*Warning*

This email originated from outside of Maynooth University's Mail System. Do not reply, click links or open attachments unless you recognise the sender and know the content is safe.

Hi!

I just got a new laptop at work which has a T500 nvidia graphics card and I managed to install all the cuda bells and whistles for it with some effort.

Now I am trying to have the csound cuda opcodes built, but seem to run into problems.  cmake finds the cuda installation allright, but there seems to be some trouble with the CUDA/CMakeLists.tx file. I get the following error:

CMake Error at CUDA/CMakeLists.txt:20 (set_target_properties):
  set_target_properties called with incorrect number of arguments.
Call Stack (most recent call first):
  CUDA/CMakeLists.txt:65 (make_cuda_plugin)

repeated for every call to 'make_cuda_plugin' in the CMakeLists.txt file

Is there something I have done wrong?

Best regards,

Anders

29 okt. 2022 kl. 18:35 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

On 29 Oct 2022, at 17:46, Anders Genell <anders.genell@gmail.com> wrote:

On 29 Oct 2022, at 19:10, Anders Genell <anders.genell@gmail.com> wrote:

29 okt. 2022 kl. 20:28 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

 I predict more code rot ahead.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 19:10, Anders Genell <anders.genell@gmail.com> wrote:



So, according to an answer here: https://itecnote.com/tecnote/cmake-complains-about-wrong-number-of-arguments/

sometimes symbols are resolved to empty strings, which appear as missing arguments to cmake. Adding quotes around the symbols semms to have done the trick

from

    set_target_properties(${libname} PROPERTIES
        RUNTIME_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR}
        LIBRARY_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR}
        ARCHIVE_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR})

to

    set_target_properties(${libname} PROPERTIES
        RUNTIME_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}"
        LIBRARY_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}"
        ARCHIVE_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}")

Now for the actual compiling...

Regards,

/A

On Sat, Oct 29, 2022 at 7:21 PM Anders Genell <anders.genell@gmail.com> wrote:

A nice glass of chianti and some well-fermented vintage code is something of an acquired taste, but for the initiated it is among the most exquisite.

On Sat, Oct 29, 2022 at 7:05 PM Victor Lazzarini <Victor.Lazzarini@mu.ie> wrote:

10 years of code rot, soon it will be mature like blue cheese.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 17:46, Anders Genell <anders.genell@gmail.com> wrote:



Well, one shouldn’t rush in to these things… I’m sure another few years should only make it more stable and mature. 

29 okt. 2022 kl. 18:35 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

 I guess the CMakeLists.txt might need to be fixed for the version of CMake you have. No wonder, it's about 10 years old.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 15:33, Anders Genell <anders.genell@gmail.com> wrote:



*Warning*

This email originated from outside of Maynooth University's Mail System. Do not reply, click links or open attachments unless you recognise the sender and know the content is safe.

Hi!

I just got a new laptop at work which has a T500 nvidia graphics card and I managed to install all the cuda bells and whistles for it with some effort.

Now I am trying to have the csound cuda opcodes built, but seem to run into problems.  cmake finds the cuda installation allright, but there seems to be some trouble with the CUDA/CMakeLists.tx file. I get the following error:

CMake Error at CUDA/CMakeLists.txt:20 (set_target_properties):
  set_target_properties called with incorrect number of arguments.
Call Stack (most recent call first):
  CUDA/CMakeLists.txt:65 (make_cuda_plugin)

repeated for every call to 'make_cuda_plugin' in the CMakeLists.txt file

Is there something I have done wrong?

Best regards,

Anders



Quick update - I updated the CMakeLists.txt to specifically point to the csound include library in order to get rid of an error about not finding .h files and then it seems to have built the opcodes all right. 

I cannot seem to be able to get csound to find them however. I set the OPCODE6DIR as well as the OPCODE6DIR64 to point to where the plugins are, but csound does not find them, and specifying —opcode-lib on the command line didn’t make any difference either. I shall make sure I have built the opcodes against the correct version of csound, but I don’t think I have more than one installed…

Any ideas are greatly appreciated. 

Regards,

Anders 

29 okt. 2022 kl. 20:28 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

 I predict more code rot ahead.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 19:10, Anders Genell <anders.genell@gmail.com> wrote:



So, according to an answer here: https://itecnote.com/tecnote/cmake-complains-about-wrong-number-of-arguments/

sometimes symbols are resolved to empty strings, which appear as missing arguments to cmake. Adding quotes around the symbols semms to have done the trick

from

    set_target_properties(${libname} PROPERTIES
        RUNTIME_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR}
        LIBRARY_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR}
        ARCHIVE_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR})

to

    set_target_properties(${libname} PROPERTIES
        RUNTIME_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}"
        LIBRARY_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}"
        ARCHIVE_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}")

Now for the actual compiling...

Regards,

/A

On Sat, Oct 29, 2022 at 7:21 PM Anders Genell <anders.genell@gmail.com> wrote:

A nice glass of chianti and some well-fermented vintage code is something of an acquired taste, but for the initiated it is among the most exquisite.

On Sat, Oct 29, 2022 at 7:05 PM Victor Lazzarini <Victor.Lazzarini@mu.ie> wrote:

10 years of code rot, soon it will be mature like blue cheese.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 17:46, Anders Genell <anders.genell@gmail.com> wrote:



Well, one shouldn’t rush in to these things… I’m sure another few years should only make it more stable and mature. 

29 okt. 2022 kl. 18:35 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

 I guess the CMakeLists.txt might need to be fixed for the version of CMake you have. No wonder, it's about 10 years old.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 15:33, Anders Genell <anders.genell@gmail.com> wrote:



*Warning*

This email originated from outside of Maynooth University's Mail System. Do not reply, click links or open attachments unless you recognise the sender and know the content is safe.

Hi!

I just got a new laptop at work which has a T500 nvidia graphics card and I managed to install all the cuda bells and whistles for it with some effort.

