The main speedup is accomplished by avoiding a heap allocation in the common case where the final string length is less than 128. Inlining stringf & vstringf adds an additional improvement.
The main speedup comes from swithing from using a SHA1 hash to std::hash<std::string>. There is no need to use an expensive cryptographic hash for fingerprinting in this context.
Compiling on GCC hid this bug as it optimized the nullptr call away as
undefined behavior, but running the SBY tests with a clang build hits
this error.
This adds the -noinitstate option which is required to simulate
counterexamples to induction with yw-cosim. Also add handling for
$initstate cells for non-co-simulation.
Instead of passing around in_lvalue/in_param flags to simplify, we make
the flags into properties of the AST nodes themselves. After the tree
is first parsed, we once do
ast->fixup_hierarchy_flags(true)
to walk the full hierarchy and set the flags to their initial correct
values. Then as long as one is using ->clone(), ->cloneInto() and the
AstNode constructor (with children passed to it) to modify the tree, the
flags will be kept in sync automatically. On the other hand if we are
modifying the children list of an existing node, we may need to call
node->fixup_hierarchy_flags()
to do a localized fixup. That fixup will update the flags on the node's
children, and will propagate the change down the tree if necessary.
clone() doesn't always retain the flags of the subtree being cloned. It
will produce a tree with a consistent setting of the flags, but the
root doesn't have in_param/in_lvalue set unless it's intrinsic to the
type of node being cloned (e.g. AST_PARAMETER). cloneInto() will make
sure the cloned subtree has the flags consistent with the new placement
in a hierarchy.
Add asserts to make sure the old and new way of determining the flags
agree.
The following commit will replace the way in_lvalue/in_param is being
tracked in the simplify code. Make tweaks in advance so that it will
be easier to make the old way and the new way agree.
These changes all should be innocuous.
For the basic single-bit operations, opt for gate cells (`$_AND_` etc.)
instead of the coarse cells (`$and` etc.). For the emission of cells
move to the conventional module methods (`module->addAndGate`) away
from the local helpers. While at it, touch on the surrounding code.
To represent intermediate signals use the `SigBit`/`SigSpec` classes as
is customary in the Yosys codebase. Do not pass around `Wire` pointers
unless we have special reason to.
* Speed up the autoname pass by 2x. This is accomplished by only constructing IdString objects for plain strings that have a higher score.
* Defer creating IdStrings even further. This increases the speedup to 3x.
This PR speeds up by roughly 17% across a wide spectrum of designs
tested at Google. Particularly for the mux generation pass.
Co-authored-by: Rasmus Larsen <rmlarsen@google.com>
Signed-off-by: Ethan Mahintorabi <ethanmoon@google.com>
Jannis Harder
Rasmus Munk Larsen
github-actions[bot]
Martin Povišer
N. Engelhardt
Miodrag Milanović
Wanda
Ethan Mahintorabi