[Vm-dev] Is there a way to get the simulated/possible machine
code of a CompiledMethod ?
Mariano Martinez Peck
marianopeck at gmail.com
Mon Apr 18 07:27:24 UTC 2011
On Mon, Apr 18, 2011 at 4:13 AM, Eliot Miranda <eliot.miranda at gmail.com>wrote:
>
>
>
> On Sun, Apr 17, 2011 at 5:03 PM, Mariano Martinez Peck <
> marianopeck at gmail.com> wrote:
>
>>
>>
>>
>> On Sun, Apr 17, 2011 at 3:24 AM, Eliot Miranda <eliot.miranda at gmail.com>wrote:
>>
>>>
>>>
>>>
>>> On Sat, Apr 16, 2011 at 4:50 PM, Mariano Martinez Peck <
>>> marianopeck at gmail.com> wrote:
>>>
>>>>
>>>> Hi Eliot. For learning purposes, I would LOVE to be able to do something
>>>> like:
>>>>
>>>>
>>>> (StackInterpreter >> #lookupMethodInClass:) simulatedMachineCode
>>>>
>>>
>>> Its easy to add a primitive to get the code. What's problematic is
>>> decorating that code. Decorating it involves knowing the names for the
>>> variables whose addresses appear in the machine code, and that's tedious to
>>> write. The best way I know of getting the code right now is to use the
>>> VMMaker and Bochs plugin and the interface that allows one to supply a
>>> method in the current image to the simulator and have it translate it as if
>>> it were in the image being simulated (if that makes sense). The
>>> VMMaker/Bochs plugin pair decorates the code beautifully. e.g.
>>>
>>> Cogit chooseCogitClass genAndDis: (Point >> #setX:setY:)
>>>
>>>
>> Hi Eliot. I don't pretend the decorating. A minimal, approximated
>> representation is more than enough.
>> I tried but first error was because #primitiveErrorTable is not
>> implemented in SmalltalkImage and:
>>
>>
>> CurrentImageCoInterpreterFacade >> initializeObjectMap
>> objectMap := IdentityDictionary new.
>> { nil. false. true. Smalltalk primitiveErrorTable. Float } do:
>> [:o| self oopForObject: o]
>>
>>
>> So...I implemented. But then there where several erros because "memory"
>> was a ByteArray and not a Bitmap, and several messages were implemented in
>> BitMap, like #unsignedLongAt: #unsignedByteAt:, etc.
>>
>> I tried with latest code commited.
>>
>
> OK. Give me a few days. I need to track down the relevant extensions I
> have in my base image. Sorry.
>
>
Yes, I imagined that. No problem. When you publish, I test, just let me
know. The browser is almost ready ;)
Thanks in advance,
Mariano
> best,
> Eliot
>
>
>>
>> Thanks in advance
>>
>> Mariano
>>
>>
>> generates the following if you choose SimpleStackBasedCogit:
>>>
>>> E68
>>> objhdr: 1003
>>> nArgs: 2 type: 2
>>> blksiz: C0
>>> method: 100040
>>> mthhdr: 4100801
