The state transition matrix of a node contains transition probabilities of supported types along with other numeric information. The rows of that matrix biuniquely correspond to quadruples where each quadruple consists of a source transition state, a user or mixed-type instruction invoked in the source transition state, the outcome of this instruction, and the content of look-ahead signal segment. The columns of that matrix biuniquely correspond to target transition states. Use the following function to dump a state transition matrix to a stream.
This function dumps the state transition matrix of a node of a multinode model to a stream filep according to parameters specified in *desc_p.
The argument node specifies the identifier of this node.
If node is equal to QSMM_SIG_INVALID, the function dumps the state transition matrices of all nodes of this multinode model.
If desc_p is NULL, the function uses default dumping parameters.
The argument rez1 is for future use and must be equal to 0.
In the current implementation, the function does not modify *desc_p if desc_p is not NULL.
However, in a future implementation, the function may modify *desc_p, for example, to store there statistics on the dumping process.
The function returns a non-negative value on success or a negative error code on failure. Currently, the function can return the following error codes.
QSMM_ERR_NOTFOUNDThe argument node is not QSMM_SIG_INVALID, and a node with identifier node does not exist.
QSMM_ERR_INVALThe argument desc_p is not NULL, and parameters in *desc_p are invalid.
QSMM_ERR_UNTIMELYThe model instance does not exist.
QSMM_ERR_CALLBACKA helper function for computing the relative probability of an output signal assigned to the environment state identification engine reported an error by returning NaN.
The function qsmm_actor_calc_action_prob calls the helper function.
QSMM_ERR_STORAGEA Storage API function reported storage failure.
QSMM_ERR_STATSInconsistent statistics on an action choice state or cycle type detected.
QSMM_ERR_ILSEQUnable to convert a multibyte string to a wide string or vice versa according to a current locale.
QSMM_ERR_NOMEMThere was not enough memory to perform the operation.
Errors QSMM_ERR_STORAGE, QSMM_ERR_STATS, QSMM_ERR_ILSEQ, and QSMM_ERR_NOMEM can leave the model instance in inconsistent state.
The description of a structure specifying dumping parameters is below.
This structure specifies the parameters of dumping the state transition matrix of a node. The structure contains the following fields.
An array specifying the types of probabilities to dump.
The indices of this array are the elements of qsmm_prob_e enumeration (except for its last element QSMM_PROB_COUNT) described in Emitting an Output Signal.
If an element of this array is non-zero, the function qsmm_mat_goto_dump_v2 dumps probabilities with a corresponding type.
The default is to dump probabilities of all types.
Left indent—the number of spaces to print at the beginning of each line of output. Must be a non-negative value. The default is to use indent 0.
The number of digits after the decimal point to print for probabilities. If that number is positive, use fixed-point notation. If that number is negative, use exponential notation with the number of digits after the decimal point equal to the absolute value of this field. If that number is zero, use exponential notation with 15 digits after the decimal point; this is the default mode.
The minimum value of fq field of an instance of qsmm_cycle_s structure for a matrix cell.
The function qsmm_mat_goto_dump_v2 does not output information on instances with lesser values of fq field (i.e. with lesser frequency).
The default is to use minimum frequency 0.
To improve compatibility with future versions of QSMM library, zero by the function memset an instance of qsmm_dump_mat_goto_desc_s structure before setting the fields of this instance passed to the function qsmm_mat_goto_dump_v2.
Below there is an example fragment of a dump.
When generating it, the element QSMM_PROB_LEARNED of do_print_prob field of qsmm_dump_mat_goto_desc_s structure passed to qsmm_mat_goto_dump_v2 was non-zero, and all other elements of this field were zero.
The example contains truncated fractional parts of numbers in exponential notation to make lines shorter.
