Smalltalk › Pharo › Pharo Smalltalk Developers

About <primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

12 messages Options

Stéphane Ducasse

About <primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>

I recall that clement told me that returning 1,2 or 3 instead of negative, zero, positive was slow.

And I wonder if the primitive got change to the logic clement proposed?

Could we not introduce another primitive and use it from the image?

compare: string1 with: string2 collated: order

"Return 1, 2 or 3, if string1 is <, =, or > string2, with the collating order of characters given by the order array."

| len1 len2 c1 c2 |

<primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>

<var: #string1 declareC: 'unsigned char *string1'>

<var: #string2 declareC: 'unsigned char *string2'>

<var: #order declareC: 'unsigned char *order'>

len1 := string1 size.

len2 := string2 size.

1 to: (len1 min: len2) do:

[:i |

c1 := order at: (string1 basicAt: i) + 1.

c2 := order at: (string2 basicAt: i) + 1.

c1 = c2 ifFalse:

[c1 < c2 ifTrue: [^ 1] ifFalse: [^ 3]]].

len1 = len2 ifTrue: [^ 2].

len1 < len2 ifTrue: [^ 1] ifFalse: [^ 3].

--------------------------------------------

Stéphane Ducasse

http://stephane.ducasse.free.fr

http://www.synectique.eu / http://www.pharo.org

03 59 35 87 52

Assistant: Julie Jonas

FAX 03 59 57 78 50

TEL 03 59 35 86 16

S. Ducasse - Inria

40, avenue Halley,

Parc Scientifique de la Haute Borne, Bât.A, Park Plaza

Villeneuve d'Ascq 59650

France

CyrilFerlicot

Re: About <primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>

On Wed, Apr 10, 2019 at 10:42 AM Stéphane Ducasse
<[hidden email]> wrote:
>
> Hi
>
> I recall that clement told me that returning 1,2 or 3 instead of negative, zero, positive was slow.
> And I wonder if the primitive got change to the logic clement proposed?
>
> Could we not introduce another primitive and use it from the image?

Hi,

I think Sophie already did most of the work to introduce a new
primitive. The missing steps to use the new optimized way to compare
strings are:
- Add the primitive to the primitive table VM side for Pharo/Squeak and Newspeak
- Use the new primitive in the image and call this one for string comparison

With this new primitive performances on string comparison can be
improved around x2.5 to x5 times faster.

>
> compare: string1 with: string2 collated: order
> "Return 1, 2 or 3, if string1 is <, =, or > string2, with the collating order of characters given by the order array."
>
> | len1 len2 c1 c2 |
> <primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>
> <var: #string1 declareC: 'unsigned char *string1'>
> <var: #string2 declareC: 'unsigned char *string2'>
> <var: #order declareC: 'unsigned char *order'>
>
> len1 := string1 size.
> len2 := string2 size.
> 1 to: (len1 min: len2) do:
> [:i |
> c1 := order at: (string1 basicAt: i) + 1.
> c2 := order at: (string2 basicAt: i) + 1.
> c1 = c2 ifFalse:
> [c1 < c2 ifTrue: [^ 1] ifFalse: [^ 3]]].
> len1 = len2 ifTrue: [^ 2].
> len1 < len2 ifTrue: [^ 1] ifFalse: [^ 3].
>
> --------------------------------------------
> Stéphane Ducasse
> http://stephane.ducasse.free.fr
> http://www.synectique.eu / http://www.pharo.org
> 03 59 35 87 52
> Assistant: Julie Jonas
> FAX 03 59 57 78 50
> TEL 03 59 35 86 16
> S. Ducasse - Inria
> 40, avenue Halley,
> Parc Scientifique de la Haute Borne, Bât.A, Park Plaza
> Villeneuve d'Ascq 59650
> France
>

--
Cyril Ferlicot
https://ferlicot.fr

ducasse

Re: About <primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>

On 10 Apr 2019, at 11:42, Cyril Ferlicot <[hidden email]> wrote:

On Wed, Apr 10, 2019 at 10:42 AM Stéphane Ducasse
<[hidden email]> wrote:

Hi

I recall that clement told me that returning 1,2 or 3 instead of negative, zero, positive was slow.
And I wonder if the primitive got change to the logic clement proposed?

Could we not introduce another primitive and use it from the image?