>>> selctr: 100048=setX:setY:
>>> blkentry: 0
>>> stackCheckOffset: 53/EBB
>>> 00000e80: xorl %edx, %edx : 31 D2
>>> 00000e82: call .+0xfffffb61 (0x000009e8=ceMethodAbort) : E8 61 FB FF FF
>>> 00000e87: nop : 90
>>> entry:
>>> 00000e88: movl %edx, %eax : 89 D0
>>> 00000e8a: andl $0x00000001, %eax : 83 E0 01
>>> 00000e8d: jnz .+0x00000010 (0x00000e9f=setX:setY:@37) : 75 10
>>> 00000e8f: movl %ds:(%edx), %eax : 8B 02
>>> 00000e91: shrl $0x0a, %eax : C1 E8 0A
>>> 00000e94: andl $0x0000007c, %eax : 83 E0 7C
>>> 00000e97: jnz .+0x00000006 (0x00000e9f=setX:setY:@37) : 75 06
>>> 00000e99: movl %ds:0xfffffffc(%edx), %eax : 8B 42 FC
>>> 00000e9c: andl $0xfffffffc, %eax : 83 E0 FC
>>> 00000e9f: cmpl %ecx, %eax : 39 C8
>>> 00000ea1: jnz .+0xffffffdf (0x00000e82=setX:setY:@1A) : 75 DF
>>> noCheckEntry:
>>> 00000ea3: pushl %ebp : 55
>>> 00000ea4: movl %esp, %ebp : 89 E5
>>> 00000ea6: pushl $0x00000e68=setX:setY:@0 : 68 68 0E 00 00
>>> IsAbsPCReference:
>>> 00000eab: movl $0x00100000=nil, %ebx : BB 00 00 10 00
>>> 00000eb0: pushl %ebx : 53
>>> 00000eb1: pushl %edx : 52
>>> 00000eb2: movl %ds:0x20000=stackLimit, %eax : A1 00 00 02 00
>>> 00000eb7: cmpl %eax, %esp : 39 C4
>>> 00000eb9: jb .+0xffffffc5 (0x00000e80=setX:setY:@18) : 72 C5
>>> HasBytecodePC:
>>> 00000ebb: movl 12(%ebp), %eax : 8B 45 0C
>>> 00000ebe: pushl %eax : 50
>>> 00000ebf: movl -12(%ebp), %edx : 8B 55 F4
>>> 00000ec2: popl %ecx : 59
>>> 00000ec3: movl %ecx, %ds:0x4(%edx) : 89 4A 04
>>> 00000ec6: movl %ecx, %eax : 89 C8
>>> 00000ec8: andl $0x00000001, %eax : 83 E0 01
>>> 00000ecb: jnz .+0x0000001a (0x00000ee7=setX:setY:@7F) : 75 1A
>>> 00000ecd: movl %ds:0x4e5e400, %eax : A1 00 E4 E5 04
>>> 00000ed2: cmpl %eax, %edx : 39 C2
>>> 00000ed4: jnb .+0x00000011 (0x00000ee7=setX:setY:@7F) : 73 11
>>> 00000ed6: cmpl %eax, %ecx : 39 C1
>>> 00000ed8: jb .+0x0000000d (0x00000ee7=setX:setY:@7F) : 72 0D
>>> 00000eda: movb %ds:0x3(%edx), %al : 8A 42 03
>>> 00000edd: andl $0x00000040, %eax : 83 E0 40
>>> 00000ee0: jnz .+0x00000005 (0x00000ee7=setX:setY:@7F) : 75 05
>>> 00000ee2: call .+0xfffffda1 (0x00000c88=ceStoreCheckTrampoline) : E8 A1
>>> FD FF FF
>>> IsRelativeCall:
>>> 00000ee7: movl 8(%ebp), %eax : 8B 45 08
>>> 00000eea: pushl %eax : 50
>>> 00000eeb: movl -12(%ebp), %edx : 8B 55 F4
>>> 00000eee: popl %ecx : 59
>>> 00000eef: movl %ecx, %ds:0x8(%edx) : 89 4A 08
>>> 00000ef2: movl %ecx, %eax : 89 C8
>>> 00000ef4: andl $0x00000001, %eax : 83 E0 01
>>> 00000ef7: jnz .+0x0000001a (0x00000f13=setX:setY:@AB) : 75 1A
>>> 00000ef9: movl %ds:0x4e5e400, %eax : A1 00 E4 E5 04
