Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
marty1885
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
4 ms
·
1.
▲
by
marty1885
2y ago
They are moving to 22.04 and there's a push for 24.04. See https://github.com/tenstorrent/tt-metal/issues/10469
2.
▲
by
marty1885
2y ago
Don't we already do that with forwarding?
3.
▲
by
marty1885
3y ago
Yes. "Coarse-grain dataflow processor". It's like an FPGA. But LUTs are replaced with RISC-V cores.
4.
▲
by
marty1885
3y ago
Grayskull is make before LLMs being a thing. And their plan is like Groq, to distribute the compute graph across multiple processors to get higher effective memory and throughput by pipelining. But better-ish by having RAM so you can fit mo
5.
▲
by
marty1885
3y ago
Not fully, 8 bits has 256 values. It's easy to keep a look up table in the L1 cache of any CPU and constant cache of any GPU. For ASICs and FPGAs, it's a simple 256-value LUT. It's not ideal, yes, but not a deal breaker. Epic
6.
▲
by
marty1885
6y ago
I've wanted to play with large VLIW processors (Itanium is a disaster what we don't talk about. And there's litte support anyway). It's almost my dream. I hope this is real!