[llvm-commits] [llvm] r147912 - in /llvm/trunk: include/llvm/CodeGen/SlotIndexes.h lib/CodeGen/SplitKit.cpp test/CodeGen/X86/2012-01-10-UndefExceptionEdge.ll
Eli Friedman
eli.friedman at gmail.com
Tue Jan 10 18:43:23 PST 2012
On Tue, Jan 10, 2012 at 6:07 PM, Jakob Stoklund Olesen <stoklund at 2pi.dk> wrote:
> Author: stoklund
> Date: Tue Jan 10 20:07:05 2012
> New Revision: 147912
>
> URL: http://llvm.org/viewvc/llvm-project?rev=147912&view=rev
> Log:
> Detect when a value is undefined on an edge to a landing pad.
>
> Consider this code:
>
> int h() {
> int x;
> try {
> x = f();
> g();
> } catch (...) {
> return x+1;
> }
> return x;
> }
>
> The variable x is undefined on the first edge to the landing pad, but it
> has the f() return value on the second edge to the landing pad.
>
> SplitAnalysis::getLastSplitPoint() would assume that the return value
> from f() was live into the landing pad when f() throws, which is of
> course impossible.
>
> Detect these cases, and treat them as if the landing pad wasn't there.
> This allows spill code to be inserted after the function call to f().
>
> <rdar://problem/10664933>
The included test appears to be crashing on Linux builders
(http://lab.llvm.org:8011/builders/llvm-x86_64-linux,
http://lab.llvm.org:8011/builders/llvm-x86_64-ubuntu). Please look
into it.
-Eli
> Added:
> llvm/trunk/test/CodeGen/X86/2012-01-10-UndefExceptionEdge.ll
> Modified:
> llvm/trunk/include/llvm/CodeGen/SlotIndexes.h
> llvm/trunk/lib/CodeGen/SplitKit.cpp
>
> Modified: llvm/trunk/include/llvm/CodeGen/SlotIndexes.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/SlotIndexes.h?rev=147912&r1=147911&r2=147912&view=diff
> ==============================================================================
> --- llvm/trunk/include/llvm/CodeGen/SlotIndexes.h (original)
> +++ llvm/trunk/include/llvm/CodeGen/SlotIndexes.h Tue Jan 10 20:07:05 2012
> @@ -208,6 +208,12 @@
> return A.lie.getPointer() == B.lie.getPointer();
> }
>
> + /// isEarlierInstr - Return true if A refers to an instruction earlier than
> + /// B. This is equivalent to A < B && !isSameInstr(A, B).
> + static bool isEarlierInstr(SlotIndex A, SlotIndex B) {
> + return A.entry().getIndex() < B.entry().getIndex();
> + }
> +
> /// Return the distance from this index to the given one.
> int distance(SlotIndex other) const {
> return other.getIndex() - getIndex();
>
> Modified: llvm/trunk/lib/CodeGen/SplitKit.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SplitKit.cpp?rev=147912&r1=147911&r2=147912&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/SplitKit.cpp (original)
> +++ llvm/trunk/lib/CodeGen/SplitKit.cpp Tue Jan 10 20:07:05 2012
> @@ -62,13 +62,14 @@
> const MachineBasicBlock *MBB = MF.getBlockNumbered(Num);
> const MachineBasicBlock *LPad = MBB->getLandingPadSuccessor();
> std::pair<SlotIndex, SlotIndex> &LSP = LastSplitPoint[Num];
> + SlotIndex MBBEnd = LIS.getMBBEndIdx(MBB);
>
> // Compute split points on the first call. The pair is independent of the
> // current live interval.
> if (!LSP.first.isValid()) {
> MachineBasicBlock::const_iterator FirstTerm = MBB->getFirstTerminator();
> if (FirstTerm == MBB->end())
> - LSP.first = LIS.getMBBEndIdx(MBB);
> + LSP.first = MBBEnd;
> else
> LSP.first = LIS.getInstructionIndex(FirstTerm);
>
> @@ -89,10 +90,24 @@
>
> // If CurLI is live into a landing pad successor, move the last split point
> // back to the call that may throw.
> - if (LPad && LSP.second.isValid() && LIS.isLiveInToMBB(*CurLI, LPad))
> - return LSP.second;
> - else
> + if (!LPad || !LSP.second || !LIS.isLiveInToMBB(*CurLI, LPad))
> + return LSP.first;
> +
> + // Find the value leaving MBB.
