L1 Terminal Fault - The Latest Speculative Execution Side Channel Attack

[schmidtbag]

starshipeleven
I see you're up to your usual antics of:
* Taking things too literally
* Turning the situation into a personal matter
* Going way too aggressive with your criticisms
* Being needlessly antagonistic over what could have been an interesting and intelligent discussion

I don't know enough about the topic to say who is right or wrong, but even if cybertraveler is wrong, he/she doesn't deserve your treatment. As you can usually be summarized:

[starshipeleven]

Being needlessly antagonistic over what could have been an interesting and intelligent discussion

It can't be an interesting and intelligent discussion if it's based on wild speculation.

I'm sick and tired of bullshit conspiracy theories, it's all noise I have to wad through to get at the info I seek.

I know that quite a bit of it is even spread maliciously as a smokescreen.

he/she doesn't deserve your treatment.

Your opinion has been noted, thank you for sharing it.

[schmidtbag]

It can't be an interesting and intelligent discussion if it's based on wild speculation.

Sure it can, but what makes something interesting is relative. If it doesn't interest you, don't join in the discussion. It's as simple as that.

I'm sick and tired of bullshit conspiracy theories, it's all noise I have to wad through to get at the info I seek.

You seem to have no problem getting the info you seek. So although I can admit wading through the BS is annoying, I don't consider this a valid argument on your part; your antagonism effectively contributes toward it.

I know that quite a bit of it is even spread maliciously as a smokescreen.

It's fine to dispute such things, but again, you don't have to be such an asshole about it.

Your opinion has been noted, thank you for sharing it.

Sarcasm or otherwise, it's a shame you won't do anything about it. You're easily the most toxic person on these forums who isn't a troll.

[Weasel]

It can't be an interesting and intelligent discussion if it's based on wild speculation.

I'm sick and tired of bullshit conspiracy theories, it's all noise I have to wad through to get at the info I seek.

I know that quite a bit of it is even spread maliciously as a smokescreen.

Your opinion has been noted, thank you for sharing it.

Relevant: https://en.wikipedia.org/wiki/I%27m_..._to_my_opinion (to support your argument)

[schmidtbag]

Relevant: https://en.wikipedia.org/wiki/I%27m_..._to_my_opinion (to support your argument)

I don't see the relevancy here, or how specifically that supports him.
The gist of Wikipedia article is describing someone who is trying to dispute facts by claiming an opinion; it's a rejection of logic and/or facts. In the context you've quoted, that does not apply.

[starshipeleven]

Sure it can, but what makes something interesting is relative. If it doesn't interest you, don't join in the discussion. It's as simple as that.

I react only on things I'm interested in, of course.

You seem to have no problem getting the info you seek.

I'm sorry to inform that your completely baseless opinion is wrong. Seriously WTF? How can you say that? You know nothing about me, John Snow.

So although I can admit wading through the BS is annoying, I don't consider this a valid argument on your part; your antagonism effectively contributes toward it.

It terminated this "discussion", and threw some useful info out there for him and others reading. That's the best I can do for a stranger over the internet.

It's fine to dispute such things, but again, you don't have to be such an asshole about it.

I attacked only the statements, and pointed out the mistakes he is making in his reasoning, Of course I'm just a man so I can't provide a 100% emotionless reaction to things.

Sarcasm or otherwise, it's a shame you won't do anything about it. You're easily the most toxic person on these forums who isn't a troll.

Instructions unclear, process terminated. I need more in-depth instructions than "do something about it".

I'll have to inform you that you are doing all this for yourself, to think you are good and all, not really to help people.

I do my best to explain how to fix the issues I'm pointing out, with all limits and flaws I can have.
They don't want to read, fine.
At least I'm not saying "you are wrong, you should do better, hurr durr", which serves no other purpose than checking a checkbox on your "good person" list, and helps none.

I don't see the relevancy here, or how specifically that supports him.
The gist of Wikipedia article is describing someone who is trying to dispute facts by claiming an opinion; it's a rejection of logic and/or facts. In the context you've quoted, that does not apply.

I have to admit that I'm not 100% sure here, but I would read that as "Weasel agrees that you are mostly using opinions to dispute facts".

Which imho you keep doing even now (see above)

[schmidtbag]

I react only on things I'm interested in, of course.

Except... you just said earlier "It can't be an interesting and intelligent discussion if it's based on wild speculation" meaning you are responding to these people and arguing with them (so, not holding a friendly discussion) specifically because it doesn't interest you. A bit hypocritical, eh?

