AVX-512 error?

[aw27]

Is there any reason why this code  builds fine with UASM but produces an error with the emulator:
SDE ERROR: Could not read memory at location 0x7f366222 nbytes= 4
For more detailed error message please run with -ptr-check and compile application with debug information
at (no-file):49 Function (no-func)

[johnsa]

Can you assemble just the avx-512 portion to a -bin output file ?
I'll have to manually check the opcodes/encoding.

I'm not sure how much testing has happened with avx512 in 32bit mode to be honest, the bulk of the testing was 64bit.

[aw27]

All right, I attach what you asked for, but probably you can just see from here:

Code Select Expand


004010EF CC                   int         3 
mov ebx, 0FFFFh
004010F0 BB FF FF 00 00       mov         ebx,0FFFFh 
kmovw k1, ebx
004010F5 C5 F8 92 CB          kmovw       k1,ebx 
vmovdqa32 zmm0 {k1} , ZMMWORD PTR Index
004010F9 62 F1 7D 49 6F 85 00 68 40 00 vmovdqa32   zmm0{k1},zmmword ptr Index (0406800h)[ebp] 
; ************ TRANSPOSING CODE START ***********************
@anotherCol:
mov ebx, 0FFFFh
00401103 BB FF FF 00 00       mov         ebx,0FFFFh 
kmovw k1, ebx
00401108 C5 F8 92 CB          kmovw       k1,ebx 
mov ebx, 0
0040110C BB 00 00 00 00       mov         ebx,0 
mov eax, 0
00401111 B8 00 00 00 00       mov         eax,0 
.while ebx<vertLoads
00401116 EB 19                jmp         main+0B0h (0401131h) 
-------------------------------------------------------------------------------
VPGATHERDD zmm1{k1}, [edx+zmm0]
00401118 62 F2 7D 49 90 8C 00 00 62 F1 7E vpgatherdd  zmm1{k1},dword ptr [eax+zmm0+7EF16200h] 
vmovdqu32 ZMMWORD PTR [edi+eax], zmm1
00401123 48                   dec         eax 

00401124 7F 0C                jg          main+0B1h (0401132h) 
00401126 38 43 81             cmp         byte ptr [ebx-7Fh],al 
add edx, REGWIDTH*COLS*4
00401129 C2 00 04             ret         400h 
0040112C 00 00                add         byte ptr [eax],al 
--------------------------------------------------------------------------------
add eax, REGWIDTH*4
0040112E 83 C0 40             add         eax,40h 
.endw


Note this part:
-------------------------------------------------------------------------------
      VPGATHERDD zmm1{k1}, [edx+zmm0]
00401118 62 F2 7D 49 90 8C 00 00 62 F1 7E vpgatherdd  zmm1{k1},dword ptr [eax+zmm0+7EF16200h] 
      vmovdqu32 ZMMWORD PTR [edi+eax], zmm1
00401123 48                   dec         eax 

00401124 7F 0C                jg          main+0B1h (0401132h) 
00401126 38 43 81             cmp         byte ptr [ebx-7Fh],al 
      add edx, REGWIDTH*COLS*4
00401129 C2 00 04             ret         400h 
0040112C 00 00                add         byte ptr [eax],al 
--------------------------------------------------------------------------------

May be is easier to see from here:

 

[johnsa]

Got it. 2.47 is about ready so this will be included.

[aw27]

 :t