Iain Buclaw

fast.math: Is inline assembler really faster than compiler builtin?

1

GDC feeds all overflow intrinsics to the __builtin_xxx functions. https://github.com/D-Programming-GDC/GDC/blob/16a15523c01e996660a0ad8b8043dd97d36535ee/gcc/d/d-codegen.cc#L3455-L3463

std.math.exponential: Implement log/log2/log10 for double- and float-precision

4

Software implementations are unchanged - only conditionally compiling in the `log(x) = z + z^^3 R(z) / S(z)` branch depending on the float format (as can be seen when reviewing...

std.complex: Add member function to returnComplex as a native type

5

So that `std.complex` can be passed to extern(C) complex functions where ABI compatibility is required.

std.math.exponential: Implement pow() for double precision

2

Double precision is enough for most cases, and results in [faster code than x87](https://d.godbolt.org/z/3E6n7ba4c). Not doing float precision, because existing tests don't account for an even higher loss of [precision](https://github.com/dlang/phobos/blob/ea805977656309e1d8f23b30130ecf8632056b1b/std/math/exponential.d#L147)....

std.math: Add RealRep union for accessing floating point bits via a union.

11

As I'd like to see this move forwards, but am in the middle of finishing up work on new ports (making sure that 128-bit is working properly for both IEEE...

[big feature] Implement core.math.pow() for ^^ operator

2

Adds two new functions for the power operator: ``` typeof(X * Y) pow(X, Y) if (IntOrFloat!X && Int!Y); typeof(X * Y) pow(X, Y) if (IntOrFloat!X && Float!Y); ``` Lowers `x...

Move usage fields from Param to DMDparam

9

These are driver-specific, rather than language tuning.

Move issue 23112 check to front-end semantic

1

This code should fail even when just checking syntax (`-o-`), not at the codegen stage.

fix Issue 22567, 22568, 22571 - Remove -target option from CLI

22

This option is not useful for DMD given its limited range of supported targets. The option itself has been made mostly redundant by the combination of `-mcpu=`, `-m32`, `-m64`, and...