Implement SSE3 support in V86 #906

spetterman66 · 2023-09-01T14:27:18Z

No description provided.

gen/x86_table.js

src/rust/cpu/instructions_0f.rs

spetterman66 · 2023-09-04T16:52:34Z

nice :/

spetterman66 · 2023-09-06T13:03:43Z

SSE3 SUPPORT IN V86 WE DID IT BOYS!! @copy check it out :D

spetterman66 · 2023-09-06T17:21:10Z

hmm... but why doesn't sse3 show up on cat /proc/cpuinfo?

copy · 2023-09-07T00:26:17Z

hmm... but why doesn't sse3 show up on cat /proc/cpuinfo?

Good question. It does show up in other OSes (e.g. KolibriOS).

copy · 2023-09-07T00:27:28Z

Did you test Windows 8 and/or Android with this?

Ryan8085309853409 · 2023-09-07T05:44:04Z

Don't have time rn, will do it later (I use this account on my phone)

spetterman66 · 2023-09-07T11:49:28Z

(here i am btw)

spetterman66 · 2023-09-07T12:31:24Z

well, now we have sse3 support, but we have problems with APIC i guess

spetterman66 · 2023-09-07T12:40:05Z

android-x86 4.3 without ACPI fails here. I am going to implement FISTTP

SuperMaxusa · 2023-09-07T14:43:03Z

It does show up in other OSes (e.g. KolibriOS).

Agreed, in CPU-Z detects SSE3, but in Linux's /proc/cpuinfo pni (Prescott New Instructions, second name of SSE3) is not detected

Also, Alpine Linux 3.18.2 (iso) won't booting for me (on 5664eea worked properly):

Log

[   62.679000] rcu: INFO: rcu_preempt detected stalls on CPUs/tasks:
[   62.679000]  (detected by 0, t=6002 jiffies, g=-1195, q=1 ncpus=1)
[   62.679000] rcu: All QSes seen, last rcu_preempt kthread activity 6002 (-23924--29926), jiffies_till_next_fqs=1, root ->qsmask 0x0
[   62.679000] rcu: rcu_preempt kthread starved for 6002 jiffies! g-1195 f0x2 RCU_GP_WAIT_FQS(5) ->state=0x0 ->cpu=0
[   62.679000] rcu:     Unless rcu_preempt kthread gets sufficient CPU time, OOM is now expected behavior.
[   62.679000] rcu: RCU grace-period kthread stack dump:
[   62.679000] task:rcu_preempt     state:R  running task     stack:0     pid:15    ppid:2      flags:0x00004000
[   62.679000] Call Trace:
[   62.679000]  __schedule+0x38d/0x1140
[   62.679000]  schedule+0x56/0xf0
[   62.679000]  schedule_timeout+0x6c/0x110
[   62.679000]  ? __bpf_trace_tick_stop+0x30/0x30
[   62.679000]  rcu_gp_fqs_loop+0xfa/0x450
[   62.679000]  ? rcu_gp_init+0x2a2/0x4d0
[   62.679000]  rcu_gp_kthread+0xc5/0x170
[   62.679000]  kthread+0xd5/0x100
[   62.679000]  ? rcu_gp_init+0x4d0/0x4d0
[   62.679000]  ? kthread_complete_and_exit+0x20/0x20
[   62.679000]  ret_from_fork+0x1c/0x28
[   62.679000] rcu: Stack dump where RCU GP kthread last ran:
[   62.679000] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 6.1.51-0-virt #1-Alpine
[   62.679000] EIP: delay_tsc+0x6a/0xd0
[   62.679000] Code: 00 39 d8 75 3b 0f ae e8 0f 31 8b 4d ec 89 c6 89 d7 2b 45 e4 1b 55 e8 39 c8 89 d0 1b 45 f0 73 38 b8 01 00 00 00 e8 66 2e 97 ff <64> a1 34 08 c0 ca 85 c0 75 bc e8 f7 2c 04 00 eb b5 8d 74 26 00 90
[   62.679000] EAX: 00000000 EBX: 00000000 ECX: 00000000 EDX: 00000000
[   62.679000] ESI: c36c4a80 EDI: 00000014 EBP: c10cff08 ESP: c10cfeec
[   62.679000] DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0068 EFLAGS: 00000246
[   62.679000] CR0: 80050033 CR2: ff7ff000 CR3: 0ac16000 CR4: 000006b0
[   62.679000] DR0: 00000000 DR1: 00000000 DR2: 00000000 DR3: 00000000
[   62.679000] DR6: 00000000 DR7: 00000000
[   62.679000] Call Trace:
[   62.679000]  
[   62.679000]  ? show_regs.cold+0x14/0x1a
[   62.679000]  ? dump_cpu_task+0x61/0x70
[   62.679000]  ? rcu_check_gp_kthread_starvation.cold+0x157/0x15d
[   62.679000]  ? rcu_sched_clock_irq+0x9a1/0x9d0
[   62.679000]  ? update_process_times+0x63/0x90
[   62.679000]  ? tick_periodic+0x31/0x90
[   62.679000]  ? tick_handle_periodic+0x1e/0x70
[   62.679000]  ? get_stack_info+0x140/0x140
[   62.679000]  ? timer_interrupt+0xf/0x20
[   62.679000]  ? __handle_irq_event_percpu+0x3b/0x160
[   62.679000]  ? handle_irq_event+0x29/0x70
[   62.679000]  ? handle_fasteoi_nmi+0x100/0x100
[   62.679000]  ? handle_level_irq+0x8f/0x170
[   62.679000]  ? __handle_irq+0x9e/0xc0
[   62.679000]  
[   62.679000]  ? __common_interrupt+0x54/0xf0
[   62.679000]  ? common_interrupt+0x34/0x50
[   62.679000]  ? asm_common_interrupt+0x102/0x140
[   62.679000]  ? delay_tsc+0x6a/0xd0
[   62.679000]  ? delay_tsc+0x6a/0xd0
[   62.679000]  __const_udelay+0x2e/0x40
[   62.679000]  test_nmi_ipi.constprop.0+0x94/0xe1
[   62.679000]  local_ipi+0x36/0x52
[   62.679000]  dotest.constprop.0+0xc/0xb6
[   62.679000]  nmi_selftest+0x7e/0x19e
[   62.679000]  native_smp_cpus_done+0x8a/0x191
[   62.679000]  smp_init+0x9f/0xbc
[   62.679000]  kernel_init_freeable+0x13b/0x291
[   62.679000]  ? rest_init+0xc0/0xc0
[   62.679000]  kernel_init+0x12/0x100
[   62.679000]  ret_from_fork+0x1c/0x28

