GCC 10 Introduces A Static Analyzer - Static Analysis On C Code With "-fanalyzer" Option

[pal666]

It had 8 "integer clusters". An integer cluster is not a core.

because some random imbecile from internet says so? ok, since integer cluster contains 4 pipelines, we can call it 4 cores and rebrand bulldozers as 32 core cpus.

It only had 1 instruction decoder per module.

it has decoder and dispatch capable of decoding and dispatching 4 instructions per cycle. which is strictly superior to two separate 2-wide decoders and dispatchers. i.e. it's also an optimization, though i guess it's required to feed shared fpus

1 FPU per module

it had two 128 bit fpus per module, which can be combined to one 256 bit fpu

. Etc. Only about 12% of the module was duplicated.

it's a lie like rest of your post. for example module had shared 64k of il1. which is strictly better than 2x32k of il1 per 2 cores of competition. but clueless brainwashed idiots will call 64k of amd "shared" and "false advertising" compared to 2x32k of intel

[skeevy420]

Anyone care to wager on which person in this thread is defending their purchase of an 8350?

[skeevy420]

like by what metric, moron? by output of random command? or by cost? or by performance? (hint: nothing of the above has any relevance to number of cores)

So a command like "inxi -C" to list my CPUs and thread counts has no relevance to the number or cores?

I show that I have an 8c/16ht system and that has no relevance to the number of cores?

And then you want to start off with calling me a moron.

Trash talk and brushing off factual data -- you clearly lost the dumbest argument on Phoronix from this week.

[smitty3268]

just i said, internet is full of idiots.

Oh nice, personal attacks.

hyperthreading doesn't double resources, hyperthreading masks cache misses.

Sorry, but this just shows you don't know what you're talking about. Hyperthreading does require additional silicon that allows it to run two threads on the same core. You don't get it for free, it's not a software feature. It requires additional silicon added to the chip to work. And no, it's not anywhere near "double" the silicon - the point is you can add a couple percent of silicon in order to get it to work. Just like bulldozer added 12% silicon to allow their "module" to run a 2nd thread.

by having two fpus shared by two cores you will have better performance when not every instruction in every thread uses fpu. alternative is to have one fpu per core - same number of transistors, same performance in worst case, worse performance on average.

The fact that you're so focused on the shared fpu just shows that you don't understand the real issue here. How can something be a core if doesn't have it's own frontend? The fpu is more of a minor detail than that.

. Etc. Only about 12% of the module was duplicated.

it's a lie like rest of your post.

The 12% number came straight from AMD's documentation. If it's a lie, go tell them.

[pal666]

Even AMD ended up basically admitting this, when they settled the false advertising lawsuit.

amd basically admitting you are brainwashed imbecile, their official statement reads "AMD is pleased to have reached a settlement of this lawsuit. While we believe the allegations are without merit, we also believe that eliminating the distraction and settling the litigation is in our best interest."

[pal666]

So a command like "inxi -C" to list my CPUs and thread counts has no relevance to the number or cores?

exactly. for example my number of cores according to it "bash: inxi: command not found..."

I show that I have an 8c/16ht system and that has no relevance to the number of cores?

yes, it has no relevance to the number of cores of bulldozer

And then you want to start off with calling me a moron.

because you are. program will output what is passed to printf. you can output anything you want, it does not change reality

[pal666]

Sorry, but this just shows you don't know what you're talking about. Hyperthreading does require additional silicon that allows it to run two threads on the same core. You don't get it for free, it's not a software feature. It requires additional silicon added to the chip to work. And no, it's not anywhere near "double" the silicon - the point is you can add a couple percent of silicon in order to get it to work. Just like bulldozer added 12% silicon to allow their "module" to run a 2nd thread.

well, the one showing lack of any clue is certainly not me. i never claimed hyperthreading doesn't require silicon. (btw, imbecile, i already told you that hyperthreading runs only one thread at a time, the other is sleeping in cache miss, that's the difference between hyperthreading and real cores). everything requires silicon. what i'm trying to explain to band of imbeciles is "amount of silicon has zero relevance to the number of cores". there are cores using different amounts of silicon per core and all of them are cores. old processors had orders of magnitude less silicon than one core of current models, does that mean they were using fractional number of cores?

The fact that you're so focused on the shared fpu just shows that you don't understand the real issue here. How can something be a core if doesn't have it's own frontend? The fpu is more of a minor detail than that.

what relevance sharedness of frontend has to number of cores? is this bullshit written in holy book of yours?

The 12% number came straight from AMD's documentation. If it's a lie, go tell them.

i'm sure amd documentation has many numbers none of which average moron can interpret properly

[pal666]

Anyone care to wager on which person in this thread is defending their purchase of an 8350?

the difference between me and average internet imbecile is that i don't buy cpu based on number of cores, either real or claimed by marketing. i buy cpu based on benchmark results and cost. and as it happens bulldozers were crushing competition in most relevant for me benchmark at the time - parallel builds (hint: this workload benefits heavily from additional cores). so i wager there are two kinds of people claiming "false cores" on amd behalf: imbeciles trying to defend purchase of cpu with half performance per dollar and imbeciles selecting videocards based on amount of videomemory. ok, there's third kind: lawyers who extorted several millions from amd

[skeevy420]

exactly. for example my number of cores according to it "bash: inxi: command not found..."

If you installed it you wouldn't get that error.

And you're calling me a moron...

the difference between me and average internet imbecile is that i don't buy cpu based on number of cores, either real or claimed by marketing. i buy cpu based on benchmark results and cost. and as it happens bulldozers were crushing competition in most relevant for me benchmark at the time - parallel builds (hint: this workload benefits heavily from additional cores). so i wager there are two kinds of people claiming "false cores" on amd behalf: imbeciles trying to defend purchase of cpu with half performance per dollar and imbeciles selecting videocards based on amount of videomemory. ok, there's third kind: lawyers who extorted several millions from amd

There's a fourth kind: People using them as the basis for jokes and having fun with the subsequent arguments that follow.