<div class="socmaildefaultfont" dir="ltr" style="font-family:Arial, Helvetica, sans-serif;font-size:10.5pt" ><div dir="ltr" >Sorry about this Eric. My local builds have compiler-rt in tree so I'm not sure why I didn't see any failures in check-all but I will rerun and see if I can reproduce them. Hiroshi actually already encountered a failure due to this patch and has <a href="https://reviews.llvm.org/D40893" >https://reviews.llvm.org/D40893</a> up to fix it. Maybe see if that fixes the asan/ubsan issues as well.</div>
<div dir="ltr" > </div>
<div dir="ltr" >Thanks</div>
<div dir="ltr" >Sean</div>
<div dir="ltr" > </div>
<div dir="ltr" > </div>
<div dir="ltr" > </div>
<blockquote data-history-content-modified="1" dir="ltr" style="border-left:solid #aaaaaa 2px; margin-left:5px; padding-left:5px; direction:ltr; margin-right:0px" >----- Original message -----<br>From: Eric Christopher <echristo@gmail.com><br>To: Sean Fertile <sfertile@ca.ibm.com><br>Cc: llvm-commits@lists.llvm.org<br>Subject: Re: [llvm] r319218 - [PowerPC] Allow tail calls of fastcc functions from C CallingConv functions.<br>Date: Thu, Dec 7, 2017 5:33 PM<br>
<div dir="ltr" >Hi Sean,
<div> </div>
<div>I've temporarily reverted this here:</div>
<div> </div>
<div><div>echristo@athyra ~/s/llvm> git svn dcommit</div>
<div>Committing to <a href="https://urldefense.proofpoint.com/v2/url?u=https-3A__llvm.org_svn_llvm-2Dproject_llvm_trunk&d=DwMFaQ&c=jf_iaSHvJObTbx-siA1ZOg&r=vE89RddfSAMwAmW9D-c51sqQCl8tXHenTEyBrBeQeeA&m=TOBFZlP7SHQipHSsdmLZhRQP2gHxIp0U-FwX36i3ipE&s=QLKYUKbjh8ziMOOMF03X8ulonjDvq8IGyu1FqAM8ZjQ&e=" target="_blank" >https://llvm.org/svn/llvm-project/llvm/trunk</a> ...</div>
<div>M<span style="white-space:pre" > </span>lib/Target/PowerPC/PPCISelLowering.cpp</div>
<div>M<span style="white-space:pre" > </span>test/CodeGen/PowerPC/duplicate-returns-for-tailcall.ll</div>
<div>Committed r320106</div></div>
<div> </div>
<div>since we were showing it as breaking asan and ubsan tests. Unfortunately this wasn't on the bots as they're currently down, but it was evident with an asan build.</div>
<div> </div>
<div>I've approved patches to get the external bots up so we'll see if we can reproduce there and I'll work with you on getting some sort of test case here.</div>
<div> </div>
<div>-eric</div></div>
<div><div dir="ltr" >On Tue, Nov 28, 2017 at 12:26 PM Sean Fertile via llvm-commits <<a href="mailto:llvm-commits@lists.llvm.org" target="_blank" >llvm-commits@lists.llvm.org</a>> wrote:</div>
<blockquote style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex" >Author: sfertile<br>Date: Tue Nov 28 12:25:58 2017<br>New Revision: 319218<br><br>URL: <a href="https://urldefense.proofpoint.com/v2/url?u=http-3A__llvm.org_viewvc_llvm-2Dproject-3Frev-3D319218-26view-3Drev&d=DwMFaQ&c=jf_iaSHvJObTbx-siA1ZOg&r=vE89RddfSAMwAmW9D-c51sqQCl8tXHenTEyBrBeQeeA&m=TOBFZlP7SHQipHSsdmLZhRQP2gHxIp0U-FwX36i3ipE&s=KQHA212Nct9xA5l_-TVetu7YTD7T4PnP3EVuapTHpb8&e=" rel="noreferrer" target="_blank" >http://llvm.org/viewvc/llvm-project?rev=319218&view=rev</a><br>Log:<br>[PowerPC] Allow tail calls of fastcc functions from C CallingConv functions.<br><br>Allow fastcc callees to be tail-called from ccc callers.<br><br>Differential Revision: <a href="https://urldefense.proofpoint.com/v2/url?u=https-3A__reviews.llvm.org_D40355&d=DwMFaQ&c=jf_iaSHvJObTbx-siA1ZOg&r=vE89RddfSAMwAmW9D-c51sqQCl8tXHenTEyBrBeQeeA&m=TOBFZlP7SHQipHSsdmLZhRQP2gHxIp0U-FwX36i3ipE&s=J2Cz7PInl5hSqpIUkRrp0AS_pUXsMkzeaKb-6Q9pEkk&e=" rel="noreferrer" target="_blank" >https://reviews.llvm.org/D40355</a><br><br>Modified:<br> llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp<br> llvm/trunk/test/CodeGen/PowerPC/duplicate-returns-for-tailcall.ll<br><br>Modified: llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp<br>URL: <a