mirror of
https://github.com/grin-compiler/grin.git
synced 2024-09-11 15:37:54 +03:00
102 lines
2.3 KiB
Plaintext
102 lines
2.3 KiB
Plaintext
* - ghc/stg frontend
|
|
* - compile original benchmark suite ; diagrams/machine code: ghc vs grin
|
|
* - HPT inline support ; context and function support
|
|
* - sharing analysis for HPT
|
|
* - HPT llvm backend
|
|
* - adaptive inliner
|
|
* - arity raising
|
|
* - pipeline polish and testing with all transformation ; make it to work
|
|
* - module handling
|
|
* - parallel transformations ; parallel recursion schemes
|
|
- grin type annotations in syntax ; with type hole support
|
|
- grin linter
|
|
- name handling
|
|
- explicit product type support ; maybe alter tagged nodes to be tag + product type
|
|
- simplify LLVM backend ; refactor as GRIN transformations ; readd simplification transformations
|
|
- precompiled HPT with partial result for each module
|
|
- compile simple programs from hackage
|
|
- agda frontend ; includes dead data elimination
|
|
- asap memory manager
|
|
- tracing gc
|
|
|
|
benchmark tool ; instruction counter
|
|
|
|
perf stat -e instructions:u mv
|
|
|
|
apr
|
|
may
|
|
|
|
jun : asap + gc
|
|
jul : spmd compiler
|
|
aug : cache friendly chunky memory
|
|
|
|
- fast jit + parallel compile then migarte heap data
|
|
- bidirectional source mapping
|
|
- rts from ghc
|
|
|
|
zurihac demo milestone:
|
|
- ghc frontend
|
|
|
|
- complete missing transformations ; polish transformation framework
|
|
- fancy benchmark framework
|
|
|
|
- HPT inline support ; sharing analysis
|
|
- HPT llvm backend
|
|
- adaptive inliner
|
|
|
|
- module support
|
|
|
|
- idris frontend
|
|
- parallel transformations
|
|
|
|
apr
|
|
05 T - ghc simplifier phd
|
|
06 F
|
|
07 S
|
|
09 S
|
|
|
|
10 M
|
|
11 T
|
|
12 W
|
|
13 T
|
|
14 F
|
|
15 S
|
|
16 S
|
|
|
|
17 M
|
|
18 T
|
|
19 W
|
|
20 T
|
|
21 F
|
|
22 S
|
|
23 S
|
|
|
|
24 M
|
|
25 T
|
|
26 W
|
|
27 T
|
|
28 F
|
|
29 S
|
|
30 S
|
|
|
|
may - 4 weeks
|
|
|
|
this moment:
|
|
i have plenty of time, many many months.
|
|
i can achieve a lot, optimise HPT to be extremely efficient and support for incremental analysis
|
|
i also can add multiple frontends to grin, like ghc, agda and idris
|
|
i also know how to extend whole program analyisis to collect static heap information and do compile time garbage collection
|
|
i could reimplement lambdacube with gpgpu support including a node editor
|
|
i can finish lambdacube quake 3
|
|
april, may, june, july, august, september
|
|
|
|
large tasks
|
|
grin phd level
|
|
frontends: ghc, idris, agda
|
|
asap
|
|
spmd support
|
|
reboot lambdacube with as new grin frontend
|
|
network support for lambdacube quake3
|
|
particle system support for lambdacube quake3
|
|
full game logic for lambdacube quake3
|