Make it possible to interpret a model #1220
Commits on Apr 8, 2024
Commits on Apr 9, 2024
Commits on Apr 10, 2024
Commits on Apr 11, 2024
-
Generator: extracted the code that generates the code to compute the …
…model. So that we can reuse them for the interpreter later on.
-
-
-
-
Commits on Apr 15, 2024
-
-
-
GeneratorInterpreter: merged generateDoubleOrConstantVariableNameCode…
…() into generateInitialisationCode(). To make it easier to work with both the Generator and Interpreter classes.
Commits on Apr 16, 2024
-
GeneratorProfile: don't define the min() and max() functions for the …
…C profile. There are indeed C functions to get the minimum/maximum of two values: fmin() and fmax().
-
-
-
Commits on Apr 17, 2024
-
-
Debug: some minor improvements to the printing of an AST / Interprete…
…rStatement as a tree.
-
-
Debug: make it easier to check the name of CI elements when printing …
…an InterpreterStatement object as a tree.
Commits on Apr 18, 2024
-
-
-
-
-
-
-
-
InterpreterStatement: removed the DIFF type.
Indeed, it's accounted for in the CI type and the `rate()` method.
-
InterpreterStatement: re-enabled the evaluation of some types.
We had disabled them for coverage reasons.
-
GeneratorProfile: simplified the definition of some trigonometric fun…
…ctions. ... and actually fixed the one for acoth()!
-
Analyser: allow a
piecewiseelement to have no children or only an ……`otherwise` element. See https://www.w3.org/TR/MathML2/chapter4.html#contm.piecewise for more details. Basically, a `piecewise` element can have 0+ `piece` child elements and optionally 1 `otherwise` child element. If there are no child elements then it's undefined, which in our case means that our InterpreterStatement class will evaluate such a `piecewise` element to NAN.
-
InterpreterStatement: don't need the
OTHERWISEtype.When evaluating it, we would have one extra call for nothing while we could directly evaluate the otherwise part.
-
GeneratorInterpreter: the default profile doesn't have an
XORopera……tor. So, we need to construct our statement in the other branch!
-
Utils: moved areNearlyEqual() to src/commonutils.cpp.
So that it can be used in tests/test_utils.cpp.
Commits on Apr 19, 2024
-
-
-
-
-
-
-
InterpreterStatement: cache a variable's index.
This means fewer calls when evaluating a model.
-
-
-
Commits on Apr 21, 2024
-
Interpreter: return the states, rates, and variables by reference.
So that the end-user can modify their values, if needed (e.g., for integration purposes).
-
Commits on Apr 22, 2024
-
-
GeneratorInterpreter: removed NLA system statements.
Since we don't currently support the interpretation of DAE models.
-
InterpretStatement: renamed to InterpretAstStatement.
So that we can compare it to our future InterpreterRpnStatement.
-
Commits on Apr 24, 2024
-
Interpreter: removed the RPN-based interpreter.
Indeed, it is 15 to 20 times slower than the AST-based interpreter. Clearly, handling a stack is more costly than handling recursion.
-
Generator tests: don't compare the compiled and interpreted models.
This was only done to check that we do indeed get the same results... and we do.
-
Coverage tests: check the values we get from the interpreter.
Very important to make sure that we do indeed get the results we expect for all aspects of the interpreter.
Commits on Apr 25, 2024
-
-
-
-
-
-
Interpreter: directly use
double *.Rather than `std::vector<double>` (and then retrieve the data using `data()`. Indeed, some solvers may want to use temporary arrays (e.g., CVODE uses a temporary array of doubles for states and rates), which means that we need to be able to set them, if needed.
Commits on Apr 26, 2024
-
-
Interpreter: don't keep track of the VOI, states, rates, and variables.
We now rely on the user providing them since some solvers may use temporary arrays to compute rates (e.g., CVODE uses temporary arrays for states and rates).
-
-
Commits on Apr 30, 2024
Commits on May 11, 2024
Commits on May 16, 2024
Commits on Aug 5, 2024
Commits on Aug 7, 2024
-
Code generator: use different arrays for constants, computed constant…
…s, and algebraic variables.
-
Generator: use different arrays for constants, computed constants, an…
…d algebraic variables for initialiseVariables().
-
Generator: use different arrays for constants, computed constants, an…
…d algebraic variables for computeRates().
-
Generator: use different arrays for constants, computed constants, an…
…d algebraic variables for computeVariables().
-
Generator: use different arrays for constants, computed constants, an…
…d algebraic variables for computeComputedConstants().
-
Commits on Aug 8, 2024
-
Tests: automatically generate the expected file contents.
To do it manually is very time consuming while here it gets done automatically. From there, we can quickly confirm, using git, whether the new expected file contents is correct.
-
Generator profile: added setters/getters for constants, computed cons…
…tants, and algebraic arrays.