Hi,

I think Sophie already did most of the work to introduce a new
primitive. The missing steps to use the new optimized way to compare
strings are:
- Add the primitive to the primitive table VM side for Pharo/Squeak and Newspeak

If you know that it is done from the VM side let us know.

- Use the new primitive in the image and call this one for string comparison

With this new primitive performances on string comparison can be
improved around x2.5 to x5 times faster.

compare: string1 with: string2 collated: order
"Return 1, 2 or 3, if string1 is <, =, or > string2, with the collating order of characters given by the order array."

| len1 len2 c1 c2 |
<primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>
<var: #string1 declareC: 'unsigned char *string1'>
<var: #string2 declareC: 'unsigned char *string2'>
<var: #order declareC: 'unsigned char *order'>

len1 := string1 size.
len2 := string2 size.
1 to: (len1 min: len2) do:
[:i |
c1 := order at: (string1 basicAt: i) + 1.
c2 := order at: (string2 basicAt: i) + 1.
c1 = c2 ifFalse:
[c1 < c2 ifTrue: [^ 1] ifFalse: [^ 3]]].
len1 = len2 ifTrue: [^ 2].
len1 < len2 ifTrue: [^ 1] ifFalse: [^ 3].

--------------------------------------------
Stéphane Ducasse
http://stephane.ducasse.free.fr
http://www.synectique.eu / http://www.pharo.org
03 59 35 87 52
Assistant: Julie Jonas
FAX 03 59 57 78 50
TEL 03 59 35 86 16
S. Ducasse - Inria
40, avenue Halley,
Parc Scientifique de la Haute Borne, Bât.A, Park Plaza
Villeneuve d'Ascq 59650
France

--
Cyril Ferlicot
https://ferlicot.fr

Nicolas Cellier

Re: About <primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>

VMMaker.oscog-sk.2367
Author: sk
Time: 19 April 2018, 12:02:35.661622 pm
UUID: 0c2401e3-1450-4f73-8e81-958f50171595
Ancestors: VMMaker.oscog- nice.2366

** new primitive to compare strings (slang + JIT)
answers negative smi, 0 or positive smi (instead of 1, 2 or 3 in the MiscPlugin)

* Slang (primitiveCompareWith)
order is optionnal.
comparison loop performed in rawCompare: string1 length: strLength1 with: string2 length: strLength2 accessBlock: accessBlock

* JIT (genPrimitiveStringCompareWith)
the JIT primitive does not take order as parameter (assumed asciiOrder)
quick jump if one of the strings is empty

Le mer. 10 avr. 2019 à 11:56, ducasse <[hidden email]> a écrit :

On 10 Apr 2019, at 11:42, Cyril Ferlicot <[hidden email]> wrote:

On Wed, Apr 10, 2019 at 10:42 AM Stéphane Ducasse
<[hidden email]> wrote:

Hi

I recall that clement told me that returning 1,2 or 3 instead of negative, zero, positive was slow.
And I wonder if the primitive got change to the logic clement proposed?

Could we not introduce another primitive and use it from the image?

Hi,

I think Sophie already did most of the work to introduce a new
primitive. The missing steps to use the new optimized way to compare
strings are:
- Add the primitive to the primitive table VM side for Pharo/Squeak and Newspeak

If you know that it is done from the VM side let us know.

- Use the new primitive in the image and call this one for string comparison

With this new primitive performances on string comparison can be
improved around x2.5 to x5 times faster.

compare: string1 with: string2 collated: order
"Return 1, 2 or 3, if string1 is <, =, or > string2, with the collating order of characters given by the order array."

| len1 len2 c1 c2 |
<primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>
<var: #string1 declareC: 'unsigned char *string1'>
<var: #string2 declareC: 'unsigned char *string2'>
<var: #order declareC: 'unsigned char *order'>

len1 := string1 size.
len2 := string2 size.
1 to: (len1 min: len2) do:
[:i |
c1 := order at: (string1 basicAt: i) + 1.
c2 := order at: (string2 basicAt: i) + 1.
c1 = c2 ifFalse:
[c1 < c2 ifTrue: [^ 1] ifFalse: [^ 3]]].
len1 = len2 ifTrue: [^ 2].
len1 < len2 ifTrue: [^ 1] ifFalse: [^ 3].