> + const VNInfo *VNI = CurLI->getVNInfoBefore(MBBEnd);
> + if (!VNI)
> return LSP.first;
> +
> + // If the value leaving MBB was defined after the call in MBB, it can't
> + // really be live-in to the landing pad. This can happen if the landing pad
> + // has a PHI, and this register is undef on the exceptional edge.
> + // <rdar://problem/10664933>
> + if (!SlotIndex::isEarlierInstr(VNI->def, LSP.second) && VNI->def < MBBEnd)
> + return LSP.first;
> +
> + // Value is properly live-in to the landing pad.
> + // Only allow splits before the call.
> + return LSP.second;
> }
>
> MachineBasicBlock::iterator
>
> Added: llvm/trunk/test/CodeGen/X86/2012-01-10-UndefExceptionEdge.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/2012-01-10-UndefExceptionEdge.ll?rev=147912&view=auto
> ==============================================================================
> --- llvm/trunk/test/CodeGen/X86/2012-01-10-UndefExceptionEdge.ll (added)
> +++ llvm/trunk/test/CodeGen/X86/2012-01-10-UndefExceptionEdge.ll Tue Jan 10 20:07:05 2012
> @@ -0,0 +1,155 @@
> +; RUN: llc < %s -disable-fp-elim
> +target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32-S128"
> +target triple = "i386-apple-macosx10.7"
> +
> +; This test case has a landing pad with two predecessors, and a variable that
> +; is undef on the first edge while carrying the first function return value on
> +; the second edge.
> +;
> +; Live range splitting tries to isolate the block containing the first function
> +; call, and it is important that the last split point is after the function call
> +; so the return value can spill.
> +;
> +; <rdar://problem/10664933>
> +
> + at Exception = external unnamed_addr constant { i8*, i8* }
> +
> +declare void @llvm.memset.p0i8.i32(i8* nocapture, i8, i32, i32, i1) nounwind
> +
> +define void @f(i32* nocapture %arg, i32* nocapture %arg1, i32* nocapture %arg2, i32* nocapture %arg3, i32 %arg4, i32 %arg5) optsize ssp {
> +bb:
> + br i1 undef, label %bb6, label %bb7
> +
> +bb6: ; preds = %bb
> + %tmp = select i1 false, i32 0, i32 undef
> + br label %bb7
> +
> +bb7: ; preds = %bb6, %bb
> + %tmp8 = phi i32 [ %tmp, %bb6 ], [ 0, %bb ]
> + %tmp9 = shl i32 %tmp8, 2
> + %tmp10 = invoke noalias i8* @_Znam(i32 undef) optsize
> + to label %bb11 unwind label %bb20
> +
> +bb11: ; preds = %bb7
> + %tmp12 = ptrtoint i8* %tmp10 to i32
> + %tmp13 = bitcast i8* %tmp10 to i32*
> + %tmp14 = shl i32 %tmp8, 2
> + %tmp15 = getelementptr i32* %tmp13, i32 undef
> + %tmp16 = getelementptr i32* %tmp13, i32 undef
> + %tmp17 = zext i32 %tmp9 to i64
> + %tmp18 = add i64 %tmp17, -1
> + %tmp19 = icmp ugt i64 %tmp18, 4294967295
> + br i1 %tmp19, label %bb29, label %bb31
> +
> +bb20: ; preds = %bb43, %bb41, %bb29, %bb7
> + %tmp21 = phi i32 [ undef, %bb7 ], [ %tmp12, %bb43 ], [ %tmp12, %bb29 ], [ %tmp12, %bb41 ]
> + %tmp22 = landingpad { i8*, i32 } personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*)
> + catch i8* bitcast ({ i8*, i8* }* @Exception to i8*)
> + br i1 undef, label %bb23, label %bb69
> +
> +bb23: ; preds = %bb38, %bb20
> + %tmp24 = phi i32 [ %tmp12, %bb38 ], [ %tmp21, %bb20 ]
> + %tmp25 = icmp eq i32 %tmp24, 0
> + br i1 %tmp25, label %bb28, label %bb26
> +
