search for: tyeing

> On Aug 17, 2016, at 2:08 PM, Zhuravlyov, Konstantin <Konstantin.Zhuravlyov at amd.com> wrote: > > >Why not going with a metadata attachment directly and kill the "singlethread" keyword? Something like: > >Something like: > > cmpxchg i32* %addr, i32 42, i32 0 monotonic monotonic, 3, !memory.scope{!42} > > cmpxchg i32* %addr, i32 42, i32 0 monotonic

Hi, I have updated the review here: https://reviews.llvm.org/D21723 As Sameer pointed out, the motivation is: In OpenCL 2.x, two atomic operations on the same atomic object need to have the same scope to prevent a data race. This derives from the definition of "inclusive scope" in OpenCL 2.x. Encoding OpenCL 2.x scope as metadata in LLVM IR would be a problem because there cannot be a

Hi all Ok, maybe it''s just me, but - how on earth do I add recipes to the wiki...?? When I go into the list of recipes, I don''t find a "Edit this page button". How should I add to the list? -- Med venlig hilsen Juri Rischel Jensen Fab:IT ApS Vesterbrogade 50 DK-1620 København Tlf: 70 202 407 / Fax: 33 313 640 www.fab-it.dk / juri@fab-it.dk

We believe that it would be best that this is added to the LLVM IR atomic memory instruction as fields on atomic instructions rather than using meta data. The reasoning is that this information is similar to other information that is represented as instruction fields. For example, the indication that memory operations are atomic rather than non-atomic, the memory ordering of atomics, and whether

Hi Scott, It's possible they've missed it, so I've explicitly CC'ed a number of the usual DWARF suspects, at least some of whom are on the standards committee. I don't have anything specific to add myself. James On Mon, 27 Apr 2020 at 15:25, via llvm-dev <llvm-dev at lists.llvm.org> wrote: > I don't know what an acceptable ping rate on an RFC is, but I also

I will comment - as one of the few people actually working on llvm's atomic implementation with any regularity - that I am opposed to extending the instructions without a strong motivating case. I don't care anywhere near as much about metadata based schemes, but extending the instruction semantics imposes a much larger burden on the rest of the community. That burden has to be well

> On Aug 21, 2016, at 11:14 AM, Philip Reames <listmail at philipreames.com> wrote: > > On 08/17/2016 03:05 PM, Mehdi Amini wrote: >> >>> On Aug 17, 2016, at 2:08 PM, Zhuravlyov, Konstantin <Konstantin.Zhuravlyov at amd.com <mailto:Konstantin.Zhuravlyov at amd.com>> wrote: >>> >>> >Why not going with a metadata attachment directly

I am trying to create such module with API (it's equivalent to c++: const char* ss[] = {"s1","s2"};): @ss = global [2 x i8*] [i8* getelementptr inbounds ([3 x i8]* @.str1, i32 0, i32 0), i8* getelementptr inbounds ([3 x i8]* @.str2, i32 0, i32 0)] ; <[2 x i8*]*> [#uses=0] @.str1 = private constant [3 x i8] c"s1\00", align 1 ; <[3 x i8]*> [#uses=1]

(Apologies to klibc@zytor.com for the re-send) This patch makes gen_init_cpio generate the initramfs_data.cpio from a file which contains a list of entries: file, dir, nod. I swapped the order of filename/location for the file arguments so that it would be more uniform with the dir and nod tyes. [thayne@torch linux-2.6.8]$ usr/gen_init_cpio --help ERROR: unable to open '--help': No such

> Let me rephrase: why is it preferable to add first class instruction support for opaque scope rather than using intrinsics? Given that LLVM core now supports atomic instructions, it would seem desirable to use them for all languages supported by LLVM. This would allow optimizations to be aware of the memory semantics. By using intrinsics this information no longer becomes available in a

Hi all, On 02.05.2016 17:46, Tom Stellard via llvm-dev wrote: >> Why not going with a metadata attachment directly and kill the "singlethread" keyword? Something like: >> >Something like: >> > >> > cmpxchg i32* %addr, i32 42, i32 0 monotonic monotonic, 3, !memory.scope{!42} >> > cmpxchg i32* %addr, i32 42, i32 0 monotonic monotonic, 3,

On Wed, Aug 31, 2016 at 12:23:34PM -0700, Justin Lebar via llvm-dev wrote: > > Some optimizations that are related to a single thread could be done without needing to know the actual memory scope. > > Right, it's clear to me that there exist optimizations that you cannot > do if we model these ops as target-specific intrinsics. > > But what I think Mehdi and I were

> Since the scope is “opaque” and target specific, can you elaborate what kind of generic optimization can be performed? Some optimizations that are related to a single thread could be done without needing to know the actual memory scope. For example, an atomic acquire can restrict reordering memory operations after it, but allow reordering of memory operations (except another atomic acquire)

I would like to be able to double click on an R source file and by that start R and have the file sourced in R. I am using WinXP Pro. So I created an association for file type R and tye following Run action: "C:\Program Files\R\rw1090beta\bin\Rgui.exe" R_PROFILE=%1 I have an R file with the following lines at the beginning. library(vcd) votes<-array(0,10*7*9*2)

Hi all, I'm wondering whether it is possible to construct a confidence interval using only the mean, variance, skewness and kurtosis, i.e. without any of the population? If anyone could help with this it'd be much appreciated (even if just a confirmation of it being impossible!). Thanks. -- View this message in context:

The patch makes gen_init_cpio generate the initramfs_data.cpio from a file which contains a list of entries: file, dir, nod. I swapped the order of filename/location for the file arguments so that it would be more uniform with the dir and nod tyes. [thayne@torch linux-2.6.8]$ usr/gen_init_cpio --help ERROR: unable to open '--help': No such file or directory Usage:

Thomas Schmitt wrote: > David Christensen wrote: > > # fdisk -l /dev/sdc > > WARNING: GPT (GUID Partition Table) detected on '/dev/sdc'! The util > > fdisk doesn't support GPT. > > That's actually a bug in fdisk. The GPT is invalid because it is not > announced by a "Protective MBR". > > > Device Boot Start End

I''m trying to setup some mildly complex associations for a project we''re working on and can''t seem to find much documentation on n-way has_many :through associations. I have the following models: Person, PhysicalAddress, EmailAddress, PhoneNumber. Each person can have multiple PhysicalAddresses, EmailAddresses, and PhoneNumbers, and multiple people can share the same

Hi, David Christensen wrote: > I use this USB flash drive for bootable installer images; I do not > attempt to mount it. I don't know that I could, even if I wanted to: mkdir /mnt/iso mount /dev/sdc /mnt/iso or, because partition 1 starts at block 0, you may mount it too mount /dev/sdc1 /mnt/iso > # fdisk -l /dev/sdc > WARNING: GPT (GUID Partition Table) detected on

http://bugzilla.mindrot.org/show_bug.cgi?id=910 mindrot at askneil.com changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |mindrot at askneil.com ------- You are receiving this mail because: ------- You are the assignee for the bug, or are watching the