--------------------------------------------
Stéphane Ducasse
http://stephane.ducasse.free.fr
http://www.synectique.eu / http://www.pharo.org
03 59 35 87 52
Assistant: Julie Jonas
FAX 03 59 57 78 50
TEL 03 59 35 86 16
S. Ducasse - Inria
40, avenue Halley,
Parc Scientifique de la Haute Borne, Bât.A, Park Plaza
Villeneuve d'Ascq 59650
France

--
Cyril Ferlicot
https://ferlicot.fr

ducasse

Re: About <primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>

Thanks nicolas so we can use them then.

On 10 Apr 2019, at 11:59, Nicolas Cellier <[hidden email]> wrote:

VMMaker.oscog-sk.2367
Author: sk
Time: 19 April 2018, 12:02:35.661622 pm
UUID: 0c2401e3-1450-4f73-8e81-958f50171595
Ancestors: VMMaker.oscog- nice.2366

** new primitive to compare strings (slang + JIT)
answers negative smi, 0 or positive smi (instead of 1, 2 or 3 in the MiscPlugin)

* Slang (primitiveCompareWith)
order is optionnal.
comparison loop performed in rawCompare: string1 length: strLength1 with: string2 length: strLength2 accessBlock: accessBlock

* JIT (genPrimitiveStringCompareWith)
the JIT primitive does not take order as parameter (assumed asciiOrder)
quick jump if one of the strings is empty

Le mer. 10 avr. 2019 à 11:56, ducasse <[hidden email]> a écrit :

On 10 Apr 2019, at 11:42, Cyril Ferlicot <[hidden email]> wrote:

On Wed, Apr 10, 2019 at 10:42 AM Stéphane Ducasse
<[hidden email]> wrote:

Hi

I recall that clement told me that returning 1,2 or 3 instead of negative, zero, positive was slow.
And I wonder if the primitive got change to the logic clement proposed?

Could we not introduce another primitive and use it from the image?

Hi,

I think Sophie already did most of the work to introduce a new
primitive. The missing steps to use the new optimized way to compare
strings are:
- Add the primitive to the primitive table VM side for Pharo/Squeak and Newspeak

If you know that it is done from the VM side let us know.

- Use the new primitive in the image and call this one for string comparison

With this new primitive performances on string comparison can be
improved around x2.5 to x5 times faster.

compare: string1 with: string2 collated: order
"Return 1, 2 or 3, if string1 is <, =, or > string2, with the collating order of characters given by the order array."

| len1 len2 c1 c2 |
<primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>
<var: #string1 declareC: 'unsigned char *string1'>
<var: #string2 declareC: 'unsigned char *string2'>
<var: #order declareC: 'unsigned char *order'>

len1 := string1 size.
len2 := string2 size.
1 to: (len1 min: len2) do:
[:i |
c1 := order at: (string1 basicAt: i) + 1.
c2 := order at: (string2 basicAt: i) + 1.
c1 = c2 ifFalse:
[c1 < c2 ifTrue: [^ 1] ifFalse: [^ 3]]].
len1 = len2 ifTrue: [^ 2].
len1 < len2 ifTrue: [^ 1] ifFalse: [^ 3].

--------------------------------------------
Stéphane Ducasse
http://stephane.ducasse.free.fr
http://www.synectique.eu / http://www.pharo.org
03 59 35 87 52
Assistant: Julie Jonas
FAX 03 59 57 78 50
TEL 03 59 35 86 16
S. Ducasse - Inria
40, avenue Halley,
Parc Scientifique de la Haute Borne, Bât.A, Park Plaza
Villeneuve d'Ascq 59650
France

--
Cyril Ferlicot
https://ferlicot.fr

ducasse

Re: About <primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>

Nicolas

I tried to find the new primitive and I did not find it here

https://github.com/OpenSmalltalk/opensmalltalk-vm/blob/Cog/src/plugins/MiscPrimitivePlugin/MiscPrimitivePlugin.c

I found the old one.

Did I look in the wrong place?

Stef

On 10 Apr 2019, at 13:46, ducasse <[hidden email]> wrote:

Thanks nicolas so we can use them then.

