<div dir="ltr"><div>Hi</div><div><br></div><div>[nataraj@dflyHP ~]$ fpc ttt.pas</div>Free Pascal Compiler version 3.2.2 [2023/07/04] for x86_64<br>Copyright (c) 1993-2021 by Florian Klaempfl and others<br>Target OS: DragonFly for x86-64<br>Compiling ttt.pas<br>Linking ttt<br>/usr/local/bin/ld.bfd: warning: /usr/local/lib/fpc/3.2.2/units/x86_64-dragonfly/rtl/prt0.o: missing .note.GNU-stack section implies executable stack<br>/usr/local/bin/ld.bfd: NOTE: This behaviour is deprecated and will be removed in a future version of the linker<br>121 lines compiled, 14.9 sec<br>[nataraj@dflyHP ~]$ ./ttt <br> Pascal control case: 6.7 ns/call<br> Using LEA instruction: 4.2 ns/call<br>Using ADD instructions: 4.0 ns/call<br><div><div dir="ltr" class="gmail_signature" data-smartmail="gmail_signature"><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><br></div><div dir="ltr"><br></div><div dir="ltr">Nataraj S Narayan<div>Synergy Info Systems</div><div>Software & Technology Consultants</div><div>Ettumanoor, INDIA</div><div>Ph:+91 9443211326</div></div></div></div></div></div></div></div><br></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Sat, Oct 7, 2023 at 9:39 PM J. Gareth Moreton via fpc-devel <<a href="mailto:fpc-devel@lists.freepascal.org">fpc-devel@lists.freepascal.org</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">That's interesting; I am interested to see the assembly output for the <br>
Pascal control cases. As for the 64-bit version, that was my fault <br>
since the assembly language is for Microsoft's ABI rather than the <br>
System V ABI, so it was checking a register with an undefined value. <br>
Find attached the fixed test.<br>
<br>
Kit<br>
<br>
P.S. Results on my Intel(R) Core(TM) i7-10750H<br>
<br>
Pascal control case: 2.0 ns/call<br>
Using LEA instruction: 1.7 ns/call<br>
Using ADD instructions: 1.3 ns/call<br>
<br>
On 07/10/2023 16:51, Tomas Hajny via fpc-devel wrote:<br>
> On 2023-10-07 03:57, J. Gareth Moreton via fpc-devel wrote:<br>
><br>
><br>
> Hi Kit,<br>
><br>
>> Do you think this should suffice? Originally it ran for 1,000,000<br>
>> repetitions but I fear that will take way too long on a 486, so I<br>
>> reduced it to 10,000.<br>
><br>
> OK, I tried it now. First of all, after turning on the old machine, I <br>
> realized that it wasn't Intel but AMD 486 DX4 - sorry for my bad <br>
> memory. :-( I compiled and ran the test under OS/2 there (I was too <br>
> lazy to boot it to DOS ;-) ), but I assume that it shouldn't make any <br>
> substantial difference. The ADD and LEA results were basically the <br>
> same there, both around 100 ns / call. The Pascal result was around <br>
> twice as long. Interestingly, the Pascal result for FPC 3.2.2 was <br>
> around 10% longer than the same source compiled with FPC 2.0.3 (the <br>
> assembler versions were obviously the same for both FPC versions; I <br>
> tried compiling it also with FPC 1.0.10 and the assembler versions <br>
> were more than three times slower due to missing support for the <br>
> nostackframe directive).<br>
><br>
> I tested it under the AMD Athlon 1 GHz machine as well and again, the <br>
> results for LEA and ADD are basically equal (both 3.1 ns/call) and the <br>
> result for Pascal slightly more than twice (7.3 ns/call). However, <br>
> rather surprisingly for me, the overall test run was _much_ longer <br>
> there?! Finally, I tried compiling the test on a 64-bit machine (AMD <br>
> A9-9425) with Linux (compiled for 64-bits with FPC 3.2.3 compiled from <br>
> a fresh 3.2 branch). The Pascal version shows about 4 ns/call, but the <br>
> assembler version runs forever - well, certainly much longer than my <br>
> patience lasts. I haven't tried to analyze the reasons, but that's <br>
> what I get.<br>
><br>
> Tomas<br>
><br>
><br>
><br>
>><br>
>> On 03/10/2023 06:30, Tomas Hajny via fpc-devel wrote:<br>
>>> On October 3, 2023 03:32:34 +0200, "J. Gareth Moreton via fpc-devel" <br>
>>> <<a href="mailto:fpc-devel@lists.freepascal.org" target="_blank">fpc-devel@lists.freepascal.org</a>> wrote:<br>
>>><br>
>>><br>
>>> Hii Kit,<br>
>>><br>
>>>> This is mainly to Florian, but also to anyone else who can answer <br>
>>>> the question - at which point did a complex LEA instruction (using <br>
>>>> all three input operands and some other specific circumstances) get <br>
>>>> slow? Preliminary research suggests the 486 was when it gained <br>
>>>> extra latency, and then Sandy Bridge when it got particularly bad. <br>
>>>> Icy Lake seems to be the architecture where faster LEA instructions <br>
>>>> are reintroduced, but I'm not sure about AMD processors.<br>
>>> I cannot answer your question, but if you prepare a test program, I <br>
>>> can run it on an Intel 486 DX2 100 Mhz and AMD Athlon 1 GHz machines <br>
>>> if it helps you in any way (at least I hope the 486 DX2 machine <br>
>>> should be still able to start ;-) ).<br>
>>><br>
>>> Tomas<br>
>>><br>
>>> _______________________________________________<br>
>>> fpc-devel maillist - <a href="mailto:fpc-devel@lists.freepascal.org" target="_blank">fpc-devel@lists.freepascal.org</a><br>
>>> <a href="https://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-devel" rel="noreferrer" target="_blank">https://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-devel</a><br>
>>><br>
>> _______________________________________________<br>
>> fpc-devel maillist - <a href="mailto:fpc-devel@lists.freepascal.org" target="_blank">fpc-devel@lists.freepascal.org</a><br>
>> <a href="https://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-devel" rel="noreferrer" target="_blank">https://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-devel</a><br>
> _______________________________________________<br>
> fpc-devel maillist - <a href="mailto:fpc-devel@lists.freepascal.org" target="_blank">fpc-devel@lists.freepascal.org</a><br>
> <a href="https://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-devel" rel="noreferrer" target="_blank">https://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-devel</a><br>
>_______________________________________________<br>
fpc-devel maillist - <a href="mailto:fpc-devel@lists.freepascal.org" target="_blank">fpc-devel@lists.freepascal.org</a><br>
<a href="https://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-devel" rel="noreferrer" target="_blank">https://lists.freepascal.org/cgi-bin/mailman/listinfo/fpc-devel</a><br>
</blockquote></div>