I'm sorry to inform that your completely baseless opinion is wrong. Seriously WTF? How can you say that? You know nothing about me, John Snow.
It terminated this "discussion", and threw some useful info out there for him and others reading. That's the best I can do for a stranger over the internet.

"blah blah blah baseless opinion blah blah blah <insert over-used meme>"
It's quite easy for me to say that - whenever you are in one of these arguments, you apparently know everything so it's blatantly obvious that the "BS" that you have to sift through isn't really getting in your way of real information.

I attacked only the statements, and pointed out the mistakes he is making in his reasoning, Of course I'm just a man so I can't provide a 100% emotionless reaction to things.

You are also an asshole who takes things way too seriously and approaches them with too much hostility. That is not inherent of being an emotional man.
EDIT:
Despite being "just a man", everyone else here is just people too. People make mistakes, and you are very unforgiving of them.

Instructions unclear, process terminated. I need more in-depth instructions than "do something about it".

The instructions are actually quite clear: be more polite and less aggressive when correcting people. It's really not hard - most people on these forums do a great job at that.

I'll have to inform you that you are doing all this for yourself, to think you are good and all, not really to help people.

If anything is to be deemed unclear here, it's whatever you meant by that sentence.

I do my best to explain how to fix the issues I'm pointing out, with all limits and flaws I can have.
They don't want to read, fine.

And often, you are actually pretty good at explaining. But explaining doesn't require condescension, insults, and an all-around dismissive or "all-knowing" attitude. Like I said, I know you're not a troll because very often you're right - you're just an asshole. I'm tired of it and I know I'm not the only one. If you can't politely inform someone how they may be wrong without attacking their intellect, then you might as well just move on. As I've told you before, nobody cares if you're right if you piss them off, so what do you gain if the information you provide goes ignored?

At least I'm not saying "you are wrong, you should do better, hurr durr", which serves no other purpose than checking a checkbox on your "good person" list, and helps none.

You're right - instead, you say equally as useless, unproductive, and/or antagonistic things such as:
* "baseless comment"
* "wild speculations"
* "why you seem to support all bullshit tinfoil theories. The most likely reason is stupidity"
* "I'm a truthseeker, please. I don't think freely, I search truth. Thinking freely makes you fall into the pits of your own mind, your main enemy is yourself."
* " I have no way of knowing why you post bullshit."
All of these were from the past 2 pages. I don't think I need to keep going, that's plenty of proof right there.

I have to admit that I'm not 100% sure here, but I would read that as "Weasel agrees that you are mostly using opinions to dispute facts".

Except in this argument with me, I'm the one with the facts here - it is indisputable that you are frequently needlessly antagonistic (case in point - the quotes above). Everything else that I said that's opinionated you have also replied to with opinions, thereby removing any significance to Weasel's comment (again, in this context).

[starshipeleven]

Except... you just said earlier "It can't be an interesting and intelligent discussion if it's based on wild speculation" meaning you are responding to these people and arguing with them (so, not holding a friendly discussion) specifically because it doesn't interest you. A bit hypocritical, eh?

The topic they are discussing ("why there are these flaws in Intel processors") interests me. Are we done with word-based chess? Or are you just trying to bore me to death so I drop and go away?

It's quite easy for me to say that - whenever you are in one of these arguments, you apparently know everything so it's blatantly obvious that the "BS" that you have to sift through isn't really getting in your way of real information.

Knowing when to pick fights and posturing does not prove that I know all I wanted.

You are also an asshole who takes things way too seriously and approaches them with too much hostility. That is not inherent of being an emotional man.

Yeah, you aren't good at psychology. Is anger not an emotion? Seriously?

The instructions are actually quite clear: be more polite and less aggressive when correcting people. It's really not hard - most people on these forums do a great job at that.

No, that's the effect. You are telling me "make a great lemon cake", not how to do it.

If anything is to be deemed unclear here, it's whatever you meant by that sentence.

You engaged in this discussion just because it makes you feel "good", but you aren't actually helping me.

Like the "thoughts and prayers" thing, or "wishing good luck". Makes the person feel good, but isn't actually helping.

If you can't politely inform someone how they may be wrong without attacking their intellect, then you might as well just move on. As I've told you before, nobody cares if you're right if you piss them off, so what do you gain if the information you provide goes unnoticed?

I'm tired of telling you the same fucking answer to the same fucking questions, I'm starting to think you truly are some kind of stateless bot:
When I go in like this it's because I'm not even hoping the person will change his mind, but I'm mostly shutting him up or poking holes in his statements so that others can see that for what it truly is.

