alastairreid / mra_tools Goto Github PK

After building regs.asl, PMCEID's fields look odd,

$ grep PMCEID[0-1]_EL0 regs.asl
__register 64 { 63:32 IDhi, 63:32 ID } PMCEID1_EL0;
__register 64 { 31:0 IDhi, 31:0 ID } PMCEID0_EL0;

it seems to me like this should be { 63:32 IDhi, 31:0 ID } for both,
like in the xml field_lsb and field_msb are

but the script generating regs.asl is interpreting the names "IDhi[62:32]" and "IDhi[31:0]" as bit slices,
but the name and the fields, do not appear to be in harmony. I can imagine situations in which a simulator might not notice such an issue unless tests are interleaving changes to both fields.

Is this working as intended?

Hi,

This problem is more about the specification than the parsing tools. Is a standalone slice like in <X, Y> valid ASL? The specification says nothing about that case but the SHA1C instruction has pseudocode containing <Y, X> = ROL(Y:X, 32);. To me that seems like the person meant to write (Y, X) = ROL(Y:X, 32); or Y:X = ROL(Y:X, 32); (if concatenations are assignable).

The other problem is that the new slice syntax overlaps with the array-style function syntax. A free token that comes to my mind would be <<< and >>> or { and }.

PS: If this question is off-topic feel free to close it and/or direct me to the appropriate place to ask the question.

This function (the second one in t.asl) is failing on the last line

boolean HaveEL(bits(2) el)
    if el IN {EL1,EL0} then
        return TRUE;                             // EL1 and EL0 must exist
    return boolean IMPLEMENTATION_DEFINED;

I think the problem is the handling of nested statements in the if-statement

Hi,

I have been trying to understand how to use the scripts. - mainly instr2asl.py to no avail.

I'm mainly interested in understanding and possibly converting the ASL to C or python...

for example, I have this code: - can I use the script to decode it?

len = HighestSetBit(immN:NOT(imms));
bits(6)` levels;
levels = ZeroExtend(Ones(len), 6);

Many thanks
Hoss

In ASL, "ab+cd" should parse as "(ab)+(cd)".
This is not yet implemented.
This will become more of a problem once an AST is generated.

$ make all
bin/instrs2asl.py v8.2/ISA_v82A_AArch32_xml_00bet3.2 v8.2/ISA_v82A_A64_xml_00bet3.2
Traceback (most recent call last):
File "bin/instrs2asl.py", line 404, in
sys.exit(main())
File "bin/instrs2asl.py", line 319, in main
(shared,names) = readShared([ f for d in args.dir for f in glob.glob(os.path.join(d, 'shared_pseudocode.xml'))])
File "bin/instrs2asl.py", line 131, in readShared
r.defs.add(m[2])
TypeError: '_sre.SRE_Match' object is not subscriptable
make: *** [arch.asl] Error 1