| Age | Commit message (Collapse) | Author |
|---|
|
# Conflicts:
# src/monomorphise.ml
|
|
It now pushes casts into lets and constructor applications, and so
supports the case needed for RISC-V.
|
|
|
|
|
|
Define initial values for record types once instead of repeating them in
the initial register state.
|
|
Don't ask Z3 to simplify them, as flow typing might lead to different
results in different if-branches, for example, leading to type errors in
Lem.
|
|
|
|
|
|
For example in RISC-V for the translation table walk:
$optimize unroll 2
val walk32 ...
function walk32 ...
would create two extra copies of the walk_32 function,
walk_32_unroll_1 and walk_32_unroll_2, with only walk_32_unroll_2
being recursive. Currently we only support the case where we have
$optimize unroll, directly followed by a valspec, then a function, but
this should be generalised in future.
This optimization nearly doubles the performance of RISC-V
It is implemented using a new Optimize.recheck rewrite that replaces
the ordinary recheck_defs pass. It uses a new typechecker
check_with_envs function that allows re-writes to utilise intermediate
typechecking environments to minimize the amount of AST checking that
occurs, for performance reasons.
Note that older Sail versions including the current OPAM release will
complain about the optimize pragma, so this cannot be used until they
become up to date with this change.
|
|
RISC-V
|
|
|
|
|
|
|
|
since riscv is no longer in this repository, and we use the RISC-V
duopod as an example, you need to build as:
make RISCV=directory manual.pdf
if directory is not equal to ../../sail-riscv (which is where it would
be if sail and sail-riscv are checked out in the same respository
together)
|
|
We should maybe just make the -latex option behavior the default to avoid this kind of thing
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
There are situations when we really want a more refined expression,
such as 8 * n instead of 64 (when we know n = 8 from a case split), but
we might not be able to generate it. For now we generate an underscore
and let Coq figure it out from the context.
|
|
|
|
With the new comment syntax, Isabelle seems to barf on that comment,
apparently due to the backslashes.
|
|
|
|
(Needed for current CHERI.)
|
|
|
|
Annotating non-recursive functions as recursive in Lem output is
allowed, but will make Lem use "fun"/"function" commands when generating
Isabelle output, which is much slower to process than "definiton".
|
|
Also output termination measures in Sail printer
|
|
Add a file nl_flow.ml which can analyse a block of Sail expressions
and insert constraints for flow-typing rules which do not follow the
lexical structure of the code (and therefore the syntax-directed
typing rules can't do any flow-typing for). A common case found in ASL
translated Sail would be something like
function decode(Rt: bits(4)) = {
if Rt == 0xF then {
throw(Error_see("instruction"));
};
let t = unsigned(Rt);
execute(t)
}
which would currently fail is execute has a 0 <= t <= 14 constraint
for a register it writes to. However if we spot this pattern and add
an assertion automatically:
let t = unsigned(Rt);
assert(t != 15);
execute(t)
Then everything works, because the assertion is in the correct place
for regular flow typing. Currently it only works for this specific
use-case, and is turned on using the -non_lexical_flow flag
|
|
Some of the output from the tests scripts is odd on Jenkins, try to
fix this by flushing stdout more regularly in the test scripts
|
|
bool_of_bit and bit_of_bool in sail_lib
|
|
|
|
|
|
|
|
subrange_subrange_concat does a zero extension internally, so another
zero extension of its result is redundant and can lead to a type error
in Lem (because Lem's type system cannot calculate the length of the
intermediate result of subrange_subrange_concat).
|
|
Sizeof-rewriting could introduce extra arguments to functions that
instantiate_simple_equations could fill in with overly complicated
types, causing unification to fail when building lem.
|
|
|
|
|
|
|
|
|
|
Now all ARM, RISC-V, and CHERI-MIPS all build successfully with
type-checking changes. All typechecker/c/ocaml/lem/builtin/riscv/arm
tests are now working as well.
Now the python test scripts can run sequentially with TEST_PAR=1 there's
no reason to keep the old shell versions around anymore.
|
|
Spawning a process for every test and running every test in parallel
is quite RAM intensive (up to about 8gb) especially when running
valgrind on every test in parallel. Now we only run up to TEST_PAR
tests in parallel (default 4).
|
|
It now includes updating the effects so that morally pure recursive
functions can be turned into this impure termination-by-assertion form.
|
|
|
|
test/typecheck/pass/tautology.sail constaints tests of various boolean
properties, e.g.
// de Morgan
_prove(constraint(not('p | 'q) <--> not('p) & not('q)));
_prove(constraint(not('p & 'q) <--> not('p) | not('q)));
introduce a new _not_prove case which allows us to assert in tests
that a constraint is not provable. This test essentially tests that
constraints map to sensible problems in the SMT solver, without
testing flow typing or any other features.
Add a script test/typecheck/update_errors.sh, which regenerates the
expected error messages. Testing that type-checking failures is
important, but can be brittle when the error messages change for
inconsequential reasons. This script automates fixing this.
Also ensure that this test case works correctly in Lem
|