tested commit: 6123197

spetterman66 · 2023-09-07T15:12:21Z

aaaaaaand the ci has failed, looks like some issue with the fpu :/

spetterman66 · 2023-09-07T18:27:18Z

then theres smth wrong with the fisttp implementation, idk what else lmao

spetterman66 · 2023-09-08T17:35:14Z

I think this CI thing has an error:

/home/runner/work/v86/v86/tests/nasm/run.js:4
process.on("unhandledRejection", exn => { throw exn; });
                                          ^

[LinkError: WebAssembly.instantiate(): Import #15 module="e" function="fpu_fisttp" error: function import requires a callable]

Node.js v18.17.1
Worker error code: 1```

copy · 2023-09-09T16:50:50Z

I think this CI thing has an error

That's a bug in the code, fpu_fisttp isn't defined.

spetterman66 · 2023-09-09T19:15:36Z

works now?

spetterman66 · 2023-09-10T06:59:21Z

works now?

yes it does!

spetterman66 · 2023-09-10T07:35:16Z

we fkn did it 💯

spetterman66 · 2023-09-11T12:03:08Z

@copy sorry for being inpatient, but can you merge this pull request, since we have sse3 support?

copy

sorry for being inpatient, but can you merge this pull request, since we have sse3 support?

I cleared the cache, because the nasmtests needed to be generated, and it found a couple of bugs.

copy · 2023-09-11T21:13:30Z

src/rust/cpu/fpu.rs

+#[no_mangle]
+pub unsafe fn fpu_fisttpm64(addr: i32) {
+    return_on_pagefault!(writable_or_pagefault(addr, 8));
+    let v = fpu_convert_to_i64(fpu_get_st0());


Use .trunc() instead of fpu_convert_to_i64 (also in the other functions).

src/rust/cpu/fpu.rs

copy · 2023-09-11T21:23:40Z

src/rust/cpu/instructions_0f.rs

+            destination.f32[0] + destination.f32[1],
+            destination.f32[2] + destination.f32[3],
+            source.f32[0] + source.f32[1],
+            source.f32[2] + source.f32[3],


This implementation isn't correct. See https://www.felixcloutier.com/x86/movshdup

copy · 2023-09-11T21:24:47Z

src/rust/jit_instructions.rs

@@ -5684,7 +5693,26 @@ pub fn instr_F30F10_reg_jit(ctx: &mut JitContext, r1: u32, r2: u32) {
    ctx.builder
        .store_aligned_i32(global_pointers::get_reg_xmm_offset(r2));
 }
-
+pub fn instr_F30F12_mem_jit(ctx: &mut JitContext, modrm_byte: ModrmByte, r: u32) {
+    instr_660F6E_mem_jit(ctx, modrm_byte, r)


This implementation isn't correct. See https://www.felixcloutier.com/x86/movsldup

src/rust/jit_instructions.rs

copy · 2023-09-11T21:26:32Z

src/rust/cpu/instructions_0f.rs

+            destination.f32[0] + destination.f32[1],
+            destination.f32[2] + destination.f32[3],
+            source.f32[0] + source.f32[1],
+            source.f32[2] + source.f32[3],


https://www.felixcloutier.com/x86/movddup

copy · 2023-09-11T21:27:03Z

src/rust/cpu/instructions_0f.rs

+            destination.f32[0] + destination.f32[1],
+            destination.f32[2] + destination.f32[3],
+            source.f32[0] + source.f32[1],
+            source.f32[2] + source.f32[3],


https://www.felixcloutier.com/x86/hsubpd

copy · 2023-09-11T21:43:32Z

Agreed, in CPU-Z detects SSE3, but in Linux's /proc/cpuinfo pni (Prescott New Instructions, second name of SSE3) is not detected

pni is set when using this PR (I wasn't aware of the alternative name, thanks!). So the cpuid part seems to work fine.

Ryan8085309853409 · 2023-09-12T06:14:16Z

Ok, will work on this when I'm home

spetterman66 · 2023-09-12T07:56:07Z

(or maybe at school lmfao)

spetterman66 · 2023-09-12T19:08:37Z

we got it down to 9 errors :)

BelleNottelling · 2023-09-13T20:56:13Z

RIP to anyone who follows this repo and didn't unfollow this PR when it was originally created. (That's 161 emails from this PR)

if this aint working then idk lmao

..

spetterman66 · 2023-09-16T17:16:10Z

when will this be merged

copy · 2023-09-17T01:27:20Z

Merged, cheers.

cyberplant · 2023-09-17T15:17:38Z

Well done, congratulations @spetterman66 !

Ryan8085309853409 · 2023-09-18T06:20:28Z

You would have never thought a web dev could implement sse3 in an emulator :D

spetterman66 mentioned this pull request Sep 1, 2023

SSE3 support #844

Closed

5 tasks

copy reviewed Sep 2, 2023

View reviewed changes

gen/x86_table.js Outdated Show resolved Hide resolved

copy reviewed Sep 4, 2023

View reviewed changes

gen/x86_table.js Outdated Show resolved Hide resolved

gen/x86_table.js Outdated Show resolved Hide resolved

gen/x86_table.js Outdated Show resolved Hide resolved

gen/x86_table.js Outdated Show resolved Hide resolved

src/rust/cpu/instructions_0f.rs Outdated Show resolved Hide resolved

spetterman66 closed this Sep 8, 2023

spetterman66 reopened this Sep 8, 2023

spetterman66 changed the title ~~SSE3 support?~~ Implement SSE3 support in V86 Sep 10, 2023

copy reviewed Sep 11, 2023

View reviewed changes

spetterman66 added 24 commits September 15, 2023 17:01

Update instructions_0f.rs

4f0a797

Update instructions_0f.rs

2bb13ac

Update instructions_0f.rs

f442d08

Update instructions_0f.rs

b6cf834

Update instructions_0f.rs

d7ef854

Update instructions_0f.rs

cf5629f

Update instructions_0f.rs

d76a05b

Update instructions_0f.rs

df16927

Update instructions_0f.rs

b03640e

Update instructions_0f.rs

f868ecc

Update instructions_0f.rs

ab53f26

Update instructions_0f.rs

c1cb511

Update instructions_0f.rs

a869575

Update instructions_0f.rs

90efc9f

Update instructions_0f.rs

f8bf9b8

Update instructions_0f.rs

1da2d1f

Update instructions_0f.rs

2c3798a

Update instructions_0f.rs

b347961

Update instructions_0f.rs

a45b953

if this aint working then idk lmao

Update instructions_0f.rs

463aa88

..

