llvm dev - Jul 2016 - Representing MIPS ABI information in the triple as ARM/X86 do for EABI/EABIHF/X32

Hi,

Having recently enabled IAS by default for the MIPS O32 ABI, I'm now trying
to do the same thing for the MIPS N64 ABI. Unfortunately, it is not currently
possible to enable IAS by default for the N64 ABI without also enabling it for
the N32 ABI because this information is not reflected in the triple and
that's the only information MipsMCAsmInfo has. This would be fine if it N32
was also in a good state but the current N32 ABI support for IAS is badly broken
and will likely take considerable effort to fix (and fixing it also requires
solving the same key problem as enabling IAS for just N64). I therefore want to
separate the two ABI's so that I can finish off N64 and then fix N32
afterwards.

I've posted a series of patches that allow us to distinguish N32 and N64 so
that we can enable IAS for only N64  (D21465, D21467, D21069 for LLVM along with
D21070, D21072 for clang). During the review of D21467, Rafael asked me to
explain my approach here. I'm hoping that people will agree that this is an
acceptable approach to take and that it's no different from how other
targets handle certain ABI's and how we already handle
-m32/-m64/-EL/-EB/etc.

The approach I'm using is to encode the ABI information into the triple in
the same way that ARM does for EABI and hardfloat, and the same way that X86
does for X32. These targets define variants of Triple::GNU for each of these
ABI's in the form of the Triple::GNUEABI, Triple::GNUEABIHF, and
Triple::GNUX32 values of the environment component of the triple and it's up
to the frontend and/or API-user to pass the right thing to the backend.

For MIPS, I'm defining Triple::GNUABI32, Triple::GNUABIN32, and
Triple::GNUABI64. All three of these are supported by tools like binutils (by
virtue of a wildcard match '*linux-gnu*') and of these three,
Triple::GNUABI32 is not in general use, Triple::GNUABIN32 will be used by Debian
if/when we do an N32 port, and Triple::GNUABI64 is currently used by Debian for
our N64 port. Once the ABI is in our backend triples, it can be used by
MipsMCAsmInfo to enable IAS by default for N64 without also enabling it for N32.
In later patches, such as D1292 this same information will also be used to fix
the N32 support for IAS.

At this point, you may be wondering what happens to Triple::GNU for triples like
mips-linux-gnu and mips64-linux-gnu. The answer to this is that the API-user
(e.g. clang) is expected to normalize such triples down to one of
Triple::GNUABI32/GNUABIN32/GNUABI64. This can be as simple as adding:
  Triple ABITriple = TT.getABIVariant(ABIName); // ABIName can be the empty
string to get the default ABI.
  if (ABITriple.getArch() != Triple::UnknownArch) {
    TT = ABITriple;
    ABIName = ""; // <- Only needed if this would end up in
MCTargetOptions::ABIName.
  }
to the appropriate place in the caller. This is the same way clang handles the
-m32, -m64, -EL, and -EB options with those options using get32BitArchVariant(),
get64BitArchVariant(), getLittleEndianArchVariant(), and
getBigEndianArchVariant() respectively to transform the triple for the backend.

In summary, human end users of tools such as clang will continue using the same
triples as they always have. These tools will then resolve them down to the
effective triple (for example, in functions like clang's
ComputeEffectiveClangTriple(), ComputeLLVMTriple(), and computeTargetTriple()).
The MIPS backend will only see the triples that have a specific known ABI.
I'd like to stress that this process is the same as the one we already use
today for other options. For example 'x86_64-linux-gnu-clang -m32'
transforms the triple with get32BitArchVariant() and passes
'i386-linux-gnu' to the backend.

So far I've only been talking about Linux/GNU but the patch series also
covers Linux/Musl, Linux/Android, and FreeBSD. I haven't talked about those
because they are handled in the same way as the Linux/GNU case described above
except that they use Triple::ABI32/ABIN32/ABI64 (or Android32/Android64 in the
case of Android) instead of Triple::GNUABI32/GNUABIN32/GNUABI64.

Why is the Triple approach better for MIPS than MCTargetOptions::ABIName?
========================================================================
A major pragmatic reason is that it works today and the MCTargetOption::ABIName
approach doesn't. It's also in keeping with other targets whereas until
fairly recently MIPS was the sole user of MCTargetOptions::ABIName. There is
currently a single PowerPC test and six ARM tests that use the -target-abi
option that feeds MCTargetOptions::ABIName at the moment and both of these
targets are only using it for relatively minor tweaks to the ABI.

As things are today, MCTargetOptions::ABIName is good at fairly small tweaks to
the ABI such as calling convention changes but it's unable to deal with more
fundamental changes such as selecting between ELFCLASS32/ELFCLASS64 or enabling
IAS. Some of these limitations look fixable if given months-years of effort but
I can see potential problems once the obvious plumbing problem is fixed. Some of
these problems such as preventing the linking of incompatible IR files in the
IRLinker would require me to represent the ABI in the LLVM-IR (it's
currently handled solely by the triple) which has been strongly opposed in
earlier discussions about triples.

I can provide more detail on some of the specific problems if required but this
boils down to a choice between an easy paved road that other people use and that
I know will end in a working state, or an unexplored path with known
difficulties and an uncertain result.

On 24 June 2016 at 11:02, Daniel Sanders via llvm-dev
<llvm-dev at lists.llvm.org> wrote:
> The approach I'm using is to encode the ABI information into the triple
in the same way that ARM does for EABI and hardfloat, and the same way that X86
does for X32. These targets define variants of Triple::GNU for each of these
ABI's in the form of the Triple::GNUEABI, Triple::GNUEABIHF, and
Triple::GNUX32 values of the environment component of the triple and it's up
to the frontend and/or API-user to pass the right thing to the backend.
Hi Daniel,

The precedent is there, so I don't think this should be a contentious issue.

> For MIPS, I'm defining Triple::GNUABI32, Triple::GNUABIN32, and
Triple::GNUABI64. All three of these are supported by tools like binutils (by
virtue of a wildcard match '*linux-gnu*') and of these three,
Triple::GNUABI32 is not in general use, Triple::GNUABIN32 will be used by Debian
if/when we do an N32 port, and Triple::GNUABI64 is currently used by Debian for
our N64 port. Once the ABI is in our backend triples, it can be used by
MipsMCAsmInfo to enable IAS by default for N64 without also enabling it for N32.
In later patches, such as D1292 this same information will also be used to fix
the N32 support for IAS.
This "should" work.

> In summary, human end users of tools such as clang will continue using the
same triples as they always have. These tools will then resolve them down to the
effective triple (for example, in functions like clang's
ComputeEffectiveClangTriple(), ComputeLLVMTriple(), and computeTargetTriple()).
The MIPS backend will only see the triples that have a specific known ABI.
I'd like to stress that this process is the same as the one we already use
today for other options. For example 'x86_64-linux-gnu-clang -m32'
transforms the triple with get32BitArchVariant() and passes
'i386-linux-gnu' to the backend.
We do similar, but not completely.

For instance, "-target arm-linux-gnueabi -mthumb -mfloat-abi=hard
-mcpu=cortex=a15" will change the triple to
"thumbv7--linux-gnueabi"
not the expected "thumbv7--linux-gnueabihf".