Now I am trying to have the csound cuda opcodes built, but seem to run into problems.  cmake finds the cuda installation allright, but there seems to be some trouble with the CUDA/CMakeLists.tx file. I get the following error:

CMake Error at CUDA/CMakeLists.txt:20 (set_target_properties):
  set_target_properties called with incorrect number of arguments.
Call Stack (most recent call first):
  CUDA/CMakeLists.txt:65 (make_cuda_plugin)

repeated for every call to 'make_cuda_plugin' in the CMakeLists.txt file

Is there something I have done wrong?

Best regards,

Anders



1) The headers were found so all should be fine on that front

2) I did set the full path to the plugins but to no avail. It seems though that the cuda opcodes are built as static .a libs if that matters in any way. 

I will continue to investigate if there is something I have missed. 

Regards,

/A

sön 30 okt. 2022 kl. 19:16 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

A couple of things:

1) opcodes do not link to Csound. You only need the headers to build them.

2) with --opcode-lib= you need the full path so if the library is in the same directory you have

--opcode-lib=./library.so  

etc 

That one should make Csound acknowledge the library has been found.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 30 Oct 2022, at 15:09, Anders Genell <anders.genell@gmail.com> wrote:



Quick update - I updated the CMakeLists.txt to specifically point to the csound include library in order to get rid of an error about not finding .h files and then it seems to have built the opcodes all right. 

I cannot seem to be able to get csound to find them however. I set the OPCODE6DIR as well as the OPCODE6DIR64 to point to where the plugins are, but csound does not find them, and specifying —opcode-lib on the command line didn’t make any difference either. I shall make sure I have built the opcodes against the correct version of csound, but I don’t think I have more than one installed…

Any ideas are greatly appreciated. 

Regards,

Anders 

29 okt. 2022 kl. 20:28 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

 I predict more code rot ahead.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 19:10, Anders Genell <anders.genell@gmail.com> wrote:



So, according to an answer here: https://itecnote.com/tecnote/cmake-complains-about-wrong-number-of-arguments/

sometimes symbols are resolved to empty strings, which appear as missing arguments to cmake. Adding quotes around the symbols semms to have done the trick

from

    set_target_properties(${libname} PROPERTIES
        RUNTIME_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR}
        LIBRARY_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR}
        ARCHIVE_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR})

to

    set_target_properties(${libname} PROPERTIES
        RUNTIME_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}"
        LIBRARY_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}"
        ARCHIVE_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}")

Now for the actual compiling...

Regards,

/A

On Sat, Oct 29, 2022 at 7:21 PM Anders Genell <anders.genell@gmail.com> wrote:

A nice glass of chianti and some well-fermented vintage code is something of an acquired taste, but for the initiated it is among the most exquisite.

On Sat, Oct 29, 2022 at 7:05 PM Victor Lazzarini <Victor.Lazzarini@mu.ie> wrote:

10 years of code rot, soon it will be mature like blue cheese.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 17:46, Anders Genell <anders.genell@gmail.com> wrote:



Well, one shouldn’t rush in to these things… I’m sure another few years should only make it more stable and mature. 

29 okt. 2022 kl. 18:35 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

 I guess the CMakeLists.txt might need to be fixed for the version of CMake you have. No wonder, it's about 10 years old.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 15:33, Anders Genell <anders.genell@gmail.com> wrote:



*Warning*

This email originated from outside of Maynooth University's Mail System. Do not reply, click links or open attachments unless you recognise the sender and know the content is safe.

Hi!

I just got a new laptop at work which has a T500 nvidia graphics card and I managed to install all the cuda bells and whistles for it with some effort.

Now I am trying to have the csound cuda opcodes built, but seem to run into problems.  cmake finds the cuda installation allright, but there seems to be some trouble with the CUDA/CMakeLists.tx file. I get the following error:

CMake Error at CUDA/CMakeLists.txt:20 (set_target_properties):
  set_target_properties called with incorrect number of arguments.
Call Stack (most recent call first):
  CUDA/CMakeLists.txt:65 (make_cuda_plugin)

repeated for every call to 'make_cuda_plugin' in the CMakeLists.txt file

Is there something I have done wrong?

Best regards,

Anders



Ok, that is not something I think I will manage to figure out…

sön 30 okt. 2022 kl. 19:41 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

No, that sounds wrong. They can't be static libs. There's something not right with the cmake build.

The result should be a .so lib on Linux, a .dylib on MacOS and .dll on Windows. Anything else is wrong.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 30 Oct 2022, at 18:25, Anders Genell <anders.genell@gmail.com> wrote:



1) The headers were found so all should be fine on that front

2) I did set the full path to the plugins but to no avail. It seems though that the cuda opcodes are built as static .a libs if that matters in any way. 

I will continue to investigate if there is something I have missed. 

Regards,

/A

sön 30 okt. 2022 kl. 19:16 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

A couple of things:

1) opcodes do not link to Csound. You only need the headers to build them.

2) with --opcode-lib= you need the full path so if the library is in the same directory you have

--opcode-lib=./library.so  

etc 

That one should make Csound acknowledge the library has been found.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 30 Oct 2022, at 15:09, Anders Genell <anders.genell@gmail.com> wrote:



Quick update - I updated the CMakeLists.txt to specifically point to the csound include library in order to get rid of an error about not finding .h files and then it seems to have built the opcodes all right. 

I cannot seem to be able to get csound to find them however. I set the OPCODE6DIR as well as the OPCODE6DIR64 to point to where the plugins are, but csound does not find them, and specifying —opcode-lib on the command line didn’t make any difference either. I shall make sure I have built the opcodes against the correct version of csound, but I don’t think I have more than one installed…

Any ideas are greatly appreciated. 

Regards,

Anders 

29 okt. 2022 kl. 20:28 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

 I predict more code rot ahead.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 19:10, Anders Genell <anders.genell@gmail.com> wrote:



So, according to an answer here: https://itecnote.com/tecnote/cmake-complains-about-wrong-number-of-arguments/

sometimes symbols are resolved to empty strings, which appear as missing arguments to cmake. Adding quotes around the symbols semms to have done the trick

from

    set_target_properties(${libname} PROPERTIES
        RUNTIME_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR}
        LIBRARY_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR}
        ARCHIVE_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR})

to

    set_target_properties(${libname} PROPERTIES
        RUNTIME_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}"
        LIBRARY_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}"
        ARCHIVE_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}")

Now for the actual compiling...

