Editing Emotion Engine (section)

=== Vector processing units ===
The majority of the Emotion Engine's [[floating point]] performance is provided by two [[Vector processor|vector processing]] units (VPU), designated VPU0 and VPU1. These were essentially [[Digital signal processor|DSP]]s tailored for 3D math, and the forerunner to [[Shader#vertex shader|hardware vertex shader pipelines]]. Each VPU features 32&nbsp;[[128-bit]] vector SIMD [[Processor registers|registers]] (holding [[4D vector]] data), 16&nbsp;16-bit fixed-point registers, four [[floating point multiply-accumulate]] (FMAC) units, a floating point divide (FDIV) unit and a [[scratchpad memory|local data memory]]. The data memory for VPU0 is 4&nbsp;KB in size, while VPU1 features a 16&nbsp;KB data memory.

To achieve high bandwidth, the VPU's data memory is connected directly to the GIF, and both of the data memories can be read directly by the [[Direct memory access|DMA]] unit. A single vector instruction consists of four 32-bit [[single-precision]] floating-point values which are distributed to the four single-precision (32-bit) FMAC units for processing. This scheme is similar to the [[Streaming SIMD Extensions|SSEx]] extensions by Intel.

The FMAC units take four cycles to execute one instruction, but as the units have a six-stage [[Pipeline (computing)|pipeline]], they have a throughput of one instruction per cycle. The FDIV unit has a nine-stage pipeline and can execute one instruction every seven cycles.