Jorgen Lundman

OK ` add $(0x10), \DATA_OFFSET` doesn't work, but ``` add $(0x10), \DATA_OFFSET 47cd: 49 83 c3 10 addq $16, %r11 ``` Now it crashes further down - lemme walk it...

Ah no, that was the only change needed: ``` diff --git a/module/icp/asm-x86_64/modes/isalc_gcm_sse.S b/module/icp/asm-x86_64/modes/isalc_gcm_sse.S index d8bc585fc..c19443687 100644 --- a/module/icp/asm-x86_64/modes/isalc_gcm_sse.S +++ b/module/icp/asm-x86_64/modes/isalc_gcm_sse.S @@ -641,7 +641,8 @@ movdqu 16*j(\KEYSCHED), \T_key // encrypt with...

Windows compile: ``` C:\src\openzfs\out\build\x64-Debug\clang -cc1as : fatal error : error in backend: unable to evaluate offset for variable 'breg' ```

Not at all; macOS is Apple's clang: `Apple clang version 12.0.0 (clang-1200.0.32.29) Target: x86_64-apple-darwin19.6.0` and Windows is `clang version 12.0.0 Target: i686-pc-windows-msvc` Note that `clang-cl.exe` makes it run in Windows-compatibility...

OK replacing `clang-cl.exe` with just `clang.exe` did not make any difference.

The llvm guys said: > The issue is that the code uses the breg symbol like a macro, but on Windows the assembler also tries to emit the symbol itself,...

Yep, breg no longer complain in my test.S, I'll try it in ZFS as well (and the other one) Wonder if we can #ifdef _WIN32 around it too, so other...

Not sure I see how the xmmreg macros are generated, I'll play with it

Oh rats, need to figure out how Linux adds header locations.

Sheesh getting the includes to work after moving the header file is a pain. It's mostly because there is 1 set of assembler files (setjmp) in ZFS.