>>> 00000efe: cmpl %eax, %edx : 39 C2
>>> 00000f00: jnb .+0x00000011 (0x00000f13=setX:setY:@AB) : 73 11
>>> 00000f02: cmpl %eax, %ecx : 39 C1
>>> 00000f04: jb .+0x0000000d (0x00000f13=setX:setY:@AB) : 72 0D
>>> 00000f06: movb %ds:0x3(%edx), %al : 8A 42 03
>>> 00000f09: andl $0x00000040, %eax : 83 E0 40
>>> 00000f0c: jnz .+0x00000005 (0x00000f13=setX:setY:@AB) : 75 05
>>> 00000f0e: call .+0xfffffd75 (0x00000c88=ceStoreCheckTrampoline) : E8 75
>>> FD FF FF
>>> IsRelativeCall:
>>> 00000f13: movl -12(%ebp), %edx : 8B 55 F4
>>> 00000f16: movl %ebp, %esp : 89 EC
>>> 00000f18: popl %ebp : 5D
>>> 00000f19: ret $0x000c : C2 0C 00
>>> 00000f1c: nop : 90
>>> 00000f1d: nop : 90
>>> 00000f1e: nop : 90
>>> 00000f1f: nop : 90
>>> 00000f20: nop : 90
>>> startpc: 12
>>> 16rEAB IsAbsPCReference (16rF27)
>>> 16rEBB HasBytecodePC (16rF26, bc: 12)
>>> 16rEE7 IsRelativeCall (16rF24)
>>> 16rF13 IsRelativeCall (16rF22)
>>>
>>> and the following if you choose StackToRegisterMappingCogit:
>>>
>>>
>>> 1128
>>> objhdr: 1003
>>> nArgs: 2 type: 2
>>> blksiz: B8
>>> method: 100040
>>> mthhdr: 4100801
>>> selctr: 100048=setX:setY:
>>> blkentry: 0
>>> stackCheckOffset: 53/117B
>>> 00001140: xorl %edx, %edx : 31 D2
>>> 00001142: call .+0xfffff9b1 (0x00000af8=ceMethodAbortNArgs) : E8 B1 F9 FF
>>> FF
>>> 00001147: nop : 90
>>> entry:
>>> 00001148: movl %edx, %eax : 89 D0
>>> 0000114a: andl $0x00000001, %eax : 83 E0 01
>>> 0000114d: jnz .+0x00000010 (0x0000115f=setX:setY:@37) : 75 10
>>> 0000114f: movl %ds:(%edx), %eax : 8B 02
>>> 00001151: shrl $0x0a, %eax : C1 E8 0A
>>> 00001154: andl $0x0000007c, %eax : 83 E0 7C
>>> 00001157: jnz .+0x00000006 (0x0000115f=setX:setY:@37) : 75 06
>>> 00001159: movl %ds:0xfffffffc(%edx), %eax : 8B 42 FC
>>> 0000115c: andl $0xfffffffc, %eax : 83 E0 FC
>>> 0000115f: cmpl %ecx, %eax : 39 C8
>>> 00001161: jnz .+0xffffffdf (0x00001142=setX:setY:@1A) : 75 DF
>>> noCheckEntry:
>>> 00001163: pushl %ebp : 55
>>> 00001164: movl %esp, %ebp : 89 E5
>>> 00001166: pushl $0x00001128=setX:setY:@0 : 68 28 11 00 00
>>> IsAbsPCReference:
>>> 0000116b: movl $0x00100000=nil, %ebx : BB 00 00 10 00
>>> 00001170: pushl %ebx : 53
>>> 00001171: pushl %edx : 52
>>> 00001172: movl %ds:0x20000=stackLimit, %eax : A1 00 00 02 00
>>> 00001177: cmpl %eax, %esp : 39 C4
>>> 00001179: jb .+0xffffffc5 (0x00001140=setX:setY:@18) : 72 C5
>>> HasBytecodePC:
>>> 0000117b: movl 12(%ebp), %ecx : 8B 4D 0C
>>> 0000117e: movl -12(%ebp), %edx : 8B 55 F4
>>> 00001181: movl %ecx, %ds:0x4(%edx) : 89 4A 04
>>> 00001184: movl %ecx, %eax : 89 C8
>>> 00001186: andl $0x00000001, %eax : 83 E0 01
