Your help needed: List of LLVM Open Projects 2017

Hi folks,

   Happy new year!

   Last LLVM Developers' Meeting I had a BoF: 'Raising Next Generation LLVM Developers'. It was suggested that we should update our open projects page and possibly restructure it a little bit.

   I volunteered to do this work and I need your help.

   Chandler and I started working on a google doc [1]. We pinged few code owners asking them to list of work items we should get done in 2017 but we do not have the manpower. Now we would like to ask for your input, too.

   I believe an up to date list can serve as a good entry point for students, interns and new contributors.

   Feel free to propose a new item or comment under an existing one. I expect to start gradually updating the page beginning of Feb.

-- Vassil

[1] https://docs.google.com/document/d/1YLK_xINSg1Ei0w8w39uAMR1n0dlf6wrzfypiX0YDQBc/edit?usp=sharing

LLDB Folks,

FYI, see below, the LLDB section is empty right now.

Best,

Mehdi

Do we have any open projects on LLD?

I know we usually try to avoid any big “projects” and mainly add/fix things in response to user needs, but just wondering if somebody has any ideas.

Some really generic/simple stuff I can think of:

  1. trying out LLD on a large program corpus and reporting/reducing/fixing bugs (e.g. contributing to the FreeBSD effort or trying to build a bunch of packages from a linux distro like Debian or Gentoo)
  2. performance analysis and optimization of LLD
  3. getting LLD to link a bootable Linux kernel and/or GRUB
  4. write an input verifier such that LLD can survive intensive fuzzing with no crashes / fatal errors [1] when the verifier says the input is okay. This will allow us to measure what the overhead of doing this actually is.

[1] As of the latest LLD discussion (in the thread “[llvm-dev] LLD status update and performance chart”) it sounds like people are okay with LLD treating fatal errors the same way that LLVM uses assertions; for inputs from the C++ API, we can document to not pass corrupted object files. For inputs read from files, there is still community interest in at least having the option to run a “verifier” to validate the inputs. I think the best way to approach the verifier is to essentially follow the approach suggested by Peter (in the context of “hardening”) in https://llvm.org/bugs/show_bug.cgi?id=30540#c5 i.e. getting to the point where LLD can survive intensive fuzzing.

– Sean Silva

The list can’t ommit clang-tidy.
There are many ideas about new checks on llvm bugzilla.
https://llvm.org/bugs/buglist.cgi?product=clang-tools-extra&component=clang-tidy&resolution=—&list_id=110936

Everything matching “.Feature Request.

Piotr

Please submit patches to Open Projects pages! Winter^WSummer of Code is coming!

Do we have any open projects on LLD?

I know we usually try to avoid any big "projects" and mainly add/fix things
in response to user needs, but just wondering if somebody has any ideas.

Some really generic/simple stuff I can think of:
1. trying out LLD on a large program corpus and reporting/reducing/fixing
bugs (e.g. contributing to the FreeBSD effort or trying to build a bunch of
packages from a linux distro like Debian or Gentoo)

From Rafael's last Poudriere ports build I think about 98% of the

packages are building with LLD, and some of the missing ones are those
that were skipped (e.g. do not build on amd64, or the upstream
distfiles have gone away). I think some next steps here for FreeBSD
include:

* Ensure we're running the test suites in packages that have them
* Actually install and use the resulting packages for a smoke test
* Address the WIP patches / workarounds currently in use
* Triage the few hundred failures

From the FreeBSD perspective there's one key LLD task of interest:

* Bring other architecture support to parity with amd64/x86_64. For us
the next one in importance is AArch64/arm64, then i386 and 32-bit arm,
and 32- and 64-bit MIPS, PowerPC, and RISC-V.

I'm not particularly active in lld anymore, but the last big item I'd
like to see implemented is Pettis-Hansen layout.
http://perso.ensta-paristech.fr/~bmonsuez/Cours/B6-4/Articles/papers15.pdf
(mainly because it improves performances of the final executable).
GCC/gold have an implementation of the algorithm that can be used as
base. I'll expand if anybody is interested.
Side note: I'd like to propose a couple of llvm projects as well, I'll
sit down later today and write them.

