<div dir="ltr"><div class="gmail_default" style="font-family:arial,helvetica,sans-serif;font-size:small">Hi Jeroen,</div><div class="gmail_default" style="font-family:arial,helvetica,sans-serif;font-size:small"><br></div><div class="gmail_default" style="font-size:small"><span style="font-family:arial,helvetica,sans-serif">Does that mean in this case, even though the frontend correctly interprets </span><font face="monospace">__restrict</font><span style="font-family:arial,helvetica,sans-serif"> on struct fields, the backend is only considering </span><font face="monospace">nolias</font><font face="arial, helvetica, sans-serif"> metadata on function arguments?</font></div><div class="gmail_default" style="font-size:small"><font face="arial, helvetica, sans-serif"><br></font></div><div class="gmail_default" style="font-size:small"><font face="arial, helvetica, sans-serif">Thanks,</font></div><div class="gmail_default" style="font-size:small"><font face="arial, helvetica, sans-serif">Bandhav</font></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Wed, Jun 24, 2020 at 1:32 AM Jeroen Dobbelaere <<a href="mailto:Jeroen.Dobbelaere@synopsys.com">Jeroen.Dobbelaere@synopsys.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<div lang="EN-US">
<div class="gmail-m_1939310555962317815WordSection1">
<p class="MsoNormal">Hi Bandhav,<u></u><u></u></p>
<p class="MsoNormal"><u></u> <u></u></p>
<p class="MsoNormal">I did notice in the previous example, that the vectorizer used the noalias information, but that it also got stripped during the vectorization.<u></u><u></u></p>
<p class="MsoNormal">That is certainly one of the places where the noalias handling can be improved.<u></u><u></u></p>
<p class="MsoNormal"><u></u> <u></u></p>
<p class="MsoNormal">It would be interesting to see if the necessary information gets through to the MIR level when vectorization is disabled.<u></u><u></u></p>
<p class="MsoNormal">That should be visible with the 'ptr_provenance' field.<u></u><u></u></p>
<p class="MsoNormal"><u></u> <u></u></p>
<p class="MsoNormal">Greetings,<u></u><u></u></p>
<p class="MsoNormal"><u></u> <u></u></p>
<p class="MsoNormal">Jeroen Dobbelaere<u></u><u></u></p>
<p class="MsoNormal"><u></u> <u></u></p>
<p class="MsoNormal"><u></u> <u></u></p>
<p class="MsoNormal"><u></u><u></u></p>
<div style="border-top:none;border-right:none;border-bottom:none;border-left:1.5pt solid blue;padding:0in 0in 0in 4pt">
<div>
<div style="border-right:none;border-bottom:none;border-left:none;border-top:1pt solid rgb(225,225,225);padding:3pt 0in 0in">
<p class="MsoNormal"><b>From:</b> Bandhav Veluri <<a href="mailto:bandhav.veluri00@gmail.com" target="_blank">bandhav.veluri00@gmail.com</a>> <br>
<b>Sent:</b> Wednesday, June 24, 2020 10:12<br>
<b>To:</b> Jeroen Dobbelaere <<a href="mailto:dobbel@synopsys.com" target="_blank">dobbel@synopsys.com</a>><br>
<b>Cc:</b> Neil Henning <<a href="mailto:neil.henning@unity3d.com" target="_blank">neil.henning@unity3d.com</a>>; <a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a>; Kruse, Michael <<a href="mailto:michael.kruse@anl.gov" target="_blank">michael.kruse@anl.gov</a>><br>
<b>Subject:</b> Re: FW: [llvm-dev] Restrict qualifier on class members<u></u><u></u></p>
</div>
</div>
<p class="MsoNormal"><u></u> <u></u></p>
<div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif">Hi Jeroen,<u></u><u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif"><u></u> <u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif">Sorry, I missed that. I tried the patch, and this program:<u></u><u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt"><br>
