Texas Instruments Corporation

John M Johnsen

Manisha Agarwala

Lewis Nardini

Maria B H Gill

Bryan Joseph Thome

Gary L Swoboda

Jose Luis Flores

Donald Edward Steiss

Joel J Graber

James O Bondi

Trace data streams are generated for tracing target processor activity. Various trace data streams are synchronized using markers called sync points. The sync points provide a unique identifier field and a context to the data that will follow it. All trace data streams may generate a sync point with this unique identifier. These unique identifiers allow synchronization between multiple trace data streams. When multiple trace data streams are on, it is possible that the data input rate may be higher than the data output rate. If synchronization is lost in such a case, there must be a scheme to resynchronize the streams. This invention is a technique for this needed resynchronization.

Maintaining coherent synchronization between data streams on detection of overflow

Data streams are generated for tracing target processor activity. When multiple streams are on, they are written at different times into their individual FIFO. It is possible that for a specific stream, the length and fields of the data that should be exported vary. This invention is a scheme to send out only the relevant fields.

Dynamic data trace output scheme

In the case of tracing processor activity and generating data streams multiple triggers can be generated at the same time. The issue is further complicated in a protected pipeline where certain locations are considered as in illegal instruction boundary. During those cycles certain information is invalid and cannot be transmitted to the user. Thus a received trace trigger cannot begin. This invention resolves all ambiguities related to multiple triggers so that the user has a known predictable behavior based on the setup of the triggers.

Trigger ordering for trace streams when multiple triggers accumulate

A method of trace collection in a data processor begins trace data collection even if a trace trigger is received during an interval when a central processing unit is stalled. Trace data collection is deferred if a trace trigger is received during an interval of an invalid instruction boundary until a valid instruction boundary.

Precise detection of triggers and trigger ordering for asynchronous events

A saturating count counts received event signals up to a first predetermined number. An overflow counter counts overflows up to a second predetermined number. The counter indicates overflow when the overflow count is non-zero and saturates and stops counting at a maximum count when the overflow count reaches the second predetermined number. The counter can be read via a register read operation. The sum of the sum of the first predetermined number of bits and the second predetermined number of bits being an integral multiple of 8 bits.

Reporting a saturated counter value

During a trace the timing stream has the active and stall information, PC stream has all the discontinuity information, and the data stream has all the data log information. The various streams are synchronized using markers called sync points. The sync points provide a unique identifier field and a context to the data that will follow it. After data corruption a sync point is inserted into the data stream. The ID of this sync point may repeat a previous sync point ID.

Method for generating timing data packet

A method of trace data compression receives trace data on a first port and a second port stores a prior data value. If trace data is received on only one port, then that trace data is transmitted as an indication of matching and non-matching sections between the current trace data and the stored data value and the non-matching sections of the current trace data on the one port. If trace data is received on both ports, then the first port trace data is transmitted relative to the prior stored value and the second port trace data is transmitted relative to the first port trace data. The stored prior data is reset to zero upon each initiation or termination of trace data on either port. The stored prior value is set to the second port value or the first port value if no second port value is received.

Multi-port trace data handling

This invention provides trace address compression by comparing respective bytes of a current trace address with a stored comparison address. Only the least significant bytes of the current trace address that do not match the comparison address or are less significant than any section of the current trace address that does not match the comparison address are transmitted. This sometimes reduces the amount of data that needs to be transmitted. The comparison address is specified by a central processing unit via a memory mapped register write operation.

Programmable extended compression mask for dynamic trace

A method of tracing data processor activity with recover from detection of trace stream corruption. If the first trace data following detection of corruption is not a program counter sync point, then the trace transmits an indication of the current program counter address in an offset format from the program counter address of a last transmitted program counter sync point and then transmits trace data in event offset format. If the first trace data following detection of corruption is a program counter sync point, then the trace transmits trace data in event offset format.

Recovery from corruption using event offset format in data trace

A system comprising a processor adapted to execute software code comprising branch instructions and a trace logic coupled to the processor and adapted to generate a branch packet comprising branch bits. At least some of the branch bits are associated with branch instructions executed by the processor. The trace logic flushes invalid branch bits in the branch packet with a common bit, the common bit an inverse of a valid branch bit. The trace logic outputs the branch packet with an indicator comprising the valid branch bit.

Efficient transfer of branch information

This invention provides trace address compression by comparing respective bytes of a current trace address with a stored prior trace address. Only the least significant bytes of the current trace address that do not match the stored prior trace address or are less significant than any section of the current trace address that does not match the stored prior trace address are transmitted. This sometimes reduces the amount of data that needs to be transmitted. The prior trace address may be updated with the current trace address if there is a complete mismatch.

Progressive extended compression mask for dynamic trace

A method of preventing trace data first-in-first-out buffer overflow in a pipelined data processor stops new instructions when a trace data first-in-first-out buffer is in danger of overflowing. The method also stalls a predetermined number of pipeline stages in the pipeline ahead of the first pipeline stage. The trace data first-in-first-out buffer is emptied while the pipeline is stalled. On restart, the stalled pipeline stages are restarted ahead of re-enabling new instructions. Asynchronous trigger events received during the stall may be buffered and unrolled in order or merely stored and applied simultaneously on restart.

Growing community of inventors

Richardson, TX, United States of America

John M Johnsen