On 10 Apr 2019, at 11:59, Nicolas Cellier <[hidden email]> wrote:

VMMaker.oscog-sk.2367
Author: sk
Time: 19 April 2018, 12:02:35.661622 pm
UUID: 0c2401e3-1450-4f73-8e81-958f50171595
Ancestors: VMMaker.oscog- nice.2366

** new primitive to compare strings (slang + JIT)
answers negative smi, 0 or positive smi (instead of 1, 2 or 3 in the MiscPlugin)

* Slang (primitiveCompareWith)
order is optionnal.
comparison loop performed in rawCompare: string1 length: strLength1 with: string2 length: strLength2 accessBlock: accessBlock

* JIT (genPrimitiveStringCompareWith)
the JIT primitive does not take order as parameter (assumed asciiOrder)
quick jump if one of the strings is empty

Le mer. 10 avr. 2019 à 11:56, ducasse <[hidden email]> a écrit :

On 10 Apr 2019, at 11:42, Cyril Ferlicot <[hidden email]> wrote:

On Wed, Apr 10, 2019 at 10:42 AM Stéphane Ducasse
<[hidden email]> wrote:

Hi

I recall that clement told me that returning 1,2 or 3 instead of negative, zero, positive was slow.
And I wonder if the primitive got change to the logic clement proposed?

Could we not introduce another primitive and use it from the image?

Hi,

I think Sophie already did most of the work to introduce a new
primitive. The missing steps to use the new optimized way to compare
strings are:
- Add the primitive to the primitive table VM side for Pharo/Squeak and Newspeak

If you know that it is done from the VM side let us know.

- Use the new primitive in the image and call this one for string comparison

With this new primitive performances on string comparison can be
improved around x2.5 to x5 times faster.

compare: string1 with: string2 collated: order
"Return 1, 2 or 3, if string1 is <, =, or > string2, with the collating order of characters given by the order array."

| len1 len2 c1 c2 |
<primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>
<var: #string1 declareC: 'unsigned char *string1'>
<var: #string2 declareC: 'unsigned char *string2'>
<var: #order declareC: 'unsigned char *order'>

len1 := string1 size.
len2 := string2 size.
1 to: (len1 min: len2) do:
[:i |
c1 := order at: (string1 basicAt: i) + 1.
c2 := order at: (string2 basicAt: i) + 1.
c1 = c2 ifFalse:
[c1 < c2 ifTrue: [^ 1] ifFalse: [^ 3]]].
len1 = len2 ifTrue: [^ 2].
len1 < len2 ifTrue: [^ 1] ifFalse: [^ 3].

--------------------------------------------
Stéphane Ducasse
http://stephane.ducasse.free.fr
http://www.synectique.eu / http://www.pharo.org
03 59 35 87 52
Assistant: Julie Jonas
FAX 03 59 57 78 50
TEL 03 59 35 86 16
S. Ducasse - Inria
40, avenue Halley,
Parc Scientifique de la Haute Borne, Bât.A, Park Plaza
Villeneuve d'Ascq 59650
France

--
Cyril Ferlicot
https://ferlicot.fr

CyrilFerlicot

Re: About <primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>

On Wed, Apr 10, 2019 at 1:52 PM ducasse <[hidden email]> wrote:

>
> Nicolas
>
> I tried to find the new primitive and I did not find it here
>
> https://github.com/OpenSmalltalk/opensmalltalk-vm/blob/Cog/src/plugins/MiscPrimitivePlugin/MiscPrimitivePlugin.c
>
> I found the old one.
> Did I look in the wrong place?
>

If I am not wrong it is directly in Slang and will be a numbered primitive.

The primitive was added in the commit Nicolas quoted but I'm not sure
the registration of the primitive in the primitive table was done.

> Stef
>
>
>

--
Cyril Ferlicot
https://ferlicot.fr

Nicolas Cellier

Re: About <primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>

In reply to this post by ducasse

Hi Stephane,

If you google 'vm-dev VMMaker.oscog-sk.2367'

and click one of the first hits https://marc.info/?l=squeak-vm-dev&m=152413936110744&w=2

you'll see the changes from Sophie:

The new InterpreterPrimitives>>primitiveCompareWith should be linked to primitive 158.

(and only for ByteString comparison).

I did not try it, I have not much time to invest at the moment.

Le mer. 10 avr. 2019 à 13:52, ducasse <[hidden email]> a écrit :

Nicolas

I tried to find the new primitive and I did not find it here

https://github.com/OpenSmalltalk/opensmalltalk-vm/blob/Cog/src/plugins/MiscPrimitivePlugin/MiscPrimitivePlugin.c

I found the old one.
Did I look in the wrong place?

