Hi,
On Sat, Nov 12, 2011 at 2:25 PM, Martin Storsjö <[email protected]> wrote:
> On Sat, 12 Nov 2011, Ronald S. Bultje wrote:
>
>> On Sat, Nov 12, 2011 at 2:18 PM, Martin Storsjö <[email protected]> wrote:
>>>
>>> On Fri, 11 Nov 2011, Ronald S. Bultje wrote:
>>>
>>>> Hi,
>>>>
>>>> On Fri, Nov 11, 2011 at 3:27 PM, Martin Storsjö <[email protected]>
>>>> wrote:
>>>>>
>>>>> On Fri, 11 Nov 2011, Ronald S. Bultje wrote:
>>>>>
>>>>>> As said, this is for the case where we read planar YUV, then
>>>>>> SwsContext->convertData[] is not used, instead the planar pointer is
>>>>>> used directly without any conversion. This is used-provided and may
>>>>>> thus be unaligned (we currently make no alignment requirements).
>>>>>
>>>>> Hmm, since this commit last Saturday
>>>>>
>>>>>> Commit: c435653627529e22d74214c2266f571255e404d6
>>>>>>
>>>>>> Author: Ronald S. Bultje <[email protected]>
>>>>>> Committer: Ronald S. Bultje <[email protected]>
>>>>>> Date: Sat Nov 5 17:31:40 2011 -0700
>>>>>>
>>>>>> swscale: write yuv2plane1 MMX/SSE2/SSE4/AVX functions.
>>>>>
>>>>> code that passes unaligned buffers to swscale started crashing. Do I
>>>>> read
>>>>> the the comment above correctly, this is a bug - calling code shouldn't
>>>>> need
>>>>> to provide any particular alignment?
>>>>
>>>> Yeah, if you tell me which instruction fails I'll fix it, or a
>>>> backtrace + disass is fine also.
>>>
>>> On one machine, I get this backtrace + disassembly:
>>>
>>> #0 ff_yuv2plane1_8_avx.loop_a () at libswscale/x86/scale.asm:804
>>> #1 0x00000000007acdd1 in swScale (c=0xed3300, src=0x7fffffffe4e0,
>>> srcStride=0x7fffffffe500, srcSliceY=0, srcSliceH=<value optimized
>>> out>,
>>> dst=0x7fffffffe4c0, dstStride=0x7fffffffe510) at
>>> libswscale/swscale.c:2510
>>> #2 0x0000000000792d5d in sws_scale (c=0xed3300, srcSlice=0xec5c00,
>>> srcStride=<value optimized out>, srcSliceY=0, srcSliceH=19,
>>> dst=<value optimized out>, dstStride=0xecb240)
>>> at libswscale/swscale_unscaled.c:913
>>>
>>> Dump of assembler code for function ff_yuv2plane1_8_avx.loop_a:
>>> 0x00000000007c264e <+0>: vpaddsw (%rdi,%rdx,2),%xmm2,%xmm0
>>> 0x00000000007c2653 <+5>: vpaddsw 0x10(%rdi,%rdx,2),%xmm3,%xmm1
>>> 0x00000000007c2659 <+11>: psraw $0x7,%xmm0
>>> 0x00000000007c265e <+16>: psraw $0x7,%xmm1
>>> 0x00000000007c2663 <+21>: packuswb %xmm1,%xmm0
>>> => 0x00000000007c2667 <+25>: movdqa %xmm0,(%rsi,%rdx,1)
>>> 0x00000000007c266c <+30>: add $0x10,%rdx
>>> 0x00000000007c2670 <+34>: jl 0x7c264e
>>> <ff_yuv2plane1_8_avx.loop_a>
>>> 0x00000000007c2672 <+36>: repz retq
>>> End of assembler dump.
>>
>> info all-registers? I wonder what rsi/rdx are here.
>
> rax 0xed9370 15569776
> rbx 0xed3300 15545088
> rcx 0x9212f0 9573104
> rdx 0xffffffffffffff9c -100
Ah, your width is not a multiple of 16. Does adding these two lines at
the bottom of scale.asm fix it?
%macro yuv2plane1_fn 3
cglobal yuv2plane1_%1, %3, %3, %2
+ add r2, mmsize - 1
+ and r2, ~(mmsize - 1)
%if %1 == 8
add r1, r2
Ronald
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel
