Windows 11 General Discussion

[gurusmi]

Cool. The whole point is this does it in one instruction. Cheaper.

Really? Loading the value from an adreess into a register, set a second register to the mask value, "and" them and at least write the result back to the memory at an adress? Are you dreaming? There are at least 4 cycles to get that done. Equal to the CPU statement. Both sides, at least 4 cycles. One as one order- The other as four.

Oops.

 

[R-T-B]

Really? Loading the value from an adreess into a register, set a second register to the mask value, "and" them and at least write the result back to the memory at an adress? Are you dreaming?

No I obviously meant the summing. POPCNT is literally a single instruction. I don't know why this is so confusing.

 

[iameatingjam]

So... I'm trying to figure out why my system randomly rebooted a bunch of times last week. I did briefly get a bsod but it flashed on screen quickly before powering off so I couldn't get the error message in time.

I looked on windows event viewer and every time it was error 41. Which I understand is pretty broad. But I did read somewhere that changing power plan to high performance might be able to help, and it seemed to. I haven't got it once since changing it, though not sure if its coincidence. Also I don't really want my cpu running at max boost when it doesn't have to.

I looked at errors previous to the shutdowns and found this common offender:

Error 11: The driver detected a controller error on \Device\RaidPort3.

Does this maybe mean there's something wrong with my raid configuration? I was installing something from the raid drive to a different drive and the shutdowns happened either while I was doing this or shortly after it crashed part way through. Could a corrupted file, non-essential to windows, cause this? And for the two drives in the raid 0, one shows 100% health, the other shows 99%. Both sata 1tb mx500.

Any ideas?

I don't think its psu related because only the cpu was drawing power in significant amounts at the time ( file decompression) which shouldn't be anywhere near the limit, and I can run the gpu with even higher power draw without any issue. But I could be wrong idk.

 

[lexluthermiester]

Cool. The whole point is this does it in one instruction. Cheaper.

I've been looking into this. They haven't done much here. Even the First Gen Corei lineup has that instruction, as does everything mobile from that timeframe. Everything from 988/1156/1366 on up is still able to run.

What is microsoft doing? And why?

I don't know why this is so confusing.

Because it seems stupid..

Does this maybe mean there's something wrong with my raid configuration?

Any ideas?

Yes run a check on your RAID array to make sure all is in order. You need to use whatever tools/utilities that came with the controller itself, if they exist. If not, you will need to backup all your data and test each drive individually one at a time.

 

[iameatingjam]

Yes run a check on your RAID array to make sure all is in order. You need to use whatever tools/utilities that came with the controller itself, if they exist. If not, you will need to backup all your data and test each drive individually one at a time.

Sorry I forgot to mention, I made this within windows. If memory serves I went to Computer Management>storages and used 'new stripped volume' or something like.

 

[lexluthermiester]

Sorry I forgot to mention, I made this within windows. If memory serves I went to Computer Management>storages and used 'new stripped volume' or something like.

Ah, well that should be easy then. The management tools have a check function that can be easier used. Have you tried it and what result did you get?

 

[ThrashZone]

Hi,
Oh my someone uses storage spaces

 

[lexluthermiester]

Hi,
Oh my someone uses storage spaces

It's not so bad if used properly. Not the best option, but not terrible.

 

[iameatingjam]

Ah, well that should be easy then. The management tools have a check function that can be easier used. Have you tried it and what result did you get?

Sorry, I don't know what you mean really. Is that something within the computer management gui? Or something done from the command prompt? I'm running a 'chkdsk G: /f /r' right now and looks like it will take a while.

Is... that what you asked for?

 

[R-T-B]

Because it seems stupid..

I was of the opinion it might have been for improved performance but sounds more like a temporary compiler bug in Visual Studio.

 

[lexluthermiester]

Sorry, I don't know what you mean really. Is that something within the computer management gui? Or something done from the command prompt? I'm running a 'chkdsk G: /f /r' right now and looks like it will take a while.

Is... that what you asked for?

You can do that too. But no, I was talking about in the Management console.

I was of the opinion it might have been for improved performance but sounds more like a temporary compiler bug in Visual Studio.

I may have misunderstood you. Weren't we still talking about the POPCNT thing?

 

[Klemc]

My Eleven build is 2023-12-12 KB, latest is KB5034765, i'll see if i install march update instead.

 

[R-T-B]

I may have misunderstood you. Weren't we still talking about the POPCNT thing?

The fact that it is now a requirement (at least temporarily), yes?

 

[lexluthermiester]

The fact that it is now a requirement (at least temporarily), yes?

Ok, yeah. Honestly, I'm at a loss. Who knows why this is going on..

 

[R-T-B]

Ok, yeah. Honestly, I'm at a loss. Who knows why this is going on..