Stef

On 10 Apr 2019, at 13:46, ducasse <[hidden email]> wrote:

Thanks nicolas so we can use them then.

On 10 Apr 2019, at 11:59, Nicolas Cellier <[hidden email]> wrote:

VMMaker.oscog-sk.2367
Author: sk
Time: 19 April 2018, 12:02:35.661622 pm
UUID: 0c2401e3-1450-4f73-8e81-958f50171595
Ancestors: VMMaker.oscog- nice.2366

** new primitive to compare strings (slang + JIT)
answers negative smi, 0 or positive smi (instead of 1, 2 or 3 in the MiscPlugin)

* Slang (primitiveCompareWith)
order is optionnal.
comparison loop performed in rawCompare: string1 length: strLength1 with: string2 length: strLength2 accessBlock: accessBlock

* JIT (genPrimitiveStringCompareWith)
the JIT primitive does not take order as parameter (assumed asciiOrder)
quick jump if one of the strings is empty

Le mer. 10 avr. 2019 à 11:56, ducasse <[hidden email]> a écrit :

On 10 Apr 2019, at 11:42, Cyril Ferlicot <[hidden email]> wrote:

On Wed, Apr 10, 2019 at 10:42 AM Stéphane Ducasse
<[hidden email]> wrote:

Hi

I recall that clement told me that returning 1,2 or 3 instead of negative, zero, positive was slow.
And I wonder if the primitive got change to the logic clement proposed?

Could we not introduce another primitive and use it from the image?

Hi,

I think Sophie already did most of the work to introduce a new
primitive. The missing steps to use the new optimized way to compare
strings are:
- Add the primitive to the primitive table VM side for Pharo/Squeak and Newspeak

If you know that it is done from the VM side let us know.

- Use the new primitive in the image and call this one for string comparison

With this new primitive performances on string comparison can be
improved around x2.5 to x5 times faster.

compare: string1 with: string2 collated: order
"Return 1, 2 or 3, if string1 is <, =, or > string2, with the collating order of characters given by the order array."

| len1 len2 c1 c2 |
<primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>
<var: #string1 declareC: 'unsigned char *string1'>
<var: #string2 declareC: 'unsigned char *string2'>
<var: #order declareC: 'unsigned char *order'>

len1 := string1 size.
len2 := string2 size.
1 to: (len1 min: len2) do:
[:i |
c1 := order at: (string1 basicAt: i) + 1.
c2 := order at: (string2 basicAt: i) + 1.
c1 = c2 ifFalse:
[c1 < c2 ifTrue: [^ 1] ifFalse: [^ 3]]].
len1 = len2 ifTrue: [^ 2].
len1 < len2 ifTrue: [^ 1] ifFalse: [^ 3].

--------------------------------------------
Stéphane Ducasse
http://stephane.ducasse.free.fr
http://www.synectique.eu / http://www.pharo.org
03 59 35 87 52
Assistant: Julie Jonas
FAX 03 59 57 78 50
TEL 03 59 35 86 16
S. Ducasse - Inria
40, avenue Halley,
Parc Scientifique de la Haute Borne, Bât.A, Park Plaza
Villeneuve d'Ascq 59650
France

--
Cyril Ferlicot
https://ferlicot.fr

ducasse

Re: About <primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>

In reply to this post by CyrilFerlicot

> On 10 Apr 2019, at 14:01, Cyril Ferlicot <[hidden email]> wrote:
>
> On Wed, Apr 10, 2019 at 1:52 PM ducasse <[hidden email]> wrote:
>>
>> Nicolas
>>
>> I tried to find the new primitive and I did not find it here
>>
>> https://github.com/OpenSmalltalk/opensmalltalk-vm/blob/Cog/src/plugins/MiscPrimitivePlugin/MiscPrimitivePlugin.c
>>
>> I found the old one.
>> Did I look in the wrong place?
>>
>
> If I am not wrong it is directly in Slang and will be a numbered primitive.

I do not know :)
Now the old one is there so…. I thought that the new one should be around.

> The primitive was added in the commit Nicolas quoted but I'm not sure
> the registration of the primitive in the primitive table was done.
>
>> Stef
>>
>>
>>
>
>
> --
> Cyril Ferlicot
> https://ferlicot.fr
>

ducasse

Re: About <primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>

In reply to this post by Nicolas Cellier

Thanks I did not know it was on a mailing-list.