* State 83
A0 |mn| O7 L0 : tmd0=183117, tmc0=6E+4, state_next=45, spur[0].val0=-4E+5, spur[1].val0=4E+2
S45 : pl=1.0E+00, spur[0].ds=-3E+05, spur[1].ds=4E+02, fq=376, ps_d=148416, ps_c=5E+04
S79 : pl=4.2E-267, spur[0].ds=-4E+04, spur[1].ds=1E+01, fq=3, ps_d=23976, ps_c=8E+03
S92 : pl=9.1E-283, spur[0].ds=-2E+04, spur[1].ds=4E+00, fq=2, ps_d=9516, ps_c=3E+03
A0 |mn| O7 L1 : tmd0=183219, tmc0=6E+4, state_next=120, spur[0].val0=-4E+5, spur[1].val0=4E+2
S63 : pl=2.2E-245, spur[0].ds=-4E+04, spur[1].ds=1E+01, fq=5, ps_d=22740, ps_c=8E+03
S87 : pl=2.0E-292, spur[0].ds=-3E+02, spur[1].ds=0E+00, fq=1, ps_d=156, ps_c=5E+01
S120 : pl=1.0E+00, spur[0].ds=-3E+05, spur[1].ds=4E+02, fq=119, ps_d=160185, ps_c=5E+04
Below there are descriptions of output pieces of information. See Structures for Accessing Storage, for details about referenced structure fields.
AiAn instruction class invoked by the node in a source transition state.
Index i uniquely identifies the instruction class in the instruction class set of the node.
The name of the instruction class enclosed between the characters ‘|’ follows the token Ai.
LiLook-ahead signal i that was in the look-ahead signal segment when the node was in a source transition state.
The number of Li tokens is equal to look-ahead signal segment length.
The positions of Li tokens are the positions of elements in the look-ahead signal segment.
OiOutcome i of an instruction class invoked by the node in a source transition state.
SiThe description of a transition to target state i.
If the state has a name assigned by the argument of stt assembler instruction, the name enclosed in double quotes follows the token Si.
RSTThis keyword can replace the tokens ‘Ax |name| Oy’ for source transition state 0. The keyword indicates that specified transitions to target states are initial transitions performed just after transferring control to the node. In this situation, the node did not yet invoke instructions, so values x and y are unknown.
fqThe value of fq field of qsmm_cycle_s structure.
paProbability of QSMM_PROB_AGGR type.
pfProbability of QSMM_PROB_FQ type.
plProbability of QSMM_PROB_LEARNED type.
ppProbability of QSMM_PROB_PROFILE type.
ps_cThe value of period_sum_c field of qsmm_cycle_s structure.
ps_dThe value of period_sum_d field of qsmm_cycle_s structure.
spur[i].dsThe value of delta_sum field of qsmm_cspur_s structure for spur type i.
Spur type 0 is usually the automatic spur.
spur[i].val0The value of val0 field of qsmm_sspur_s structure for spur type i.
Spur type 0 is usually the automatic spur.
StateThe index of a source transition state.
If the state has a name assigned by the argument of stt assembler instruction, the name enclosed in double quotes follows that index.
state_nextThe index of target state of last transition made from a source state.
That index corresponds to the value of sig_cycle_next field of qsmm_state_s structure.
The special value ‘N’ corresponds to the value QSMM_SIG_INVALID of this field.
tmc0The value of tmc0 field of qsmm_state_s structure.
tmd0The value of tmd0 field of qsmm_state_s structure.
The function qsmm_mat_goto_dump_v2 dumps only the descriptions of state transitions that have some information in statistics storage.
This approach helps reduce the length of output when dumping a sparse state transition matrix.
When the descriptions of some state transitions are absent in the output because of a positive value of fq_min field of qsmm_dump_mat_goto_desc_s structure or because statistics storage does not contain information on some state transitions, the sum of state transition probabilities with a specific type may be less than 1 in a matrix row.
If the field fq_min is zero, every not dumped transition has probability (1-p)/n, where p is the sum of probabilities of dumped transitions, and n is the number of not dumped transitions.