diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index 2d78aa4be3455..5aaed70f5895f 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -499,8 +499,9 @@ static void computeKnownBitsAddSub(bool Add, const Value *Op0, const Value *Op1,
   KnownOut = KnownBits::computeForAddSub(Add, NSW, NUW, Known2, KnownOut);
 
   if (!Add && NSW && !KnownOut.isNonNegative() &&
-      isImpliedByDomCondition(ICmpInst::ICMP_SLE, Op1, Op0, Q.CxtI, Q.DL)
-          .value_or(false))
+      (isImpliedByDomCondition(ICmpInst::ICMP_SLE, Op1, Op0, Q.CxtI, Q.DL)
+           .value_or(false) ||
+       match(Op1, m_c_SMin(m_Specific(Op0), m_Value()))))
     KnownOut.makeNonNegative();
 
   if (Add)
diff --git a/llvm/test/Transforms/InstCombine/sext-nonneg-sub.ll b/llvm/test/Transforms/InstCombine/sext-nonneg-sub.ll
new file mode 100644
index 0000000000000..b1121637c83b4
--- /dev/null
+++ b/llvm/test/Transforms/InstCombine/sext-nonneg-sub.ll
@@ -0,0 +1,63 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
+; RUN: opt < %s -passes=instcombine -S | FileCheck %s
+
+; Test that b - smin(b, a) is recognized as non-negative
+define i64 @func1(i32 %a, i32 %b) {
+; CHECK-LABEL: define i64 @func1(
+; CHECK-SAME: i32 [[A:%.*]], i32 [[B:%.*]]) {
+; CHECK-NEXT:  [[ENTRY:.*:]]
+; CHECK-NEXT:    [[SPEC_SELECT:%.*]] = tail call i32 @llvm.smin.i32(i32 [[B]], i32 [[A]])
+; CHECK-NEXT:    [[SUB:%.*]] = sub nsw i32 [[B]], [[SPEC_SELECT]]
+; CHECK-NEXT:    [[CONV:%.*]] = zext nneg i32 [[SUB]] to i64
+; CHECK-NEXT:    ret i64 [[CONV]]
+;
+entry:
+  %spec.select = tail call i32 @llvm.smin.i32(i32 %b, i32 %a)
+  %sub = sub nsw i32 %b, %spec.select
+  %conv = sext i32 %sub to i64
+  ret i64 %conv
+}
+
+; Test commutative smin pattern: b - smin(a, b) should also optimize
+define i64 @smin_commutative(i32 %a, i32 %b) {
+; CHECK-LABEL: define i64 @smin_commutative(
+; CHECK-SAME: i32 [[A:%.*]], i32 [[B:%.*]]) {
+; CHECK-NEXT:    [[MIN:%.*]] = call i32 @llvm.smin.i32(i32 [[A]], i32 [[B]])
+; CHECK-NEXT:    [[SUB:%.*]] = sub nsw i32 [[B]], [[MIN]]
+; CHECK-NEXT:    [[EXT:%.*]] = zext nneg i32 [[SUB]] to i64
+; CHECK-NEXT:    ret i64 [[EXT]]
+;
+  %min = call i32 @llvm.smin.i32(i32 %a, i32 %b)
+  %sub = sub nsw i32 %b, %min
+  %ext = sext i32 %sub to i64
+  ret i64 %ext
+}
+
+; NEGATIVE TEST: mismatched operands should not optimize
+define i64 @smin_mismatch(i32 %a, i32 %b, i32 %c) {
+; CHECK-LABEL: define i64 @smin_mismatch(
+; CHECK-SAME: i32 [[A:%.*]], i32 [[B:%.*]], i32 [[C:%.*]]) {
+; CHECK-NEXT:    [[MIN:%.*]] = call i32 @llvm.smin.i32(i32 [[B]], i32 [[C]])
+; CHECK-NEXT:    [[SUB:%.*]] = sub nsw i32 [[A]], [[MIN]]
+; CHECK-NEXT:    [[EXT:%.*]] = sext i32 [[SUB]] to i64
+; CHECK-NEXT:    ret i64 [[EXT]]
+;
+  %min = call i32 @llvm.smin.i32(i32 %b, i32 %c)
+  %sub = sub nsw i32 %a, %min
+  %ext = sext i32 %sub to i64
+  ret i64 %ext
+}
+
+
+; NEGATIVE TEST: unguarded subtraction should NOT optimize
+define i64 @neg_unguarded_sub(i32 %a, i32 %b) {
+; CHECK-LABEL: define i64 @neg_unguarded_sub(
+; CHECK-SAME: i32 [[A:%.*]], i32 [[B:%.*]]) {
+; CHECK-NEXT:    [[SUB:%.*]] = sub nsw i32 [[B]], [[A]]
+; CHECK-NEXT:    [[EXT:%.*]] = sext i32 [[SUB]] to i64
+; CHECK-NEXT:    ret i64 [[EXT]]
+;
+  %sub = sub nsw i32 %b, %a
+  %ext = sext i32 %sub to i64
+  ret i64 %ext
+}