Thread: [Ikvm-developers] HashMap Performance Issue
Brought to you by:
jfrijters
Menu
▾
▴
ikvm-developers
|
From: xiang l. <sea...@gm...> - 2011-05-09 07:23:42
|
Hi,
I am looking at a HashMap performance issue. In the Java code, the key
used in the Hashmap has its hashCode overridden. But when I looked
into the reversed C# code of java.util.HashMap implementation. Since
the Key is java.lang.Object, the overridden version is not used, see
below. It actuall calls
Object.instancehelper_hashCode. which calls GetHashCode() in ture if
the object is not a String, equals method has the same issue in
HashMap .NET implementation.
[MethodImpl(MethodImplOptions.NoInlining),
Signature("(Ljava/lang/Object;)TV;"), LineNumberTable(new byte[] {
160, 0xc9, 0x63, 0x69, 0x6c, 0x75, 0xa3, 0x7d, 0xe7, 0x3d, 0xe9, 0x45
})]
public override object get(object key)
{
if (key == null)
{
return this.getForNullKey();
}
int num = hash(Object.instancehelper_hashCode(key));
for (Entry entry = this.table[indexFor(num, this.table.Length)];
entry != null; entry = entry.next)
{
object obj2;
if ((entry.hash == num) && (((obj2 = entry.key) == key) ||
Object.instancehelper_equals(key, obj2)))
{
return entry.value;
}
}
return null;
}
I would want the .NET version use the same hashCode and equals
function, but I have no idea of resolving this issue, Is this a bug or
some issue we already know? Could a map.xml be used to preserve
hashCode method which means no rediection of the call?
Any idea? Please help.
Thanks,
Sean
|
|
From: Jeroen F. <je...@su...> - 2011-05-09 07:41:27
|
Hi,
The overridden version *is* used, because java.lang.Object.hashCode() overrides System.Object.GetHashCode().
Regards,
Jeroen
> -----Original Message-----
> From: xiang liu [mailto:sea...@gm...]
> Sent: Monday, May 09, 2011 9:24 AM
> To: Ikv...@li...
> Subject: [Ikvm-developers] HashMap Performance Issue
>
> Hi,
>
> I am looking at a HashMap performance issue. In the Java code, the key
> used in the Hashmap has its hashCode overridden. But when I looked into
> the reversed C# code of java.util.HashMap implementation. Since the Key
> is java.lang.Object, the overridden version is not used, see below. It
> actuall calls
>
> Object.instancehelper_hashCode. which calls GetHashCode() in ture if the
> object is not a String, equals method has the same issue in HashMap .NET
> implementation.
>
> [MethodImpl(MethodImplOptions.NoInlining),
> Signature("(Ljava/lang/Object;)TV;"), LineNumberTable(new byte[] { 160,
> 0xc9, 0x63, 0x69, 0x6c, 0x75, 0xa3, 0x7d, 0xe7, 0x3d, 0xe9, 0x45 })]
> public override object get(object key) {
> if (key == null)
> {
> return this.getForNullKey();
> }
> int num = hash(Object.instancehelper_hashCode(key));
> for (Entry entry = this.table[indexFor(num, this.table.Length)];
> entry != null; entry = entry.next)
> {
> object obj2;
> if ((entry.hash == num) && (((obj2 = entry.key) == key) ||
> Object.instancehelper_equals(key, obj2)))
> {
> return entry.value;
> }
> }
> return null;
> }
>
>
> I would want the .NET version use the same hashCode and equals function,
> but I have no idea of resolving this issue, Is this a bug or some issue
> we already know? Could a map.xml be used to preserve hashCode method
> which means no rediection of the call?
>
> Any idea? Please help.
>
> Thanks,
>
> Sean
>
> ------------------------------------------------------------------------
> ------
> WhatsUp Gold - Download Free Network Management Software The most
> intuitive, comprehensive, and cost-effective network management toolset
> available today. Delivers lowest initial acquisition cost and overall
> TCO of any competing solution.
> http://p.sf.net/sfu/whatsupgold-sd
> _______________________________________________
> Ikvm-developers mailing list
> Ikv...@li...
