Us programmers always fuck up our dev machines, don't worry about it.
I've managed to get it so that Windows essentially deadlocks itself when starting new processes, still not sure how I did it. Had to reinstall to fix it
I am so glad I pushed back on moving my platform at work to ec2 when I was pushed to do so. I would be in a world of shit right now.
Really wouldn't have been your fault that Amazon decided that they should place all the storage for a single region in one data centre rather than spread it out across the eastern seaboard. When you purchase time across the multiple clusters in a region you would expect them to be geographically distinct.
Wouldn't have mattered to the customer responsible for me still having a job, though.
My platform I develop and maintain used to be in London with a whole team there. It was falling the fuck apart, had multiple instances of 8-10hr down time and 1 or 2 instances of a couple of days of down time within the last year or so. Everyone over there got let go. I was told "move this stuff somewhere else" at the point that our biggest customer, and one of very few left using that platform, was considering dropping us due to the instability of the platform. I moved it all to new servers here in the US back in February.
The CEO was wanting me to put it on ec2 where our own customer facing storefront lives, on the east coast instance, in the zone that is STILL down (although possibly could be back up if other people were on top of things and had proper backups, but I'm staying out of that as I don't manage that at all). That big customer would be doing up paperwork to drop us right now. No hard feelings toward me about it, but business is business. Without that customer, my platform costs too much to maintain and the CEO would be doing up the paperwork to get rid of me within days of them dropping us.
Hey guys, I'm having some route error issues with ASP.NET MVC 2, and I'm not sure if it's because I'm in debugging mode on my dev machine, or because something else is going on.
I'm trying to purposely invoke a 404 response. Instead, I get YSODs. If I supply a correctly formed URL, but a parameter is invalid (cannot be used to obtain data because it doesn't match anything that's in the db), I get a View exception since the view model's properties are null. If I supply a malformed URL, I get a Ninject exception telling me it cannot find the right controller. Applying a [HandleError] attribute doesn't fix the problem.
Any ideas on what's going on, or if this is par for the course since I'm in debugging mode?
Nightslyr on
PSN/XBL/Nintendo/Origin/Steam: Nightslyr 3DS: 1607-1682-2948 Switch: SW-3515-0057-3813 FF XIV: Q'vehn Tia
Us programmers always fuck up our dev machines, don't worry about it.
I've managed to get it so that Windows essentially deadlocks itself when starting new processes, still not sure how I did it. Had to reinstall to fix it
It seems super-handy for diagnosing problems like this. And I like to err on the side of fixing things rather then reinstalling around them.
No, actual process startup was taking around 30 seconds or so, just to get to the start of user code. The loader lock was being held by some random thread for a long time
void inner4(vec_ptr u, vec_ptr v, data t *dest)
{
long int i;
int length = vec_length(u);
data_t *udata = get_vec_start(u);
data_t *vdata = get_vec_start(v);
data_t sum = (data_t) 0;
for (i = 0; i < length; i++){
sum = sum + udata[i] * vdata[i];
}
*dest = sum;
}
I have to write a version of the "inner product procedure" that uses "four way loop unrolling."
This is my code:
void combine5(vec_ptr v, data_t *dest) {
long int length = vec_length(v);
long int limit = length-4;
data_t *data = get_vec_start(v);
data_t acc = IDENT;
long int i;
for (i = 0; i < limit; i+=3)
acc = acc OP data[i] OP data[i+1] OP data[i+2];
}
for (; i < length; i++)
acc = acc OP data[i];
}
*dest = acc;
}
Is this right?
EDIT: I made some revisions to my code, if anyone can verify if its right or give me some additional input it would be greatly appreciated. I am forever grateful to you guys in this thread. I like to think I've learned a considerable amount from here in addition to the classes I am taking
Huh. A lot of changes there, I'd have figured all you wanted to do was unroll the loop:
void inner4(vec_ptr u, vec_ptr v, data t *dest)
{
long int i;
int length = vec_length(u);
data_t *udata = get_vec_start(u);
data_t *vdata = get_vec_start(v);
data_t sum = (data_t) 0;
for (i = 0; i < length; i+=4){
sum = sum + udata[i] * vdata[i];
sum = sum + udata[i+1] * vdata[i+1];
sum = sum + udata[i+2] * vdata[i+2];
sum = sum + udata[i+3] * vdata[i+3];
}
*dest = sum;
}
That as a first step, then something to take into account if the limit isn't a multiple of 4. That's your basic loop unrolling (assuming 4-way unrolling here means unrolling into batches of 4).
