1 |
On Tue, Sep 5, 2017 at 10:28 PM, Grant <emailgrant@×××××.com> wrote: |
2 |
>>>> Now I'm running into "trap invalid opcode" errors on the older |
3 |
>>>> systems. Can I disable some of the newer CPU instruction sets on the |
4 |
>>>> master laptop when compiling to hopefully generate binaries that will |
5 |
>>>> work on the older systems? |
6 |
>>> |
7 |
>>> Yes. You need to find out CPU_FLAGS_X86 and "-march=" values the |
8 |
>>> machines have, and use that in make.conf. Run the commands... |
9 |
>>> |
10 |
>>> cpuinfo2cpuflags-x86 |
11 |
>>> gcc -c -Q -march=native --help=target | grep march= |
12 |
>>> |
13 |
>>> ...on the target machines. This will tell you what "native" is and |
14 |
>>> what CPU_FLAGS_X86 values to use. |
15 |
>>> |
16 |
>>> https://gcc.gnu.org/onlinedocs/gcc-6.3.0/gcc/x86-Options.html#x86-Options |
17 |
>>> lists available "march=" options, and what instruction sets they support. |
18 |
>>> E.g. my old core2 desktop shows... |
19 |
>>> |
20 |
>>> [d531][waltdnes][~] cpuinfo2cpuflags-x86 |
21 |
>>> CPU_FLAGS_X86="mmx mmxext sse sse2 sse3 ssse3" |
22 |
>>> |
23 |
>>> [d531][waltdnes][~] gcc -c -Q -march=native --help=target | grep march= |
24 |
>>> -march= core2 |
25 |
>>> |
26 |
>>> |
27 |
>>> Unless the laptops are really old, you can probably get away with... |
28 |
>>> CFLAGS="-O2 march=core2 -mfpmath=sse -fopenmp -fomit-frame-pointer -pipe -fno-unwind-tables -fno-asynchronous-unwind-tables" |
29 |
>>> |
30 |
>>> booby trap 1) Unless all machines are Intel "Atom" family, do *NOT* use |
31 |
>>> a "march=" that implements the "movbe" instruction. |
32 |
>>> |
33 |
>>> booby trap 2) If you throw in any AMD-based machines proceed with care. |
34 |
>>> |
35 |
>>> Can you post the output of... |
36 |
>>> gcc -c -Q -march=native --help=target | grep march= |
37 |
>>> ...for all the target machines? |
38 |
>> |
39 |
>> |
40 |
>> Let's see how -mtune=native goes and resort to the above if necessary. |
41 |
>> It doesn't look too bad though. |
42 |
> |
43 |
> |
44 |
> emerge -e world has finished and pushed and -mtune=native seems to |
45 |
> have solved the issue for now. |
46 |
> |
47 |
|
48 |
You might be interested in "-march=x86-64 -mtune=generic" though this |
49 |
will mean you might miss out on some optimizations. |