diff options
| author | Matthew Bentham <matthew.bentham@arm.com> | 2021-07-05 15:45:25 +0100 |
|---|---|---|
| committer | Sadik Armagan <sadik.armagan@arm.com> | 2021-07-16 14:39:09 +0000 |
| commit | 94df84d1e4964f00fbeaa29a720ef1822ba033d6 (patch) | |
| tree | 15487f86c06dd3a2e563030cb5c1ee68f41f0684 | |
| parent | c013bc85f45add8c4bc8c7163316bbb90b37a15e (diff) | |
| download | armnn-94df84d1e4964f00fbeaa29a720ef1822ba033d6.tar.gz | |
Avoid undefined behaviour in NumericCast unit test
Because the maximum value of an int32_t isn't exactly representable
by a 32-bit float, casting it to a float produces a greater value,
and then casting that back to an int32_t is undefined. (In tested
implementations the cast saturates, so accidentally the correct
result is returned).
Using double for the intermediate value instead gives a high
enough precision that this test case can be safely expressed.
Silences warning from Undefioned Behaviour Sanitiser.
Signed-off-by: Matthew Bentham <matthew.bentham@arm.com>
Change-Id: I6fe3e89cb6ac695049c92f3b75aca0c14d0e14aa
| -rw-r--r-- | src/armnn/test/UtilityTests.cpp | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/src/armnn/test/UtilityTests.cpp b/src/armnn/test/UtilityTests.cpp index b3b6c3bb3d..7911fd2608 100644 --- a/src/armnn/test/UtilityTests.cpp +++ b/src/armnn/test/UtilityTests.cpp @@ -202,7 +202,7 @@ TEST_CASE("NumericCast") CHECK_NOTHROW(numeric_cast<int8_t>(static_cast<float>(int8_max))); CHECK_NOTHROW(numeric_cast<int16_t>(static_cast<float>(int16_max))); - CHECK_NOTHROW(numeric_cast<int32_t>(static_cast<float>(int32_max))); + CHECK_NOTHROW(numeric_cast<int32_t>(static_cast<double>(int32_max))); CHECK_THROWS_AS(numeric_cast<int8_t>(float_max), std::bad_cast); CHECK_THROWS_AS(numeric_cast<int16_t>(float_max), std::bad_cast); |