The vast majority of your changes don't seem to have anything to do with loop unrolling, and I'm not sure what they're meant to do - the given implementation does inner products for you and it looks to me like you've broken some important parts of it.
I am so glad I pushed back on moving my platform at work to ec2 when I was pushed to do so. I would be in a world of shit right now.
Really wouldn't have been your fault that Amazon decided that they should place all the storage for a single region in one data centre rather than spread it out across the eastern seaboard. When you purchase time across the multiple clusters in a region you would expect them to be geographically distinct.
Wouldn't have mattered to the customer responsible for me still having a job, though.
My platform I develop and maintain used to be in London with a whole team there. It was falling the fuck apart, had multiple instances of 8-10hr down time and 1 or 2 instances of a couple of days of down time within the last year or so. Everyone over there got let go. I was told "move this stuff somewhere else" at the point that our biggest customer, and one of very few left using that platform, was considering dropping us due to the instability of the platform. I moved it all to new servers here in the US back in February.
The CEO was wanting me to put it on ec2 where our own customer facing storefront lives, on the east coast instance, in the zone that is STILL down (although possibly could be back up if other people were on top of things and had proper backups, but I'm staying out of that as I don't manage that at all). That big customer would be doing up paperwork to drop us right now. No hard feelings toward me about it, but business is business. Without that customer, my platform costs too much to maintain and the CEO would be doing up the paperwork to get rid of me within days of them dropping us.
The good news is that you can get an ubuntu cloud cluster with 3-4 machines up and simulate an ec2 environment and make it easy to migrate back and forth. I mean if the store was expecting tons of traffic this might not be optimal but at least downtime is less "oh shit oh shit oh shit oh shit" and more "meh it's kinda slow today"
bowen on
not a doctor, not a lawyer, examples I use may not be fully researched so don't take out of context plz, don't @ me
0
jackalFuck Yes. That is an orderly anal warehouse.Registered Userregular
Hey guys, I'm having some route error issues with ASP.NET MVC 2, and I'm not sure if it's because I'm in debugging mode on my dev machine, or because something else is going on.
I'm trying to purposely invoke a 404 response. Instead, I get YSODs. If I supply a correctly formed URL, but a parameter is invalid (cannot be used to obtain data because it doesn't match anything that's in the db), I get a View exception since the view model's properties are null. If I supply a malformed URL, I get a Ninject exception telling me it cannot find the right controller. Applying a [HandleError] attribute doesn't fix the problem.
Any ideas on what's going on, or if this is par for the course since I'm in debugging mode?
In web.config under <system.web> add <customErrors mode="On" />.
So, branches in git. I made a branch to work on a pretty big change to some code, so I could have the working codebase intact to deploy while I worked on the new stuff in the branch.
I got done, merged the branch into the main trunk. But I still have the branch there in git. Now what? I'm done working on the branch, can I just delete it? Every commit I did in the branch got merged into the main trunk, right?
Wouldn't have mattered to the customer responsible for me still having a job, though.
My platform I develop and maintain used to be in London with a whole team there. It was falling the fuck apart, had multiple instances of 8-10hr down time and 1 or 2 instances of a couple of days of down time within the last year or so. Everyone over there got let go. I was told "move this stuff somewhere else" at the point that our biggest customer, and one of very few left using that platform, was considering dropping us due to the instability of the platform. I moved it all to new servers here in the US back in February.
The CEO was wanting me to put it on ec2 where our own customer facing storefront lives, on the east coast instance, in the zone that is STILL down (although possibly could be back up if other people were on top of things and had proper backups, but I'm staying out of that as I don't manage that at all). That big customer would be doing up paperwork to drop us right now. No hard feelings toward me about it, but business is business. Without that customer, my platform costs too much to maintain and the CEO would be doing up the paperwork to get rid of me within days of them dropping us.