It's also fun to poke holes in things.

You're right - instead, you say equally as useless, unproductive, and/or antagonistic things such as:

I repeat, I have flaws but I did what I could to explain things too in the same posts you took these quotes from. You would see it if you didn't take them out of their sentence.

You don't. You ask over and over the same questions and tell people to "be a good boy" as if someone that is misbehaving could just flip a switch and be good. People don't work like that.

Except in this argument with me, I'm the one with the facts here - it is indisputable that you are frequently needlessly antagonistic (case in point - the quotes above). Everything else that I said that's opinionated you have also replied with opinions, thereby removing any significance to, Weasel's comment.

Sorry what?

I react only on things I'm interested in, of course. FACT, if you didn't understand the first time it's your own issue.

I'm sorry to inform that your completely baseless opinion is wrong. FACT, yes I happen to know better than you what I would like to know and I don't because there is so much noise, thank you very much.

I attacked only the statements, and pointed out the mistakes he is making in his reasoning, Of course I'm just a man so I can't provide a 100% emotionless reaction to things. FACT, and you also kinda acknowledge this when you say that I'm good at explaining but I'm an asshole.

Instructions unclear, process terminated. I need more in-depth instructions than "do something about it". FACT, "be good", "do something" and such are not in-depth instructions, they at best describe the goal.

At least I'm not saying "you are wrong, you should do better, hurr durr", which serves no other purpose than checking a checkbox on your "good person" list, and helps none. FACT, same as above. Not providing meaningful assistence, that's not helping. I can concede the second part, the purpose, I can't say with 100% certainty, but it is the classical reason people do it.


-----

To be fair, I'm confident you will not understand a damn of what I wrote and keep repeating yourself over and over. That's why I'm not really guaranteeing I'll feel like answering you again, this time anyway. But don't be sad, I'll repeat all I said here some time in the future when we have another chat exactly like this.

[schmidtbag]

The topic they are discussing ("why there are these flaws in Intel processors") interests me. Are we done with word-based chess? Or are you just trying to bore me to death so I drop and go away?

Moving goalposts eh? Your favorite past-time.

Yeah, you aren't good at psychology. Is anger not an emotion? Seriously?

What does anger have anything to do with what I said? It's not hard to control your anger - most people do it on a regular basis.

No, that's the effect. You are telling me "make a great lemon cake", not how to do it.

I'm telling you to make a great lemon cake because you keep talking shit about everyone else's pastries. If you are so certain about what you have to say, it shouldn't be that hard to concisely and politely make your point (or your own cake).

You engaged in this discussion just because it makes you feel "good", but you aren't actually helping me.

Baseless claims!!!! You know nothing!
Despite me saying this is true in this context, see how annoying and useless that is to say?
I engaged in this conversation in the hopes that you stop doing what you do in the future. There's nothing that feels good about going into community forums only to see you derailing the thread.
For almost a full year, you've actually been doing pretty good. You were politely making concise points that people actually listened to and respected (and for the ones who retaliated with hostility, you, understandably, reacted accordingly). But basically since summer started, you've just completely turned around to your old annoying self. I think you need to just take a breath once in a while.

I'm tired of telling you the same fucking answer to the same fucking questions, I'm starting to think you truly are some kind of stateless bot:
When I go in like this it's because I'm not even hoping the person will change his mind, but I'm mostly shutting him up or poking holes in his statements so that others can see that for what it truly is.

Except you don't ever answer the same fucking questions because you know damn well that isn't actually the answer, because it so rarely works the way you expect it to. You don't succeed in shutting people up, you don't succeed in poking holes (or are you too self-centered to realize that the only ones who cares what you have to say is yourself?), and you derail the thread so much that nobody else is reading what you have to say. For example, you know very well that I'm stubborn, and more often than not, I'm the one who gets the last word. For example: when was the last time you succeeded in getting me to shut up? And yet, you keep going, just like right now.
So I will ask yet again - what do you hope to accomplish? Because I don't believe you're stupid enough to believe what you just told me.

I repeat, I have flaws but I did what I could to explain things too in the same posts you took these quotes from. You would see it if you didn't take them out of their sentence.

No, you didn't. You started out explaining something in a condescending way, you got a fitting response, and then you escalated the situation from there. Insults and condescension are not a requirement of explaining things.

and tell people to "be a good boy" as if someone that is misbehaving could just flip a switch and be good. People don't work like that.

