blob: e2d22e053a764e807e148c98cd6510ae930b77dd [file] [log] [blame]
/*******************************************************************************
* Tracealyzer v2.6.0 Recorder Library
* Percepio AB, www.percepio.com
*
* trcConfig.h
*
* Configuration parameters for the trace recorder library. Before using the
* trace recorder library, please check that the default settings are
* appropriate for your system, and if necessary adjust these. Most likely, you
* will need to adjust the NTask, NISR, NQueue, NMutex and NSemaphore values to
* reflect the number of such objects in your system. These may be
* over-approximated, although larger values values implies more RAM usage.
*
* Terms of Use
* This software is copyright Percepio AB. The recorder library is free for
* use together with Percepio products. You may distribute the recorder library
* in its original form, including modifications in trcHardwarePort.c/.h
* given that these modification are clearly marked as your own modifications
* and documented in the initial comment section of these source files.
* This software is the intellectual property of Percepio AB and may not be
* sold or in other ways commercially redistributed without explicit written
* permission by Percepio AB.
*
* Disclaimer
* The trace tool and recorder library is being delivered to you AS IS and
* Percepio AB makes no warranty as to its use or performance. Percepio AB does
* not and cannot warrant the performance or results you may obtain by using the
* software or documentation. Percepio AB make no warranties, express or
* implied, as to noninfringement of third party rights, merchantability, or
* fitness for any particular purpose. In no event will Percepio AB, its
* technology partners, or distributors be liable to you for any consequential,
* incidental or special damages, including any lost profits or lost savings,
* even if a representative of Percepio AB has been advised of the possibility
* of such damages, or for any claim by any third party. Some jurisdictions do
* not allow the exclusion or limitation of incidental, consequential or special
* damages, or the exclusion of implied warranties or limitations on how long an
* implied warranty may last, so the above limitations may not apply to you.
*
* Copyright Percepio AB, 2013.
* www.percepio.com
******************************************************************************/
#ifndef TRCCONFIG_H
#define TRCCONFIG_H
/*******************************************************************************
* CONFIGURATION RELATED TO CAPACITY AND ALLOCATION
******************************************************************************/
/*******************************************************************************
* EVENT_BUFFER_SIZE
*
* Macro which should be defined as an integer value.
*
* This defines the capacity of the event buffer, i.e., the number of records
* it may store. Each registered event typically use one record (4 byte), but
* vTracePrintF may use multiple records depending on the number of data args.
******************************************************************************/
#define EVENT_BUFFER_SIZE 4000 /* Adjust wrt. to available RAM */
/*******************************************************************************
* USE_LINKER_PRAGMA
*
* Macro which should be defined as an integer value, default is 0.
*
* If this is 1, the header file "recorderdata_linker_pragma.h" is included just
* before the declaration of RecorderData (in trcBase.c), i.e., the trace data
* structure. This allows the user to specify a pragma with linker options.
*
* Example (for IAR Embedded Workbench and NXP LPC17xx):
* #pragma location="AHB_RAM_MEMORY"
*
* This example instructs the IAR linker to place RecorderData in another RAM
* bank, the AHB RAM. This can also be used for other compilers with a similar
* pragmas for linker options.
*
* Note that this only applies if using static allocation, see below.
******************************************************************************/
#define USE_LINKER_PRAGMA 0
/*******************************************************************************
* SYMBOL_TABLE_SIZE
*
* Macro which should be defined as an integer value.
*
* This defines the capacity of the symbol table, in bytes. This symbol table
* stores User Events labels and names of deleted tasks, queues, or other kernel
* objects. Note that the names of active objects not stored here but in the
* Object Table. Thus, if you don't use User Events or delete any kernel
* objects you set this to a very low value, e.g. 4, but not zero (0) since
* this causes a declaration of a zero-sized array, for which the C compiler
* behavior is not standardized and may cause misaligned data.
******************************************************************************/
#define SYMBOL_TABLE_SIZE 800
#if (SYMBOL_TABLE_SIZE == 0)
#error "SYMBOL_TABLE_SIZE may not be zero!"
#endif
/*******************************************************************************
* USE_SEPARATE_USER_EVENT_BUFFER
*
* Macro which should be defined as an integer value.