> Do we have any open projects on LLD?
>
> I know we usually try to avoid any big "projects" and mainly add/fix
things
> in response to user needs, but just wondering if somebody has any ideas.
>
> Some really generic/simple stuff I can think of:
> 1. trying out LLD on a large program corpus and reporting/reducing/fixing
> bugs (e.g. contributing to the FreeBSD effort or trying to build a bunch
of
> packages from a linux distro like Debian or Gentoo)

From Rafael's last Poudriere ports build I think about 98% of the
packages are building with LLD, and some of the missing ones are those
that were skipped (e.g. do not build on amd64, or the upstream
distfiles have gone away).

I thought most of the skipped stuff was due to dependencies on packages
that failed? Or is that no longer the case?

I think some next steps here for FreeBSD
include:

* Ensure we're running the test suites in packages that have them
* Actually install and use the resulting packages for a smoke test
* Address the WIP patches / workarounds currently in use

Are these collected somewhere / is there a status page to reference?

* Triage the few hundred failures

Are these collected somewhere / is there a status page to reference?

From the FreeBSD perspective there's one key LLD task of interest:

* Bring other architecture support to parity with amd64/x86_64. For us
the next one in importance is AArch64/arm64, then i386 and 32-bit arm,
and 32- and 64-bit MIPS, PowerPC, and RISC-V.

Architecture porting might be challenging for new contributors, since the
will usually require access to "unusual" hardware, right? Or are there
emulator options available? If so, it would be good to document those
options because it will greatly expand the number of people that can work
on these tasks.

-- Sean Silva

> Do we have any open projects on LLD?
>
> I know we usually try to avoid any big "projects" and mainly add/fix
things
> in response to user needs, but just wondering if somebody has any ideas.
>

I'm not particularly active in lld anymore, but the last big item I'd
like to see implemented is Pettis-Hansen layout.
http://perso.ensta-paristech.fr/~bmonsuez/Cours/B6-4/Articles/papers15.pdf
(mainly because it improves performances of the final executable).
GCC/gold have an implementation of the algorithm that can be used as
base. I'll expand if anybody is interested.
Side note: I'd like to propose a couple of llvm projects as well, I'll
sit down later today and write them.

For FullLTO it is conceptually pretty easy to get profile data we need for
this, but I'm not sure about the ThinLTO case.

Teresa, Mehdi,

Are there any plans (or things already working!) for getting profile data
from ThinLTO in a format that the linker can use for code layout? I assume
that profile data is being used already to guide importing, so it may just
be a matter of siphoning that off.

Or maybe that layout code should be inside LLVM; maybe part of the general
LTO interface? It looks like the current gcc plugin calls back into gcc for
the actual layout algorithm itself (function call
find_pettis_hansen_function_layout) rather than the reordering logic living
in the linker:
https://android.googlesource.com/toolchain/gcc/+/3f73d6ef90458b45bbbb33ef4c2b174d4662a22d/gcc-4.6/function_reordering_plugin/function_reordering_plugin.c

-- Sean Silva

My idea was exactly to have the reordering logic living in LLVM rather
than lld, FWIW.

Do we have any open projects on LLD?

I know we usually try to avoid any big “projects” and mainly add/fix things
in response to user needs, but just wondering if somebody has any ideas.

I’m not particularly active in lld anymore, but the last big item I’d
like to see implemented is Pettis-Hansen layout.
http://perso.ensta-paristech.fr/~bmonsuez/Cours/B6-4/Articles/papers15.pdf
(mainly because it improves performances of the final executable).
GCC/gold have an implementation of the algorithm that can be used as
base. I’ll expand if anybody is interested.
Side note: I’d like to propose a couple of llvm projects as well, I’ll
sit down later today and write them.

I’m not sure, can you confirm that such layout optimization on ELF requires -ffunction-sections?

Also, for clang on OSX the best layout we could get is to order functions in the order in which they get executed at runtime.

For FullLTO it is conceptually pretty easy to get profile data we need for this, but I’m not sure about the ThinLTO case.

Teresa, Mehdi,

Are there any plans (or things already working!) for getting profile data from ThinLTO in a format that the linker can use for code layout? I assume that profile data is being used already to guide importing, so it may just be a matter of siphoning that off.

I’m not sure what kind of “profile information” is needed, and what makes it easier for MonolithicLTO compared to ThinLTO?

