[fpc-devel] Optimization theory

[J. Gareth Moreton]

 If I had to start somewhere though... this is an extract of the
disassembly from the pass 1 optimizer from the SVN head:

 x86_64aoptcpu.pas:105   result:=OptPass1OP(p);
 488b16                   mov    (%rsi),%rdx
 4889d9                   mov    %rbx,%rcx
 e8414f0000               callq  0x100189a30 
 4088c7                   mov    %al,%dil
 e951000000               jmpq   0x100184b48 

 90                       nop
 x86_64aoptcpu.pas:110   result:=OptPass1MOVXX(p);
 4889f2                   mov    %rsi,%rdx
 4889d9                   mov    %rbx,%rcx
 e8cd4c0000               callq  0x1001897d0 
 4088c7                   mov    %al,%dil
 eb40                     jmp    0x100184b48 

 x86_64aoptcpu.pas:112   result:=OptPass1LEA(p);
 4889f2                   mov    %rsi,%rdx
 4889d9                   mov    %rbx,%rcx
 e8ad500000               callq  0x100189bc0 
 4088c7                   mov    %al,%dil
 eb30                     jmp    0x100184b48 

 x86_64aoptcpu.pas:114   result:=OptPass1Sub(p);
 4889f2                   mov    %rsi,%rdx
 4889d9                   mov    %rbx,%rcx
 e89d580000               callq  0x10018a3c0 
 4088c7                   mov    %al,%dil
 eb20                     jmp    0x100184b48 

 There are some interesting things to note here.  OptPass1OP has "const p:
tai" rather than "var p: tai" as a parameter, and this causes an
unfortunate difference in the set-up, namely "mov (%rsi),%rdx" instead of
"mov %rsi,%rdx".  If these were fixed  to all be "var p: tai" (I might
actually do this, since it's a very simple change), then all the case nodes
will be identical, save the call address, and would be a better
experimental case for optimisation improvements.  Additionally, after the
call, ther lies "mov %al,%dil" and an identical jump (to the end of the
case block).  The MOV could easily be moved to after the jump with some
careful code analysis, which might then allow for futher optimisation,
depending on what is then done with %dil.

 Gareth aka. Kit
 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freepascal.org/pipermail/fpc-devel/attachments/20180624/ba7dac46/attachment.html>