<div dir="ltr">Dear Community, <div><br></div><div>Please find summary of work done during this week as follow:</div><div><br></div><div>Implementation:</div><div>============</div><div><p style="margin:0px;line-height:normal;font-family:Helvetica">During this week we have identified a bug in IPRA due to not considering RegMask of function calls in given machine function. The same bug on AArch64 has been reported by Chad Rosier and more detailed description can be found at <a href="https://llvm.org/bugs/show_bug.cgi?id=28144">https://llvm.org/bugs/show_bug.cgi?id=28144</a> . To fix this bug RegMask calculation have been modified to consider RegMask of function call in a Machine Function. The patch is here <a href="http://reviews.llvm.org/D21395">http://reviews.llvm.org/D21395</a>. </p>
<p style="margin:0px;line-height:normal;font-family:Helvetica">AsmPrinter.cpp is modified to print call preserved registers in comments at call site in generated assembly file. This suggestion was by Quentin Colombet to improve readability of asm files while experimenting RegMask and calling convention etc. This simple patch can be found here <a href="http://reviews.llvm.org/D21490">http://reviews.llvm.org/D21490</a>.</p>
<p style="margin:0px;line-height:normal;font-family:Helvetica">We have also experimented a simple improvement to IPRA by setting callee saved registers to none for local function and we have found some performance improvement.</p></div><div><br></div><div>Testing:</div><div>======</div><div><p style="margin:0px;line-height:normal;font-family:Helvetica">After bug 28144 fix there is no runtime failures in test suite. Also due to bug 28144 there was about 60 run time failures and total time taken for test suite compilation was 30% more compare to with out IPRA. After bug fix with IPRA total compile time improvement compare to without IPRA is about 4 to 8 minutes.</p><p style="margin:0px;line-height:normal;font-family:Helvetica"><br></p><p style="margin:0px;line-height:normal;font-family:Helvetica">Study:</p><p style="margin:0px;line-height:normal;font-family:Helvetica">=====</p><p style="margin:0px;line-height:normal;font-family:Helvetica">This week I study code responsible for adding spill and restore for callee saved registers. Also studied how calling convention is defined in target specific .td files. I studied AsmPrinter.cpp and specifically emitComments() method which is responsible for adding comments in llvm generated assembly files. I also studied about some linkage type in LLMV IR like ‘internal’ which represent local function in module. </p><p style="margin:0px;line-height:normal;font-family:Helvetica"><br></p><p style="margin:0px;line-height:normal;font-family:Helvetica">Plan for next week:</p><p style="margin:0px;line-height:normal;font-family:Helvetica">1) Submit patch related to local function optimization for review </p><p style="margin:0px;line-height:normal;font-family:Helvetica">2) Find more possible improvements </p><p style="margin:0px;line-height:normal;font-family:Helvetica">3) Get active patches committed </p><p style="margin:0px;line-height:normal;font-family:Helvetica">4) Compile large software with IPRA enabled </p><p style="margin:0px;line-height:normal;font-family:Helvetica"><br></p><p style="margin:0px;line-height:normal;font-family:Helvetica">Sincerely,</p><p style="margin:0px;line-height:normal;font-family:Helvetica">Vivek</p></div><div class="gmail_extra"><br><div class="gmail_quote">On Wed, Jun 15, 2016 at 8:45 AM, vivek pandya <span dir="ltr"><<a href="mailto:vivekvpandya@gmail.com" target="_blank">vivekvpandya@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><br><div class="gmail_extra"><br><div class="gmail_quote"><span class="">On Wed, Jun 15, 2016 at 8:40 AM, vivek pandya <span dir="ltr"><<a href="mailto:vivekvpandya@gmail.com" target="_blank">vivekvpandya@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><br><div class="gmail_extra"><br><div class="gmail_quote"><span>On Wed, Jun 15, 2016 at 6:16 AM, Quentin Colombet <span dir="ltr"><<a href="mailto:qcolombet@apple.com" target="_blank">qcolombet@apple.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="auto"><div>Hi Vivek,</div><div><br></div><div>How much of the slow down on runtime comes from the different layout of the function in the asm file? (I.e., because of the dummy scc pass.)