Documentation

A weaker notion of equality of CmmTypes than cmmEqType, used (only) in Cmm Lint.

Why "weaker"? Because:

• we don't distinguish GcPtr vs NonGcPtr, because the the RTS files are not yet well-typed wrt pointers,
• for vectors, we only compare the widths, because in practice things like X86 xmm registers support different types of data (e.g. 4xf32, 2xf64, 2xu64 etc).

A width in bits.

A width in bytes.

widthFromBytes (widthInBytes w) === w

log_2 of the width in bytes, useful for generating shifts.

• Partial* A width from the number of bytes.

The width of the current platform's word size.

The width of the current platform's half-word size.

A bit-mask for the lower half-word of current platform.

Narrow a signed or unsigned value to the given width. The result will reside in [0, +2^width).

>>> narrowU W8 256    == 256
>>> narrowU W8 255    == 255
>>> narrowU W8 128    == 128
>>> narrowU W8 127    == 127
>>> narrowU W8 0      == 0
>>> narrowU W8 (-127) == 129
>>> narrowU W8 (-128) == 128
>>> narrowU W8 (-129) == 127
>>> narrowU W8 (-255) == 1
>>> narrowU W8 (-256) == 0

Narrow a signed value to the given width. The result will reside in [-2^(width-1), +2^(width-1)).

>>> narrowS W8 256    == 0
>>> narrowS W8 255    == -1
>>> narrowS W8 128    == -128
>>> narrowS W8 127    == 127
>>> narrowS W8 0      == 0
>>> narrowS W8 (-127) == -127
>>> narrowS W8 (-128) == -128
>>> narrowS W8 (-129) == 127
>>> narrowS W8 (-255) == 1
>>> narrowS W8 (-256) == 0

is -falignment-sanitisation enabled?