The good news is that you can get an ubuntu cloud cluster with 3-4 machines up and simulate an ec2 environment and make it easy to migrate back and forth. I mean if the store was expecting tons of traffic this might not be optimal but at least downtime is less "oh shit oh shit oh shit oh shit" and more "meh it's kinda slow today"
Sort of. At a sane company, sure. The whole reason we have anything on ec2 is because it's cheap with no up front hardware cost and slight less maintenance/sys admin type cost as long as you're not using the load balancing and keeping instances synched between regions. With that in mind, you can probably determine what the chances are the CEO would approve buying a few machines to have sitting around and paying monthly hosting costs on them "just in case".
Since we're still not back up yet, I have to assume we couldn't even be bothered to take snapshots of the instances since you've been able to use a snapshot to start up instances in a new zone for somewhere around 18 hours now.
So yeah, I'm just really glad I fought to get my shit on our own hardware in a data center 20 minutes away. In an absolute worst case scenario where the data center just couldn't get us back up, I could go pull everything out of the rack and bring it into the office, crank the AC, change DNS, and call it a day. All VMs are mirrored across two physical machines, network storage is all mirrored albeit the backup mirror is low end shit that would barely hold up.
Yeah That's the best solution Jimmy. I am not a fan of "let us take care of shit for you" because 9 times out of 10 as soon as they start taking care of things, someone takes a giant dump on the servers and they hand waive for 48 hours.
bowen on
not a doctor, not a lawyer, examples I use may not be fully researched so don't take out of context plz, don't @ me
So, branches in git. I made a branch to work on a pretty big change to some code, so I could have the working codebase intact to deploy while I worked on the new stuff in the branch.
I got done, merged the branch into the main trunk. But I still have the branch there in git. Now what? I'm done working on the branch, can I just delete it? Every commit I did in the branch got merged into the main trunk, right?
If the branch was really just a feature branch, then yeah, deleting it is fine.
Now when you say "git" do you actually mean "git" or do you mean "github"? Deleting a local branch is easy, but deleting a remote branch is a little more complicated. You need to push nothing to the remote branch via git push REMOTE :BRANCH.
jonxp on
Every time you write parallel fifths, Bach kills a kitten.
3DS Friend Code: 2707-1614-5576 PAX Prime 2014 Buttoneering!
So hey, what is a sensible way to take a whole bunch of annoyingly formatted text files (not quite space delimited, not quite tab delimited... just sort of presented like someone wanted it to look pretty and then forgot no one fucking cares and we just want sensible text files), grab the info I want, and slap that into an Excel spreadsheet?
I've got about 200 of these godawful useless files that some poor intern has been opening up and doing manual data-entry with, and I don't see any reason for anyone to be doing that ever again.
My first instinct is to just make some sort of batch file to take care of it, but I'm not really certain what that would entail as I just realized that I have 0 experience with scripting outside of Linux.
Maybe I should write something in C. I haven't done that in like 2 years. I remember something about editing text files in C, though. Man, I need some sort of refresher course every 6 months for all the languages I "know".
Actually watch the video. The program looks really really nice, and I would most likely try it out. But we the number of submodules I currently work with and the full on branchy work flow I don't know if it would be worth while.
Okay, let's pretend I can't actually install anything that will allow me to use C# on this apparently.
I might bring a file home to play with, but what's my other option?
I guess I'd need to know your platform, most PCs have .NET so it's a matter of writing and compiling and running it on the remote system with these files?
Your other option is to do it by hand, or, I guess do C++? Though you'll need just as much to compile a C/C++ app as you will a C# app. Maybe slightly less if you use notepad and just download gcc or something.
bowen on
not a doctor, not a lawyer, examples I use may not be fully researched so don't take out of context plz, don't @ me
Python would be my option. It is way nicer for general file IO, and file formatting that C++. Than use python2exe to make a single self extracting executable to process all the files.
Okay, let's pretend I can't actually install anything that will allow me to use C# on this apparently.
I might bring a file home to play with, but what's my other option?
I guess I'd need to know your platform, most PCs have .NET so it's a matter of writing and compiling and running it on the remote system with these files?
Your other option is to do it by hand, or, I guess do C++? Though you'll need just as much to compile a C/C++ app as you will a C# app. Maybe slightly less if you use notepad and just download gcc or something.
Oh, wait, no. I've gotten around the permissions business.
Okay, I'll install .NET, try my hand with C#.
Or, now that I look at Ethea's answer, maybe Python would work. Ooh, now that I know I can actually install things I can move on up to Notepad++.