Or maybe that layout code should be inside LLVM; maybe part of the general LTO interface? It looks like the current gcc plugin calls back into gcc for the actual layout algorithm itself (function call find_pettis_hansen_function_layout) rather than the reordering logic living in the linker: https://android.googlesource.com/toolchain/gcc/+/3f73d6ef90458b45bbbb33ef4c2b174d4662a22d/gcc-4.6/function_reordering_plugin/function_reordering_plugin.c

I was thinking about this: could this be done by reorganizing the module itself for LTO?

That wouldn’t help non-LTO and ThinLTO though.

> Do we have any open projects on LLD?
>
> I know we usually try to avoid any big "projects" and mainly add/fix
> things
> in response to user needs, but just wondering if somebody has any ideas.
>

I'm not particularly active in lld anymore, but the last big item I'd
like to see implemented is Pettis-Hansen layout.
http://perso.ensta-paristech.fr/~bmonsuez/Cours/B6-4/Articles/papers15.pdf
(mainly because it improves performances of the final executable).
GCC/gold have an implementation of the algorithm that can be used as
base. I'll expand if anybody is interested.
Side note: I'd like to propose a couple of llvm projects as well, I'll
sit down later today and write them.

I’m not sure, can you confirm that such layout optimization on ELF requires
-ffunction-sections?

For the non-LTO case, I think so.

Also, for clang on OSX the best layout we could get is to order functions in
the order in which they get executed at runtime.

That's what we already do for lld. We collect and order file (run a
profiler) and pass that to the linker that lays out functions
accordingly.
This is to improve startup time for a class of startup-time-sensitive
operations. The algorithm proposed by Pettis (allegedly) aims to
reduce the TLB misses as it tries to lay out hot functions (or
functions that are likely to be called together near in the final
binary).

For FullLTO it is conceptually pretty easy to get profile data we need for
this, but I'm not sure about the ThinLTO case.

Teresa, Mehdi,

Are there any plans (or things already working!) for getting profile data
from ThinLTO in a format that the linker can use for code layout? I assume
that profile data is being used already to guide importing, so it may just
be a matter of siphoning that off.

I’m not sure what kind of “profile information” is needed, and what makes it
easier for MonolithicLTO compared to ThinLTO?

Or maybe that layout code should be inside LLVM; maybe part of the general
LTO interface? It looks like the current gcc plugin calls back into gcc for
the actual layout algorithm itself (function call
find_pettis_hansen_function_layout) rather than the reordering logic living
in the linker:
https://android.googlesource.com/toolchain/gcc/+/3f73d6ef90458b45bbbb33ef4c2b174d4662a22d/gcc-4.6/function_reordering_plugin/function_reordering_plugin.c

I was thinking about this: could this be done by reorganizing the module
itself for LTO?

That wouldn’t help non-LTO and ThinLTO though.

This is a dimension that I think can be explored. The fact that it
wouldn't help with other modes of operation is completely orthogonal,
in particular until it's proven that this kind of optimization makes
sense with ThinLTO (and if it doesn't, it can be an optimization ran
only during full LTO).

> Do we have any open projects on LLD?
>
> I know we usually try to avoid any big "projects" and mainly add/fix
things
> in response to user needs, but just wondering if somebody has any ideas.
>

I'm not particularly active in lld anymore, but the last big item I'd
like to see implemented is Pettis-Hansen layout.
http://perso.ensta-paristech.fr/~bmonsuez/Cours/B6-4/Article
s/papers15.pdf
(mainly because it improves performances of the final executable).
GCC/gold have an implementation of the algorithm that can be used as
base. I'll expand if anybody is interested.
Side note: I'd like to propose a couple of llvm projects as well, I'll
sit down later today and write them.

I’m not sure, can you confirm that such layout optimization on ELF
requires -ffunction-sections?

In order for a standard ELF linker to safely be able to reorder sections at
function granularity, -ffunction-sections would be required. This isn't a
problem during LTO since the code generation is set up by the linker :slight_smile:

Also, for clang on OSX the best layout we could get is to order functions
in the order in which they get executed at runtime.

What the optimal layout may be for given apps is a bit of a separate
question. Right now we're mostly talking about how to plumb everything
together so that we can do the reordering of the final executable.

