Tuesday @ 1130 ISA Shootout – a Comparison of RISC V, ARM, and x86 Chris Celio, UC Berkeley V2

Channel:RISC-V International · Length:32:08 · Published:2016-07-26 · Views:50178

Watch on YouTube

Summary

Chris Celio of UC Berkeley compares the RISC-V, ARM and x86 instruction sets, with quantitative data on instruction counts and code density. It examines how the compressed extension fares against Thumb and x86.

Why HN recommends it

Cited as primary-source evidence that RISC-V code density is competitive with or better than ARM and x86.

You should watch this video — HN, 2016

Computer architecture & CPUs

Subtitles (excerpt)

my name is chriso I'm a graduate student UC Berkeley advised by christovich and David Patterson and today I'm going to be comparing a handful different instruction set architectures against one another using the industry standard spec Benchmark suite and let's just see what happens um this is um a companion to a tech report that we just released uh where based on this data that I'm going to present uh we're going to make a renewed case for reduced instruction set computers um and uh if you like my talk or you want to see more data please go and give that a read uh and if you don't like my talk you should still read it anyway uh I do want to Define my terms today I'm talking about comparing instruction set architectures isas that's the boundary between the software and the hardware um I like to think of that as the language that the processor speaks um to make That explicit if you have some source code for an application you care about like bzip you tell your C compiler please target the risk 5 RV 64g instruction set architecture and once you have that binary any processor that can speak RV 64g can run that binary whether it's a single issue norder core like rocket or the super SK core like boom uh the user w't actually be

On Hacker News

> If you look at functionality per instruction byte, RISC is fairly low. You should watch this video: [VIDEO] The most important information therein is that the difference in code density is minuscule between x86, ARM and RISC-V. The differences between GCC versions is even bigger. > The equations that you have to resolve to issue an X86 instruction are simply more complex than for ARM, and all t… — legulere, 2016-08
At best, that's a simplistic and misleading view, see the following talk for an in depth analysis: "Comparison of RISC V, ARM, and x86" [VIDEO] — dkjaudyeqooe, 2020-12
As of early 2016, with the GCC port at that time, RV32GC was as dense as Thumb, and RV64GC was denser than AArch64 and every other major 64-bit ISA, including AMD64. Though RV64G (no C) was in some extreme cases up to 50% larger than AArch64 (due to inlining memcpy and memset, which are a bit larger without compressed instructions), but usually around the same (except MIPS64, which is way larger … — microcolonel, 2019-02
Code density is claimed to be similar to x86 and better than ARM with thumb, see the bottom of this slide: [VIDEO] More information on the original design (not quite the final design) of the Compressed extension can be found in the original research paper: [LINK] And ... SPARC. Seriously? Register windows were a failed experiment. — rwmj, 2018-11
> poor vector instruction implementations All I have heard is that the people who know about vectors absolutely love the extension. Why is your problem with it? > bad code density That is just plain false. Its actually extremely good, RV32C and RV64C easily beat every other ISA. Even without the 'C' extension the difference to the other ISA are minimal. See: [VIDEO] > The other general point I re… — nickik, 2018-07

Related videos