Notes
-----

Friday, 19 February
	is it possible to read in the files in the order in which we'll
	output them:  ipbbbb, pbbb, etc.
	(this uses less memory)

	but might files be given sequentially, rather than in a name
	list?  (do we have random access guaranteed?)

	...

	need to DECODE I-frame so can get true motion vectors, right?	

	should separate I-frame macroblock into procedure, since it
	can be used in P-frame generation

Monday, 22 February
	motion vectors:  forward_f code is passed to genMBHead, but not
	written there -- it's written in genPictHead.  Have to fix this!

	fixed the motion vectors by working backwards from decoding

Wednesday, 24 February
	DCT:  can't just use y_blocks because in I-frames, this has been
		DCT'd.

	DCT:  is it set up just to work with positive numbers?!?!?!
		if so, have trouble doing DCT for diffs
		(SEE PAGE D-38!!!)

Wednesday, 3 March
	headers.c:  why is vbv_delay a 32-bit integer, when only 16 bits
			are used?
	COMPRESSION:  I-frames give a 12-to-1 compression rate for wally
	ComputeDiff:	need to make it work with non-integers!!!!

Thursday, 4 March
	change program to allow input to be bunch of rgb frames


prev motion vectors reset when:  SliceHeader
				 when no forward motion vector
				 when block is skipped
				 beginning of new frame

Wednesday, 10 March
	should the program have any say in which frames should be I, B, P?
	dependencies:
		when get an I frame, can output the I frame, output all
		previous B frames, delete the previous reference frame, and
		save the I frame

		when get a P frame, can output the P frame, then output all
		previous B frames, delete the previous reference frame, and
		save the P frame

		when get a B frame, just save it (have to wait for future
		reference frame)

	so, need a linked list of B-frames, and a previousRefFrame, and a
	currentFrame

Monday, 15 March
	assuming that skipped macroblocks at end of slice don't need some
	end macroblock to be sent.


I:  12:1 compression...P:  24:1 compression...B:  0:1 compression
Elapsed times (s):  104.5 user, 1.5 system, 106.0 total
PROCESSED:  191197 Bits per second
I Frames:  669621 BPS....P Frames:  102940 BPS....B Frames: 0 BPS


Monday, 22 March
	Something slightly wrong with DCT coefficients -- getting blue
	stuff in flower garden

	Are the DC predicted values reset to 128 on skipped macroblocks?
	Yes, they are!


Tuesday, 23 March
	skipped macroblocks have different meanings in B and P frames


Monday, 5 April
	Change q-scale for I-blocks inside of P-frames???

Monday, 19 April
	Does last block have to be transmitted (not skipped) for others to
	be processed???
		Tuesday -- Yes!


Friday, 23 April
	Need to take care of Forced Updating (page 46 of MPEG-1 doc)

	DO THIS:  Make program SORT input file names, if
			numeric

Wednesday, 28 April
	DO THIS:  Have option to take gif files and calls giftoppm
		  appropriately


Wednesday, 2 June
	motion vector for chrominance is HALF normal motion!!!

Friday, 18 June
	ComputeSubSampled... is 0.03 ms per call, whereas ComputeMotionLum...
	is 0.06 per call.  Should be 1/4, only 1/2.  See if you can do
	somethinga about this

	When dividing by 2, they usually mean // (where 0.5 becomes 1)
	So notice the calculation of half-pixel motion and of B-frame motion

Tuesday, 13 July
	GOING TO ASSUME, for now, that all I, P frames need to be decoded
	if referenceType == DECODED_FRAME (not true, but a safe bet)


Thursday, 22 July
	Endianness:
		Big		Little
		---		------
		Sun, HP		DEC

NEXT RELEASE:  DELETE ALL makedepend STUFF from MAKEFILE