]> granicus.if.org Git - libvpx/commit
Save NEON registers in VP8 NEON functions
authorYunqing Wang <yunqingwang@google.com>
Mon, 28 Apr 2014 21:42:23 +0000 (14:42 -0700)
committerYunqing Wang <yunqingwang@google.com>
Mon, 28 Apr 2014 21:51:53 +0000 (14:51 -0700)
commit33df6d1fc1d268b4901b74b4141f83594266f041
tree87031fef033550336f5d32c3cfe55ae4cf6d2e09
parent5ba44e37a470be7ec74f717c293cfcb864c84a0d
Save NEON registers in VP8 NEON functions

The recent compiler can generate optimized code that uses NEON registers
for various operations besides floating-point operations. Therefore,
only saving callee-saved registers d8 - d15 at the beginning of the
encoder/decoder is not enough anymore. This patch added register saving
code in VP8 NEON functions that use those registers.

Change-Id: Ie9e44f5188cf410990c8aaaac68faceee9dffd31
21 files changed:
vp8/common/arm/neon/buildintrapredictorsmby_neon.asm
vp8/common/arm/neon/idct_dequant_0_2x_neon.asm
vp8/common/arm/neon/idct_dequant_full_2x_neon.asm
vp8/common/arm/neon/loopfilter_neon.asm
vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.asm
vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.asm
vp8/common/arm/neon/mbloopfilter_neon.asm
vp8/common/arm/neon/sad16_neon.asm
vp8/common/arm/neon/sad8_neon.asm
vp8/common/arm/neon/shortidct4x4llm_neon.asm
vp8/common/arm/neon/sixtappredict16x16_neon.asm
vp8/common/arm/neon/sixtappredict4x4_neon.asm
vp8/common/arm/neon/sixtappredict8x4_neon.asm
vp8/common/arm/neon/sixtappredict8x8_neon.asm
vp8/common/arm/neon/variance_neon.asm
vp8/common/arm/neon/vp8_subpixelvariance16x16_neon.asm
vp8/common/arm/neon/vp8_subpixelvariance16x16s_neon.asm
vp8/common/arm/neon/vp8_subpixelvariance8x8_neon.asm
vp8/encoder/arm/neon/subtract_neon.asm
vp8/encoder/arm/neon/vp8_memcpy_neon.asm
vp8/encoder/arm/neon/vp8_mse16x16_neon.asm