On 10 Apr 2019, at 14:02, Nicolas Cellier <[hidden email]> wrote:

Hi Stephane,
If you google 'vm-dev VMMaker.oscog-sk.2367'
and click one of the first hits https://marc.info/?l=squeak-vm-dev&m=152413936110744&w=2

you'll see the changes from Sophie:
The new InterpreterPrimitives>>primitiveCompareWith should be linked to primitive 158.
(and only for ByteString comparison).

I did not try it, I have not much time to invest at the moment.

Le mer. 10 avr. 2019 à 13:52, ducasse <[hidden email]> a écrit :
Nicolas

I tried to find the new primitive and I did not find it here

https://github.com/OpenSmalltalk/opensmalltalk-vm/blob/Cog/src/plugins/MiscPrimitivePlugin/MiscPrimitivePlugin.c

I found the old one.
Did I look in the wrong place?

Stef

On 10 Apr 2019, at 13:46, ducasse <[hidden email]> wrote:

Thanks nicolas so we can use them then.

On 10 Apr 2019, at 11:59, Nicolas Cellier <[hidden email]> wrote:

VMMaker.oscog-sk.2367
Author: sk
Time: 19 April 2018, 12:02:35.661622 pm
UUID: 0c2401e3-1450-4f73-8e81-958f50171595
Ancestors: VMMaker.oscog- nice.2366

** new primitive to compare strings (slang + JIT)
answers negative smi, 0 or positive smi (instead of 1, 2 or 3 in the MiscPlugin)

* Slang (primitiveCompareWith)
order is optionnal.
comparison loop performed in rawCompare: string1 length: strLength1 with: string2 length: strLength2 accessBlock: accessBlock

* JIT (genPrimitiveStringCompareWith)
the JIT primitive does not take order as parameter (assumed asciiOrder)
quick jump if one of the strings is empty

Le mer. 10 avr. 2019 à 11:56, ducasse <[hidden email]> a écrit :

On 10 Apr 2019, at 11:42, Cyril Ferlicot <[hidden email]> wrote:

On Wed, Apr 10, 2019 at 10:42 AM Stéphane Ducasse
<[hidden email]> wrote:

Hi

I recall that clement told me that returning 1,2 or 3 instead of negative, zero, positive was slow.
And I wonder if the primitive got change to the logic clement proposed?

Could we not introduce another primitive and use it from the image?

Hi,

I think Sophie already did most of the work to introduce a new
primitive. The missing steps to use the new optimized way to compare
strings are:
- Add the primitive to the primitive table VM side for Pharo/Squeak and Newspeak

If you know that it is done from the VM side let us know.

- Use the new primitive in the image and call this one for string comparison

With this new primitive performances on string comparison can be
improved around x2.5 to x5 times faster.

compare: string1 with: string2 collated: order
"Return 1, 2 or 3, if string1 is <, =, or > string2, with the collating order of characters given by the order array."

| len1 len2 c1 c2 |
<primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>
<var: #string1 declareC: 'unsigned char *string1'>
<var: #string2 declareC: 'unsigned char *string2'>
<var: #order declareC: 'unsigned char *order'>

len1 := string1 size.
len2 := string2 size.
1 to: (len1 min: len2) do:
[:i |
c1 := order at: (string1 basicAt: i) + 1.
c2 := order at: (string2 basicAt: i) + 1.
c1 = c2 ifFalse:
[c1 < c2 ifTrue: [^ 1] ifFalse: [^ 3]]].
len1 = len2 ifTrue: [^ 2].
len1 < len2 ifTrue: [^ 1] ifFalse: [^ 3].

--------------------------------------------
Stéphane Ducasse
http://stephane.ducasse.free.fr
http://www.synectique.eu / http://www.pharo.org
03 59 35 87 52
Assistant: Julie Jonas
FAX 03 59 57 78 50
TEL 03 59 35 86 16
S. Ducasse - Inria
40, avenue Halley,
Parc Scientifique de la Haute Borne, Bât.A, Park Plaza
Villeneuve d'Ascq 59650
France

--
Cyril Ferlicot
https://ferlicot.fr

SophieKaleba

Re: About <primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>

HI,

We actually talked about this very primitive with Cyril not that long ago. I'll have a look at it with him this week so it can be used !

Sophie

Le mer. 10 avr. 2019 à 14:10, ducasse <[hidden email]> a écrit :

Thanks I did not know it was on a mailing-list.

