angr's power comes not from it being an emulator, but from being able to execute with what we call symbolic variables.
potential target for low latency (or sparse?) stuff
gonna need this sooner or later
just useful for learning about BB (what about interior point?)
is this the actual basis of PIP?
for hackerrank interviews...
can i use this for a better python ast? also https://plugins.jetbrains.com/docs/intellij/psi.html
implement for CIRCT?
Circle is a C++ bare metal programming environment for the Raspberry Pi.
basically does rewrite on import which enables python files to host alternative syntax. kind of cute but not really what i want
static memory allocation problem can be formulated as LRA opt + XOR? also https://optimathsat.disi.unitn.it/.
not sure
Chunky Loop Analyzer: A Polyhedral Representation Extraction Tool for High Level Programs
compiling is annoying af
- python bindings
- use clan to analyze python for loops
A Next-Generation Compiler Compiler
- use it to parse python into MLIR (using mll)
- extract the BDD and combine with PIP
- use the python frontend to lower to affine
if i'm gonna solve things symbolically (ie DSA) this might be useful
just take for a test drive
use in combination with yices?
how is it different from ISL? sage bindings?. also python bindings
can this solve DSA?
somewhere in here there's an implementation of a PIP solver
Polyhedral projection is a main operation of the polyhedron abstract domain. It can be computed via parametric linear programming (PLP), which is more efficient than the classic Fourier-Motzkin elimination method.
^ hmm is that really true? i thought one relied on the other?
related to this paper (An efficient parametric linear programming solver and application to polyhedral projection)
A naive implementation of the Gomory cutting plane algorithm
is this faster than solving the CSP?
The basic concept is to get the normal compiler front end like clang to emit its LLVM Intermediate Representation (IR) as a bitcode file, then run a custom LLVM pass over this to inject DRTI code at the appropriate places.
also https://software.cs.uni-koeln.de/scil/documentation/html/main.html
We present a novel algorithm for finite domain constraint optimization that generalizes branch-and-bound search by reasoning about sets of assignments rather than individual assignments.
the next step of the low latency work
don't quite understand this but symbolic virtual machine is what i need for various parts of the static memory planning thing
faster way to solve PIPs?
approx method for solving ILPs (learning). probably not that easy to implement (gym env?)
Efficiently Solving Repeated Integer Linear Programming Problems by Learning Solutions of Similar Linear Programming Problems using Boosting Trees
approx method for solving ILPs (learning). probably easy to prototype using xgboost
isl official tutorial (read after completing pollylabs tuts)
Interestingly, AlphaTensor finds algorithms with a larger number of additions compared with Strassen-square (or equivalently, denser decompositions), but the discovered algorithms generate individual operations that can be efficiently fused by the specific XLA33 grouping procedure and thus are more tailored towards the compiler stack we use.
A Scalable Approach to Exact Resource-Constrained Scheduling Based on a Joint SDC and SAT Formulation
implement in CIRCT
maybe this can be done using or-tools?
maybe this can be done with https://github.com/siala/Hybrid-Mistral
just good to know
can this solve the DSA problem?
just good to know
Clockwork: Resource-Efficient Static Scheduling for Multi-Rate Image Processing Applications on FPGAs
implement in CIRCT
read after learning racket
racket class on implementing small languages
probably useful for lockfree?
just good to know
required reading but i'll probably never get to it?
required reading but i'll probably never get to it?
definitely need to know this for HFT
suitable allocation algo for nns?
re the MLIR pass plugin project. also https://developer.ibm.com/tutorials/l-dynamic-libraries/
maybe for bare metal arm but also maybe bare metal gpu
this is the next stage of the low latency project
also https://forums.developer.nvidia.com/t/boosting-inline-packet-processing-using-dpdk-and-gpudev-with-gpus/212943)) and Benchmarking GPUDirect RDMA on Modern Server Platforms
can this solve the DSA problem?
should probably know this after teaching rohan
learn in order to use rust a compiler. another source.
read in prep for luminous
some useful discussion on symbolic autodiff
useful for ... everything?
A look at the performance of expression templates in C++: Eigen vs Blaze vs Fastor vs Armadillo vs XTensor
probably inflated but worth a read
compile MLIR to wasm in order to run in browser? maybe that's useless and just use torch-mlir wheels
hack pcie to get lower than 0.5us latency
cs.theory discussing how presburger formulas actually work - including some discussion of BDDs