In fact, standard ELF linking semantics generally require input sections to
be concatenated in command line order (this is e.g. how .init_array/.ctors
build up their arrays of pointers to initializers; a crt*.o file at the
beginning/end has a sentinel value and so the order matters). So the linker
will generally need blessing from the compiler to do most sorts of
reorderings as far as I'm aware.

Other signals besides profile info, such as a startup trace, might be
useful too, and we should make sure we can plug that into the design.
My understanding of the clang on OSX case is based on a comparison of the
`form_by_*` functions in clang/utils/perf-training/perf-helper.py which
offer a relatively simple set of algorithms, so I think the jury is still
out on the best approach (that script also uses a data collection method
that is not part of LLVM's usual instrumentation or sampling workflows for
PGO, so we may not be able to provide the same signals out of the box as
part of our standard offering in the compiler)
I think that once we have this ordering capability integrated more deeply
into the compiler, we'll be able to evaluate more complicated algorithms
like Pettis-Hansen, have access to signals like global profile info, do
interesting call graph analyses, etc. to find interesting approaches.

For FullLTO it is conceptually pretty easy to get profile data we need for
this, but I'm not sure about the ThinLTO case.

Teresa, Mehdi,

Are there any plans (or things already working!) for getting profile data
from ThinLTO in a format that the linker can use for code layout? I assume
that profile data is being used already to guide importing, so it may just
be a matter of siphoning that off.

I’m not sure what kind of “profile information” is needed, and what makes
it easier for MonolithicLTO compared to ThinLTO?

For MonolithicLTO I had in mind that a simple implementation would be:

std::vector<std::string> Ordering;
auto Pass = make_unique<LayoutModulePass>(&Ordering);
addPassToLTOPipeline(std::move(Pass))

The module pass would just query the profile data directly on IR
datastructures and get the order out. This would require very little
"plumbing".

Or maybe that layout code should be inside LLVM; maybe part of the general
LTO interface? It looks like the current gcc plugin calls back into gcc for
the actual layout algorithm itself (function call
find_pettis_hansen_function_layout) rather than the reordering logic
living in the linker: https://android.googlesource.com/toolchain/gcc/+/
3f73d6ef90458b45bbbb33ef4c2b174d4662a22d/gcc-4.6/function_
reordering_plugin/function_reordering_plugin.c

I was thinking about this: could this be done by reorganizing the module
itself for LTO?

For MonolithicLTO that's another simple approach.

That wouldn’t help non-LTO and ThinLTO though.

I think we should ideally aim for something that works uniformly for
Monolithic and Thin. For example, GCC emits special sections containing the
profile data and the linker just reads those sections; something analogous
in LLVM would just happen in the backend and be common to Monolithic and
Thin. If ThinLTO already has profile summaries in some nice form though, it
may be possible to bypass this.

Another advantage of using special sections in the output like GCC does is
that you don't actually need LTO at all to get the function reordering. The
profile data passed to the compiler during per-TU compilation can be
lowered into the same kind of annotations. (though LTO and function
ordering are likely to go hand-in-hand most often for peak-performance
builds).

-- Sean Silva

>
>
>
>
>>
>> > Do we have any open projects on LLD?
>> >
>> > I know we usually try to avoid any big "projects" and mainly add/fix
>> > things
>> > in response to user needs, but just wondering if somebody has any
ideas.
>> >
>>
>> I'm not particularly active in lld anymore, but the last big item I'd
>> like to see implemented is Pettis-Hansen layout.
>> http://perso.ensta-paristech.fr/~bmonsuez/Cours/B6-4/
Articles/papers15.pdf
>> (mainly because it improves performances of the final executable).
>> GCC/gold have an implementation of the algorithm that can be used as
>> base. I'll expand if anybody is interested.
>> Side note: I'd like to propose a couple of llvm projects as well, I'll
>> sit down later today and write them.
>
>
>
> I’m not sure, can you confirm that such layout optimization on ELF
requires
> -ffunction-sections?
>

For the non-LTO case, I think so.

