Skip to content
Go to file

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time


Ternary functions for SSE and AVX2


AVX512F introduced instruction vpternlog (intrinsic _mm512_ternarylogic_epi{32,64}), which evaluates a three-argument boolean function for each bit of operands. The function is given as an integer number from 0 to 255.

This repository contains a kind of library which allows a programmer to use ternary functions in SSE, XOP, AVX2 and x86 in similar manner, by choosing the function by its number.

In native AVX512F code you would write:

// 0x96 = A xor B xor C
const __m512i res = _mm512_ternarylogic_epi32(A, B, C, 0x96);

With help of the library you can write an AVX2 program:

const __m256i res = ternarylogic::avx2::ternary<0x96>(A, B, C);

or a SSE program:

const __m128i res = ternarylogic::sse::ternary<0x96>(A, B, C);


The function number comes from the result column of the function's truth table. The number is formed from bits a, b, ... h. You can find more details and examples in my article.

A B C function
0 0 0 a
0 0 1 b
0 1 0 c
0 1 1 d
1 0 0 e
1 0 1 f
1 1 0 g
1 1 1 h

Intel has published list of all functions in form of logical expression in their monumental document "Intel® 64 and IA-32 Architectures Software Developer’s Manual Volume 2 (2A, 2B, 2C & 2D): Instruction Set Reference, A-Z".

I extracted these functions and created application which generates SSE and AVX2 code. List of function is in the file py/data/intel.txt, I also put there more details. Later I manually optimized some of these function. And in the end I developed a program which found many better representations of functions for both Intel and AMD flavours of SIMD ISA.


You need python and make.

Type make, then following files will be generated:

  • ternary_sse.cpp,
  • ternary_avx2.cpp,
  • ternary_avx512.cpp (use only two-argument logic instructions),
  • ternary_xop.cpp,
  • ternary_x86_32.cpp,
  • ternary_x86_64.cpp.

You can include them directly into your application.

Programs validate_sse, validate_avx2 and validate_xop test if all generated functions are correct.

See also


Support for ternary logic in SSE, XOP, AVX2 and x86 programs




No releases published
You can’t perform that action at this time.