</div><div><br></div></div></blockquote></span><div>Hello Quentin, </div><div><br></div><div>Please do not consider previous results as there was a major bug in RegMask calculation due to not considering RegMasks of callee in MF body while calculating register usage information, that has been fixed now ( as discussed with Matthias Braun and Mehdi Amini ) and after this bugfix I have run test-suite with and without IPRA. Yes there is runtime slow down for some test cases ranging from 1% to 64% similarly compile time slow down is ranging from 1% to 48%. The runtime performance improvement is ranging from 1% to 35% and surprisingly there is also compile time improvement in a range from 1% to 60% . I would request you to go through complete results at <a href="https://docs.google.com/document/d/1cavn-POrZdhw-rrdPXV8mSvyppvOWs2rxmLgaOnd6KE/edit?usp=sharing" target="_blank">https://docs.google.com/document/d/1cavn-POrZdhw-rrdPXV8mSvyppvOWs2rxmLgaOnd6KE/edit?usp=sharing</a></div><div><br></div></div></div></div></blockquote></span><div>In above result baseline is IPRA and current is without IPRA. So actually data with background red is actual improvement and green is regression. </div><span class="HOEnZb"><font color="#888888"><div>-Vivek </div></font></span><div><div class="h5"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div class="gmail_extra"><div class="gmail_quote"><div></div><div>Also there is not extra failure due to IPRA now so in the result above I have removed failures.</div><div><br></div><div>Sincerely,</div><div>Vivek</div><div><div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="auto"><div></div><div>Cheers,</div><div>Q</div><div><div><div><br>Le 11 juin 2016 à 21:49, vivek pandya via llvm-dev <<a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a>> a écrit :<br><br></div><blockquote type="cite"><div><div dir="ltr">Dear Community,<div><br></div><div>The patch for Interprocedural Register Allocation has been committed now , thanks to Mehdi Amini for that. We would like you to play with it and let us know your views and more importantly ideas to improve it. </div><div><br></div><div>The test-suite run has indicated some non trivial issue that results in run time failure of the programs, we will be investigating it more. Here are some stats :</div><div><p style="margin:0px;line-height:normal;font-family:Helvetica">test-suite has been tested with IPRA enabled and overall results are not much encouraging. On average 30% increase in compile time. Many programs have also increase in execution time ( average 20%) that is really serious concern for us. About 60 tests have failed on run time this indicates error in compilation. how ever 3 tests have improvement in their runtime and that is 7% average.</p><p style="margin:0px;line-height:normal;font-family:Helvetica"><br></p></div><div>This week I think good thing for me to learn is to setup llvm development environment properly other wise one can end up wasting too much time building the llvm it self.</div><div><br></div><div>So here is brief summary:</div><div>Implementation:</div><div>============</div><div><p style="margin:0px;line-height:normal;font-family:Helvetica">The patch has been split into analysis and transformation passes. The pass responsible for register usage propagation has been made target independent. A print method and command line option -print-regusage has been added so that RegMaks details can be printed in Release builds also, this enables lit test case to be testable in Release build too. Other minor changes to adhere coding and naming conventions.</p><p style="margin:0px;line-height:normal;font-family:Helvetica"><br></p><p style="margin:0px;line-height:normal;font-family:Helvetica">Testing:</p><p style="margin:0px;line-height:normal;font-family:Helvetica">======</p><p style="margin:0px;line-height:normal;font-family:Helvetica">test-suite has been tested with IPRA enabled.</p><p style="margin:0px;line-height:normal;font-family:Helvetica"><br></p><p style="margin:0px;line-height:normal;font-family:Helvetica">Study and other:</p><p style="margin:0px;line-height:normal;font-family:Helvetica">=============</p><p style="margin:0px;line-height:normal;font-family:Helvetica">Learned about LNT, test-suite for LLVM, Inline assembly in LLVM IR, fastcc, local functions, MCStream class. In C++ I leaned about emplace family of methods in STL and perfect forwarding introduced in C++11.</p><p style="margin:0px;line-height:normal;font-family:Helvetica"><br></p><p style="margin:0px;line-height:normal;font-family:Helvetica">Plan for next week:</p><p style="margin:0px;line-height:normal;font-family:Helvetica">1) Investigate issue related to functional correctness that leads to run time failures</p><p style="margin:0px;line-height:normal;font-family:Helvetica">2) profile the compilation process to verify increase in time due to IPRA</p><p style="margin:0px;line-height:normal;font-family:Helvetica">3) Improve IPRA by instructing codegen to not save register for local function.</p><p style="margin:0px;line-height:normal;font-family:Helvetica">4) Make the pass emit asm comments to indicate register clobbered by function call at call site in generated ASM file.