https://community.idera.com/developer-tools/programming-languages/f/c-libraries/72223/inline-assembly
this code compiles well
int UTF8_avance(unsigned char*& srcUTF8, int nombre)
{
__asm {
push esi
push ebx
mov eax, nombre
mov esi, [srcUTF8]
nvCaractere:
movzx ebx, BYTE PTR [esi] // le caractère courant
jz finChaine
inc esi
s0:
bt ebx, 7
jnc finCar // un seul octet
and ebx, 0xC0 // b7-6
cmp ebx, 0xC0 // premier octet d'une suite commence par 0b11xxxxxx
je s1
neg eax // erreur
jmp fin
s1:
movzx ebx, BYTE PTR [esi] // le caractère courant
jz finChaine
inc esi
and ebx, 0xC0 // b7-6
cmp ebx, 0x80 // tous les octets aprés le premier d'une suite commencent par 0b10xxxxxx
je s1
jmp s0
finCar:
dec eax
jne nvCaractere
finChaine:
mov [srcUTF8], esi
fin:
pop esi
pop ebx
}
}
see line "[srcUTF8], esi"
int UTF8_trouveExtrait(unsigned char* srcUTF8, unsigned char*& debutUTF8, unsigned char*& finUTF)
{
__asm {
push esi
push ebx
push ecx
mov esi, [debutUTF8]
mov edi, [srcUTF8]
mov ecx, 120 // la longeur de la première chaine
cld
REPNE cmpsb
jnz s1
mov eax, -1
jmp sortie
s1:
mov ebx, eax
mov esi, [finUTF]
push esi
push eax
// call strrchr
pop ecx // nettoyer la pile on modifier directement esp
pop ecx
cmp eax, 0
jnz s2
mov eax, -2
jmp sortie
s2:
mov [debutUTF8], ebx
mov [finUTF], eax
xor eax, eax
sortie:
pop ecx
pop ebx
pop esi
}
}
this one indicates an error (MSBUILD : error E1059: instantiated into assembly here
Anyone have an explanation? I'm moving from Visual Studio to CAD Builder but problems like this are many !
Why is this written in assembly to begin with? This can easily be written in C and run just as efficiently, and be easier to management and port between compilers.
Votre réponse indique une certaine soumission aux "standards" mais elle à le mérite d'être. Il se trouve que j'aime programmer en assembleur (et ce n'est guère plus long qu'autre chose) et que j'ai des dizaines de routines qui fonctionnent bien, alors avant de passez à un autre compilateur de VS je regardes ce qu'il en est. Quand à penser que du C sera aussi rapide que de l'assembleur, c'est inexact, même si la différence dans ce cas sera faible. Mais j'ai du écrire un solveur il y a quelques années, je vous assure que la version QT, puis C étaient largement distancée par la version en assembleur.
Your answer indicates a low submission at "standards" but it exist ! I to like to program in assembler (and it's not much longer than anything else) and I have dozens of routines that work well, so before I switch from VS to another compiler I look at what's going on. Well comes to thinking that C will be as fast as the assembler, it's inaccurate, even if the difference in this case will be small. But I had to write a solver a few years ago, I assure you that the QT version, then C were largely distanced by the assembler version.
"instantiated into assembly here" is a Clang error message. C++Builder's Clang-based compilers have caveats with using inline assembly, see Embarcadero's documentation:
Differences Between Clang-enhanced C++ Compilers and Previous-Generation C++ Compilers: Inline_Assembly
In particular:
"Clang-enhanced C++ compilers allow inline assembly with these caveats:
You cannot mix assembly with C++ code.
Do not touch the stack pointer (RSP).
The assembler must use the AT&T line-by-line syntax, not the more familiar block-of-Intel syntax. For more information, see GCC-Inline-Assembly-HOWTO.
http://docwiki.embarcadero.com/RADStudio/Rio/en/Clang-enhanced_C%2B%2B_Compilers
http://www.ibiblio.org/gferg/ldp/GCC-Inline-Assembly-HOWTO.html