Index: lib/Transform/ScheduleOptimizer.cpp
===================================================================
--- lib/Transform/ScheduleOptimizer.cpp
+++ lib/Transform/ScheduleOptimizer.cpp
@@ -1014,6 +1014,14 @@
   int Car = floor(
       (FirstCacheLevelAssociativity - 1) /
       (1 + static_cast<double>(MicroKernelParams.Nr) / MicroKernelParams.Mr));
+
+  // Car can be computed to be zero since it is floor to int.
+  // On Mac OS, division by 0 does not raise a signal. This causes negative
+  // tile sizes to be computed. Prevent division by 0 Cac by early returning
+  // if this happens.
+  if (Car == 0)
+    return {1, 1, 1};
+
   auto ElementSize = getMatMulAlignTypeSize(MMI);
   assert(ElementSize > 0 && "The element size of the matrix multiplication "
                             "operands should be greater than zero.");
@@ -1022,8 +1030,13 @@
   double Cac =
       static_cast<double>(Kc * ElementSize * SecondCacheLevelAssociativity) /
       SecondCacheLevelSize;
+
   int Mc = floor((SecondCacheLevelAssociativity - 2) / Cac);
   int Nc = PollyPatternMatchingNcQuotient * MicroKernelParams.Nr;
+
+  assert(Mc > 0 && Nc > 0 && Kc > 0 &&
+         "Matrix block sizes should be "
+         " greater than zero");
   return {Mc, Nc, Kc};
 }
 
Index: test/ScheduleOptimizer/ensure-correct-tile-sizes.ll
===================================================================
--- /dev/null
+++ test/ScheduleOptimizer/ensure-correct-tile-sizes.ll
@@ -0,0 +1,84 @@
+; Test that Polly does not crash due to configurations that can lead to
+; incorrect tile size computations.
+; The parameters are setup such that Car in `getMacroKernelParams` is evaluated
+; to 0.
+;
+; RUN: opt  -polly-process-unprofitable  -polly-remarks-minimal  -polly-opt-isl  -polly-pattern-matching-based-opts=true -polly-target-throughput-vector-fma=1  -polly-target-latency-vector-fma=1  -polly-ast -polly-target-vector-register-bitwidth=4096 -polly-target-1st-cache-level-associativity=3 < %s
+;
+;    static const int N = 3000;
+;
+;    void f(int A[N][N], int B[N][N], int C[N][N]) {
+;
+;      for (int i = 0; i < N; i++) {
+;        for (int j = 0; j < N; j++) {
+;          A[i][j] = 0;
+;          for (int k = 0; k < N; k++) {
+;            A[i][j] += B[i][k] * C[k][j];
+;          }
+;        }
+;      }
+;    }
+;
+source_filename = "testbed.c"
+target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
+
+define void @f([3000 x i32]* %A, [3000 x i32]* %B, [3000 x i32]* %C) {
+entry:
+  br label %for.cond
+
+for.cond:                                         ; preds = %for.inc24, %entry
+  %indvars.iv4 = phi i64 [ %indvars.iv.next5, %for.inc24 ], [ 0, %entry ]
+  %exitcond6 = icmp ne i64 %indvars.iv4, 3000
+  br i1 %exitcond6, label %for.body, label %for.end26
+
+for.body:                                         ; preds = %for.cond
+  br label %for.cond1
+
+for.cond1:                                        ; preds = %for.inc21, %for.body
+  %indvars.iv1 = phi i64 [ %indvars.iv.next2, %for.inc21 ], [ 0, %for.body ]
+  %exitcond3 = icmp ne i64 %indvars.iv1, 3000
+  br i1 %exitcond3, label %for.body3, label %for.end23
+
+for.body3:                                        ; preds = %for.cond1
+  %arrayidx5 = getelementptr inbounds [3000 x i32], [3000 x i32]* %A, i64 %indvars.iv4, i64 %indvars.iv1
+  store i32 0, i32* %arrayidx5, align 4
+  br label %for.cond6
+
+for.cond6:                                        ; preds = %for.inc, %for.body3
+  %indvars.iv = phi i64 [ %indvars.iv.next, %for.inc ], [ 0, %for.body3 ]
+  %exitcond = icmp ne i64 %indvars.iv, 3000
+  br i1 %exitcond, label %for.body8, label %for.end
+
+for.body8:                                        ; preds = %for.cond6
+  %arrayidx12 = getelementptr inbounds [3000 x i32], [3000 x i32]* %B, i64 %indvars.iv4, i64 %indvars.iv
+  %tmp = load i32, i32* %arrayidx12, align 4
+  %arrayidx16 = getelementptr inbounds [3000 x i32], [3000 x i32]* %C, i64 %indvars.iv, i64 %indvars.iv1
+  %tmp7 = load i32, i32* %arrayidx16, align 4
+  %mul = mul nsw i32 %tmp, %tmp7
+  %arrayidx20 = getelementptr inbounds [3000 x i32], [3000 x i32]* %A, i64 %indvars.iv4, i64 %indvars.iv1
+  %tmp8 = load i32, i32* %arrayidx20, align 4
+  %add = add nsw i32 %tmp8, %mul
+  store i32 %add, i32* %arrayidx20, align 4
+  br label %for.inc
+
+for.inc:                                          ; preds = %for.body8
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  br label %for.cond6
+
+for.end:                                          ; preds = %for.cond6
+  br label %for.inc21
+
+for.inc21:                                        ; preds = %for.end
+  %indvars.iv.next2 = add nuw nsw i64 %indvars.iv1, 1
+  br label %for.cond1
+
+for.end23:                                        ; preds = %for.cond1
+  br label %for.inc24
+
+for.inc24:                                        ; preds = %for.end23
+  %indvars.iv.next5 = add nuw nsw i64 %indvars.iv4, 1
+  br label %for.cond
+
+for.end26:                                        ; preds = %for.cond
+  ret void
+}