similar to: [LLVMdev] [lld] alias atoms and LayoutPass

Hi, I think we are overloading the Layout references for garbage collection. If you are creating a reference (kindLayoutAfter) from A to B, that may not mean that you cannot garbage collect B for the end user. My thought on Layout references was that it only guarantees that atoms appear in Layout reference order. Why are we overloading this for Garbage collection (aside from saving space/code)

In r226336 I shove off 1.2 seconds out of 9.8 seconds for lld to link lld. That's done by parallelizing archive member parsing. But I realized that was not the slowest pass. The single slowest pass in LLD is LayoutPass. Only sort() at the last of Layoutpass::perform takes about 3 seconds (one third of total execution time). It is because the comparison function passed to sort, compareAtoms,

Hi, I think it would be great if we could subclass the LayoutPass. All the generic functionality could be in the parent to run the layout-after/in-group/ and currently the layout-before passes(until it gets removed). Flavors can *choose to run the layout passes* that they use and determine the way things get ordered in the layout pass. The compare function in the LayoutPass would call

Ping ? Do you think that we need to have an API in LinkingContext to return the next ordinal available, so that files created by passes can be assigned ordinals ? Thanks Shankar Easwaran On 10/6/2013 11:07 PM, Shankar Easwaran wrote: > In addition I think the LayoutPass std::stable_sort be replaced with > std::sort as total ordering is guaranteed as each File would get an >

In addition I think the LayoutPass std::stable_sort be replaced with std::sort as total ordering is guaranteed as each File would get an ordinal and each atom would get an ordinal too, after the below problem is fixed. Thanks Shankar Easwaran On 10/6/2013 10:54 PM, Shankar Easwaran wrote: > Hi, > > It looks like the the ELFPassFile doesnot get an ordinal value > assigned, as its

Thank you for quick responses! As to dead stripping, if dead stripping is the only pass we need bi-directional edges, we might want the dead stripping pass to construct internal data structure by reversing the graph to construct layout-before edges from layout-after edges. This should be less error prone than maintaining two reverse-directional edges throughout all passes. Of course it will make

Hi, It looks like the the ELFPassFile doesnot get an ordinal value assigned, as its added in a pass. Is there a way to assign a file ordinal for the files added by Passes ? Till that time, I am going to XFAIL the ifunc test. More tests should fail, and am not sure why they are not failing. Thanks Shankar Easwaran -- Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted

+llvmdev On Fri, Mar 21, 2014 at 3:43 PM, Rui Ueyama <ruiu at google.com> wrote: > Hi all, > > I'm trying to debug an issue that LLD sometimes get into an infinite loop > in setChainRoot() in LayoutPass.cpp. It looks like the cause is either > buildPrecededByTable() handles layoutBefore edges in a wrong way or we > construct a contradictory layout-before/layout-after

On Mon, Oct 7, 2013 at 11:51 AM, Shankar Easwaran <shankare at codeaurora.org>wrote: > Ping ? > > Do you think that we need to have an API in LinkingContext to return the > next ordinal available, so that files created by passes can be assigned > ordinals ? > That API may work, but I don't think you always want to assign the largest file ordinal for a file created in

On 10/7/2013 3:43 PM, Rui Ueyama wrote: > On Mon, Oct 7, 2013 at 11:51 AM, Shankar Easwaran > <shankare at codeaurora.org>wrote: > >> Ping ? >> >> Do you think that we need to have an API in LinkingContext to return the >> next ordinal available, so that files created by passes can be assigned >> ordinals ? >> > That API may work, but I

Hi Nick, --start-group/--end-group functionality with the Gnu flavor currently works only if the --start-group/--end-group contains archive files. If they contain dynamic libraries, the undefined atoms from the dynamic libraries are processed whenever the group is iterated again. I am trying to find out a way to make the resolver add undefined atoms from the shared library only *once*, when

Hi, Can we add deadStrip() to undefinedAtoms as well ? This will enable to choose whether we want to set the property deadStripNormal or deadStripNever on them. Also I think it will be cleaner for atoms to be added to deadStripRoot set using a single API. Thanks Shankar Easwaran -- Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by the Linux Foundation

Hi Rui, I thought about this, but there is a catch. The undefined symbols still need to be resolved with exports from dynamic libraries even the second time, or any number of times. For example :- ld 1.o --start-group libc.so myprintf.a --end-group 1.o has a undef for myprintf, which is in myprintf.a and myprintf.a calls puts. The first time when libc.so, undef atoms would be added, no

I'd do that with the nextFile abstraction like this: On the first iteration, a Group would return its member every time getNextFile() is called (the same as the current behavior). On the second and further iterations, the Group should skip all the members whose type is not Archive. By doing this, the core linker sees dynamic libraries (or regular object files) only once even if they are in

Currently, member functions like File::defined() return an iterator for const atoms. But we don't actually treat atoms as consts -- we updates atoms in many places including the core resolver. We have too many const_casts in our code. It just doesn't make sense. I'm making a change to make atoms non-const. Please hit reply if you have any concerns. -------------- next part

Hi Nick, Waiting for your feedback on this. Thanks Shankar Easwaran On 11/16/2012 10:03 AM, Shankar Easwaran wrote: > Hi Nick, > > Thanks for your reply. > > The usecase here is just trying to construct a valid ELF. The lld > linker needs to handle all sorts of code written in assembly as well > as 'C'. The usecase is just one example of it. > > I have also

Hi Nick, Thanks for your reply. The usecase here is just trying to construct a valid ELF. The lld linker needs to handle all sorts of code written in assembly as well as 'C'. The usecase is just one example of it. I have also seen similiar code in http://lxr.free-electrons.com/source/arch/powerpc/kernel/head_32.S?a=powerpc which has global and local labels. You are right, that the

Hi Nick, On 8/27/2013 12:45 AM, Nick Kledzik wrote: > On Aug 26, 2013, at 10:20 PM, Shankar Easwaran wrote: >> Can we add deadStrip() to undefinedAtoms as well ? >> >> This will enable to choose whether we want to set the property deadStripNormal or deadStripNever on them. >> >> Also I think it will be cleaner for atoms to be added to deadStripRoot set using a

+ Nick Rui, Does PECOFF writer need the filename in the writer as well, I am not sure if linker scripts are supported with PECOFF though. If PECOFF also needs it, I think it makes sense to store the filename in the Atom as the native format needs to store that information. The only option for the ELF writer to know this information is to use References if other flavors dont need the

Hi Nick, I need the original path of the files, from where the atoms were parsed. The path for the atoms are completely useless after the round trip pass. This is needed to support linker scripts with the Gnu flavor. I think there are more uses too like diagnostics (relocation overflows etc) I was planning to add a patch so that the native writer, stores the original path for atoms Another