On Sat, Feb 07, 2015 at 04:14:11PM +0100, Christophe Gisquet wrote:
Hi,
2015-02-06 18:33 GMT+01:00 James Almer jamr...@gmail.com:
Yes, that should do it.
And for that matter you were right, it should be 4, 6. I was reading it
as regs
loaded when the second argument is regs needed.
Hi,
2015-02-06 18:33 GMT+01:00 James Almer jamr...@gmail.com:
Yes, that should do it.
And for that matter you were right, it should be 4, 6. I was reading it as
regs
loaded when the second argument is regs needed.
Ok, here's an updated patch.
I also fixed the unneeded mov on UNIX64, as
2015-02-06 16:18 GMT+01:00 James Almer jamr...@gmail.com:
mov r5d, denomm is loading it on Win64. It's also a superfluous instruction
on
UNIX64, where it translates to mov r5d, r5d.
At least if i'm reading this right.
movifnidn then?
--
Christophe
On 06/02/15 6:14 AM, Christophe Gisquet wrote:
2015-02-05 20:20 GMT+01:00 Christophe Gisquet christophe.gisq...@gmail.com:
-cglobal hevc_put_hevc_bi_w%1_%2, 5, 7, 10, dst, dststride, src,
srcstride, src2, height, denom, wx0, wx1, ox0, ox1
-mov r6d, denomm
+cglobal
On 06/02/15 1:05 PM, Christophe Gisquet wrote:
2015-02-06 16:18 GMT+01:00 James Almer jamr...@gmail.com:
mov r5d, denomm is loading it on Win64. It's also a superfluous
instruction on
UNIX64, where it translates to mov r5d, r5d.
At least if i'm reading this right.
movifnidn then?
Yes,
The second stride is always the internal buffer one, MAX_PB_SIZE (times 2 to
get the value in bytes).
---
libavcodec/x86/hevc_mc.asm| 30 +++---
libavcodec/x86/hevcdsp.h | 4 ++--
libavcodec/x86/hevcdsp_init.c | 16
3 files changed, 25