compile without AVX512 instructions

[schnusch]

When running nixfmt on a VPS I get an illegal instruction signal. I initially tried nixfmt from NixOS/nixpkgs@107d5ef (EDIT I also tried NixOS/nixpkgs@a0f3e10, see below) but then tried building it:

$ git checkout 8d4bd69
$ nix-shell
$ cabal new-build
$ ./dist-newstyle/build/x86_64-linux/ghc-9.4.8/nixfmt-0.6.0/x/nixfmt/build/nixfmt/nixfmt < /etc/nixos/*.nix
Illegal instruction (core dumped)

I dug a little, recompiled it with cabal new-build --enable-debug-info=3, and ran it with gdb:

(gdb) continue
[...]
Thread 1 "nixfmt" received signal SIGILL, Illegal instruction.
0x00000000005753b7 in measure_off_avx ()
(gdb) disassemble 
Dump of assembler code for function measure_off_avx:
   0x00000000005753a0 <+0>:     mov    rax,rdi
   0x00000000005753a3 <+3>:     mov    rdi,rsi
   0x00000000005753a6 <+6>:     lea    rsi,[rsi-0x3f]
   0x00000000005753aa <+10>:    mov    rcx,rdx
   0x00000000005753ad <+13>:    cmp    rax,rsi
   0x00000000005753b0 <+16>:    jae    0x5753e8 <measure_off_avx+72>
   0x00000000005753b2 <+18>:    mov    edx,0xffffffbf
=> 0x00000000005753b7 <+23>:    vpbroadcastb zmm0,edx
   0x00000000005753bd <+29>:    jmp    0x5753cc <measure_off_avx+44>
   0x00000000005753bf <+31>:    nop
   0x00000000005753c0 <+32>:    add    rax,0x40
   0x00000000005753c4 <+36>:    sub    rcx,rdx
   0x00000000005753c7 <+39>:    cmp    rax,rsi
   0x00000000005753ca <+42>:    jae    0x5753e8 <measure_off_avx+72>
[...]

This answer on StackOverflow says vpbroadcastb zmm0,edx is a AVX512 instruction. Judging by /proc/cpuinfo my CPU only supports AVX2:

VPS's /proc/cpuinfo

processor       : 0                                                 
vendor_id       : AuthenticAMD                                      
cpu family      : 23                                                
model           : 49                                                                                                                                                                                                                                                              
model name      : AMD EPYC 7282 16-Core Processor                                                                                                                                                                                                                                 
stepping        : 0                                                 
microcode       : 0x830107a                                                                                                              
cpu MHz         : 2794.748                                          
cache size      : 512 KB                                            
physical id     : 0                                                 
siblings        : 4                                                 
core id         : 0                                                 
cpu cores       : 4                                                 
apicid          : 0                                                 
initial apicid  : 0                                                 
fpu             : yes                                               
fpu_exception   : yes                                               
cpuid level     : 16                                                
wp              : yes                                               
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm rep_good nopl cpuid extd_apicid tsc_known_freq pni pclmulqdq ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm cmp_legacy cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw perfctr_core ssbd ibrs ibpb stibp vmmcall fsgsbase tsc_adjust bmi1 avx2 smep bmi2 rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 clzero xsaveerptr wbnoinvd arat umip rdpid arch_capabilities          
bugs            : sysret_ss_attrs null_seg spectre_v1 spectre_v2 spec_store_bypass retbleed smt_rsb srso                                 
bogomips        : 5589.49                                           
TLB size        : 1024 4K pages                                     
clflush size    : 64                                                
cache_alignment : 64                                                
address sizes   : 40 bits physical, 48 bits virtual                 
power management:

Weirdly I can run nixfmt from NixOS/nixpkgs@a0f3e10 on my Notebook, which does not seem to support AVX512 either:

