Arm Cortex A510: two children in a trench coat

Arm series 5 double moves slowly. THE Cortex A53 And A55 each served through several generations of series 7 companions, And both were 2-width in order nuclei running has weak clocks. When your architecture East only expected has handle undemanding background Tasks, You don't do it need has push THE performance envelope. But as Arm evolved THE ARM instruction together, they necessary A little heart companion that could keep pace. THE Cortex A55 necessary A successor, And Arm took that opportunity has shake up THE series 5 programming.

THE resulting Cortex A510 present A stack of changes. He breaks THE 2-width in order formula by going 3 wide, but stay in order. Two A510 nuclei can be regroup And share A L2 hidden, FPU, And L2 TLB. Arm calls This A merged heart configuration, And hope How's it going increase area efficiency with minimal performance impact. Qualcomm has selected THE merged heart option In their Snapdragon 8+ Gen. 1, And I will be using that has test THE A510.

High Level

Cortex A510 East A 3-width in order heart. Cores can be regroup In pairs And share resources has to safeguard area, Or can be implemented separately. Arm offers A variety of configuration choice For heart hidden sizes And THE sharing FPU.

A example configuration watch. Grouping And hidden size configurations can differ

A510 has A eight scene pipeline a lot as THE A55, but THE pipeline layout has amended has accommodate THE wider decoders. Decoding NOW takes three pipeline steps instead of one.

Cortex A510 And A55 pipeline steps Since THE respective optimization guides

However, A510 can redirect THE pipeline Since THE First of all execution scene if A bifurcate East poorly predicted, while A55 could not TO DO SO until THE rewriting scene. SO, both nuclei to have A...