> Also, for clang on OSX the best layout we could get is to order
functions in
> the order in which they get executed at runtime.
>

That's what we already do for lld. We collect and order file (run a
profiler) and pass that to the linker that lays out functions
accordingly.
This is to improve startup time for a class of startup-time-sensitive
operations. The algorithm proposed by Pettis (allegedly) aims to
reduce the TLB misses as it tries to lay out hot functions (or
functions that are likely to be called together near in the final
binary).

IIRC from when I looked at the paper a while ago, it is mostly just a
"huffman tree construction" type algorithm (agglomerating based on highest
probability) and assumes that if two functions are hot then they are likely
to be needed together. This is not always the case.

E.g. consider a server that accepts RPC requests and based on those
requests either does Foo or Bar which are largely disjoint. It's entirely
possible for the top two functions of the profile to be one in Foo and one
in Bar, but laying them out near each other doesn't make sense since there
is never locality (for a given RPC, either Foo or Bar gets run). A static
call graph analysis can provide the needed signals to handle this case
better.

-- Sean Silva

>
>
>
>
>>
>> > Do we have any open projects on LLD?
>> >
>> > I know we usually try to avoid any big "projects" and mainly add/fix
>> > things
>> > in response to user needs, but just wondering if somebody has any
ideas.
>> >
>>
>> I'm not particularly active in lld anymore, but the last big item I'd
>> like to see implemented is Pettis-Hansen layout.
>> http://perso.ensta-paristech.fr/~bmonsuez/Cours/B6-4/Article
s/papers15.pdf
>> (mainly because it improves performances of the final executable).
>> GCC/gold have an implementation of the algorithm that can be used as
>> base. I'll expand if anybody is interested.
>> Side note: I'd like to propose a couple of llvm projects as well, I'll
>> sit down later today and write them.
>
>
>
> I’m not sure, can you confirm that such layout optimization on ELF
requires
> -ffunction-sections?
>

For the non-LTO case, I think so.

> Also, for clang on OSX the best layout we could get is to order
functions in
> the order in which they get executed at runtime.
>

That's what we already do for lld. We collect and order file (run a
profiler) and pass that to the linker that lays out functions
accordingly.
This is to improve startup time for a class of startup-time-sensitive
operations. The algorithm proposed by Pettis (allegedly) aims to
reduce the TLB misses as it tries to lay out hot functions (or
functions that are likely to be called together near in the final
binary).

IIRC from when I looked at the paper a while ago, it is mostly just a
"huffman tree construction" type algorithm (agglomerating based on highest
probability) and assumes that if two functions are hot then they are likely
to be needed together. This is not always the case.

E.g. consider a server that accepts RPC requests and based on those
requests either does Foo or Bar which are largely disjoint. It's entirely
possible for the top two functions of the profile to be one in Foo and one
in Bar, but laying them out near each other doesn't make sense since there
is never locality (for a given RPC, either Foo or Bar gets run). A static
call graph analysis can provide the needed signals to handle this case
better.

Hence you said "allegedly" :slight_smile: I know we've talked about this before. Just
wanted to put the backstory of the "allegedly" on the list.

-- Sean Silva

Do we have any open projects on LLD?

I know we usually try to avoid any big “projects” and mainly add/fix things
in response to user needs, but just wondering if somebody has any ideas.

I’m not particularly active in lld anymore, but the last big item I’d
like to see implemented is Pettis-Hansen layout.
http://perso.ensta-paristech.fr/~bmonsuez/Cours/B6-4/Articles/papers15.pdf
(mainly because it improves performances of the final executable).
GCC/gold have an implementation of the algorithm that can be used as
base. I’ll expand if anybody is interested.
Side note: I’d like to propose a couple of llvm projects as well, I’ll
sit down later today and write them.

I’m not sure, can you confirm that such layout optimization on ELF requires -ffunction-sections?

In order for a standard ELF linker to safely be able to reorder sections at function granularity, -ffunction-sections would be required. This isn’t a problem during LTO since the code generation is set up by the linker :slight_smile:

Also, for clang on OSX the best layout we could get is to order functions in the order in which they get executed at runtime.

What the optimal layout may be for given apps is a bit of a separate question. Right now we’re mostly talking about how to plumb everything together so that we can do the reordering of the final executable.

Yes, I was raising this exactly with the idea of “we may want to try different algorithm based on different kind of data”.

