Commit message (Collapse) | Author | Age | |
---|---|---|---|
* | d/rules: Enable CONFIG_LOWBITDEPTH | James Cowgill | 2019-09-01 |
| | | | | | | This enables the use of the faster 8-bit pipeline if possible. It also seems to help avoid crashes when using the ARM NEON intrinsics, but I'm unsure about the underlying cause of those. | ||
* | d/rules: Manually set AOM_NEON_INTRIN_FLAG on armhf | James Cowgill | 2019-09-01 |
| | | | | | | | | | | | | | | | | | aom contains some ARM optimisations which use NEON intrinsics. Before 1.0.0.eratta1-1, these were disabled due to a typo in the target CPU, but now we set the CPU properly they attempt to build but fail with errors such as: /<<PKGBUILDDIR>>/aom_dsp/arm/sad4d_neon.c: In function ‘horizontal_long_add_16x8’: /usr/lib/gcc/arm-linux-gnueabihf/9/include/arm_neon.h:6188:1: error: inlining failed in call to always_inline ‘vget_lane_u32’: target specific option mismatch 6188 | vget_lane_u32 (uint32x2_t __a, const int __b) | ^~~~~~~~~~~~~ This happens because we need to pass the "-mfpu=neon" flag to those files. For some reason, upstream sets this flag in its CMake toolchain files which are not used by Debian. Temporarily fix this by setting the flag in debian/rules. | ||
* | d/rules: Manually specify target cpu and enable multithreading | James Cowgill | 2019-08-29 |
| | | | | | | | | | In v1.0.0-errata1, the build system has been changed to always default to a non-optimised build, so we now have to set the target cpu manually for all platforms. Also re-enable multithreading to revert another defaults change in v1.0.0-errata1. | ||
* | d/rules: Disable AltiVec on powerpcspe | John Paul Adrian Glaubitz | 2018-10-04 |
| | | | | Closes: #909950 | ||
* | d/rules: Disable optimizations on x32 to fix FTBFS | James Cowgill | 2018-09-20 |
| | |||
* | d/rules: workaround broken 32-bit ARM architecture detection | James Cowgill | 2018-09-20 |
| | |||
* | Add initial Debian packaging | James Cowgill | 2018-08-12 |