Backport updates from dav1d 1.5.1 #1372
Description
The following are the commits from 1.5.1. Note that this is a raw list of commits and so some of these may not be relevant for rav1d.
- Include <string.h> for memcpy()
- AArch64: Add Neon implementation of load_tmvs
- mc: Reduce stack use in {put,prep}scaled{bilin,8tap}
- riscv: Fix building on non-Linux OS's
- obu: don't print warnings for Metadata OBUs of types "Unregistered user private"
- arm32: looprestoration: Rewrite the wiener functions
- looprestoration: Use only 6 row buffer for wiener, like NEON/x86
- looprestoration: Make the C wiener h filter more optimizable for the compiler
- looprestoration: Rewrite the C version of the wiener filter
- Replace Av1Block with pal_sz in read_pal_indices
- Explicitly use uint8_t for the order_palette() scratch buffer
- flush: Reset f->task_thread.error
- Fix ClangTidy misc-include-cleaner warnings
- checkasm: looprestoration: Do strict bounds checking of the output
- riscv: Enable FreeBSD / OpenBSD elf_aux_info() support
- arm32: looprestoration: Rewrite the SGR functions
- arm32: looprestoration: Apply simplifications to align with C code
- arm64: looprestoration: Fix a comment typo
- arm: looprestoration: Fix the single line loop in sgr_weighted2
- looprestoration: Rewrite the C version of the SGR filter
- arm: looprestoration: Give symbols and defines unique names
- arm: looprestoration: Add spacing around operators
- arm: looprestoration: Get rid of unnecessary rotate_ab_N intermediate functions
- arm: looprestoration: Apply 'const' more consistently on parameters
- checkasm: add loongarch GAS file to checkasm_asm_sources
- Fix comments on Dav1dMasteringDisplay
- arm64: looprestoration: Fix use of the wrong register
- arm: Use /proc/cpuinfo on linux if getauxval is unavailable
- ci: Raise the timeout multipliers for jobs that run in QEMU
- arm64: looprestoration: Remove an unnecessary duplicate parameter in dav1d_sgr_weighted2_Xbpc_neon
- arm64: looprestoration: Apply simplifications to align with C code
- arm: looprestoration: Split an overly long line
- x86: Use the decl and init macros for put_8tap and prep_8tap
- loongarch: Use the decl and init macros for put_8tap and prep_8tap
- mc: Factor out the decl and init macros
- ppc: use a jumptable for the blends
- ppc: blend_h pwr9 implementation
- ppc: blend_v pwr9 implementation
- ppc: blend pwr9 implementation
- Provide dav1d_getauxval() wrapper for getauxvaul() and elf_aux_info()
- riscv64/mc: Only process w*3/4 elements in blend_v
- riscv64/mc16: Unroll 16bpc RVV blend_v 2x
- riscv64/mc16: Branchless vsetvl in blend_v function
- riscv64/mc16: Add VLEN=256 8bpc RVV blend_v function
- riscv64/mc16: Add 16bpc RVV blend_v function
- riscv64/mc16: Unroll 16bpc RVV blend 2x
- riscv64/mc16: Branchless vsetvl in blend function
- riscv64/mc16: Add VLEN=256 8bpc RVV blend function
- meson: Move riscv64 8bpc only files into bitdepth sources
- riscv64/mc16: Add 16bpc RVV blend function
- riscv64/ipred16: Fix build error with -Dbitdepths=16
- x86: Improve SSSE3 SGR asm
- checkasm: Remove leftover comment
- meson: Test support for aarch64 extensions with gas-preprocessor too
- meson: Move checks for gas-preprocessor earlier