On 10 Apr 2019, at 14:02, Nicolas Cellier <[hidden email]> wrote:

Hi Stephane,
If you google 'vm-dev VMMaker.oscog-sk.2367'
and click one of the first hits https://marc.info/?l=squeak-vm-dev&m=152413936110744&w=2

you'll see the changes from Sophie:
The new InterpreterPrimitives>>primitiveCompareWith should be linked to primitive 158.
(and only for ByteString comparison).

I did not try it, I have not much time to invest at the moment.

Le mer. 10 avr. 2019 à 13:52, ducasse <[hidden email]> a écrit :
Nicolas

I tried to find the new primitive and I did not find it here

https://github.com/OpenSmalltalk/opensmalltalk-vm/blob/Cog/src/plugins/MiscPrimitivePlugin/MiscPrimitivePlugin.c

I found the old one.
Did I look in the wrong place?

Stef

On 10 Apr 2019, at 13:46, ducasse <[hidden email]> wrote:

Thanks nicolas so we can use them then.

On 10 Apr 2019, at 11:59, Nicolas Cellier <[hidden email]> wrote:

VMMaker.oscog-sk.2367
Author: sk
Time: 19 April 2018, 12:02:35.661622 pm
UUID: 0c2401e3-1450-4f73-8e81-958f50171595
Ancestors: VMMaker.oscog- nice.2366

** new primitive to compare strings (slang + JIT)
answers negative smi, 0 or positive smi (instead of 1, 2 or 3 in the MiscPlugin)

* Slang (primitiveCompareWith)
order is optionnal.
comparison loop performed in rawCompare: string1 length: strLength1 with: string2 length: strLength2 accessBlock: accessBlock

* JIT (genPrimitiveStringCompareWith)
the JIT primitive does not take order as parameter (assumed asciiOrder)
quick jump if one of the strings is empty

Le mer. 10 avr. 2019 à 11:56, ducasse <[hidden email]> a écrit :

On 10 Apr 2019, at 11:42, Cyril Ferlicot <[hidden email]> wrote:

On Wed, Apr 10, 2019 at 10:42 AM Stéphane Ducasse
<[hidden email]> wrote:

Hi

I recall that clement told me that returning 1,2 or 3 instead of negative, zero, positive was slow.
And I wonder if the primitive got change to the logic clement proposed?

Could we not introduce another primitive and use it from the image?

Hi,

I think Sophie already did most of the work to introduce a new
primitive. The missing steps to use the new optimized way to compare
strings are:
- Add the primitive to the primitive table VM side for Pharo/Squeak and Newspeak

If you know that it is done from the VM side let us know.

- Use the new primitive in the image and call this one for string comparison

With this new primitive performances on string comparison can be
improved around x2.5 to x5 times faster.

compare: string1 with: string2 collated: order
"Return 1, 2 or 3, if string1 is <, =, or > string2, with the collating order of characters given by the order array."

| len1 len2 c1 c2 |
<primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>
<var: #string1 declareC: 'unsigned char *string1'>
<var: #string2 declareC: 'unsigned char *string2'>
<var: #order declareC: 'unsigned char *order'>

len1 := string1 size.
len2 := string2 size.
1 to: (len1 min: len2) do:
[:i |
c1 := order at: (string1 basicAt: i) + 1.
c2 := order at: (string2 basicAt: i) + 1.
c1 = c2 ifFalse:
[c1 < c2 ifTrue: [^ 1] ifFalse: [^ 3]]].
len1 = len2 ifTrue: [^ 2].
len1 < len2 ifTrue: [^ 1] ifFalse: [^ 3].

--------------------------------------------
Stéphane Ducasse
http://stephane.ducasse.free.fr
http://www.synectique.eu / http://www.pharo.org
03 59 35 87 52
Assistant: Julie Jonas
FAX 03 59 57 78 50
TEL 03 59 35 86 16
S. Ducasse - Inria
40, avenue Halley,
Parc Scientifique de la Haute Borne, Bât.A, Park Plaza
Villeneuve d'Ascq 59650
France

--
Cyril Ferlicot
https://ferlicot.fr

ducasse

Re: About <primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>

tx!

On 11 Apr 2019, at 14:42, Sophie Kaleba <[hidden email]> wrote:

HI,
We actually talked about this very primitive with Cyril not that long ago. I'll have a look at it with him this week so it can be used !