> https://lists.sourceforge.net/lists/listinfo/ikvm-developers
|
|
From: Sean L. <sea...@gm...> - 2011-05-09 07:56:27
|
Hi, Jeroen, Thanks for your quick reply. I am not sure I understood what you meant. Here is the code I copied from .NET Reflector. Both of the methods simply call System.Object.GetHashCode(). [HideFromJava <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://IKVM.Runtime:0.47.4142.0/IKVM.Attributes.HideFromJavaAttribute/.ctor%28%29>,EditorBrowsable <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://System:2.0.0.0:b77a5c561934e089/System.ComponentModel.EditorBrowsableAttribute/.ctor%28System.ComponentModel.EditorBrowsableState%29>(EditorBrowsableState <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://System:4.0.0.0:b77a5c561934e089/System.ComponentModel.EditorBrowsableState>.Never <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://System:4.0.0.0:b77a5c561934e089/System.ComponentModel.EditorBrowsableState/Never>)] public int <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://mscorlib:2.0.0.0:b77a5c561934e089/System.Int32> *GetHashCode <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://IKVM.OpenJDK.Core:0.47.4142.0/java.lang.Object/GetHashCode%28%29:Int32>*() { return this.GetHashCode <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://mscorlib:2.0.0.0:b77a5c561934e089/System.Object/GetHashCode%28%29:Int32>(); } public virtual int <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://mscorlib:2.0.0.0:b77a5c561934e089/System.Int32> *hashCode <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://IKVM.OpenJDK.Core:0.47.4142.0/java.lang.Object/hashCode%28%29:Int32>*() { return base.GetHashCode <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://mscorlib:2.0.0.0:b77a5c561934e089/System.Object/GetHashCode%28%29:Int32>(); } Has anyone reported that IKVM HashMap is much slower than the original java implementation. In my case, IKVM version took 5 times as the Java version did. Thanks, Sean On 5/9/2011 3:41 PM, Jeroen Frijters wrote: > Hi, > > The overridden version *is* used, because java.lang.Object.hashCode() overrides System.Object.GetHashCode(). > > Regards, > Jeroen > >> -----Original Message----- >> From: xiang liu [mailto:sea...@gm...] >> Sent: Monday, May 09, 2011 9:24 AM >> To: Ikv...@li... >> Subject: [Ikvm-developers] HashMap Performance Issue >> >> Hi, >> >> I am looking at a HashMap performance issue. In the Java code, the key >> used in the Hashmap has its hashCode overridden. But when I looked into >> the reversed C# code of java.util.HashMap implementation. Since the Key >> is java.lang.Object, the overridden version is not used, see below. It >> actuall calls >> >> Object.instancehelper_hashCode. which calls GetHashCode() in ture if the >> object is not a String, equals method has the same issue in HashMap .NET >> implementation. >> >> [MethodImpl(MethodImplOptions.NoInlining), >> Signature("(Ljava/lang/Object;)TV;"), LineNumberTable(new byte[] { 160, >> 0xc9, 0x63, 0x69, 0x6c, 0x75, 0xa3, 0x7d, 0xe7, 0x3d, 0xe9, 0x45 })] >> public override object get(object key) { >> if (key == null) >> { >> return this.getForNullKey(); >> } >> int num = hash(Object.instancehelper_hashCode(key)); >> for (Entry entry = this.table[indexFor(num, this.table.Length)]; >> entry != null; entry = entry.next) >> { >> object obj2; >> if ((entry.hash == num)&& (((obj2 = entry.key) == key) || >> Object.instancehelper_equals(key, obj2))) >> { >> return entry.value; >> } >> } >> return null; >> } >> >> >> I would want the .NET version use the same hashCode and equals function, >> but I have no idea of resolving this issue, Is this a bug or some issue >> we already know? Could a map.xml be used to preserve hashCode method >> which means no rediection of the call? >> >> Any idea? Please help. >> >> Thanks, >> >> Sean >> >> ------------------------------------------------------------------------ >> ------ >> WhatsUp Gold - Download Free Network Management Software The most >> intuitive, comprehensive, and cost-effective network management toolset >> available today. Delivers lowest initial acquisition cost and overall >> TCO of any competing solution. >> http://p.sf.net/sfu/whatsupgold-sd >> _______________________________________________ >> Ikvm-developers mailing list >> Ikv...@li... >> https://lists.sourceforge.net/lists/listinfo/ikvm-developers |
|
From: Jeroen F. <je...@su...> - 2011-05-09 08:00:40
|
Hi,
Reflector is not a helpful tool to understand IKVM.NET, it does not accurately reflect the IL. Use ILDASM instead.
