<div dir="ltr"><br><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Aug 27, 2015 at 4:20 PM, Hal Finkel <span dir="ltr"><<a href="mailto:hfinkel@anl.gov" target="_blank">hfinkel@anl.gov</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Hi David,<br>
<br>
I'd really like this to be a correct change, but I'm not sure that it is. I'd like it to be correct because we add readnone for functions that, at the C level, have __attribute__((const)), and I believe those can be speculated (assuming my understanding of GCC's semantics is correct). However, in terms of readnone itself, we need to be careful about undefined behavior. Specifically,<br>
<br>
int divide(int x, int y) { return x/y; }<br>
<br>
is readnone, according to our definition, but could trap if speculated with y == 0.<br>
<br>
Thanks again,<br>
Hal<br></blockquote><div><br></div><div>Would it be correct for a function which non-deterministically traps (but which does not read or write memory) to be marked readnone? If so, I think our optimizer is in quite a bit of trouble.</div><div><br></div><div>I'll also note that we go out of our way not to mark things like rdrand/rdseed as readnone despite them not reading any memory. I'd prefer that readnone should imply that speculation, CSE, etc. are safe. Would it be problematic to let 'divide' not be readnone?</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div class="HOEnZb"><div class="h5"><br>
----- Original Message -----<br>
> From: "David Majnemer via llvm-commits" <<a href="mailto:llvm-commits@lists.llvm.org">llvm-commits@lists.llvm.org</a>><br>
> To: <a href="mailto:llvm-commits@lists.llvm.org">llvm-commits@lists.llvm.org</a><br>
> Sent: Thursday, August 27, 2015 6:03:01 PM<br>
> Subject: [llvm] r246232 - [ValueTracking] readnone CallInsts are fair game for speculation<br>
><br>
> Author: majnemer<br>
> Date: Thu Aug 27 18:03:01 2015<br>
> New Revision: 246232<br>
><br>
> URL: <a href="http://llvm.org/viewvc/llvm-project?rev=246232&view=rev" rel="noreferrer" target="_blank">http://llvm.org/viewvc/llvm-project?rev=246232&view=rev</a><br>
> Log:<br>
> [ValueTracking] readnone CallInsts are fair game for speculation<br>
><br>
> Any call which is side effect free is trivially OK to speculate. We<br>
> already had similar logic in EarlyCSE and GVN but we were missing it<br>
> from isSafeToSpeculativelyExecute.<br>
><br>
> This fixes PR24601.<br>
><br>
> Modified:<br>
> llvm/trunk/lib/Analysis/ValueTracking.cpp<br>
> llvm/trunk/test/Transforms/SimplifyCFG/speculate-math.ll<br>
><br>
> Modified: llvm/trunk/lib/Analysis/ValueTracking.cpp<br>
> URL:<br>
> <a href="http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ValueTracking.cpp?rev=246232&r1=246231&r2=246232&view=diff" rel="noreferrer" target="_blank">http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ValueTracking.cpp?rev=246232&r1=246231&r2=246232&view=diff</a><br>
> ==============================================================================<br>
> --- llvm/trunk/lib/Analysis/ValueTracking.cpp (original)<br>
> +++ llvm/trunk/lib/Analysis/ValueTracking.cpp Thu Aug 27 18:03:01<br>
> 2015<br>
> @@ -3147,46 +3147,10 @@ bool llvm::isSafeToSpeculativelyExecute(<br>
> LI->getPointerOperand(), LI->getAlignment(), DL, CtxI, DT,<br>
> TLI);<br>
> }<br>
> case Instruction::Call: {<br>
> - if (const IntrinsicInst *II = dyn_cast<IntrinsicInst>(Inst)) {<br>
> - switch (II->getIntrinsicID()) {<br>
> - // These synthetic intrinsics have no side-effects and just<br>
> mark<br>
> - // information about their operands.<br>
> - // FIXME: There are other no-op synthetic instructions that<br>
> potentially<br>
> - // should be considered at least *safe* to speculate...<br>
> - case Intrinsic::dbg_declare:<br>
> - case Intrinsic::dbg_value:<br>
> - return true;<br>
> -<br>
> - case Intrinsic::bswap:<br>
> - case Intrinsic::ctlz:<br>
> - case Intrinsic::ctpop:<br>