Regards,

/A

On Sat, Oct 29, 2022 at 7:21 PM Anders Genell <anders.genell@gmail.com> wrote:

A nice glass of chianti and some well-fermented vintage code is something of an acquired taste, but for the initiated it is among the most exquisite.

On Sat, Oct 29, 2022 at 7:05 PM Victor Lazzarini <Victor.Lazzarini@mu.ie> wrote:

10 years of code rot, soon it will be mature like blue cheese.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 17:46, Anders Genell <anders.genell@gmail.com> wrote:



Well, one shouldn’t rush in to these things… I’m sure another few years should only make it more stable and mature. 

29 okt. 2022 kl. 18:35 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

 I guess the CMakeLists.txt might need to be fixed for the version of CMake you have. No wonder, it's about 10 years old.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 15:33, Anders Genell <anders.genell@gmail.com> wrote:



*Warning*

This email originated from outside of Maynooth University's Mail System. Do not reply, click links or open attachments unless you recognise the sender and know the content is safe.

Hi!

I just got a new laptop at work which has a T500 nvidia graphics card and I managed to install all the cuda bells and whistles for it with some effort.

Now I am trying to have the csound cuda opcodes built, but seem to run into problems.  cmake finds the cuda installation allright, but there seems to be some trouble with the CUDA/CMakeLists.tx file. I get the following error:

CMake Error at CUDA/CMakeLists.txt:20 (set_target_properties):
  set_target_properties called with incorrect number of arguments.
Call Stack (most recent call first):
  CUDA/CMakeLists.txt:65 (make_cuda_plugin)

repeated for every call to 'make_cuda_plugin' in the CMakeLists.txt file

Is there something I have done wrong?

Best regards,

Anders



Hm. I may have spoken too soon. It seems there are no segfaults but there is also no sound...

Score finishes with overall amps of 0.0

Oh well...

Regards,

/A

On Sun, Oct 30, 2022 at 9:02 PM Anders Genell <anders.genell@gmail.com> wrote:

Ok, after a bit of googling and a bit of hacking I changed to build.sh script as follows:

#!/bin/sh
# use -Xptxas="-v" to check register usage and --maxrregcount 32 to limit it
echo "building cuda opcodes ..."
#nvcc -O3 -shared -o libcudaop1.dylib adsyn.cu -use_fast_math -I../../debug/CsoundLib64.framework/Headers -arch=sm_30 -I/usr/local/cuda/include -L/usr/local/cuda/lib
#nvcc -O3 -shared -o libcudaop2.dylib pvsops.cu -use_fast_math -g -I../../debug/CsoundLib64.framework/Headers -arch=sm_30 -I/usr/local/cuda/include -L/usr/local/cuda/lib -lcufft
#nvcc -O3 -shared -o libcudaop3.dylib slidingm.cu -use_fast_math  -I../../debug/CsoundLib64.framework/Headers -arch=sm_30 -I/usr/local/cuda/include -L/usr/local/cuda/lib
#nvcc -O3 -shared -o libcudaop4.dylib conv.cu   -I../../debug/CsoundLib64.framework/Headers -arch=sm_30 -I/usr/local/cuda/include -L/usr/local/cuda/lib
#nvcc -O3 -shared -o libcudaop5.dylib pconv.cu -I../../debug/CsoundLib64.framework/Headers -arch=sm_30 -I/usr/local/cuda/include -L/usr/local/cuda/lib -lcufft
nvcc -O3 -Xcompiler -fPIC -shared -o libcudaop1.so adsyn.cu -use_fast_math -I/usr/local/include/csound -arch=sm_75 -I/usr/local/cuda/include -L/usr/local/cuda/lib
nvcc -O3 -Xcompiler -fPIC -shared -o libcudaop2.so pvsops.cu -use_fast_math -g -I/usr/local/include/csound -arch=sm_75 -I/usr/local/cuda/include -L/usr/local/cuda/lib -lcufft
nvcc -O3 -Xcompiler -fPIC -shared -o libcudaop3.so slidingm.cu -use_fast_math  -I/usr/local/include/csound -arch=sm_75 -I/usr/local/cuda/include -L/usr/local/cuda/lib
nvcc -O3 -Xcompiler -fPIC -shared -o libcudaop4.so conv.cu   -I/usr/local/include/csound -arch=sm_75 -I/usr/local/cuda/include -L/usr/local/cuda/lib
nvcc -O3 -Xcompiler -fPIC -shared -o libcudaop5.so pconv.cu -I/usr/local/include/csound -arch=sm_75 -I/usr/local/cuda/include -L/usr/local/cuda/lib -lcufft
echo "...done"

Before I added -Xcompiler -fPIC I got errors saying

/usr/bin/ld: /tmp/tmpxft_0000e6be_00000000-11_adsyn.o: warning: relocation against `_Z6samplePfS_fPlS_if' in read-only section `.text.startup'
/usr/bin/ld: /tmp/tmpxft_0000e6be_00000000-11_adsyn.o: relocation R_X86_64_PC32 against symbol `_Z6samplePfS_fPlS_if' can not be used when making a shared object; recompile with -fPIC
/usr/bin/ld: final link failed: bad value
collect2: error: ld returned 1 exit status

but some googling said what to add and then the build ran fine. cudapconv seems to work, and if checking with nvidia-smi while running csound with cudapconv I get the following output:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 520.61.05    Driver Version: 520.61.05    CUDA Version: 11.8     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA T500         On   | 00000000:01:00.0 Off |                  N/A |
| N/A   48C    P8    N/A /  N/A |     84MiB /  4096MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A      2648      G   /usr/bin/gnome-shell                1MiB |
|    0   N/A  N/A     59817      C   csound                             80MiB |
+-----------------------------------------------------------------------------+

So it seems to work, but there are no guarantees there aren't any nasties I missed so someone who actually knows what they're doing should definately have a look at things...