</span><span style="font-size:12pt;font-family:"Courier New"">#include <stdint.h><br>
<br>
#define __remote __attribute__((address_space(1)))<br>
<br>
__remote int* A;<br>
__remote int* B;<br>
<br>
void vec_add(__remote int* __restrict a,<br>
__remote int* __restrict b,<br>
int n) {<br>
#pragma unroll 4<br>
for(int i=0; i<n; ++i) {<br>
a[i] += b[i];<br>
}<br>
}<br>
<br>
int main(int argc, char** argv) {<br>
__remote int* __restrict a = A;<br>
__remote int* __restrict b = B;<br>
<br>
#pragma unroll 4<br>
for(int i=0; i<4; ++i) {<br>
a[i] += b[i];<br>
}<br>
<br>
return 0;<br>
}</span><span style="font-size:12pt"><u></u><u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt"><u></u> <u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:"Courier New"">vec_add</span><span style="font-size:12pt"> give following schedule:<u></u><u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt"><u></u> <u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:"Courier New"">*** Final schedule for %bb.8 ***<br>
SU(0): %33:gpr = LW %56:gpr, -8 :: (load 4 from %ir.scevgep8, !tbaa !14, !noalias !13, addrspace 1)<br>
SU(1): %34:gpr = LW %55:gpr, -8 :: (load 4 from %ir.scevgep14, !tbaa !14, !noalias !13, addrspace 1)<br>
SU(4): %36:gpr = LW %56:gpr, -4 :: (load 4 from %ir.scevgep10, !tbaa !14, !noalias !13, addrspace 1)<br>
SU(5): %37:gpr = LW %55:gpr, -4 :: (load 4 from %ir.scevgep16, !tbaa !14, !noalias !13, addrspace 1)<br>
SU(8): %39:gpr = LW %56:gpr, 0 :: (load 4 from %ir.lsr.iv6, !tbaa !14, !noalias !13, addrspace 1)<br>
SU(9): %40:gpr = LW %55:gpr, 0 :: (load 4 from %ir.lsr.iv12, !tbaa !14, !noalias !13, addrspace 1)<br>
SU(12): %42:gpr = LW %56:gpr, 4 :: (load 4 from %ir.scevgep9, !tbaa !14, !noalias !13, addrspace 1)<br>
SU(13): %43:gpr = LW %55:gpr, 4 :: (load 4 from %ir.scevgep15, !tbaa !14, !noalias !13, addrspace 1)<br>
SU(2): %35:gpr = nsw ADD %34:gpr, %33:gpr<br>
SU(3): SW %35:gpr, %55:gpr, -8 :: (store 4 into %ir.scevgep14, !tbaa !14, !noalias !13, addrspace 1)<br>
SU(6): %38:gpr = nsw ADD %37:gpr, %36:gpr<br>
SU(7): SW %38:gpr, %55:gpr, -4 :: (store 4 into %ir.scevgep16, !tbaa !14, !noalias !13, addrspace 1)<br>
SU(10): %41:gpr = nsw ADD %40:gpr, %39:gpr<br>
SU(11): SW %41:gpr, %55:gpr, 0 :: (store 4 into %ir.lsr.iv12, !tbaa !14, !noalias !13, addrspace 1)<br>
SU(14): %44:gpr = nsw ADD %43:gpr, %42:gpr<br>
SU(15): SW %44:gpr, %55:gpr, 4 :: (store 4 into %ir.scevgep15, !tbaa !14, !noalias !13, addrspace 1)<br>
SU(16): %57:gpr = nuw nsw ADDI %57:gpr, 4<br>
SU(17): %56:gpr = ADDI %56:gpr, 16<br>
SU(18): %55:gpr = ADDI %55:gpr, 16</span><span style="font-size:12pt"><u></u><u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt"><u></u> <u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif">And
</span><span style="font-size:12pt;font-family:"Courier New"">main </span><span style="font-size:12pt;font-family:Arial,sans-serif">gives following schedule:</span><span style="font-size:12pt"><u></u><u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt"><u></u> <u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:"Courier New""> *** Final schedule for %bb.0 ***<br>
SU(0): %2:gpr = LUI target-flags(riscv-hi) @A<br>
SU(2): %4:gpr = LUI target-flags(riscv-hi) @B<br>
SU(3): %5:gpr = LW %4:gpr, target-flags(riscv-lo) @B :: (dereferenceable load 4 from @B, !tbaa !9, !noalias !22)<br>
