Milestones

Cost of milestones were calculated by taking the effort allocated per consortium member for the particular quarter the milestone is in and allocating a proportion of that quarter's costs based on the amount of effort required in that quarter (shown in the table as quarter effort). The cost to milestone is then calculated by using the projects Gantt Chart and calculating the cumulative costs of all work required for that milestone. For the costs attributed to Work Package 9, these are ongoing throughout the project and have been allocated as such.

Ref	Title	Work Package	Responsibility	Due Date	Quarter Effort	Cost of Milestone (£)	Cost to Milestone (£)	Notes
WP3/1	Board Design Documentation and working hardware	WP3	UoB	2013-07-31	100%	17009.00	17009.00	Completed, v2 board produced and distributed at EEC SIG workshop and FOSDEM with TSB support. New batch being ordered end of September 2014, to support industry interest in evaluation.
WP1/1	Decision of GCC/LLVM for first implementation	WP1	Emb.	2013-08-31	35%	5134.80	5134.80	GCC chosen.
WP1/2	Design documentation for compiler integration, machine learning interface and available features	WP1	Emb.	2013-08-31	65%	9536.07	9536.07	Iterative design, live document
WP4/1	Training set and case study sources	WP4	UoB	2013-09-30	33.3%	2510.93	2510.93	BEEBS v2.0 now released with 80 benchmarks, data variants and per-function data to be used. Three real-world case studies identified.
WP4/2	Embedded systems set up for testing	WP4	UoB	2013-09-30	33.3%	3010.93	22530.85	Complete, documentation as Workshop on how to use the system.
WP5/2	Selection of core learning algorithm(s), training approach and API	WP5	UoB	2013-09-30	33%	2510.93	8212.45	Output of working days. Complete.
WP2/1	First and second iterations of implementation, testing, documentation with two compilers	WP2	Emb.	2013-11-30	100%	10289.30	33172.62	Complete for GCC and LLVM.
WP5/1	Literature Review	WP5	UoB	2013-12-31	100%	5701.53	5701.53	Complete Review by Munaaf Ghumran as part of his MEng. May be appropriate for publication.
WP6/1	Working training infrastructure (existing optimisations)	WP6	Emb.	2014-02-28	35%	7556.52	7556.52	Complete. Data currently being gathered for ATMega328.
WP6/2	Results from proof of concept training and test (existing optimisations)	WP6	Emb.	2014-02-28	25%	5397.51	68657.49	Demonstrated at GNU Tools Cauldron
WP7/1	Design and implement optimisation passes in GCC	WP7	Emb.	2014-02-28	40%	8636.02	8626.02	Theoretical basis being presented at conferences in Q6. Implementation unlikely, due to illness.
WP6/3	Working training infrastructure (new optimisations)	WP6	Emb.	2014-05-31	33.3% Emb, 50% UoB	9579.99	86873.50	Not possible due to dependency on WP 7/1,
WP6/4	Results from proof of concept training and test (new optimisations)	WP6	Emb.	2014-05-31	33.3% Emb, 50% UoB	9579.99	96453.50	Not possible due to dependency on WP 7/1,
WP7/2	Implement optimisation passes in LLVM	WP7	Emb.	2014-05-31	33.3%	7171.74	15807.76	Not possible due to dependency on WP 7/1,
WP5/3	Review of approach	WP5	UoB	2014-07-31	33%	8281.77	41454.38	Blog post by Greg Cawthorne, which may be suitable for future publication.
WP8/1	Trained and tested complete system with full case studies	WP8	UoB	2014-09-30	33.3% Emb, 50% UoB	14030.86	109443.65	In progress.
WP8/2	Draft paper presenting results	WP8	UoB	2014-09-30	33.3% Emb, 50% UoB	14030.86	139968.72	In progress.
WP8/3	Final paper with final modifications and result generation	WP8	Emb.	2014-11-30	100% Emb, 100% UoB	25426.28	165395.00	This is the ultimate report and it is anticipated that it will take some time to develop. Depends on WP8/2.