FMA instruction set

FMA x86 refers to a set of instructions for microprocessors from Intel and AMD to support the Fused Multiply - Add technique ( FMA). This the first time there was a reaction in the " Bulldozer" CPUs from AMD. Intel announced an implementation in future processors.

There are two non- compatible versions, FMA4 and FMA3:

  • FMA4 the full version which allows an operation of the form;
  • FMA3 requires, however, that the destination register is one of the operand registers, and this will be overwritten.

New Features

FMA extends the possibilities and vector operations can be seen as an extension of the AVX commands.

New instructions

CPUs with FMA4

  • Intel Z. Z. is still unclear whether Intel adjoins FMA4 or will stay with FMA3.
  • Bulldozer - based processors (AMD FX), Q4/2011
  • AMD Piledriver - based processors (AMD FX, Trinity and Richland APUs ), Q2/2012
  • AMD Steamroller - based processors (4th generation A- series processors, Kaveri APUs ) Q1/2014

CPUs with FMA3

  • Intel Haswell - based processors (4th generation Core i processors)
  • AMD Piledriver - based processors (2nd generation FX CPUs, Trinity and Richland APUs ), Q2/2012
  • AMD Steamroller - based processors (4th generation A- series processors, Kaveri APUs ) Q1/2014

Application

  • Gleitkommaintensive useful for calculation, particularly in multi-media, scientific and financial calculations. Integer operations are to follow later.
  • Increases parallelism and throughput of floating point SIMD calculations
  • Reduces the register load by non-destructive four- operand form ( in the case of FMA4 )

CPUs with x86 FMA

  • Intel Processors with Haswell architecture
  • AMD: Bulldozer processors since Q4 2011.
  • Piledriver processors since Q2 of 2012.
  • Steamroller processors since Q1 2014
Multiply–accumulate operation#Fused multiply.E2.80.93add Bulldozer (microarchitecture) Haswell (microarchitecture) Advanced Vector Extensions
340677
de