Code compiled with -m64 fails but with -m32 works just fine Are there any special moments to remembe

[Romant]

Yes. But you are working with 64 bit system. Do you?

I use XP64 … don’t have plans to migrate to 7 as lots of stuff will stop working after it. Seems like I have a dilemma - XP64 without Nsight or 7 with Nsight but with lots of migration problems also.

[Romant]

Yes. But you are working with 64 bit system. Do you?

I use XP64 … don’t have plans to migrate to 7 as lots of stuff will stop working after it. Seems like I have a dilemma - XP64 without Nsight or 7 with Nsight but with lots of migration problems also.

[Romant]

Yes. But you are working with 64 bit system. Do you?

In addition … if I understand right, you questions has a purpose to make sure that I don’t try to compile the kernel with -m64 under 32bit system ?

[Romant]

Yes. But you are working with 64 bit system. Do you?

In addition … if I understand right, you questions has a purpose to make sure that I don’t try to compile the kernel with -m64 under 32bit system ?

[Lev]

I just thought that it was rare case xp64. Though after you told you are using 64 bit OS and told that nsight does not work on XP, I suggested that you use linux64 and windowx xp32 bit. Now I got you are on xp64. Maybe it is some cuda issue on this OS.

[Lev]

I just thought that it was rare case xp64. Though after you told you are using 64 bit OS and told that nsight does not work on XP, I suggested that you use linux64 and windowx xp32 bit. Now I got you are on xp64. Maybe it is some cuda issue on this OS.

[tera]

To me this very much sounds like you are running out of registers, although I’m not sure about the error code returned.

[tera]

To me this very much sounds like you are running out of registers, although I’m not sure about the error code returned.

[Romant]

To me this very much sounds like you are running out of registers, although I’m not sure about the error code returned.

Code 4 (unspecified launch failure). How can I run out of registers ? Kernel would not run at all if too many resources requested, I guess.

[Romant]

To me this very much sounds like you are running out of registers, although I’m not sure about the error code returned.

Code 4 (unspecified launch failure). How can I run out of registers ? Kernel would not run at all if too many resources requested, I guess.

[tera]

Kernel would not run at all if too many resources requested, I guess.

That is what I suspected, yes.

[tera]

Kernel would not run at all if too many resources requested, I guess.

That is what I suspected, yes.

[jan.heckman]

I have a kernel that used to work normally for almost 1.5 years, it has been continuously maintained and improved for all this time. Main project can be compiled for x86 and x64 so I use both -m32 and -m64 keys.

After addition of two code snippets into the kernel (added code has been checked and rechecked hundred of times, it is absolutely similar to already present code that used to work fine) it began to fail with “Unspecified launch failure” error. It is very surprising that this kernel (with added code) works without any errors when compiled with -m32 but fails with -m64.

I have nailed the problem down to the condition when single line of code being commented makes the code work without “unspecified launch failures” but this line of code has absolutely nothing with code snippets that has been actually added. In general, commenting one case in a switch statement makes the kernel work - but the statement that needs to be commented is not called even a single time, it is a code that does not work under testing conditions.

All the facts make me think that it is a bug in nvcc, however, the goal is to make the kernel work, not to point that there are bugs in nvcc. The fact that I can’t extract small piece of code that reproduces the issue is also not helpful. I would appreciate any suggestions if someone has already fixed something similar (-m32 that works and -m64 that don’t).

Thanks in advance.

You can try “Verbose PTXAS output” option to determine register pressure, anyway.

Which GPU architecture are you compiling for?

480 → cache options?

Just some thougths.

Oh, migration. Why don’t you put another harddisk in your system with win 7 64 on it and try things out in your own good time? Made migration lots easier for me.

[jan.heckman]

I have a kernel that used to work normally for almost 1.5 years, it has been continuously maintained and improved for all this time. Main project can be compiled for x86 and x64 so I use both -m32 and -m64 keys.

After addition of two code snippets into the kernel (added code has been checked and rechecked hundred of times, it is absolutely similar to already present code that used to work fine) it began to fail with “Unspecified launch failure” error. It is very surprising that this kernel (with added code) works without any errors when compiled with -m32 but fails with -m64.

I have nailed the problem down to the condition when single line of code being commented makes the code work without “unspecified launch failures” but this line of code has absolutely nothing with code snippets that has been actually added. In general, commenting one case in a switch statement makes the kernel work - but the statement that needs to be commented is not called even a single time, it is a code that does not work under testing conditions.

All the facts make me think that it is a bug in nvcc, however, the goal is to make the kernel work, not to point that there are bugs in nvcc. The fact that I can’t extract small piece of code that reproduces the issue is also not helpful. I would appreciate any suggestions if someone has already fixed something similar (-m32 that works and -m64 that don’t).

Thanks in advance.

You can try “Verbose PTXAS output” option to determine register pressure, anyway.

Which GPU architecture are you compiling for?

480 → cache options?

Just some thougths.

Oh, migration. Why don’t you put another harddisk in your system with win 7 64 on it and try things out in your own good time? Made migration lots easier for me.