SU(1): %3:gpr = LW %2:gpr, target-flags(riscv-lo) @A :: (dereferenceable load 4 from @A, !tbaa !9, !noalias !22)<br>
SU(4): %6:gpr = LW %5:gpr, 0 :: (load 4 from %ir.3, !tbaa !14, !noalias !22, addrspace 1)<br>
SU(5): %7:gpr = LW %3:gpr, 0 :: (load 4 from %ir.1, !tbaa !14, !noalias !22, addrspace 1)<br>
SU(6): %8:gpr = nsw ADD %7:gpr, %6:gpr<br>
SU(7): SW %8:gpr, %3:gpr, 0 :: (store 4 into %ir.1, !tbaa !14, !noalias !22, addrspace 1)<br>
SU(8): %9:gpr = LW %5:gpr, 4 :: (load 4 from %ir.arrayidx.1, !tbaa !14, !noalias !22, addrspace 1)<br>
SU(9): %10:gpr = LW %3:gpr, 4 :: (load 4 from %ir.arrayidx1.1, !tbaa !14, !noalias !22, addrspace 1)<br>
SU(10): %11:gpr = nsw ADD %10:gpr, %9:gpr<br>
SU(11): SW %11:gpr, %3:gpr, 4 :: (store 4 into %ir.arrayidx1.1, !tbaa !14, !noalias !22, addrspace 1)<br>
SU(12): %12:gpr = LW %5:gpr, 8 :: (load 4 from %ir.arrayidx.2, !tbaa !14, !noalias !22, addrspace 1)<br>
SU(13): %13:gpr = LW %3:gpr, 8 :: (load 4 from %ir.arrayidx1.2, !tbaa !14, !noalias !22, addrspace 1)<br>
SU(14): %14:gpr = nsw ADD %13:gpr, %12:gpr<br>
SU(15): SW %14:gpr, %3:gpr, 8 :: (store 4 into %ir.arrayidx1.2, !tbaa !14, !noalias !22, addrspace 1)<br>
SU(16): %15:gpr = LW %5:gpr, 12 :: (load 4 from %ir.arrayidx.3, !tbaa !14, !noalias !22, addrspace 1)<br>
SU(17): %16:gpr = LW %3:gpr, 12 :: (load 4 from %ir.arrayidx1.3, !tbaa !14, !noalias !22, addrspace 1)<br>
SU(18): %17:gpr = nsw ADD %16:gpr, %15:gpr<br>
SU(20): $x10 = COPY $x0<br>
SU(19): SW %17:gpr, %3:gpr, 12 :: (store 4 into %ir.arrayidx1.3, !tbaa !14, !noalias !22, addrspace 1)</span><span style="font-size:12pt"><u></u><u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif"><u></u> <u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif">This is great! Memory accesses are marked
</span><span style="font-size:12pt;font-family:"Courier New"">noalias. </span><span style="font-size:12pt;font-family:Arial,sans-serif">I wanted memory accesses to be annotated as noalias to basically remove loop-carried dependencies so that I can reorder
them for efficient scheduling. But when I look at Schedule DAG,</span><span style="font-size:12pt"><u></u><u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt"><u></u> <u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif">For
</span><span style="font-size:12pt;font-family:"Courier New"">vec_add</span><span style="font-size:12pt;font-family:Arial,sans-serif"> I see something like this (note
</span><span style="font-size:12pt;font-family:"Courier New"">BotQ.A</span><span style="font-size:12pt;font-family:Arial,sans-serif">, scheduler can choose any of those => no loop carried dependence):</span><span style="font-size:12pt"><u></u><u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:"Courier New""> - Latency limited.<br>
** ScheduleDAGMILive::schedule picking next node<br>
Queue BotQ.P:<br>
Queue BotQ.A: 16 15 11 7 3<br>
Cand SU(16) ORDER<br>
Pick Bot ORDER</span><span style="font-size:12pt"><u></u><u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif"><u></u> <u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif">For
</span><span style="font-size:12pt;font-family:"Courier New"">main</span><span style="font-size:12pt;font-family:Arial,sans-serif">, at best I see something like this:</span><span style="font-size:12pt"><u></u><u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:"Courier New"">** ScheduleDAGMILive::schedule picking next node<br>
Cycle: 45 BotQ.A<br>
Queue BotQ.P:<br>
Queue BotQ.A: 12 13<br>
Cand SU(12) ORDER<br>