Update instructions_0f.rs

5d2b9f5

Update instructions_0f.rs

860ef95

Update instructions_0f.rs

5653e48

Update instructions_0f.rs

1ec9248

fix

1a38137

copy merged commit c634ae2 into copy:master Sep 17, 2023
3 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Implement SSE3 support in V86 #906

Implement SSE3 support in V86 #906

spetterman66 commented Sep 1, 2023

spetterman66 commented Sep 4, 2023

spetterman66 commented Sep 6, 2023

spetterman66 commented Sep 6, 2023

copy commented Sep 7, 2023

copy commented Sep 7, 2023

Ryan8085309853409 commented Sep 7, 2023

spetterman66 commented Sep 7, 2023

spetterman66 commented Sep 7, 2023

spetterman66 commented Sep 7, 2023 •

edited

Loading

SuperMaxusa commented Sep 7, 2023

spetterman66 commented Sep 7, 2023

spetterman66 commented Sep 7, 2023

spetterman66 commented Sep 8, 2023

copy commented Sep 9, 2023

spetterman66 commented Sep 9, 2023

spetterman66 commented Sep 10, 2023

spetterman66 commented Sep 10, 2023

spetterman66 commented Sep 11, 2023

copy left a comment

copy Sep 11, 2023

copy Sep 11, 2023

copy Sep 11, 2023

copy Sep 11, 2023

copy Sep 11, 2023

copy commented Sep 11, 2023

Ryan8085309853409 commented Sep 12, 2023

spetterman66 commented Sep 12, 2023

spetterman66 commented Sep 12, 2023

BelleNottelling commented Sep 13, 2023

spetterman66 commented Sep 16, 2023

copy commented Sep 17, 2023

cyberplant commented Sep 17, 2023

Ryan8085309853409 commented Sep 18, 2023

Implement SSE3 support in V86 #906

Implement SSE3 support in V86 #906

Conversation

spetterman66 commented Sep 1, 2023

spetterman66 commented Sep 4, 2023

spetterman66 commented Sep 6, 2023

spetterman66 commented Sep 6, 2023

copy commented Sep 7, 2023

copy commented Sep 7, 2023

Ryan8085309853409 commented Sep 7, 2023

spetterman66 commented Sep 7, 2023

spetterman66 commented Sep 7, 2023

spetterman66 commented Sep 7, 2023 • edited Loading

SuperMaxusa commented Sep 7, 2023

spetterman66 commented Sep 7, 2023

spetterman66 commented Sep 7, 2023

spetterman66 commented Sep 8, 2023

copy commented Sep 9, 2023

spetterman66 commented Sep 9, 2023

spetterman66 commented Sep 10, 2023

spetterman66 commented Sep 10, 2023

spetterman66 commented Sep 11, 2023

copy left a comment

Choose a reason for hiding this comment

copy Sep 11, 2023

Choose a reason for hiding this comment

copy Sep 11, 2023

Choose a reason for hiding this comment

copy Sep 11, 2023

Choose a reason for hiding this comment

copy Sep 11, 2023

Choose a reason for hiding this comment

copy Sep 11, 2023

Choose a reason for hiding this comment

copy commented Sep 11, 2023

Ryan8085309853409 commented Sep 12, 2023

spetterman66 commented Sep 12, 2023

spetterman66 commented Sep 12, 2023

BelleNottelling commented Sep 13, 2023

spetterman66 commented Sep 16, 2023

copy commented Sep 17, 2023

cyberplant commented Sep 17, 2023

Ryan8085309853409 commented Sep 18, 2023

spetterman66 commented Sep 7, 2023 •

edited

Loading