[fpc-pascal] Procedural generics question

[Sven Barth]

Am 25.08.2020 um 23:47 schrieb Benito van der Zander via fpc-pascal:
> Hi,
>
> that is generating rather odd code (r40721)
>
> project1.lpr:9                            begin
> 0000000000401090 55                       push   %rbp
> 0000000000401091 4889e5                   mov    %rsp,%rbp
> 0000000000401094 488d6424f0               lea -0x10(%rsp),%rsp
> 0000000000401099 48895df8                 mov %rbx,-0x8(%rbp)
> project1.lpr:10                           if TypeInfo(T) = 
> TypeInfo(Integer) then WriteLn('an integer');
> 000000000040109D 488d1584af0700           lea 
> 0x7af84(%rip),%rdx        # 0x47c028 <RTTI_$SYSTEM_$$_LONGINT>
> 00000000004010A4 488d057daf0700           lea 
> 0x7af7d(%rip),%rax        # 0x47c028 <RTTI_$SYSTEM_$$_LONGINT>
> 00000000004010AB 4839c2                   cmp    %rax,%rdx
> 00000000004010AE 752b                     jne    0x4010db 
> <ADD$1$CRC713F463B+75>
> 00000000004010B0 e8cbcf0100               callq  0x41e080 <fpc_get_output>
> 00000000004010B5 4889c3                   mov    %rax,%rbx
> 00000000004010B8 488d15412f0600           lea 
> 0x62f41(%rip),%rdx        # 0x464000 <_$PROJECT1$_Ld1>
> 00000000004010BF 4889de                   mov    %rbx,%rsi
> 00000000004010C2 31ff                     xor    %edi,%edi
> 00000000004010C4 e867d30100               callq  0x41e430 
> <fpc_write_text_shortstr>
> 00000000004010C9 e8d2700100               callq  0x4181a0 <fpc_iocheck>
> 00000000004010CE 4889df                   mov    %rbx,%rdi
> 00000000004010D1 e85ad20100               callq  0x41e330 
> <fpc_writeln_end>
> 00000000004010D6 e8c5700100               callq  0x4181a0 <fpc_iocheck>
> project1.lpr:11                           if TypeInfo(T) = 
> TypeInfo(String) then WriteLn('a string');
> 00000000004010DB 488d0546af0700           lea 
> 0x7af46(%rip),%rax        # 0x47c028 <RTTI_$SYSTEM_$$_LONGINT>
> 00000000004010E2 488d1537b20700           lea 
> 0x7b237(%rip),%rdx        # 0x47c320 <RTTI_$SYSTEM_$$_SHORTSTRING>
> 00000000004010E9 4839d0                   cmp    %rdx,%rax
> 00000000004010EC 752b                     jne    0x401119 
> <ADD$1$CRC713F463B+137>
> 00000000004010EE e88dcf0100               callq  0x41e080 <fpc_get_output>
> 00000000004010F3 4889c3                   mov    %rax,%rbx
> 00000000004010F6 488d15132f0600           lea 
> 0x62f13(%rip),%rdx        # 0x464010 <_$PROJECT1$_Ld2>
> 00000000004010FD 4889de                   mov    %rbx,%rsi
> 0000000000401100 31ff                     xor    %edi,%edi
> 0000000000401102 e829d30100               callq  0x41e430 
> <fpc_write_text_shortstr>
> 0000000000401107 e894700100               callq  0x4181a0 <fpc_iocheck>
> 000000000040110C 4889df                   mov    %rbx,%rdi
> 000000000040110F e81cd20100               callq  0x41e330 
> <fpc_writeln_end>
> 0000000000401114 e887700100               callq  0x4181a0 <fpc_iocheck>
> project1.lpr:12                           end;
>
> The compiler should know everything about TypeInfo(T) and optimize it 
> away where possible

For the compiler these are simply ordinary pointers, it does not (yet) 
have any knowledge about "constant" pointers. Though there is a bug 
report about that exact problem somewhere...

Regards,
Sven
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freepascal.org/pipermail/fpc-pascal/attachments/20200826/f2aa2898/attachment-0001.htm>