Actually, people with a fully developed prefrontal cortex do a very good job of that, as can be exemplified by the vast majority of forum members in their mid 20s and later. Don't try to make your behaviors out to be normal - even if your answers to my "fucking questions" are accurate, most people don't think that way. Most people aren't that malicious and understand how to maintain a proper social life. Most people who explain things are trying to help, even if poorly.

I react only on things I'm interested in, of course. FACT, if you didn't understand the first time it's your own issue.

Actually, it's not my problem - this is a failure on your part. You directed that explanation toward me, in response to something I said that bothered you. You failed to explain yourself and it is you who has the interest in defending what you deem is a fact. I couldn't care less what's a fact in this context - all I care about is getting you to be less of a pain in the ass in future discussions.

I'm sorry to inform that your completely baseless opinion is wrong. FACT, yes I happen to know better than you what I would like to know and I don't because there is so much noise, thank you very much.

So, I guess I just have to take your word for it? As though it bears any significance, or as though you have any credentials worth noting? You're nothing more than just a random neckbeard on the internet. Saying "baseless claims" and treating your own as fact because it's all in YOUR head doesn't mean a damn thing.

I attacked only the statements, and pointed out the mistakes he is making in his reasoning, Of course I'm just a man so I can't provide a 100% emotionless reaction to things. FACT, and you also kinda acknowledge this when you say that I'm good at explaining but I'm an asshole.

Take note how I never said you were wrong. I'm fine with acknowledging it; doesn't change my point.
As I've already implied, I'm not here to prove you wrong about anything except how your approaches to handling people accomplishes nothing, not even the "goals" that you claim to have.

To be fair, I'm confident you will not understand a damn of what I wrote and keep repeating yourself over and over. That's why I'm not really guaranteeing I'll feel like answering you again, this time anyway. But don't be sad, I'll repeat all I said here some time in the future when we have another chat exactly like this.

Could it be that I do understand, but your reasons for your behavior are delusional? If you're so confident we're going to do this again, why not just get this over with?

[oiaohm]

Which is completely useless for general purpose computing and that is the reason you won't find these CPUs there.

I wonder if you guys realize that supercomputers execute stupidly simple code, but a lot of it. They don't execute complex code with thousands of branches and other complexities: they execute simple equations and brute force calculate them in parallel. Yet a lot of people think that supercomputers, somehow, must be extremely complex beasts and executing stuff that's so complicated for the feeble desktop minds. In fact a high-end desktop CPU is likely to be faster than a supercomputer's CPU at a single-threaded task and that's the kind of tasks you need for general purpose computing.

Which is a completely different thing than single-threaded performance, and that's what you need for general purpose computing. Otherwise, just use a GPU for general purpose computing and see where it gets you?

Speculative execution exists to increase SINGLE THREADED performance. Stop this bullshit with 8 threads per core, that is NOT SINGLE THREADED performance so it's NOT an option. Literally comparing apples with oranges.

Really read before posting.

This is different the a cray barrel engine/thread management core the big difference between what IBM did and what risc-v prototype does is that you in fact have thread priorities in the risc-v one just like the cray barrel. So higher priority threads in the risc-v prototype gets faster access back to cpu for processing once the issue that caused them to stall has passed.
This is a very different beast to what IBM did in power with hyperthreadingx8.

Barrel multi threading means if you have only 1 thread you cpu performs like its the inorder version of the cpu core. Yes performs like the inorder version of the cpu core without OS present there is no scheduler overhead. Why having a thread management engine in the cpu means you don't have to perform preventive multi tasking in the OS scheduler instead the cpu is taken care of this transparently.

Inorder with multi instruction processing is not that slow particularly once you start considering how heavy the overheads dealing with speculative issues in software. Please note some of the fixes to speculative is basically force sections of code back to in order processing so those areas of your operating system you are now not getting any gain from speculative and in fact wasting cpu time telling the cpu not to be speculative. .

You general OS workload running single thread application is not that single threaded either. With speculative designed cpus you will be performing context switches between the OS and the application as part of thread management.

Think hardware event has come in while you are running your single threaded task this can be a disruption with context switches in your x86 as the cpu core is taken for kernel usage. Something with a thread management engine in hardware it is different being hardware event just sees a high priority thread woken up you single thread can still keep on progressing forward slower filling in what ever the hardware event is not using in the cpu.

So yes speculative gets single threading faster in some cases. The problem is modern day operating systems spend very little time in single thread processing. Providing hardware that removed overhead from being multi threaded moves a lot of things .

