HCenc 022 released

[Nomolu]

Like many of the other posts, I'm glad that SMP support has been added to HCenc. It made it possible to find the following problem in half the time. :sly: Sometimes the MPEG2 stream created by HCenc 0.22 results in the following error when the audio and video are being muxed by the dvdauthoring s/w I use (which is GUI for DVDauthor in the Muxman mode):

P-STD buffer underflow by 2639 bytes at 25736978, sector 109169.
P-STD buffer underflow by 6032 bytes at 25739981, sector 109189.
...
Resulting DVD will probably not play smoothly.


If I switch GfD back to it's "dvdauthor" mode where it uses mplex to do the muxing, I get the following warnings:

++ WARN: [???] Stream e0: data will arrive too late sent(SCR)=25730779 required(DTS)=0
++ WARN: [???] Video e0: buf= 193632 frame=008564 sector=00105224
++ WARN: [???] Audio bd: buf= 7363 frame=008940 sector=00003406
...
**ERROR: [???] MUX STATUS: Frame data under-runs detected!

And I'm left with a corrupted, unplayable DVD files.

My system has two Althon MP processors, and I've tried running HC 0.22 with SMP disabled. But it didn't fix the problem. Running HC 0.21 produced a fine/valid MPEG2 file. It just took twice as long. :(

By the process of elimination, I found that there is one minute of frantic action in the video that HC 0.22 is having trouble encoding. Is this the result of a bug within the VBV portion of HC?

Here are my encoding settings:

profile: BEST
frames: 0 8990
framerate: 29.970
aspect ratio: 4:3
bitrate Kb/s: 6000
max. bitrate Kb/s: 8000
pulldown: no
closed gops: no
VBV check: yes
scene change det.: yes
interlaced: yes, TFF
goplen,B-pic: AUTO
dc_precision: 9
scan method: alt
bias: 0
chapter frames: 0
time code: 0 0 0 0
CPU: MMX
priority: idle
SMP active: yes
matrix: MPEG
luminance gain: no

FYI: It makes no difference if I specify interlaced or if I set HC to auto. The only MPEG profile which produces a valid MPEG2 stream is the "normal" setting. Fast & Best produce corrupt streams.

BTW, I prefer the old VB Verifer tally that was displayed in versions 0.20 and earlier to the new flickering LED type of display. Please consider changing it back. At the very least, could the tally be written out to the log file again? Thanks.

[Mazer Rackham]

HC Enc 21 gave me an error message I've never had before, that's why I searched and found the new HC Enc 22. Same error message unfortunately:

ERROR, can't read from dbs file, aborting...

It happens during 2nd pass:

---------------------
| encoding - pass 1 |
---------------------

pass 1 encoding time: 1:48:32 (6511.76 s)
fps: 27.6

--------------------------------
| encoding - intermediate pass |
--------------------------------

bitrate set to: 4000 kb/s
est. outfile length: 3506836 kB
intermediate encoding time: 1.0 s

---------------------
| encoding - pass 2 |
---------------------


*** ERROR, can't read from dbs file, aborting...

*** ERROR, can't read from dbs file, aborting...

---
Any ideas what's causing the error?

[blutach]

A belated thank you from me Hank!

Having no problems at all using AVISynth 2.5.8

Very much appreciate the full use of multi cored CPUs.

Regards

[pandy]

Thank You very much. Superb piece of software!

[hank315]

@Nomolu
I tried to replicate this with some difficult sources, all muxed OK.
Seems I need your source for it, could you make it available?

@Mazer Rackham
It means HC can't read the database file which was written in the first pass, most probable cause the disk is full??
With HC022 you can redirect the dbs file to another disk.

[Mazer Rackham]

@Mazer Rackham
It means HC can't read the database file which was written in the first pass, most probable cause the disk is full??
With HC022 you can redirect the dbs file to another disk.

Thanks for your reply, hank315!

Hmm, strange. The video I try to encode is about 4 Gig big and I have about 11 Gig disc space left. I will increase the amount of free disc space and try again though.

Edit: Okay, freeing more disc space solved the problem. Thanks again! Good to know: 11 Gig free disc space is not enough for a 2-pass encoding of a 4 Gig video.

[Nomolu]

@Nomolu
I tried to replicate this with some difficult sources, all muxed OK.
Seems I need your source for it, could you make it available?


hank315: I just sent you a PM with the upload links.

