forked from Github_Repos/cvw
39 lines
2.0 KiB
Plaintext
39 lines
2.0 KiB
Plaintext
Many of the scripts to build the linux ram.txt and trace files have changed over the Summer.
|
|
Specifically the parsed*.txt files have all been replaced by a single all.txt file which contains
|
|
all of the changes concurrent with a specific instruction.
|
|
Each line of all.txt is encoded in the following way.
|
|
|
|
The tokens are space deliminted (limitation the parsing function in system verilog). This could be
|
|
improved with some effort.
|
|
|
|
<Token> denotes a required token.
|
|
|
|
()? is an optional set of tokens. Exactly 0 or 1 of this pattern will occur.
|
|
The register update, memory operation, and CSR update are all possilbe but not present on all operations.
|
|
()+ is used to denote a variable number of this pattern with at least 1 instance of the pattern.
|
|
All integers are in hex and not zero extended.
|
|
|
|
<PC> <instruction bits> <instruction text> (<GPR> <Reg Number> <Value>)? (<MemR|MemW|MemRW> <Address> <WriteData if valid> <ReadData if valid>)? (<CSR> (<Name> <Value>)+)?
|
|
|
|
Example
|
|
|
|
1010 182b283 ld_t0,24(t0) GPR 5 80000000 MemR 1018 0 80000000
|
|
|
|
PC = 0x1010
|
|
Instruction encoding = 0x182_b283
|
|
instruction pneumonic (text) = ld_t0,24(t0)
|
|
Updating x5 to 0x8000_0000
|
|
Memory read at address 0x8000_0000 with read data of 0x8000_0000
|
|
|
|
CSR updates can occur in more than once for a single instruction. The multiple sets will appear as pairs of regsiter name followed by value.
|
|
|
|
**** This trace is generated using the CreateTrace.sh script.
|
|
|
|
Generation of ram.txt has not changed. Still use logBuildrootMem.sh
|
|
|
|
Only the all.txt and ram.txt are required to run modelsim's linux simulation. However there are three additional files will aid
|
|
in the debugging process. logBuildrootMem.sh was modified to also create an object dump from the vmlinux image. Using
|
|
extractFunctionRadix.sh the objdump is converted into two files vmlinux.objdump.addr and vmlinux.objdump.lab which contain
|
|
the addresses and labels of global functions in the linux binarary. The linux test bench is configured to uses these two files
|
|
to tell the user which function is currently being executed in modelsim.
|