diff --git a/llvm/include/llvm/CodeGen/GlobalISel/MIPatternMatch.h b/llvm/include/llvm/CodeGen/GlobalISel/MIPatternMatch.h
--- a/llvm/include/llvm/CodeGen/GlobalISel/MIPatternMatch.h
+++ b/llvm/include/llvm/CodeGen/GlobalISel/MIPatternMatch.h
@@ -53,6 +53,24 @@
 
 inline ConstantMatch m_ICst(int64_t &Cst) { return ConstantMatch(Cst); }
 
+/// Matcher for a specific constant value.
+struct SpecificConstantMatch {
+  int64_t RequestedVal;
+  SpecificConstantMatch(int64_t RequestedVal) : RequestedVal(RequestedVal) {}
+  bool match(const MachineRegisterInfo &MRI, Register Reg) {
+    int64_t MatchedVal;
+    return mi_match(Reg, MRI, m_ICst(MatchedVal)) && MatchedVal == RequestedVal;
+  }
+};
+
+/// Matches a constant equal to \p RequestedValue.
+inline SpecificConstantMatch m_SpecificICst(int64_t RequestedValue) {
+  return SpecificConstantMatch(RequestedValue);
+}
+
+/// Matches an integer 0.
+inline SpecificConstantMatch m_ZeroInt() { return SpecificConstantMatch(0); }
+
 // TODO: Rework this for different kinds of MachineOperand.
 // Currently assumes the Src for a match is a register.
 // We might want to support taking in some MachineOperands and call getReg on
@@ -425,6 +443,14 @@
                          TargetOpcode::G_INSERT_VECTOR_ELT>(Src0, Src1, Src2);
 }
 
+/// Matches a register negated by a G_SUB.
+/// G_SUB 0, %negated_reg
+template <typename SrcTy>
+inline BinaryOp_match<SpecificConstantMatch, SrcTy, TargetOpcode::G_SUB>
+m_Neg(const SrcTy &&Src) {
+  return m_GSub(m_ZeroInt(), Src);
+}
+
 } // namespace GMIPatternMatch
 } // namespace llvm
 
diff --git a/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp b/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp
--- a/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp
@@ -2614,9 +2614,7 @@
   // ((0-A) + B) -> B - A
   // (A + (0-B)) -> A - B
   auto CheckFold = [&](Register &MaybeSub, Register &MaybeNewLHS) {
-    int64_t Cst;
-    if (!mi_match(MaybeSub, MRI, m_GSub(m_ICst(Cst), m_Reg(NewRHS))) ||
-        Cst != 0)
+    if (!mi_match(MaybeSub, MRI, m_Neg(m_Reg(NewRHS))))
       return false;
     NewLHS = MaybeNewLHS;
     return true;
diff --git a/llvm/lib/CodeGen/GlobalISel/Utils.cpp b/llvm/lib/CodeGen/GlobalISel/Utils.cpp
--- a/llvm/lib/CodeGen/GlobalISel/Utils.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/Utils.cpp
@@ -682,9 +682,7 @@
   const unsigned NumOps = MI.getNumOperands();
   for (unsigned I = 1; I != NumOps; ++I) {
     Register Element = MI.getOperand(I).getReg();
-    int64_t ElementValue;
-    if (!mi_match(Element, MRI, m_ICst(ElementValue)) ||
-        ElementValue != SplatValue)
+    if (!mi_match(Element, MRI, m_SpecificICst(SplatValue)))
       return false;
   }
 
diff --git a/llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp b/llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp
--- a/llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp
+++ b/llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp
@@ -1837,10 +1837,7 @@
   // Also take the opportunity here to try to do some optimization.
   // Try to convert this into a G_SUB if the offset is a 0-x negate idiom.
   Register NegatedReg;
-  int64_t Cst;
-  if (!mi_match(I.getOperand(2).getReg(), MRI,
-                m_GSub(m_ICst(Cst), m_Reg(NegatedReg))) ||
-      Cst != 0)
+  if (!mi_match(I.getOperand(2).getReg(), MRI, m_Neg(m_Reg(NegatedReg))))
     return true;
   I.getOperand(2).setReg(NegatedReg);
   I.setDesc(TII.get(TargetOpcode::G_SUB));
diff --git a/llvm/lib/Target/AArch64/GISel/AArch64PostLegalizerLowering.cpp b/llvm/lib/Target/AArch64/GISel/AArch64PostLegalizerLowering.cpp
--- a/llvm/lib/Target/AArch64/GISel/AArch64PostLegalizerLowering.cpp
+++ b/llvm/lib/Target/AArch64/GISel/AArch64PostLegalizerLowering.cpp
@@ -291,8 +291,7 @@
     return false;
 
   // Match the index constant 0.
-  int64_t Index = 0;
-  if (!mi_match(InsMI->getOperand(3).getReg(), MRI, m_ICst(Index)) || Index)
+  if (!mi_match(InsMI->getOperand(3).getReg(), MRI, m_ZeroInt()))
     return false;
 
   MatchInfo = ShuffleVectorPseudo(AArch64::G_DUP, MI.getOperand(0).getReg(),
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp b/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
--- a/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
@@ -633,7 +633,6 @@
 
   Register ShiftSrc0;
   Register ShiftSrc1;
-  int64_t ShiftAmt;
 
   // With multiple uses of the shift, this will duplicate the shift and
   // increase register pressure.
@@ -645,14 +644,11 @@
   // (build_vector_trunc $src0, $src1)
   //  => (S_PACK_LL_B32_B16 $src0, $src1)
 
