchmarks on macOS * `#18689 <https://github.com/scipy/scipy/pull/18689>`__: CI: test with Python 3.12-beta * `#18690 <https://github.com/scipy/scipy/pull/18690>`__: DOC: mention that \`genlogistic\` is one of several types of... * `#18692 <https://github.com/scipy/scipy/pull/18692>`__: MAINT: signal: replace \`np.r_["-1", arrays]\` by a more sane... * `#18694 <https://github.com/scipy/scipy/pull/18694>`__: DOC: signal: Updated Chebyshev 2 documentation * `#18695 <https://github.com/scipy/scipy/pull/18695>`__: ENH: override halflogistic fit for free parameters * `#18696 <https://github.com/scipy/scipy/pull/18696>`__: MAINT Fix broken link in scipy.optimize._differentialevolution.py * `#18699 <https://github.com/scipy/scipy/pull/18699>`__: DEP: integrate: deprecate old aliases * `#18705 <https://github.com/scipy/scipy/pull/18705>`__: MAINT Optimize link in scipy.optimize._differentialevolution.py * `#18706 <https://github.com/scipy/scipy/pull/18706>`__: DOC: linalg: various doc improvements * `#18708 <https://github.com/scipy/scipy/pull/18708>`__: MAINT: signal/dlsim: avoid using interp1d; use make_interp_spline... * `#18710 <https://github.com/scipy/scipy/pull/18710>`__: MAINT: remove np.r_["-1", ...] * `#18712 <https://github.com/scipy/scipy/pull/18712>`__: MAINT: bump minimal supported NumPy version to 1.22.4 * `#18713 <https://github.com/scipy/scipy/pull/18713>`__: MAINT: fix linter error in \`_fitpack_impl.py\` * `#18714 <https://github.com/scipy/scipy/pull/18714>`__: ENH: Improve beta entropy when one argument is large * `#18715 <https://github.com/scipy/scipy/pull/18715>`__: MAINT: signal: remove duplicated _atleast_2d_or_none * `#18718 <https://github.com/scipy/scipy/pull/18718>`__: TST: mark two tests of \`lsq_linear\` as xslow * `#18719 <https://github.com/scipy/scipy/pull/18719>`__: MAINT: optimize._chandrupatla: result object fixup * `#18720 <https://github.com/scipy/scipy/pull/18720>`__: DOC: Fix notational variation in _lbfgsb_py.py: "pg_i" and "proj... * `#18723 <https://github.com/scipy/scipy/pull/18723>`__: MAINT:sparse.linalg: Use _NoValue for deprecated kwargs * `#18726 <https://github.com/scipy/scipy/pull/18726>`__: ENH: improve halflogistic distribution fitting with fixed parameters * `#18727 <https://github.com/scipy/scipy/pull/18727>`__: MAINT: Add skip CI link to PR template * `#18728 <https://github.com/scipy/scipy/pull/18728>`__: MAINT: optimize._chandrupatla: refactor for code reuse * `#18729 <https://github.com/scipy/scipy/pull/18729>`__: DOC: Add note connecting weibull_min to standard exponential. * `#18734 <https://github.com/scipy/scipy/pull/18734>`__: MAINT: Update codeowners * `#18742 <https://github.com/scipy/scipy/pull/18742>`__: ENH: differential_evolution callback accepts intermediate_result * `#18744 <https://github.com/scipy/scipy/pull/18744>`__: TST:sparse.linalg:Relax test_hermitian_modes tolerances * `#18746 <https://github.com/scipy/scipy/pull/18746>`__: MAINT: forward port 1.11.0 relnotes * `#18755 <https://github.com/scipy/scipy/pull/18755>`__: ENH: override isf for Burr distribution * `#18756 <https://github.com/scipy/scipy/pull/18756>`__: DOC: stats.boxcox_normmax: correct minimize -> maximize * `#18758 <https://github.com/scipy/scipy/pull/18758>`__: DOC: Document performance cliff for scipy.sparse.random * `#18760 <https://github.com/scipy/scipy/pull/18760>`__: ENH: override halfnorm fit * `#18764 <https://github.com/scipy/scipy/pull/18764>`__: ENH: stats: add informative error message to \`boxcox_normmax\`... * `#18771 <https://github.com/scipy/scipy/pull/18771>`__: Fix typo from #18758 * `#18777 <https://github.com/scipy/scipy/pull/18777>`__: DOC: Clarify Prewitt filter * `#18783 <https://github.com/scipy/scipy/pull/18783>`__: ENH: stats: Implement _sf and _isf for hypsecant. * `#18786 <https://github.com/scipy/scipy/pull/18786>`__: MAINT: forward port 1.11.1 relnotes * `#18794 <https://github.com/scipy/scipy/pull/18794>`__: MAINT: fix \`halflogistic.fit\` for bad location guess * `#18795 <https://github.com/scipy/scipy/pull/18795>`__: BUG, DOC: Correct \`adsurl\` value and add a general \`url\`... * `#18799 <https://github.com/scipy/scipy/pull/18799>`__: ENH: simplify \`gausshyper.pdf\` * `#18802 <https://github.com/scipy/scipy/pull/18802>`__: MAINT: Use \`sparse.diags\` instead of \`spdiags\` internally. * `#18803 <https://github.com/scipy/scipy/pull/18803>`__: MAINT: \`rv_discrete\` should raise with duplicate \`xk\` in... * `#18807 <https://github.com/scipy/scipy/pull/18807>`__: ENH: remove unnecessary root-find from skewnorm * `#18808 <https://github.com/scipy/scipy/pull/18808>`__: TST: test for gh-18800 * `#18812 <https://github.com/scipy/scipy/pull/18812>`__: DEP: linalg: sharpen deprecation warning for pinv {,r}cond * `#18814 <https://github.com/scipy/scipy/pull/18814>`__: DOC: update windows instructions and move conda/mamba * `#18815 <https://github.com/scipy/scipy/pull/18815>`__: ENH:stats: Add _isf method to loglaplace * `#18816 <https://github.com/scipy/scipy/pull/18816>`__: ENH:stats: Add _isf method to lognorm * `#18817 <https://github.com/scipy/scipy/pull/18817>`__: DOC: Fix examples in randint description * `#18818 <https://github.com/scipy/scipy/pull/18818>`__: MAINT: Simplify codespaces env activation * `#18819 <https://github.com/scipy/scipy/pull/18819>`__: TST: stats.dgamma.pdf: adjust test that fails intermittently * `#18820 <https://github.com/scipy/scipy/pull/18820>`__: ENH:stats: Add _isf method to pareto * `#18822 <https://github.com/scipy/scipy/pull/18822>`__: ENH:stats: Add _sf and _isf methods to kappa3 * `#18823 <https://github.com/scipy/scipy/pull/18823>`__: ENH:stats: Add _isf method to lomax * `#18824 <https://github.com/scipy/scipy/pull/18824>`__: ENH: override halfcauchy distribution fit * `#18826 <https://github.com/scipy/scipy/pull/18826>`__: DEP: linalg: use _NoValue for eigh/eigvalsh positional argument... * `#18829 <https://github.com/scipy/scipy/pull/18829>`__: ENH: optimize: vectorized minimization of univariate functions * `#18830 <https://github.com/scipy/scipy/pull/18830>`__: DOC: soften wording on import guidelines, mention lazy loading * `#18833 <https://github.com/scipy/scipy/pull/18833>`__: ENH: optimize: release the GIL while computing the LSAP solution * `#18835 <https://github.com/scipy/scipy/pull/18835>`__: DEP: scipy.stats.stats: clarify deprecation warnings * `#18840 <https://github.com/scipy/scipy/pull/18840>`__: MAINT: stats.burr12: moments are undefined when c\*d <= order * `#18841 <https://github.com/scipy/scipy/pull/18841>`__: MAINT: \*sctype\* replace NumPy 2.0 * `#18843 <https://github.com/scipy/scipy/pull/18843>`__: DEP: vendor sklearn's mechanism to deprecate passing kwargs positionally * `#18846 <https://github.com/scipy/scipy/pull/18846>`__: MAINT: Reduce file size of the SVG files included in \`signal.rst\` * `#18847 <https://github.com/scipy/scipy/pull/18847>`__: MAINT: ptp no method for NumPy 2.0 * `#18848 <https://github.com/scipy/scipy/pull/18848>`__: ENH: add compile and link args to \`show_config\` output * `#18850 <https://github.com/scipy/scipy/pull/18850>`__: ENH: improve performance of \`SphericalVoronoi\` area calculation * `#18852 <https://github.com/scipy/scipy/pull/18852>`__: BUG: fix overflow in stats.yeojohnson * `#18853 <https://github.com/scipy/scipy/pull/18853>`__: BENCH: Remove factorial benchmarks * `#18854 <https://github.com/scipy/scipy/pull/18854>`__: MAINT: signal: lighten the notation in lsim * `#18856 <https://github.com/scipy/scipy/pull/18856>`__: MAINT: Remove more runtests.py stuff * `#18858 <https://github.com/scipy/scipy/pull/18858>`__: MAINT: clean up views/strides/dtypes utilities in \`cluster.hierarcy\` * `#18861 <https://github.com/scipy/scipy/pull/18861>`__: MAINT: avoid \`np.deprecate\` and \`np.core\`, add \`normalize_axis_index\`... * `#18862 <https://github.com/scipy/scipy/pull/18862>`__: ENH: sparse arrays for hstack, vstack, bmat, block_diag. New... * `#18863 <https://github.com/scipy/scipy/pull/18863>`__: DEP: remove extra np.deprecate and add docs back * `#18872 <https://github.com/scipy/scipy/pull/18872>`__: DOC: Add relationship between Fisk (log-logistic) and logistic. * `#18873 <https://github.com/scipy/scipy/pull/18873>`__: DOC: add more cross compilation details * `#18885 <https://github.com/scipy/scipy/pull/18885>`__: DOC: minor issues in install commands * `#18886 <https://github.com/scipy/scipy/pull/18886>`__: MAINT/DOC: remove docstring issues in ndimage * `#18887 <https://github.com/scipy/scipy/pull/18887>`__: DOC: tab sync based on same tab name using \`\`sync\`\` statement * `#18888 <https://github.com/scipy/scipy/pull/18888>`__: MAINT: Set RAM requirement for Codespaces * `#18889 <https://github.com/scipy/scipy/pull/18889>`__: CI: Add prerelease build with 64bit OpenBLAS nightly * `#18893 <https://github.com/scipy/scipy/pull/18893>`__: ENH: Geometric quality metrics for random sampling * `#18898 <https://github.com/scipy/scipy/pull/18898>`__: DOC: fix sparse docs "matrix" and "array" mixup * `#18905 <https://github.com/scipy/scipy/pull/18905>`__: DOC: Clarify DOF in f-distribution notes * `#18910 <https://github.com/scipy/scipy/pull/18910>`__: DOC: from CONTRIBUTING.rst link to SciPy website * `#18911 <https://github.com/scipy/scipy/pull/18911>`__: DOC/MAINT: \`special\` doc fixes * `#18914 <https://github.com/scipy/scipy/pull/18914>`__: ENH: Enhance \`dev.py\` by adding command to query PYTHONPATH * `#18916 <https://github.com/scipy/scipy/pull/18916>`__: DEP: stats: improve deprecation of private but present modules * `#18917 <https://github.com/scipy/scipy/pull/18917>`__: ENH: move {c, q}spline_1d to use sosfilt/lfilter * `#18920 <https://github.com/scipy/scipy/pull/18920>`__: DOC: indicate functions with multiple valid solutions in csgraph * `#18927 <https://github.com/scipy/scipy/pull/18927>`__: TST: use Hypothesis for property-based tests * `#18929 <https://github.com/scipy/scipy/pull/18929>`__: ENH: change sparse dok from subclass of dict to have attribute... * `#18930 <https://github.com/scipy/scipy/pull/18930>`__: ENH: _lib._util: make _lazywhere compatible with Array API * `#18931 <https://github.com/scipy/scipy/pull/18931>`__: MAINT: cleanup redundant tests for \`vonmises.fit\` * `#18932 <https://github.com/scipy/scipy/pull/18932>`__: DEP: deprecate positional arguments for some methods in signal._fir_f… * `#18933 <https://github.com/scipy/scipy/pull/18933>`__: DEP: deprecate positional arguments for special.comb * `#18934 <https://github.com/scipy/scipy/pull/18934>`__: DEP: deprecate positional arguments for some methods in sparse.linalg… * `#18935 <https://github.com/scipy/scipy/pull/18935>`__: DEP: deprecate positional arguments for linalg.pinv * `#18936 <https://github.com/scipy/scipy/pull/18936>`__: DEP: deprecate positional arguments for stats.kendalltau * `#18942 <https://github.com/scipy/scipy/pull/18942>`__: CI: Change OpenBLAS nightly wheel location * `#18943 <https://github.com/scipy/scipy/pull/18943>`__: DEP: sparse.linalg: deprecate positional arguments for gcrotmk,... * `#18946 <https://github.com/scipy/scipy/pull/18946>`__: DEP: linalg: deprecate positional args for eigh/eigvalsh * `#18948 <https://github.com/scipy/scipy/pull/18948>`__: ENH: Jones and Faddy Skew-T distribution * `#18949 <https://github.com/scipy/scipy/pull/18949>`__: DOC: stats.FitResult.plot: add example * `#18952 <https://github.com/scipy/scipy/pull/18952>`__: MAINT: Remove \`._is_array\` since there's now: \`isinstance(x,... * `#18954 <https://github.com/scipy/scipy/pull/18954>`__: BENCH: update/rewrite lobpcg benchmark sparse_linalg_lobpcg.py * `#18955 <https://github.com/scipy/scipy/pull/18955>`__: MAINT/DOC: spatial: Document and test the double cover property... * `#18958 <https://github.com/scipy/scipy/pull/18958>`__: MAINT: optimize.differential_evolution: clarify that bounds must... * `#18962 <https://github.com/scipy/scipy/pull/18962>`__: Update name of coverage action * `#18963 <https://github.com/scipy/scipy/pull/18963>`__: ENH:sparse.linalg: Add LaplacianNd class to special sparse arrays * `#18967 <https://github.com/scipy/scipy/pull/18967>`__: MAINT: redefine \`PyArray_MAX\`/\`PyArray_MIN\` because they... * `#18968 <https://github.com/scipy/scipy/pull/18968>`__: DEP: optimize: improve deprecation of private modules named without... * `#18970 <https://github.com/scipy/scipy/pull/18970>`__: DEP: integrate: improve deprecation of private modules named... * `#18973 <https://github.com/scipy/scipy/pull/18973>`__: MAINT: linalg: remove a stray np.cast * `#18975 <https://github.com/scipy/scipy/pull/18975>`__: ENH: Port scipy to use new numpy complex types * `#18977 <https://github.com/scipy/scipy/pull/18977>`__: ENH: Expose c1 and c2 of scalar_search_wolfe to fmin_bfgs * `#18986 <https://github.com/scipy/scipy/pull/18986>`__: ENH: add possibility to fix parameters in multivariate normal... * `#18992 <https://github.com/scipy/scipy/pull/18992>`__: DEP: interpolate: improve deprecation of private modules named... * `#18995 <https://github.com/scipy/scipy/pull/18995>`__: DOC: stats.lognorm: add example of the relationship between norm... * `#18996 <https://github.com/scipy/scipy/pull/18996>`__: ENH: optimize.curve_fit: reduce overhead of lightweight memoization * `#18997 <https://github.com/scipy/scipy/pull/18997>`__: DOC: add \`noexcept\` to \`cython_optimize\` docs * `#19000 <https://github.com/scipy/scipy/pull/19000>`__: MAINT: fft: rename \`test_numpy.py\` to \`test_basic.py\` * `#19005 <https://github.com/scipy/scipy/pull/19005>`__: ENH: fft: support array API standard * `#19007 <https://github.com/scipy/scipy/pull/19007>`__: BLD: Add venv to .gitignore * `#19010 <https://github.com/scipy/scipy/pull/19010>`__: DOC/MAINT: typo: change Azure reference to Cirrus * `#19012 <https://github.com/scipy/scipy/pull/19012>`__: DOC: stats.mannwhitneyu: address apparent sign inconsistency... * `#19013 <https://github.com/scipy/scipy/pull/19013>`__: DOCS: Fixed typo in _fitpack2.py * `#19014 <https://github.com/scipy/scipy/pull/19014>`__: MAINT: copy, array-api compatible utility function * `#19015 <https://github.com/scipy/scipy/pull/19015>`__: ENH: Update numpy exceptions imports * `#19018 <https://github.com/scipy/scipy/pull/19018>`__: TST/BUG: fix array API test skip decorators * `#19023 <https://github.com/scipy/scipy/pull/19023>`__: ENH: special: dispatch to array library * `#19030 <https://github.com/scipy/scipy/pull/19030>`__: MAINT: replace \`optparse\` usage with \`argparse\` * `#19033 <https://github.com/scipy/scipy/pull/19033>`__: MAINT: Fix codespaces setup.sh script * `#19035 <https://github.com/scipy/scipy/pull/19035>`__: DEP: linalg: improve deprecation of private modules named without... * `#19045 <https://github.com/scipy/scipy/pull/19045>`__: MAINT: Change \`add_newdoc\` import and remove \`NINF\` * `#19047 <https://github.com/scipy/scipy/pull/19047>`__: Corrected a grammatical error in the docstring of class DOP853 * `#19050 <https://github.com/scipy/scipy/pull/19050>`__: CI/BLD: make nightly wheels daily closes #19048 * `#19051 <https://github.com/scipy/scipy/pull/19051>`__: ENH: array agnostic \`cov\`, used in \`cluster\` * `#19052 <https://github.com/scipy/scipy/pull/19052>`__: BENCH: update asv, fail benchmark CI on errors, fix fallout * `#19053 <https://github.com/scipy/scipy/pull/19053>`__: CI: cirrus mods * `#19055 <https://github.com/scipy/scipy/pull/19055>`__: MAINT: use \`copy\` utility from #19014 in \`cluster\` * `#19057 <https://github.com/scipy/scipy/pull/19057>`__: ENH: add computing a selected number m of extreme eigenpairs... * `#19060 <https://github.com/scipy/scipy/pull/19060>`__: MAINT: port minpack2.dcsrch from Fortran to Python, remove Fortran... * `#19061 <https://github.com/scipy/scipy/pull/19061>`__: DOC: spell NaN -> nan * `#19062 <https://github.com/scipy/scipy/pull/19062>`__: BLD: Replace complex occurences with singlecomplex in SuperLU * `#19064 <https://github.com/scipy/scipy/pull/19064>`__: TST: bump tolerance of \`TestTruncexpon.test_is_isf\` a little * `#19066 <https://github.com/scipy/scipy/pull/19066>`__: Update dfovec.py case 20 for zero indexing * `#19067 <https://github.com/scipy/scipy/pull/19067>`__: DEP: remove all deprecated \`scipy.<numpy-func>\` objects * `#19070 <https://github.com/scipy/scipy/pull/19070>`__: DEP: odr: improved deprecation of private modules * `#19073 <https://github.com/scipy/scipy/pull/19073>`__: DOC: remove reference to no longer imported numpy funcs * `#19074 <https://github.com/scipy/scipy/pull/19074>`__: TST: add regression test for sqrtm unexpectedly converting to... * `#19078 <https://github.com/scipy/scipy/pull/19078>`__: ENH: Reflect changes from numpy namespace refactor part 3 * `#19081 <https://github.com/scipy/scipy/pull/19081>`__: ENH: optimize: c1 and c2 of scalar_search_wolfe to cg and newton-cg * `#19083 <https://github.com/scipy/scipy/pull/19083>`__: MAINT: forward port 1.11.2 relnotes * `#19086 <https://github.com/scipy/scipy/pull/19086>`__: BLD: tweaks to build dependencies and 3.12 classifier in pyproject.toml * `#19087 <https://github.com/scipy/scipy/pull/19087>`__: DOC: add info about usage of ccache for benchmarks * `#19089 <https://github.com/scipy/scipy/pull/19089>`__: DOC: add comment on args in solve_ivp (#18876) * `#19091 <https://github.com/scipy/scipy/pull/19091>`__: MAINT: optimize.root_scalar: handle args when method is newton... * `#19093 <https://github.com/scipy/scipy/pull/19093>`__: MAINT: stats.zscore: return correct result when 0th element is... * `#19094 <https://github.com/scipy/scipy/pull/19094>`__: BUG: optimize: update incorrect c1 value * `#19095 <https://github.com/scipy/scipy/pull/19095>`__: ENH: optimize: add inv Hess estimate in BFGS * `#19100 <https://github.com/scipy/scipy/pull/19100>`__: MAINT: Move \`trapezoid\` implementation to SciPy * `#19104 <https://github.com/scipy/scipy/pull/19104>`__: MAINT: fix the CT subclass benchmark * `#19110 <https://github.com/scipy/scipy/pull/19110>`__: MAINT: Remove usages of np.row_stack and np.in1d * `#19113 <https://github.com/scipy/scipy/pull/19113>`__: MAINT, TST: default_rng few tests * `#19115 <https://github.com/scipy/scipy/pull/19115>`__: ENH, TST: Use explicit imports in docstring examples * `#19127 <https://github.com/scipy/scipy/pull/19127>`__: DOC: linalg: fix typos in pinv docs * `#19128 <https://github.com/scipy/scipy/pull/19128>`__: ENH: optimize: add method attribute to result object from root_scalar... * `#19129 <https://github.com/scipy/scipy/pull/19129>`__: DOC:Added Examples To \`combine_pvalues\` and \`fhtoffset\` * `#19136 <https://github.com/scipy/scipy/pull/19136>`__: MAINT: remove NumPy infty alias * `#19141 <https://github.com/scipy/scipy/pull/19141>`__: MAINT: more NumPy API shims * `#19143 <https://github.com/scipy/scipy/pull/19143>`__: MAINT: remove asfarray usage * `#19146 <https://github.com/scipy/scipy/pull/19146>`__: DOC: sparse.linalg: Enhance lobpcg documentation with updated... * `#19152 <https://github.com/scipy/scipy/pull/19152>`__: DOC: fix predeces(s)or typo. * `#19157 <https://github.com/scipy/scipy/pull/19157>`__: TST: add array-agnostic assertions * `#19162 <https://github.com/scipy/scipy/pull/19162>`__: ENH: Allow \`sparse_mtx.transpose(axes=(1, 0))\` * `#19165 <https://github.com/scipy/scipy/pull/19165>`__: DEP: improve deprecation warnings for constants module * `#19166 <https://github.com/scipy/scipy/pull/19166>`__: MAINT: Reflect changes from \`numpy\` namespace refactor Part... * `#19168 <https://github.com/scipy/scipy/pull/19168>`__: BLD: special: fix dependencies for \`_ellip_harm_2\` * `#19170 <https://github.com/scipy/scipy/pull/19170>`__: MAINT: fix a number of issues in Cython code * `#19171 <https://github.com/scipy/scipy/pull/19171>`__: ENH: constructors for sparse arrays * `#19172 <https://github.com/scipy/scipy/pull/19172>`__: MAINT: Spherical Voronoi sort simplify * `#19173 <https://github.com/scipy/scipy/pull/19173>`__: MAINT: integrate._tanhsinh: improvements after further testing * `#19175 <https://github.com/scipy/scipy/pull/19175>`__: DOC: stats: Correct levy_stable loc-scale note * `#19178 <https://github.com/scipy/scipy/pull/19178>`__: MAINT/DOC: add \`fit\` to multivariate normal methods in documentation * `#19179 <https://github.com/scipy/scipy/pull/19179>`__: DOC: Added example for tf2zpk function * `#19183 <https://github.com/scipy/scipy/pull/19183>`__: MAINT: Import \`fromarrays\` from \`rec\` * `#19184 <https://github.com/scipy/scipy/pull/19184>`__: TST: fix sparse constructor test with large memory footprint * `#19186 <https://github.com/scipy/scipy/pull/19186>`__: TST: _lib: improve array API assertions * `#19187 <https://github.com/scipy/scipy/pull/19187>`__: BUG: accept \`np\`-coercible array-likes with array API flag... * `#19191 <https://github.com/scipy/scipy/pull/19191>`__: BLD: update minimum versions of meson-python and pythran * `#19192 <https://github.com/scipy/scipy/pull/19192>`__: TST: test tolerance bumps to fix reported failures * `#19194 <https://github.com/scipy/scipy/pull/19194>`__: BUG/TST: fix \`test_array_api\` for \`cupy\` * `#19195 <https://github.com/scipy/scipy/pull/19195>`__: DOC: Fixed typo in scipy.stats.Covariance * `#19198 <https://github.com/scipy/scipy/pull/19198>`__: DOC: optimize.minimize_scalar: note limitations and suggest alternatives * `#19201 <https://github.com/scipy/scipy/pull/19201>`__: DOC: missing backticks * `#19202 <https://github.com/scipy/scipy/pull/19202>`__: DOC: remove old mention to Python 2. * `#19203 <https://github.com/scipy/scipy/pull/19203>`__: DOC: Fix _nan_allsame example prompt. * `#19204 <https://github.com/scipy/scipy/pull/19204>`__: DOC: missing (and stray) backtick in mannwhitneyu * `#19206 <https://github.com/scipy/scipy/pull/19206>`__: TST, MAINT: skip fix for array API tests * `#19217 <https://github.com/scipy/scipy/pull/19217>`__: BLD: start building wheels against numpy 2.0 [wheel build] * `#19218 <https://github.com/scipy/scipy/pull/19218>`__: DOC: Added examples for zpk2tf, tf2sos, sos2tf, lp2lp_zpk, lp2hp_zpk,... * `#19219 <https://github.com/scipy/scipy/pull/19219>`__: ENH: Implemented custom mutation / crossover functions for DifferentialEvolution * `#19222 <https://github.com/scipy/scipy/pull/19222>`__: DOC: clarify guidance on style modifications * `#19224 <https://github.com/scipy/scipy/pull/19224>`__: BUG: integrate: expose ODEintWarning * `#19226 <https://github.com/scipy/scipy/pull/19226>`__: DEP: improve message for deprecated private API for \`sparse\` * `#19227 <https://github.com/scipy/scipy/pull/19227>`__: DOC: Trailing character after closing backquote. * `#19229 <https://github.com/scipy/scipy/pull/19229>`__: DOC: Fix repeated typos in dev.py * `#19233 <https://github.com/scipy/scipy/pull/19233>`__: DOC: Fix character after trailing backticks * `#19236 <https://github.com/scipy/scipy/pull/19236>`__: BUG: Remove removed item from \`__all__\` * `#19237 <https://github.com/scipy/scipy/pull/19237>`__: DOC: Missing import in private _nan_allsame function docs. * `#19240 <https://github.com/scipy/scipy/pull/19240>`__: DEP: spatial: improve deprecation of private modules without... * `#19243 <https://github.com/scipy/scipy/pull/19243>`__: DOC: stats: fix documentation of warnings raised by pearsonr * `#19244 <https://github.com/scipy/scipy/pull/19244>`__: MAINT: scipy.io NEP 50 shims * `#19247 <https://github.com/scipy/scipy/pull/19247>`__: MAINT: NumPy version check removals * `#19251 <https://github.com/scipy/scipy/pull/19251>`__: TST/MAINT: cluster: use new array API assertions * `#19253 <https://github.com/scipy/scipy/pull/19253>`__: MAINT: Cleanup expired ndarray methods * `#19256 <https://github.com/scipy/scipy/pull/19256>`__: DOC: Added examples for correlation, num_obs_dm, num_obs_y functions. * `#19259 <https://github.com/scipy/scipy/pull/19259>`__: DOC: outline workaround when precision losses occur in BFGS [skip... * `#19261 <https://github.com/scipy/scipy/pull/19261>`__: MAINT: fft: refactor basic backend with \`_execute_1D\` * `#19262 <https://github.com/scipy/scipy/pull/19262>`__: MAINT: fft: clean up test-skips * `#19263 <https://github.com/scipy/scipy/pull/19263>`__: ENH: fft: GPU support for non-standard basic transforms * `#19264 <https://github.com/scipy/scipy/pull/19264>`__: MAINT/TST: fft: remove duplicate namespace and dtype checks * `#19265 <https://github.com/scipy/scipy/pull/19265>`__: MAINT: array API: rename \`arg_err_msg\` and move to \`_lib\` * `#19266 <https://github.com/scipy/scipy/pull/19266>`__: ENH: use inplace add in _minimize_newtoncg * `#19267 <https://github.com/scipy/scipy/pull/19267>`__: MAINT: use linalg.norm in \`_minimize_newtoncg\` * `#19270 <https://github.com/scipy/scipy/pull/19270>`__: MAINT: fix -Wdeprecated-non-prototype warnings in Cephes code * `#19271 <https://github.com/scipy/scipy/pull/19271>`__: DOC: add example for \`ndimage.shift\` * `#19275 <https://github.com/scipy/scipy/pull/19275>`__: MAINT: Add warning filter for \`numpy.core\` rename * `#19278 <https://github.com/scipy/scipy/pull/19278>`__: DOC: use \`np.copysign()\` instead of \`np.sign()\` * `#19281 <https://github.com/scipy/scipy/pull/19281>`__: DOC: add \`fft\` to list of array API supporting modules * `#19282 <https://github.com/scipy/scipy/pull/19282>`__: MAINT: fft: clean up assertions * `#19284 <https://github.com/scipy/scipy/pull/19284>`__: ENH: update SuperLU to version 6.0.1 * `#19285 <https://github.com/scipy/scipy/pull/19285>`__: MAINT fix more \`-Wdeprecate-non-prototype\` warnings * `#19287 <https://github.com/scipy/scipy/pull/19287>`__: ENH: add exact=False support for stirling2 * `#19289 <https://github.com/scipy/scipy/pull/19289>`__: BLD: stats: improve build config for \`unuran_wrapper\` * `#19293 <https://github.com/scipy/scipy/pull/19293>`__: MAINT: optimize._bracket_root: refactor and comment for clarity * `#19294 <https://github.com/scipy/scipy/pull/19294>`__: MAINT: replace IOError alias with OSError * `#19295 <https://github.com/scipy/scipy/pull/19295>`__: DOC: fix typos found by codespell * `#19296 <https://github.com/scipy/scipy/pull/19296>`__: MAINT: fix linter issues * `#19298 <https://github.com/scipy/scipy/pull/19298>`__: MAINT: replace io.open alias with built-in open * `#19302 <https://github.com/scipy/scipy/pull/19302>`__: MAINT: Fix/parametrize Rotation tests. * `#19304 <https://github.com/scipy/scipy/pull/19304>`__: DOC: fix page typo in ks_2samp reference * `#19310 <https://github.com/scipy/scipy/pull/19310>`__: MAINT: Remove \`np.int_\` and \`np.uint\` * `#19311 <https://github.com/scipy/scipy/pull/19311>`__: DOC: invalid interpreted-text (missing space after closing backtick). * `#19312 <https://github.com/scipy/scipy/pull/19312>`__: DOC: Misc formatting error. * `#19313 <https://github.com/scipy/scipy/pull/19313>`__: DOC: And new line at end of line in CPP code. * `#19314 <https://github.com/scipy/scipy/pull/19314>`__: DOC: Rst don't like list without blanklines. * `#19315 <https://github.com/scipy/scipy/pull/19315>`__: MAINT: forward port 1.11.3 relnotes * `#19317 <https://github.com/scipy/scipy/pull/19317>`__: MAINT: remove support for building with setup.py * `#19327 <https://github.com/scipy/scipy/pull/19327>`__: ENH: Add a numerically stable logpdf function for the skewnorm... * `#19331 <https://github.com/scipy/scipy/pull/19331>`__: MAINT, BLD: more setup.py cleanups * `#19333 <https://github.com/scipy/scipy/pull/19333>`__: ENH: Fit skewnorms with MLE more robustly * `#19340 <https://github.com/scipy/scipy/pull/19340>`__: DEP: deprecate scipy.signal wavelets * `#19341 <https://github.com/scipy/scipy/pull/19341>`__: MAINT: Remove remaining \`numpy.int_\` and filter \`np.long\`... * `#19346 <https://github.com/scipy/scipy/pull/19346>`__: MAINT: stats: fix \`differential_entropy\` too small error behavior * `#19347 <https://github.com/scipy/scipy/pull/19347>`__: BUG: fix ValueError in stats.truncpareto * `#19349 <https://github.com/scipy/scipy/pull/19349>`__: MAINT: Minor fix to shim import of scipy.signal. * `#19350 <https://github.com/scipy/scipy/pull/19350>`__: ENH: stats: add support for masked arrays, \`nan_policy\`, and... * `#19358 <https://github.com/scipy/scipy/pull/19358>`__: TST: SphericalVoronoi region type * `#19360 <https://github.com/scipy/scipy/pull/19360>`__: BLD, MAINT: more build system cleanups * `#19367 <https://github.com/scipy/scipy/pull/19367>`__: CI: avoid using Cython 3.0.3 * `#19371 <https://github.com/scipy/scipy/pull/19371>`__: ENH: stats.ansari: add axis / nan_policy / keepdims support * `#19376 <https://github.com/scipy/scipy/pull/19376>`__: TST: bump tolerance of \`test_al_mohy_higham_2012_experiment_1\` * `#19380 <https://github.com/scipy/scipy/pull/19380>`__: DEV/CI: use scipy-openblas32 wheels in dev.py and some CI jobs * `#19385 <https://github.com/scipy/scipy/pull/19385>`__: ENH: beta negative binomial distribution * `#19390 <https://github.com/scipy/scipy/pull/19390>`__: DOC, MAINT: more distutils/build cleanups * `#19391 <https://github.com/scipy/scipy/pull/19391>`__: MAINT: spatial: Change error message in KDTree to be more informative. * `#19392 <https://github.com/scipy/scipy/pull/19392>`__: ENH: stats.bartlett: add axis / nan_policy / keepdims support * `#19394 <https://github.com/scipy/scipy/pull/19394>`__: MAINT: Partially revert \`np.int_\` changes * `#19395 <https://github.com/scipy/scipy/pull/19395>`__: DEP: ndimage: improve the deprecation of private modules named... * `#19399 <https://github.com/scipy/scipy/pull/19399>`__: TST: signal: add missing assertion in test_filter_design.py * `#19402 <https://github.com/scipy/scipy/pull/19402>`__: MAINT: l-bfgs-b remove intermediate bounds array * `#19409 <https://github.com/scipy/scipy/pull/19409>`__: MAINT: fix a C++ build error with GCC 13 * `#19410 <https://github.com/scipy/scipy/pull/19410>`__: DOC: interpolate: add note to BarycentricInterpolator.add_xi * `#19412 <https://github.com/scipy/scipy/pull/19412>`__: ENH: stats: add support for masked arrays for circular statistics... * `#19414 <https://github.com/scipy/scipy/pull/19414>`__: BLD: with-scipy-openblas instead of use-scipy-openblas * `#19419 <https://github.com/scipy/scipy/pull/19419>`__: ENH: SphericalVoronoi sort faster * `#19422 <https://github.com/scipy/scipy/pull/19422>`__: DOC: Fix trivial typo. * `#19425 <https://github.com/scipy/scipy/pull/19425>`__: ENH: stats: add masked array, axis tuple, and nan policy support... * `#19426 <https://github.com/scipy/scipy/pull/19426>`__: MAINT: Update \`fft.helper\` import * `#19428 <https://github.com/scipy/scipy/pull/19428>`__: ENH: stats.levene/fligner: add axis / nan_policy / keepdims support * `#19429 <https://github.com/scipy/scipy/pull/19429>`__: DOC: Fix documentation of callback function signature of scipy.optimize.minimize * `#19431 <https://github.com/scipy/scipy/pull/19431>`__: MAINT: more NEP 50 shims * `#19433 <https://github.com/scipy/scipy/pull/19433>`__: DOC: Add example for \`stats.hdquantiles\` * `#19435 <https://github.com/scipy/scipy/pull/19435>`__: MAINT: Translate lambertw_scalar into C++ * `#19439 <https://github.com/scipy/scipy/pull/19439>`__: BLD: update build dependency versions in pyproject.toml * `#19440 <https://github.com/scipy/scipy/pull/19440>`__: DOC: remove unused parameter in spline_filter docstring * `#19443 <https://github.com/scipy/scipy/pull/19443>`__: TST: Fix #19442 minimally * `#19445 <https://github.com/scipy/scipy/pull/19445>`__: TST: Remove some unnecessary tuple conversions * `#19449 <https://github.com/scipy/scipy/pull/19449>`__: DOC/MAINT: fft: fix signature for next_fast_len * `#19453 <https://github.com/scipy/scipy/pull/19453>`__: Added parenthesis for accurate calculation of frequencies w when... * `#19458 <https://github.com/scipy/scipy/pull/19458>`__: DOC: spatial: Rotation docs updates * `#19461 <https://github.com/scipy/scipy/pull/19461>`__: DOC: dev: update max line length to 88 characters * `#19465 <https://github.com/scipy/scipy/pull/19465>`__: DOC: fix broken NetCDF URL * `#19466 <https://github.com/scipy/scipy/pull/19466>`__: MAINT: NumPy int type shims * `#19469 <https://github.com/scipy/scipy/pull/19469>`__: DEP: special: improve deprecation of private modules named without... * `#19471 <https://github.com/scipy/scipy/pull/19471>`__: MAINT: Translate binom to C++ * `#19472 <https://github.com/scipy/scipy/pull/19472>`__: MAINT: Update fast_matrix_market to 1.7.4 * `#19482 <https://github.com/scipy/scipy/pull/19482>`__: MAINT: Make use of sf_error conditional on flag in scipy::special... * `#19483 <https://github.com/scipy/scipy/pull/19483>`__: ENH: add max dist to NearestNDInterpolator * `#19484 <https://github.com/scipy/scipy/pull/19484>`__: CI: limit scipy-openblas32 wheel to 0.3.23.293.2 * `#19487 <https://github.com/scipy/scipy/pull/19487>`__: ENH: stats.cramervonmises/epps_singleton: add axis / nan_policy... * `#19489 <https://github.com/scipy/scipy/pull/19489>`__: MAINT: lint: ignore E501 (line length) by subpackage * `#19491 <https://github.com/scipy/scipy/pull/19491>`__: MAINT/STY: misc: remove E501 (line length) lint ignore * `#19494 <https://github.com/scipy/scipy/pull/19494>`__: MAINT/STY: \`conftest.py\`: remove E501 (line length) lint ignore * `#19495 <https://github.com/scipy/scipy/pull/19495>`__: MAINT: get rid of Python 2 left-overs * `#19497 <https://github.com/scipy/scipy/pull/19497>`__: CI: add '[lint only]', '[docs only]' skip tags * `#19501 <https://github.com/scipy/scipy/pull/19501>`__: TST: skip test_bootstrap_against_theory * `#19503 <https://github.com/scipy/scipy/pull/19503>`__: MAINT/STY: fftpack: remove E501 (line length) lint ignore * `#19505 <https://github.com/scipy/scipy/pull/19505>`__: MAINT/STY: \`scipyoptdoc.py\`: remove E501 (line length) lint... * `#19506 <https://github.com/scipy/scipy/pull/19506>`__: MAINT: sparse: update tests to switch to format and toarray from... * `#19507 <https://github.com/scipy/scipy/pull/19507>`__: MAINT/DOC: stats: fix lint errors * `#19508 <https://github.com/scipy/scipy/pull/19508>`__: MAINT/DOC/BENCH: optimize: fix linter errors * `#19509 <https://github.com/scipy/scipy/pull/19509>`__: MAINT: fix remaining lint errors * `#19510 <https://github.com/scipy/scipy/pull/19510>`__: DEP: integrate: deprecate romberg and quadrature * `#19514 <https://github.com/scipy/scipy/pull/19514>`__: MAINT/STY: odr: remove E501 (line length) lint ignore * `#19515 <https://github.com/scipy/scipy/pull/19515>`__: MAINT: Restructure extra_special (working name) as internal library * `#19516 <https://github.com/scipy/scipy/pull/19516>`__: MAINT: lint: enable UP rules * `#19519 <https://github.com/scipy/scipy/pull/19519>`__: DOC: Fix example used in stats tutorial * `#19520 <https://github.com/scipy/scipy/pull/19520>`__: MAINT/STY: fft: remove E501 (line length) lint ignore, \`noqa\`... * `#19523 <https://github.com/scipy/scipy/pull/19523>`__: MAINT/STY: constants: remove E501 (line length) lint ignore * `#19525 <https://github.com/scipy/scipy/pull/19525>`__: DOC, CI: refguide may vary locs * `#19527 <https://github.com/scipy/scipy/pull/19527>`__: DOC, MAINT: more refguide fixes * `#19529 <https://github.com/scipy/scipy/pull/19529>`__: MAINT: blanket \`noqa\` vet * `#19533 <https://github.com/scipy/scipy/pull/19533>`__: DOC: fix more typos found by codespell * `#19534 <https://github.com/scipy/scipy/pull/19534>`__: MAINT: get rid of more Python 2 left-overs * `#19540 <https://github.com/scipy/scipy/pull/19540>`__: DOC: _binomtest.py: Add clearer error messages * `#19546 <https://github.com/scipy/scipy/pull/19546>`__: MAINT: Fix interpolate.make_smoothing_spline's size validation... * `#19556 <https://github.com/scipy/scipy/pull/19556>`__: DOC: revert part of bcbb18f / #19533 * `#19558 <https://github.com/scipy/scipy/pull/19558>`__: MAINT: forward port 1.11.4 relnotes * `#19562 <https://github.com/scipy/scipy/pull/19562>`__: MAINT: Fix typos in Rotation.from_mrp referenced paper * `#19564 <https://github.com/scipy/scipy/pull/19564>`__: MAINT: Mark in \`.gitignore\` the directories instead of their... * `#19570 <https://github.com/scipy/scipy/pull/19570>`__: DOC: Fixed build instructions in index.rst (#19568) * `#19574 <https://github.com/scipy/scipy/pull/19574>`__: DOC: update Building from source docs for editable installs * `#19576 <https://github.com/scipy/scipy/pull/19576>`__: BUG: Use uint32 for cost in NI_WatershedElement * `#19578 <https://github.com/scipy/scipy/pull/19578>`__: BUG: stats._axis_nan_policy_factory: respect too_small when detecting... * `#19579 <https://github.com/scipy/scipy/pull/19579>`__: BLD:MAINT:Update dev.py message * `#19581 <https://github.com/scipy/scipy/pull/19581>`__: BUG: interpolate/ndbspline: fix OOB access for len(tx) != len(ty)... * `#19583 <https://github.com/scipy/scipy/pull/19583>`__: ENH: Use \`np.dot\` to speedup \`spatial.distance.correlation\` * `#19585 <https://github.com/scipy/scipy/pull/19585>`__: DEP: improve deprecation of remaining private modules named without... * `#19588 <https://github.com/scipy/scipy/pull/19588>`__: BLD:Add missing cstdin include in fast matrix market * `#19589 <https://github.com/scipy/scipy/pull/19589>`__: ENH: Use \`np.dot\` to speedup \`spatial.distance.hamming\` * `#19590 <https://github.com/scipy/scipy/pull/19590>`__: MAINT:special:Add more noexcept to Cython signatures * `#19591 <https://github.com/scipy/scipy/pull/19591>`__: MAINT: interpolate/RGI: avoid unnecessary validation of data... * `#19593 <https://github.com/scipy/scipy/pull/19593>`__: Returning NotImplemented in comparisons with sparse arrays * `#19597 <https://github.com/scipy/scipy/pull/19597>`__: TST: special._sinpi/_cospi: skip failing \`test_intermediate_overflow\` * `#19600 <https://github.com/scipy/scipy/pull/19600>`__: Fix sparse multiply * `#19601 <https://github.com/scipy/scipy/pull/19601>`__: ENH: Make special C++ implementations work on CUDA (and beyond!) * `#19603 <https://github.com/scipy/scipy/pull/19603>`__: DOC: doc shim for eig normalization * `#19604 <https://github.com/scipy/scipy/pull/19604>`__: BUG: fix overflow in stats.boxcox_normmax when method='mle' * `#19606 <https://github.com/scipy/scipy/pull/19606>`__: MAINT: array types: restrict to boolean & numerical dtypes * `#19608 <https://github.com/scipy/scipy/pull/19608>`__: MAINT/DOC: refguide-check fixes * `#19609 <https://github.com/scipy/scipy/pull/19609>`__: MAINT: lint: enable line-length check (package-wide) * `#19611 <https://github.com/scipy/scipy/pull/19611>`__: DOC: clarify interpolator termionolgy * `#19613 <https://github.com/scipy/scipy/pull/19613>`__: MAINT: Added clang-format formatting for special C++ files * `#19617 <https://github.com/scipy/scipy/pull/19617>`__: fix numerical stability * `#19618 <https://github.com/scipy/scipy/pull/19618>`__: MAINT: qmc permutations win type * `#19622 <https://github.com/scipy/scipy/pull/19622>`__: CI: undo cython==0.29.35 pin for the 32-bit Linux job, remove... * `#19623 <https://github.com/scipy/scipy/pull/19623>`__: MAINT: lint: enable \`stacklevel\` warnings check * `#19624 <https://github.com/scipy/scipy/pull/19624>`__: MAINT/TST: _lib: use value instead of deprecated s * `#19626 <https://github.com/scipy/scipy/pull/19626>`__: MAINT: more SciPy windows int shims * `#19628 <https://github.com/scipy/scipy/pull/19628>`__: DOC: 1.12.0 release notes * `#19635 <https://github.com/scipy/scipy/pull/19635>`__: MAINT: simplify Nakagami mean calculation * `#19637 <https://github.com/scipy/scipy/pull/19637>`__: DOC: Clarify integration error bound in \`integrate\` tutorial * `#19648 <https://github.com/scipy/scipy/pull/19648>`__: MAINT: simplify chi distribution mean calculation * `#19651 <https://github.com/scipy/scipy/pull/19651>`__: MAINT: stats.entropy: silence mypy * `#19656 <https://github.com/scipy/scipy/pull/19656>`__: BUG: Remove unnecessary incref on unrelated type * `#19658 <https://github.com/scipy/scipy/pull/19658>`__: MAINT: git blame ignores for lint clean-ups * `#19660 <https://github.com/scipy/scipy/pull/19660>`__: STY: special: use indent width of 4 in clang-format * `#19661 <https://github.com/scipy/scipy/pull/19661>`__: CI: fix pre-release job by correct version pin for scipy-openblas32 * `#19677 <https://github.com/scipy/scipy/pull/19677>`__: DOC: array types: mention partial support in \`special\` * `#19686 <https://github.com/scipy/scipy/pull/19686>`__: TST: fix incorrect signal.sosfilt tests * `#19690 <https://github.com/scipy/scipy/pull/19690>`__: BLD: avoid fast-math for oneAPI compilers, fix up handling of... * `#19691 <https://github.com/scipy/scipy/pull/19691>`__: BUG: fix negative overflow in stats.boxcox_normmax * `#19693 <https://github.com/scipy/scipy/pull/19693>`__: BUG: Prevent mutation of \`w\` parameter in \`spatial.distance.