[MrC]

@hank315

Is it possible to have the "AVSRELOAD" argument from the CLI in the next release?

:thanks:

Bye

[manolito]

Even though I have been following HC's development from the very beginning I have rarely used it for productive tasks, simply because my computer is rather slow and HC does not offer a true one-pass mode. But now it looks like Santa will have a new computer for me, so I did a couple of test encodes with HC 0.22 during the last few days. There were no problems whatsoever, great quality, and Muxman did not complain one single time even though I always used the "Fast" profile.

I use HC with DVD2SVCD and with DVD-RB free. In both cases HC is controlled by QuEnc commandline parameters, and I have full control over the HC.ini settings.

I still could not find answers in the forum for a couple of questions I have. Here they are in random order:

1. HC 0.21 had some issues with scene change detection. Have these problems been addressed in the current version? (The change log does not mention it)

2. Some users reported that HC 0.22 needs Avisynth 2.57 or higher. In my tests I used 2.56a without any problems. Does HC 0.22 contain any code which will not work with Avisynth 2.56 ?

3. When using HC with DVD-RB I remember that there were problems when HC inserted a sequence end code after every segment. It was recommended to use the "*NOSEQ_ENDCODE" statement in HC.ini to avoid these problems. Is this still true?

4. In previous versions I usually used a bias setting of 30 or higher to give static scenes more bitrate. When I now use the "*LUMGAIN" parameter (set to 2 or 3) is it recommended to set bias to 0 (full VBR) ?

5. Does HC somehow allow to have the TFF flag set for a progressive encode? I know that in progressive mode this flag is meaningless, but CCE does set the TFF flag in this case, and it really helps me for some conversions (Movie is progressive, end credits are interlaced TFF).


Anyway, thanks Hank for a great encoder!

Cheers
manolito

[hank315]

HC 0.21 had some issues with scene change detection. Have these problems been addressed in the current version? (The change log does not mention it)
Yes, the scene change detection algorithm is changed, it's one of the minor changes that never made it to the change log.

Some users reported that HC 0.22 needs Avisynth 2.57 or higher. In my tests I used 2.56a without any problems. Does HC 0.22 contain any code which will not work with Avisynth 2.56 ?
I've seen this reported also a couple of times, here it's working OK with Avisynth 2.5.6a and 2.5.7.
Rather strange because nothing really changed in the code which creates the script environment for Avisynth.

When using HC with DVD-RB I remember that there were problems when HC inserted a sequence end code after every segment. It was recommended to use the "*NOSEQ_ENDCODE" statement in HC.ini to avoid these problems. Is this still true?
By default it still inserts a sequence endcode because it's the proper way to end a sequence but this issue was addressed by jdobbs long time ago, DVD-RB simply removes the end code (00 00 01 B7) from the stream.

In previous versions I usually used a bias setting of 30 or higher to give static scenes more bitrate. When I now use the "*LUMGAIN" parameter (set to 2 or 3) is it recommended to set bias to 0 (full VBR) ?
I still use bias settings of approx. 30 in combination with lumgain 3, also depends on the source.

Does HC somehow allow to have the TFF flag set for a progressive encode? I know that in progressive mode this flag is meaningless, but CCE does set the TFF flag in this case, and it really helps me for some conversions (Movie is progressive, end credits are interlaced TFF).
For a real progressive encode the answer is no, BFF is set as default (has no real meaning) and the progressive sequence flag is set.
AFAIK CCE never sets the progressive sequence flag, it only sets flags for frames (progressive/interlaced).
If you set HC to 'auto detect', the progressive sequence flag isn't set and TFF is default, progressive frames are encoded progressive and interlaced frames are encoded interlaced.

Is it possible to have the "AVSRELOAD" argument from the CLI in the next release?
Will do, no problem.

[boywondr16]

Please forgive me if this has been answered before, but in using HCenc with DVD Forger or AVS to DVD, I can't seem to figure out how to pass a command to set the resolution size.

That said, Hank, you've done a great job with HC. I use it exclusively.

[Boulder]

HC doesn't do any resizing.

[MrC]

Please forgive me if this has been answered before, but in using HCenc with DVD Forger or AVS to DVD, I can't seem to figure out how to pass a command to set the resolution size.

