We plan a change in the semantics of stack_solve_algo. Currently, values 0 and 1 actually implement the same algorithm (sim1/simk don’t actually implement the original relaxation algorithm). We propose to associate to value 1 the original relaxation algorithm, which will be available in bytecode at first (slower, but less memory-consuming).
As of 5f0dac24, the test file block_bytecode/ls2003.mod crashes under Octave
We should probably swap values 5 and 8 for solve_algo, for consistency with Dynare 4.1
The new values for stack_solve_algo and solve_algo (in combination with block/bytecode) need to be documented on the wiki and in the reference manual (on the wiki, we should keep separate sections on Dynare 4.1 and Dynare 4.2)
The new values for stack_solve_algo and solve_algo (in combination with block/bytecode) need to be documented on the wiki and in the reference manual (on the wiki, we should keep separate sections on Dynare 4.1 and Dynare 4.2)