Here's what ILDASM shows for java.lang.Object.hashCode():
.method public hidebysig newslot strict virtual
instance int32 hashCode() cil managed
{
.override [mscorlib]System.Object::GetHashCode
.maxstack 8
IL_0000: ldarg.0
IL_0001: call instance int32 [mscorlib]System.Object::GetHashCode()
IL_0006: ret
}
The relevant line is ".override [mscorlib]System.Object::GetHashCode".
If you have a performance problem, please try to provide an isolated test case, that gives the highest chance of getting it improved.
Regards,
Jeroen
> -----Original Message-----
> From: Sean Liu [mailto:sea...@gm...]
> Sent: Monday, May 09, 2011 9:56 AM
> To: Jeroen Frijters
> Cc: Ikv...@li...
> Subject: Re: [Ikvm-developers] HashMap Performance Issue
>
> Hi, Jeroen,
>
> Thanks for your quick reply.
>
> I am not sure I understood what you meant. Here is the code I copied
> from .NET Reflector. Both of the methods simply call
> System.Object.GetHashCode().
>
>
> [HideFromJava
> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://IKVM.Runt
> ime:0.47.4142.0/IKVM.Attributes.HideFromJavaAttribute/.ctor%28%29> ,
> EditorBrowsable
> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://System:2.
> 0.0.0:b77a5c561934e089/System.ComponentModel.EditorBrowsableAttribute/.c
> tor%28System.ComponentModel.EditorBrowsableState%29>
> (EditorBrowsableState
> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://System:4.
> 0.0.0:b77a5c561934e089/System.ComponentModel.EditorBrowsableState>
> .Never
> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://System:4.
> 0.0.0:b77a5c561934e089/System.Com
> ponentModel.EditorBrowsableState/Never> )] public int
> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://mscorlib:
> 2.0.0.0:b77a5c561934e089/System.Int32> GetHashCode
> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://IKVM.Open
> JDK.Core:0.47.4142.0/java.lang.Object/GetHashCode%28%29:Int32> () {
> return this.GetHashCode
> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://mscorlib:
> 2.0.0.0:b77a5c561934e089/System.Object/GetHashCode%28%29:Int32> (); }
>
> public virtual int
> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://mscorlib:
> 2.0.0.0:b77a5c561934e089/System.Int32> hashCode
> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://IKVM.Open
> JDK.Core:0.47.4142.0/java.lang.Object/hashCode%28%29:Int32> () {
> return base.GetHashCode
> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://mscorlib:
> 2.0.0.0:b77a5c561934e089/System.Object/GetHashCode%28%29:Int32> (); }
>
> Has anyone reported that IKVM HashMap is much slower than the original
> java implementation. In my case, IKVM version took 5 times as the Java
> version did.
>
> Thanks,
>
> Sean
>
> On 5/9/2011 3:41 PM, Jeroen Frijters wrote:
>
> Hi,
>
> The overridden version *is* used, because
> java.lang.Object.hashCode() overrides System.Object.GetHashCode().
>
> Regards,
> Jeroen
>
>
> -----Original Message-----
> From: xiang liu [mailto:sea...@gm...]
> Sent: Monday, May 09, 2011 9:24 AM
> To: Ikv...@li...
> Subject: [Ikvm-developers] HashMap Performance Issue
>
> Hi,
>
> I am looking at a HashMap performance issue. In the Java
> code, the key
> used in the Hashmap has its hashCode overridden. But when I
> looked into
> the reversed C# code of java.util.HashMap implementation.
> Since the Key
> is java.lang.Object, the overridden version is not used, see
> below. It
> actuall calls
>
> Object.instancehelper_hashCode. which calls GetHashCode() in
> ture if the
> object is not a String, equals method has the same issue in
> HashMap .NET
> implementation.