If for resolution size you mean video frame width and height, you have to change it by editing the AviSynth script (in AVStoDVD click on "AVS Script" icon).

;)

Bye

[Chumbo]

I wanted to report that I too have received:*** ERROR, source mismatch in pass 2 starting at frame: 70251
using 022. My .ini file does contain*AVSRELOAD

I use avisynth 2.57. For this encode, my sources were muxed into one mkv file first and then fed into hc. Normally, I just encode the parts separately, so I don't know if that had anything to do with this problem cropping up.

[AlanHK]

If for resolution size you mean video frame width and height, you have to change it by editing the AviSynth script (in AVStoDVD click on "AVS Script" icon).


IN more detail: you would add a resize command to the AVS. Eg:

AVISource("sja-100.avi")
LanczosResize(720,576,0,0,640,352)

See Avisynth docs for details.
Or use the excellent FitCD utility.

[Chumbo]

I wanted to report that I too have received:*** ERROR, source mismatch in pass 2 starting at frame: 70251
using 022. My .ini file does contain*AVSRELOAD

I use avisynth 2.57. For this encode, my sources were muxed into one mkv file first and then fed into hc. Normally, I just encode the parts separately, so I don't know if that had anything to do with this problem cropping up.
I wanted to follow up to this. I just tried reencoding part 1 directly from the EVO, via a graphedt file through an avs script, and the same error popped up on pass #2. Again, the reload avisynth is checked. This time, though, it started at frame 0:*** ERROR, source mismatch in pass 2 starting at frame: 0

I'm going to retry this in 021.

[MrC]

IN more detail: you would add a resize command to the AVS. Eg:

AVISource("sja-100.avi")
LanczosResize(720,576,0,0,640,352)

See Avisynth docs for details.
Or use the excellent FitCD utility.

Ehm, AVStoDVD just does that job.... and then you can change the script automatically generated.

;)

Bye

[kolak]

Why I have only about 30% usage of my 8 core PC (110 FPS) when encoding with HC?

I use 2.57 avisynth with 2 lines: avisource line and convertyv12 line.
Source is uncompressed PAL YUY2.

kolak

[totya]

Why I have only about 30% usage of my 8 core PC (110 FPS) when encoding with HC?

I use 2.57 avisynth with 2 lines: avisource line and convertyv12 line.
Source is uncompressed PAL YUY2.

kolak

8 core???

[Chumbo]

I wanted to follow up to this. I just tried reencoding part 1 directly from the EVO, via a graphedt file through an avs script, and the same error popped up on pass #2. Again, the reload avisynth is checked. This time, though, it started at frame 0:*** ERROR, source mismatch in pass 2 starting at frame: 0

I'm going to retry this in 021.
With 0.21, HC kept crashing on the first EVO, and would either continue with the 2nd or crash with the 2nd. I just have a batch file that runs the command line with the .ini file. But even with the GUI, the first EVO kept crashing HC. I tried to change the .grf file a bit, i.e., add the audio decoder to the audio out pins (both with and without the audio renderer) and still crashed. I even rebooted the PC and cleaned out the temp folder and that didn't help.

I was able to get it going by muxing the video into a single MKV container and used it as a source and now 0.21 is encoding. This should work fine w/out any errors as usual.

[six13]

Can someone please expaline this to me? I was reading some posts by Boulder and jdobbs and was trying to understand if using this setting with HCenc 22 would benefit my output.

My source is a NTSC Sat. rec SD TV interlaced capture recorded with MS Media Center and the .dvr-ms file is converted to MPEG2 for demuxing and authoring to process with DVD-RBa nd HCenc 22.

Considering my source would this benfit my output and if so which setting to start with? Would this feature benfit a progressice source DVD rip and if so which setting to start with? I would alter HCenc with the GUI.

Thanks to everyone.

Six13

[mikenadia]

Excellent program, Hank. I noticed that when I click on suspend , the internal clock does not stop, the average FPS decreases and the time left to encode will increase when you click on resume.Just a notice.

[hank315]

Why I have only about 30% usage of my 8 core PC (110 FPS) when encoding with HC?