</p><p style="margin:0px;line-height:normal;font-family:Helvetica"><br></p><p style="margin:0px;line-height:normal;font-family:Helvetica">Sincerely,</p><p style="margin:0px;line-height:normal;font-family:Helvetica">Vivek</p></div></div><div class="gmail_extra"><br><div class="gmail_quote">On Sun, Jun 5, 2016 at 8:48 AM, vivek pandya <span dir="ltr"><<a href="mailto:vivekvpandya@gmail.com" target="_blank">vivekvpandya@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="ltr">Dear Community, <div><br></div><div>This week I got my patch reviewed by mentors and based on that I have done changes. Also we have identified a problem with callee-saved registers being marked as clobbered registers so we fixed that bug. I have described other minor changes in following section.</div><div><br></div><div>It was expected to get the patch committed by end of this week but due to unexpected mistake I was not able to complete writing test cases. Sorry for that.</div><div><span style="font-size:13px">I had build llvm with ipra enable by default and that build files were on my path ! Due to that next time I tried to build llvm it was terribly slow (almost 1 hour for 10% build ). I spend to much time on fixing this by playing around with environment variables, cmake options etc.</span></div><div style="font-size:13px">But I think this is a serious concern, we need to think verify this time complexity other wise building a large software with IPRA enable would be very time consuming.</div><div style="font-size:13px"><br></div><div style="font-size:13px">The toughest part for this week was to get lit and FileCheck work as you expect them to work, specially when analysis pass prints info on stdio and there is also a output file generated by llc or opt command. <br></div><div style="font-size:13px"><br></div><div style="font-size:13px">So here is brief summary :</div><div style="font-size:13px"><br></div><div style="font-size:13px">Implementation:</div><div style="font-size:13px">============</div><div><p style="margin:0px;line-height:normal;font-family:Helvetica"></p>RegUsageInfoCollector is now Calling Convention aware so that RegMask does not mark callee saved register as clobbered register. Due to this register allocator can use callee saved register for caller. <br>PhysicalRegisterUsageInfo.cpp renamed to RegisterUsageInfo.cpp.<br>StringMap used in RegisterUsageInfo.cpp is replaced by DenseMap of GlobalVariable * to RegMask.<br>DummyCGSCCPass moved from TargetPassConfig.cpp to CallGraphSCCPass.h.<br>Minor correction in comments, changes to adhere coding standards of LLVM.<br><p></p>
</div><div>Testing:</div><div>=====</div><div><p style="margin:0px;line-height:normal;font-family:Helvetica">The above mentioned changes has been tested with SNU-Realtime benchmarks.</p>
<p style="margin:0px;line-height:normal;font-family:Helvetica">Studied lit and FileCheck tool and written simple test to verify functionality of coding. </p><p style="margin:0px;line-height:normal;font-family:Helvetica"><br></p><p style="margin:0px;line-height:normal;font-family:Helvetica">Study and other:</p><p style="margin:0px;line-height:normal;font-family:Helvetica">============</p><p style="margin:0px;line-height:normal;font-family:Helvetica">Studied some examples of lit compatible llvm IR with comments to RUN test cases, FileCheck tool syntax and how to use it with in lit infrastructure. </p><p style="margin:0px;line-height:normal;font-family:Helvetica">I also understand X86 calling convention in more details. </p><p style="margin:0px;line-height:normal;font-family:Helvetica">I also studied basic concepts in llvm IR language while reading .ll files written for lit.</p><p style="margin:0px;line-height:normal;font-family:Helvetica">
</p><p style="margin:0px;line-height:normal;font-family:Helvetica">I learned about rvalue references and move semantics introduced in C++11.</p><p style="margin:0px;line-height:normal;font-family:Helvetica"><br></p><p style="margin:0px;line-height:normal;font-family:Helvetica">Plan for next week:</p><p style="margin:0px;line-height:normal;font-family:Helvetica">1) Get the patch committed along with proper tets cases.</p><p style="margin:0px;line-height:normal;font-family:Helvetica">2) Analyse time complexity of the approach.</p><p style="margin:0px;line-height:normal;font-family:Helvetica">3) Make target specific pass to CodeGen as it seems it is not required to be target specific.</p><p style="margin:0px;line-height:normal;font-family:Helvetica">4) If possible build a large application with ipra enable and analyze the impact.