Generate rust code

README

@@ -1,5 +1,34 @@
 A simple script for parsing TRM register definition...
 
-Experimenting
+Pipe the `pdftotext -layout` output to the python script stdin, and specify the
+starting and ending address (absolute) of the registers.
 
-pdftotext -f 1436 -l 1438 -layout ug585-Zynq-7000-TRM.pdf - | python main.py
+This script can handle fields generation and description. This script is very
+hacky, have a lot of assumptions on the format of the PDF, so use with care,
+better check the output before you trust it.
+
+Example:
+
+pdftotext -f 1435 -l 1437 -layout ug585-Zynq-7000-TRM.pdf - | python main.py 0xF8F00000 0xF8F01FFF
+
+#[repr(C)]
+pub struct RegisterBlock {
+    /// SCU Control Register
+    pub scu_control_register: ScuControlRegister,
+    /// SCU Configuration Register
+    pub scu_configuration_register: ScuConfigurationRegister,
+}
+
+register!(scu_control_register, ScuControlRegister, RW, u32);
+register_bit!(scu_control_register, ic_standby_enable, 6);
+register_bit!(scu_control_register, scu_standby_enable, 5);
+register_bit!(scu_control_register, force_all_device_to_po, 4);
+register_bit!(scu_control_register, scu_speculative_linefil, 3);
+register_bit!(scu_control_register, scu_rams_parity_enab, 2);
+register_bit!(scu_control_register, address_filtering_enabl, 1);
+register_bit!(scu_control_register, scu_enable, 0);
+
+register!(scu_configuration_register, ScuConfigurationRegister, RO, u32);
+register_bits!(scu_configuration_register, tag_ram_sizes,u8, 8, 15);
+register_bits!(scu_configuration_register, cpus_smp,u8, 4, 7);
+register_bits!(scu_configuration_register, cpu_number,u8, 0, 1);

main.py

@@ -167,18 +167,39 @@ def interpret(reg):
     assert len(expected) == 0
     return result
 
-def snake_to_camel(name: str):
+def to_camel(name: str):
     result = []
-    start = True
+    parts = name.split('_')
-    for c in name:
+    for p in parts:
-        if c == '_':
+        if len(p) == 0:
-            start = True
+            continue
-        elif start:
+        p = p[0].upper() + p[1:]
-            start = False
+        result.append(p)
-            result.append(c.upper())
+    return ''.join(result)
+
+def to_snake(text: str):
+    if len(text) == 0:
+        return ''
+    if '_' in text:
+        return text.lower()
+
+    prev_upper = text[-1].isupper()
+    result = []
+    for c in reversed(text):
+        current_upper = c.isupper()
+        if prev_upper and not current_upper:
+            if len(result) > 0 and result[-1] != '_':
+                result.append('_')
+            result.append(c.lower())
+        elif not prev_upper and current_upper:
+            result.append(c.lower())
+            result.append('_')
         else:
             result.append(c.lower())
-    return ''.join(result)
+        prev_upper = current_upper
+    if result[-1] == '_':
+        result.pop()
+    return ''.join(reversed(result))
 