Though, that's probably a bug. :)

cheers,
--renato

On 24 June 2016 at 06:02, Daniel Sanders <Daniel.Sanders at imgtec.com>
wrote:
> Hi,
>
> Having recently enabled IAS by default for the MIPS O32 ABI, I'm now
trying to do the same thing for the MIPS N64 ABI. Unfortunately, it is not
currently possible to enable IAS by default for the N64 ABI without also
enabling it for the N32 ABI because this information is not reflected in the
triple and that's the only information MipsMCAsmInfo has.
What would it take to pass more information to MipsMCAsmInfo?
> For MIPS, I'm defining Triple::GNUABI32, Triple::GNUABIN32, and
Triple::GNUABI64. All three of these are supported by tools like binutils (by
virtue of a wildcard match '*linux-gnu*') and of these three,
Triple::GNUABI32 is not in general use, Triple::GNUABIN32 will be used by Debian
if/when we do an N32 port, and Triple::GNUABI64 is currently used by Debian for
our N64 port. Once the ABI is in our backend triples, it can be used by
MipsMCAsmInfo to enable IAS by default for N64 without also enabling it for N32.
In later patches, such as D1292 this same information will also be used to fix
the N32 support for IAS.
I would be perfectly fine with this if instead of just hitting an
wildcard these different triples actually meant different things to
binutils.

Also, do you plan to deprecate the other ways of passing mips abi info
and use only a normalized triple? If so I am OK with adding these
extra triples.

Eric, what do you think?

Cheers,
Rafael

Thanks Renato and Rafael.
> On 24 June 2016 at 06:02, Daniel Sanders <Daniel.Sanders at
imgtec.com>
> wrote:
> > Hi,
> >
> > Having recently enabled IAS by default for the MIPS O32 ABI, I'm
now
> trying to do the same thing for the MIPS N64 ABI. Unfortunately, it is not
> currently possible to enable IAS by default for the N64 ABI without also
> enabling it for the N32 ABI because this information is not reflected in
the
> triple and that's the only information MipsMCAsmInfo has.
> 
> What would it take to pass more information to MipsMCAsmInfo?
All callers of createMCAsmInfo() and LLVMCreateDisasmCPUFeatures() would need to
specify the ABI. In the case of LLVMCreateDisasmCPUFeatures(), they'd move
to a new variant of this function and some backward compatibility fixups would
be left in the pre-existing version to maintain its current level of
functionality where only the default ABI for the triple is available.
IRObjectFile would need some means of determining the ABI otherwise
IRObjectFile::CollectAsmUndefinedRefs() would produce a slightly different
result for some assembly inputs. This is because a small number of directives
introduce references on particular ABI's and not on others. I think this
information would be best delivered via the LLVM-IR but adding the ABI in
LLVM-IR files has been strongly opposed in previous discussions. Also,
IRObjectFile may need the CPU due to some changes to relocation rules for
MIPS32R6 that prevent elimination of relocations that can be evaluated locally
which would lead to the same issue.

The other callers of createMCAsmInfo() either already have the ABI by one means
or another, or would be able to figure it out if we extend ObjectFile to detect
the ABI from the object*. Some of the tools that call createMCAsmInfo() link
multiple object files (e.g. llvm-rtdyld, and llvm-dwp) and may therefore need to
implement the ABI inter-linking rules.

Looking beyond enabling IAS for N64 towards fixing N32 support for IAS, the same
requirements would apply to createMCSubtargetInfo(), createAsmPrinter() (because
the Mips, PowerPC, AMDGPU, and WebAssembly targets call createMCSubtargetInfo()
in their implementations), createMCCodeEmitter(), and createMCAsmBackend().
IRLinker would also need to handle the incompatibility between the O32 and
N32/N64 ABI's IR-level since frontends can generate different IR for
different calling conventions. Other targets handle this incompatibility by
having different triples. I think N32 and N64 are compatible at the LLVM-IR
level but I haven't proven this and it's moot since they aren't
compatible with O32 which should also be allowed on MIPS64.

* This bit is slightly different from the description in
http://reviews.llvm.org/D20916 where I believed many tools needed to provide the
ABI but lacked the information to do so. Since writing that, I have added a
facility to ObjectFile that allows llvm-objdump to detect the feature bits from
the Object header. The same technique can be used for other tools that handle
ObjectFiles.
> > For MIPS, I'm defining Triple::GNUABI32, Triple::GNUABIN32, and
> > Triple::GNUABI64. All three of these are supported by tools like
binutils (by
> > virtue of a wildcard match '*linux-gnu*') and of these three,
> > Triple::GNUABI32 is not in general use, Triple::GNUABIN32 will be used
by
> > Debian if/when we do an N32 port, and Triple::GNUABI64 is currently
used
> > by Debian for our N64 port. Once the ABI is in our backend triples, it
can be
> > used by MipsMCAsmInfo to enable IAS by default for N64 without also
> > enabling it for N32. In later patches, such as D1292 this same
information will
> > also be used to fix the N32 support for IAS.
> 
> I would be perfectly fine with this if instead of just hitting an
> wildcard these different triples actually meant different things to
> binutils.
That doesn't seem necessary since 'gnueabi' and 'gnueabihf'
seem to be matched by a 'gnu*' wildcard too but I'm in favour of
defining these triples in binutils if that's necessary to fix LLVM. I'd
appreciate it if this patch series didn't have to wait until the triples are
present in the binutils source though. Are you ok with us going ahead with this
patch series while a corresponding binutils patch is being put together?

I should also mention that Debian's binutils package already contains
downstream patches to define 'gnuabi64' and select the appropriate
default behaviour. I assume this will be upstreamed in the near future.
> Also, do you plan to deprecate the other ways of passing mips abi info
> and use only a normalized triple? If so I am OK with adding these
> extra triples.
Yes, http://reviews.llvm.org/D21069 makes it mandatory to pass the ABI in the
triple.
> Eric, what do you think?
> 
> Cheers,
> Rafael

Hi Daniel,

While I'm sympathetic to the amount of work associated this appears to be
attempting to work around an API problem by changing triples needlessly.
There's already some ABI support in MCTargetOptions that I used when fixing
up the ARM port a while back and this use can either be extended or changed
depending on what's necessary. It may require plumbing, but that's
expected
with an API being expanded for new uses. It appears that the command line
syntax is sufficient for the needs of enumerating target plus abi and if we
need to add more defaults then it shouldn't be too bad to do so.

Even if it weren't an API problem I'm not sure it'd be a good idea.
I think
we could add a with-abi configuration option to match the current gcc one
rather than tacking it onto triple parsing.

That said, I'm more than happy to help you detangle the existing API and
come up with a good strategy for the MC level API.

Thanks!

-eric

ps. It also doesn't appear that Matthew (who took over my gcc
responsibilities, thanks!) has added support for this in gcc yet either.

echristo at dzur ~/b/build-gcc-mips64abin64> .../configure
--target=mips64-linux-gnuabi64 --enable-languages=c
echristo at dzur ~/tmp> ~/builds/build-gcc-mips64abin64/gcc/cc1 foo.c -o -
.file 1 "foo.c"
.section .mdebug.abiN32
...

