mirrors/cosmopolitan
1
0
Fork 0
mirror of https://github.com/jart/cosmopolitan.git synced 2025-07-12 14:09:12 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Improve AVX512 feature detection

Browse source
This commit is contained in:
Justine Tunney 2024-05-05 05:45:52 -07:00
parent 317c8bc312
commit 5fd7b07fac
No known key found for this signature in database
GPG key ID: BE714B4575D6E328
7 changed files with 492 additions and 144 deletions
Download patch file Download diff file Expand all files Collapse all files

33
libc/nexgen32e/kcpuids.S
View file

@ -93,18 +93,37 @@ kCpuids:.long 0,0,0,0 // EAX=0 (Basic Processor Info)
add $4*4,%rdi
jmp 2b
3: nop
#if !X86_NEED(AVX2)
// test if cpu supports avx
testb X86_HAVE(AVX)(%r8)
jz 5f
jz 7f
testb X86_HAVE(OSXSAVE)(%r8)
jz 4f
jz 5f
xor %ecx,%ecx
xgetbv
mov %eax,%ecx
// test if operating system saves avx registers
and $XCR0_SSE|XCR0_AVX,%eax
cmp $XCR0_SSE|XCR0_AVX,%eax
je 5f
4: btr $X86_BIT(AVX),X86_WORD(AVX)(%r8)
jne 5f
// test if operating system saves avx512 registers
and $XCR0_OPMASK|XCR0_ZMM_HI256|XCR0_HI16_ZMM,%ecx
cmp $XCR0_OPMASK|XCR0_ZMM_HI256|XCR0_HI16_ZMM,%ecx
jne 6f
je 7f
// operating system doesn't support avx
5: btr $X86_BIT(AVX),X86_WORD(AVX)(%r8)
btr $X86_BIT(AVX2),X86_WORD(AVX2)(%r8)
#endif
5: pop %rbx
// operating system supports avx but not avx512
6: andl $~(1<<30|1<<28|1<<17|1<<27|1<<16|1<<21|1<<26|1<<31),KCPUIDS(7H, EBX)(%r8)
andl $~(1<<1|1<<12|1<<6|1<<11|1<<14),KCPUIDS(7H, ECX)(%r8)
andl $~(1<<2|1<<3|1<<8),KCPUIDS(7H, EDX)(%r8)
andl $~(1<<5),KCPUIDS(7H_1H, EAX)(%r8)
// we're done
7: pop %rbx
.init.end 201,_init_kCpuids