intrinsics for scatter/gather

intrinsics for scatter/gather

there is a discussion on the AVX forum about the needs for scattter/gather instruction in the ISA

http://software.intel.com/en-us/forums/showthread.php?t=79768&o=a&s=lr

in my opinionthere is no reason to wait for an hardware implementationbeforeto start writing code using such constructs, the most important concerns being to have a clean maintenable source code and a guarenteed backward compatibility in future (hopefully faster) implementations

soI supposethata nice new feature for some upcoming release ofIntel C++will beto provide scatter/gather intrinsicsgeneratingoptimized sequences of instructions, much like _mm256_set_ps but with a base address and 8 packedindices in a __m256i argument instead of the 8 floats with_mm256_set_ps

1 post / 0 new
For more complete information about compiler optimizations, see our Optimization Notice.