[llvm-dev] The state of IRPGO (3 remaining work items)

Xinliang David Li via llvm-dev llvm-dev at lists.llvm.org
Wed May 25 17:13:31 PDT 2016


On Wed, May 25, 2016 at 4:34 PM, Sean Silva <chisophugis at gmail.com> wrote:

>
>
> On Wed, May 25, 2016 at 2:17 PM, Xinliang David Li <davidxl at google.com>
> wrote:
>
>>
>>
>> On Wed, May 25, 2016 at 12:12 PM, Sean Silva <chisophugis at gmail.com>
>> wrote:
>>
>>>
>>>
>>> On Mon, May 23, 2016 at 8:23 PM, Sean Silva <chisophugis at gmail.com>
>>> wrote:
>>>
>>>> Jake and I have been integrating IRPGO on PS4, and we've identified 3
>>>> remaining work items.
>>>>
>>>>
>>>> - Driver changes
>>>>
>>>> We'd like to make IRPGO the default on PS4. We also think that it would
>>>> be beneficial to make IRPGO the default PGO on all platforms (coverage
>>>> would continue to use FE instr as it does currently, of course). In
>>>> previous conversations (e.g. http://reviews.llvm.org/D15829) it has
>>>> come up that Apple have requirements that would prevent them from moving to
>>>> IRPGO as the default PGO, at least without a deprecation period of one or
>>>> two releases.
>>>>
>>>> I'd like to get consensus on a path forward.
>>>> As a point of discussion, how about we make IRPGO the default on all
>>>> platforms except Apple platforms. I really don't like fragmenting things
>>>> like this (e.g. if a third-party tests "clang's" PGO they will get
>>>> something different depending on the platform), but I don't see another way
>>>> given Apple's constraints.
>>>>
>>>>
>>>> - Pre-instrumentation passes
>>>>
>>>> Pre-instrumentation optimization has been critical for reducing the
>>>> overhead of PGO for the PS4 games we tested (as expected). However, in our
>>>> measurements (and we are glad to provide more info) the main benefit was
>>>> inlining (also as expected). A simple pass of inlining at threshold 100
>>>> appeared to give all the benefits. Even inlining at threshold 0 gave almost
>>>> all the benefits. For example, the passes initially proposed in
>>>> http://reviews.llvm.org/D15828 did not improve over just inlining with
>>>> threshold 100.
>>>>
>>>> (due to PR27299 we also need to add simplifycfg after inlining to clean
>>>> up, but this doesn't affect the instrumentation overhead in our
>>>> measurements)
>>>>
>>>> Bottom line: for our use cases, inlining does all the work, but we're
>>>> not opposed to having more passes, which might be beneficial for non-game
>>>> workloads (which is most code).
>>>>
>>>>
>>>> - Warnings
>>>>
>>>> We identified 3 classes of issues which manifest as spammy warnings
>>>> when applying profile data with IRPGO (these affect FEPGO also I believe,
>>>> but we looked in depth at IRPGO):
>>>>
>>>> 1. The main concerning one is that getPGOFuncName mangles the filename
>>>> into the counter name. This causes us to get
>>>> instrprof_error::unknown_function when the pgo-use build is done in a
>>>> different build directory from the training build (which is a reasonable
>>>> thing to support). In this situation, PGO data is useless for all `static`
>>>> functions (and as a byproduct results in a huge volume of warnings).
>>>>
>>>
>>>
>>> Rong, I was just looking at implementing a fix for this, but noticed
>>> something. Can we get rid of the "InLTO" argument to getPGOFuncName if we
>>> unconditionally apply the funcname metadata to all functions?
>>>
>>
>> Are you proposing always emitting the meta data for internal functions?
>>
>
> Yes; actually for all functions. If I understand correctly, the InLTO
> argument fixes a specific case of a general problem: the PGOFuncName is
> currently constructed via an algorithm that depends on the current state of
> the module and the result can therefore change as the module is transformed.
> To solve this, the idea is to only run the algorithm at a single point of
> truth (the module as seen by prof-gen/prof-use) and freeze the result in
> metadata. All other accesses to PGOFuncName simply read that metadata, so
> there is no possibility for getting out of date.
>


>
> Currently, it seems the meaning of the InLTO flag is basically "we are
> accessing the PGOFuncName at a point that is not prof-gen/prof-use, so use
> the metadata". I just want to make that clearer.
>

This problem is kind of special to value profiling in LTO mode. In this
mode, the transformation needs to be delayed (when all function bodies are
available) so we end up having compute the pgonames at two different points
with different program states.  Your suggestion of doing this
unconditionally will work but will require more memory at compile time.

David



>
> -- Sean Silva
>
>
>>
>> David
>>
>>
>>
>>>
>>> -- Sean Silva
>>>
>>>
>>>>
>>>> 2. In different TU's, pre-instr inlining might make different inlining
>>>> decisions (for example, different functions may be available for inlining),
>>>> causing hash mismatch errors (instrprof_error::hash_mismatch). In building
>>>> a large game, we only saw 8 instance of this, so it is not as severe as 1,
>>>> but would be good to fix.
>>>>
>>>> 3. A .cpp file may be compiled and put into an archive, but then not
>>>> selected by the linker and will therefore not result in a counter in the
>>>> profraw. When compiling this file with pgo-use,
>>>> instrprof_error::unknown_function will result and a warning will be emitted.
>>>>
>>>> Case 1 can be fixed using a function hash or other unique identifier
>>>> instead of a file path. David, in D20195 you mentioned that Rong was
>>>> working on a patch that would fix 2; we are looking forward to that.
>>>>
>>>> For 3, I unfortunately do not know of any solution. I don't think there
>>>> is a way for us to make this warning reliable in the face of this
>>>> circumstance. So my conclusion is that instrprof_error::unknown_function at
>>>> least must be defaulted to off unfortunately.
>>>>
>>>> -- Sean Silva
>>>>
>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160525/28a89f3e/attachment.html>


More information about the llvm-dev mailing list