Logic Poet Forum - All Forums

test of spaces

lukmac — Sat, 10 Jul 2010 08:29:14 -0700

This problem happens when posting:
When i prefix a word with many spaces, the spaces are all cut out. For example, there should be several spaces before the first word in the following line but there arent:
first word.

Bug of SystemC or my code?

lukmac — Sat, 10 Jul 2010 08:14:35 -0700

My purpose is to use 'sc_pending_activity_at_current_time()' to detect if there are other processes scheduled at the current simulation time. So I coded like the following:

Code:

...

void thread_kid1()

{

    string nm = " >thread_kid1<";

    cout << nm<<": Pending act at now: "<< sc_pending_activity_at_current_time()<<"\n";             

}

void thread_kid2()

{

    string nm = " >thread_kid2<";

    cout << nm<<": Pending act at now: "<< sc_pending_activity_at_current_time()<<"\n";

    event_mom.notify(100, SC_NS);

}

void thread_mom()

{

    string nm = " >thread_mom<";

    cout << nm<<": Pending act at now: "<< sc_pending_activity_at_current_time()<<"\n";

    event_kid.notify(10,SC_NS);

}

...

...

SC_THREAD(thread_kid1)

    dont_initialize();

    sensitive << event_kid;

SC_THREAD(thread_kid2)

    dont_initialize();

    sensitive << event_kid;

SC_THREAD(thread_mom)

    dont_initialize();

    sensitive << event_mom;

event_mom.notify(100,SC_NS);

1. My main question is: why i always got a '0' from 'sc_pending_activity_at_current_time()'?
The definition of this API is at 'sc_simcontext.cpp', which is very short:

Code:

// Return indication if there are more processes to execute in this delta phase

bool

sc_pending_activity_at_current_time()

{

    sc_simcontext* c_p = sc_get_curr_simcontext();

    return (c_p->m_delta_events.size() != 0) ||

            !c_p->m_runnable->is_empty() ||

        c_p->m_prim_channel_registry->pending_updates();

}

I think the 'c_p->m_runnable->is_empty()' should return a '0' for the first executed kid thread and hence '1' should be returned. But as said, i always got a 0, so what is wrong?

2. My minor question is 'event_mom.notify(100, SC_NS);' does not evoke the 'thread_mom' , strange enough, why?

So my poets, which part of the story does not have a good rhythm, systemc or my code?
Thanks for any help!

SystemC-AMS Beta 1 Is Out

Jeff — Wed, 12 May 2010 06:36:01 -0700

I updated the SystemC Suite to provide the official 1.0 Beta 1 release of the AMS library, replacing the old RC 0.15 Proof of Concept that had been available in the installer. You can find the new installer on the downloads page here : http://www.logicpoet.com/downloads

The older installer with the prior AMS library is still available at :
http://www.logicpoet.com/releases/system...te_1.1.dmg

Enjoy and let me know if you have any issues.

Jeff

approximately timed

Rasoul — Thu, 14 Jan 2010 07:30:05 -0800

I am a little stuck understanding the abstraction levels of TLM2 and i need some advice..

According to the Gajski Diagram. Communications are modeled in approximately timed abstraction level if delays are modeled.

If I want to model just a functional model which describes the RTL system with more implementation details, even the communications among the system modules in my point of view, the communications can be modeled on approximately timed even if i did not model any delay time..

is that what i am thinking about right??

Best regards
Rasoul

What free video tools do mac users use?

Abigail — Sun, 08 Nov 2009 23:33:07 -0800

Hi Everyone..

While thinking about buying a mac mini as replacement for my old PC I was wondering what free tools do mac users use for their video editing/processing/playback needs?

Snow Leopard library problems

rduncan — Tue, 13 Oct 2009 15:00:06 -0700

I've just installed the latest release, and I'm getting the following error when trying to use it under Snow Leopard:

ld: warning: in /Library/SystemC/Current/lib-macosx/libsystemc.a, missing required architecture x86_64 in file

Am I doing something wrong? How can I confirm which release I have installed?

Thanks,

Rob.

SystemC Suite with Snow Leopard

Jeff — Tue, 13 Oct 2009 08:08:04 -0700

For those who are using SystemC Suite 1.0 and have upgraded to Snow Leopard, you may notice an issue compiling with the pre-defined templates. This is due to the fact that the Xcode project templates were deigned to support 10.4 (Tiger) as well, and are set up to compile against that baseline SDK. The standard Snow Leopard developer tools installation doe not include the 10.4 SDK (it is an option though), so compilation of SystemC projects will fail since this SDK is missing.