-
-
-
-
Analyser: keep track of computed constants and external variables for…
… an equation. We will also need to keep track of states and constants.
Commits on Aug 9, 2024
-
-
AnalyserModel: keep track of the constants, computed constants, and e…
…xternal variables.
Commits on Aug 12, 2024
Commits on Aug 14, 2024
-
Generator: generate the CONSTANT_INFO, COMPUTED_CONSTANT_INFO, ALGEBR…
…AIC_INFO, and EXTERNAL_INFO arrays rather than just the VARIABLE_INFO array.
-
Commits on Aug 15, 2024
-
-
Generator: initialise things in initialiseVariables() in the order of…
… its parameters.
Commits on Aug 16, 2024
-
-
Generator: don't initialise our external variables in initialiseVaria…
…bles(). There was never a need to initialise them there in the first place. In fact, an external variable is effectively an algebraic variable which is computed by an end user. So, it needs to be computed in computeRates() (if available) and in computeVariables(). Also fixed an issue with initialiseVariables() not always including the VOI.
-
Analyser: a variable using an NLA should always be an algebraic varia…
…ble. Even if the NLA equation only uses constant variables and constant values.
-
Analyser: some minor cleaning up.
Constants are indexed using makeConstant().
-
-
Commits on Aug 17, 2024
-
Generator: make sure that constants are initialised when they are nee…
…ded to initialise a state.
Commits on Aug 19, 2024
-
Generator: renamed the "external" array to "externals".
Oversight on my part!
-
Generator: "initialise" computed constants in initialiseVariables().
This effectively reverts commit b1b7536. Indeed, some computed constants are initialised using an equation (e.g., x = 3 rather than x with an initial value of 3). So, there is no need to compute those computed constants many times (using computeComputedConstants()). Instead, it's much better to compute them in initialiseVariables(). Not to mention that we might have "computedConstants[1] = 1.0;" and then "computedConstants[0] = computedConstants[1];" so if we were to compute both of them in computeComputedConstants(), we would need to compute "computedConstants[1] = 1.0;" before computing "computedConstants[0] = computedConstants[1];", which would make things slightly more complicated to generate.
-
Generator: initialise things in initialiseVariables() in the order of…
… its parameters.
-
-
Generator: make sure that constants are initialised when they are nee…
…ded to initialise a state.
-
GeneratorProfile: use the constants, computed constants, and algebrai…
…c arrays rather than the variables array with findRoot() & Co.
-
GeneratorProfile: use the constants, computed constants, and algebrai…
…c arrays rather than the variables array to compute external variables.
Commits on Aug 20, 2024
Commits on Aug 21, 2024
-
-
GeneratorProfile: use the externals array to retrieve the value of ex…
…ternal variables.
Commits on Aug 30, 2024
Commits on Sep 3, 2024
Commits on Sep 5, 2024
-
-
Tests: added a test to generate some code for the DAE version of the …
…HH52 model with various external variables.
Commits on Sep 6, 2024
-
Tests: changed the external variable in algebraicEqnWithOneNonIsolate…
…dUnknownWithExternalVariable. This is so that we can test an algebraic model with an external variable (in https://github.com/agarny/libCellMLGeneratedCodeTests).
Commits on Sep 13, 2024
Commits on Oct 14, 2024
-
AnalyserEquationAst: don't include "libcellml/undefines.h" in the .cp…
…p file. It's already included in the corresponding .h file.
-
SWIG: make it work with Clang 19
Indeed, from https://releases.llvm.org/19.1.0/tools/clang/docs/ReleaseNotes.html: Added the INFINITY and NAN macros to Clang’s <float.h> freestanding implementation; these macros were defined in <math.h> in C99 but C23 added them to <float.h> in WG14 N2848.
Commits on Oct 15, 2024
Commits on Oct 16, 2024
Commits on Oct 17, 2024
-
CMake: don't use hints to find llvm-cov and llvm-profdata.
Someone might have a version of llvm-cov and llvm-profdata that they use and which is not located in /Library/Developer/CommandLineTools/usr/bin. This means that they will need to provide LLVM_BIN_DIR which is not ideal, not least since there are PREFERRED_LLVM_COV_NAMES and PREFERRED_LLVM_PROFDATA_NAMES to play with, if needed.
Commits on Oct 21, 2024
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
CMake: use the PATH as a hint to find llvm-cov and llvm-profdata.
Indeed, someone might have a version of llvm-cov and llvm-profdata that they use and which is not located in /Library/Developer/CommandLineTools/usr/bin.
-
-
-
-
Commits on Oct 22, 2024
-
-
-
-
-
-
-
-
-
-
Analyser: account for the case where a piecewise statement doesn't ha…
…ve any children.
-
-
-
-
-
-
-
-
-