Yeah it's sort of confusing right now.

 

[lexluthermiester]

Yeah it's sort of confusing right now.

I think some people have made some assumptions. The folks at microsoft themselves haven't made any official statement. Time will tell..

 

[gurusmi]

No I obviously meant the summing. POPCNT is literally a single instruction. I don't know why this is so confusing.

One command taking at least 4 cycles or 4 commands taking one cycle each. Which command is faster?

 

[lexluthermiester]

One command taking at least 4 cycles or 4 commands taking one cycle each. Which command is faster?

That doesn't really matter.

 

[R-T-B]

One command taking at least 4 cycles or 4 commands taking one cycle each. Which command is faster?

You have no way of knowing that POPCNT is 4 cycles. That'd shock me frankly.

 

[lexluthermiester]

You have no way of knowing that POPCNT is 4 cycles. That'd shock me frankly.

There is a way to calculate that, but it is code, compile and routine dependant. In theory, the shortest amount of time that a CPU can run the POPCNT function seems to be 4 cycles if it's only counting a bit shift. I've been bouncing all over the net reading about this and don't remember where I read that(so I'm sorry, can't cite it), but it does seem about right. However, your implication holds true, as a rule it will take much longer(hundreds of cycles) to run the function as a part of a full routine.

 

[R-T-B]

There is a way to calculate that, but it is code, compile and routine dependant. In theory, the shortest amount of time that a CPU can run the POPCNT function seems to be 4 cycles if it's only counting a bit shift. I've been bouncing all over the net reading about this and don't remember where I read that(so I'm sorry, can't cite it), but it does seem about right. However, your implication holds true, as a rule it will take much longer(hundreds of cycles) to run the function as a part of a full routine.

All I know is they don't add instructions for fun, so there likely is a performance related rational to it.

 

[lexluthermiester]

All I know is they don't add instructions for fun, so there likely is a performance related rational to it.

Agreed.

 

[gurusmi]

There is a way to calculate that, but it is code, compile and routine dependant. In theory, the shortest amount of time that a CPU can run the POPCNT function seems to be 4 cycles if it's only counting a bit shift. I've been bouncing all over the net reading about this and don't remember where I read that(so I'm sorry, can't cite it), but it does seem about right. However, your implication holds true, as a rule it will take much longer(hundreds of cycles) to run the function as a part of a full routine.

I just cite:

jcranmer on Sept 9, 2019 | root | parent | next [–] A popcnt/test/jmp cycle should have roughly 4 cycles of latency: a test+jmp can be fused into one uop, and popcnt would be 3 cycles of latency. n&(n-1) == 0 would compile down into a DEC, TEST, JMP, which is 2 cycles of latency (again, test+jmp are fused into one uop), as dec is just 1 cycle of latency. So n&(n-1) is faster for checking if it's a power of two.

I'm sorry. POPCNT IS slower!

All I know is they don't add instructions for fun, so there likely is a performance related rational to it.

They do. POPCNT could be also done by n&(n-1) == 0

 

[lexluthermiester]

I just cite:

jcranmer on Sept 9, 2019 | root | parent | next [–] A popcnt/test/jmp cycle should have roughly 4 cycles of latency: a test+jmp can be fused into one uop, and popcnt would be 3 cycles of latency. n&(n-1) == 0 would compile down into a DEC, TEST, JMP, which is 2 cycles of latency (again, test+jmp are fused into one uop), as dec is just 1 cycle of latency. So n&(n-1) is faster for checking if it's a power of two.

I'm sorry. POPCNT IS slower!

That not the same article I read, but it's similar. So yeah, that about covers it.

 

[Onasi]

So, an update on the Defender situation I described yesterday. Seems like I was indeed correct when I thought that the January 24th Update borked something in my Windows Update behaviour. The system just... doesn't check for updates automatically anymore. If I click it manually all is fine and dandy and it works and downloads. But there is no automatic check on a fresh boot after *checks time* around 18-19 hours of the PC being off. This is... weird, seeing how it worked properly before. I haven't changed anything, no Group Policies affecting the WU had been changed. The SFC comes out normal, DISM check too, Windows Update troubleshooter doesn't do anything. I have no idea what MS borked up here. It's not too big of a deal since I can just check manually at my leasure, but the whole issue is just bizzare.
Hilariously, Defender itself still checks for and downloads new definitions from the Security Center. It's a mess.

Edit: Oh wait, it JUST checked after PC being on for a while. And, interestingly, after Defender itself downloaded its definitions. I have... no idea. Maybe the tasks of Defender and WU started conflicting and one holds up the other? Maybe. So Defender checked on 2:58 PM (my local time, natch) and at 3:03 Update finally woke up and checked itself. What the actual f**k MS.