Age | Commit message (Collapse) | Author |
|
Add aarch64 architecture support in configure.ac.
|
|
Add small neon code snipet to check the compiler neon support in
acinclude.m4
|
|
|
|
Add armv8 simd supported function list using macro
"ARMV8_NEON_SIMULATION". All decoder related simd functions
are supported.
|
|
Add armv8 yuv-rgb conversion, macros generate_jsimd_ycc_rgb_convert_neon
have been modified to support armv8 instruction and register literals.
RTSM integer saturation instruction issue workaround added.
|
|
Modified existing idct_2x2 armv7 implementation to support armv8
|
|
Armv8 instruction changes ported to idct_4x4 armv7 implementation
idct_helper macro changed to accomodate the new instruction and
register literals.
|
|
Idct fast armv8 instruction changes, additional mov instructions
are added due to the changes in armv8 lower to higher register
overlapping. vswp instruction of armv7 is removed in armv8 and
swap in armv8 port done manually.
|
|
Armv8 spcific instruction changes incorprated on armv7 neon idct_slow
code, instructions like orrs doenot exist in armv8 and cpsr flags
needs to be updated with alternate instruction like cmp.
RTSM simulator specific changes are incorprated as some instructions
particularly related to integer saturation doenot work as expected.
vtrans( armv7 instruction) is split into trn1 and trn2 due to which
transpose_4x4 macro is modified and a new transpose macro replaces
the orginal vtrn1 instruction for armv8. Etra ins instruction are
added to move data to higher half of 128bit registers due to non
overlap of lower registers to higher registers as in armv7. Data
movement to higher half of registers can be optimized further.
|
|
Baselined libjpeg-turbo 1.3 release version for android
refresh
|