Opened 2 years ago
Last modified 22 months ago
#9779 new defect
ffmpeg compile with ffnvcodec
| Reported by: | santosh | Owned by: | |
|---|---|---|---|
| Priority: | normal | Component: | undetermined |
| Version: | 4.4.3 | Keywords: | msvc |
| Cc: | santosh | Blocked By: | |
| Blocking: | ffmpeg exe is not compiled | Reproduced by developer: | yes |
| Analyzed by developer: | no |
Description
Summary of the bug:
I am trying to compile ffmpeg 4.4.2 or 4.4.1 with ffnvcodec to use hardware acceleration.
Platform: WINDOWS 10
How to reproduce:
./configure --enable-nonfree --disable-shared --enable-ffnvcodec --enable-cuda-llvm --enable-gpl --disable-debug --enable-cuda-nvcc --enable-libnpp --toolchain=msvc --extra-cflags=-I../nv_sdk --extra-ldflags=-libpath:../nv_sdk --nvccflags="-gencode arch=compute_61,code=sm_61 -O2"
make -j 8
OUTPUT ERROR received in the end:
libavfilter/vf_scale_cuda_bicubic.ptx.c(2842): fatal error C1060: compiler is out of heap space
make: * [ffbuild/common.mak:67: libavfilter/vf_scale_cuda_bicubic.ptx.o] Error 2
similar existing ticket #9019(avfilter:new): CUDA Compile Broken Using MSVC
provided patch fix with xxd. But it not working for me on windows.
Please provide a fix or patch for above issue.
Change History (2)
comment:2 by , 22 months ago
| Priority: | important → normal |
|---|
Not a regression or a crash, so does not qualify as important.
4.4.2 or 4.4.1 are not supported here, only master branch is. And this is a duplicate.
Is a bug in the compiler. Moreover it is strange since windows heap is much better than linux. Like it can expand stack and use heap for stack.
Anyway, should not you use clang or nvcc for PTX code? if you are using nvcc (looks like) then this is a bug nvidia+msvc even.