> +bb26: ; preds = %bb23
> + %tmp27 = inttoptr i32 %tmp24 to i8*
> + br label %bb28
> +
> +bb28: ; preds = %bb26, %bb23
> + ret void
> +
> +bb29: ; preds = %bb11
> + invoke void @OnOverFlow() optsize
> + to label %bb30 unwind label %bb20
> +
> +bb30: ; preds = %bb29
> + unreachable
> +
> +bb31: ; preds = %bb11
> + %tmp32 = bitcast i32* %tmp15 to i8*
> + %tmp33 = zext i32 %tmp8 to i64
> + %tmp34 = add i64 %tmp33, -1
> + %tmp35 = icmp ugt i64 %tmp34, 4294967295
> + %tmp36 = icmp sgt i32 %tmp8, 0
> + %tmp37 = add i32 %tmp9, -4
> + br label %bb38
> +
> +bb38: ; preds = %bb67, %bb31
> + %tmp39 = phi i32 [ %tmp68, %bb67 ], [ undef, %bb31 ]
> + %tmp40 = icmp sgt i32 %tmp39, undef
> + br i1 %tmp40, label %bb41, label %bb23
> +
> +bb41: ; preds = %bb38
> + invoke void @Pjii(i32* %tmp16, i32 0, i32 %tmp8) optsize
> + to label %bb42 unwind label %bb20
> +
> +bb42: ; preds = %bb41
> + tail call void @llvm.memset.p0i8.i32(i8* %tmp32, i8 0, i32 %tmp9, i32 1, i1 false) nounwind
> + br i1 %tmp35, label %bb43, label %bb45
> +
> +bb43: ; preds = %bb42
> + invoke void @OnOverFlow() optsize
> + to label %bb44 unwind label %bb20
> +
> +bb44: ; preds = %bb43
> + unreachable
> +
> +bb45: ; preds = %bb57, %bb42
> + %tmp46 = phi i32 [ %tmp58, %bb57 ], [ 255, %bb42 ]
> + %tmp47 = icmp slt i32 undef, 0
> + br i1 %tmp47, label %bb48, label %bb59
> +
> +bb48: ; preds = %bb45
> + tail call void @llvm.memset.p0i8.i32(i8* %tmp32, i8 0, i32 %tmp9, i32 1, i1 false) nounwind
> + br i1 %tmp36, label %bb49, label %bb57
> +
> +bb49: ; preds = %bb49, %bb48
> + %tmp50 = phi i32 [ %tmp55, %bb49 ], [ 0, %bb48 ]
> + %tmp51 = add i32 %tmp50, undef
> + %tmp52 = add i32 %tmp50, undef
> + %tmp53 = getelementptr i32* %tmp13, i32 %tmp52
> + %tmp54 = load i32* %tmp53, align 4, !tbaa !0
> + %tmp55 = add i32 %tmp50, 1
> + %tmp56 = icmp eq i32 %tmp55, %tmp8
> + br i1 %tmp56, label %bb57, label %bb49
> +
> +bb57: ; preds = %bb49, %bb48
> + %tmp58 = add i32 %tmp46, -1
> + br label %bb45
> +
> +bb59: ; preds = %bb45
> + %tmp60 = ashr i32 %tmp46, 31
> + tail call void @llvm.memset.p0i8.i32(i8* null, i8 0, i32 %tmp37, i32 1, i1 false) nounwind
> + br i1 %tmp36, label %bb61, label %bb67
> +
> +bb61: ; preds = %bb61, %bb59
> + %tmp62 = phi i32 [ %tmp65, %bb61 ], [ 0, %bb59 ]
> + %tmp63 = add i32 %tmp62, %tmp14
> + %tmp64 = getelementptr i32* %tmp13, i32 %tmp63
> + store i32 0, i32* %tmp64, align 4, !tbaa !0
> + %tmp65 = add i32 %tmp62, 1
> + %tmp66 = icmp eq i32 %tmp65, %tmp8
> + br i1 %tmp66, label %bb67, label %bb61
> +
> +bb67: ; preds = %bb61, %bb59
> + %tmp68 = add i32 %tmp39, -1
> + br label %bb38
> +
> +bb69: ; preds = %bb20
> + resume { i8*, i32 } %tmp22
> +}
> +
> +declare i32 @__gxx_personality_v0(...)
> +
> +declare noalias i8* @_Znam(i32) optsize
> +
> +declare void @Pjii(i32*, i32, i32) optsize
> +
> +declare i32 @llvm.eh.typeid.for(i8*) nounwind readnone
> +
> +declare void @OnOverFlow() noreturn optsize ssp align 2
> +
> +!0 = metadata !{metadata !"int", metadata !1}
> +!1 = metadata !{metadata !"omnipotent char", metadata !2}
> +!2 = metadata !{metadata !"Simple C/C++ TBAA", null}
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
More information about the llvm-commits
mailing list