* Default is zero (0).
*
* This enables and disables the use of the separate user event buffer.
*
* Note: When using the separate user event buffer, you may get an artificial
* task instance named "Unknown actor". This is added as a placeholder when the
* user event history is longer than the task scheduling history.
******************************************************************************/
#define USE_SEPARATE_USER_EVENT_BUFFER 0
/*******************************************************************************
* USER_EVENT_BUFFER_SIZE
*
* Macro which should be defined as an integer value.
*
* This defines the capacity of the user event buffer, in number of slots.
* A single user event can use between 1 and X slots, depending on the data.
*
* Only in use if USE_SEPARATE_USER_EVENT_BUFFER is set to 1.
******************************************************************************/
#define USER_EVENT_BUFFER_SIZE 10
/*******************************************************************************
* USER_EVENT_CHANNELS
*
* Macro which should be defined as an integer value.
*
* This defines the number of allowed user event channels.
*
* Only in use if USE_SEPARATE_USER_EVENT_BUFFER is set to 1.
******************************************************************************/
#define CHANNEL_FORMAT_PAIRS 32
/*******************************************************************************
* NTask, NISR, NQueue, NSemaphore, NMutex
*
* A group of Macros which should be defined as an integer value of zero (0)
* or larger.
*
* This defines the capacity of the Object Property Table - the maximum number
* of objects active at any given point within each object class.
*
* NOTE: In case objects are deleted and created during runtime, this setting
* does not limit the total amount of objects, only the number of concurrently
* active objects.
*
* Using too small values will give an error message through the vTraceError
* routine, which makes the error message appear when opening the trace data
* in Tracealyzer. If you are using the recorder status monitor task,
* any error messages are displayed in console prints, assuming that the
* print macro has been defined properly (vConsolePrintMessage).
*
* It can be wise to start with very large values for these constants,
* unless you are very confident on these numbers. Then do a recording and
* check the actual usage in Tracealyzer. This is shown by selecting
* View -> Trace Details -> Resource Usage -> Object Table
*
* NOTE 2: Remember to account for all tasks and other objects created by
* the kernel, such as the IDLE task, any timer tasks, and any tasks created
* by other 3rd party software components, such as communication stacks.
* Moreover, one task slot is used to indicate "(startup)", i.e., a fictive
* task that represent the time before the scheduler starts.
* NTask should thus be at least 2-3 slots larger than your application task count.
*
******************************************************************************/
#define NTask 15
#define NISR 5
#define NQueue 10
#define NSemaphore 10
#define NMutex 10
#define NTimer 2
#define NEventGroup 2
/* Maximum object name length for each class (includes zero termination) */
#define NameLenTask 15
#define NameLenISR 15
#define NameLenQueue 15
#define NameLenSemaphore 15
#define NameLenMutex 15
#define NameLenTimer 15
#define NameLenEventGroup 15
/******************************************************************************
* TRACE_DESCRIPTION
*
* Macro which should be defined as a string.
*
* This string is stored in the trace and displayed in Tracealyzer. Can be
* used to store, e.g., system version or build date. This is also used to store
* internal error messages from the recorder, which if occurs overwrites the
* value defined here. This may be maximum 256 chars.
*****************************************************************************/
#define TRACE_DESCRIPTION "Tracealyzer Recorder Test Program"
/******************************************************************************
* TRACE_DESCRIPTION_MAX_LENGTH
*
* The maximum length (including zero termination) for the TRACE_DESCRIPTION
* string. Since this string also is used for internal error messages from the
* recorder do not make it too short, as this may truncate the error messages.
* Default is 80.
* Maximum allowed length is 256 - the trace will fail to load if longer.
*****************************************************************************/
#define TRACE_DESCRIPTION_MAX_LENGTH 80
/******************************************************************************
* TRACE_DATA_ALLOCATION
*
* This defines how to allocate the recorder data structure, i.e., using a
* static declaration or using a dynamic allocation in runtime (malloc).