Notebook's /proc/cpuinfo

processor       : 0                                                                                                                                                                                                                                                               
vendor_id       : AuthenticAMD                                                                                                                                                                                                                                                    
cpu family      : 23                                                                                                                                                                                                                                                              
model           : 24                                                                                                                                                                                                                                                              
model name      : AMD Ryzen 3 3200U with Radeon Vega Mobile Gfx                                                                                                                                                                                                                   
stepping        : 1                                                                                                                                                                                                                                                               
microcode       : 0x8108102                                                                                                                                                                                                                                                       
cpu MHz         : 3378.967                                                                                                                                                                                                                                                        
cache size      : 512 KB                                                                                                                                                                                                                                                          
physical id     : 0                                                                                                                                                                                                                                                               
siblings        : 4                                                                                                                                                                                                                                                               
core id         : 0                                                                                                                                                                                                                                                               
cpu cores       : 2                                                                                                                                                                                                                                                               
apicid          : 0                                                                                                                                                                                                                                                               
initial apicid  : 0                                                                                                                                                                                                                                                               
fpu             : yes                                                                                                                                                                                                                                                             
fpu_exception   : yes                                                                                                                                                                                                                                                             
cpuid level     : 13                                                                                                                                                                                                                                                              
wp              : yes                                                                                                                                                                                                                                                             
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb hw_pstate ssbd ibpb vmmcall fsgsbase bmi1 avx2 smep bmi2 rdseed adx smap clflushopt sha_ni xsaveopt xsavec xgetbv1 clzero irperf xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif overflow_recov succor smca sev sev_es                             
bugs            : sysret_ss_attrs null_seg spectre_v1 spectre_v2 spec_store_bypass retbleed smt_rsb srso div0 ibpb_no_ret                                                                                                                                                         
bogomips        : 5190.28                                                                                                                                                                                                                                                         
TLB size        : 2560 4K pages                                                                                                                                                                                                                                                   
clflush size    : 64                                                                                                                                                                                                                                                              
cache_alignment : 64                                                                                                                                                                                                                                                              address sizes   : 43 bits physical, 48 bits virtual                                                                               
power management: ts ttp tm hwpstate eff_freq_ro [13] [14]

---

How do I tell cabal to build nixfmt without AVX512? (or ideally how would I override pkgs.nixfmt-rfc-style to do that?)

[schnusch]

It seems wrong cpuid extended feature flags are reported for/by the AMD EPYC 7282 processor.

measure_off_avx comes from haskell's text package, bundled with ghc.

https://github.com/haskell/text/blob/1ae86be323ff0561a313810303f779add0f7da76/cbits/measure_off.c#L46-L52

The call to __get_cpuid_count seems to report wrong results, because the AMD EPYC 7282 is not supposed to support AVX512 per its specs, but the following test program returns support for AVX512:

#include <cpuid.h>
#include <stdint.h>
#include <stdio.h>

static void print_register_bits(uint32_t x, const char *s) {
    for(size_t i = 0; i < 32; ++i) {
        printf("%s & (1 << %zu) = %lu\n", s, i, (unsigned long)(x & ((uint32_t)1 << i)));
    }
}

int main(void) {
    uint32_t eax = 0, ebx = 0, ecx = 0, edx = 0;
    __get_cpuid_count(7, 0, &eax, &ebx, &ecx, &edx);
    // https://en.wikipedia.org/wiki/CPUID#EAX=7,_ECX=0:_Extended_Features
    print_register_bits(ebx, "ebx");
    return 0;
}

Namely the following extended feature flags which should not be supported:

• avx512-f ebx & (1 << 16)
• avx512-dq ebx & (1 << 17)
• avx512-ifma ebx & (1 << 21)
• avx512-cd ebx & (1 << 28)
• avx512-bw ebx & (1 << 30)
• avx512-vl ebx & (1 << 31)

---

My current fix is to override pkgs.haskell.packages.ghc966 and add a patch where has_avx512_vl_bw always reports no support for AVX512.

final: prev: {
  haskell = prev.haskell // {
    packages = prev.haskell.packages // {
      ghc966 = prev.haskell.packages.ghc966.override (prevArgs: {
        ghc = prevArgs.ghc.overrideAttrs (prevAttrs: {
          patches = (prevAttrs.patches or [ ]) ++ [
            ./haskell-text-no-avx512.patch
          ];
        });
      });
    };
  };
}

diff --git a/cbits/measure_off.c b/cbits/measure_off.c
index 5f098f8..231f4d3 100644
--- a/libraries/text/cbits/measure_off.c
+++ b/libraries/text/cbits/measure_off.c
@@ -42,6 +42,7 @@
 
 #if defined(__x86_64__) && defined(COMPILER_SUPPORTS_AVX512)
 bool has_avx512_vl_bw() {
+  return false;
 #if (__GNUC__ >= 7 || __GNUC__ == 6 && __GNUC_MINOR__ >= 3) || defined(__clang_major__)
   uint32_t eax = 0, ebx = 0, ecx = 0, edx = 0;
   __get_cpuid_count(7, 0, &eax, &ebx, &ecx, &edx);

Overriding Haskell packages proved quite complicated (pkgs.haskell.packages.text is null because text is actually provided by GHC) and I am still not totally satisfied with the above overlay.

[infinisil]

Great detective work!