>>> 00001189: jnz .+0x0000001a (0x000011a5=setX:setY:@7D) : 75 1A
>>> 0000118b: movl %ds:0x4e5e400, %eax : A1 00 E4 E5 04
>>> 00001190: cmpl %eax, %edx : 39 C2
>>> 00001192: jnb .+0x00000011 (0x000011a5=setX:setY:@7D) : 73 11
>>> 00001194: cmpl %eax, %ecx : 39 C1
>>> 00001196: jb .+0x0000000d (0x000011a5=setX:setY:@7D) : 72 0D
>>> 00001198: movb %ds:0x3(%edx), %al : 8A 42 03
>>> 0000119b: andl $0x00000040, %eax : 83 E0 40
>>> 0000119e: jnz .+0x00000005 (0x000011a5=setX:setY:@7D) : 75 05
>>> 000011a0: call .+0xfffffd53 (0x00000ef8=ceStoreCheckTrampoline) : E8 53
>>> FD FF FF
>>> IsRelativeCall:
>>> 000011a5: movl 8(%ebp), %ecx : 8B 4D 08
>>> 000011a8: movl %ecx, %ds:0x8(%edx) : 89 4A 08
>>> 000011ab: movl %ecx, %eax : 89 C8
>>> 000011ad: andl $0x00000001, %eax : 83 E0 01
>>> 000011b0: jnz .+0x0000001a (0x000011cc=setX:setY:@A4) : 75 1A
>>> 000011b2: movl %ds:0x4e5e400, %eax : A1 00 E4 E5 04
>>> 000011b7: cmpl %eax, %edx : 39 C2
>>> 000011b9: jnb .+0x00000011 (0x000011cc=setX:setY:@A4) : 73 11
>>> 000011bb: cmpl %eax, %ecx : 39 C1
>>> 000011bd: jb .+0x0000000d (0x000011cc=setX:setY:@A4) : 72 0D
>>> 000011bf: movb %ds:0x3(%edx), %al : 8A 42 03
>>> 000011c2: andl $0x00000040, %eax : 83 E0 40
>>> 000011c5: jnz .+0x00000005 (0x000011cc=setX:setY:@A4) : 75 05
>>> 000011c7: call .+0xfffffd2c (0x00000ef8=ceStoreCheckTrampoline) : E8 2C
>>> FD FF FF
>>> IsRelativeCall:
>>> 000011cc: movl -12(%ebp), %edx : 8B 55 F4
>>> 000011cf: movl %ebp, %esp : 89 EC
>>> 000011d1: popl %ebp : 5D
>>> 000011d2: ret $0x000c : C2 0C 00
>>> 000011d5: nop : 90
>>> 000011d6: nop : 90
>>> 000011d7: nop : 90
>>> 000011d8: nop : 90
>>> startpc: 12
>>> 16r116B IsAbsPCReference (16r11DF)
>>> 16r117B HasBytecodePC (16r11DE, bc: 12)
>>> 16r11A5 IsRelativeCall (16r11DC)
>>> 16r11CC IsRelativeCall (16r11DA)
>>>
>>> (spot the difference?)
>>>
>>>
>>>> and that answers a possible machine code for that compiled method ? is
>>>> that possible ? Sorry for my ignorance if this is absurd.
>>>>
>>>
>>> Far from it.
>>>
>>>
>>>>
>>>> if you provide me such method I plan to build a little browser on top of
>>>> that.
>>>>
>>>
>>> Go for it.
>>>
>>>
>>>>
>>>> Thanks
>>>>
>>>> --
>>>> Mariano
>>>> http://marianopeck.wordpress.com
>>>>
>>>>
>>>>
>>>
>>>
>>
>>
>> --
>> Mariano
>> http://marianopeck.wordpress.com
>>
>>
>>
>
>
--
Mariano
http://marianopeck.wordpress.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.squeakfoundation.org/pipermail/vm-dev/attachments/20110418/75d46e89/attachment-0001.htm
More information about the Vm-dev
mailing list