Cand SU(13) ORDER</span><span style="font-size:12pt"><u></u><u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt"><u></u> <u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif">In theory, schedules for vec_add and main should be the same right? Is there anything else I should do to make the
</span><span style="font-size:12pt;font-family:"Courier New"">__restrict</span><span style="font-size:12pt;font-family:Arial,sans-serif"> remove loop-carried dependence in
</span><span style="font-size:12pt;font-family:"Courier New"">main</span><span style="font-size:12pt;font-family:Arial,sans-serif">?</span><span style="font-size:12pt"><u></u><u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt"><u></u> <u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif">Attaching IR and scheduler log for reference...</span><span style="font-size:12pt"><u></u><u></u></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif"><u></u> <u></u></span></p>
</div>
<p class="MsoNormal"><u></u> <u></u></p>
<div>
<div>
<p class="MsoNormal">On Mon, Jun 22, 2020 at 3:03 PM Jeroen Dobbelaere <<a href="mailto:Jeroen.Dobbelaere@synopsys.com" target="_blank">Jeroen.Dobbelaere@synopsys.com</a>> wrote:<u></u><u></u></p>
</div>
<blockquote style="border-top:none;border-right:none;border-bottom:none;border-left:1pt solid rgb(204,204,204);padding:0in 0in 0in 6pt;margin-left:4.8pt;margin-right:0in">
<div>
<div>
<p class="MsoNormal">Hi Bandhav,<u></u><u></u></p>
<p class="MsoNormal"> <u></u><u></u></p>
<p class="MsoNormal">as mentioned in the summary of
<a href="https://urldefense.com/v3/__https:/reviews.llvm.org/D69542__;!!A4F2R9G_pg!OmBKB80qBacnrqnvVVCAWFkwISI7m6FZ1z4B8XOkDZKUndeom24G2GmH_KktbhCAqzAnX4K4$" target="_blank">
https://reviews.llvm.org/D69542</a> :<u></u><u></u></p>
<p class="MsoNormal"> <u></u><u></u></p>
<p class="MsoNormal"> The base version is
<a href="https://urldefense.com/v3/__https:/reviews.llvm.org/rGb2a37cfe2bda0bc8c4d2e981922b5ac59c429bdc__;!!A4F2R9G_pg!OmBKB80qBacnrqnvVVCAWFkwISI7m6FZ1z4B8XOkDZKUndeom24G2GmH_KktbhCAq3tq-EY3$" target="_blank">
b2a37cfe2bda0bc8c4d2e981922b5ac59c429bdc</a> (June 12, 2020)<u></u><u></u></p>
<p class="MsoNormal"> <u></u><u></u></p>
<p class="MsoNormal">Greetings,<u></u><u></u></p>
<p class="MsoNormal"> <u></u><u></u></p>
<p class="MsoNormal">Jeroen Dobbelaere<u></u><u></u></p>
<p class="MsoNormal"> <u></u><u></u></p>
<p class="MsoNormal"> <u></u><u></u></p>
<p class="MsoNormal"> <u></u><u></u></p>
<p class="MsoNormal"><b>From:</b> llvm-dev <<a href="mailto:llvm-dev-bounces@lists.llvm.org" target="_blank">llvm-dev-bounces@lists.llvm.org</a>>
<b>On Behalf Of </b>Bandhav Veluri via llvm-dev<br>
<b>Sent:</b> Monday, June 22, 2020 18:32<br>
<b>To:</b> Neil Henning <<a href="mailto:neil.henning@unity3d.com" target="_blank">neil.henning@unity3d.com</a>><br>
<b>Cc:</b> llvm-dev <<a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a>>; Kruse, Michael <<a href="mailto:michael.kruse@anl.gov" target="_blank">michael.kruse@anl.gov</a>><br>
<b>Subject:</b> Re: [llvm-dev] Restrict qualifier on class members<u></u><u></u></p>
<p class="MsoNormal"> <u></u><u></u></p>
<div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif">Hi Jeroen,</span><u></u><u></u></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif"> </span><u></u><u></u></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif">That's great! I was trying to use the patch, what's the latest version of the project we could apply it on?</span><u></u><u></u></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif"> </span><u></u><u></u></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif">Hi Neil,</span><u></u><u></u></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif"> </span><u></u><u></u></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif">That seems like what I can do as well! Do you happen to have some examples lying around? Maybe a pointer to the planned
presentation, if that's okay?</span><u></u><u></u></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif"> </span><u></u><u></u></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif">Thank you,</span><u></u><u></u></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:12pt;font-family:Arial,sans-serif">Bandhav</span><u></u><u></u></p>
</div>
</div>
<p class="MsoNormal"> <u></u><u></u></p>
<div>
<div>
<p class="MsoNormal">On Mon, Jun 22, 2020 at 1:55 AM Neil Henning <<a href="mailto:neil.henning@unity3d.com" target="_blank">neil.henning@unity3d.com</a>> wrote:<u></u><u></u></p>
</div>
<blockquote style="border-top:none;border-right:none;border-bottom:none;border-left:1pt solid rgb(204,204,204);padding:0in 0in 0in 6pt;margin:5pt 0in 5pt 4.8pt">
<div>
<div>
<p class="MsoNormal">I was originally going to cover this in my now defunct EuroLLVM talk but... we had this exact same problem on Unity's HPC# Burst compiler - how to track no-aliasing on structs.
We were constrained in that we had to make it work with LLVM versions all the way back to shipped LLVM 6, so what we did was:<u></u><u></u></p>
</div>
<div>
<ul type="disc">
<li class="MsoNormal">
Add module-level metadata that tracked whether a given struct member field was no-alias.<u></u><u></u></li><li class="MsoNormal">
Added our own alias analysis using createExternalAAWrapperPass to register it in the pass pipeline.<u></u><u></u></li></ul>
<div>
<p class="MsoNormal">This allowed us to have zero modifications to LLVM and do something useful with aliasing. The one 'issue' with it is if you have a stack-allocated struct that is SROA'ed you will
lose the info that it was a struct, or if you are in a private/internal linkage function that has the struct as an argument, the opt passes can modify the function signature to lose the struct too. We had to do some mitigations here to get perfect aliasing
on our usecases.<u></u><u></u></p>
</div>
<div>
<p class="MsoNormal"> <u></u><u></u></p>
</div>
<div>
<p class="MsoNormal">Hope this helps,<u></u><u></u></p>
</div>
<div>
<p class="MsoNormal">-Neil.<u></u><u></u></p>
</div>
</div>
</div>
<p class="MsoNormal"> <u></u><u></u></p>
<div>
<div>
<p class="MsoNormal">On Mon, Jun 22, 2020 at 5:44 AM Michael Kruse via llvm-dev <<a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a>> wrote:<u></u><u></u></p>
</div>
<blockquote style="border-top:none;border-right:none;border-bottom:none;border-left:1pt solid rgb(204,204,204);padding:0in 0in 0in 6pt;margin:5pt 0in 5pt 4.8pt">
<p class="MsoNormal">Unfortunately
<a href="https://urldefense.com/v3/__https:/llvm.org/docs/LangRef.html*llvm-loop-parallel-accesses-metadata__;Iw!!A4F2R9G_pg!KbQb7EC98K_vFVBRDQoSveXuMcvOvKcWvTers1QW_g1LmGPsOh-wI0r2mV_c2G1yLJxlIVzl$" target="_blank">
https://llvm.org/docs/LangRef.html#llvm-loop-parallel-accesses-metadata</a><br>
is not a solution here. A loop-parallel access does not imply<br>
non-aliasing. The obvious case is when only reading from a location,<br>
but even when a location is written to I'd be careful to deduce that<br>
