Enable LTO and use __restrict on Degrain_C: 9-13% faster Degrain6 on 16-bit sources

[adworacz]

As it says on the tin - this turns on Link Time Optimization officially (which has been the default on Arch Linux for a while now).

Additionally, added the __restrict keyword to the Degrain_C function, which enables compilers to better optimize memory access/register usage. This only impacts 16-bit sources, as 8-bit uses the SSE2 implementation and isn't effected.

All outputs are identical between the master-clang, lto-clang, and lto-restrict-clang libraries. Used Clang as the compiler.

Tests were run with blocksize 8 and 16, on 4k, 1080p, and 540p content, with Degrain6 and Analyse (no Recalculate).

-o 1 == blocksize 8 -o 2 == blocksize 16

Results:

1080p and 540p:

Command	Mean [s]	Min [s]	Max [s]	Relative
vspipe -p -e 3000 -o 1 --arg mvversion=master-clang --arg src=test-1080p.dgi tester.vpy /dev/null	148.705 ± 0.281	148.385	148.913	7.30 ± 0.02
vspipe -p -e 3000 -o 1 --arg mvversion=lto-clang --arg src=test-1080p.dgi tester.vpy /dev/null	132.927 ± 0.056	132.867	132.978	6.53 ± 0.01
vspipe -p -e 3000 -o 1 --arg mvversion=lto-restrict-clang --arg src=test-1080p.dgi tester.vpy /dev/null	131.146 ± 0.173	131.022	131.343	6.44 ± 0.01
vspipe -p -e 3000 -o 2 --arg mvversion=master-clang --arg src=test-1080p.dgi tester.vpy /dev/null	76.785 ± 0.046	76.755	76.838	3.77 ± 0.00
vspipe -p -e 3000 -o 2 --arg mvversion=lto-clang --arg src=test-1080p.dgi tester.vpy /dev/null	72.265 ± 0.012	72.257	72.279	3.55 ± 0.00
vspipe -p -e 3000 -o 2 --arg mvversion=lto-restrict-clang --arg src=test-1080p.dgi tester.vpy /dev/null	70.463 ± 0.014	70.446	70.471	3.46 ± 0.00
vspipe -p -e 3000 -o 1 --arg mvversion=master-clang --arg src=test-540p.dgi tester.vpy /dev/null	41.942 ± 0.026	41.917	41.969	2.06 ± 0.00
vspipe -p -e 3000 -o 1 --arg mvversion=lto-clang --arg src=test-540p.dgi tester.vpy /dev/null	38.550 ± 0.028	38.532	38.583	1.89 ± 0.00
vspipe -p -e 3000 -o 1 --arg mvversion=lto-restrict-clang --arg src=test-540p.dgi tester.vpy /dev/null	37.955 ± 0.024	37.941	37.982	1.86 ± 0.00
vspipe -p -e 3000 -o 2 --arg mvversion=master-clang --arg src=test-540p.dgi tester.vpy /dev/null	21.950 ± 0.012	21.937	21.958	1.08 ± 0.00
vspipe -p -e 3000 -o 2 --arg mvversion=lto-clang --arg src=test-540p.dgi tester.vpy /dev/null	21.101 ± 0.022	21.083	21.126	1.04 ± 0.00
vspipe -p -e 3000 -o 2 --arg mvversion=lto-restrict-clang --arg src=test-540p.dgi tester.vpy /dev/null	20.362 ± 0.019	20.346	20.383	1.00

4k:

Command	Mean [s]	Min [s]	Max [s]	Relative
vspipe -p -e 500 -o 1 --arg mvversion=master-clang --arg src=test-4k.dgi tester.vpy /dev/null	110.780 ± 0.016	110.770	110.799	2.05 ± 0.00
vspipe -p -e 500 -o 1 --arg mvversion=lto-clang --arg src=test-4k.dgi tester.vpy /dev/null	106.270 ± 12.193	99.224	120.349	1.96 ± 0.23
vspipe -p -e 500 -o 1 --arg mvversion=lto-restrict-clang --arg src=test-4k.dgi tester.vpy /dev/null	97.966 ± 0.076	97.887	98.039	1.81 ± 0.00
vspipe -p -e 500 -o 2 --arg mvversion=master-clang --arg src=test-4k.dgi tester.vpy /dev/null	58.903 ± 0.004	58.900	58.907	1.09 ± 0.00
vspipe -p -e 500 -o 2 --arg mvversion=lto-clang --arg src=test-4k.dgi tester.vpy /dev/null	55.515 ± 0.048	55.463	55.559	1.03 ± 0.00
vspipe -p -e 500 -o 2 --arg mvversion=lto-restrict-clang --arg src=test-4k.dgi tester.vpy /dev/null	54.124 ± 0.041	54.089	54.169	1.00

Note how the lto-restrict-clang is always the fastest of the bunch, sometimes by 3-4% just by using __restrict, while enabling LTO can have a 6-11% improvement on its own.

[dubhater]

Thanks!