On Fri, Jun 24, 2016 at 3:02 AM Daniel Sanders <Daniel.Sanders at
imgtec.com>
wrote:
> Hi,
>
> Having recently enabled IAS by default for the MIPS O32 ABI, I'm now
> trying to do the same thing for the MIPS N64 ABI. Unfortunately, it is not
> currently possible to enable IAS by default for the N64 ABI without also
> enabling it for the N32 ABI because this information is not reflected in
> the triple and that's the only information MipsMCAsmInfo has. This
would be
> fine if it N32 was also in a good state but the current N32 ABI support for
> IAS is badly broken and will likely take considerable effort to fix (and
> fixing it also requires solving the same key problem as enabling IAS for
> just N64). I therefore want to separate the two ABI's so that I can
finish
> off N64 and then fix N32 afterwards.
>
> I've posted a series of patches that allow us to distinguish N32 and
N64
> so that we can enable IAS for only N64  (D21465, D21467, D21069 for LLVM
> along with D21070, D21072 for clang). During the review of D21467, Rafael
> asked me to explain my approach here. I'm hoping that people will agree
> that this is an acceptable approach to take and that it's no different
from
> how other targets handle certain ABI's and how we already handle
> -m32/-m64/-EL/-EB/etc.
>
> The approach I'm using is to encode the ABI information into the triple
in
> the same way that ARM does for EABI and hardfloat, and the same way that
> X86 does for X32. These targets define variants of Triple::GNU for each of
> these ABI's in the form of the Triple::GNUEABI, Triple::GNUEABIHF, and
> Triple::GNUX32 values of the environment component of the triple and
it's
> up to the frontend and/or API-user to pass the right thing to the backend.
>
> For MIPS, I'm defining Triple::GNUABI32, Triple::GNUABIN32, and
> Triple::GNUABI64. All three of these are supported by tools like binutils
> (by virtue of a wildcard match '*linux-gnu*') and of these three,
> Triple::GNUABI32 is not in general use, Triple::GNUABIN32 will be used by
> Debian if/when we do an N32 port, and Triple::GNUABI64 is currently used by
> Debian for our N64 port. Once the ABI is in our backend triples, it can be
> used by MipsMCAsmInfo to enable IAS by default for N64 without also
> enabling it for N32. In later patches, such as D1292 this same information
> will also be used to fix the N32 support for IAS.
>
> At this point, you may be wondering what happens to Triple::GNU for
> triples like mips-linux-gnu and mips64-linux-gnu. The answer to this is
> that the API-user (e.g. clang) is expected to normalize such triples down
> to one of Triple::GNUABI32/GNUABIN32/GNUABI64. This can be as simple as
> adding:
>   Triple ABITriple = TT.getABIVariant(ABIName); // ABIName can be the
> empty string to get the default ABI.
>   if (ABITriple.getArch() != Triple::UnknownArch) {
>     TT = ABITriple;
>     ABIName = ""; // <- Only needed if this would end up in
> MCTargetOptions::ABIName.
>   }
> to the appropriate place in the caller. This is the same way clang handles
> the -m32, -m64, -EL, and -EB options with those options using
> get32BitArchVariant(), get64BitArchVariant(), getLittleEndianArchVariant(),
> and getBigEndianArchVariant() respectively to transform the triple for the
> backend.
>
> In summary, human end users of tools such as clang will continue using the
> same triples as they always have. These tools will then resolve them down
> to the effective triple (for example, in functions like clang's
> ComputeEffectiveClangTriple(), ComputeLLVMTriple(), and
> computeTargetTriple()). The MIPS backend will only see the triples that
> have a specific known ABI. I'd like to stress that this process is the
same
> as the one we already use today for other options. For example
> 'x86_64-linux-gnu-clang -m32' transforms the triple with
> get32BitArchVariant() and passes 'i386-linux-gnu' to the backend.
>
> So far I've only been talking about Linux/GNU but the patch series also
> covers Linux/Musl, Linux/Android, and FreeBSD. I haven't talked about
those
> because they are handled in the same way as the Linux/GNU case described
> above except that they use Triple::ABI32/ABIN32/ABI64 (or
> Android32/Android64 in the case of Android) instead of
> Triple::GNUABI32/GNUABIN32/GNUABI64.
>
> Why is the Triple approach better for MIPS than MCTargetOptions::ABIName?
>
========================================================================>
> A major pragmatic reason is that it works today and the
> MCTargetOption::ABIName approach doesn't. It's also in keeping with
other
> targets whereas until fairly recently MIPS was the sole user of
> MCTargetOptions::ABIName. There is currently a single PowerPC test and six
> ARM tests that use the -target-abi option that feeds
> MCTargetOptions::ABIName at the moment and both of these targets are only
> using it for relatively minor tweaks to the ABI.
>
> As things are today, MCTargetOptions::ABIName is good at fairly small
> tweaks to the ABI such as calling convention changes but it's unable to
> deal with more fundamental changes such as selecting between
> ELFCLASS32/ELFCLASS64 or enabling IAS. Some of these limitations look
> fixable if given months-years of effort but I can see potential problems
> once the obvious plumbing problem is fixed. Some of these problems such as
> preventing the linking of incompatible IR files in the IRLinker would
> require me to represent the ABI in the LLVM-IR (it's currently handled
> solely by the triple) which has been strongly opposed in earlier
> discussions about triples.
>
> I can provide more detail on some of the specific problems if required but
> this boils down to a choice between an easy paved road that other people
> use and that I know will end in a working state, or an unexplored path with
> known difficulties and an uncertain result.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20160701/e4aece72/attachment.html>

Hi Eric,

It's the unsolved problems on the pass-MCTargetOptions-everywhere path that
are my main concern with that approach rather than the amount of work. The first
problem is that the result of IRObjectFile::CollectAsmUndefinedRefs() depends on
the ABI but IRObjectFile doesn't know it. How would you deliver the ABI to
IRObjectFile? The second problem is that IRLinker will link incompatible IR
together because it doesn't know the ABI's involved in the link. It
thinks it's sufficient to check that the triples are the same. How would you
prevent it from linking IR intended for O32 with IR intended for N64?

I'm also worried about the risk of undiscovered problems on the
pass-MCTargetOptions-everywhere path but it's difficult to discuss those for
obvious reasons. However, I think it's safe to say that there could be some
major problems waiting to be discovered given that we don't have any targets
that implement a major ABI with this approach.
> There's already some ABI support in MCTargetOptions that I used when
fixing up the ARM port a while back and this use can either be extended or
changed depending on what's necessary
That's correct and r224492 is quite similar to what I'm trying to do in
this patch series (see below). I should mention that even that commit relies on
the ABI being encoded in the triple though. In computeTargetABI() from
ARMTargetMachine.cpp, MCTargetOptions::ABIName specifies the ABI if it is a
non-empty string but the code otherwise falls back on extracting the ABI from
the triple. The clang driver has similar code, it determines the value for the
–target-abi option from –mabi if it's given and falls back on calculating it
from the triple using the same code as computeTargetABI(). The overall effect is
that the ABI information is delivered to some parts of LLVM via both
MCTargetOptions::ABIName and the triple, and other parts solely via the triple
(e.g. ARMAsmPrinter and ARMISelLowering). Both copies of this information always
agree when –mabi is not given but they may differ when –mabi is given.

As I mentioned above, I'm trying to do a very similar thing to r224492 but I
have an additional problem to deal with. In my scenario, clang doesn't start
off with a triple that clearly specifies the ABI (mips64-linux-gnu can be N32 or
N64 depending on –mabi) which makes it impossible for the bits of LLVM that
can't see MCTargetOptions to know what to do.  I therefore start by
normalizing the triple in clang to one that does specify the ABI clearly and is
consistent with the –mabi option. With that problem solved, the rest of the
patch series is consistent with r224492 except that MCTargetOptions::ABIName and
the triple can never disagree. Personally, I would stop the patch series at this
point and not go on to duplicate the ABI information into
MCTargetOptions::ABIName (because this permits inconsistency and it doesn't
make sense to put N64 code inside an O32 ELF) but if you feel strongly about
this then I can add another patch that causes clang to pass the information in
both the triple and the –target-abi option and have the Mips backend accept
this. With that additional patch, this patch series will be completely
consistent with r224492.
> Even if it weren't an API problem I'm not sure it'd be a good
idea. I think we could add a with-abi configuration option to match the current
gcc one rather than tacking it onto triple parsing.
I thought you were strongly against configure-time options based on previous
discussions. Has this changed?
> ps. It also doesn't appear that Matthew (who took over my gcc
responsibilities, thanks!) has added support for this in gcc yet either.
> echristo at dzur ~/b/build-gcc-mips64abin64> .../configure
--target=mips64-linux-gnuabi64 --enable-languages=c
> echristo at dzur ~/tmp> ~/builds/build-gcc-mips64abin64/gcc/cc1 foo.c -o
-
>          .file      1 "foo.c"
>          .section .mdebug.abiN32
It's true that upstream GCC doesn't know about this triple yet but
that's because downstream users of GCC don't need to push their changes
upstream when they redefine triples. Fedora and Gentoo both use --with-abi while
Debian uses source patches. Upstream GCC only contains the default-defaults
which may be different from the actual-defaults that are compiled in to the
toolchain. In contrast, LLVM requires all the (often conflicting) meanings of a
triple to be upstream. This is part of a separate problem that I plan to address
in a later discussion so I'll leave this for another thread.