I have done very little text manipulating with C#, but Python is leagues ahead of C++ in this area. I would also read http://docs.python.org/library/csv.html as this might module might solve most of your issues durandal (at least for writing at a min).
Huh. A lot of changes there, I'd have figured all you wanted to do was unroll the loop:
void inner4(vec_ptr u, vec_ptr v, data t *dest)
{
long int i;
int length = vec_length(u);
data_t *udata = get_vec_start(u);
data_t *vdata = get_vec_start(v);
data_t sum = (data_t) 0;
for (i = 0; i < length; i+=4){
sum = sum + udata[i] * vdata[i];
sum = sum + udata[i+1] * vdata[i+1];
sum = sum + udata[i+2] * vdata[i+2];
sum = sum + udata[i+3] * vdata[i+3];
}
*dest = sum;
}
That as a first step, then something to take into account if the limit isn't a multiple of 4. That's your basic loop unrolling (assuming 4-way unrolling here means unrolling into batches of 4).
The vast majority of your changes don't seem to have anything to do with loop unrolling, and I'm not sure what they're meant to do - the given implementation does inner products for you and it looks to me like you've broken some important parts of it.
I initially thought that well, since I'm doing a 4 way loop unrolling then the length would have to be 4 right?
I was wrong in that regard. Would 3 be a more appropriate number? (Same length as the Cycles Per Element)
Thanks for the help. Ive got a long way to go
Huh. A lot of changes there, I'd have figured all you wanted to do was unroll the loop:
void inner4(vec_ptr u, vec_ptr v, data t *dest)
{
long int i;
int length = vec_length(u);
data_t *udata = get_vec_start(u);
data_t *vdata = get_vec_start(v);
data_t sum = (data_t) 0;
for (i = 0; i < length; i+=4){
sum = sum + udata[i] * vdata[i];
sum = sum + udata[i+1] * vdata[i+1];
sum = sum + udata[i+2] * vdata[i+2];
sum = sum + udata[i+3] * vdata[i+3];
}
*dest = sum;
}
That as a first step, then something to take into account if the limit isn't a multiple of 4. That's your basic loop unrolling (assuming 4-way unrolling here means unrolling into batches of 4).
The vast majority of your changes don't seem to have anything to do with loop unrolling, and I'm not sure what they're meant to do - the given implementation does inner products for you and it looks to me like you've broken some important parts of it.
I initially thought that well, since I'm doing a 4 way loop unrolling then the length would have to be 4 right?
I was wrong in that regard. Would 3 be a more appropriate number? (Same length as the Cycles Per Element)
Thanks for the help. Ive got a long way to go
The problem here is that you don't know what the length of the vector is in advance, so you can't customise the implementation for what the length of the vector is. It could be anything: there is no "appropriate number" in that sense, and for any number of operations bundled in a loop iteration, you could have a length which doesn't divide into it evenly. I haven't heard loop unrolling referred to as "n-way" so I was assuming that's what it meant: what were you taught that 4-way unrolling means?
So, in this sort of situation, in order to unroll that loop, you also need special handling code for when it doesn't divide evenly: just work out how many elements are left, and process those. That's part and parcel of dealing with loop unrolling for variable-length data.
As an aside, this project has been fucking frustrating as hell. There's a gazillion little detail things that have been tripping me up like this as I've slowly (agonizingly) progressed. There seems to be an assumption by the bloggers, tutorial writers, and, hell, even the book authors, that newbies are familiar/comfortable with playing around in Web.config, knowing what options are available in it, or knowing when Entity Framework runs its own validation, or any number of other minute things that can stop a project dead without giving much of an inkling about what's actually wrong. There always seems to be a hidden "It'll work only if you do this obvious thing we won't mention" caveat with everything.
I actually started a goddamn blog about that kind of shit this afternoon. I'd link to it if I wasn't worried about getting mod-hammered for spam.
Nightslyr on
PSN/XBL/Nintendo/Origin/Steam: Nightslyr 3DS: 1607-1682-2948 Switch: SW-3515-0057-3813 FF XIV: Q'vehn Tia
As an aside, this project has been fucking frustrating as hell. There's a gazillion little detail things that have been tripping me up like this as I've slowly (agonizingly) progressed. There seems to be an assumption by the bloggers, tutorial writers, and, hell, even the book authors, that newbies are familiar/comfortable with playing around in Web.config, knowing what options are available in it, or knowing when Entity Framework runs its own validation, or any number of other minute things that can stop a project dead without giving much of an inkling about what's actually wrong. There always seems to be a hidden "It'll work only if you do this obvious thing we won't mention" caveat with everything.