href="https://urldefense.proofpoint.com/v2/url?u=http-3A__llvm.org_viewvc_llvm-2Dproject_llvm_trunk_lib_Target_PowerPC_PPCISelLowering.cpp-3Frev-3D319218-26r1-3D319217-26r2-3D319218-26view-3Ddiff&d=DwMFaQ&c=jf_iaSHvJObTbx-siA1ZOg&r=vE89RddfSAMwAmW9D-c51sqQCl8tXHenTEyBrBeQeeA&m=TOBFZlP7SHQipHSsdmLZhRQP2gHxIp0U-FwX36i3ipE&s=5Tc7DZFZGJ5VzQENAdHNlSL--DDy8fYgYi8rV92anho&e=" rel="noreferrer" target="_blank" >http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp?rev=319218&r1=319217&r2=319218&view=diff</a><br>==============================================================================<br>--- llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp (original)<br>+++ llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp Tue Nov 28 12:25:58 2017<br>@@ -4397,13 +4397,18 @@ hasSameArgumentList(const Function *Call<br> static bool<br> areCallingConvEligibleForTCO_64SVR4(CallingConv::ID CallerCC,<br> CallingConv::ID CalleeCC) {<br>- // Tail or Sibling call optimization (TCO/SCO) needs callee and caller to<br>- // have the same calling convention.<br>- if (CallerCC != CalleeCC)<br>+ // tail calls are possible with fastcc and ccc.<br>+ auto isTailCallableCC = [] (CallingConv::ID CC){<br>+ return CC == CallingConv::C || CC == CallingConv::Fast;<br>+ };<br>+ if (!isTailCallableCC(CallerCC) || !isTailCallableCC(CalleeCC))<br> return false;<br><br>- // Tail or Sibling calls can be done with fastcc/ccc.<br>- return (CallerCC == CallingConv::Fast || CallerCC == CallingConv::C);<br>+ // We can safely tail call both fastcc and ccc callees from a c calling<br>+ // convention caller. If the caller is fastcc, we may have less stack space<br>+ // then a non-fastcc caller with the same signature so disable tail-calls in<br>+ // that case.<br>+ return CallerCC == CallingConv::C || CallerCC == CalleeCC;<br> }<br><br> bool<br><br>Modified: llvm/trunk/test/CodeGen/PowerPC/duplicate-returns-for-tailcall.ll<br>URL: <a href="https://urldefense.proofpoint.com/v2/url?u=http-3A__llvm.org_viewvc_llvm-2Dproject_llvm_trunk_test_CodeGen_PowerPC_duplicate-2Dreturns-2Dfor-2Dtailcall.ll-3Frev-3D319218-26r1-3D319217-26r2-3D319218-26view-3Ddiff&d=DwMFaQ&c=jf_iaSHvJObTbx-siA1ZOg&r=vE89RddfSAMwAmW9D-c51sqQCl8tXHenTEyBrBeQeeA&m=TOBFZlP7SHQipHSsdmLZhRQP2gHxIp0U-FwX36i3ipE&s=7pJqkTkBrJHAKFM6RuKE2_imobSQEq6fS2EQ-SNvqQI&e=" rel="noreferrer" target="_blank" >http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/PowerPC/duplicate-returns-for-tailcall.ll?rev=319218&r1=319217&r2=319218&view=diff</a><br>==============================================================================<br>--- llvm/trunk/test/CodeGen/PowerPC/duplicate-returns-for-tailcall.ll (original)<br>+++ llvm/trunk/test/CodeGen/PowerPC/duplicate-returns-for-tailcall.ll Tue Nov 28 12:25:58 2017<br>@@ -42,10 +42,10 @@ if.end4:<br> if.then6: ; preds = %if.end4<br> %call7 = tail call fastcc signext i32 @call3(i32 signext %a, i32 signext %b, i32 signext %c)<br> br label %return<br>-; No duplication here because the calling convention mismatch means we won't tail-call<br>+; tail calling a fastcc function from a ccc function is supported.<br> ; CHECK_LABEL: if.then13:<br>-; CHECK: tail call fastcc signext i32 @call3<br>-; CHECK-NEXT: br<br>+; CHECK: %[[T2:[a-zA-Z0-9]+]] = tail call fastcc signext i32 @call3<br>+; CHECK-NEXT: ret i32 %[[T2]]<br><br> return: ; preds = %if.end4, %if.then6, %if.then2, %if.then<br> %retval.0 = phi i32 [ %call, %if.then ], [ %call3, %if.then2 ], [ %call7, %if.then6 ], [ %c, %if.end4 ]<br><br><br>_______________________________________________<br>llvm-commits mailing list<br><a href="mailto:llvm-commits@lists.llvm.org" target="_blank" >llvm-commits@lists.llvm.org</a><br><a href="https://urldefense.proofpoint.com/v2/url?u=http-3A__lists.llvm.org_cgi-2Dbin_mailman_listinfo_llvm-2Dcommits&d=DwMFaQ&c=jf_iaSHvJObTbx-siA1ZOg&r=vE89RddfSAMwAmW9D-c51sqQCl8tXHenTEyBrBeQeeA&m=TOBFZlP7SHQipHSsdmLZhRQP2gHxIp0U-FwX36i3ipE&s=AWmK4VG15oZn1VtPRF72qUU9vyfYrFHGynJ9yhuaoV8&e=" rel="noreferrer" target="_blank" >http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits</a></blockquote></div></blockquote>
<div dir="ltr" > </div></div><BR>