similar to: Why are GEPs type based?

Hello, This is an RFC for adding intrinsics which perform saturating signed/unsigned left shift. There is currently a patch on Phabricator here: https://reviews.llvm.org/D83216 The intrinsics are of the form i32 @llvm.sshl.sat.i32(i32, i32) i32 @llvm.ushl.sat.i32(i32, i32) <4 x i32> @llvm.sshl.sat.v4i32(<4 x i32>, <4 x i32>) <4 x i32>

On Thu, Jun 25, 2020 at 2:43 AM Nikita Popov <nikita.ppv at gmail.com> wrote: > On Thu, Jun 25, 2020 at 11:22 AM Zachary Turner via llvm-dev < > llvm-dev at lists.llvm.org> wrote: > >> I can’t really provide a doc, but i can describe what I believe to be the >> biggest problem. >> >> In a GH PR, comments are associated with commit hashes. If a commit

On Thu, Jun 25, 2020 at 11:43 AM Nikita Popov via llvm-dev <llvm-dev at lists.llvm.org> wrote: > On Thu, Jun 25, 2020 at 11:22 AM Zachary Turner via llvm-dev <llvm-dev at lists.llvm.org> wrote: >> What this means for LLVM is that everyone will have to completely stop using history rewriting operations. No more rebase, squash, amend, etc. > > This is also incorrect. Most

> On Mar 25, 2020, at 19:42, Chris Tetreault <ctetreau at quicinc.com> wrote: > > I would like to just chime in and say that I’m fairly strongly opposed to any blanket version increases without justification. Having a low version requirement is a feature. It means that more people can build the codebase. We should increase the minimum CMake version requirement only if we need to do

Hi, Does the LLVM project perform any kind of tracking for commit-by-commit compile-time changes? It looks like LNT only tracks run-time performance (and to be honest I wasn't able to make heads or tails of the results even for that -- the interface was pretty unintuitive to me.) While it is "normal" that each new LLVM release regresses compile-time by 2-3%, LLVM 10 seems to be

I understand organization restrictions and old operating systems (I use CentOS 7 myself), but I’ll note that the only requirement for running a new CMake is the ability to download and untar a tarball; in particular, you don’t require sudo. (I understand that there may be restrictions around running arbitrary executables downloaded from the internet, which of course make sense, but I wanted to

> On Mar 25, 2020, at 13:07, Nikita Popov <nikita.ppv at gmail.com> wrote: > > On Wed, Mar 25, 2020 at 5:01 PM Tom Stellard via llvm-dev <llvm-dev at lists.llvm.org <mailto:llvm-dev at lists.llvm.org>> wrote: > On 03/25/2020 06:20 AM, Louis Dionne wrote: > > > > > >> On Mar 25, 2020, at 00:47, Tom Stellard <tstellar at redhat.com

Hi Nikita, Glad to hear that Rust code can benefit a lot from this. I have put patches to enable merge-similar functions with thinLTO. https://reviews.llvm.org/D52896 etc. <https://reviews.llvm.org/D52896> This is more powerful than existing merge-functions pass and all we need to do is port these patches to trunk llvm. I'd be happy to help with this effort. -Aditya

Hi, The IRBuilder is currently templated over a constant folder, and an instruction inserter. https://reviews.llvm.org/D73835 proposes to move this towards using virtual dispatch instead. As this is a larger design change, I would like to get some feedback on this. The current templated design of IRBuilder has a couple of problems: 1. It's not possible to share code between use-sites that

Hello all. This is a proposal to introduce 5 new integer intrinsics: * absolute value * signed min * signed max * unsigned min * unsigned max This is motivated by the fact that we keep working around not having these intrinsics, and that constantly leads us into having more workarounds, and causes infinite combine loops. Here's a (likely incomplete!) list of motivational bugs: infinite

On Fri, Jul 17, 2020 at 10:52 PM Nikita Popov via llvm-dev <llvm-dev at lists.llvm.org> wrote: > > On Fri, Jul 17, 2020 at 9:30 PM Jonathan Roelofs via llvm-dev <llvm-dev at lists.llvm.org> wrote: >> >> I’m digging through a build failure [1], and it looks like the loop idiom recognizer adds some instructions [2], and then removes them again [3]. I don’t understand

Hi, It’s been a few years now since I added some intrinsics for doing vector reductions. We’ve been using them exclusively on AArch64, and I’ve seen some traffic a while ago on list for other targets too. Sander did some work last year to refine the semantics after some discussion. Are we at the point where we can drop the “experimental” from the name? IMO all target should begin to transition

I would be interested in learning what the set of used semantics for float-to-int conversion is. If the only two used are 1) undefined behavior if unrepresentable and 2) saturate to int_{min,max} with NaN going to zero, then I think it makes sense to expose both of those natively in the IR. If the set is much larger, I think separate intrinsics for each behavior would make sense. It would be nice

I’m digging through a build failure [1], and it looks like the loop idiom recognizer adds some instructions [2], and then removes them again [3]. I don’t understand why yet, but the LegacyPassManager detects that the structural hash of the function has changed, and complains that the pass didn’t correctly report that it changed the function [4] (even though materially, it didn’t). This raises a

The thing is, we *don't* "not demand" those high bits. We *don't* not care what's in those bits - IR shifts don't mask their shift amounts. I.e we can't replace `x >> (32-y)` with `x >> (-y)`, which would be legal transform should we not demand those bits. We very much demand them. We just know those bits to be zero. And i'm not sure how to convey

Hi everyone! The fptoui/fptosi instructions are currently specified to return a poison value if the rounded-towards-zero floating point number cannot be represented by the target integer type. The motivation for this behavior is that overflowing float to int casts in C are undefined behavior. However, many newer languages prefer to have a float to integer cast that is well-defined for all input

I was looking at the SymbolTable code in LLD COFF and ELF recently, and I’m confused by the use of CachedHashStringRef. From what I understand, a CachedHashStringRef combines a StringRef with a computed hash. There’s no caching going on in the CachedHashStringRef itself; that is, if you construct CachedHashStringRef("foo"), and then construct a second

On Mon, Oct 7, 2019 at 11:32 AM Roman Lebedev <lebedev.ri at gmail.com> wrote: > > Bump. Any further thoughts here? > > To recap - i don't really see how this can be a demandedbits problem - we do > demand all those bits, we just know they must be zero. > (i would love to be proven wrong though!) > > Roman. > > On Tue, Oct 1, 2019 at 11:17 PM Roman Lebedev

I can’t really provide a doc, but i can describe what I believe to be the biggest problem. In a GH PR, comments are associated with commit hashes. If a commit hash ceases to exist, so do all comments associated with it. The comments are quite literally destroyed and irretrievable. What this means for LLVM is that everyone will have to completely stop using history rewriting operations. No

On Sat, Feb 9, 2019 at 6:25 PM Simon Pilgrim <llvm-dev at redking.me.uk> wrote: > The add/sub (+mul) overflow intrinsics are being updated to support > vectors to match the related add/sub saturation intrinsics. We haven't > updated the docs yet as legalization, vectorization and various minor bits > of plumbing still need to be finished before it can be officially supported