*
* Should be one of these two options:
* - TRACE_DATA_ALLOCATION_STATIC (default)
* - TRACE_DATA_ALLOCATION_DYNAMIC
*
* Using static allocation has the benefits of compile-time errors if the buffer
* is too large (too large constants in trcConfig.h) and no need to call the
* initialization routine (xTraceInitTraceData).
*
* Using dynamic allocation may give more flexibility in some cases.
*****************************************************************************/
#define TRACE_DATA_ALLOCATION TRACE_DATA_ALLOCATION_STATIC
/******************************************************************************
* CONFIGURATION REGARDING WHAT CODE/FEATURES TO INCLUDE
*****************************************************************************/
/******************************************************************************
* USE_TRACE_ASSERT
*
* Macro which should be defined as either zero (0) or one (1).
* Default is 0.
*
* If this is one (1), the TRACE_ASSERT macro will verify that a condition is
* true. If the condition is false, vTraceError() will be called.
*****************************************************************************/
#define USE_TRACE_ASSERT 1
/******************************************************************************
* INCLUDE_FLOAT_SUPPORT
*
* Macro which should be defined as either zero (0) or one (1).
* Default is 1.
*
* If this is zero (0), all references to floating point values are removed,
* in case floating point values are not supported by the platform used.
* Floating point values are only used in vTracePrintF and its subroutines, to
* store float (%f) or double (%lf) argments.
*
* Note: vTracePrintF can still be used with integer and string arguments in
* either case.
*****************************************************************************/
#define INCLUDE_FLOAT_SUPPORT 1
/******************************************************************************
* INCLUDE_USER_EVENTS
*
* Macro which should be defined as either zero (0) or one (1).
* Default is 1.
*
* If this is zero (0) the code for creating User Events is excluded to
* reduce code size. User Events are application-generated events, like
* "printf" but for the trace log instead of console output. User Events are
* much faster than a printf and can therefore be used in timing critical code.
* See vTraceUserEvent() and vTracePrintF() in trcUser.h
*
* Note that User Events are not displayed in FreeRTOS+Trace Free Edition.
*****************************************************************************/
#define INCLUDE_USER_EVENTS 1
/*****************************************************************************
* INCLUDE_READY_EVENTS
*
* Macro which should be defined as either zero (0) or one (1).
* Default is 1.
*
* If this is zero (0), the code for recording Ready events is
* excluded. Note, this will make it impossible to calculate the correct
* response times.
*****************************************************************************/
#define INCLUDE_READY_EVENTS 1
/*****************************************************************************
* INCLUDE_NEW_TIME_EVENTS
*
* Macro which should be defined as either zero (0) or one (1).
* Default is 0.
*
* If this is zero (1), events will be generated whenever the os clock is
* increased.
*****************************************************************************/
#define INCLUDE_NEW_TIME_EVENTS 0
/*****************************************************************************
* INCLUDE_ISR_TRACING
*
* Macro which should be defined as either zero (0) or one (1).
* Default is 1.
*
* If this is zero (0), the code for recording Interrupt Service Routines is
* excluded to reduce code size.
*
* Note, if the kernel has no central interrupt dispatcher, recording ISRs
* require that you insert calls to vTraceStoreISRBegin and vTraceStoreISREnd
* in your interrupt handlers.
*****************************************************************************/
#define INCLUDE_ISR_TRACING 1
/******************************************************************************
* INCLUDE_OBJECT_DELETE
*
* Macro which should be defined as either zero (0) or one (1).
* Default is 1.
*
* This must be enabled (1) if tasks, queues or other
* traced kernel objects are deleted at runtime. If no deletes are made, this
* can be set to 0 in order to exclude the delete-handling code.
*****************************************************************************/
#define INCLUDE_OBJECT_DELETE 1
/******************************************************************************
* INCLUDE_MEMMANG_EVENTS
*
* Macro which should be defined as either zero (0) or one (1).
* Default is 1.
*
* This controls if malloc and free calls should be traced. Set this to zero to
* exclude malloc/free calls from the tracing.