The easiest fix for this is to download the latest 1.1 version of the SystemC Suite installer which was posted on 10/10/09. Those Xcode templates are now set to compile against the 10.5 (Leopard) SDK which is installed by default with Snow Leopard. For those Tiger users (really?), this new installer may cause you some issues so steer clear.

The new installer also includes the latest version of the GreenSocs libraries including some new ones that were not in the old version. If you haven't checked these out before, spend a little time over at GreenSocs and see what they provide. There are a number of handy time-savers in their utilities.

For those that want to fix the 10.4 SDK problem without the new installation, you can simply set the SDK version in the General tab of the Project settings. Change it to 10.5 or 10.6. Alternatively you can install the SN developer tools with the 10.4 support option enabled which will give you that SDK.

As usual, each of the included library binaries were compiled as a Universal Binary and should work fine on PPC as well as Intel. But I'll be honest, I am not firing up the PPC Power Mac much these days so verification is hit or miss. Let me know if you have any issues though.

Cheers,
Jeff

ignorable extension

Dovak — Wed, 09 Sep 2009 15:35:44 -0700

Hi all,

i try to write a small nonblocking example with ignorable extension and i have the following problem. So i have one initiator and one tareget. from initiator i generate the trans and set the value of extension variable befor sending it to the target by calling nb_transport_fw()..

initiator
----------
generate_trans(){

tlm_generic_payload trans;
my_extension* ext;

ext = new my_extension;

ext->my_attribute = 5; // default value is 0.

trans.set_extension(ext);

init_sock->nb_transport_fw(trans,...,...);
}

target
-------
tlm_generic_payload* m_trans; // member variable
m_trans = new tlm_generic_payload;

nb_transport_fw(trans,...,...){

my_extension* ext;
trans.get_extension(ext);

cout<< "my attribute"<my_attribute; // output 5 -- correct
cout << "ext address is " << ext; //output -- 0x103b741

m_trans = &trans;
event.notify() // triggering target_process_method
}

target_process_method{

my_extension* ext;

m_trans.get_extension(ext);

cout <<"my attribute"<my_attribute; //output 0 --error
cout << "ext address is " << ext; //output -- 0x103b783
}

my question now, why i can check the value of my_attribute correctly in the nb_transport_fw but not in the method process (or another member function). As you see i got difference address value of ext ? why?
what i have done wrong in this code?

thanks alot for each support
Dovak

Renaming a New Group Wont Work

raarce — Fri, 21 Aug 2009 09:01:43 -0700

Hi,
I am using Scansion 1.02 on OSX 10.5.8 to do debugging on some VCD files. When I create a New Group, the "New Group" label is created but when I double click it to rename it won't accept my key strokes. All I can do is erase the "New Group" label but not rename it.
Thanks for your help.
Rafael

Register Callback

Rasoul — Wed, 05 Aug 2009 10:17:52 -0700

Hi everyone,

Actually, i can not understand the principle of register callback that used in TLM. the register function is defined in the target socket and as function's argument is a a call of another function that is defined in the bus ..

but how does it work?

thanks in advance..

best regards!
Rasoul

bus protocol

Rasoul — Sun, 02 Aug 2009 03:15:05 -0700

Hi all,

what is the difference between the bus protocol and the non-bus protocol in general and according to TLM modelling style? can you give me one example for each?

thanks in advance..

best regards
Rasoul

Interconect component

Rasoul — Fri, 31 Jul 2009 16:02:25 -0700

Hi all,

The interconnect model can be considered as a bus, arbiter or somthing like that. It tkaes simply the packets from the start point(initiator) to the end point(target). when i want to follow the TLM rules which defined in the reference manual. It was strongly recommended that the interconnect model is not allowed to generate any transaction form TLM_generic_payload that is instantiated inside the interconnect model.

but, when i want to model a spesific protocol which contains e.g. of three different Massege-types, something like header, data and a lock massege. All this types are inhireted from the TLM_base_protocol to keep the system interoperability. Now i want to model a specific system so that when the interconnect component(in this case i do not mean the bus but a model which can transmit and receive too) receive one header from one initaitor, then it has to store it in a spesific queue and then send TLM_Completed as return back to the caller. In the same time the interconnect model should lock all other initiator through the lock massege until all data packets are received form the the initaitor's owner of the header packet.

According to this i have two questions:

1- Do i break fundamentally the TLM rules when i instance a gp inside the interconnect model to generate spesific Massages.

2- Is TLM modeling style able to define several Masseges which are inhirted from a spesific protocol to fulfill a spesific communication?

thanks in advance..
best regards
Rasoul

memory mapped bus

Dovak — Fri, 31 Jul 2009 09:43:40 -0700

Hi everyone,

I am really confused about the right meaning of the memory mapped bus which is mentioned several times in the TLM user manaul.

waht i ve understood it as the following, all HW/SW modells can be bounded to the bus and then accessed through the adress rang that is deined for each modell.

but i think that is not the same meaning that is mentioned in TLM user manual.

thanks for any advance..

Dovak MCJohn.

events

kdv — Tue, 21 Jul 2009 02:53:48 -0700

First of all, thanks for providing a MacOS waveform viewer. It's been a long wait.

I'm wondering whether Scansion "does the right thing" when showing vcd files with events. Right now Scansion seems to show the value of the event, while IEEE 1364 says: (par. 18.2.2)

Quote:Events are dumped in the same format as scalars. ... For events, however, the value ... is irrelevant. Only the identifier code ... is significant. It appears in the VCD file as a marker to indicate the event was triggered during the time step.

Some waveform viewers show an event as "high" when the event appears in the time step, "low" otherwise. This seems a safe approach.
Alternatively, one could show events as markers. YMMV.

regards,

koen

Scansion 1.0 Released

Jeff — Fri, 17 Jul 2009 11:44:27 -0700

We are very pleased to report that Scansion version 1.0 is now available for download.

This release now requires a license to remove some limitations on larger files, as I mentioned in a previous post that it would. This is done by capping the maximum number of events allowed in the unlicensed version. For files which exceed this, you'll get a notification about the excess and the file will be displayed except that it will be truncated to stay within the event limit. For logic files this is set at 250,000 events (state change on a signal or a vector). For the transaction viewing, the limit is 10,000 events; it is smaller since those files tend to contain significantly fewer events.

That said, you will get the most of Scansion with a purchased license. Besides the added functionality you will be enabling future development of this and other applications in the pipeline. Thanks in advance for your support.

New in Scansion 1.0

Analog display is now available for vectors (not just for real valued data).
ASCII Radix Support
Better support for preserving window layout from session to session.

Bug Fixes in 1.0

Modules read from SCNX transactions files could display incorrect hierarchy.
Reloading an SCNX transaction file could lead to application instability or crash.

Scansion 0.95 Released

Jeff — Thu, 18 Jun 2009 09:04:40 -0700

Greetings,

I am pleased to tell you that, after what must seem like a long time, a new version of Scansion is available. You can download the new version here or you will be prompted to update the next time you start Scansion (or you can just select Check for Updates in the menu if you have it running).

Big changes here are in the support for larger files and iChat Theater support; the latter allows you to share your window with others in a video iChat session. A list of the changes and bug fixes in the new release is at the end of this post.

This is a time-limted beta again and will expire on 9/1/09. We are getting very close to releasing 1.0 (finally) and want to get this version out for some wider testing before we release the next major revision. There are many violations of the VCD standards amongst the many tools that generate these, which makes it difficult for us to be sure that there are no issues out there so your feedback is greatly appreciated when issues do come up.

There will still be a free version available once 1.0 comes out. The plan is to release 1.0 with a free version that supports a reasonable but limited number of events. The unlimited version will require a license and we'll announce the details of that shortly. We'll keep the costs reasonable though.

Below are some details on what is going on with the development. If you don't care, skip to the bottom for the bullet list of changes and enjoy the new version.

As I mentioned before I was pretty unhappy with the way Scansion was holding up with medium to large files, and there have been a lot of steps taken in this new release to remedy this.

Under the hood, the entire memory model has been reworked; we moved away from garbage collection entirely which improves performance and our ability to keep the memory footprint in check. This made it easier/possible to address the numerous memory leaks that were in previous version, especially during reload. Memory leaks while using garbage collection? Yes, it unfortunately pretty easy to do and in my opinion harder to find and fix.

Another very significant change is support for 64 bit addressing for those systems that support it (most new-ish Macs will). That allows us to shoot past the memory limits imposed by 32b systems, and I am having pretty good success and performance using 2+ GB VCD files on a machine with 4GB of RAM. You mileage will vary but its looking pretty good; I'd still recommend sticking to smaller files if you can due to load time etc though.

You'll see other things listed below as well to help the memory usage but all in all you should see a memory footprint of around 50% of the last version even with 64 bit (which increases memory usage since some important data types like pointers get bigger). We are working on some more dynamic file loading to make using enormous files faster and easier but this is still a little ways out. Meanwhile, the new features in 0.95 should make big files usable on most machines.

New in Scansion 0.95

64 bit support for much better handling of very large files (for 64 bit systems).
iChat Theater support to allow interactive sharing sessions via iChat.
Reading of large files can be stopped before the file has completed, with the partial results displayed.
Background and font color customization options.
Specify max expandable vector sizes to reduce memory footprint and improve performance.
Overall performance and memory usage improvements.

Bug Fixes in 0.95

Reloading would cause memory leaks (memory consumption roughly doubled).
Large file sizes would crash the application (fixed for 64 bit systems).
Mentor's Questa generates VCD scope types with non-standard "uknown" for SystemVerilog interfaces, which was not accepted by the VCD parser.
Using the Inspector without traces selected could occasionally cause a crash.

Hope you enjoy it, and please send us any issues you may find.

Cheers,
Jeff

might be some problem about the lib

cyqh1028 — Thu, 21 May 2009 18:25:01 -0700

Hi, jeff
I found there are some path like /Users/jeff/Logic Poet in the binary file libawms.a,liblptracer.a,libscv.a,libsystemc-ams.a,libsystemc.a,
which I think it should be /Library/SystemC/Current.

Compiler complains about sc_spawn

Jeff — Wed, 29 Apr 2009 04:47:01 -0700

Just wanted to post a reminder about something that trips people up fairly frequently. If you start seeing loads of compiler issues complaining about sc_spawn, you need to remember to set the pre-processor macro SC_INCLUDE_DYNAMIC_PROCESSES. A snippet of the error you might see is:

/Library/SystemC/Current/include/tlm_utils/simple_target_socket.h:173: error: invalid use of undefined type 'struct sc_core::sc_spawn_options'
/Library/SystemC/Current/include/sysc/kernel/sc_process.h:238: error: forward declaration of 'struct sc_core::sc_spawn_options'
/Library/SystemC/Current/include/tlm_utils/simple_target_socket.h:174: error: 'sc_spawn' is not a member of 'sc_core'

This pre-processor macro is needed in situations where you are creating new sc_methods or sc_thread on the fly (sc_spawn) or forking the thread (SC_FORK, SC_JOIN).

Its pretty easy to remember to set this when doing it in your own code, but it can sometimes be confusing when you get the error when using someone else's libraries or examples.

For reference, the GreenSocs project template for XCode included in the SystemC Suite installation has this macro set since those libraries rely on it. The other templates don't define this, so set it your self in the project or target build settings under "Preprocessor Macros"

Cheers,
Jeff

where is the examples of greensocs

cyqh1028 — Sat, 28 Feb 2009 05:16:42 -0800

i could not find it

TLM examples

billit — Sun, 01 Feb 2009 22:24:03 -0800

Thanks a lot for putting this stuff together for Mac OS.

In order to run a TLM example, I had to set a few defaults that could be patched in the Makefile.config, since your installer always puts it at /Library/SystemC/current you could consider patching it so it installs with defaults that would usually work.

examples/tlm/build-unix/Makefile.config needs environment vars as follows:

TARGET_ARCH=macosx
SYSTEMC_HOME=/Library/SystemC/current
TLM_HOME=/Library/SystemC/current
BOOST_HOME=/Library/SystemC/current/boost

these get the make in the at_ooo example to work. You put the TLM includes in a different directory than the osci tlm tar does, but it works because they get picked up by $SYSTEMC_INC_DIR in Makefile.config which makes the $TLM_INC_DIR unused, it sets that to $TLM_HOME/include/tlm which doesn't exist in your install (that is where the osci tar puts tlm stuff). No big deal since it does work with environment vars set as above.