[llvm] [APInt] Add default-disabled assertion to APInt constructor (PR #106524)

[via llvm-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-llvm-adt

@llvm/pr-subscribers-llvm-support

Author: Nikita Popov (nikic)

<details>
<summary>Changes</summary>

If the uint64_t constructor is used, assert that the value is actually a signed or unsigned N-bit integer depending on whether the isSigned flag is set. Provide an implicitTrunc flag to restore the previous behavior, where the argument is silently truncated instead.

In this commit, implicitTrunc is enabled by default, which means that the new assertions are disabled and no actual change in behavior occurs. The plan is to flip the default once all places violating the assertion have been fixed. See #<!-- -->80309 for the scope of the necessary changes.

The primary motivation for this change is to avoid incorrectly specified isSigned flags. A recurring problem we have is that people write something like `APInt(BW, -1)` and this works perfectly fine -- until the code path is hit with `BW > 64`. Most of our i128 specific miscompilations are caused by variants of this issue.

The cost of the change is that we have to specify the correct isSigned flag (and make sure there are no excess bits) for uses where BW is always <= 64 as well.

---
Full diff: https://github.com/llvm/llvm-project/pull/106524.diff


3 Files Affected:

- (modified) llvm/include/llvm/ADT/APInt.h (+17-2) 
- (modified) llvm/lib/Support/APInt.cpp (+9-5) 
- (modified) llvm/unittests/ADT/APIntTest.cpp (+3-2) 


``````````diff
diff --git a/llvm/include/llvm/ADT/APInt.h b/llvm/include/llvm/ADT/APInt.h
index 65ba3f15305c78..a42dae8887392d 100644
--- a/llvm/include/llvm/ADT/APInt.h
+++ b/llvm/include/llvm/ADT/APInt.h
@@ -106,11 +106,26 @@ class [[nodiscard]] APInt {
   /// \param numBits the bit width of the constructed APInt
   /// \param val the initial value of the APInt
   /// \param isSigned how to treat signedness of val
-  APInt(unsigned numBits, uint64_t val, bool isSigned = false)
+  /// \param implicitTrunc allow implicit truncation of non-zero/sign bits of
+  ///                      val beyond the range of numBits
+  APInt(unsigned numBits, uint64_t val, bool isSigned = false,
+        bool implicitTrunc = true)
       : BitWidth(numBits) {
+    if (!implicitTrunc) {
+      if (BitWidth == 0) {
+        assert(val == 0 && "Value must be zero for 0-bit APInt");
+      } else if (isSigned) {
+        assert(llvm::isIntN(BitWidth, val) &&
+               "Value is not an N-bit signed value");
+      } else {
+        assert(llvm::isUIntN(BitWidth, val) &&
+               "Value is not an N-bit unsigned value");
+      }
+    }
     if (isSingleWord()) {
       U.VAL = val;
-      clearUnusedBits();
+      if (implicitTrunc || isSigned)
+        clearUnusedBits();
     } else {
       initSlowCase(val, isSigned);
     }
diff --git a/llvm/lib/Support/APInt.cpp b/llvm/lib/Support/APInt.cpp
index fe22e9ba04b6f5..78d573966c6c99 100644
--- a/llvm/lib/Support/APInt.cpp
+++ b/llvm/lib/Support/APInt.cpp
@@ -234,7 +234,8 @@ APInt& APInt::operator-=(uint64_t RHS) {
 APInt APInt::operator*(const APInt& RHS) const {
   assert(BitWidth == RHS.BitWidth && "Bit widths must be the same");
   if (isSingleWord())
-    return APInt(BitWidth, U.VAL * RHS.U.VAL);
+    return APInt(BitWidth, U.VAL * RHS.U.VAL, /*isSigned=*/false,
+                 /*implicitTrunc=*/true);
 
   APInt Result(getMemory(getNumWords()), getBitWidth());
   tcMultiply(Result.U.pVal, U.pVal, RHS.U.pVal, getNumWords());
@@ -455,7 +456,8 @@ APInt APInt::extractBits(unsigned numBits, unsigned bitPosition) const {
          "Illegal bit extraction");
 
   if (isSingleWord())
-    return APInt(numBits, U.VAL >> bitPosition);
+    return APInt(numBits, U.VAL >> bitPosition, /*isSigned=*/false,
+                 /*implicitTrunc=*/true);
 
   unsigned loBit = whichBit(bitPosition);
   unsigned loWord = whichWord(bitPosition);
@@ -463,7 +465,8 @@ APInt APInt::extractBits(unsigned numBits, unsigned bitPosition) const {
 
   // Single word result extracting bits from a single word source.
   if (loWord == hiWord)
-    return APInt(numBits, U.pVal[loWord] >> loBit);
+    return APInt(numBits, U.pVal[loWord] >> loBit, /*isSigned=*/false,
+                 /*implicitTrunc=*/true);
 
   // Extracting bits that start on a source word boundary can be done
   // as a fast memory copy.
@@ -907,7 +910,8 @@ APInt APInt::trunc(unsigned width) const {
   assert(width <= BitWidth && "Invalid APInt Truncate request");
 
   if (width <= APINT_BITS_PER_WORD)
-    return APInt(width, getRawData()[0]);
+    return APInt(width, getRawData()[0], /*isSigned=*/false,
+                 /*implicitTrunc=*/true);
 
   if (width == BitWidth)
     return *this;
@@ -955,7 +959,7 @@ APInt APInt::sext(unsigned Width) const {
   assert(Width >= BitWidth && "Invalid APInt SignExtend request");
 
   if (Width <= APINT_BITS_PER_WORD)
-    return APInt(Width, SignExtend64(U.VAL, BitWidth));
+    return APInt(Width, SignExtend64(U.VAL, BitWidth), /*isSigned=*/true);
 
   if (Width == BitWidth)
     return *this;
diff --git a/llvm/unittests/ADT/APIntTest.cpp b/llvm/unittests/ADT/APIntTest.cpp
index eb4b847185f53b..fff29d24a05299 100644
--- a/llvm/unittests/ADT/APIntTest.cpp
+++ b/llvm/unittests/ADT/APIntTest.cpp
@@ -220,11 +220,12 @@ TEST(APIntTest, i256) {
 }
 
 TEST(APIntTest, i1) {
-  const APInt neg_two(1, static_cast<uint64_t>(-2), true);
+  const APInt neg_two(1, static_cast<uint64_t>(-2), true,
+                      /*implicitTrunc=*/true);
   const APInt neg_one(1, static_cast<uint64_t>(-1), true);
   const APInt zero(1, 0);
   const APInt one(1, 1);
-  const APInt two(1, 2);
+  const APInt two(1, 2, false, /*implicitTrunc=*/true);
 
   EXPECT_EQ(0, neg_two.getSExtValue());
   EXPECT_EQ(-1, neg_one.getSExtValue());

``````````

</details>


https://github.com/llvm/llvm-project/pull/106524