guix/cpu should parse /sys/devices/system/cpu/modalias

I saw a post by Marcan¹ of Asahi Linux the other day about not parsing

/proc/cpuinfo and remembered that's what we're doing in (guix cpu).

I've included a couple of dumps below and some notes regarding future

implementation:

// My machine

(ins)efraim@3900XT ~$ cat /sys/devices/system/cpu/modalias

cpu:type:x86,ven0002fam0017mod0071:feature:,0000,0001,0002,0003,0004,0005,0006,0007,0008,0009,000B,000C,000D,000E,000F,0010,0011,0013,0017,0018,0019,001A,001C,0020,0021,0022,0023,0024,0025,0026,0027,0028,0029,002B,002C,002D,002E,002F,0030,0031,0034,0036,0037,0038,0039,003A,003B,003D,0064,0068,006E,0070,0074,0075,0076,0078,0079,007A,007C,007D,0080,0081,0083,0089,008C,008D,0093,0094,0095,0096,0097,0099,009A,009B,009C,009D,009E,00C0,00C1,00C3,00C4,00C5,00C6,00C7,00C8,00C9,00CA,00CC,00CD,00D1,00D6,00D7,00D8,00DA,00DC,00DD,00DE,00E2,00E4,00E6,00E8,00EA,00ED,00F0,00F1,00F2,00F3,00F5,00FA,00FB,00FC,010F,0120,0123,0125,0127,0128,012C,012F,0132,0133,0134,0137,0138,013D,0140,0141,0142,0160,0161,0162,0163,0165,016C,016E,016F,0170,0178,017C,01A0,01A1,01A2,01A4,01A6,01A9,01AA,01AC,01AD,01AF,01B2,01B3,01B8,01C2,01E0,01E1,01E2,01E3,01E4,01E5,01E6,01E7,01EA,01EB,01EC,01ED,01EF,01F0,01F1,01F4,0202,0216,0220,0221,0223,0224,0261,0262,0263,0270,0282,029C

// Berlin, aka ci.guix.gnu.org

(ins)efraim@berlin ~$ cat /sys/devices/system/cpu/modalias

cpu:type:x86,ven0002fam0017mod0001:feature:,0000,0001,0002,0003,0004,0005,0006,0007,0008,0009,000B,000C,000D,000E,000F,0010,0011,0013,0017,0018,0019,001A,001C,0020,0021,0022,0023,0024,0025,0026,0027,0028,0029,002B,002C,002D,002E,002F,0030,0031,0034,0036,0037,0038,0039,003A,003B,003D,0064,0068,006E,0070,0074,0075,0078,0079,007A,007B,007C,007D,0080,0081,0083,0089,008C,008D,0093,0094,0096,0097,0099,009A,009B,009C,009D,009E,00C0,00C1,00C2,00C3,00C4,00C5,00C6,00C7,00C8,00C9,00CC,00CD,00D1,00D6,00D7,00D8,00DA,00DC,00DD,00E2,00E8,00EA,00ED,00F1,00F3,00F5,00F8,00FA,00FC,010F,0120,0123,0125,0127,0128,0132,0133,0134,0137,013D,0140,0141,0142,0143,0165,016C,016E,016F,0170,0178,017F,01A0,01A1,01A2,01AC,01C2,01E0,01E1,01E2,01E3,01E4,01E5,01E6,01E7,01EA,01EB,01EC,01ED,01EF,01F0,0220,0221,0223,0224,0261,0262,0263,0282,029C

// pine64, kickstarter addition

(ins)efraim@3900XT ~$ ssh pine64 cat /sys/devices/system/cpu/modalias

cpu:type:aarch64:feature:,0000,0001,0002,0003,0004,0005,0006,0007,000B

// rock64

(ins)efraim@3900XT ~$ ssh rock64 cat /sys/devices/system/cpu/modalias

cpu:type:aarch64:feature:,0000,0001,0002,0003,0004,0005,0006,0007,000B

// raspberry pi 5b

(ins)efraim@3900XT ~$ ssh ubuntu.unicorn-typhon.ts.net cat /sys/devices/system/cpu/modalias

efraim@ubuntu.unicorn-typhon.ts.net: Permission denied (publickey).