</p><p style="margin:0px;line-height:normal;font-family:Helvetica"><br></p><p style="margin:0px;line-height:normal;font-family:Helvetica">Sincerely,</p><p style="margin:0px;line-height:normal;font-family:Helvetica">Vivek</p></div><div><br></div></div><div><div><div class="gmail_extra"><br><div class="gmail_quote">On Sat, May 28, 2016 at 7:31 PM, vivek pandya <span dir="ltr"><<a href="mailto:vivekvpandya@gmail.com" target="_blank">vivekvpandya@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="ltr">Dear community,<div><br></div><div>This is to brief you the progress of Interprocedural Register Allocation, for those who are interested to see the progress in terms of code please consider <a href="http://reviews.llvm.org/D20769" target="_blank">http://reviews.llvm.org/D20769</a> </div><div>This patch contains simple infrastructure to propagate register usage information of callee to caller in call graph. The code generation order is changed to follow bottom up order on call graph , Thanks to Mehdi Amini for the patch for the same ! I will write a blog on this very soon. </div><div><br></div><div>So during this week as per the schedule proposed it should be study related infrastructure in LLVM and finalizing an approach for IPRA, but instead I am able to implement a working (may not be fully correct) prototype because I have used community bonding period to discuss and learn related stuffs from the mentors and also due to patch for CodeGen reordering was provided by dear mentor Mehdi Amini.</div><div><br></div><div>So I conclude the work done during this week as follows:</div><div><div>Implementation :</div><div>============</div><div>Following passes have been implemented during this week:
An immutable pass to store competed RegMask, a machine function pass that iterates through each registers and check if it is used or not and based on that details create a RegMask and a target specific machine function pass that uses the RegMask created by second pass and propagates information by updating call instructions RegMask. To update the RegMask of MI , setRegMask() function has been added to MachineOperand, a command line option -enable-ipra and debug type -debug-only=“ipra" has been added to control the optimization through llc.</div><div><br></div><div>Testing: </div><div>=====</div><div>The above mentioned implementation has been tested over SNU-Real-Time benchmark suit (<a href="http://www.cprover.org/goto-cc/examples/snu.html" target="_blank">http://www.cprover.org/goto-cc/examples/snu.html</a>) and some simple programs that uses library function ( for a library function register allocation is not done by LLVM so this optimization will simply skip them)<span style="white-space:pre-wrap"> </span></div><div><br></div><div>Study and Other:</div><div>=============</div><div>I have learned following things in LLVM, how it stores reg clobbering information? how it is used by Reg allocators through LivePhysRegs, LiveRegMatrix and other related passes? How to schedule a pass using TargetPassConfig and TargetMachine? What are called callee saved registers? What is an Immutable Pass?
Apart from that I have also learned how to use phabricator to send review request.
I have also read some related literatures.<span style="white-space:pre-wrap"> </span></div></div><div><br></div><div>During this week though task was to schedule the passes in proper order so that dependencies of related passes are satisfied.</div><div><br></div><div>Plan for next week:</div><div>1) Perform more testing and debug any known issue</div><div>2) Fine ture the implementation so as to eliminate any unnecessary work</div><div>3) During the testing from the stats I have observed that IPRA does not always improve the work of IntraProcedural register allocators and it is also observer that the amount of benefit (in terms of spilled live ranges ) is not deterministic. So I would like to find reasons for this behavior.</div><div>4) Start implementing target specific pass for other targets if review passes properly with no major bugs.</div><div><br></div><div>Please provide any feedback/suggestion including for format of this email. </div><div><br></div><div>I would also like to thanks my mentors Mehdi Amini , Hal Finkel, <span style="font-size:13px">Quentin Colombet, </span><span style="font-size:13px">Matthias Braun</span><span style="font-size:13px"> and other community members for providing quick help every time when I asked ( I have got replies even after 8 PM ( PDT) ! ) .</span></div><div><span style="font-size:13px"> </span></div><div>Sincerely,</div><div>Vivek</div></div>
</blockquote></div><br></div>
</div></div></blockquote></div><br></div>
</div></blockquote></div></div><blockquote type="cite"><div><span>_______________________________________________</span><br><span>LLVM Developers mailing list</span><br><span><a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a></span><br><span><a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev" target="_blank">http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev</a></span><br></div></blockquote></div></blockquote></div></div></div><br></div></div>
</blockquote></div></div></div><br></div></div>
</blockquote></div><br></div></div>