In fact, standard ELF linking semantics generally require input sections to be concatenated in command line order (this is e.g. how .init_array/.ctors build up their arrays of pointers to initializers; a crt*.o file at the beginning/end has a sentinel value and so the order matters). So the linker will generally need blessing from the compiler to do most sorts of reorderings as far as I’m aware.

Other signals besides profile info, such as a startup trace, might be useful too, and we should make sure we can plug that into the design.

My understanding of the clang on OSX case is based on a comparison of the form_by_* functions in clang/utils/perf-training/perf-helper.py which offer a relatively simple set of algorithms, so I think the jury is still out on the best approach (that script also uses a data collection method that is not part of LLVM’s usual instrumentation or sampling workflows for PGO, so we may not be able to provide the same signals out of the box as part of our standard offering in the compiler)

Yes, I was thinking that some Xray-based instrumentation could be used to provided the same data.

I think that once we have this ordering capability integrated more deeply into the compiler, we’ll be able to evaluate more complicated algorithms like Pettis-Hansen, have access to signals like global profile info, do interesting call graph analyses, etc. to find interesting approaches.

For FullLTO it is conceptually pretty easy to get profile data we need for this, but I’m not sure about the ThinLTO case.

Teresa, Mehdi,

Are there any plans (or things already working!) for getting profile data from ThinLTO in a format that the linker can use for code layout? I assume that profile data is being used already to guide importing, so it may just be a matter of siphoning that off.

I’m not sure what kind of “profile information” is needed, and what makes it easier for MonolithicLTO compared to ThinLTO?

For MonolithicLTO I had in mind that a simple implementation would be:

std::vector<std::string> Ordering;
auto Pass = make_unique<LayoutModulePass>(&Ordering);
addPassToLTOPipeline(std::move(Pass))

The module pass would just query the profile data directly on IR datastructures and get the order out. This would require very little “plumbing”.

Or maybe that layout code should be inside LLVM; maybe part of the general LTO interface? It looks like the current gcc plugin calls back into gcc for the actual layout algorithm itself (function call find_pettis_hansen_function_layout) rather than the reordering logic living in the linker: https://android.googlesource.com/toolchain/gcc/+/3f73d6ef90458b45bbbb33ef4c2b174d4662a22d/gcc-4.6/function_reordering_plugin/function_reordering_plugin.c

I was thinking about this: could this be done by reorganizing the module itself for LTO?

For MonolithicLTO that’s another simple approach.

That wouldn’t help non-LTO and ThinLTO though.

I think we should ideally aim for something that works uniformly for Monolithic and Thin. For example, GCC emits special sections containing the profile data and the linker just reads those sections; something analogous in LLVM would just happen in the backend and be common to Monolithic and Thin. If ThinLTO already has profile summaries in some nice form though, it may be possible to bypass this.

Another advantage of using special sections in the output like GCC does is that you don’t actually need LTO at all to get the function reordering. The profile data passed to the compiler during per-TU compilation can be lowered into the same kind of annotations. (though LTO and function ordering are likely to go hand-in-hand most often for peak-performance builds).

Yes I agree with all of this :slight_smile:
That makes it for interesting design trade-off!

>
>
>
>
>>
>> > Do we have any open projects on LLD?
>> >
>> > I know we usually try to avoid any big "projects" and mainly add/fix
>> > things
>> > in response to user needs, but just wondering if somebody has any
ideas.
>> >
>>
>> I'm not particularly active in lld anymore, but the last big item I'd
>> like to see implemented is Pettis-Hansen layout.
>> http://perso.ensta-paristech.fr/~bmonsuez/Cours/B6-4/Article
s/papers15.pdf
>> (mainly because it improves performances of the final executable).
>> GCC/gold have an implementation of the algorithm that can be used as
>> base. I'll expand if anybody is interested.
>> Side note: I'd like to propose a couple of llvm projects as well, I'll
>> sit down later today and write them.
>
>
>
> I’m not sure, can you confirm that such layout optimization on ELF
requires
> -ffunction-sections?
>

For the non-LTO case, I think so.

