1908f3ea29
Fixed minor compilation errors 2
5090538c64
Fixed minor compilation errors
61cd257385
Reimplemented non-trigonometric functions for increased numerical stability, mathematical accuracy, and precision (1e-11 on all functions, 1e-12 on all functions other than pow). Changes caused a...
4e986c7b70
Reimplemented entire common functions, added extra utilies, improved performance for some, improved numerical stability and precision
95e2410b02
Implemented error functions
b6023fd22d
Finished trig suite
2d0d84c8c4
Implemented trig and reimplemented inverse trig functions
bade3150a3
Added some raw timings for benchmarked functions
6c28909e43
Added explanation for major speedups.
f4f51fc71f
Fixed 'Getting Started' section, added testing information
5a0437cfab
Added splatting example
a712fe5a0b
Slight sanity changes to README
7092f60113
Added 'Getting Started' and 'Attribution' sections
7585b6f33f
Fixed N-Ary operator benchmark table.
ebed58ee6e
Added ILP variant flag to
hypot description.
8d96772235
Updated N-Ary operator content
10704a3fb5
Fixed N-ary operators and added
ilp variants which attempt to leverage instruction-level parallelism
be068f0f42
Added simple tests to ensure logic functions compile correctly
edefada05b
Updated Zigstd parity table to be more helpful
ca40c7522c
Added comment so that it is clear what
GenerateOps does, just in case.
c0a234bb8e
Fixed bitwise operator return types.
4b058253fc
Added logic operators, cleared up README a tiny bit
7cfcbe71ec
Removed redundant code and simplified repo structure. Enabled operations on arbitrary comptime-known vector lengths.
89b18ad5f6
Fixed a few markdown issues, addede explanation for the interpunct (I think that's what it's called)
f329b2d771
Added
f64x8, f64xN to README. Clarified a few other points.
5e9b6c603d
Simplified testing suite and enabled 512-bit / autodetected vector width testing
503522a798
Implemented operations for automatically-detected optimal SIMD vector widths
5bf8cac7c7
Implemented operations for AVX-512, or 512-bit wide SIMD vectors
70cf067e56
Added 512-bit and autodetected vector width options
4aa7d4eee6
Fixed
atan and asin implementations to pass the 1e-10 relative error threshold.
std.math.acosh of large negative numbers does not return nan
I'm not too familiar with the specifics of implementing this as part of the math library, as I started using Zig very recently, but is there a particular reason all the conditions need to be...