1. Marat Dukhan
  2. Yeppp

Source

Yeppp /

Filename Size Date modified Message
binaries/android/yeppp
bindings
build
codegen
docs
drivers
examples
installer/windows
library
runtime
tools
138 B
Repository: add .hgignore and .hgeol
925 B
Build: new build system based on Ninja + Python
59 B
Merged the tags mess
9.2 KB
Build instructions: require Python 2.7 for build
1.7 KB
License: moved license text to Yeppp! root directory
3.3 KB
Documentation: new PowerPC options
2.3 KB
Setup scripts: set variables CPATH and LIBRARY_PATH for gcc on Windows
16.1 KB
Build: new build system based on Ninja + Python
19.0 KB
Unix setup script: converted CR+LF into LF
4.3 KB
Windows: added yeppp.ico resource

Yeppp! F.A.Q

What is Yeppp!?

Yeppp! is a high-performance SIMD-optimized mathematical library for x86, ARM, PowerPC, and MIPS processors on Windows, Android, Mac OS X, and GNU/Linux systems. Yeppp! contains versions of its functions for multiple architectures and instruction sets and chooses the optimal implementation during initialization to guarantee the best performance on the host machine.

Which vector operations are supported?

Yeppp! aims to provide an orthogonal set of vector operations. By orthogonal we mean that Yeppp! supports all possible variations of operands. For example, for subtraction Yeppp! has 7 variants:

  • Subtraction of two arrays (z[i] := x[i] - y[i])
  • Subtraction of constant from array (z[i] := x[i] - c)
  • Subtraction of array from constant (z[i] := c - x[i])
  • In-place subtraction of two arrays, overwrite first array (x[i] := x[i] - y[i])
  • In-place subtraction of two arrays, overwrite second array (y[i] := x[i] - y[i])
  • In-place subtraction of constant from array (x[i] := x[i] - c)
  • In-place subtraction of array from constant (x[i] := c - x[i])

Moreover, there is a complete set of the above functions for different data types:

  • 8-bit, 16-bit, 32-bit, and 64-bit integers (signed and unsigned)
  • Single precision (32-bit) and double precision (64-bit) floating-point numbers

Besides basic arithmetic operations (addition, subtraction, multiplication) Yeppp! provides high-performance mathematical functions, such as log, exp, sin, which operate on vectors.

What other functionality is supported?

Yeppp! provides a number of functions to retrieve information about the processor (supported instruction extensions, architecture and microarchitecture, canonical processor name), measure processor cycles, access high-performance system timer, and processor energy counters.

Do I have to write in C++ to use Yeppp?

No. Yeppp! functions can be called from C, C++, C#, Java, FORTRAN.

Which architectures and operating systems are supported?

As of Yeppp! 1.0.1, the library supports 15 platforms:

  • Windows on x86 (32-bit) architecture
  • Windows on x86-64 (64-bit) architecture
  • Mac OS X on x86 (Intel 32 bit) architecture
  • Mac OS X on x86-64 (Intel 64 bit) architecture
  • Linux on x86 architecture
  • Linux on x86-64 architecture
  • Linux on armel ABI (ARMv5T architecture, Soft-Float EABI)
  • Linux on armhf ABI (ARMv7-A architecture, Hard-Float EABI)
  • Linux on k1om architecture (Xeon Phi)
  • Linux on PowerPC 64 architecture
  • Linux on Blue Gene/Q architecture
  • Android "armeabi" ABI (ARMv5TE architecture)
  • Android "armeabi-v7a" ABI (ARMv7-A architecture)
  • Android "x86" ABI
  • Android "mips" ABI

What is the license?

Yeppp! is licensed under the New BSD license (AKA 3-clause BSD), an OSI-approved permissive open source license. In particular, the license allows you to

  • Use Yeppp! in closed-source projects
  • Distribute Yeppp! as a part of your commercial product
  • Modify Yeppp! source code