This is a subset of #68288, with hopefully narrower scope. It does not
support bitcasting to non-integral types yet.
Bitfields are supported, but only if they result in a full byte-sized
final buffer. It does not support casting from null-pointers yet or
casting from indeterminate bits.
The tests are from #68288 and partially from #74775.
The `BitcastBuffer` struct is currently always working in single bits,
but I plan to (try to) optimize this for the common full-byte case.
... tried their initializer already. In that case, diagnose the
non-const initializer instead of the reference to a non-constexpr
variable later. This is used in a lot of openmp tests.
According to [P0533R9](https://wg21.link/P0533R9), the C++ standard
library functions corresponding to the C macros in `[c.math.abs]` are
now `constexpr`.
To implement this feature in libc++, we must make the built-in abs
function `constexpr`. This patch adds the implementation of a
`constexpr` abs function for the current constant evaluator and the new
bytecode interpreter.
It is important to note that in 2's complement systems, the absolute
value of the most negative value is out of range. In gcc, it will result
in an out-of-range error and will not be evaluated as constants. We
follow the same approach here.
The global scope we create when evaluating expressions might free some
of the dynamic memory allocations, so we can't check for memory leaks
before destroying it.
This fixes all the places that hit the new assertion added in
https://github.com/llvm/llvm-project/pull/106524 in tests. That is,
cases where the value passed to the APInt constructor is not an N-bit
signed/unsigned integer, where N is the bit width and signedness is
determined by the isSigned flag.
The fixes either set the correct value for isSigned, set the
implicitTrunc flag, or perform more calculations inside APInt.
Note that the assertion is currently still disabled by default, so this
patch is mostly NFC.
```cpp
const int V33 = 4;
const int V34 = 0;
const int V35 = 2;
constexpr int V36 = V33 / V34;
// expected-error@-1 {{constexpr variable 'V36' must be initialized by a constant expression}}
constexpr int V37 = V33 / V35;
// expected-error@-1 {{constexpr variable 'V37' must be initialized by a constant expression}}
```
---------
Signed-off-by: yronglin <yronglin777@gmail.com>
Disable optimizations when building clang/lib/AST/ByteCode/Interp.cpp
with Microsoft's compiler as it has a bug that causes excessive build
times. We do this only when NDEBUG is not defined on the assumption that
building without asserts indicates that a user is strongly invested in
runtime performance.
Partially addresses: https://github.com/llvm/llvm-project/issues/102513.
Once the bug is addressed in the Microsoft compiler this can be removed.
Co-authored-by: dyung