 def access_to_type(access: str):
     access = access.upper()
@@ -192,14 +213,15 @@ def fields_to_rust(reg):
     fields = reg['fields']
     name_pattern = re.compile(r'(.+\w)\d+')
     access = access_to_type(reg['access'])
+    assert reg['width'] in [8, 16, 32]
     if fields == []:
         return (f'{access}<u{reg["width"]}>', [])
     if len(fields) == 1:
         bits = fields[0]['bits']
         if bits[1] - bits[0] + 1 == reg['width']:
             return (f'{access_to_type(reg["access"])}<u{reg["width"]}>', [])
-    namespace = reg['name'].lower()
+    namespace = to_snake(reg['name'])
-    name = snake_to_camel(reg['name'])
+    name = to_camel(reg['name'])
     if 'similar' in reg:
         # remove the trailing digits
         name = name_pattern.fullmatch(name).group(1)
@@ -208,13 +230,16 @@ def fields_to_rust(reg):
     has_wtc = False
     lines = []
     for f in fields:
-        field_name = f['name'].lower()
+        field_name = to_snake(f['name'])
+        if field_name in ['na', 'reserved']:
+            continue
         field_access = f['access'].upper()
         [low, high] = f['bits']
         assert low <= high
         if field_access == 'WTC':
             has_wtc = True
         else:
+            assert field_access in ['RO', 'RW', 'WO']
             for i in range(high - low + 1):
                 bitmask |= 1 << (i + low)
         if low == high:
@@ -232,7 +257,7 @@ def fields_to_rust(reg):
                 # we did not implement WTC for multiple bits, but could be done
                 raise ValueError()
             lines.append(f'register_bits!({namespace}, {field_name},'
-                         f'{value_range}, {low}, {high});')
+                         f' {value_range}, {low}, {high});')
     if has_wtc:
         lines.insert(0, f'register!({namespace}, {name}, {access},'
                      f' u{reg["width"]}, {bitmask});')
@@ -245,49 +270,71 @@ def emit_rust(base_addr, ending_addr, registers):
     current_addr = base_addr
     reserved_id = 0
     code = []
+    fields = []
+    def advance_to(addr, width):
+        nonlocal current_addr, reserved_id, code
+        padding = addr - current_addr
+        assert padding >= 0
+        if padding > 0:
+            if padding % 4 == 0:
+                code.append(f'    unused{reserved_id}: [u32; {padding // 4}],')
+            else:
+                code.append(f'    unused{reserved_id}: [u8; {padding}],')
+            reserved_id += 1
+        assert width in [8, 16, 32]
+        current_addr += padding + width // 8
+
     for reg in registers:
-        addr = int(reg['address'], 16)
+        reg = interpret(reg)
+        (typename, lines) = fields_to_rust(reg)
+        description = reg['description']
+        addr = reg['abs']
+        if addr > ending_addr:
+            break
+        if 'similar' in reg:
+            for r in reg['similar']:
+                addr = r['abs']
                 if addr > ending_addr:
                     break
                 if addr < base_addr:
                     continue
-        padding = addr - current_addr
+                advance_to(addr, reg['width'])
-        if padding > 0:
+                # add description for the first one
-            if padding % 4 == 0:
+                if description is not None:
-                code.append(f'unused{reserved_id}: [RO<u32>; {padding // 4}],')
+                    code.append(f'    /// {description}')
+                    description = None
+                if len(lines) > 0:
+                    fields.append('')
+                    fields += lines
+                    lines = []
+                code.append(f'    pub {reg["name"].lower()}: {typename},')
         else:
-                code.append(f'unused{reserved_id}: [RO<u8>; {padding}],')
+            addr = reg['abs']
-            reserved_id += 1
+            if addr > ending_addr:
-        access = ''
+                break
-        unknown = False
+            if addr < base_addr:
-        if reg['type'] == 'ro':
+                continue
-            access = 'RO'
+            advance_to(addr, reg['width'])
-        elif reg['type'] == 'wo':
+            code.append(f'    /// {description}')
-            access = 'WO'
+            code.append(f'    pub {reg["name"].lower()}: {typename},')
-        elif reg['type'] in ['rw', 'mixed']:
+            if len(lines) > 0:
-            access = 'RW'
+                fields.append('')
-        else:
+                fields += lines
-            access = reg['type']
+    code.insert(0, '#[repr(C)]')
-            unknown = True
+    code.insert(1, 'pub struct RegisterBlock {')
-        size = int(reg['size'])
+    code.append('}')
-        if size not in [8, 16, 32]:
+    code += fields
-            unknown = True
-        current_addr += padding + size // 8
-        line = f"pub {reg['id'].lower()}: {access}<u{size}>,"
-        if unknown:
-            line = '// FIXME: ' + line
-        code.append(f"/// {reg['description']}")
-        code.append(line)
     return code
 
-parser = parse_registers()
+if __name__ == '__main__':
-for line in sys.stdin:
+    if len(sys.argv) != 3:
+        print("Please read the README")
+        exit(0)
+
+    parser = parse_registers()
+    for line in sys.stdin:
         parser.send(line)
-v = end_iterator(parser)
+    v = end_iterator(parser)
-for reg in v:
+    for line in emit_rust(int(sys.argv[1], 0), int(sys.argv[2], 0), v):
-    reg = interpret(reg)
+        print(line)
-    (name, lines) = fields_to_rust(reg)
+
-    print(reg)
-    print(name)
-    print(lines)
-    print('----')