Sophie

Le mer. 10 avr. 2019 à 14:10, ducasse <[hidden email]> a écrit :
Thanks I did not know it was on a mailing-list.

On 10 Apr 2019, at 14:02, Nicolas Cellier <[hidden email]> wrote:

Hi Stephane,
If you google 'vm-dev VMMaker.oscog-sk.2367'
and click one of the first hits https://marc.info/?l=squeak-vm-dev&m=152413936110744&w=2

you'll see the changes from Sophie:
The new InterpreterPrimitives>>primitiveCompareWith should be linked to primitive 158.
(and only for ByteString comparison).

I did not try it, I have not much time to invest at the moment.

Le mer. 10 avr. 2019 à 13:52, ducasse <[hidden email]> a écrit :
Nicolas

I tried to find the new primitive and I did not find it here

https://github.com/OpenSmalltalk/opensmalltalk-vm/blob/Cog/src/plugins/MiscPrimitivePlugin/MiscPrimitivePlugin.c

I found the old one.
Did I look in the wrong place?

Stef

On 10 Apr 2019, at 13:46, ducasse <[hidden email]> wrote:

Thanks nicolas so we can use them then.

On 10 Apr 2019, at 11:59, Nicolas Cellier <[hidden email]> wrote:

VMMaker.oscog-sk.2367
Author: sk
Time: 19 April 2018, 12:02:35.661622 pm
UUID: 0c2401e3-1450-4f73-8e81-958f50171595
Ancestors: VMMaker.oscog- nice.2366

** new primitive to compare strings (slang + JIT)
answers negative smi, 0 or positive smi (instead of 1, 2 or 3 in the MiscPlugin)

* Slang (primitiveCompareWith)
order is optionnal.
comparison loop performed in rawCompare: string1 length: strLength1 with: string2 length: strLength2 accessBlock: accessBlock

* JIT (genPrimitiveStringCompareWith)
the JIT primitive does not take order as parameter (assumed asciiOrder)
quick jump if one of the strings is empty

Le mer. 10 avr. 2019 à 11:56, ducasse <[hidden email]> a écrit :

On 10 Apr 2019, at 11:42, Cyril Ferlicot <[hidden email]> wrote:

On Wed, Apr 10, 2019 at 10:42 AM Stéphane Ducasse
<[hidden email]> wrote:

Hi

I recall that clement told me that returning 1,2 or 3 instead of negative, zero, positive was slow.
And I wonder if the primitive got change to the logic clement proposed?

Could we not introduce another primitive and use it from the image?

Hi,

I think Sophie already did most of the work to introduce a new
primitive. The missing steps to use the new optimized way to compare
strings are:
- Add the primitive to the primitive table VM side for Pharo/Squeak and Newspeak

If you know that it is done from the VM side let us know.

- Use the new primitive in the image and call this one for string comparison

With this new primitive performances on string comparison can be
improved around x2.5 to x5 times faster.

compare: string1 with: string2 collated: order
"Return 1, 2 or 3, if string1 is <, =, or > string2, with the collating order of characters given by the order array."

| len1 len2 c1 c2 |
<primitive: 'primitiveCompareString' module: 'MiscPrimitivePlugin'>
<var: #string1 declareC: 'unsigned char *string1'>
<var: #string2 declareC: 'unsigned char *string2'>
<var: #order declareC: 'unsigned char *order'>

len1 := string1 size.
len2 := string2 size.
1 to: (len1 min: len2) do:
[:i |
c1 := order at: (string1 basicAt: i) + 1.
c2 := order at: (string2 basicAt: i) + 1.
c1 = c2 ifFalse:
[c1 < c2 ifTrue: [^ 1] ifFalse: [^ 3]]].
len1 = len2 ifTrue: [^ 2].
len1 < len2 ifTrue: [^ 1] ifFalse: [^ 3].

--------------------------------------------
Stéphane Ducasse
http://stephane.ducasse.free.fr
http://www.synectique.eu / http://www.pharo.org
03 59 35 87 52
Assistant: Julie Jonas
FAX 03 59 57 78 50
TEL 03 59 35 86 16
S. Ducasse - Inria
40, avenue Halley,
Parc Scientifique de la Haute Borne, Bât.A, Park Plaza
Villeneuve d'Ascq 59650
France

--
Cyril Ferlicot
https://ferlicot.fr