(ins)efraim@3900XT ~$ ssh ubuntu@ubuntu.unicorn-typhon.ts.net cat /sys/devices/system/cpu/modalias

cpu:type:aarch64:feature:,0000,0001,0002,0003,0004,0005,0006,0007,0008,0009,000A,000B,000C,000F,0010,0014

// power9 machine from OSU for ci.guix.gnu.org

(ins)efraim@3900XT ~$ ssh guixp9 cat /sys/devices/system/cpu/modalias

cpu:type:power9:feature:,0001,0006,0007,0008,000A,000D,000E,001A,001B,001C,001E,001F,0034,0035,0036,0037,0039,003A,003B,003C,003D,003F

// rpi 1b

(ins)efraim@3900XT ~$ ssh 192.168.68.50 cat /sys/devices/system/cpu/modalias

cpu:type:v6l:feature:,0001,0002,0004,0006,0007,0008,000F

// hifive unmatched, riscv64

(ins)efraim@3900XT ~$ ssh 192.168.68.51 cat /sys/devices/system/cpu/modalias

cat: /sys/devices/system/cpu/modalias: No such file or directory

// macbook 4,1 I think

(ins)efraim@3900XT ~$ ssh israel.unicorn-typhon.ts.net cat /sys/devices/system/cpu/modalias

cpu:type:x86,ven0000fam0006mod0017:feature:,0000,0001,0002,0003,0004,0005,0006,0007,0008,0009,000B,000C,000D,000E,000F,0010,0011,0013,0015,0016,0017,0018,0019,001A,001C,001D,001F,002B,0034,003D,0068,006B,006C,006D,006F,0070,0072,0074,0075,0079,007C,0080,0082,0083,0084,0085,0087,0088,0089,008D,008E,008F,0093,00C0,00EB,00ED,00F3,00FD,00FF,0100,0101,0102,0104,0165,016C,0172,01C0

arch/x86/include/asm/cpufeatures.h

list of CPU features for x86 and x86_64

I haven't been able to find the table for other architectures yet.

¹ https://social.treehouse.systems/@marcan/113548772069263164

--

Efraim Flashner <efraim@flashner.co.il> ????? ?????

GPG key = A28B F40C 3E55 1372 662D 14F7 41AA E7DC CA3D 8351

Confidentiality cannot be guaranteed on emails sent or received unencrypted

Hello,

Efraim Flashner <efraim@flashner.co.il> skribis:

I believe it initially mimicked what GCC does for ‘-march=native’, but

it does seem to be less exhaustive than /sys/devices/cpu/modalias.

I wonder if there are version/portability issues with the latter.

Ludo’.

On Mon, Dec 09, 2024 at 06:37:22PM +0100, Ludovic Courtès wrote:

I've been thinking more about it over the past few days.

I think ultimately we need to match both parts; the kernel interfaces

with the hardware and knows (or at least should know) exactly what the

hardware can support, and gcc (or the other compilers) need to support

optimizing for those hardware bits specifically.

Also, parsing /proc/cpuinfo is already reading what the kernel says it

supports, except in an incomplete manner.

Starting with x86_64 (and i686), I think we keep GCC's definition of

what each sub-architecture is but parse the modalias file instead of

parsing cpuinfo.

I already have some code to parse modalias and some ideas for just

reusing the math bits from the kernel (it looks like ((11*32)+28) for

zen2) to adjust the if-flags to match the numerical equivalent of the

feature.

For x86_64 the kernel does have a policy of reusing flags if they've

been deprecated and unused long enough, but hopefully that won't be a

problem, even with some of our older LTS kernels.

I think for the start it would make most sense to read modalias in

addition to cpuinfo, sort based on the data in modialis, and then

fallback to cpuinfo if it doesn't match. This helps for architectures

without the modalias file (like riscv64), and I don't know what the Hurd

provides, but I'd be surprised if it had the same modalias file for

parsing.

I'll try to work on it a bit over the coming days.

--

Efraim Flashner <efraim@flashner.co.il> ????? ?????

GPG key = A28B F40C 3E55 1372 662D 14F7 41AA E7DC CA3D 8351

Confidentiality cannot be guaranteed on emails sent or received unencrypted