undefined | Better HN

0 pointssteelframe1y ago0 comments

I think Apple Silicon has shown us that x86 doesn't have the monopoly potential it once had.

0 comments

5 comments · 1 top-level

alganet1y ago· 4 in thread

Apple Silicon was designed to be efficient at emulating x86-64.

If you take that away, it becomes irrelevant (like many other ARM-based processors that struggle to be a good product because of compatibility).

Apple has a promising path for x86 liberation, but it is not there yet.

alganet1y ago

> One of the key reasons why Rosetta 2 provides such a high level of translation efficiency is the support of x86-64 memory ordering in the Apple M1 SOC.

https://www.sciencedirect.com/science/article/pii/S138376212...

ryao1y ago

TSO might be slower than ARM’s memory ordering, but implementing it in hardware is considered faster than implementing it in software.

astrange1y ago

It was designed to be efficient (performance/watt) for everything. Most users aren't using Rosetta.

It's not the only ARM CPU with TSO support though, some server platforms also do it.

alganet1y ago

"efficient for everything": that is a bold claim, can you elaborate more on it?

Also, can you name these other platforms?

Also, can you back up the information about most users not leveraging Rosetta?

1 more reply

j / k navigate · click thread line to collapse

0 comments

5 comments · 1 top-level

alganet1y ago· 4 in thread

Apple Silicon was designed to be efficient at emulating x86-64.

If you take that away, it becomes irrelevant (like many other ARM-based processors that struggle to be a good product because of compatibility).

Apple has a promising path for x86 liberation, but it is not there yet.

alganet1y ago

> One of the key reasons why Rosetta 2 provides such a high level of translation efficiency is the support of x86-64 memory ordering in the Apple M1 SOC.

https://www.sciencedirect.com/science/article/pii/S138376212...

ryao1y ago

TSO might be slower than ARM’s memory ordering, but implementing it in hardware is considered faster than implementing it in software.

astrange1y ago

It was designed to be efficient (performance/watt) for everything. Most users aren't using Rosetta.

It's not the only ARM CPU with TSO support though, some server platforms also do it.

alganet1y ago

"efficient for everything": that is a bold claim, can you elaborate more on it?

Also, can you name these other platforms?

Also, can you back up the information about most users not leveraging Rosetta?

1 more reply

j / k navigate · click thread line to collapse