RISCV64 SoC Architecture diagram

[Pintu Agarwal]

Hi All,

For the education purpose, we need RISCV SoC internal architecture
diagram which can give a pictorial overview of RISCV internals.
As part of study, we want to compare this with ARMv8 architecture.

If anybody have it please share.

Also, if anybody have risv64 benchmarking data with arm64, please
share this as well.


Thanks,
Pintu

[Bruce Hoult]

RISC-V is an instruction set. It can be implemented by many different internal architectures. SiFive alone currently has three different microarchitectures that can each support 32 or 64 bit RISC-V with varying area vs performance trade-offs. Other organisations and individuals have developed many more.

The same applies to ARMv8. ARM themselves have I think nearly half a dozen different ARMv8 microarchitectures now, plus there are other independent implementations from Apple, Samsung, Qualcomm, NVidia, Cavium and probably others I can't think of now.

So, the question is meaningless.

You might want to compare some specific RISC-V design against some specific ARMv8 design. For example the SiFive U74 against the ARM A53 or A55,

You can see some (3rd party curated) information about these at:

https://en.wikichip.org/wiki/sifive/microarchitectures/7_series

https://en.wikichip.org/wiki/arm_holdings/microarchitectures/cortex-a53

--
You received this message because you are subscribed to the Google Groups "RISC-V SW Dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sw-dev+un...@groups.riscv.org.
To post to this group, send email to sw-...@groups.riscv.org.
Visit this group at https://groups.google.com/a/groups.riscv.org/group/sw-dev/.
To view this discussion on the web visit https://groups.google.com/a/groups.riscv.org/d/msgid/sw-dev/CAOuPNLiQWiqwkhKXyKV08mWz-Zfis8O3dHB0W5UVfroLU%3DW4Rw%40mail.gmail.com.

[MitchAlsup]

To reinforce Allen's point: I have a chapter in a book on computer architecture 

called pipelining. In this chapter I present a single data path of classic Stanford

RISC, and show how it can be pipelined into different implementations 7 different

ways. In all cases, it remained a 1-wide in order Scalar pipelined processor.

Now imagine doing a CDC Scoreboard version

Then Imagine doing a Tomasulo version

Then imagine 2-wide SuperScalar, 3-wide Superscalar

Then imagene a trace cache front end

Then imagine a vector back end

All of the above are orthogonal vectors of implementation.

Then, perhaps, you can see why one diagram is utterly useless in describing

some architecture which can be implemented in as many ways as you and your 

brother can think.

Mitch