It's a different thing for GPU. Compilers are not yet smart enough yet to do that level of vectorization.

I know gpu is much more difficult.
But if cpu extension (for example SSE2) is "so easy", why not?
R@H admins are not interested??

---

It's a different thing for GPU. Compilers are not yet smart enough yet to do that level of vectorization.

I know gpu is much more difficult.
But if cpu extension (for example SSE2) is "so easy", why not?
R@H admins are not interested??

Oh come on, give them a break! It’s not a case of the developers not being interested but the compiler writers not having worked out how to code the required optimisations.

If the compiler can’t do it the coders can’t either.

Oh come on, give them a break!

I don't want to disturb the admins. It's just a question.

If the compiler can’t do it the coders can’t either.

From Laurent message:

It was already available in most better compilers sometimes around 2000-2005

---

Oh come on, give them a break!

I don't want to disturb the admins. It's just a question.

If the compiler can’t do it the coders can’t either.

From Laurent message:

It was already available in most better compilers sometimes around 2000-2005

That was specifically compilers for CPUs, NOT compilers for GPUs. The point was made that GPU compilers did not have that ability.

That was specifically compilers for CPUs, NOT compilers for GPUs. The point was made that GPU compilers did not have that ability.

This thread is about cpu, not gpu.
And i cannot understand if some extensions are present (in linux app)

---

Interesting: CPP-Taskflow

---

This could help to optimize the code (from RosettaCommons paper):

The Rosetta codebase as of 2019 contains about 3.1 million lines of code. Although Rosetta can solve a variety of complex problems at the interface between biology, chemistry, physics and mathematics, we estimate that the codebase could be reduced by half without a significant loss of functionality

---

Linus Torvalds

I hope AVX512 dies a painful death, and that Intel starts fixing real problems instead of trying to create magic instructions to then create benchmarks that they can look good on.

We are still waiting for SSE2 on Rosetta, AVX512 is not a problem :-P

---

It's a different thing for GPU. Compilers are not yet smart enough yet to do that level of vectorization.

I know gpu is much more difficult.
But if cpu extension (for example SSE2) is "so easy", why not?
R@H admins are not interested??

They've said in the past that it won't fit into the memory of the gpu making it take forever as it has to go back and forth to the cpu to get the data it needs, now that gpu's have more than 2 or 3 gb of ram they may have to relook at that BUT more doesn't always mean bigger.

They've said in the past that it won't fit into the memory of the gpu making it take forever as it has to go back and forth to the cpu to get the data it needs, now that gpu's have more than 2 or 3 gb of ram they may have to relook at that BUT more doesn't always mean bigger.

Now 2 or 3 gb of ram is an entry level gpu at less than 100$.
I think we will never see a gpu version of R@H in our pc (and it's a pity).

But this thread is dedicated to cpu optimiziations.

---

AMD releases AOCC 2.2

---

C++20 Draft approved as major upgrade.

C++20 adds to the language concepts, modules, the "spaceship operator" for three-way comparisons, coroutines, designated initializers, new standard attributes, and much more. The C++20 library standard also adds ranges, feature test macros, bit operations, and more.

---

They've said in the past that it won't fit into the memory of the gpu making it take forever as it has to go back and forth to the cpu to get the data it needs, now that gpu's have more than 2 or 3 gb of ram they may have to relook at that BUT more doesn't always mean bigger.

Now 2 or 3 gb of ram is an entry level gpu at less than 100$.
I think we will never see a gpu version of R@H in our pc (and it's a pity).

But this thread is dedicated to cpu optimiziations.

YES but optimizations can lead to other apps as well.

But this thread is dedicated to cpu optimiziations.

YES but optimizations can lead to other apps as well.

What do you mean?

---

But this thread is dedicated to cpu optimiziations.

YES but optimizations can lead to other apps as well.

What do you mean?

Even though they compute differently they both compute, so if you find a way to speed up cpu computing maybe it can start the process of figuring out how to do it on gpu's too. NO not every idea will lead down that path, but the next idea might, it just takes the right idea to spark the right person to think 'what if' and BOOM things change.

Even though they compute differently they both compute, so if you find a way to speed up cpu computing maybe it can start the process of figuring out how to do it on gpu's too. NO not every idea will lead down that path, but the next idea might, it just takes the right idea to spark the right person to think 'what if' and BOOM things change.

All these years, all these discussions....i think we will never have a gpu app.
But i'll continue to update these discussions :-)

---

Even though they compute differently they both compute, so if you find a way to speed up cpu computing maybe it can start the process of figuring out how to do it on gpu's too. NO not every idea will lead down that path, but the next idea might, it just takes the right idea to spark the right person to think 'what if' and BOOM things change.

All these years, all these discussions....i think we will never have a gpu app.
But i'll continue to update these discussions :-)

Thank you...hope springs eternal!!!
Without Sci-Fi writers would be ever have cell phones that fit in our pockets?

I'm curious to see performances on SSE/AVX with new Ryzen 5xxx.
Seems to see monster cpus!!

---

I'm curious to see performances on SSE/AVX with new Ryzen 5xxx.
Seems to see monster cpus!!

Supposedly it also does a better job of confing tasks to it's own CCX too if the cpu has more than one and if you adjust your settings correctly and run multi-threaded tasks.

Supposedly it also does a better job of confing tasks to it's own CCX too if the cpu has more than one and if you adjust your settings correctly and run multi-threaded tasks.

It's a pity that we cannot take full advantage (using cpu extensions) of these little monsters.

---