*****************************************************************************/
#define INCLUDE_MEMMANG_EVENTS 1
/******************************************************************************
* CONFIGURATION RELATED TO BEHAVIOR
*****************************************************************************/
/******************************************************************************
* TRACE_RECORDER_STORE_MODE
*
* Macro which should be defined as one of:
* - TRACE_STORE_MODE_RING_BUFFER
* - TRACE_STORE_MODE_STOP_WHEN_FULL
* Default is TRACE_STORE_MODE_RING_BUFFER.
*
* With TRACE_RECORDER_STORE_MODE set to TRACE_STORE_MODE_RING_BUFFER, the events are
* stored in a ring buffer, i.e., where the oldest events are overwritten when
* the buffer becomes full. This allows you to get the last events leading up
* to an interesting state, e.g., an error, without having a large trace buffer
* for string the whole run since startup. In this mode, the recorder can run
* "forever" as the buffer never gets full, i.e., in the sense that it always
* has room for more events.
*
* To fetch the trace in mode TRACE_STORE_MODE_RING_BUFFER, you need to first halt the
* system using your debugger and then do a RAM dump, or to explicitly stop the
* recorder using vTraceStop() and then store/upload the trace data using a
* task that you need to provide yourself. The trace data is found in the struct
* RecorderData, initialized in trcBase.c.
*
* Note that, if you upload the trace using a RAM dump, i.e., when the system is
* halted on a breakpoint or by a debugger command, there is no need to stop the
* recorder first.
*
* When TRACE_RECORDER_STORE_MODE is TRACE_STORE_MODE_STOP_WHEN_FULL, the recording is
* stopped when the buffer becomes full. When the recorder stops itself this way
* vTracePortEnd() is called which allows for custom actions, such as triggering
* a task that stores the trace buffer, i.e., in case taking a RAM dump
* using an on-chip debugger is not possible. In the Windows port, vTracePortEnd
* saves the trace to file directly, but this is not recommended in a real-time
* system since the scheduler is blocked during the processing of vTracePortEnd.
*****************************************************************************/
#define TRACE_RECORDER_STORE_MODE TRACE_STORE_MODE_RING_BUFFER
/******************************************************************************
* STOP_AFTER_N_EVENTS
*
* Macro which should be defined as an integer value, or not defined.
* Default is -1
*
* STOP_AFTER_N_EVENTS is intended for tests of the ring buffer mode (when
* RECORDER_STORE_MODE is STORE_MODE_RING_BUFFER). It stops the recording when
* the specified number of events has been observed. This value can be larger
* than the buffer size, to allow for test of the "wrapping around" that occurs
* in ring buffer mode . A negative value (or no definition of this macro)
* disables this feature.
*****************************************************************************/
#define STOP_AFTER_N_EVENTS -1
/******************************************************************************
* USE_IMPLICIT_IFE_RULES
*
* Macro which should be defined as either zero (0) or one (1).
* Default is 1.
*
* ### Instance Finish Events (IFE) ###
*
* For tasks with "infinite" main loops (non-terminating tasks), the concept
* of a task instance has no clear definition, it is an application-specific
* thing. Tracealyzer allows you to define Instance Finish Events (IFEs),
* which marks the point in a cyclic task when the "task instance" ends.
* The IFE is a blocking kernel call, typically in the main loop of a task
* which typically reads a message queue, waits for a semaphore or performs
* an explicit delay.
*
* If USE_IMPLICIT_IFE_RULES is one (1), the kernel macros (trcKernelPort.h)
* will define what kernel calls are considered by default to be IFEs.
*
* However, Implicit IFEs only applies to blocking kernel calls. If a
* service reads a message without blocking, it does not create a new
* instance since no blocking occurred.
*
* Moreover, the actual IFE might sometimes be another blocking call. We
* therefore allow for user-defined Explicit IFEs by calling
*
* vTraceTaskInstanceIsFinished()
*
* right before the kernel call considered as IFE. This does not create an
* additional event but instead stores the service code and object handle
* of the IFE call as properties of the task.
*
* If using Explicit IFEs and the task also calls an Implicit IFE, this may
* result in additional incorrect task instances.
* This is solved by disabling the Implicit IFEs for the task, by adding
* a call to
*
* vTraceTaskSkipDefaultInstanceFinishedEvents()
*
* in the very beginning of that task. This allows you to combine Explicit IFEs
* for some tasks with Implicit IFEs for the rest of the tasks, if
* USE_IMPLICIT_IFE_RULES is 1.
