[PATCH] D30781: [ValueTracking] Out of range shifts might be undef
Oliver Stannard via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Thu Mar 9 05:56:30 PST 2017
olista01 created this revision.
If it is possible for the RHS of a shift operation to be greater than or equal to the bit-width, then the result might be undef, and we can't report any known bits.
In some cases, this was allowing a transformation in instcombine which widened an undef value from i1 to i32, increasing the range of values that a function could return.
Repository:
rL LLVM
https://reviews.llvm.org/D30781
Files:
lib/Analysis/ValueTracking.cpp
test/Transforms/InstCombine/zext-or-icmp.ll
Index: test/Transforms/InstCombine/zext-or-icmp.ll
===================================================================
--- test/Transforms/InstCombine/zext-or-icmp.ll
+++ test/Transforms/InstCombine/zext-or-icmp.ll
@@ -19,3 +19,33 @@
; CHECK-NEXT: ret i8 %zext
}
+; Here, widening the or from i1 to i32 and removing one of the icmps would
+; widen an undef value (created by the out-of-range shift), increasing the
+; range of valid values for the return, so we can't do it.
+define i32 @dont_widen_undef() {
+entry:
+ br label %block2
+
+block1:
+ br label %block2
+
+block2:
+ %m.011 = phi i32 [ 33, %entry ], [ 0, %block1 ]
+ %cmp.i = icmp ugt i32 %m.011, 1
+ %m.1.op = lshr i32 1, %m.011
+ %sext.mask = and i32 %m.1.op, 65535
+ %cmp115 = icmp ne i32 %sext.mask, 0
+ %cmp1 = or i1 %cmp.i, %cmp115
+ %conv2 = zext i1 %cmp1 to i32
+ ret i32 %conv2
+
+; CHECK-LABEL: dont_widen_undef(
+; CHECK: %m.011 = phi i32 [ 33, %entry ], [ 0, %block1 ]
+; CHECK-NEXT: %cmp.i = icmp ugt i32 %m.011, 1
+; CHECK-NEXT: %m.1.op = lshr i32 1, %m.011
+; CHECK-NEXT: %sext.mask = and i32 %m.1.op, 65535
+; CHECK-NEXT: %cmp115 = icmp ne i32 %sext.mask, 0
+; CHECK-NEXT: %cmp1 = or i1 %cmp.i, %cmp115
+; CHECK-NEXT: %conv2 = zext i1 %cmp1 to i32
+; CHECK-NEXT: ret i32 %conv2
+}
Index: lib/Analysis/ValueTracking.cpp
===================================================================
--- lib/Analysis/ValueTracking.cpp
+++ lib/Analysis/ValueTracking.cpp
@@ -848,6 +848,14 @@
computeKnownBits(I->getOperand(1), KnownZero, KnownOne, Depth + 1, Q);
+ // If the shift amount could be greater than or equal to the bit-width of the LHS, the
+ // value could be undef, so we don't know anything about it.
+ if ((~KnownZero).uge(BitWidth)) {
+ KnownZero.clearAllBits();
+ KnownOne.clearAllBits();
+ return;
+ }
+
// Note: We cannot use KnownZero.getLimitedValue() here, because if
// BitWidth > 64 and any upper bits are known, we'll end up returning the
// limit value (which implies all bits are known).
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D30781.91168.patch
Type: text/x-patch
Size: 2054 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170309/d7090851/attachment.bin>
More information about the llvm-commits
mailing list