[fpc-devel] Producing assembly with less branches?
Stefan Glienke
sglienke at dsharp.org
Sun Jul 19 19:24:39 CEST 2020
Hi,
not sure if anything significantly changed in trunk compared to 3.2 wrt
to optimized code being generated but I am quite disappointed that fpc
(checked win64 with -O3 and -O4) does not use cmovxx instructions and
alike for the most basic things and produces terrible code like this:
unit1.pas:49 if left < right then
000000010002E3C0 39ca cmp edx,ecx
000000010002E3C2 7e06 jle 0x10002e3ca <COMPAREINT+10>
unit1.pas:50 Result := -1
000000010002E3C4 b8ffffffff mov eax,0xffffffff
000000010002E3C9 c3 ret
unit1.pas:51 else if left > right then
000000010002E3CA 39ca cmp edx,ecx
000000010002E3CC 7d06 jge 0x10002e3d4 <COMPAREINT+20>
unit1.pas:52 Result := 1
000000010002E3CE b801000000c3 mov eax,0x1
unit1.pas:54 Result := 0;
000000010002E3D4 31c0 xor eax,eax
unit1.pas:55 end;
000000010002E3D6 c3 ret
Similar for even simpler things:
unit1.pas:43 if i < 0 then
000000010002E3A1 85c0 test eax,eax
000000010002E3A3 7d03 jge 0x10002e3a8
<BUTTON1CLICK+72>
unit1.pas:44 i := 0;
000000010002E3A5 31c0 xor eax,eax
000000010002E3A7 90 nop
Imo someone should work at that and make the compiler produce less
branches. Not sure if that is on your list but it should be looked at.
--
Diese E-Mail wurde von Avast Antivirus-Software auf Viren geprüft.
https://www.avast.com/antivirus
More information about the fpc-devel
mailing list