Thank you again!

Regards,

Anders

On Sun, Oct 30, 2022 at 8:06 PM Anders Genell <anders.genell@gmail.com> wrote:

I’ll give that a go and see what I can manage.

Thanks!

Regards,

/A

sön 30 okt. 2022 kl. 19:52 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

alternatively you could ditch cmake and build the opcodes without. There is a script there showing how to do in on macos, it should be a case of adapting it to your linux paths and library name.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 30 Oct 2022, at 18:44, Anders Genell <anders.genell@gmail.com> wrote:



Ok, that is not something I think I will manage to figure out…

sön 30 okt. 2022 kl. 19:41 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

No, that sounds wrong. They can't be static libs. There's something not right with the cmake build.

The result should be a .so lib on Linux, a .dylib on MacOS and .dll on Windows. Anything else is wrong.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 30 Oct 2022, at 18:25, Anders Genell <anders.genell@gmail.com> wrote:



1) The headers were found so all should be fine on that front

2) I did set the full path to the plugins but to no avail. It seems though that the cuda opcodes are built as static .a libs if that matters in any way. 

I will continue to investigate if there is something I have missed. 

Regards,

/A

sön 30 okt. 2022 kl. 19:16 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

A couple of things:

1) opcodes do not link to Csound. You only need the headers to build them.

2) with --opcode-lib= you need the full path so if the library is in the same directory you have

--opcode-lib=./library.so  

etc 

That one should make Csound acknowledge the library has been found.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 30 Oct 2022, at 15:09, Anders Genell <anders.genell@gmail.com> wrote:



Quick update - I updated the CMakeLists.txt to specifically point to the csound include library in order to get rid of an error about not finding .h files and then it seems to have built the opcodes all right. 

I cannot seem to be able to get csound to find them however. I set the OPCODE6DIR as well as the OPCODE6DIR64 to point to where the plugins are, but csound does not find them, and specifying —opcode-lib on the command line didn’t make any difference either. I shall make sure I have built the opcodes against the correct version of csound, but I don’t think I have more than one installed…

Any ideas are greatly appreciated. 

Regards,

Anders 

29 okt. 2022 kl. 20:28 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

 I predict more code rot ahead.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 19:10, Anders Genell <anders.genell@gmail.com> wrote:



So, according to an answer here: https://itecnote.com/tecnote/cmake-complains-about-wrong-number-of-arguments/

sometimes symbols are resolved to empty strings, which appear as missing arguments to cmake. Adding quotes around the symbols semms to have done the trick

from

    set_target_properties(${libname} PROPERTIES
        RUNTIME_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR}
        LIBRARY_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR}
        ARCHIVE_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR})

to

    set_target_properties(${libname} PROPERTIES
        RUNTIME_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}"
        LIBRARY_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}"
        ARCHIVE_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}")

Now for the actual compiling...

Regards,

/A

On Sat, Oct 29, 2022 at 7:21 PM Anders Genell <anders.genell@gmail.com> wrote:

A nice glass of chianti and some well-fermented vintage code is something of an acquired taste, but for the initiated it is among the most exquisite.

On Sat, Oct 29, 2022 at 7:05 PM Victor Lazzarini <Victor.Lazzarini@mu.ie> wrote:

10 years of code rot, soon it will be mature like blue cheese.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 17:46, Anders Genell <anders.genell@gmail.com> wrote:



Well, one shouldn’t rush in to these things… I’m sure another few years should only make it more stable and mature. 

29 okt. 2022 kl. 18:35 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

 I guess the CMakeLists.txt might need to be fixed for the version of CMake you have. No wonder, it's about 10 years old.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 15:33, Anders Genell <anders.genell@gmail.com> wrote:



*Warning*

This email originated from outside of Maynooth University's Mail System. Do not reply, click links or open attachments unless you recognise the sender and know the content is safe.

Hi!

I just got a new laptop at work which has a T500 nvidia graphics card and I managed to install all the cuda bells and whistles for it with some effort.

Now I am trying to have the csound cuda opcodes built, but seem to run into problems.  cmake finds the cuda installation allright, but there seems to be some trouble with the CUDA/CMakeLists.tx file. I get the following error:

CMake Error at CUDA/CMakeLists.txt:20 (set_target_properties):
  set_target_properties called with incorrect number of arguments.
Call Stack (most recent call first):
  CUDA/CMakeLists.txt:65 (make_cuda_plugin)

repeated for every call to 'make_cuda_plugin' in the CMakeLists.txt file

Is there something I have done wrong?

Best regards,

Anders

On 31 Oct 2022, at 16:29, Anders Genell <anders.genell@gmail.com> wrote:

31 okt. 2022 kl. 18:45 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:



Ah, yes…

I wonder how/where I can find what sm_XX setting to use?

If XX should correspond to compute ability I believe it is correct now, as I have compute ability 7.5 (according to nvidia-smi) and have set sm_75 in the build options. 

Are there other possibilities?

Regards,

Anders

31 okt. 2022 kl. 18:45 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

 maybe check that the build options match your device

Prof. Victor Lazzarini

Maynooth University

Ireland

On 31 Oct 2022, at 16:29, Anders Genell <anders.genell@gmail.com> wrote:



Well, unfortunately ftconv gives me plenty of amplitude, so it seems to be the cudapconv opcode that doesn't work properly here...

Regards,

Anders

On Sun, Oct 30, 2022 at 10:18 PM Victor Lazzarini <Victor.Lazzarini@mu.ie> wrote:

no sound could be to do with the impulse response (if it's cudapconv). If you look at the csd there is ftconv there to test alongside it.

You should be able to swap one for the other and test.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 30 Oct 2022, at 20:09, Anders Genell <anders.genell@gmail.com> wrote:



Hm. I may have spoken too soon. It seems there are no segfaults but there is also no sound...

Score finishes with overall amps of 0.0

Oh well...