*
* By setting USE_IMPLICIT_IFE_RULES to zero (0), the implicit IFEs are disabled
* for all tasks. Tasks will then be considered to have a single instance only,
* covering all execution fragments, unless you define an explicit IFE in each
* task by calling vTraceTaskInstanceIsFinished before the blocking call.
*****************************************************************************/
#define USE_IMPLICIT_IFE_RULES 1
/******************************************************************************
* USE_16BIT_OBJECT_HANDLES
*
* Macro which should be defined as either zero (0) or one (1).
* Default is 0.
*
* If set to 0 (zero), the recorder uses 8-bit handles to identify kernel
* objects such as tasks and queues. This limits the supported number of
* concurrently active objects to 255 of each type (object class).
*
* If set to 1 (one), the recorder uses 16-bit handles to identify kernel
* objects such as tasks and queues. This limits the supported number of
* concurrent objects to 65535 of each type (object class). However, since the
* object property table is limited to 64 KB, the practical limit is about
* 3000 objects in total.
*
* NOTE: An object with a high ID (> 255) will generate an extra event
* (= 4 byte) in the event buffer.
*
* NOTE: Some internal tables in the recorder gets larger when using 16-bit
* handles. The additional RAM usage is 5-10 byte plus 1 byte per kernel object
*, i.e., task, queue, semaphore, mutex, etc.
*****************************************************************************/
#define USE_16BIT_OBJECT_HANDLES 0
/****** Port Name ******************** Code ** Official ** OS Platform ******
* PORT_APPLICATION_DEFINED -2 - -
* PORT_NOT_SET -1 - -
* PORT_HWIndependent 0 Yes Any
* PORT_Win32 1 Yes FreeRTOS Win32
* PORT_Atmel_AT91SAM7 2 No Any
* PORT_Atmel_UC3A0 3 No Any
* PORT_ARM_CortexM 4 Yes Any
* PORT_Renesas_RX600 5 Yes Any
* PORT_Microchip_dsPIC_AND_PIC24 6 Yes Any
* PORT_TEXAS_INSTRUMENTS_TMS570 7 No Any
* PORT_TEXAS_INSTRUMENTS_MSP430 8 No Any
* PORT_MICROCHIP_PIC32 9 No Any
* PORT_XILINX_PPC405 10 No FreeRTOS
* PORT_XILINX_PPC440 11 No FreeRTOS
* PORT_XILINX_MICROBLAZE 12 No Any
* PORT_NXP_LPC210X 13 No Any
*****************************************************************************/
#define SELECTED_PORT PORT_ARM_CortexM
#if (SELECTED_PORT == PORT_NOT_SET)
#error "You need to define SELECTED_PORT here!"
#endif
/******************************************************************************
* USE_PRIMASK_CS (for Cortex M devices only)
*
* An integer constant that selects between two options for the critical
* sections of the recorder library.
*
* 0: The default FreeRTOS critical section (BASEPRI) - default setting
* 1: Always disable ALL interrupts (using PRIMASK)
*
* Option 0 uses the standard FreeRTOS macros for critical sections.
* However, on Cortex-M devices they only disable interrupts with priorities
* below a certain configurable level, while higher priority ISRs remain active.
* Such high-priority ISRs may not use the recorder functions in this mode.
*
* Option 1 allows you to safely call the recorder from any ISR, independent of
* the interrupt priority. This mode may however cause higher IRQ latencies
* (some microseconds) since ALL configurable interrupts are disabled during
* the recorder's critical sections in this mode, using the PRIMASK register.
******************************************************************************/
#define USE_PRIMASK_CS 0
/******************************************************************************
* HEAP_SIZE_BELOW_16M
*
* An integer constant that can be used to reduce the buffer usage of memory
* allocation events (malloc/free). This value should be 1 if the heap size is
* below 16 MB (2^24 byte), and you can live with addresses truncated to the
* lower 24 bit. Otherwise set it to 0 to get the full 32-bit addresses.
******************************************************************************/
#define HEAP_SIZE_BELOW_16M 0
#endif