Command Line Options¶
- Table of contents
- Command Line Options
- Overview
Overview¶
The Trace Matcher command line looks as follows:
usage: traceutils [-h] [-d] [-fds] [-sa START_ADDR] [-se] [-sr] [-w [W_SIZE]] FILE [FILE ...] positional arguments: FILE Path to trace file optional arguments: -h, --help show this help message and exit -d, --debug Print for debug messages (default: disabled) -fds, --first-dif-stop Stop the tool on the first diversity (default: disabled) -bs B_SIZE, --boot-size B_SIZE Number of instruction records at the trace beginning that avoid matching because they are of boot section (default: unused) -sa START_ADDR, --start-addr START_ADDR Set the start address from which trace records are to be compared (default: unused) -se, --skip-equal Skip subsequent records if they differ in time field only (default: disabled) -sr, --skip-rest Skip matching for the rest of the longest trace (default: disabled) -w [W_SIZE], --window-size [W_SIZE] Matching window size in ticks (default: 1)
The detailed descriptions of some specific options are given below.
Stopping when first difference¶
When an appropriate option is enabled, the tool ceases matching process when the first non-matched record has been detected.
Boot size¶
Size of the boot section at which (or right after) start address should be located.
Start address¶
One of the possible use cases for Trace Matcher
tool is a comparison of traces with different beginnings but with the same main part. Such situation may appear at microprocessor execution traces, when the same test program has been run on two models of microprocessor (or on model\emulator and on RTL itself), but in one case trace includes preceding boot program trace and does not contain it in another case.
To reduce the number of false discrepancies, start address
is introduced. Having this option enabled, the tool will skip all the parsed records at the trace until an instruction with the specified address will not be found. An option is global for all traces.
There may be the following situations:
- If the option is not set, the matching process is started from the first record
- If the option is set but there is no such instruction record at some trace - this trace is treated as incorrect and an error is reported
Skip matching for the rest¶
There are not so many reasons to continue the matching process when there is only one not completely parsed trace. When enabled, the appropriate option disables matching in such case.
Window size¶
For example, there are two input trace files that are called log1.txt
and log2.txt
. The log1.txt
file contains the following records:
1 clk 0 IT (1) 00000004 3c080001 A svc : lui t0,0x1 1 clk R r8 00000000
The log2.txt
file contains the following records:
1 clk R r8 00000000 1 clk 0 IT (1) 00000004 3c080001 A svc : lui t0,0x1
Trace files to compare may contain same records (lines), but at different positions relative to each other. To take this effect into account and do not treat actually simultaneous but formally logged in different order events, a window concept is introduced. The idea is not to compare trace files record by record, but "record block" by "records block" with the aim to find equal but reordered records. Such block is called as "window". Window is characterized by it's size that is a number of records inside the window. By default, window size is set to 1. It means that record by record comparison will be made.
Updated by Sergey Smolov about 2 years ago · 17 revisions