I use 2.57 avisynth with 2 lines: avisource line and convertyv12 line.
Source is uncompressed PAL YUY2.
That's as expected, input via Avisynth or d2v takes about 30%, encoding and output is 70%.
Decoding and encoding are done simultaneous, encoding is also frame based multi-threaded.
In the case of uncompressed sources or heavy filtering, the encoder probably has to wait for input.
At best the whole process will max out at approx. 80% CPU on a quad core and at 40% on 8 cores.

@mikenadia
I will put it on the todo list.

[Chumbo]

If you think worthwhile, it would be great to be able to change the priority while HC is encoding. That way rather than suspend, we can set it to Idle while doing something else for a bit and then set it back to normal or high or whatever. :)

[Irakli]

If you think worthwhile, it would be great to be able to change the priority while HC is encoding. That way rather than suspend, we can set it to Idle while doing something else for a bit and then set it back to normal or high or whatever. :)

You can do this with HC 0.22 (it's in 'Settings 3').

[Chumbo]

You can do this with HC 0.22 (it's in 'Settings 3').
You must read it as it's written young padawan. ;) The key words are "while HC is encoding" which you cannot do. You can set it, true, prior to the encoding starting, but once it starts, you cannot change it.

A good example of programs that do that are VritualDub and Nero Vision.

[Irakli]

You must read it as it's written young padawan. ;) The key words are "while HC is encoding" which you cannot do. You can set it, true, prior to the encoding starting, but once it starts, you cannot change it.

A good example of programs that do that are VritualDub and Nero Vision.

Sorry, I misread your post. I guess that's because I was trying to prepare for my university presentation and browse doom9.org at the same time.

Yes, that could be really nice feature to add (although I am perfectly fine with CTRL+ALT+Del approach for setting priority).

Regards,
Irakli

[Chumbo]

Sorry, I misread your post. I guess that's because I was trying to prepare for my university presentation and browse doom9.org at the same time.

Yes, that could be really nice feature to add (although I am perfectly fine with CTRL+ALT+Del approach for setting priority).

Regards,
Irakli
No problem. Yeah, it's just more convenient than Task Manager. ;)

[kolak]

That's as expected, input via Avisynth or d2v takes about 30%, encoding and output is 70%.
Decoding and encoding are done simultaneous, encoding is also frame based multi-threaded.
In the case of uncompressed sources or heavy filtering, the encoder probably has to wait for input.
At best the whole process will max out at approx. 80% CPU on a quad core and at 40% on 8 cores.



It's a simple encoding, no filtering or other preprocessing.
Encoder doesn't have to wait for data because it comes from RAID which can deliver 400MB/s.

Elecard Studio Converter encodes 12 times faster than RT with 50% CPU usage.

x264 uses 100% all the time :)

[Mr_Odwin]

Hey Hank.
I seem to be taking a speed hit with this release, without doing anything different. Would there be any reason for that?

In particular I would run multiple instances of HC; now when I run multiple instances, (regardless of SMP setting) I get much lower speeds (about 2/3 what it was).

[Mr_Odwin]

I have some actual numbers and details.
Using an Intel e4300 overclocked to 2.4GHz with 1GB RAM.
Using a fairly standard avs script that serves a 720x576 file at 23.976 FPS. Ini file:

*CQ_MAXBITRATE 8
*MAXBITRATE 8000
*PROFILE FAST
*DC_PREC 8
*MATRIX MPEG
*ASPECT 16:9
*AUTOGOP 12
*NOSCD
*WAIT 0

Command line is just avs location, desired location of m2v and where the ini file is.

HC 0.22: 3 Instances, No SMP = 87 FPS
HC 0.22: 3 Instances, SMP = 77 FPS
HC 0.22: 1 Instance, No SMP = 45 FPS
HC 0.22: 1 Instance, SMP = 73 FPS

HC 0.21: 1 Instance = 71 FPS
HC 0.21: 3 Instances = 184 FPS

There's quite a difference.

[Boulder]

Out of interest, what does Task Manager show regarding CPU usage?

[Mr_Odwin]

0.22 3 Instances, No SMP = 100%
0.22 3 Instances, SMP = 100%
0.22 1 Instance, No SMP = 50%
0.22 1 Instance, SMP = about 90%

0.21 3 instances = 100% CPU
0.21 1 instance = 50% CPU

[G_M_C]

Ive been trying HC 022 on SMP, and im very pleased with the results. Having "switched over" from CCE SP i must say that HC really is a good alternative to it, speed-wise and quality-wise.