> - case Intrinsic::cttz:<br>
> - case Intrinsic::objectsize:<br>
> - case Intrinsic::sadd_with_overflow:<br>
> - case Intrinsic::smul_with_overflow:<br>
> - case Intrinsic::ssub_with_overflow:<br>
> - case Intrinsic::uadd_with_overflow:<br>
> - case Intrinsic::umul_with_overflow:<br>
> - case Intrinsic::usub_with_overflow:<br>
> - return true;<br>
> - // Sqrt should be OK, since the llvm sqrt intrinsic isn't<br>
> defined to set<br>
> - // errno like libm sqrt would.<br>
> - case Intrinsic::sqrt:<br>
> - case Intrinsic::fma:<br>
> - case Intrinsic::fmuladd:<br>
> - case Intrinsic::fabs:<br>
> - case Intrinsic::minnum:<br>
> - case Intrinsic::maxnum:<br>
> - return true;<br>
> - // TODO: some fp intrinsics are marked as having the same<br>
> error handling<br>
> - // as libm. They're safe to speculate when they won't error.<br>
> - // TODO: are convert_{from,to}_fp16 safe?<br>
> - // TODO: can we list target-specific intrinsics here?<br>
> - default: break;<br>
> - }<br>
> - }<br>
> + if (cast<CallInst>(Inst)->doesNotAccessMemory())<br>
> + return true;<br>
> return false; // The called function could have undefined<br>
> behavior or<br>
> - // side-effects, even if marked readnone nounwind.<br>
> + // side-effects.<br>
> }<br>
> case Instruction::VAArg:<br>
> case Instruction::Alloca:<br>
><br>
> Modified: llvm/trunk/test/Transforms/SimplifyCFG/speculate-math.ll<br>
> URL:<br>
> <a href="http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SimplifyCFG/speculate-math.ll?rev=246232&r1=246231&r2=246232&view=diff" rel="noreferrer" target="_blank">http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SimplifyCFG/speculate-math.ll?rev=246232&r1=246231&r2=246232&view=diff</a><br>
> ==============================================================================<br>
> --- llvm/trunk/test/Transforms/SimplifyCFG/speculate-math.ll<br>
> (original)<br>
> +++ llvm/trunk/test/Transforms/SimplifyCFG/speculate-math.ll Thu Aug<br>
> 27 18:03:01 2015<br>
> @@ -6,6 +6,7 @@ declare float @llvm.fmuladd.f32(float, f<br>
> declare float @llvm.fabs.f32(float) nounwind readonly<br>
> declare float @llvm.minnum.f32(float, float) nounwind readonly<br>
> declare float @llvm.maxnum.f32(float, float) nounwind readonly<br>
> +declare float @llvm.copysign.f32(float, float) nounwind readonly<br>
><br>
> ; CHECK-LABEL: @sqrt_test(<br>
> ; CHECK: select<br>
> @@ -107,4 +108,21 @@ test_maxnum.exit:<br>
> %cond.i = phi float [ %0, %cond.else.i ], [ 0x7FF8000000000000,<br>
> %entry ]<br>
> store float %cond.i, float addrspace(1)* %out, align 4<br>
> ret void<br>
> +}<br>
> +<br>
> +; CHECK-LABEL: @copysign_test(<br>
> +; CHECK: select<br>
> +define void @copysign_test(float addrspace(1)* noalias nocapture<br>
> %out, float %a, float %b) nounwind {<br>
> +entry:<br>
> + %cmp.i = fcmp olt float %a, 0.000000e+00<br>
> + br i1 %cmp.i, label %test_copysign.exit, label %cond.else.i<br>
> +<br>
> +cond.else.i: ; preds = %entry<br>
> + %0 = tail call float @llvm.copysign.f32(float %a, float %b)<br>
> nounwind readnone<br>
> + br label %test_copysign.exit<br>
> +<br>
> +test_copysign.exit: ; preds =<br>
> %cond.else.i, %entry<br>
> + %cond.i = phi float [ %0, %cond.else.i ], [ 0x7FF8000000000000,<br>
> %entry ]<br>
> + store float %cond.i, float addrspace(1)* %out, align 4<br>
> + ret void<br>
> }<br>
><br>
><br>
> _______________________________________________<br>
> llvm-commits mailing list<br>
> <a href="mailto:llvm-commits@lists.llvm.org">llvm-commits@lists.llvm.org</a><br>
> <a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits" rel="noreferrer" target="_blank">http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits</a><br>
><br>
<br>
</div></div><span class="HOEnZb"><font color="#888888">--<br>
Hal Finkel<br>
Assistant Computational Scientist<br>
Leadership Computing Facility<br>
Argonne National Laboratory<br>
</font></span></blockquote></div><br></div></div>