Regards,

/A

On Sun, Oct 30, 2022 at 9:02 PM Anders Genell <anders.genell@gmail.com> wrote:

Ok, after a bit of googling and a bit of hacking I changed to build.sh script as follows:

#!/bin/sh
# use -Xptxas="-v" to check register usage and --maxrregcount 32 to limit it
echo "building cuda opcodes ..."
#nvcc -O3 -shared -o libcudaop1.dylib adsyn.cu -use_fast_math -I../../debug/CsoundLib64.framework/Headers -arch=sm_30 -I/usr/local/cuda/include -L/usr/local/cuda/lib
#nvcc -O3 -shared -o libcudaop2.dylib pvsops.cu -use_fast_math -g -I../../debug/CsoundLib64.framework/Headers -arch=sm_30 -I/usr/local/cuda/include -L/usr/local/cuda/lib -lcufft
#nvcc -O3 -shared -o libcudaop3.dylib slidingm.cu -use_fast_math  -I../../debug/CsoundLib64.framework/Headers -arch=sm_30 -I/usr/local/cuda/include -L/usr/local/cuda/lib
#nvcc -O3 -shared -o libcudaop4.dylib conv.cu   -I../../debug/CsoundLib64.framework/Headers -arch=sm_30 -I/usr/local/cuda/include -L/usr/local/cuda/lib
#nvcc -O3 -shared -o libcudaop5.dylib pconv.cu -I../../debug/CsoundLib64.framework/Headers -arch=sm_30 -I/usr/local/cuda/include -L/usr/local/cuda/lib -lcufft
nvcc -O3 -Xcompiler -fPIC -shared -o libcudaop1.so adsyn.cu -use_fast_math -I/usr/local/include/csound -arch=sm_75 -I/usr/local/cuda/include -L/usr/local/cuda/lib
nvcc -O3 -Xcompiler -fPIC -shared -o libcudaop2.so pvsops.cu -use_fast_math -g -I/usr/local/include/csound -arch=sm_75 -I/usr/local/cuda/include -L/usr/local/cuda/lib -lcufft
nvcc -O3 -Xcompiler -fPIC -shared -o libcudaop3.so slidingm.cu -use_fast_math  -I/usr/local/include/csound -arch=sm_75 -I/usr/local/cuda/include -L/usr/local/cuda/lib
nvcc -O3 -Xcompiler -fPIC -shared -o libcudaop4.so conv.cu   -I/usr/local/include/csound -arch=sm_75 -I/usr/local/cuda/include -L/usr/local/cuda/lib
nvcc -O3 -Xcompiler -fPIC -shared -o libcudaop5.so pconv.cu -I/usr/local/include/csound -arch=sm_75 -I/usr/local/cuda/include -L/usr/local/cuda/lib -lcufft
echo "...done"

Before I added -Xcompiler -fPIC I got errors saying

/usr/bin/ld: /tmp/tmpxft_0000e6be_00000000-11_adsyn.o: warning: relocation against `_Z6samplePfS_fPlS_if' in read-only section `.text.startup'
/usr/bin/ld: /tmp/tmpxft_0000e6be_00000000-11_adsyn.o: relocation R_X86_64_PC32 against symbol `_Z6samplePfS_fPlS_if' can not be used when making a shared object; recompile with -fPIC
/usr/bin/ld: final link failed: bad value
collect2: error: ld returned 1 exit status

but some googling said what to add and then the build ran fine. cudapconv seems to work, and if checking with nvidia-smi while running csound with cudapconv I get the following output:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 520.61.05    Driver Version: 520.61.05    CUDA Version: 11.8     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA T500         On   | 00000000:01:00.0 Off |                  N/A |
| N/A   48C    P8    N/A /  N/A |     84MiB /  4096MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A      2648      G   /usr/bin/gnome-shell                1MiB |
|    0   N/A  N/A     59817      C   csound                             80MiB |
+-----------------------------------------------------------------------------+

So it seems to work, but there are no guarantees there aren't any nasties I missed so someone who actually knows what they're doing should definately have a look at things...

Thank you again!

Regards,

Anders

On Sun, Oct 30, 2022 at 8:06 PM Anders Genell <anders.genell@gmail.com> wrote:

I’ll give that a go and see what I can manage.

Thanks!

Regards,

/A

sön 30 okt. 2022 kl. 19:52 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

alternatively you could ditch cmake and build the opcodes without. There is a script there showing how to do in on macos, it should be a case of adapting it to your linux paths and library name.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 30 Oct 2022, at 18:44, Anders Genell <anders.genell@gmail.com> wrote:



Ok, that is not something I think I will manage to figure out…

sön 30 okt. 2022 kl. 19:41 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

No, that sounds wrong. They can't be static libs. There's something not right with the cmake build.

The result should be a .so lib on Linux, a .dylib on MacOS and .dll on Windows. Anything else is wrong.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 30 Oct 2022, at 18:25, Anders Genell <anders.genell@gmail.com> wrote:



1) The headers were found so all should be fine on that front

2) I did set the full path to the plugins but to no avail. It seems though that the cuda opcodes are built as static .a libs if that matters in any way. 

I will continue to investigate if there is something I have missed. 

Regards,

/A

sön 30 okt. 2022 kl. 19:16 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

A couple of things:

1) opcodes do not link to Csound. You only need the headers to build them.

2) with --opcode-lib= you need the full path so if the library is in the same directory you have

--opcode-lib=./library.so  

etc 

That one should make Csound acknowledge the library has been found.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 30 Oct 2022, at 15:09, Anders Genell <anders.genell@gmail.com> wrote:



Quick update - I updated the CMakeLists.txt to specifically point to the csound include library in order to get rid of an error about not finding .h files and then it seems to have built the opcodes all right. 

I cannot seem to be able to get csound to find them however. I set the OPCODE6DIR as well as the OPCODE6DIR64 to point to where the plugins are, but csound does not find them, and specifying —opcode-lib on the command line didn’t make any difference either. I shall make sure I have built the opcodes against the correct version of csound, but I don’t think I have more than one installed…

Any ideas are greatly appreciated. 