they do not alias since it might be a "benign data race" or the value<br>
never used. Additionally, LLVM's loop unroller is known to now handle<br>
noalias metadata correctly as it just copies it.<br>
<br>
There has been a discussion here:<br>
<a href="https://urldefense.com/v3/__http:/lists.llvm.org/pipermail/llvm-dev/2020-May/141587.html__;!!A4F2R9G_pg!KbQb7EC98K_vFVBRDQoSveXuMcvOvKcWvTers1QW_g1LmGPsOh-wI0r2mV_c2G1yLAvIBb7x$" target="_blank">http://lists.llvm.org/pipermail/llvm-dev/2020-May/141587.html</a><br>
<br>
Michael<br>
<br>
<br>
Am So., 21. Juni 2020 um 12:24 Uhr schrieb Johannes Doerfert via<br>
llvm-dev <<a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a>>:<br>
><br>
> Hi Bandhav,<br>
><br>
><br>
> Jeroen Dobbelaere (CC'ed) is currently working on support for restrict qualified local variables and struct members.<br>
><br>
> The patches exist but are not merged yet. If you want to give it a try apply <a href="https://urldefense.com/v3/__https:/reviews.llvm.org/D69542__;!!A4F2R9G_pg!KbQb7EC98K_vFVBRDQoSveXuMcvOvKcWvTers1QW_g1LmGPsOh-wI0r2mV_c2G1yLLzXOdAP$" target="_blank">
https://reviews.llvm.org/D69542</a>.<br>
><br>
><br>
> Initially I could only think of this solution for your problem: <a href="https://urldefense.com/v3/__https:/godbolt.org/z/6WtPXJ__;!!A4F2R9G_pg!KbQb7EC98K_vFVBRDQoSveXuMcvOvKcWvTers1QW_g1LmGPsOh-wI0r2mV_c2G1yLEju0wQ6$" target="_blank">
https://godbolt.org/z/6WtPXJ</a><br>
><br>
> Michael (CC'ed) might now another annotation to get `llvm.access` metadata for the loop, which should do what you intend.<br>
><br>
><br>
> Cheers,<br>
><br>
> Johannes<br>
><br>
><br>
> On 6/21/20 11:56 AM, Bandhav Veluri via llvm-dev wrote:<br>
><br>
> Hi,<br>
><br>
> I'm trying to abstract some special pointers with a class, like in the<br>
> example program below:<br>
><br>
> 1 #define __remote __attribute__((address_space(1)))<br>
> 2 #include <stdint.h><br>
> 3<br>
> 4 __remote int* A;<br>
> 5 __remote int* B;<br>
> 6<br>
> 7 class RemotePtr {<br>
> 8 private:<br>
> 9 __remote int* __restrict a;<br>
> 10<br>
> 11 public:<br>
> 12 RemotePtr(__remote int* a) : a(a) {}<br>
> 13<br>
> 14 __remote int& at(int n) {<br>
> 15 return a[n];<br>
> 16 }<br>
> 17 };<br>
> 18<br>
> 19 int main(int argc, char** argv) {<br>
> 20 RemotePtr a(A);<br>
> 21 RemotePtr b(B);<br>
> 22<br>
> 23 #pragma unroll 4<br>
> 24 for(int i=0; i<4; ++i) {<br>
> 25 <a href="https://urldefense.com/v3/__http:/a.at__;!!A4F2R9G_pg!KbQb7EC98K_vFVBRDQoSveXuMcvOvKcWvTers1QW_g1LmGPsOh-wI0r2mV_c2G1yLOY-9qOq$" target="_blank">a.at</a>(i) +=
<a href="https://urldefense.com/v3/__http:/b.at__;!!A4F2R9G_pg!KbQb7EC98K_vFVBRDQoSveXuMcvOvKcWvTers1QW_g1LmGPsOh-wI0r2mV_c2G1yLBIWnTL5$" target="_blank">
b.at</a>(i);<br>
> 26 }<br>
> 27<br>
> 28 return 0;<br>
> 29 }<br>
><br>
> It's given that pointer a, in each object of the class RemotePtr, is the<br>
> only pointer that can access the array pointed by it. So, I tried __remote<br>
> int* __restrict a; (line 9) construct to tell Clang the same. This doesn't<br>
> seem to work and I see no noliass in the generated IR. Specifically, I want<br>
> lines 23-26 optimized assuming no aliasing between A and B. Any reason why<br>
> Clang shouldn't annotate memory accesses in lines 23-26 with noaliass<br>
> taking line 9 into account?<br>
><br>
> The higher level problem is this: is there a way to compile lines 23-26<br>
> assuming no aliasing between A and B, by just doing something in the<br>
> RemotePtr class (so that main is clear of ugly code)? If that's not<br>