> Also, for clang on OSX the best layout we could get is to order
functions in
> the order in which they get executed at runtime.
>

That's what we already do for lld. We collect and order file (run a
profiler) and pass that to the linker that lays out functions
accordingly.
This is to improve startup time for a class of startup-time-sensitive
operations. The algorithm proposed by Pettis (allegedly) aims to
reduce the TLB misses as it tries to lay out hot functions (or
functions that are likely to be called together near in the final
binary).

IIRC from when I looked at the paper a while ago, it is mostly just a
"huffman tree construction" type algorithm (agglomerating based on highest
probability) and assumes that if two functions are hot then they are likely
to be needed together. This is not always the case.

E.g. consider a server that accepts RPC requests and based on those
requests either does Foo or Bar which are largely disjoint. It's entirely
possible for the top two functions of the profile to be one in Foo and one
in Bar, but laying them out near each other doesn't make sense since there
is never locality (for a given RPC, either Foo or Bar gets run). A static
call graph analysis can provide the needed signals to handle this case
better.

Hence you said "allegedly" :slight_smile: I know we've talked about this before. Just
wanted to put the backstory of the "allegedly" on the list.

Looks like I remembered this wrong. The algorithm in section 3.2 of the
paper is call-graph aware. It does do greedy coalescing like a Huffman tree
construction algorithms, but constrains the available coalescing operations
at each step by call graph adjacency (in fact, what it is "greedy" about is
the hotness of the edges between call graph nodes and not the nodes
themselves).

-- Sean Silva

> Do we have any open projects on LLD?
>
> I know we usually try to avoid any big "projects" and mainly add/fix
things
> in response to user needs, but just wondering if somebody has any
ideas.
>

I'm not particularly active in lld anymore, but the last big item I'd
like to see implemented is Pettis-Hansen layout.
http://perso.ensta-paristech.fr/~bmonsuez/Cours/B6-4/Article
s/papers15.pdf
(mainly because it improves performances of the final executable).
GCC/gold have an implementation of the algorithm that can be used as
base. I'll expand if anybody is interested.
Side note: I'd like to propose a couple of llvm projects as well, I'll
sit down later today and write them.

I’m not sure, can you confirm that such layout optimization on ELF
requires -ffunction-sections?

In order for a standard ELF linker to safely be able to reorder sections
at function granularity, -ffunction-sections would be required. This isn't
a problem during LTO since the code generation is set up by the linker :slight_smile:

Also, for clang on OSX the best layout we could get is to order functions
in the order in which they get executed at runtime.

What the optimal layout may be for given apps is a bit of a separate
question. Right now we're mostly talking about how to plumb everything
together so that we can do the reordering of the final executable.

Yes, I was raising this exactly with the idea of “we may want to try
different algorithm based on different kind of data”.

In fact, standard ELF linking semantics generally require input sections
to be concatenated in command line order (this is e.g. how
.init_array/.ctors build up their arrays of pointers to initializers; a
crt*.o file at the beginning/end has a sentinel value and so the order
matters). So the linker will generally need blessing from the compiler to
do most sorts of reorderings as far as I'm aware.

Other signals besides profile info, such as a startup trace, might be
useful too, and we should make sure we can plug that into the design.
My understanding of the clang on OSX case is based on a comparison of the
`form_by_*` functions in clang/utils/perf-training/perf-helper.py which
offer a relatively simple set of algorithms, so I think the jury is still
out on the best approach (that script also uses a data collection method
that is not part of LLVM's usual instrumentation or sampling workflows for
PGO, so we may not be able to provide the same signals out of the box as
part of our standard offering in the compiler)

Yes, I was thinking that some Xray-based instrumentation could be used to
provided the same data.

I hadn't though of using Xray for this! Good idea! (I haven't been
following Xray very closely; I should look at it more...)

-- Sean Silva

Google GCC records profile data (dynamic callgraph) in a special named section in ELF object file to be consumed by the plugin. Those sections will be discarded later by the linker.

There are pros and cons of using xray for layout purpose. The call trace from xray is certainly more powerful for layout purpose, but it adds addtional complexity to the optimized build process. You would need to collect xray trace profile on the optimized binary (presumably built with PGO already) and rebuild without xray nop insertion and function layout.

David

Would it make sense for xray instrumentation be part of -fprofile-generate? PGO will affect inlining decisions etc for the optimized binary, but the collected traces during the instrumented build would still have quite a bit of useful information.

– Sean Silva