Use none SSE functions when data is not 16 byte aligned

Use none SSE functions when data is not 16 byte aligned

BUG=
R=jabdelmalek@google.com

Committed: https://pdfium.googlesource.com/pdfium/+/0db35e8

[Bo Xu, 2014-06-04 06:58:10]

Please review the change, thanks

[Lei Zhang, 2014-06-04 09:11:28]

The openjpeg code tries to make sure the wavelet is actually aligned. It calls
opj_aligned_malloc() rather than just opj_malloc().

However, with _FOXIT_MEM_MANAGER_ defined in opj_malloc.h and j2k.c,
opj_aligned_malloc() is equivalent to opj_malloc(), which just calls
FXMEM_DefaultAlloc() and gives back non-aligned memory allocations.

I'm not familiar with FXMEM_DefaultAlloc() internals, but is there a way to tell
it to align the allocated memory? If not, you'll need to do actually implement
opj_aligned_malloc() and opj_aligned_free(), so that they can take the memory
allocated by FXMEM_DefaultAlloc() and do the alignment.

[Bo Xu, 2014-06-05 18:14:48]

Ok, we are looking at implementing aligned malloc in the memory manager.

[jabdelmalek, 2014-06-06 17:25:31]

https://codereview.chromium.org/318593002/diff/20001/core/src/fxcodec/fx_libo...
File core/src/fxcodec/fx_libopenjpeg/libopenjpeg20/dwt.c (right):

https://codereview.chromium.org/318593002/diff/20001/core/src/fxcodec/fx_libo...
core/src/fxcodec/fx_libopenjpeg/libopenjpeg20/dwt.c:144: #if 0
here and below where you comment this out, please add "http://crbug.com/373619"
in a comment so that someone looking at this knows why it's commented out

https://codereview.chromium.org/318593002/diff/20001/core/src/fxcodec/fx_libo...
File core/src/fxcodec/fx_libopenjpeg/libopenjpeg20/mct.c (left):

https://codereview.chromium.org/318593002/diff/20001/core/src/fxcodec/fx_libo...
core/src/fxcodec/fx_libopenjpeg/libopenjpeg20/mct.c:192: } else {
why remove this?

[Bo Xu, 2014-06-07 01:30:00]

On 2014/06/06 17:25:31, jabdelmalek wrote:
>
https://codereview.chromium.org/318593002/diff/20001/core/src/fxcodec/fx_libo...
> File core/src/fxcodec/fx_libopenjpeg/libopenjpeg20/dwt.c (right):
> 
>
https://codereview.chromium.org/318593002/diff/20001/core/src/fxcodec/fx_libo...
> core/src/fxcodec/fx_libopenjpeg/libopenjpeg20/dwt.c:144: #if 0
> here and below where you comment this out, please add
"http://crbug.com/373619"
> in a comment so that someone looking at this knows why it's commented out

I will add the comment. 

> 
>
https://codereview.chromium.org/318593002/diff/20001/core/src/fxcodec/fx_libo...
> File core/src/fxcodec/fx_libopenjpeg/libopenjpeg20/mct.c (left):
> 
>
https://codereview.chromium.org/318593002/diff/20001/core/src/fxcodec/fx_libo...
> core/src/fxcodec/fx_libopenjpeg/libopenjpeg20/mct.c:192: } else {
> why remove this?

This was a fix for previous security bugs, but it computes the color twice, so
the "else" part should be removed. I can actually land it on another issue. It's
kinda irrelevance here.

[jam, 2014-06-07 01:31:18]

On 2014/06/07 01:30:00, bo_xu wrote:
> On 2014/06/06 17:25:31, jabdelmalek wrote:
> >
>
https://codereview.chromium.org/318593002/diff/20001/core/src/fxcodec/fx_libo...
> > File core/src/fxcodec/fx_libopenjpeg/libopenjpeg20/dwt.c (right):
> > 
> >
>
https://codereview.chromium.org/318593002/diff/20001/core/src/fxcodec/fx_libo...
> > core/src/fxcodec/fx_libopenjpeg/libopenjpeg20/dwt.c:144: #if 0
> > here and below where you comment this out, please add
> "http://crbug.com/373619"
> > in a comment so that someone looking at this knows why it's commented out
> 
> I will add the comment. 
> 
> > 
> >
>
https://codereview.chromium.org/318593002/diff/20001/core/src/fxcodec/fx_libo...
> > File core/src/fxcodec/fx_libopenjpeg/libopenjpeg20/mct.c (left):
> > 
> >
>
https://codereview.chromium.org/318593002/diff/20001/core/src/fxcodec/fx_libo...
> > core/src/fxcodec/fx_libopenjpeg/libopenjpeg20/mct.c:192: } else {
> > why remove this?
> 
> This was a fix for previous security bugs, but it computes the color twice, so
> the "else" part should be removed. I can actually land it on another issue.
It's
> kinda irrelevance here.

yes, please put that separately. thanks.

[jabdelmalek, 2014-06-08 22:58:15]

lgtm

[Bo Xu, 2014-06-08 23:35:52]

Committed patchset #3 manually as r0db35e8 (presubmit successful).