[llvm-dev] [RFC] Propeller: A frame work for Post Link Optimizations

Wed Oct 16 15:52:18 PDT 2019

Hi Sri,

I want to clarify one thing before sending a detailed reply: did you evaluate
BOLT on Clang built with basic block sections? In the makefile you reference,
there are two versions: a “vanilla” and a default built with function sections.
High overheads you see with BOLT on Clang do not match our experience.

Thanks,
Maksim

On 10/14/19, 11:44 AM, "llvm-dev on behalf of Sriraman Tallam via llvm-dev" <llvm-dev-bounces at lists.llvm.org<mailto:llvm-dev-bounces at lists.llvm.org> on behalf of llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>> wrote:

Hello,

I wanted to consolidate all the discussions and our final thoughts on the concerns raised.  I have attached a document consolidating it.

BOLT’s performance gains inspired this work and we believe BOLT
is a great piece of engineering.  However, there are build environments where
scalability is critical and memory limits per process are tight :

* Debug Fission,  https://gcc.gnu.org/wiki/DebugFission was primarily
invented to achieve scalability and better incremental build times while
building large binaries with debug information.

* ThinLTO,
http://blog.llvm.org/2016/06/thinlto-scalable-and-incremental-lto.html<https://urldefense.proofpoint.com/v2/url?u=http-3A__blog.llvm.org_2016_06_thinlto-2Dscalable-2Dand-2Dincremental-2Dlto.html&d=DwMFaQ&c=5VD0RTtNlTh3ycd41b3MUw&r=4c9jZ8ZwYXlxUZHyw4Wing&m=BOTyGbKXpK1kdAvdQF0QoVsl4A5BCIQJMEEXJRVW6To&s=rW9yHyu5DPla9M38HolcW_w_Md8TLqe53BTWIClBxO4&e=> was
primarily invented to make LLVM’s full LTO scalable and keep the memory and
time overheads low.  ThinLTO has enabled much broader adoption of whole
program optimization, by making it non-monolithic.

* For Chromium builds,
https://chromium-review.googlesource.com/c/chromium/src/+/695714/3/build/toolcha<https://urldefense.proofpoint.com/v2/url?u=https-3A__chromium-2Dreview.googlesource.com_c_chromium_src_-2B_695714_3_build_toolcha&d=DwMFaQ&c=5VD0RTtNlTh3ycd41b3MUw&r=4c9jZ8ZwYXlxUZHyw4Wing&m=BOTyGbKXpK1kdAvdQF0QoVsl4A5BCIQJMEEXJRVW6To&s=8EBzmSqxfeVJXXXFKkx4Mzkf5d6cucxPc9pXkF36v_o&e=>
in/concurrent_links.gni, the linker process memory is set to 10GB with ThinLTO.
It was 26GB with Full LTO before that and individual processes will run of out
of memory beyond that.

* Here,
https://gotocon.com/dl/goto-chicago-2016/slides/AysyluGreenberg_BuildingADistrib<https://urldefense.proofpoint.com/v2/url?u=https-3A__gotocon.com_dl_goto-2Dchicago-2D2016_slides_AysyluGreenberg-5FBuildingADistrib&d=DwMFaQ&c=5VD0RTtNlTh3ycd41b3MUw&r=4c9jZ8ZwYXlxUZHyw4Wing&m=BOTyGbKXpK1kdAvdQF0QoVsl4A5BCIQJMEEXJRVW6To&s=lH-bp7s0QTtCyJkcSTOL8B_wOsRw-SLGrFsbZLmjaxQ&e=>
utedBuildSystemAtGoogleScale.pdf, a distributed build system at Google scale
is shown where 5 million binary and test builds are performed every day on
several thousands of machines, each  with a limitation of 12G of memory per
process and 15 minute time-out on tests. Memory overheads of 35G (clang) are
well above these thresholds.

We have developed Propeller like ThinLTO that can be used to obtain similar
performance gains like BOLT in such environments.

Thanks
Sri

On Fri, Oct 11, 2019 at 11:25 AM Xinliang David Li via llvm-dev <llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>> wrote:

On Fri, Oct 11, 2019 at 10:46 AM James Y Knight via llvm-dev <llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>> wrote:
Is there large value from deferring the block ordering to link time? That is, does the block layout algorithm need to consider global layout issues when deciding which blocks to put together and which to relegate to the far-away part of the code?

Or, could the propellor-optimized compile step instead split each function into only 2 pieces -- one containing an "optimally-ordered" set of hot blocks from the function, and another containing the cold blocks? The linker would have less flexibility in placement, but maybe it doesn't actually need that flexibility?

Apologies if this is obvious for those who actually know what they're talking about here. :)

It is a fair question.

We believe the flexibility to do fine grained layout in whole program context is important. PostLinkOptimization is aimed at getting as much performance improvement as possible (usually applied on top of ThinLTO+PGO), so the framework is designed to enable it.

In particular, it allows the linker to stitch hot bb traces from different functions to be stitched together. It also allows hot trace duplication across procedure boundaries (kind of interprocedural tailDup). Besides, code alignment decisions to minimize branch mispredictions  may require global context (e.g, too conflicting branches residing in two different functions).  Other micro-arch specific optimizations to improve processor front-end throughput may also require global context.

It is conceivable to have an option to control the level of granularity at the possible cost of performance.

thanks,

David

On Wed, Oct 2, 2019 at 6:18 PM Rafael Auler <rafaelauler at fb.com<mailto:rafaelauler at fb.com>> wrote:
You’re correct, except that, in Propeller, CFI duplication happens for every basic block as it operates with the conservative assumption that a block can be put anywhere by the linker. That’s a significant bloat that is not cleaned up later. So, during link time, if N blocks from the same function are contiguous in the final layout, as it should happen most of the time for any sane BB order, we would have several FDEs for a region that only needs one. The bloat goes to the final binary (a lot more FDEs, specifically, one FDE per basic block).

BOLT will only split a function in two parts, and only if it has profile. Most of the time, a function is not split. It also has an option not to split at all. For internally reordered basic blocks of a given function, it has CFI deduplication logic (it will interpret and build the CFI states for each block and rewrite the CFIs in a way that uses the minimum number of instructions to encode the states for each block).

From: llvm-dev <llvm-dev-bounces at lists.llvm.org<mailto:llvm-dev-bounces at lists.llvm.org>> on behalf of James Y Knight via llvm-dev <llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>>
Reply-To: James Y Knight <jyknight at google.com<mailto:jyknight at google.com>>
Date: Wednesday, October 2, 2019 at 1:59 PM
To: Maksim Panchenko <maks at fb.com<mailto:maks at fb.com>>
Cc: "llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>" <llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>>
Subject: Re: [llvm-dev] [RFC] Propeller: A frame work for Post Link Optimizations

I'm a bit confused by this subthread -- doesn't BOLT have the exact same CFI bloat issue? From my cursory reading of the propellor doc, the CFI duplication is _necessary_ to represent discontiguous functions, not anything particular to the way Propellor happens to generate those discontiguous functions.

And emitting discontiguous functions is a fundamental goal of this, right?

On Wed, Oct 2, 2019 at 4:25 PM Maksim Panchenko via llvm-dev <llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>> wrote:
Thanks for clarifying. This means once you move to the next basic block (or any other basic
block in the function) you have to execute an entirely new set of CFI instructions
except for the common CIE part. While indeed this is not as bad, on average, the overall
active memory footprint will increase.

Creating one FDE per basic block means that .eh_frame_hdr, an allocatable section,
will be bloated too. This will increase the FDE lookup time. I don’t see .eh_frame_hdr
being mentioned in the proposal.

Maksim

On 10/2/19, 12:20 PM, "Krzysztof Pszeniczny" <kpszeniczny at google.com<mailto:kpszeniczny at google.com>> wrote:

On Wed, Oct 2, 2019 at 8:41 PM Maksim Panchenko via llvm-dev <llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>> wrote:
*Pessimization/overhead for stack unwinding used by system-wide profilers and
for exception handling*

Larger CFI programs put an extra burden on unwinding at runtime as more CFI
(and thus native) instructions have to be executed. This will cause more
overhead for any profiler that records stack traces, and, as you correctly note
in the proposal, for any program that heavily uses exceptions.

The number of CFI instructions that have to be executed when unwinding any given stack stays the same. The CFI instructions for a function have to be duplicated in every basic block section, but when performing unwinding only one such a set is executed -- the copy for the current basic block. However, this copy contains precisely the same CFI instructions as the ones that would have to be executed if there were no basic block sections.

--
Krzysztof Pszeniczny
_______________________________________________
LLVM Developers mailing list
llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev<https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.llvm.org_cgi-2Dbin_mailman_listinfo_llvm-2Ddev&d=DwMFaQ&c=5VD0RTtNlTh3ycd41b3MUw&r=kx31RNFp5lAJejEYwuEQ4Zc5A6GakBit07EY08bIAvc&m=-AXqQmc2_r5LuTxyQRxmJESWGU7DLqvYjOlvwJnas_Q&s=h1mfecKZOhD5a1QaEabyI_nHKF81KAXoYRAgR0lNPvM&e=>
_______________________________________________
LLVM Developers mailing list
llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev<https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.llvm.org_cgi-2Dbin_mailman_listinfo_llvm-2Ddev&d=DwMFaQ&c=5VD0RTtNlTh3ycd41b3MUw&r=4c9jZ8ZwYXlxUZHyw4Wing&m=BOTyGbKXpK1kdAvdQF0QoVsl4A5BCIQJMEEXJRVW6To&s=nI5BAwXY8FSk9XDrnUJfuretr1UXOXiiqKNvmEDjDXo&e=>
_______________________________________________
LLVM Developers mailing list
llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev<https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.llvm.org_cgi-2Dbin_mailman_listinfo_llvm-2Ddev&d=DwMFaQ&c=5VD0RTtNlTh3ycd41b3MUw&r=4c9jZ8ZwYXlxUZHyw4Wing&m=BOTyGbKXpK1kdAvdQF0QoVsl4A5BCIQJMEEXJRVW6To&s=nI5BAwXY8FSk9XDrnUJfuretr1UXOXiiqKNvmEDjDXo&e=>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20191016/64fdafc0/attachment.html>