A cpu with barrel style thread management engine with a inorder read and process multi instructions does not take many threads to catch up to a cpu speculative without issues it about 4 threads worse is speculative with issues is about 2 threads. Please note once you go over that number the speculative is losing.

There are a few mass processing tests where you see insane difference between risc-v and x86 in big data stuff you can 1/20 of the power consume of the x86 and 1/4 of the processing time of the x86 this was just optimising the memory management system for big data using pure risc-v inorder single instruction at a time. That was a Xeon 8 cores active at 3ghz losing on wall time to a 1.5Ghz risc-v single core with thread management engine with a MMU design that allows reading 1 byte at a time from memory with single inorder instruction processing. When speculative execution hinders your performance it does it in insanely big way. This is the problem we know in mass processing that speculative execution is useless.

Now lets look at realtime processing. Speculative execution is also useless here as it introduces the jitters where time to complete a task is totally not predictable because it will depend if branch prediction got it right or wrong for how long the execution took this is not useful when you need to perform x tasks in exactly y time. Notice that 1.5 ghz risc-v from program tests does not have that problem either.

Your general x86 cpu design sits in the middle unable to be good for big data or good for real-time processing yet you see some of the new risc-v design be really good at both from a single design. There is quite a narrow window where speculative execution provides any benefit. It will not take very much change in the applications people use to get to the point where there is no benefit at all from speculative execution.

High end risc is asking the question have we got the memory system completely wrong. That high performance big data risc-v has no L2 cache at all. L3 and L1 yes. What wrong is that the cache systems on x86 are optimised that we are hitting the cache 90% of the time the big data risc-v was optimised on the idea that you miss the cache almost 100 percent of the time so price of a cache miss better be fairly cheep passing though L2 just adds extra latency. Speculative execution where you are repeatedly missing the cache is insanely performance harmful.

The hard reality is is x86 cpu cores might be design to support Speculative execution in fact head of the programs we want to run so our general performance of general applications could be in fact crippled. The work on risc-v is asking a lot questions how much in the current cpu we are using is in fact right and how much is in fact wrong for the workloads we wish to run.

How much has been just done for legacy application support adding speculative execution could add performance to existing programs without requiring to modify them. You cannot do cpu thread management engine without the OS you are running on it supporting it.

Weasel speculative execution can improve single threaded performance it can also hinder single thread performance in huge ways. So question with the general applications people use is speculative execution in fact helping or is it causing memory to bottle neck performing actions on branches that are going to be never used. Its a good question how many games are hitting 100 percent cpu usage because of speculative execution. Complex games and big data processing have a lot in common for having processing paths branch prediction is no good with as we have seen in big data you can be losing 2000 to 1 so only extract 1/2000 of the performance out the cpu hardware in the worst cases because speculative execution branch prediction keeps on getting it wrong.

As I said ideal cpu would have a thread management engine with the option to use speculative execution only when it suits and no hardware errors in speculative execution. Speculative execution mostly only suits when you have a low number of threads(1-4 per core) to process and you are not missing the caches performing speculative then speculative gives you single thread performance benefit otherwise speculative execution is performance hindering. Speculative is performance hindering quite a bit. Current x86 cpus are basically poor performing its like the old saying everything is a nail when you only have a hammer. Hammering a screw is not the correct usage of a screw and the x86 cpu design is that stupid this is why it not suitable for big data or real-time. Of course the question is if x86/speculative execution is in fact suitable for general computing in the middle between big data and real-time and that is up in the air with the answer being possibly no it not suitable.

Weasel I know my cpu hardware designs you don't. Most of your ideas are badly wrong. Ok the ideas you have were taught before people started playing with risc-v in big data and real-time resulting in seeing that lot of the presumes are wrong. The people designing cpus are not quite sure yet how badly wrong we have had it. Its possible that no cpu designed in the past 30 years has been anywhere near right.

The hardware errors in x86 are partly because up until now we could not build proper formal validation of silicon designs the maths to-do this have only came possible in the past 3 years. Expect more errors to be found now that attackers are looking at the black box that cpus have been because their designs never had the validation. At some point new cpu designs will have to be done more complex the instruction set the longer the validation process takes.

Something as simple as a risc-v would take a year to validate on a 32 core xeon workstation kind of explains why those working on risc-v are attempting to fix the big data processing problem first as you need this to validate designs. Validating something as complex as x86 you don't want to think about. Yes due to how much time validation in fact takes there are new risc-v being made with only part validation. The fact x86 is not great with big data how could intel effectively validate their chips the answer is simple they could not.