- Marcar como nuevo
- Favorito
- Suscribir
- Silenciar
- Suscribirse a un feed RSS
- Resaltar
- Imprimir
- Informe de contenido inapropiado
Dear Intel,
I have tried using ippiFilterSeparable_32f_C1R_L but encountered the following problem:
- the function writes one row after the end of the available size in the destination buffer. I verified this by creating a bigger destination buffer with predetermined value. After the function I checked which pixels outside the allowed roi have been changed.
Here is my use case:
- IPP Version 2018.2 64 bits on Windows 10 Pro x64
- CPU Intel i7 3770
- dataType = ipp32f, kernelDataType = ipp32f, numChannels = 1, roiSize = { 192, 192 }, kernelSize = { 7, 7 }, srcStep 768
Is this forum the correct place to submit such issues?
Regards Nikolai
Enlace copiado
- Marcar como nuevo
- Favorito
- Suscribir
- Silenciar
- Suscribirse a un feed RSS
- Resaltar
- Imprimir
- Informe de contenido inapropiado
Hi Nikolai,
thank you, we will investigate the issue.
Pavel
- Marcar como nuevo
- Favorito
- Suscribir
- Silenciar
- Suscribirse a un feed RSS
- Resaltar
- Imprimir
- Informe de contenido inapropiado
Hello Nikolai.
I confirm your issue. It is one of latest new features of IPP. As workaround allocate please 1 additional line in destination image. We'll fix this issue in next release of IPP.
Thank you very much for information.
- Marcar como nuevo
- Favorito
- Suscribir
- Silenciar
- Suscribirse a un feed RSS
- Resaltar
- Imprimir
- Informe de contenido inapropiado
Isn't that a memory alignment issue?
Does the problem persist if you allocate the destination buffer with ippsMalloc_32f_L?
Something like:
Ipp32f* Dest = ippsMalloc_32f_L(height*width*sizeof(Ipp32f));

- Suscribirse a un feed RSS
- Marcar tema como nuevo
- Marcar tema como leído
- Flotar este Tema para el usuario actual
- Favorito
- Suscribir
- Página de impresión sencilla