From: Eric Christopher [mailto:echristo at gmail.com]
Sent: 01 July 2016 23:05
To: Daniel Sanders; llvm-dev at lists.llvm.org
Cc: rafael.espindola at gmail.com; Matthew Fortune
Subject: Re: Representing MIPS ABI information in the triple as ARM/X86 do for
EABI/EABIHF/X32

Hi Daniel,

While I'm sympathetic to the amount of work associated this appears to be
attempting to work around an API problem by changing triples needlessly.
There's already some ABI support in MCTargetOptions that I used when fixing
up the ARM port a while back and this use can either be extended or changed
depending on what's necessary. It may require plumbing, but that's
expected with an API being expanded for new uses. It appears that the command
line syntax is sufficient for the needs of enumerating target plus abi and if we
need to add more defaults then it shouldn't be too bad to do so.

Even if it weren't an API problem I'm not sure it'd be a good idea.
I think we could add a with-abi configuration option to match the current gcc
one rather than tacking it onto triple parsing.

That said, I'm more than happy to help you detangle the existing API and
come up with a good strategy for the MC level API.

Thanks!

-eric

ps. It also doesn't appear that Matthew (who took over my gcc
responsibilities, thanks!) has added support for this in gcc yet either.

echristo at dzur ~/b/build-gcc-mips64abin64> .../configure
--target=mips64-linux-gnuabi64 --enable-languages=c
echristo at dzur ~/tmp> ~/builds/build-gcc-mips64abin64/gcc/cc1 foo.c -o -
          .file      1 "foo.c"
          .section .mdebug.abiN32
...

On Fri, Jun 24, 2016 at 3:02 AM Daniel Sanders <Daniel.Sanders at
imgtec.com<mailto:Daniel.Sanders at imgtec.com>> wrote:
Hi,

Having recently enabled IAS by default for the MIPS O32 ABI, I'm now trying
to do the same thing for the MIPS N64 ABI. Unfortunately, it is not currently
possible to enable IAS by default for the N64 ABI without also enabling it for
the N32 ABI because this information is not reflected in the triple and
that's the only information MipsMCAsmInfo has. This would be fine if it N32
was also in a good state but the current N32 ABI support for IAS is badly broken
and will likely take considerable effort to fix (and fixing it also requires
solving the same key problem as enabling IAS for just N64). I therefore want to
separate the two ABI's so that I can finish off N64 and then fix N32
afterwards.

I've posted a series of patches that allow us to distinguish N32 and N64 so
that we can enable IAS for only N64  (D21465, D21467, D21069 for LLVM along with
D21070, D21072 for clang). During the review of D21467, Rafael asked me to
explain my approach here. I'm hoping that people will agree that this is an
acceptable approach to take and that it's no different from how other
targets handle certain ABI's and how we already handle
-m32/-m64/-EL/-EB/etc.

The approach I'm using is to encode the ABI information into the triple in
the same way that ARM does for EABI and hardfloat, and the same way that X86
does for X32. These targets define variants of Triple::GNU for each of these
ABI's in the form of the Triple::GNUEABI, Triple::GNUEABIHF, and
Triple::GNUX32 values of the environment component of the triple and it's up
to the frontend and/or API-user to pass the right thing to the backend.

For MIPS, I'm defining Triple::GNUABI32, Triple::GNUABIN32, and
Triple::GNUABI64. All three of these are supported by tools like binutils (by
virtue of a wildcard match '*linux-gnu*') and of these three,
Triple::GNUABI32 is not in general use, Triple::GNUABIN32 will be used by Debian
if/when we do an N32 port, and Triple::GNUABI64 is currently used by Debian for
our N64 port. Once the ABI is in our backend triples, it can be used by
MipsMCAsmInfo to enable IAS by default for N64 without also enabling it for N32.
In later patches, such as D1292 this same information will also be used to fix
the N32 support for IAS.

At this point, you may be wondering what happens to Triple::GNU for triples like
mips-linux-gnu and mips64-linux-gnu. The answer to this is that the API-user
(e.g. clang) is expected to normalize such triples down to one of
Triple::GNUABI32/GNUABIN32/GNUABI64. This can be as simple as adding:
  Triple ABITriple = TT.getABIVariant(ABIName); // ABIName can be the empty
string to get the default ABI.
  if (ABITriple.getArch() != Triple::UnknownArch) {
    TT = ABITriple;
    ABIName = ""; // <- Only needed if this would end up in
MCTargetOptions::ABIName.
  }
to the appropriate place in the caller. This is the same way clang handles the
-m32, -m64, -EL, and -EB options with those options using get32BitArchVariant(),
get64BitArchVariant(), getLittleEndianArchVariant(), and
getBigEndianArchVariant() respectively to transform the triple for the backend.

In summary, human end users of tools such as clang will continue using the same
triples as they always have. These tools will then resolve them down to the
effective triple (for example, in functions like clang's
ComputeEffectiveClangTriple(), ComputeLLVMTriple(), and computeTargetTriple()).
The MIPS backend will only see the triples that have a specific known ABI.
I'd like to stress that this process is the same as the one we already use
today for other options. For example 'x86_64-linux-gnu-clang -m32'
transforms the triple with get32BitArchVariant() and passes
'i386-linux-gnu' to the backend.

So far I've only been talking about Linux/GNU but the patch series also
covers Linux/Musl, Linux/Android, and FreeBSD. I haven't talked about those
because they are handled in the same way as the Linux/GNU case described above
except that they use Triple::ABI32/ABIN32/ABI64 (or Android32/Android64 in the
case of Android) instead of Triple::GNUABI32/GNUABIN32/GNUABI64.

Why is the Triple approach better for MIPS than MCTargetOptions::ABIName?
========================================================================
A major pragmatic reason is that it works today and the MCTargetOption::ABIName
approach doesn't. It's also in keeping with other targets whereas until
fairly recently MIPS was the sole user of MCTargetOptions::ABIName. There is
currently a single PowerPC test and six ARM tests that use the -target-abi
option that feeds MCTargetOptions::ABIName at the moment and both of these
targets are only using it for relatively minor tweaks to the ABI.

As things are today, MCTargetOptions::ABIName is good at fairly small tweaks to
the ABI such as calling convention changes but it's unable to deal with more
fundamental changes such as selecting between ELFCLASS32/ELFCLASS64 or enabling
IAS. Some of these limitations look fixable if given months-years of effort but
I can see potential problems once the obvious plumbing problem is fixed. Some of
these problems such as preventing the linking of incompatible IR files in the
IRLinker would require me to represent the ABI in the LLVM-IR (it's
currently handled solely by the triple) which has been strongly opposed in
earlier discussions about triples.

