Commit d5dd8c7b authored by Janne Grunau's avatar Janne Grunau

aarch64: h264 qpel NEON optimizations

Ported from ARMv7 NEON.
parent 8438b3f0
OBJS-$(CONFIG_H264CHROMA) += aarch64/h264chroma_init_aarch64.o
OBJS-$(CONFIG_H264DSP) += aarch64/h264dsp_init_aarch64.o
OBJS-$(CONFIG_H264QPEL) += aarch64/h264qpel_init_aarch64.o
OBJS-$(CONFIG_RV40_DECODER) += aarch64/rv40dsp_init_aarch64.o
OBJS-$(CONFIG_VC1_DECODER) += aarch64/vc1dsp_init_aarch64.o
NEON-OBJS-$(CONFIG_H264CHROMA) += aarch64/h264cmc_neon.o
NEON-OBJS-$(CONFIG_H264DSP) += aarch64/h264idct_neon.o
NEON-OBJS-$(CONFIG_H264QPEL) += aarch64/h264qpel_neon.o
This diff is collapsed.
This diff is collapsed.
......@@ -16,6 +16,70 @@
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
.macro transpose_8x8B r0, r1, r2, r3, r4, r5, r6, r7, r8, r9
trn1 \r8\().8B, \r0\().8B, \r1\().8B
trn2 \r9\().8B, \r0\().8B, \r1\().8B
trn1 \r1\().8B, \r2\().8B, \r3\().8B
trn2 \r3\().8B, \r2\().8B, \r3\().8B
trn1 \r0\().8B, \r4\().8B, \r5\().8B
trn2 \r5\().8B, \r4\().8B, \r5\().8B
trn1 \r2\().8B, \r6\().8B, \r7\().8B
trn2 \r7\().8B, \r6\().8B, \r7\().8B
trn1 \r4\().4H, \r0\().4H, \r2\().4H
trn2 \r2\().4H, \r0\().4H, \r2\().4H
trn1 \r6\().4H, \r5\().4H, \r7\().4H
trn2 \r7\().4H, \r5\().4H, \r7\().4H
trn1 \r5\().4H, \r9\().4H, \r3\().4H
trn2 \r9\().4H, \r9\().4H, \r3\().4H
trn1 \r3\().4H, \r8\().4H, \r1\().4H
trn2 \r8\().4H, \r8\().4H, \r1\().4H
trn1 \r0\().2S, \r3\().2S, \r4\().2S
trn2 \r4\().2S, \r3\().2S, \r4\().2S
trn1 \r1\().2S, \r5\().2S, \r6\().2S
trn2 \r5\().2S, \r5\().2S, \r6\().2S
trn2 \r6\().2S, \r8\().2S, \r2\().2S
trn1 \r2\().2S, \r8\().2S, \r2\().2S
trn1 \r3\().2S, \r9\().2S, \r7\().2S
trn2 \r7\().2S, \r9\().2S, \r7\().2S
.endm
.macro transpose_8x16B r0, r1, r2, r3, r4, r5, r6, r7, t0, t1
trn1 \t0\().16B, \r0\().16B, \r1\().16B
trn2 \t1\().16B, \r0\().16B, \r1\().16B
trn1 \r1\().16B, \r2\().16B, \r3\().16B
trn2 \r3\().16B, \r2\().16B, \r3\().16B
trn1 \r0\().16B, \r4\().16B, \r5\().16B
trn2 \r5\().16B, \r4\().16B, \r5\().16B
trn1 \r2\().16B, \r6\().16B, \r7\().16B
trn2 \r7\().16B, \r6\().16B, \r7\().16B
trn1 \r4\().8H, \r0\().8H, \r2\().8H
trn2 \r2\().8H, \r0\().8H, \r2\().8H
trn1 \r6\().8H, \r5\().8H, \r7\().8H
trn2 \r7\().8H, \r5\().8H, \r7\().8H
trn1 \r5\().8H, \t1\().8H, \r3\().8H
trn2 \t1\().8H, \t1\().8H, \r3\().8H
trn1 \r3\().8H, \t0\().8H, \r1\().8H
trn2 \t0\().8H, \t0\().8H, \r1\().8H
trn1 \r0\().4S, \r3\().4S, \r4\().4S
trn2 \r4\().4S, \r3\().4S, \r4\().4S
trn1 \r1\().4S, \r5\().4S, \r6\().4S
trn2 \r5\().4S, \r5\().4S, \r6\().4S
trn2 \r6\().4S, \t0\().4S, \r2\().4S
trn1 \r2\().4S, \t0\().4S, \r2\().4S
trn1 \r3\().4S, \t1\().4S, \r7\().4S
trn2 \r7\().4S, \t1\().4S, \r7\().4S
.endm
.macro transpose_4x4H r0, r1, r2, r3, r4, r5, r6, r7
trn1 \r4\().4H, \r0\().4H, \r1\().4H
trn2 \r5\().4H, \r0\().4H, \r1\().4H
......
......@@ -78,6 +78,8 @@ av_cold void ff_h264qpel_init(H264QpelContext *c, int bit_depth)
break;
}
if (ARCH_AARCH64)
ff_h264qpel_init_aarch64(c, bit_depth);
if (ARCH_ARM)
ff_h264qpel_init_arm(c, bit_depth);
if (ARCH_PPC)
......
......@@ -31,6 +31,7 @@ typedef struct H264QpelContext {
void ff_h264qpel_init(H264QpelContext *c, int bit_depth);
void ff_h264qpel_init_aarch64(H264QpelContext *c, int bit_depth);
void ff_h264qpel_init_arm(H264QpelContext *c, int bit_depth);
void ff_h264qpel_init_ppc(H264QpelContext *c, int bit_depth);
void ff_h264qpel_init_x86(H264QpelContext *c, int bit_depth);
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment