[Utils][LoongArch](5/6) Add a --bits-endian option to extract-section.py

[SixWeining]

Summary

This is a split patch of D115862 which adds a --bits-endian option to
extract-section to make it possible to print bits in specified endianness.
It means that we can print instruction encoding of some targets like LoongArch
as bits[0] to bits[31] from right to left by specifing --bits-endian little.

[SixWeining]

delete useless debugging line

[xen0n]

llvm/utils/extract-section.py

64

Fixed grammar and punctuation: Print out bits in specified endianness (little or big); defaults to big

72

This probably should be bits_endian to agree with the addition above; while looking at the surrounding it seems tool_path should be readobj_path too. I'm not entirely sure though.

88–90

While we're at it, add a space after every comma for general PEP8 cleanness?

[SixWeining]

Reply to @xen0n

llvm/utils/extract-section.py

64

No problem. I will change that. Thanks.

72

OK, I will rename endian to bits_endian. Thanks.
As to tool_path maybe it should be renamed in a seperate patch if we are entirely sure.

88–90

Good suggestion. But seems that LLVM doesn't define a coding standard for python in https://llvm.org/docs/CodingStandards.html

[SixWeining]

address comments from @xen0n

[xen0n]

My comments are all addressed, but I'm not too familiar with this piece of code, so would anyone else please take a look at this too?

[myhsu]

LGTM.
Actually you can break the dependency with other patches in this series since this one is self-contained. But it's up to you.

[SixWeining]

In D116100#3303011, @myhsu wrote:

LGTM.
Actually you can break the dependency with other patches in this series since this one is self-contained. But it's up to you.

Thanks @myhsu. You're right and I have removed the dependency with other patches but keep the dependency with it's child patch (D115862).

Hi @rengolin, could we land this patch now since author of the modified file has accepted the change or we should wait other patches to be accepted and land together?

BTW, I don't have commit access yet, please help to commit the change if possible.

[rengolin]

Sure, no problems. Let's just land all patches at once, in order, after they're all approved.

[SixWeining]

In D116100#3304044, @rengolin wrote:

Sure, no problems. Let's just land all patches at once, in order, after they're all approved.

Hi @rengolin. Now all the 6 initial patches are approved and could you help to commit them? Thanks.

llvm/utils/extract-section.py

if __name__ == '__main__':
     arg_parser.add_argument('section', metavar='<name>', type=str,
             help='Name of the section to extract')
     # Output format
     format_group = arg_parser.add_mutually_exclusive_group()
     format_group.add_argument('-b', dest='format', action='store_const', const='bits',
             help='Print out in bits')
     arg_parser.add_argument('--byte-indicator', action='store_true',
             help='Whether to print a \'.\' every 8 bits in bits printing mode')
+    arg_parser.add_argument('--bits-endian', metavar='<little/big>', type=str,
+            choices=['little', 'big'],
+            help='Print out bits in specified endianness (little or big), default to big')
     format_group.add_argument('-h', dest='format', action='store_const', const='hex',
             help='Print out in hexadecimal')
     arg_parser.add_argument('--hex-width', metavar='<# of bytes>', type=int,
             help='The width (in byte) of every element in hex printing mode')
 
     arg_parser.add_argument('--help', action='help')
     arg_parser.set_defaults(format='bits', tool_path='llvm-readobj', input_file='-',
-            byte_indicator=False, hex_width=4)
+            byte_indicator=False, hex_width=4, endian='big')
     args = arg_parser.parse_args()
 
     raw_section = get_raw_section_dump(args.tool_path, args.section, args.input_file)
 
     results = []
     for line in raw_section.splitlines(False):
         if line.startswith('Hex dump'):
             continue
         parts = line.strip().split(' ')[1:]
         for part in parts[:4]:
             # exclude any non-hex dump string
             try:
                 val = int(part, 16)
                 if args.format == 'bits':
                     # divided into bytes first
-                    for byte in [(val >> off) & 0xFF for off in (24,16,8,0)]:
+                    offsets = (24,16,8,0)
+                    if args.endian == 'little':
+                        offsets = (0,8,16,24)
+                    for byte in [(val >> off) & 0xFF for off in offsets]:
                         for bit in [(byte >> off) & 1 for off in range(7, -1, -1)]:
                             results.append(str(bit))
                         if args.byte_indicator:
                             results.append('.')
                 elif args.format == 'hex':
                     assert args.hex_width <= 4 and args.hex_width > 0
                     width_bits = args.hex_width * 8
                     offsets = [off for off in range(32 - width_bits, -1, -width_bits)]
                     mask = (1 << width_bits) - 1
                     format_str = "{:0" + str(args.hex_width * 2) + "x}"
                     for word in [(val >> i) & mask for i in offsets]:
                         results.append(format_str.format(word))
             except:
                 break
     print(' '.join(results), end='')