> possible, is there a way to tell Clang that lines 23-26 should assume no<br>
> aliasing at all, by some pragma?<br>
><br>
> Thank you,<br>
> Bandhav<br>
><br>
><br>
> _______________________________________________<br>
> LLVM Developers mailing list<br>
> <a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a><br>
> <a href="https://urldefense.com/v3/__https:/lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev__;!!A4F2R9G_pg!KbQb7EC98K_vFVBRDQoSveXuMcvOvKcWvTers1QW_g1LmGPsOh-wI0r2mV_c2G1yLD9-lkj8$" target="_blank">
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev</a><br>
><br>
> _______________________________________________<br>
> LLVM Developers mailing list<br>
> <a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a><br>
> <a href="https://urldefense.com/v3/__https:/lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev__;!!A4F2R9G_pg!KbQb7EC98K_vFVBRDQoSveXuMcvOvKcWvTers1QW_g1LmGPsOh-wI0r2mV_c2G1yLD9-lkj8$" target="_blank">
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev</a><br>
_______________________________________________<br>
LLVM Developers mailing list<br>
<a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a><br>
<a href="https://urldefense.com/v3/__https:/lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev__;!!A4F2R9G_pg!KbQb7EC98K_vFVBRDQoSveXuMcvOvKcWvTers1QW_g1LmGPsOh-wI0r2mV_c2G1yLD9-lkj8$" target="_blank">https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev</a><u></u><u></u></p>
</blockquote>
</div>
<p class="MsoNormal"><br clear="all">
<br>
-- <u></u><u></u></p>
<div>
<div>
<table border="0" cellspacing="0" cellpadding="0" width="100%" style="width:100%;border-collapse:collapse;border-spacing:0px">
<tbody>
<tr>
<td valign="top" style="padding:0in 0in 15pt">
<table border="0" cellspacing="0" cellpadding="0" align="left" style="border-collapse:collapse;border-spacing:0px">
<tbody>
<tr>
<td width="64" valign="top" style="width:48pt;padding:0in">
<p class="MsoNormal" style="line-height:18pt">
<span style="font-size:12.5pt;border:1pt solid windowtext;padding:0in"><img border="0" width="64" height="64" style="width: 0.6666in; height: 0.6666in;" id="gmail-m_1939310555962317815m_-4201325131281898321m_-197615054483451320m_-6512711895812077998gmail-m_962587869508049584_x005f_x0000_i1025" src="cid:172e7f7719d4ce8e91" alt="Image removed by sender."></span><u></u><u></u></p>
</td>
</tr>
</tbody>
</table>
</td>
</tr>
<tr>
<td valign="top" style="padding:0in">
<div>
<p class="MsoNormal" style="line-height:11.25pt">
<b><span style="font-size:10.5pt;font-family:Arial,sans-serif;color:black">Neil Henning</span></b><u></u><u></u></p>
</div>
</td>
</tr>
<tr>
<td valign="top" style="padding:0in">
<div>
<p class="MsoNormal" style="line-height:11.25pt">
<span style="font-size:10.5pt;font-family:Arial,sans-serif;color:black">Senior Software Engineer Compiler</span><u></u><u></u></p>
</div>
</td>
</tr>
<tr>
<td valign="top" style="padding:0in">
<div>
<p class="MsoNormal" style="line-height:11.25pt">
<span style="font-size:9pt;font-family:Arial,sans-serif;color:black"><a href="https://urldefense.com/v3/__http:/unity.com__;!!A4F2R9G_pg!KbQb7EC98K_vFVBRDQoSveXuMcvOvKcWvTers1QW_g1LmGPsOh-wI0r2mV_c2G1yLJhMkp2Z$" target="_blank">unity.com</a></span><u></u><u></u></p>
</div>
</td>
</tr>
</tbody>
</table>
<p class="MsoNormal"> <u></u><u></u></p>
</div>
</div>
</blockquote>
</div>
</div>
</div>
</blockquote>
</div>
</div>
</div>
</div>
</div>
</blockquote></div>