caught  division split ignores overflow              %14 = where %8, %6, %13
caught  split remainder may overflow                 AssertionError: overflowing remainder at {0: -100, 1: 1073741824, 2: 1073741824}
caught  x+0.0 -> x on floats                         AssertionError: seed 23: {0: -1073741824, 1: -10000000000.0, 2: False}
caught  float MAX is reordered                       AssertionError: seed 54: {0: -1073741824, 1: nan, 2: False}
caught  mod bucket off by one                        %12 = add %10, %11
caught  ADD range one too small                      %6 = where %5, %4, %3
caught  FLOORMOD range one too small                 %0 = const : dtype=i32 value=0
caught  x*-1 -> x                                    %1 = neg %0
caught  (x//c1)//c2 as x//(c1+c2)                    %2 = floordiv %0, %1
caught  MUL range ignores overflow                   AssertionError: seed 24: MUL marked safe but wrapped: 0 vs 4611686018427387904
caught  const fold reverses operands                 AssertionError: seed 0 {0: -4, 1: -7.006492321624085e-45, 2: False}: folded False, interpr
caught  inline picks the wrong argument              uopc.uop.UOpError: WHERE: both branches must have the same dtype
caught  C: float->int cast not saturating            AssertionError: seed 28 C original {0: -8457356, 1: 4.6577843433232165e+28, 2: True}: 2147
caught  C: floor division uses C's /                 AssertionError: seed 8 C original {0: -2147483647, 1: 8.433293342590332, 2: False}: 1 vs -
caught  C: plain / allowed when divisor may be 0     
caught  C: max of floats uses fmaxf                  AssertionError: seed 15 C original {0: 100, 1: nan, 2: True}: nan vs 0.0
caught  text format drops min/max bounds             AssertionError: seed 11
caught  hash cons key ignores sign of zero           %0 = const : dtype=f32 value=inf
caught  AFTER dropped from the renderer's order      AssertionError: seed 25: visiting order changed buffer 1
19 of 19 deliberate bugs were caught