I can provide more detail on some of the specific problems if required but this
boils down to a choice between an easy paved road that other people use and that
I know will end in a working state, or an unexplored path with known
difficulties and an uncertain result.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20160705/be5f59d3/attachment-0001.html>

>  cc1as. The main difference is that X86 is adjusting the triple in the
> driver while I'm doing it in cc1 and cc1as.
>
>
>
Rereading your email it looks like there was a bit of confusion.

As long as the interface to the driver is still --target mips-linux-gnu
-mabi=XX I care a bit less that you're mangling the abi information into
the triple. I still don't believe it's the correct long term direction
for
the project, but it appears to unblock you with a minimum of work.

Please do, however, make this consistent for every port in the tree. Mips
is not enough of a special snowflake that it needs different ABI handling
than any other port. :)

Thanks.

-eric

> At this point, I think I can safely say that the principle of what I'm
> doing is sound and has solid precedent in other targets (particularly ARM
> and X86). That is, tools and API users select a triple that specifies an
> ABI and the backend is uses the triple as a source of ABI information. Can
> everyone confirm whether we're agreed on that? Assuming we are agreed,
I
> believe the main details we need to agree on are which parts of clang
> adjust the triple and whether the ABI is also passed via
> MCTargetOptions::ABIName if an MCTargetOptions object is available. Are
> there any others?
>
>
>
> For whether the ABI is also passed via MCTargetOptions::ABIName, I
don't
> mind either way in the end but I feel I need to pass through a step where
> we don't pass the ABI name via MCTargetOptions::ABIName on my way to
the
> end point we choose. I feel that passing it both ways introduces potential
> inconsistency between two sources of information and I'd like to
stabilize
> on a consistent state before re-introducing the possibility of
> inconsistency. That said, it's not much harder to go straight to the
end
> point if that's what we prefer. Can everyone confirm which end point
they
> prefer and whether they want to go straight to it or through the
> intermediate step?
>
>
>
> For which parts of clang adjust the triple, I think we have a choice
> between doing it in the driver (like X86's current implementation),
doing
> it in cc1 and cc1as (like my implementation), and doing both. I think that
> both isn't really necessary and that cc1 and cc1as are not really
intended
> for end-users (much like llc/llvm-mc/etc), so I think the right place for
> the triple adjustment is probably the driver. The reason I chose cc1 and
> cc1as in my current patch series is because I had difficulty with the
> adjusted triple getting into various paths and breaking the toolchain. My
> original plan was to reach a working state with the cc1 and cc1as approach
> and then try to push the adjustment into the driver during the spare time
> I'll have while fixing the N32/N64 support in IAS. Can everyone confirm
> whether they think the driver, cc1+cc1as, or both should do the triple
> adjustment? If you pick 'driver' could you also confirm whether
'cc1+cc1as'
> is ok as an intermediate step. If you pick 'both', could you also
confirm
> which one should land first or if both should land at the same time?
>
>
>
> Once we've established a consensus, I'll adjust the patch series to
match
> it.
>
>
>
> Thanks
>
>
>
> *From:* llvm-dev [mailto:llvm-dev-bounces at lists.llvm.org] *On Behalf Of
*Daniel
> Sanders via llvm-dev
> *Sent:* 05 July 2016 13:54
> *To:* Eric Christopher; llvm-dev at lists.llvm.org
> *Subject:* Re: [llvm-dev] Representing MIPS ABI information in the triple
> as ARM/X86 do for EABI/EABIHF/X32
>
>
>
> Hi Eric,
>
>
>
> It's the unsolved problems on the pass-MCTargetOptions-everywhere path
> that are my main concern with that approach rather than the amount of work.
> The first problem is that the result of
> IRObjectFile::CollectAsmUndefinedRefs() depends on the ABI but IRObjectFile
> doesn't know it. How would you deliver the ABI to IRObjectFile? The
second
> problem is that IRLinker will link incompatible IR together because it
> doesn't know the ABI's involved in the link. It thinks it's
sufficient to
> check that the triples are the same. How would you prevent it from linking
> IR intended for O32 with IR intended for N64?
>
>
>
> I'm also worried about the risk of undiscovered problems on the
> pass-MCTargetOptions-everywhere path but it's difficult to discuss
those
> for obvious reasons. However, I think it's safe to say that there could
be
> some major problems waiting to be discovered given that we don't have
any
> targets that implement a major ABI with this approach.
>
>
>
> > There's already some ABI support in MCTargetOptions that I used
when
> fixing up the ARM port a while back and this use can either be extended or
> changed depending on what's necessary
>
>
>
> That's correct and r224492 is quite similar to what I'm trying to
do in
> this patch series (see below). I should mention that even that commit
> relies on the ABI being encoded in the triple though. In computeTargetABI()
> from ARMTargetMachine.cpp, MCTargetOptions::ABIName specifies the ABI if it
> is a non-empty string but the code otherwise falls back on extracting the
> ABI from the triple. The clang driver has similar code, it determines the
> value for the –target-abi option from –mabi if it's given and falls
back on
> calculating it from the triple using the same code as computeTargetABI().
> The overall effect is that the ABI information is delivered to some parts
> of LLVM via both MCTargetOptions::ABIName and the triple, and other parts
> solely via the triple (e.g. ARMAsmPrinter and ARMISelLowering). Both copies
> of this information always agree when –mabi is not given but they may
> differ when –mabi is given.
>
>
>
> As I mentioned above, I'm trying to do a very similar thing to r224492
but
> I have an additional problem to deal with. In my scenario, clang
doesn't
> start off with a triple that clearly specifies the ABI (mips64-linux-gnu
> can be N32 or N64 depending on –mabi) which makes it impossible for the
> bits of LLVM that can't see MCTargetOptions to know what to do.  I
> therefore start by normalizing the triple in clang to one that does specify
> the ABI clearly and is consistent with the –mabi option. With that problem
> solved, the rest of the patch series is consistent with r224492 except that
> MCTargetOptions::ABIName and the triple can never disagree. Personally, I
> would stop the patch series at this point and not go on to duplicate the
> ABI information into MCTargetOptions::ABIName (because this permits
> inconsistency and it doesn't make sense to put N64 code inside an O32
ELF)
> but if you feel strongly about this then I can add another patch that
> causes clang to pass the information in both the triple and the –target-abi
> option and have the Mips backend accept this. With that additional patch,
> this patch series will be completely consistent with r224492.
>
>
>
> > Even if it weren't an API problem I'm not sure it'd be a
good idea. I
> think we could add a with-abi configuration option to match the current gcc
> one rather than tacking it onto triple parsing.
>
>
>
> I thought you were strongly against configure-time options based on
> previous discussions. Has this changed?
>
>
>
> > ps. It also doesn't appear that Matthew (who took over my gcc
> responsibilities, thanks!) has added support for this in gcc yet either.
>
> > echristo at dzur ~/b/build-gcc-mips64abin64> .../configure
> --target=mips64-linux-gnuabi64 --enable-languages=c
>
> > echristo at dzur ~/tmp> ~/builds/build-gcc-mips64abin64/gcc/cc1
foo.c -o -
>
> >          .file      1 "foo.c"
>
> >          .section .mdebug.abiN32
>
>
>
> It's true that upstream GCC doesn't know about this triple yet but
that's
> because downstream users of GCC don't need to push their changes
upstream
> when they redefine triples. Fedora and Gentoo both use --with-abi while
> Debian uses source patches. Upstream GCC only contains the default-defaults
> which may be different from the actual-defaults that are compiled in to the
> toolchain. In contrast, LLVM requires all the (often conflicting) meanings
> of a triple to be upstream. This is part of a separate problem that I plan
> to address in a later discussion so I'll leave this for another thread.
>
>
>
> *From:* Eric Christopher [mailto:echristo at gmail.com]
> *Sent:* 01 July 2016 23:05
> *To:* Daniel Sanders; llvm-dev at lists.llvm.org
> *Cc:* rafael.espindola at gmail.com; Matthew Fortune
> *Subject:* Re: Representing MIPS ABI information in the triple as ARM/X86
> do for EABI/EABIHF/X32
>
>
>
> Hi Daniel,
>
>
>
> While I'm sympathetic to the amount of work associated this appears to
be
> attempting to work around an API problem by changing triples needlessly.
> There's already some ABI support in MCTargetOptions that I used when
fixing
> up the ARM port a while back and this use can either be extended or changed
> depending on what's necessary. It may require plumbing, but that's
expected
> with an API being expanded for new uses. It appears that the command line
> syntax is sufficient for the needs of enumerating target plus abi and if we
> need to add more defaults then it shouldn't be too bad to do so.
>
>
>
> Even if it weren't an API problem I'm not sure it'd be a good
idea. I
> think we could add a with-abi configuration option to match the current gcc
> one rather than tacking it onto triple parsing.
>
>
>
> That said, I'm more than happy to help you detangle the existing API
and
> come up with a good strategy for the MC level API.
>
>
>
> Thanks!
>
>
>
> -eric
>
>
>
> ps. It also doesn't appear that Matthew (who took over my gcc
> responsibilities, thanks!) has added support for this in gcc yet either.
>
>
>
> echristo at dzur ~/b/build-gcc-mips64abin64> .../configure
> --target=mips64-linux-gnuabi64 --enable-languages=c
>
> echristo at dzur ~/tmp> ~/builds/build-gcc-mips64abin64/gcc/cc1 foo.c -o
-
>
>           .file      1 "foo.c"
>
>           .section .mdebug.abiN32
>
> ...
>
>
>
> On Fri, Jun 24, 2016 at 3:02 AM Daniel Sanders <Daniel.Sanders at
imgtec.com>
> wrote:
>
> Hi,
>
> Having recently enabled IAS by default for the MIPS O32 ABI, I'm now
> trying to do the same thing for the MIPS N64 ABI. Unfortunately, it is not
> currently possible to enable IAS by default for the N64 ABI without also
> enabling it for the N32 ABI because this information is not reflected in
> the triple and that's the only information MipsMCAsmInfo has. This
would be
> fine if it N32 was also in a good state but the current N32 ABI support for
> IAS is badly broken and will likely take considerable effort to fix (and
> fixing it also requires solving the same key problem as enabling IAS for
> just N64). I therefore want to separate the two ABI's so that I can
finish
> off N64 and then fix N32 afterwards.
>
> I've posted a series of patches that allow us to distinguish N32 and
N64
> so that we can enable IAS for only N64  (D21465, D21467, D21069 for LLVM
> along with D21070, D21072 for clang). During the review of D21467, Rafael
> asked me to explain my approach here. I'm hoping that people will agree
> that this is an acceptable approach to take and that it's no different
from
> how other targets handle certain ABI's and how we already handle
> -m32/-m64/-EL/-EB/etc.
>
> The approach I'm using is to encode the ABI information into the triple
in
> the same way that ARM does for EABI and hardfloat, and the same way that
> X86 does for X32. These targets define variants of Triple::GNU for each of
> these ABI's in the form of the Triple::GNUEABI, Triple::GNUEABIHF, and
> Triple::GNUX32 values of the environment component of the triple and
it's
> up to the frontend and/or API-user to pass the right thing to the backend.
>
> For MIPS, I'm defining Triple::GNUABI32, Triple::GNUABIN32, and
> Triple::GNUABI64. All three of these are supported by tools like binutils
> (by virtue of a wildcard match '*linux-gnu*') and of these three,
> Triple::GNUABI32 is not in general use, Triple::GNUABIN32 will be used by
> Debian if/when we do an N32 port, and Triple::GNUABI64 is currently used by
> Debian for our N64 port. Once the ABI is in our backend triples, it can be
> used by MipsMCAsmInfo to enable IAS by default for N64 without also
> enabling it for N32. In later patches, such as D1292 this same information
> will also be used to fix the N32 support for IAS.
>
> At this point, you may be wondering what happens to Triple::GNU for
> triples like mips-linux-gnu and mips64-linux-gnu. The answer to this is
> that the API-user (e.g. clang) is expected to normalize such triples down
> to one of Triple::GNUABI32/GNUABIN32/GNUABI64. This can be as simple as
> adding:
>   Triple ABITriple = TT.getABIVariant(ABIName); // ABIName can be the
> empty string to get the default ABI.
>   if (ABITriple.getArch() != Triple::UnknownArch) {
>     TT = ABITriple;
>     ABIName = ""; // <- Only needed if this would end up in
> MCTargetOptions::ABIName.
>   }
> to the appropriate place in the caller. This is the same way clang handles
> the -m32, -m64, -EL, and -EB options with those options using
> get32BitArchVariant(), get64BitArchVariant(), getLittleEndianArchVariant(),
> and getBigEndianArchVariant() respectively to transform the triple for the
> backend.
>
> In summary, human end users of tools such as clang will continue using the
> same triples as they always have. These tools will then resolve them down
> to the effective triple (for example, in functions like clang's
> ComputeEffectiveClangTriple(), ComputeLLVMTriple(), and
> computeTargetTriple()). The MIPS backend will only see the triples that
> have a specific known ABI. I'd like to stress that this process is the
same
> as the one we already use today for other options. For example
> 'x86_64-linux-gnu-clang -m32' transforms the triple with
> get32BitArchVariant() and passes 'i386-linux-gnu' to the backend.
>
> So far I've only been talking about Linux/GNU but the patch series also
> covers Linux/Musl, Linux/Android, and FreeBSD. I haven't talked about
those
> because they are handled in the same way as the Linux/GNU case described
> above except that they use Triple::ABI32/ABIN32/ABI64 (or
> Android32/Android64 in the case of Android) instead of
> Triple::GNUABI32/GNUABIN32/GNUABI64.
>
> Why is the Triple approach better for MIPS than MCTargetOptions::ABIName?
>
========================================================================>
> A major pragmatic reason is that it works today and the
> MCTargetOption::ABIName approach doesn't. It's also in keeping with
other
> targets whereas until fairly recently MIPS was the sole user of
> MCTargetOptions::ABIName. There is currently a single PowerPC test and six
> ARM tests that use the -target-abi option that feeds
> MCTargetOptions::ABIName at the moment and both of these targets are only
> using it for relatively minor tweaks to the ABI.
>
> As things are today, MCTargetOptions::ABIName is good at fairly small
> tweaks to the ABI such as calling convention changes but it's unable to
> deal with more fundamental changes such as selecting between
> ELFCLASS32/ELFCLASS64 or enabling IAS. Some of these limitations look
> fixable if given months-years of effort but I can see potential problems
> once the obvious plumbing problem is fixed. Some of these problems such as
> preventing the linking of incompatible IR files in the IRLinker would
> require me to represent the ABI in the LLVM-IR (it's currently handled
> solely by the triple) which has been strongly opposed in earlier
> discussions about triples.
>
> I can provide more detail on some of the specific problems if required but
> this boils down to a choice between an easy paved road that other people
> use and that I know will end in a working state, or an unexplored path with
> known difficulties and an uncertain result.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20160708/e9537727/attachment-0001.html>

> > cc1as. The main difference is that X86 is adjusting the triple in the
driver while I'm doing it in cc1 and cc1as.
> As long as the interface to the driver is still --target mips-linux-gnu
-mabi=XX I care a bit
> less that you're mangling the abi information into the triple. I still
don't believe it's the
> correct long term direction for the project, but it appears to unblock you
with a minimum
> of work.
Thanks. Yep, the interface to the driver is unchanged. Do you have a preference
as to whether the driver, cc1+cc1as, or both adjust the triple?
> Please do, however, make this consistent for every port in the tree. Mips
is not enough of
> a special snowflake that it needs different ABI handling than any other
port. :)
Sure, I'll make sure ARM, Mips, and X86 work the same way and I'll have
another look through the others in case I've missed any that do the same
thing.