-  // FIXME: This is an inconvenient way to check a specific value
   bool Shift0 = mi_match(
-    Src0, *MRI, m_OneUse(m_GLShr(m_Reg(ShiftSrc0), m_ICst(ShiftAmt)))) &&
-    ShiftAmt == 16;
+      Src0, *MRI, m_OneUse(m_GLShr(m_Reg(ShiftSrc0), m_SpecificICst(16))));
 
   bool Shift1 = mi_match(
-    Src1, *MRI, m_OneUse(m_GLShr(m_Reg(ShiftSrc1), m_ICst(ShiftAmt)))) &&
-    ShiftAmt == 16;
+      Src1, *MRI, m_OneUse(m_GLShr(m_Reg(ShiftSrc1), m_SpecificICst(16))));
 
   unsigned Opc = AMDGPU::S_PACK_LL_B32_B16;
   if (Shift0 && Shift1) {
@@ -3474,9 +3470,7 @@
   if (Def->getOpcode() != AMDGPU::G_MERGE_VALUES)
     return false;
 
-  int64_t MergeRHS;
-  if (mi_match(Def->getOperand(2).getReg(), MRI, m_ICst(MergeRHS)) &&
-      MergeRHS == 0) {
+  if (mi_match(Def->getOperand(2).getReg(), MRI, m_ZeroInt())) {
     return Def->getOperand(1).getReg();
   }
 
diff --git a/llvm/unittests/CodeGen/GlobalISel/PatternMatchTest.cpp b/llvm/unittests/CodeGen/GlobalISel/PatternMatchTest.cpp
--- a/llvm/unittests/CodeGen/GlobalISel/PatternMatchTest.cpp
+++ b/llvm/unittests/CodeGen/GlobalISel/PatternMatchTest.cpp
@@ -392,6 +392,71 @@
   match = mi_match(MIBAdd.getReg(0), *MRI, m_OneUse(m_GAdd(m_Reg(), m_Reg())));
   EXPECT_FALSE(match);
 }
+
+TEST_F(AArch64GISelMITest, MatchSpecificConstant) {
+  setUp();
+  if (!TM)
+    return;
+
+  // Basic case: Can we match a G_CONSTANT with a specific value?
+  auto FortyTwo = B.buildConstant(LLT::scalar(64), 42);
+  EXPECT_TRUE(mi_match(FortyTwo.getReg(0), *MRI, m_SpecificICst(42)));
+  EXPECT_FALSE(mi_match(FortyTwo.getReg(0), *MRI, m_SpecificICst(123)));
+
+  // Test that this works inside of a more complex pattern.
+  LLT s64 = LLT::scalar(64);
+  auto MIBAdd = B.buildAdd(s64, Copies[0], FortyTwo);
+  EXPECT_TRUE(mi_match(MIBAdd.getReg(2), *MRI, m_SpecificICst(42)));
+
+  // Wrong constant.
+  EXPECT_FALSE(mi_match(MIBAdd.getReg(2), *MRI, m_SpecificICst(123)));
+
+  // No constant on the LHS.
+  EXPECT_FALSE(mi_match(MIBAdd.getReg(1), *MRI, m_SpecificICst(42)));
+}
+
+TEST_F(AArch64GISelMITest, MatchZeroInt) {
+  setUp();
+  if (!TM)
+    return;
+  auto Zero = B.buildConstant(LLT::scalar(64), 0);
+  EXPECT_TRUE(mi_match(Zero.getReg(0), *MRI, m_ZeroInt()));
+
+  auto FortyTwo = B.buildConstant(LLT::scalar(64), 42);
+  EXPECT_FALSE(mi_match(FortyTwo.getReg(0), *MRI, m_ZeroInt()));
+}
+
+TEST_F(AArch64GISelMITest, MatchNeg) {
+  setUp();
+  if (!TM)
+    return;
+
+  LLT s64 = LLT::scalar(64);
+  auto Zero = B.buildConstant(LLT::scalar(64), 0);
+  auto NegInst = B.buildSub(s64, Zero, Copies[0]);
+  Register NegatedReg;
+
+  // Match: G_SUB = 0, %Reg
+  EXPECT_TRUE(mi_match(NegInst.getReg(0), *MRI, m_Neg(m_Reg(NegatedReg))));
+  EXPECT_EQ(NegatedReg, Copies[0]);
+
+  // Don't match: G_SUB = %Reg, 0
+  auto NotNegInst1 = B.buildSub(s64, Copies[0], Zero);
+  EXPECT_FALSE(mi_match(NotNegInst1.getReg(0), *MRI, m_Neg(m_Reg(NegatedReg))));
+
+  // Don't match: G_SUB = 42, %Reg
+  auto FortyTwo = B.buildConstant(LLT::scalar(64), 42);
+  auto NotNegInst2 = B.buildSub(s64, FortyTwo, Copies[0]);
+  EXPECT_FALSE(mi_match(NotNegInst2.getReg(0), *MRI, m_Neg(m_Reg(NegatedReg))));
+
+  // Complex testcase.
+  // %sub = G_SUB = 0, %negated_reg
+  // %add = G_ADD = %x, %sub
+  auto AddInst = B.buildAdd(s64, Copies[1], NegInst);
+  NegatedReg = Register();
+  EXPECT_TRUE(mi_match(AddInst.getReg(2), *MRI, m_Neg(m_Reg(NegatedReg))));
+  EXPECT_EQ(NegatedReg, Copies[0]);
+}
 } // namespace
 
 int main(int argc, char **argv) {