Regards,

Anders 

29 okt. 2022 kl. 20:28 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

 I predict more code rot ahead.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 19:10, Anders Genell <anders.genell@gmail.com> wrote:



So, according to an answer here: https://itecnote.com/tecnote/cmake-complains-about-wrong-number-of-arguments/

sometimes symbols are resolved to empty strings, which appear as missing arguments to cmake. Adding quotes around the symbols semms to have done the trick

from

    set_target_properties(${libname} PROPERTIES
        RUNTIME_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR}
        LIBRARY_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR}
        ARCHIVE_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR})

to

    set_target_properties(${libname} PROPERTIES
        RUNTIME_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}"
        LIBRARY_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}"
        ARCHIVE_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}")

Now for the actual compiling...

Regards,

/A

On Sat, Oct 29, 2022 at 7:21 PM Anders Genell <anders.genell@gmail.com> wrote:

A nice glass of chianti and some well-fermented vintage code is something of an acquired taste, but for the initiated it is among the most exquisite.

On Sat, Oct 29, 2022 at 7:05 PM Victor Lazzarini <Victor.Lazzarini@mu.ie> wrote:

10 years of code rot, soon it will be mature like blue cheese.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 17:46, Anders Genell <anders.genell@gmail.com> wrote:



Well, one shouldn’t rush in to these things… I’m sure another few years should only make it more stable and mature. 

29 okt. 2022 kl. 18:35 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

 I guess the CMakeLists.txt might need to be fixed for the version of CMake you have. No wonder, it's about 10 years old.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 15:33, Anders Genell <anders.genell@gmail.com> wrote:



*Warning*

This email originated from outside of Maynooth University's Mail System. Do not reply, click links or open attachments unless you recognise the sender and know the content is safe.

Hi!

I just got a new laptop at work which has a T500 nvidia graphics card and I managed to install all the cuda bells and whistles for it with some effort.

Now I am trying to have the csound cuda opcodes built, but seem to run into problems.  cmake finds the cuda installation allright, but there seems to be some trouble with the CUDA/CMakeLists.tx file. I get the following error:

CMake Error at CUDA/CMakeLists.txt:20 (set_target_properties):
  set_target_properties called with incorrect number of arguments.
Call Stack (most recent call first):
  CUDA/CMakeLists.txt:65 (make_cuda_plugin)

repeated for every call to 'make_cuda_plugin' in the CMakeLists.txt file

Is there something I have done wrong?

Best regards,

Anders



Alas, no. Still amps 0.0

However, I also tried the other cuda opcode examples and lo and behold, the cuda_pvs_example.csd resulted in sound output, and it also announced the use of the nvidia device:

UnifiedCSD:  cuda_pvs_example.csd
Elapsed time at end of orchestra compile: real: 0.001s, CPU: 0.001s
sorting score ...
... done
Elapsed time at end of score sort: real: 0.001s, CPU: 0.001s
--Csound version 6.17 (double samples) May 29 2022
[commit: 95007834d8afc68b9ae465a6f65b93b23ae41709]
libsndfile-1.1.0
graphics suppressed, ascii substituted
sr = 44100.0, kr = 689.062, ksmps = 64
0dBFS level = 1.0, A4 tuning = 440.0
orch now loaded
audio buffered in 256 sample-frame blocks
ALSA output: total buffer size: 1024, period size: 256
writing 256 sample blks of 64-bit floats to dac
SECTION 1:
new alloc for instr 1:
diskin2: opened 'SpaceFunk20_91bpm_mono.wav':
         48000 Hz, 1 channel(s), 506373 sample frames

CUDAnal: using device NVIDIA T500 (capability 7.5)
CUDAsynth: using device NVIDIA T500 (capability 7.5)
B  0.000 .. 60.000 T 60.000 TT 60.000 M:  1.00535
number of samples out of range:       36
B 60.000 .. 60.011 T 60.011 TT 60.011 M:  0.69522
Score finished in csoundPerform().
inactive allocs returned to freespace
end of score.   overall amps:  1.00535
  overall samples out of range:       36
0 errors in performance
Elapsed time at end of performance: real: 60.062s, CPU: 2.196s
10338 256 sample blks of 64-bit floats written to dac

***************************************************************

The cuda_sliding_example.csd announced the use of the nvidia device but outputs no sound:

UnifiedCSD:  cuda_sliding_example.csd
Loading command-line libraries:
Elapsed time at end of orchestra compile: real: 0.001s, CPU: 0.001s
sorting score ...
... done
Elapsed time at end of score sort: real: 0.001s, CPU: 0.001s
--Csound version 6.17 (double samples) May 29 2022
[commit: 95007834d8afc68b9ae465a6f65b93b23ae41709]
libsndfile-1.1.0
graphics suppressed, ascii substituted
sr = 44100.0, kr = 689.062, ksmps = 64
0dBFS level = 1.0, A4 tuning = 440.0
orch now loaded
audio buffered in 256 sample-frame blocks
ALSA output: total buffer size: 1024, period size: 256
writing 256 sample blks of 64-bit floats to dac
SECTION 1:
new alloc for instr 1:
diskin2: opened 'SpaceFunk20_91bpm.wav':
         48000 Hz, 2 channel(s), 506373 sample frames

Sliding PV: using floats on device NVIDIA T500 (capability 7.5)
B  0.000 .. 60.000 T 60.000 TT 60.000 M:  0.00000
B 60.000 .. 60.011 T 60.011 TT 60.011 M:  0.00000
Score finished in csoundPerform().
inactive allocs returned to freespace
end of score.   overall amps:  0.00000
  overall samples out of range:        0
0 errors in performance
Elapsed time at end of performance: real: 60.064s, CPU: 2.208s
10338 256 sample blks of 64-bit floats written to dac

***************************************************************

Finally, cudadsyn_pvs_example.csd behaves like the cudapconv example; no announcement about nvidia device and no sound:

UnifiedCSD:  cudadsyn_pvs_example.csd
Loading command-line libraries:
Elapsed time at end of orchestra compile: real: 0.001s, CPU: 0.001s
sorting score ...
... done
Elapsed time at end of score sort: real: 0.001s, CPU: 0.001s
--Csound version 6.17 (double samples) May 29 2022
[commit: 95007834d8afc68b9ae465a6f65b93b23ae41709]
libsndfile-1.1.0
graphics suppressed, ascii substituted
sr = 44100.0, kr = 689.062, ksmps = 64
0dBFS level = 1.0, A4 tuning = 440.0
orch now loaded
audio buffered in 256 sample-frame blocks
ALSA output: total buffer size: 1024, period size: 256
writing 256 sample blks of 64-bit floats to dac
SECTION 1:
new alloc for instr 1:
diskin2: opened 'SpaceFunk20_91bpm.wav':
         48000 Hz, 2 channel(s), 506373 sample frames

B  0.000 .. 60.000 T 60.000 TT 60.000 M:  0.00000
B 60.000 .. 60.011 T 60.011 TT 60.011 M:  0.00000
Score finished in csoundPerform().
inactive allocs returned to freespace
end of score.   overall amps:  0.00000
  overall samples out of range:        0
0 errors in performance
Elapsed time at end of performance: real: 60.049s, CPU: 3.513s
10338 256 sample blks of 64-bit floats written to dac

Regards,

Anders

On Mon, Oct 31, 2022 at 8:13 PM Victor Lazzarini <Victor.Lazzarini@mu.ie> wrote:

ok I think I spotted a need for updating, maybe you can try this:

Line 244 of pconv.cu 

  {"cudapconv", sizeof(PCONV),0, 5, "a", "aii", (SUBR) pconv_init, NULL,

    (SUBR) pconv_perf},

should be

  {"cudapconv", sizeof(PCONV),0, 5, "a", "aii", (SUBR) pconv_init,  (SUBR) pconv_perf},

See if that makes any difference.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 31 Oct 2022, at 18:34, Anders Genell <anders.genell@gmail.com> wrote:



Ah, yes…

I wonder how/where I can find what sm_XX setting to use?

If XX should correspond to compute ability I believe it is correct now, as I have compute ability 7.5 (according to nvidia-smi) and have set sm_75 in the build options. 

Are there other possibilities?

Regards,

Anders

31 okt. 2022 kl. 18:45 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

 maybe check that the build options match your device

Prof. Victor Lazzarini

Maynooth University

Ireland

On 31 Oct 2022, at 16:29, Anders Genell <anders.genell@gmail.com> wrote:



Well, unfortunately ftconv gives me plenty of amplitude, so it seems to be the cudapconv opcode that doesn't work properly here...

Regards,

Anders

On Sun, Oct 30, 2022 at 10:18 PM Victor Lazzarini <Victor.Lazzarini@mu.ie> wrote:

no sound could be to do with the impulse response (if it's cudapconv). If you look at the csd there is ftconv there to test alongside it.

You should be able to swap one for the other and test.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 30 Oct 2022, at 20:09, Anders Genell <anders.genell@gmail.com> wrote:



Hm. I may have spoken too soon. It seems there are no segfaults but there is also no sound...

Score finishes with overall amps of 0.0

Oh well...

Regards,

/A

On Sun, Oct 30, 2022 at 9:02 PM Anders Genell <anders.genell@gmail.com> wrote:

Ok, after a bit of googling and a bit of hacking I changed to build.sh script as follows:

#!/bin/sh
# use -Xptxas="-v" to check register usage and --maxrregcount 32 to limit it
echo "building cuda opcodes ..."
#nvcc -O3 -shared -o libcudaop1.dylib adsyn.cu -use_fast_math -I../../debug/CsoundLib64.framework/Headers -arch=sm_30 -I/usr/local/cuda/include -L/usr/local/cuda/lib
#nvcc -O3 -shared -o libcudaop2.dylib pvsops.cu -use_fast_math -g -I../../debug/CsoundLib64.framework/Headers -arch=sm_30 -I/usr/local/cuda/include -L/usr/local/cuda/lib -lcufft
#nvcc -O3 -shared -o libcudaop3.dylib slidingm.cu -use_fast_math  -I../../debug/CsoundLib64.framework/Headers -arch=sm_30 -I/usr/local/cuda/include -L/usr/local/cuda/lib
#nvcc -O3 -shared -o libcudaop4.dylib conv.cu   -I../../debug/CsoundLib64.framework/Headers -arch=sm_30 -I/usr/local/cuda/include -L/usr/local/cuda/lib
#nvcc -O3 -shared -o libcudaop5.dylib pconv.cu -I../../debug/CsoundLib64.framework/Headers -arch=sm_30 -I/usr/local/cuda/include -L/usr/local/cuda/lib -lcufft
nvcc -O3 -Xcompiler -fPIC -shared -o libcudaop1.so adsyn.cu -use_fast_math -I/usr/local/include/csound -arch=sm_75 -I/usr/local/cuda/include -L/usr/local/cuda/lib
nvcc -O3 -Xcompiler -fPIC -shared -o libcudaop2.so pvsops.cu -use_fast_math -g -I/usr/local/include/csound -arch=sm_75 -I/usr/local/cuda/include -L/usr/local/cuda/lib -lcufft
nvcc -O3 -Xcompiler -fPIC -shared -o libcudaop3.so slidingm.cu -use_fast_math  -I/usr/local/include/csound -arch=sm_75 -I/usr/local/cuda/include -L/usr/local/cuda/lib
nvcc -O3 -Xcompiler -fPIC -shared -o libcudaop4.so conv.cu   -I/usr/local/include/csound -arch=sm_75 -I/usr/local/cuda/include -L/usr/local/cuda/lib
nvcc -O3 -Xcompiler -fPIC -shared -o libcudaop5.so pconv.cu -I/usr/local/include/csound -arch=sm_75 -I/usr/local/cuda/include -L/usr/local/cuda/lib -lcufft
echo "...done"