Thanks again.

________________________________
From: Eric Christopher [echristo at gmail.com]
Sent: 08 July 2016 18:23
To: Daniel Sanders; Rafael Espíndola (rafael.espindola at gmail.com; Matthew
Fortune; Renato Golin (renato.golin at linaro.org); Rui Ueyama (ruiu at
google.com); llvm-dev
Subject: Re: Representing MIPS ABI information in the triple as ARM/X86 do for
EABI/EABIHF/X32


 cc1as. The main difference is that X86 is adjusting the triple in the driver
while I'm doing it in cc1 and cc1as.


Rereading your email it looks like there was a bit of confusion.

As long as the interface to the driver is still --target mips-linux-gnu -mabi=XX
I care a bit less that you're mangling the abi information into the triple.
I still don't believe it's the correct long term direction for the
project, but it appears to unblock you with a minimum of work.

Please do, however, make this consistent for every port in the tree. Mips is not
enough of a special snowflake that it needs different ABI handling than any
other port. :)

Thanks.

-eric

At this point, I think I can safely say that the principle of what I'm doing
is sound and has solid precedent in other targets (particularly ARM and X86).
That is, tools and API users select a triple that specifies an ABI and the
backend is uses the triple as a source of ABI information. Can everyone confirm
whether we're agreed on that? Assuming we are agreed, I believe the main
details we need to agree on are which parts of clang adjust the triple and
whether the ABI is also passed via MCTargetOptions::ABIName if an
MCTargetOptions object is available. Are there any others?

For whether the ABI is also passed via MCTargetOptions::ABIName, I don't
mind either way in the end but I feel I need to pass through a step where we
don't pass the ABI name via MCTargetOptions::ABIName on my way to the end
point we choose. I feel that passing it both ways introduces potential
inconsistency between two sources of information and I'd like to stabilize
on a consistent state before re-introducing the possibility of inconsistency.
That said, it's not much harder to go straight to the end point if
that's what we prefer. Can everyone confirm which end point they prefer and
whether they want to go straight to it or through the intermediate step?

For which parts of clang adjust the triple, I think we have a choice between
doing it in the driver (like X86's current implementation), doing it in cc1
and cc1as (like my implementation), and doing both. I think that both isn't
really necessary and that cc1 and cc1as are not really intended for end-users
(much like llc/llvm-mc/etc), so I think the right place for the triple
adjustment is probably the driver. The reason I chose cc1 and cc1as in my
current patch series is because I had difficulty with the adjusted triple
getting into various paths and breaking the toolchain. My original plan was to
reach a working state with the cc1 and cc1as approach and then try to push the
adjustment into the driver during the spare time I'll have while fixing the
N32/N64 support in IAS. Can everyone confirm whether they think the driver,
cc1+cc1as, or both should do the triple adjustment? If you pick 'driver'
could you also confirm whether 'cc1+cc1as' is ok as an intermediate
step. If you pick 'both', could you also confirm which one should land
first or if both should land at the same time?

Once we've established a consensus, I'll adjust the patch series to
match it.

Thanks

From: llvm-dev [mailto:llvm-dev-bounces at
lists.llvm.org<mailto:llvm-dev-bounces at lists.llvm.org>] On Behalf Of
Daniel Sanders via llvm-dev
Sent: 05 July 2016 13:54
To: Eric Christopher; llvm-dev at lists.llvm.org<mailto:llvm-dev at
lists.llvm.org>
Subject: Re: [llvm-dev] Representing MIPS ABI information in the triple as
ARM/X86 do for EABI/EABIHF/X32

Hi Eric,

It's the unsolved problems on the pass-MCTargetOptions-everywhere path that
are my main concern with that approach rather than the amount of work. The first
problem is that the result of IRObjectFile::CollectAsmUndefinedRefs() depends on
the ABI but IRObjectFile doesn't know it. How would you deliver the ABI to
IRObjectFile? The second problem is that IRLinker will link incompatible IR
together because it doesn't know the ABI's involved in the link. It
thinks it's sufficient to check that the triples are the same. How would you
prevent it from linking IR intended for O32 with IR intended for N64?

I'm also worried about the risk of undiscovered problems on the
pass-MCTargetOptions-everywhere path but it's difficult to discuss those for
obvious reasons. However, I think it's safe to say that there could be some
major problems waiting to be discovered given that we don't have any targets
that implement a major ABI with this approach.
> There's already some ABI support in MCTargetOptions that I used when
fixing up the ARM port a while back and this use can either be extended or
changed depending on what's necessary
That's correct and r224492 is quite similar to what I'm trying to do in
this patch series (see below). I should mention that even that commit relies on
the ABI being encoded in the triple though. In computeTargetABI() from
ARMTargetMachine.cpp, MCTargetOptions::ABIName specifies the ABI if it is a
non-empty string but the code otherwise falls back on extracting the ABI from
the triple. The clang driver has similar code, it determines the value for the
–target-abi option from –mabi if it's given and falls back on calculating it
from the triple using the same code as computeTargetABI(). The overall effect is
that the ABI information is delivered to some parts of LLVM via both
MCTargetOptions::ABIName and the triple, and other parts solely via the triple
(e.g. ARMAsmPrinter and ARMISelLowering). Both copies of this information always
agree when –mabi is not given but they may differ when –mabi is given.

As I mentioned above, I'm trying to do a very similar thing to r224492 but I
have an additional problem to deal with. In my scenario, clang doesn't start
off with a triple that clearly specifies the ABI (mips64-linux-gnu can be N32 or
N64 depending on –mabi) which makes it impossible for the bits of LLVM that
can't see MCTargetOptions to know what to do.  I therefore start by
normalizing the triple in clang to one that does specify the ABI clearly and is
consistent with the –mabi option. With that problem solved, the rest of the
patch series is consistent with r224492 except that MCTargetOptions::ABIName and
the triple can never disagree. Personally, I would stop the patch series at this
point and not go on to duplicate the ABI information into
MCTargetOptions::ABIName (because this permits inconsistency and it doesn't
make sense to put N64 code inside an O32 ELF) but if you feel strongly about
this then I can add another patch that causes clang to pass the information in
both the triple and the –target-abi option and have the Mips backend accept
this. With that additional patch, this patch series will be completely
consistent with r224492.
> Even if it weren't an API problem I'm not sure it'd be a good
idea. I think we could add a with-abi configuration option to match the current
gcc one rather than tacking it onto triple parsing.
I thought you were strongly against configure-time options based on previous
discussions. Has this changed?
> ps. It also doesn't appear that Matthew (who took over my gcc
responsibilities, thanks!) has added support for this in gcc yet either.
> echristo at dzur ~/b/build-gcc-mips64abin64> .../configure
--target=mips64-linux-gnuabi64 --enable-languages=c
> echristo at dzur ~/tmp> ~/builds/build-gcc-mips64abin64/gcc/cc1 foo.c -o
-
>          .file      1 "foo.c"
>          .section .mdebug.abiN32
It's true that upstream GCC doesn't know about this triple yet but
that's because downstream users of GCC don't need to push their changes
upstream when they redefine triples. Fedora and Gentoo both use --with-abi while
Debian uses source patches. Upstream GCC only contains the default-defaults
which may be different from the actual-defaults that are compiled in to the
toolchain. In contrast, LLVM requires all the (often conflicting) meanings of a
triple to be upstream. This is part of a separate problem that I plan to address
in a later discussion so I'll leave this for another thread.