\*\` * `#19702 <https://github.com/scipy/scipy/pull/19702>`__: DEP: Adopt \`\*tol\` deprecations also for \`gcrotmk/lgmres/minres/tfqmr\` Checksums ========= MD5 ~~~ 3a240b8500dbd61f81c738a72912be1b Changelog fe705a1a222d86c178f060f26f39fc06 README.txt 885907212172fa7ad0fff3170f2b4f13 scipy-1.12.0rc1-cp310-cp310-macosx_10_9_x86_64.whl 03a7bf978ec11435d71030d621d40d9e scipy-1.12.0rc1-cp310-cp310-macosx_12_0_arm64.whl 9715069ca0e76b84b73688905a84a3f4 scipy-1.12.0rc1-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl 43373136c7b91834822ead0b4de2b965 scipy-1.12.0rc1-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl 3de1ecf1c2e32460303f3d218350b03a scipy-1.12.0rc1-cp310-cp310-musllinux_1_1_x86_64.whl 1cefe0b5202049e33ecd2ac0b7a1928a scipy-1.12.0rc1-cp310-cp310-win_amd64.whl 8194c755ce714ae1cba2066ce5af9281 scipy-1.12.0rc1-cp311-cp311-macosx_10_9_x86_64.whl 88cdfc9aee1c24b2476f650f69135cfa scipy-1.12.0rc1-cp311-cp311-macosx_12_0_arm64.whl f5223a5c0a835793fe06fab073713f29 scipy-1.12.0rc1-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl 4a86f000117fce12236f4ff8488f99b3 scipy-1.12.0rc1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl 43c7f91c847f451687d6159352fb1d25 scipy-1.12.0rc1-cp311-cp311-musllinux_1_1_x86_64.whl fdca37067f8abc153a582e55b2efb4bb scipy-1.12.0rc1-cp311-cp311-win_amd64.whl 9d0ef5585aa4cb8e4df4bffc634dd293 scipy-1.12.0rc1-cp312-cp312-macosx_10_9_x86_64.whl 8aeba1ecaef49cc515345cca9959dc0c scipy-1.12.0rc1-cp312-cp312-macosx_12_0_arm64.whl fc39a4e802933ee434b96189961ac60e scipy-1.12.0rc1-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl e336ba0a1460b48ce0d2a0204d8a007d scipy-1.12.0rc1-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl 2a7700fd76aac9c4682fa7e21fcff7fa scipy-1.12.0rc1-cp312-cp312-musllinux_1_1_x86_64.whl 37ac6bf2459338d8fbca3bcd1f48f445 scipy-1.12.0rc1-cp312-cp312-win_amd64.whl cbac45dcfa4e0f2f1f21d7e34a713141 scipy-1.12.0rc1-cp39-cp39-macosx_10_9_x86_64.whl 6ff21d2ed00248f11170d0926b8a9141 scipy-1.12.0rc1-cp39-cp39-macosx_12_0_arm64.whl 9cc67b976369e068de6dcb2726ea06e5 scipy-1.12.0rc1-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl 6690d768b953edebf48f99e743074a7a scipy-1.12.0rc1-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl 2d3eca17ab5b07f5b6e8daea86c0d2c0 scipy-1.12.0rc1-cp39-cp39-musllinux_1_1_x86_64.whl 30725e963b6781e1b1ace788b9d17272 scipy-1.12.0rc1-cp39-cp39-win_amd64.whl SHA256 ~~~~~~ 2e733b2385ed5e793698b7bd39a5733506c472c5f33f19bf79e0f193ac733c1b Changelog 121bbe693f7e373540abb6fb6e4ff226cba28e8191ff9c9788468e3582c2e057 README.txt 3615179981dbf46acfbb24f5883e79fd6dff2984799139ce674125f1271f9f4b scipy-1.12.0rc1-cp310-cp310-macosx_10_9_x86_64.whl 91c886f0c2fd494da4866e4e0be33325b6e5c6f8d2a7bf6eea2af7782b3f89a5 scipy-1.12.0rc1-cp310-cp310-macosx_12_0_arm64.whl 7f93c6febff595fad1f25c0303061a79a8c6e8b720ac76a9595b6b3523bea7fd scipy-1.12.0rc1-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl 0817f36403a03d37a385d9931463921c348dfb86f4051c981c49c00a3b56809d scipy-1.12.0rc1-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl 059bf15669d30f4e105aa283fb7fac8fa2f9a6de8a3939422bad265a674c537b scipy-1.12.0rc1-cp310-cp310-musllinux_1_1_x86_64.whl 7aa5d8794955db85ce3806fcbacd15a363472ddd3be776fa0deb0c01896b67af scipy-1.12.0rc1-cp310-cp310-win_amd64.whl a4056fb90600d94f598caed6a3465e9223fedd15b76bf6ab71775e3bf2bbc5d1 scipy-1.12.0rc1-cp311-cp311-macosx_10_9_x86_64.whl 59f121fdc3c9e5a21e94029f4e60f0466e3c81b1842114dcbfa6289e33f4d238 scipy-1.12.0rc1-cp311-cp311-macosx_12_0_arm64.whl c6b4455c11407996c454f05b02a51129aef3a7f92d5f8c2d01b0af496716e43b scipy-1.12.0rc1-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl dd19dff414b7c07b711b4bb584728d3803f8a09aa31e13ed137830f27a5e8dd9 scipy-1.12.0rc1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl 6ef9890afa855f20a77c9fb83db7f21366ff2fd4fbaa4204e05dc7680725f5f5 scipy-1.12.0rc1-cp311-cp311-musllinux_1_1_x86_64.whl 7bfba81e6da57a5fea65c7450de8cc043f4ac95966183c53b8ce379957f33d19 scipy-1.12.0rc1-cp311-cp311-win_amd64.whl bfd39fa37239331711194c75b57230333463630a9885385216939e9327fb3db5 scipy-1.12.0rc1-cp312-cp312-macosx_10_9_x86_64.whl cf2b879f1e81c16bfd40dd61ede0774467a07cadeaa7ba93ff52abf85e4358e5 scipy-1.12.0rc1-cp312-cp312-macosx_12_0_arm64.whl 36589310180e55537f0a42444bfc05f388155f3c9e75644fc9fabc8ea05351b4 scipy-1.12.0rc1-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl 8786144accdf673ed3b7d01510a78926eb95f7d1bb339a993fdd0a22e20f1f64 scipy-1.12.0rc1-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl b0113b938209293a1d4fbdd1fb07a5ac02314d9abc538503f38cd15c83a65363 scipy-1.12.0rc1-cp312-cp312-musllinux_1_1_x86_64.whl 7914b4a79237a20988369c77d01ba974d85795c8c84c6ebed72f3611765245ca scipy-1.12.0rc1-cp312-cp312-win_amd64.whl e5b7614be617716dec97f5865a352f3a46d50a1aa509d6c273c0f19a30b69030 scipy-1.12.0rc1-cp39-cp39-macosx_10_9_x86_64.whl e4418262eb9937af44236b677b33d378631476282bd3234209d424a5aca3c4fc scipy-1.12.0rc1-cp39-cp39-macosx_12_0_arm64.whl 65cb554a52be1ddd288f154c496e73baca2411054956ad6b073343f8428e025d scipy-1.12.0rc1-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl 7b1f66bcafffcbd1bbf52796fb1fe64d077be65792058b88eac7f1e9b51b51d8 scipy-1.12.0rc1-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl d8f574b692f5bcd204d8e23bebf113ec6ade2bb5724ee5ea7e81d5875442a008 scipy-1.12.0rc1-cp39-cp39-musllinux_1_1_x86_64.whl 6697133d20770ad289a6a23080532fd269f1ac9e9c93c2e84f3b1f1d60ff510b scipy-1.12.0rc1-cp39-cp39-win_amd64.whl

1 0