basilisk-devel

[B2-devel] B2/JIT for x86_64

[Gwenole B. <gb...@di...>]

Hi,

I revived and fixed the x86_64 port that started almost two years ago.=20=

Performance may look lower than with the 32-bit binary because it's=20
currently lacking two key features that will yield an interesting speed=20=

boost. For now, it's a way to check the runtime assembler is working=20
properly for a new PowerPC JIT.

Talking about a verifier, I currently have a test code=20
(test_codegen_x86.cpp) that covers only 5.5 million variations and is=20
lacking some others. Someone may be interested in contributing a=20
rewrite that uses a derivative of GNU disassembler's internal=20
structures instead of string compares to check things. I simply=20
implemented that rapidly for other matters. ;-)

My test machine for the port is based on Xeon EM64T but it should work=20=

fine on AMD64 processors too. Relevant code is in CVS.

Bye,
Gwenol=E9.=

Re: [B2-devel] B2/JIT for x86_64

[Bob D. <bob...@te...>]

On Mon, 2004-11-01 at 22:53, Gwenole Beauchesne wrote:
> Hi,

Fedora Core 2 on an Athlon64: compilation warnings but no errors.

Test combination of MacOS 8 CD with NewWorld ROM v. 1.0 p:

"Welcome to Mac OS" -> freezes.

This is currently the only MacOS CD I have to test - I'll look around if
I can find anything else in the following weeks.

Keep up the great work.

Bob Deblier

Re: [B2-devel] B2/JIT for x86_64

[Gwenole B. <gb...@di...>]

Hi,

> Fedora Core 2 on an Athlon64: compilation warnings but no errors.
> Test combination of MacOS 8 CD with NewWorld ROM v. 1.0 p:

Hmm, are you talking about SheepShaver?

> "Welcome to Mac OS" -> freezes.

If it's SheepShaver, freezes were supposed to be fixed in last 
snapshot. Does it freezes without JIT? It would be slow though. Anyway, 
I won't invest much time in the current SheepShaver/JIT since it's to 
be obsoleted now that the run-time assembler is being debugged within 
Basilisk II.

Re: [B2-devel] B2/JIT for x86_64

[Bob D. <bob...@te...>]

On Wed, 2004-11-03 at 09:39, Gwenole Beauchesne wrote:
> Hi,
> 
> > Fedora Core 2 on an Athlon64: compilation warnings but no errors.
> > Test combination of MacOS 8 CD with NewWorld ROM v. 1.0 p:
> 
> Hmm, are you talking about SheepShaver?

SheepShaver - sorry I got confused; it's one mailing list for both
emulators.

BasiliskII compiles without any problems, and runs fine; faster than any
68K Mac I've ever owned...

> > "Welcome to Mac OS" -> freezes.
> 
> If it's SheepShaver, freezes were supposed to be fixed in last 
> snapshot. Does it freezes without JIT? It would be slow though. Anyway, 
> I won't invest much time in the current SheepShaver/JIT since it's to 
> be obsoleted now that the run-time assembler is being debugged within 
> Basilisk II.

Latest CVS build of SheepShaver this morning; no problem - I'll hold out
for that release.

Test results without JIT: crash (SIGSEGV)

Installing drivers...
SIGSEGV
  pc 0x46929c
  ea 0x40461008
 r0 00000000   r1 25ff85f0   r2 00000000   r3 26000400
 r4 00002278   r5 fffffffa   r6 00004e75   r7 ffffffe8
 r8 00000000   r9 00000028  r10 0000006a  r11 2400694e
r12 ffffffff  r13 00000000  r14 200031cc  r15 23ffec00
r16 20078af8  r17 000013ff  r18 20077ec8  r19 40814c1e
r20 25ff8602  r21 26000190  r22 26000400  r23 00000000
r24 00001401  r25 00000020  r26 00000000  r27 ffff876e
r28 00000000  r29 40df1200  r30 40c60000  r31 68fff000
 f0 0.00000   f1 0.00000   f2 0.00000   f3 0.00000
 f4 0.00000   f5 0.00000   f6 0.00000   f7 0.00000
 f8 0.00000   f9 0.00000  f10 0.00000  f11 0.00000
f12 0.00000  f13 0.00000  f14 0.00000  f15 0.00000
f16 0.00000  f17 0.00000  f18 0.00000  f19 0.00000
f20 0.00000  f21 0.00000  f22 0.00000  f23 0.00000
f24 0.00000  f25 0.00000  f26 0.00000  f27 0.00000
f28 0.00000  f29 0.00000  f30 0.00000  f31 0.00000
 lr 40df1200  ctr 00000000   cr 40501c22  xer 00000000
 pc 40df1200 fpscr 00000000

Sincerely,

Bob Deblier

Re: [B2-devel] SheepShaver/JIT for x86_64

[Gwenole B. <gb...@di...>]

Hi,

> SheepShaver - sorry I got confused; it's one mailing list for both
> emulators.

I have now built SheepShaver on x86_64 with gcc 3.4.1. Fixes committed=20=

to CVS, please update your tree. Note, don't care about the following=20
warnings:
../kpx_cpu/src/cpu/jit/basic-dyngen-ops.cpp:43: warning: register used=20=

for two global register variables

They will be obsoleted with the new JIT anyway.

> BasiliskII compiles without any problems, and runs fine; faster than=20=

> any
> 68K Mac I've ever owned...

Even on a K6-2/300, B2/JIT was already faster than my Quadra 630, which=20=

was the goal. Now that I have an x86-64 system, the new reference Mac=20
is my Powerbook G4/400 vs. SheepShaver. This goal is achieved too. ;-)

> Test results without JIT: crash (SIGSEGV)
>
> Installing drivers...
> SIGSEGV
>   pc 0x46929c
>   ea 0x40461008
[...]
>  pc 40df1200 fpscr 00000000

You can't boot MacOS < 8.6 with a NewWorld ROM yet. It was probably=20
possible in the past but I never could get it working yet. Here is a=20
map table of currently working combinations:

MacOS < 8.6	PowerMac PCI 4 MB ROMs
MacOS >=3D 8.6	+ NewWorld ROMs <=3D v1.6

Bye,
Gwenol=E9.=

Re: [B2-devel] SheepShaver/JIT for x86_64

[Bob D. <bob...@te...>]

On Wed, 2004-11-10 at 07:58 +0100, Gwenole Beauchesne wrote:
> You can't boot MacOS < 8.6 with a NewWorld ROM yet. It was probably=20
> possible in the past but I never could get it working yet. Here is a=20
> map table of currently working combinations:
>=20
> MacOS < 8.6	PowerMac PCI 4 MB ROMs
> MacOS >=3D 8.6	+ NewWorld ROMs <=3D v1.6
>=20
> Bye,
> Gwenol=C3=A9.

I've gotten hold of a 9.0.4 CD and it runs like a charm with NewWorld
ROM (v. 1.0p).

Thanks,

Bob