But ive got a question about the adaptive matrix dll' s supplied with the release:
When i use AQ through one of the supplied DLL's or with the luminance-slider, the Q-matrix gets adapted based on luma in a GOP. But how much change does it make to the matrix, and more important; Does it revert back to my original custom matrix when a "normal scene" is found ?

[mikeytown2]

Feature request for 2 pass encodes:

On the first pass, have HC output a huffyuv/Lagarith file. Then on the second pass instead of re-running the avs script it would use the huffyuv/Lagarith stream. This has the potential to speed up 2 pass encodes.

If you do decide to implement this then some of the request to support avi files might be doable... just have the program create a temp avs file and use that to load the avi file.

IMHO only downside to using huffyuv/Lagarith for the 2nd pass would be massive space usage.

Any Thoughts?

[hank315]

@Mr_Odwin
Just did a test and couldn't find a significant difference in speed between HC021 and HC022 running multiple instances (also tested it with 3 instances).
But there's a difference how time and fps is measured, with HC021 fps is measured relative to CPU time used, with HC022 it's measured relative to wall clock time, this means the fps in HC021 isn't accurate when running multiple instances.
Do you also find a large difference if you measure the real time?
And if you're running multiple instances with HC022, make sure *NOSMP is present in the ini file.

@G_M_C
The matrix switching is done on a GOP basis, so on average it can switch 2 times every second and yes it will use the original matrix on normal scenes.
Maybe better to use the *LUMGAIN command which does almost the same.
The change depends on the average luma values in a GOP, the max change using the dlls is a factor 0.6, using *LUMGAIN 4, the max change is a factor 0.5, so half of the original matrix values.



Encoder doesn't have to wait for data because it comes from RAID which can deliver 400MB/s.
400MB/s is nice... but that's not the speed at which Avisynth delivers frames, disk speed isn't the main issue.
Getting the input takes 30% on average, processing/decoding it just takes time.
Elecard Studio Converter encodes 12 times faster than RT with 50% CPU usage.
I tested the Elecard encoder on a dual core and it didn't run faster than HC022.
Main goal was to create effective SMP encoding on main stream processors (dual/quad core).
And if you can get 12 speed out of the Elecard encoder with 8 cores, best to use that one... well, speedwise of course...

x264 uses 100% all the time :)
Cool :), sure you're getting 12 speed also with some descent quality settings :D
Just as you will get 50% or more CPU with CCE ;)

[G_M_C]

@G_M_C
The matrix switching is done on a GOP basis, so on average it can switch 2 times every second and yes it will use the original matrix on normal scenes.
Maybe better to use the *LUMGAIN command which does almost the same.
The change depends on the average luma values in a GOP, the max change using the dlls is a factor 0.6, using *LUMGAIN 4, the max change is a factor 0.5, so half of the original matrix values.


Hope there is a max to how low the matrix can be adapted, since i allread use an insane matrix;

8 10 11 12 12 13 14 15
10 11 12 13 13 15 15 16
11 12 12 14 15 15 16 17
12 13 14 15 15 16 17 18
12 13 15 15 16 17 18 19
13 15 15 16 17 18 19 19
14 15 16 17 18 19 19 20
15 16 17 18 19 19 20 20

8 8 8 9 9 9 9 10
8 8 9 9 9 9 10 10
8 9 9 9 9 10 10 10
9 9 9 9 10 10 10 10
9 9 9 10 10 10 10 11
9 9 10 10 10 10 11 11
9 10 10 10 10 11 11 11
10 10 10 10 11 11 11 11

Project=NTSC DVD, 120 minutes, 23.976fps / pull-up, 720x480 / 16:9, avg. bitrate=7000 / maximum=8200.

The project has approx 30% ~ 40% "dark scenes", and most are important to the storyline; That's why i'd like to try the option. I' ll set the LUMGAIN @ 1 and see what happens.

2 last questions though; Do the end-credits (black scene with small white text) also count as a "dark scene" for the adjustment to kick in ? And what about, for instance, scene-transition with a short moment of black screen ( a sequence of "blanc frames" ) ?

[AlanHK]

IMHO only downside to using huffyuv/Lagarith for the 2nd pass would be massive space usage.

That "massive space usage" would be a killer for me. I have to do a search and destroy session to clear space before I make every DVD as it is.

just have the program create a temp avs file and use that to load the avi file.