>
> [MethodImpl(MethodImplOptions.NoInlining),
> Signature("(Ljava/lang/Object;)TV;"), LineNumberTable(new
> byte[] { 160,
> 0xc9, 0x63, 0x69, 0x6c, 0x75, 0xa3, 0x7d, 0xe7, 0x3d, 0xe9,
> 0x45 })]
> public override object get(object key) {
> if (key == null)
> {
> return this.getForNullKey();
> }
> int num = hash(Object.instancehelper_hashCode(key));
> for (Entry entry = this.table[indexFor(num,
> this.table.Length)];
> entry != null; entry = entry.next)
> {
> object obj2;
> if ((entry.hash == num) && (((obj2 = entry.key) ==
> key) ||
> Object.instancehelper_equals(key, obj2)))
> {
> return entry.value;
> }
> }
> return null;
> }
>
>
> I would want the .NET version use the same hashCode and
> equals function,
> but I have no idea of resolving this issue, Is this a bug or
> some issue
> we already know? Could a map.xml be used to preserve hashCode
> method
> which means no rediection of the call?
>
> Any idea? Please help.
>
> Thanks,
>
> Sean
>
> -------------------------------------------------------------
> -----------
> ------
> WhatsUp Gold - Download Free Network Management Software The
> most
> intuitive, comprehensive, and cost-effective network
> management toolset
> available today. Delivers lowest initial acquisition cost
> and overall
> TCO of any competing solution.
> http://p.sf.net/sfu/whatsupgold-sd
> _______________________________________________
> Ikvm-developers mailing list
> Ikv...@li...
> https://lists.sourceforge.net/lists/listinfo/ikvm-developers
>
>
>
|
|
From: Sean L. <sea...@gm...> - 2011-05-09 08:33:48
|
Hi, Jeroen,
You were right, Refector did not show it correctly. There must be
something else which does not work correctly to cause the performance
issue, my use case is too complicated, I will try to provide a similar
but simple test case.
Thanks a lot.
Sean
On 5/9/2011 4:00 PM, Jeroen Frijters wrote:
> Hi,
>
> Reflector is not a helpful tool to understand IKVM.NET, it does not accurately reflect the IL. Use ILDASM instead.
>
> Here's what ILDASM shows for java.lang.Object.hashCode():
>
> .method public hidebysig newslot strict virtual
> instance int32 hashCode() cil managed
> {
> .override [mscorlib]System.Object::GetHashCode
> .maxstack 8
> IL_0000: ldarg.0
> IL_0001: call instance int32 [mscorlib]System.Object::GetHashCode()
> IL_0006: ret
> }
>
> The relevant line is ".override [mscorlib]System.Object::GetHashCode".
>
> If you have a performance problem, please try to provide an isolated test case, that gives the highest chance of getting it improved.
>
> Regards,
> Jeroen
>
>> -----Original Message-----
>> From: Sean Liu [mailto:sea...@gm...]
>> Sent: Monday, May 09, 2011 9:56 AM
>> To: Jeroen Frijters
>> Cc: Ikv...@li...
>> Subject: Re: [Ikvm-developers] HashMap Performance Issue
>>
>> Hi, Jeroen,
>>
>> Thanks for your quick reply.
>>
>> I am not sure I understood what you meant. Here is the code I copied
>> from .NET Reflector. Both of the methods simply call
>> System.Object.GetHashCode().
>>
>>
>> [HideFromJava
>> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://IKVM.Runt
>> ime:0.47.4142.0/IKVM.Attributes.HideFromJavaAttribute/.ctor%28%29> ,
>> EditorBrowsable
>> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://System:2.
>> 0.0.0:b77a5c561934e089/System.ComponentModel.EditorBrowsableAttribute/.c
>> tor%28System.ComponentModel.EditorBrowsableState%29>
>> (EditorBrowsableState
>> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://System:4.
>> 0.0.0:b77a5c561934e089/System.ComponentModel.EditorBrowsableState>
>> .Never
>> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://System:4.
>> 0.0.0:b77a5c561934e089/System.Com
>> ponentModel.EditorBrowsableState/Never> )] public int
>> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://mscorlib:
>> 2.0.0.0:b77a5c561934e089/System.Int32> GetHashCode
>> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://IKVM.Open
>> JDK.Core:0.47.4142.0/java.lang.Object/GetHashCode%28%29:Int32> () {
>> return this.GetHashCode
>> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://mscorlib:
>> 2.0.0.0:b77a5c561934e089/System.Object/GetHashCode%28%29:Int32> (); }
>>
>> public virtual int
>> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://mscorlib:
>> 2.0.0.0:b77a5c561934e089/System.Int32> hashCode
>> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://IKVM.Open
>> JDK.Core:0.47.4142.0/java.lang.Object/hashCode%28%29:Int32> () {
>> return base.GetHashCode
>> <http://www.aisto.com/roeder/dotnet/Default.aspx?Target=code://mscorlib:
>> 2.0.0.0:b77a5c561934e089/System.Object/GetHashCode%28%29:Int32> (); }
>>
>> Has anyone reported that IKVM HashMap is much slower than the original
>> java implementation. In my case, IKVM version took 5 times as the Java
>> version did.
>>
>> Thanks,
>>
>> Sean
>>
>> On 5/9/2011 3:41 PM, Jeroen Frijters wrote:
>>
>> Hi,
>>
>> The overridden version *is* used, because
>> java.lang.Object.hashCode() overrides System.Object.GetHashCode().
>>
>> Regards,
>> Jeroen
>>
>>
>> -----Original Message-----
>> From: xiang liu [mailto:sea...@gm...]
>> Sent: Monday, May 09, 2011 9:24 AM
>> To: Ikv...@li...
>> Subject: [Ikvm-developers] HashMap Performance Issue
>>
>> Hi,
>>
>> I am looking at a HashMap performance issue. In the Java
>> code, the key
>> used in the Hashmap has its hashCode overridden. But when I
>> looked into
>> the reversed C# code of java.util.HashMap implementation.
>> Since the Key
>> is java.lang.Object, the overridden version is not used, see
>> below. It
>> actuall calls
>>
>> Object.instancehelper_hashCode. which calls GetHashCode() in
>> ture if the
>> object is not a String, equals method has the same issue in
>> HashMap .NET
>> implementation.
>>
>> [MethodImpl(MethodImplOptions.NoInlining),
>> Signature("(Ljava/lang/Object;)TV;"), LineNumberTable(new
>> byte[] { 160,
>> 0xc9, 0x63, 0x69, 0x6c, 0x75, 0xa3, 0x7d, 0xe7, 0x3d, 0xe9,
>> 0x45 })]
>> public override object get(object key) {
>> if (key == null)
>> {
>> return this.getForNullKey();
>> }
>> int num = hash(Object.instancehelper_hashCode(key));
>> for (Entry entry = this.table[indexFor(num,
>> this.table.Length)];
>> entry != null; entry = entry.next)
>> {
>> object obj2;
>> if ((entry.hash == num)&& (((obj2 = entry.key) ==
>> key) ||
>> Object.instancehelper_equals(key, obj2)))
>> {
>> return entry.value;
>> }
>> }
>> return null;
>> }
>>
>>
>> I would want the .NET version use the same hashCode and
>> equals function,
>> but I have no idea of resolving this issue, Is this a bug or
>> some issue
>> we already know? Could a map.xml be used to preserve hashCode
>> method
>> which means no rediection of the call?
>>
>> Any idea? Please help.
>>
>> Thanks,
>>
>> Sean
>>
>> -------------------------------------------------------------
>> -----------
>> ------
>> WhatsUp Gold - Download Free Network Management Software The
>> most
>> intuitive, comprehensive, and cost-effective network
>> management toolset
>> available today. Delivers lowest initial acquisition cost
>> and overall
>> TCO of any competing solution.
>> http://p.sf.net/sfu/whatsupgold-sd
>> _______________________________________________
>> Ikvm-developers mailing list
>> Ikv...@li...
>> https://lists.sourceforge.net/lists/listinfo/ikvm-developers
>>
>>
>>
>
|
×
Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.
Thanks for helping keep SourceForge clean.
X