similar to: [LLVMdev] glibc 2.10 featuring new elf trick

"Nuno Lopes" <nunoplopes at sapo.pt> writes: > Hi, Hello, > I just wanted to brought you attention to the glibc 2.10 changelog. Take a > special attention to the bottom (to the 'Automatic use of optimized > function' section). URL: http://udrepper.livejournal.com/20948.html > This seems to be a nice way to make fat binaries. IMO, fat binaries should go

On Apr 19, 2009, at 3:31 AM, Anton Korobeynikov wrote: > Hello, Michael > >> IMO, fat binaries should go away and be replaced by LLVM bitcode, >> with >> code generation performed by the OS at load time. This would have >> many >> advantages: the compatibility with all current and future >> architectures would be guaranteed, > Unfortunately no, the

Hello, Michael > IMO, fat binaries should go away and be replaced by LLVM bitcode, with > code generation performed by the OS at load time. This would have many > advantages: the compatibility with all current and future > architectures would be guaranteed, Unfortunately no, the IR obtained from languages like C, C++, etc is not target neutral. -- With best regards, Anton

This takes advantage of glibc's MALLOC_PERTURB_. If you develop on glibc-based systems and don't yet set this via your shell's init scripts, add something like this to e.g., .bashrc: export MALLOC_PERTURB_=$(($RANDOM % 255 + 1)) This change ensures that it's enabled for anyone running "make check" and in particular for koji. There's probably documentation

Here is an update to the patch posted yesterday. The changes are: (1) Don't use inline (for mdbooth). (2) Added a comment to HACKING (for mdbooth). (3) Check for old-style and new-style glibc functions and use whatever is available (for danpb). (4) Removed the old shell_quote function instead of just commenting it out. Can anyone decode from Uli's message[1] whether these new

Is this the right mailing list ? I want to compile llvm-2.5 under Fedora 11 with gcc (GCC) 4.4.0 20090506 (Red Hat 4.4.0-4) and get In file included from Signals.cpp:31: Unix/Signals.inc: In function ‘void<unnamed>::PrintStackTrace()’: Unix/Signals.inc:81: error: invalid conversion from ‘const char*’ to ‘char*’ Unix/Signals.inc:96: error: invalid conversion from ‘const char*’ to ‘char*’

At the moment the daemon code contains an incredibly hairy function called shell_quote for safely quoting strings passed to the shell. The patch replaces that with a glibc custom printf format (actually two, but very closely related), %Q and %R. %Q is like %s but it safely shell quotes the string. %R is like %Q but it prefixes the path with /sysroot. Example usage (w/o error checks):

Sorry, we really screwed up the server migration. Alive is now working again and should be fixed for good :) Permalinks are still missing; we are working on recovering those. Apologies again for all the trouble. Nuno -----Original Message----- From: Sanjay Patel Sent: Monday, October 2, 2017 5:10 PM Subject: Re: [llvm-dev] Where did Alive go? I still can't use the web app - spins for

And now rise4fun.com doesn't work at all? ~Craig On Wed, Sep 20, 2017 at 9:53 AM, Nuno Lopes <nunoplopes at sapo.pt> wrote: > Alive is now working again. There was a migration to a new server. > Permalinks are still being copied from backup; they will work again > shortly as well. (I would probably not create new ones since they may get > replaced while the copy is in

Craig I know it's a pain compared to the web interface but Alive is pretty easy to install and run from a shell. John On 9/22/17 11:41 AM, Craig Topper via llvm-dev wrote: > And now rise4fun.com <http://rise4fun.com> doesn't work at all? > > ~Craig > > On Wed, Sep 20, 2017 at 9:53 AM, Nuno Lopes <nunoplopes at sapo.pt > <mailto:nunoplopes at

On Jun 4, 2012, at 10:37 AM, Nuno Lopes <nunoplopes at sapo.pt> wrote: > So here is a new proposal: > > !0 = metadata !{ alloc_siz_fn, offset_fn, parameters* } The parameters are a separate metadata array or the alloc_size metadata is variable length? You'll probably want to write up some docs for the website on how this is supposed to be laid out and work. -eric

Alive is now working again. There was a migration to a new server. Permalinks are still being copied from backup; they will work again shortly as well. (I would probably not create new ones since they may get replaced while the copy is in flux). Nuno Citando Nuno Lopes via llvm-dev <llvm-dev at lists.llvm.org>: > I'm investigating; thanks for the heads up. Sorry for the

We have these transforms from funnel shift to a simpler shift op: // fshl(X, 0, C) -> shl X, C // fshl(X, undef, C) -> shl X, C // fshl(0, X, C) -> lshr X, (BW-C) // fshl(undef, X, C) -> lshr X, (BW-C) These were part of: https://reviews.llvm.org/D54778 In all cases, one operand must be 0 or undef and the shift amount is a constant, so I think these are safe.

Hey Scott, There has been quite a lot of research on capture tracking (aka escape analysis) for Java and other dynamic languages. See e.g.: https://wiki.openjdk.java.net/display/HotSpot/EscapeAnalysis http://docs.oracle.com/javase/7/docs/technotes/guides/vm/performance-enhancements-7.html http://dl.acm.org/citation.cfm?doid=320384.320386 Nuno -----Original Message----- From: Scott Egerton via

>> On Thu, Mar 1, 2018 at 2:08 AM, Nuno Lopes <nunoplopes at sapo.pt> wrote: >> We can do "add %x, undef" => "undef" because for any value of %x, we can >> always find a value that when added to %x produces any value in the >> domain of integers. >> >> This is not the case with floats since with some inputs, e.g., NaNs, we

Quoting Eric Christopher <echristo at apple.com>: > On Jun 4, 2012, at 10:37 AM, Nuno Lopes <nunoplopes at sapo.pt> wrote: > >> So here is a new proposal: >> >> !0 = metadata !{ alloc_siz_fn, offset_fn, parameters* } > > The parameters are a separate metadata array or the alloc_size metadata > is variable length? Variable length. I think that's

Ah, thanks for explaining. So given that any of these ops will return NaN with a NaN operand, let's choose the undef operand value to be NaN. That means we can fold all of these to a NaN constant in the general case. But if we have 'nnan' FMF, then we can fold harder to undef? nnan - Allow optimizations to assume the arguments and result are not NaN. Such optimizations are required to

On Sat, Jun 27, 2015 at 2:16 PM, Nuno Lopes <nunoplopes at sapo.pt> wrote: > At least in C/C++ that's UB, yes. So you cannot map every turing machine to a valid C/C++ program then. :) Also, does this mean that "daemon" programs that run continuously till they're killed by the OS (using a mechanism that is not visible in C) are effectively undefined? -- Sanjoy >

Definitely have a look at the current analyses: - llvm/Transforms/IPO/FunctionAttrs.cpp - llvm/Transforms/IPO/InferFunctionAttrs.cpp Also, study the semantics of these attributes, starting with the docs: http://llvm.org/docs/LangRef.html#function-attributes Also, grep the LLVM sources for test cases that use the attributes to see examples on how they are used for optimization. Finally, have a

On Sun, Jul 16, 2017, 12:45 PM Nuno Lopes <nunoplopes at sapo.pt> wrote: > >On 07/15/2017 04:51 AM, Nuno Lopes wrote: > >>> On 07/14/2017 04:37 PM, Nuno Lopes wrote: > >>>> Thank you all for your replies. > >>>> So here seems to be an agreement that the documentation for > >>>> PartialAlias is incorrect. > >>>>