There are plenty of third-party apps that will do that, and even run HC and make your coffee while burning a DVD. I hope Hank can resist the pressure to add such features, spending time on that can only be at the expense of improving the core function of creating excellent MPEGs.

[kolak]

Cool :), sure you're getting 12 speed also with some descent quality settings :D
Just as you will get 50% or more CPU with CCE ;)


Hehehe- I'm getting 6FPS with the best settings.
CCE uses below 20% :( and the speed is 6.

If my source will be YUV12 will this increase speed?
Why do I have to go through avisynth?

Thanks

[Chumbo]

I wanted to report that I too have received:*** ERROR, source mismatch in pass 2 starting at frame: 70251
using 022. My .ini file does contain*AVSRELOAD

I use avisynth 2.57. For this encode, my sources were muxed into one mkv file first and then fed into hc. Normally, I just encode the parts separately, so I don't know if that had anything to do with this problem cropping up.
Just wanted to add another response to this. I'm still getting these errors and this weekend I had to do an encode 3 times. First with 022 with AVSRELOAD which still failed, then 022 without AVSRELOAD and then with 021 which was fine. ---------------------
| encoding - pass 2 |
---------------------

*** ERROR, source mismatch in pass 2 starting at frame: 1292

*** ERROR, source mismatch in pass 2 starting at frame: 1292

*** INFO, adjusting average bitrate: -0.48 %
*** INFO, adjusting average bitrate: -0.85 %
...
If there's anything I can do to help, extra debugging or logging or whatever, please let me know. Many thanks.

[G_M_C]

I've got a question for Hank;

Is it possible to set a "minimum bitrate", some encodes for DVD seem to require a minimum (allthough my DVD-player doesnt seem to mind).

[Mr_Odwin]

@Mr_Odwin
Just did a test and couldn't find a significant difference in speed between HC021 and HC022 running multiple instances (also tested it with 3 instances).
But there's a difference how time and fps is measured, with HC021 fps is measured relative to CPU time used, with HC022 it's measured relative to wall clock time, this means the fps in HC021 isn't accurate when running multiple instances.
Do you also find a large difference if you measure the real time?
And if you're running multiple instances with HC022, make sure *NOSMP is present in the ini file.


Real time results with a stopwatch:
(Mins:Secs)
(Same source obviously)

HC 0.22: 3 Instances, No SMP = 8:25
HC 0.22: 1 Instance, No SMP = 13:50
HC 0.22: 1 Instance, SMP = 9:16

HC 0.21: 1 Instance = 12:09
HC 0.21: 3 Instances = 7:39

Still a bit of a speed drop. Similar in magnitude between the different cases too. (i.e. 8:25/7:39 ~= 13:50/12:09)
The cli and ini files are identical, except the 0.22 ini files also contain:

*NOSMP
*PROGRESSIVE

(Or the *SMP command.)

[dragongodz]

Feature request for 2 pass encodes:

IMHO only downside to using huffyuv/Lagarith for the 2nd pass would be massive space usage.

to say the least.

frankly if someone is going to be using filters etc, which is where this would really show a difference, then they should have the knowledge to create the avi themselves and just use that from the beginning.

this is not the sort of jumping through hoops an encoder should have to do.

[mikeytown2]

this is not the sort of jumping through hoops an encoder should have to do.
Good point! Thankfully IMHO there is a good solution now for a speed boost on 2 pass encodes
http://forum.doom9.org/showthread.php?p=1073371#post1073371

[Calimari]

The Bitburners site is hosting v0.21 again? Where is v0.22?

[Fishman0919]

Just noticed that too... you can download HC v0.22 from the http://www.doom9.org/ download section

[Calimari]

There it is back.

[rendez2k]

Not sure if this has been asked before, but is there much difference between the NORMAL and BEST profile in the latest version and how much does it impact on time as a rule (is it say double the time for the best profile?).

[shirohamada]

kinda late, but,

ｷﾀ━━━━━━(ﾟ∀ﾟ)━━━━━━!!!!!

[hartford]

@Hank

Could HC22 _not_ take focus when finished with a job? Perhaps a switch to do so?

I ask because it happens that I often get to encoding a vid at the time that I play a game online. When HC22 takes focus, my game halts until it is shutdown.

Thanks.