I actually started a goddamn blog about that kind of shit this afternoon. I'd link to it if I wasn't worried about getting mod-hammered for spam.
I've been working with ASP/.NET MVC for the first time in my new job, and my experience has been the same. I've got many years experience with both C# (clientside apps) and web frameworks, but this is the first time I've had to combine the two. It's been super frustrating to get the simplest tasks done inside the framework they provide.
jonxp on
Every time you write parallel fifths, Bach kills a kitten.
3DS Friend Code: 2707-1614-5576 PAX Prime 2014 Buttoneering!
Personally, I think it would be cool to have some sort of list of F/OSS projects that people here are working on or have created on their own and may be inactive with but still may be useful or interesting to people.
Posts
Suggestion: get a program called Process Monitor.
It seems super-handy for diagnosing problems like this. And I like to err on the side of fixing things rather then reinstalling around them.
My platform I develop and maintain used to be in London with a whole team there. It was falling the fuck apart, had multiple instances of 8-10hr down time and 1 or 2 instances of a couple of days of down time within the last year or so. Everyone over there got let go. I was told "move this stuff somewhere else" at the point that our biggest customer, and one of very few left using that platform, was considering dropping us due to the instability of the platform. I moved it all to new servers here in the US back in February.
The CEO was wanting me to put it on ec2 where our own customer facing storefront lives, on the east coast instance, in the zone that is STILL down (although possibly could be back up if other people were on top of things and had proper backups, but I'm staying out of that as I don't manage that at all). That big customer would be doing up paperwork to drop us right now. No hard feelings toward me about it, but business is business. Without that customer, my platform costs too much to maintain and the CEO would be doing up the paperwork to get rid of me within days of them dropping us.
I'm trying to purposely invoke a 404 response. Instead, I get YSODs. If I supply a correctly formed URL, but a parameter is invalid (cannot be used to obtain data because it doesn't match anything that's in the db), I get a View exception since the view model's properties are null. If I supply a malformed URL, I get a Ninject exception telling me it cannot find the right controller. Applying a [HandleError] attribute doesn't fix the problem.
Any ideas on what's going on, or if this is par for the course since I'm in debugging mode?
Switch: SW-3515-0057-3813 FF XIV: Q'vehn Tia
No, actual process startup was taking around 30 seconds or so, just to get to the start of user code. The loader lock was being held by some random thread for a long time
EDIT: I made some revisions to my code, if anyone can verify if its right or give me some additional input it would be greatly appreciated. I am forever grateful to you guys in this thread. I like to think I've learned a considerable amount from here in addition to the classes I am taking
void inner4(vec_ptr u, vec_ptr v, data t *dest) { long int i; int length = vec_length(u); data_t *udata = get_vec_start(u); data_t *vdata = get_vec_start(v); data_t sum = (data_t) 0; for (i = 0; i < length; i+=4){ sum = sum + udata[i] * vdata[i]; sum = sum + udata[i+1] * vdata[i+1]; sum = sum + udata[i+2] * vdata[i+2]; sum = sum + udata[i+3] * vdata[i+3]; } *dest = sum; }That as a first step, then something to take into account if the limit isn't a multiple of 4. That's your basic loop unrolling (assuming 4-way unrolling here means unrolling into batches of 4).The vast majority of your changes don't seem to have anything to do with loop unrolling, and I'm not sure what they're meant to do - the given implementation does inner products for you and it looks to me like you've broken some important parts of it.
The good news is that you can get an ubuntu cloud cluster with 3-4 machines up and simulate an ec2 environment and make it easy to migrate back and forth. I mean if the store was expecting tons of traffic this might not be optimal but at least downtime is less "oh shit oh shit oh shit oh shit" and more "meh it's kinda slow today"
In web.config under <system.web> add <customErrors mode="On" />.
I got done, merged the branch into the main trunk. But I still have the branch there in git. Now what? I'm done working on the branch, can I just delete it? Every commit I did in the branch got merged into the main trunk, right?
