Extra short uppercase: Function UpperCaseJ Naked (s As String) As ZString Ptr Asm mov rax, QWORD Ptr [rcx] or edx, -1 _loopJ: inc edx movzx ecx, Byte Ptr [rax+rdx] jecxz _exitJ cmp ecx, 97 ' a jb _loopJ cmp ecx, 122 ' z ja _loopJ and cl, 223 mov [rax+rdx], cl jmp _loopJ _exitJ: ret End Asm End Funct...

Afaik Popcnt is a bottleneck on intel (from Sandy bridge to Skylake and derivatives). Summing 127 times (and then expanding to dword) avoids that. There is a thread on popcnt timings here . "Summing 127 times" sounds slow, but I must admit that I don't quite understand what your code does...

A variant: Function UpperCaseJ Naked (s As String) As ZString Ptr Asm mov rax, QWORD Ptr [rcx] or ecx, -1 _loopJ: inc ecx movzx edx, Byte Ptr [rax+rcx] test edx, edx jz _exitJ cmp edx, 97 ' a jb _loopJ cmp edx, 122 ' z ja _loopJ and dl, 223 mov [rax+rcx], dl jmp _loopJ _exitJ: ret End Asm End Functi...

this is more for scanning for first <x> (based on pcmpeqb) and less for counting <x>. I'm doing counting. This counts the 80h in x7: include \masm32\MasmBasic\MasmBasic.inc .data mask4 OWORD 80808080808080808080808080808080h ; 16*80h x7 OWORD 80008000000080808000800000000080h ; 7*80h Init movups xm...

Could you review the asm routine I made last week then? Advanced stuff, looks good. There is not much to gain, performance-wise; these are standard SIMD sequences, and you can rarely speed them up. You might play with aligning innermost loop to 4 or 8 bytes, but even that does not have much effect ...

aurelVZAB wrote:thanks jj !

You are welcome. I could send you a PM on Masm32 if you like.

Munair wrote:You are right

Most of the time, when it comes to Assembly ;-)
Test this:

Munair wrote:MUL doesn't overwrite EDX

MUL always overwrites edx. You can use esi edi ebx, but you must save them before and restore them after, e.g. with

It compiles fine, congrats!

using SS_OWNERDRAW is the best option. It gives us full control. But we need to draw the entire control. As mentioned earlier, you can also draw directly in the client area of the main window. Pseudocode: MakeBrush hBrush, LiteGrey MakeFont hFont, Height:-16, "Arial Black" Event Paint mov...

It might even be easier to draw on the main window's DC. No control needed.

See the doc. I'm afraid that is a case for an ownerdraw exercise...

I haven't used it for a while, but looking at old code it seems that only CDDS_PREERASE and CDDS_PREPAINT are relevant.

@everybody: cool down a bit. The format of this forum could be improved, but not much - it's already good "as is". Has anybody noticed that there has been incredible activity over the past weeks? I wish other forums would be so active.

Your code is fine, and under normal conditions there will be no flicker. What is not normal is that you have many buttons, and that means hard work for Windows serving your program. Therefore a certain delay in repainting the buttons.