Before I added -Xcompiler -fPIC I got errors saying

/usr/bin/ld: /tmp/tmpxft_0000e6be_00000000-11_adsyn.o: warning: relocation against `_Z6samplePfS_fPlS_if' in read-only section `.text.startup'
/usr/bin/ld: /tmp/tmpxft_0000e6be_00000000-11_adsyn.o: relocation R_X86_64_PC32 against symbol `_Z6samplePfS_fPlS_if' can not be used when making a shared object; recompile with -fPIC
/usr/bin/ld: final link failed: bad value
collect2: error: ld returned 1 exit status

but some googling said what to add and then the build ran fine. cudapconv seems to work, and if checking with nvidia-smi while running csound with cudapconv I get the following output:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 520.61.05    Driver Version: 520.61.05    CUDA Version: 11.8     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA T500         On   | 00000000:01:00.0 Off |                  N/A |
| N/A   48C    P8    N/A /  N/A |     84MiB /  4096MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A      2648      G   /usr/bin/gnome-shell                1MiB |
|    0   N/A  N/A     59817      C   csound                             80MiB |
+-----------------------------------------------------------------------------+

So it seems to work, but there are no guarantees there aren't any nasties I missed so someone who actually knows what they're doing should definately have a look at things...

Thank you again!

Regards,

Anders

On Sun, Oct 30, 2022 at 8:06 PM Anders Genell <anders.genell@gmail.com> wrote:

I’ll give that a go and see what I can manage.

Thanks!

Regards,

/A

sön 30 okt. 2022 kl. 19:52 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

alternatively you could ditch cmake and build the opcodes without. There is a script there showing how to do in on macos, it should be a case of adapting it to your linux paths and library name.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 30 Oct 2022, at 18:44, Anders Genell <anders.genell@gmail.com> wrote:



Ok, that is not something I think I will manage to figure out…

sön 30 okt. 2022 kl. 19:41 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

No, that sounds wrong. They can't be static libs. There's something not right with the cmake build.

The result should be a .so lib on Linux, a .dylib on MacOS and .dll on Windows. Anything else is wrong.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 30 Oct 2022, at 18:25, Anders Genell <anders.genell@gmail.com> wrote:



1) The headers were found so all should be fine on that front

2) I did set the full path to the plugins but to no avail. It seems though that the cuda opcodes are built as static .a libs if that matters in any way. 

I will continue to investigate if there is something I have missed. 

Regards,

/A

sön 30 okt. 2022 kl. 19:16 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

A couple of things:

1) opcodes do not link to Csound. You only need the headers to build them.

2) with --opcode-lib= you need the full path so if the library is in the same directory you have

--opcode-lib=./library.so  

etc 

That one should make Csound acknowledge the library has been found.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 30 Oct 2022, at 15:09, Anders Genell <anders.genell@gmail.com> wrote:



Quick update - I updated the CMakeLists.txt to specifically point to the csound include library in order to get rid of an error about not finding .h files and then it seems to have built the opcodes all right. 

I cannot seem to be able to get csound to find them however. I set the OPCODE6DIR as well as the OPCODE6DIR64 to point to where the plugins are, but csound does not find them, and specifying —opcode-lib on the command line didn’t make any difference either. I shall make sure I have built the opcodes against the correct version of csound, but I don’t think I have more than one installed…

Any ideas are greatly appreciated. 

Regards,

Anders 

29 okt. 2022 kl. 20:28 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

 I predict more code rot ahead.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 19:10, Anders Genell <anders.genell@gmail.com> wrote:



So, according to an answer here: https://itecnote.com/tecnote/cmake-complains-about-wrong-number-of-arguments/

sometimes symbols are resolved to empty strings, which appear as missing arguments to cmake. Adding quotes around the symbols semms to have done the trick

from

    set_target_properties(${libname} PROPERTIES
        RUNTIME_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR}
        LIBRARY_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR}
        ARCHIVE_OUTPUT_DIRECTORY ${BUILD_PLUGINS_DIR})

to

    set_target_properties(${libname} PROPERTIES
        RUNTIME_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}"
        LIBRARY_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}"
        ARCHIVE_OUTPUT_DIRECTORY "${BUILD_PLUGINS_DIR}")

Now for the actual compiling...

Regards,

/A

On Sat, Oct 29, 2022 at 7:21 PM Anders Genell <anders.genell@gmail.com> wrote:

A nice glass of chianti and some well-fermented vintage code is something of an acquired taste, but for the initiated it is among the most exquisite.

On Sat, Oct 29, 2022 at 7:05 PM Victor Lazzarini <Victor.Lazzarini@mu.ie> wrote:

10 years of code rot, soon it will be mature like blue cheese.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 17:46, Anders Genell <anders.genell@gmail.com> wrote:



Well, one shouldn’t rush in to these things… I’m sure another few years should only make it more stable and mature. 

29 okt. 2022 kl. 18:35 skrev Victor Lazzarini <Victor.Lazzarini@mu.ie>:

 I guess the CMakeLists.txt might need to be fixed for the version of CMake you have. No wonder, it's about 10 years old.

Prof. Victor Lazzarini

Maynooth University

Ireland

On 29 Oct 2022, at 15:33, Anders Genell <anders.genell@gmail.com> wrote:



*Warning*

This email originated from outside of Maynooth University's Mail System. Do not reply, click links or open attachments unless you recognise the sender and know the content is safe.

Hi!

I just got a new laptop at work which has a T500 nvidia graphics card and I managed to install all the cuda bells and whistles for it with some effort.

Now I am trying to have the csound cuda opcodes built, but seem to run into problems.  cmake finds the cuda installation allright, but there seems to be some trouble with the CUDA/CMakeLists.tx file. I get the following error:

CMake Error at CUDA/CMakeLists.txt:20 (set_target_properties):
  set_target_properties called with incorrect number of arguments.
Call Stack (most recent call first):
  CUDA/CMakeLists.txt:65 (make_cuda_plugin)

repeated for every call to 'make_cuda_plugin' in the CMakeLists.txt file

Is there something I have done wrong?

Best regards,

Anders