Since we're still not back up yet, I have to assume we couldn't even be bothered to take snapshots of the instances since you've been able to use a snapshot to start up instances in a new zone for somewhere around 18 hours now.
So yeah, I'm just really glad I fought to get my shit on our own hardware in a data center 20 minutes away. In an absolute worst case scenario where the data center just couldn't get us back up, I could go pull everything out of the rack and bring it into the office, crank the AC, change DNS, and call it a day. All VMs are mirrored across two physical machines, network storage is all mirrored albeit the backup mirror is low end shit that would barely hold up.
If the branch was really just a feature branch, then yeah, deleting it is fine.
Now when you say "git" do you actually mean "git" or do you mean "github"? Deleting a local branch is easy, but deleting a remote branch is a little more complicated. You need to push nothing to the remote branch via git push REMOTE :BRANCH.
3DS Friend Code: 2707-1614-5576
PAX Prime 2014 Buttoneering!
fixed
3DS Friend Code: 2707-1614-5576
PAX Prime 2014 Buttoneering!
I've got about 200 of these godawful useless files that some poor intern has been opening up and doing manual data-entry with, and I don't see any reason for anyone to be doing that ever again.
My first instinct is to just make some sort of batch file to take care of it, but I'm not really certain what that would entail as I just realized that I have 0 experience with scripting outside of Linux.
Maybe I should write something in C. I haven't done that in like 2 years. I remember something about editing text files in C, though. Man, I need some sort of refresher course every 6 months for all the languages I "know".
not really but thanks anyway
IMO, the process should be unifying their formation into a csv file, then importing the csv file into excel.
No submodule support, can't use.
Edit:
Actually watch the video. The program looks really really nice, and I would most likely try it out. But we the number of submodules I currently work with and the full on branchy work flow I don't know if it would be worth while.
I really wish they'd just all been csv files in the first place.
I might bring a file home to play with, but what's my other option?
I guess I'd need to know your platform, most PCs have .NET so it's a matter of writing and compiling and running it on the remote system with these files?
Your other option is to do it by hand, or, I guess do C++? Though you'll need just as much to compile a C/C++ app as you will a C# app. Maybe slightly less if you use notepad and just download gcc or something.
Oh, wait, no. I've gotten around the permissions business.
Okay, I'll install .NET, try my hand with C#.
Or, now that I look at Ethea's answer, maybe Python would work. Ooh, now that I know I can actually install things I can move on up to Notepad++.
I love when work finally gives me actual programming problems instead of constant tinkering with ancient Access databases.
I initially thought that well, since I'm doing a 4 way loop unrolling then the length would have to be 4 right?
I was wrong in that regard. Would 3 be a more appropriate number? (Same length as the Cycles Per Element)
Thanks for the help. Ive got a long way to go
So, in this sort of situation, in order to unroll that loop, you also need special handling code for when it doesn't divide evenly: just work out how many elements are left, and process those. That's part and parcel of dealing with loop unrolling for variable-length data.
Been using it since the beta started.
As an aside, this project has been fucking frustrating as hell. There's a gazillion little detail things that have been tripping me up like this as I've slowly (agonizingly) progressed. There seems to be an assumption by the bloggers, tutorial writers, and, hell, even the book authors, that newbies are familiar/comfortable with playing around in Web.config, knowing what options are available in it, or knowing when Entity Framework runs its own validation, or any number of other minute things that can stop a project dead without giving much of an inkling about what's actually wrong. There always seems to be a hidden "It'll work only if you do this obvious thing we won't mention" caveat with everything.
I actually started a goddamn blog about that kind of shit this afternoon. I'd link to it if I wasn't worried about getting mod-hammered for spam.
Switch: SW-3515-0057-3813 FF XIV: Q'vehn Tia
Do share with the class.
I've been working with ASP/.NET MVC for the first time in my new job, and my experience has been the same. I've got many years experience with both C# (clientside apps) and web frameworks, but this is the first time I've had to combine the two. It's been super frustrating to get the simplest tasks done inside the framework they provide.
3DS Friend Code: 2707-1614-5576
PAX Prime 2014 Buttoneering!
http://mpdevblog.blogspot.com/
Switch: SW-3515-0057-3813 FF XIV: Q'vehn Tia