From: Eric Christopher [mailto:echristo at gmail.com<mailto:echristo at
gmail.com>]
Sent: 01 July 2016 23:05
To: Daniel Sanders; llvm-dev at lists.llvm.org<mailto:llvm-dev at
lists.llvm.org>
Cc: rafael.espindola at gmail.com<mailto:rafael.espindola at gmail.com>;
Matthew Fortune
Subject: Re: Representing MIPS ABI information in the triple as ARM/X86 do for
EABI/EABIHF/X32

Hi Daniel,

While I'm sympathetic to the amount of work associated this appears to be
attempting to work around an API problem by changing triples needlessly.
There's already some ABI support in MCTargetOptions that I used when fixing
up the ARM port a while back and this use can either be extended or changed
depending on what's necessary. It may require plumbing, but that's
expected with an API being expanded for new uses. It appears that the command
line syntax is sufficient for the needs of enumerating target plus abi and if we
need to add more defaults then it shouldn't be too bad to do so.

Even if it weren't an API problem I'm not sure it'd be a good idea.
I think we could add a with-abi configuration option to match the current gcc
one rather than tacking it onto triple parsing.

That said, I'm more than happy to help you detangle the existing API and
come up with a good strategy for the MC level API.

Thanks!

-eric

ps. It also doesn't appear that Matthew (who took over my gcc
responsibilities, thanks!) has added support for this in gcc yet either.

echristo at dzur ~/b/build-gcc-mips64abin64> .../configure
--target=mips64-linux-gnuabi64 --enable-languages=c
echristo at dzur ~/tmp> ~/builds/build-gcc-mips64abin64/gcc/cc1 foo.c -o -
          .file      1 "foo.c"
          .section .mdebug.abiN32
...

On Fri, Jun 24, 2016 at 3:02 AM Daniel Sanders <Daniel.Sanders at
imgtec.com<mailto:Daniel.Sanders at imgtec.com>> wrote:
Hi,

Having recently enabled IAS by default for the MIPS O32 ABI, I'm now trying
to do the same thing for the MIPS N64 ABI. Unfortunately, it is not currently
possible to enable IAS by default for the N64 ABI without also enabling it for
the N32 ABI because this information is not reflected in the triple and
that's the only information MipsMCAsmInfo has. This would be fine if it N32
was also in a good state but the current N32 ABI support for IAS is badly broken
and will likely take considerable effort to fix (and fixing it also requires
solving the same key problem as enabling IAS for just N64). I therefore want to
separate the two ABI's so that I can finish off N64 and then fix N32
afterwards.

I've posted a series of patches that allow us to distinguish N32 and N64 so
that we can enable IAS for only N64  (D21465, D21467, D21069 for LLVM along with
D21070, D21072 for clang). During the review of D21467, Rafael asked me to
explain my approach here. I'm hoping that people will agree that this is an
acceptable approach to take and that it's no different from how other
targets handle certain ABI's and how we already handle
-m32/-m64/-EL/-EB/etc.

The approach I'm using is to encode the ABI information into the triple in
the same way that ARM does for EABI and hardfloat, and the same way that X86
does for X32. These targets define variants of Triple::GNU for each of these
ABI's in the form of the Triple::GNUEABI, Triple::GNUEABIHF, and
Triple::GNUX32 values of the environment component of the triple and it's up
to the frontend and/or API-user to pass the right thing to the backend.

For MIPS, I'm defining Triple::GNUABI32, Triple::GNUABIN32, and
Triple::GNUABI64. All three of these are supported by tools like binutils (by
virtue of a wildcard match '*linux-gnu*') and of these three,
Triple::GNUABI32 is not in general use, Triple::GNUABIN32 will be used by Debian
if/when we do an N32 port, and Triple::GNUABI64 is currently used by Debian for
our N64 port. Once the ABI is in our backend triples, it can be used by
MipsMCAsmInfo to enable IAS by default for N64 without also enabling it for N32.
In later patches, such as D1292 this same information will also be used to fix
the N32 support for IAS.

At this point, you may be wondering what happens to Triple::GNU for triples like
mips-linux-gnu and mips64-linux-gnu. The answer to this is that the API-user
(e.g. clang) is expected to normalize such triples down to one of
Triple::GNUABI32/GNUABIN32/GNUABI64. This can be as simple as adding:
  Triple ABITriple = TT.getABIVariant(ABIName); // ABIName can be the empty
string to get the default ABI.
  if (ABITriple.getArch() != Triple::UnknownArch) {
    TT = ABITriple;
    ABIName = ""; // <- Only needed if this would end up in
MCTargetOptions::ABIName.
  }
to the appropriate place in the caller. This is the same way clang handles the
-m32, -m64, -EL, and -EB options with those options using get32BitArchVariant(),
get64BitArchVariant(), getLittleEndianArchVariant(), and
getBigEndianArchVariant() respectively to transform the triple for the backend.

In summary, human end users of tools such as clang will continue using the same
triples as they always have. These tools will then resolve them down to the
effective triple (for example, in functions like clang's
ComputeEffectiveClangTriple(), ComputeLLVMTriple(), and computeTargetTriple()).
The MIPS backend will only see the triples that have a specific known ABI.
I'd like to stress that this process is the same as the one we already use
today for other options. For example 'x86_64-linux-gnu-clang -m32'
transforms the triple with get32BitArchVariant() and passes
'i386-linux-gnu' to the backend.

So far I've only been talking about Linux/GNU but the patch series also
covers Linux/Musl, Linux/Android, and FreeBSD. I haven't talked about those
because they are handled in the same way as the Linux/GNU case described above
except that they use Triple::ABI32/ABIN32/ABI64 (or Android32/Android64 in the
case of Android) instead of Triple::GNUABI32/GNUABIN32/GNUABI64.

Why is the Triple approach better for MIPS than MCTargetOptions::ABIName?
========================================================================
A major pragmatic reason is that it works today and the MCTargetOption::ABIName
approach doesn't. It's also in keeping with other targets whereas until
fairly recently MIPS was the sole user of MCTargetOptions::ABIName. There is
currently a single PowerPC test and six ARM tests that use the -target-abi
option that feeds MCTargetOptions::ABIName at the moment and both of these
targets are only using it for relatively minor tweaks to the ABI.

As things are today, MCTargetOptions::ABIName is good at fairly small tweaks to
the ABI such as calling convention changes but it's unable to deal with more
fundamental changes such as selecting between ELFCLASS32/ELFCLASS64 or enabling
IAS. Some of these limitations look fixable if given months-years of effort but
I can see potential problems once the obvious plumbing problem is fixed. Some of
these problems such as preventing the linking of incompatible IR files in the
IRLinker would require me to represent the ABI in the LLVM-IR (it's
currently handled solely by the triple) which has been strongly opposed in
earlier discussions about triples.

I can provide more detail on some of the specific problems if required but this
boils down to a choice between an easy paved road that other people use and that
I know will end in a working state, or an unexplored path with known
difficulties and an uncertain result.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20160708/a3a0caef/attachment-0001.html>