D%(2n) == D & ((2n) - 1) (with 0 <= n <= 31 and 2^n means 2 to the power of n )
Smallest positive subnormal value (i.e. DBL_TRUE_MIN).
+
Smallest positive subnormal value (i.e. DBL_TRUE_MIN).
Linear interpolation, calculating x + t*(y - x) but avoids troublesome edge cases.
- Follows the C++20 specification.
+ Follows the C++20 specification.
In particular, if x.is-finitestd/num/float64/is-finite: (d : float64) -> bool && y.is-finitestd/num/float64/is-finite: (d : float64) -> bool, then:
- exact: lerpstd/num/float64/lerp: (x : float64, y : float64, t : float64) -> float64(x,y,0.0) == x and lerpstd/num/float64/lerp: (x : float64, y : float64, t : float64) -> float64(x,y,1.0) == y
- monotonic: if x <= y and t1 <= t2, then cmp( lerpstd/num/float64/lerp: (x : float64, y : float64, t : float64) -> float64(x,y,t1), lerpstd/num/float64/lerp: (x : float64, y : float64, t : float64) -> float64(x,y,t2) ) <= Eqstd/core/types/Eq: order (and other cases)
diff --git a/doc/std_num_int32-source.html b/doc/std_num_int32-source.html
index 48442d559..05d5bb197 100644
--- a/doc/std_num_int32-source.html
+++ b/doc/std_num_int32-source.html
@@ -438,7 +438,7 @@
Of course `(min-int32 + 1) / -1` is again positive (namely `max-int32`).
See also _Division and modulus for computer scientists, Daan Leijen, 2001_
-[pdf](http://research.microsoft.com/pubs/151917/divmodnote.pdf) .
+[pdf](https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/divmodnote-letter.pdf) .
*/
pub fip fun (/)std/num/int32/(/): (x : int32, y : int32) -> int32( xx: int32 : int32std/core/types/int32: V, yy: int32 : int32std/core/types/int32: V ) : int32std/core/types/int32: V
if (yy: int32 ==std/num/int32/(==): (int32, int32) -> bool 0.int32std/core/int32: (i : int) -> int32) then returnreturn: int32 0.int32std/core/int32: (i : int) -> int32
diff --git a/doc/std_num_int32.html b/doc/std_num_int32.html
index e1350eb76..22570ac1b 100644
--- a/doc/std_num_int32.html
+++ b/doc/std_num_int32.html
@@ -80,7 +80,7 @@ 32-bit signed integers.
Of course (min-int32std/num/int32/min-int32: int32 + 1) / -1 is again positive (namely max-int32std/num/int32/max-int32: int32).
See also Division and modulus for computer scientists, Daan Leijen, 2001 -pdf . +pdf .
(min-int64std/num/int64/min-int64: int64 + 1) / -1 is again positive (namely max-int64std/num/int64/max-int64: int64).
See also Division and modulus for computer scientists, Daan Leijen, 2001 -pdf . +pdf .
The backend compiler name. +
The backend compiler name, like gcc, clang, cl, clang-cl, mingw, or icc (and js for JavaScript).
Return the main processor architecture: x64, x86, arm64, arm, riscv32, riscv64, alpha, ppc64, etc. +
Return the main processor architecture: x64, x86, arm64, arm32, riscv32, riscv64, alpha64, ppc64, etc.
Return the size of boxed values in the heap (8*sizeof(kk_box_t)). This is usually
- equal to 8*sizeof(void*) but can be less if compressed pointers are used.
+ equal to 8*sizeof(void*) but can be less if compressed pointers are used (when
+ compiled with --target=c64c for example).
get-cpu-size-bitsstd/os/env/get-cpu-size-bits: () -> ndet int and get-cpu-pointer-bitsstd/os/env/get-cpu-pointer-bits: () -> ndet int on modern cpu's
but they can differ on segmented architectures.
- For example, on the old x86 FAR-NEAR model, the addresses are 32-bit but the maximum object size is 16-bit.
+ For example, on the old x86 FAR-NEAR model, the addresses are 32-bit but the integer register size is 16-bit.
Or on the more recent-x32 ABI
- the addresses and objects are 32-bits but the architecture has 64-bit registers.
+ the addresses and object sizes are 32-bits but the architecture has 64-bit integer registers.
Return the processor maximum object size in bits (8*sizeof(size_t)). This is usually
- equal to the get-cpu-arch-bits but may be different on segmented architectures.
+ equal to the get-cpu-int-bitsstd/os/env/get-cpu-int-bits: () -> ndet int but may be different on segmented architectures.
Return the main OS name: windows, linux, macos, unix, posix, ios, tvos, watchos, unknown.
- Sometimes has a dash subsystem, like: unix-
128-bit double-double floating point numbers.
to avoid magnifying rounding errors – in that case the result for IEEE 128-bit floats becomes: -The implementation is based closely on the excellent QD C++ library [4, 5], +
The implementation is based closely on the excellent QD C++ library [4, 5], and assumes proper 64-bit IEEE
float64std/core/types/float64: Vs with correct rounding. Integers can be represented precisely up to 30 decimal digits (and a bit more… up to 2106 - 2). @@ -128,20 +128,20 @@128-bit double-double floating point numbers.