[Starpu-devel] Starpu & Trace

["SYLVAND, Guillaume" <guillaume.sylvand@airbus.com>]

Hi Girls & Boys,

I am currently using starpu to generate traces for H-matrix on my Mac, and I have some troubles...

First, it seems that the codelet name, when it appears in the paje trace, is limited to 32 characters on linux (and 24 on mac, dont know why). To overcome this limitation, I have increased FXT_MAX_PARAMS in fxt and recompiled it, but I need to patch starpu with the joint patch to set _starpu_last_codelet_symbol to its new "exact" size. The patch seems to work for me...

Second, when I use pj_dump to convert the file paje.trace, I get an error :

PajeTypeException: Type '(Type, name: Thread State, alias: S)' is not child type of container type '(Type, name: User Thread, alias: UT)' in (Line: 1408, Fields: '5, Contents: '10 1472.175387000 t123145303375872 S B')

Coming from this line in the trace:

10    1472.175387000    t123145303375872    S    B

that put t123145303375872 in the state "S", where t123145303375872 is defined as a "User Thread"

7    5479.882243000    t123145303375872    UT    p    UserThread123145303375872

The problem is that "S" in not a valid state for "UT", but for "T" (at least, this is my understanding of the bug).

As anyone allready seen this ? I have a workaround for this bug : in starpu_fxt.c line 803, calling user_thread_set_state() - that is availlable but commented as "unused" - instead of thread_set_state(). Is this a valid correction ?

Thx,

GS

-- 
Guillaume SYLVAND
Airbus Group Innovations / INRIA Bordeaux Sud-Ouest
Std : +33 (0)5 35 00 26 27
GSM : +33 (0)6 72 87 47 36
Fax : +33 (0)5 61 16 88 05
--------------------------

Postal Adress:
Inria Bordeaux – Sud-Ouest
HiePACS Team
200 avenue de la vieille tour
33405 Talence Cedex France

The information in this e-mail is confidential. The contents may not be disclosed or used by anyone other than the addressee. Access to this e-mail by anyone else is unauthorised.
If you are not the intended recipient, please notify Airbus immediately and delete this e-mail.
Airbus cannot accept any responsibility for the accuracy or completeness of this e-mail as it has been sent over public networks. If you have any concerns over the content of this message or its Accuracy or Integrity, please contact Airbus immediately.
All outgoing e-mails from Airbus are checked using regularly updated virus scanning software but you should take whatever measures you deem to be appropriate to ensure that this message and any attachments are virus free.

diff --git a/src/debug/traces/starpu_fxt.c b/src/debug/traces/starpu_fxt.c
index 749cc17..0a4d650 100644
--- a/src/debug/traces/starpu_fxt.c
+++ b/src/debug/traces/starpu_fxt.c
@@ -276,7 +278,7 @@ static unsigned get_colour_symbol_blue(char *name)
 
 static double last_codelet_start[STARPU_NMAXWORKERS];
 /* _STARPU_FUT_DO_PROBE4STR records only 4 longs */
-char _starpu_last_codelet_symbol[STARPU_NMAXWORKERS][4*sizeof(unsigned 
long)];
+char 
_starpu_last_codelet_symbol[STARPU_NMAXWORKERS][(FXT_MAX_PARAMS-4)*sizeof(unsigned
 long)];
 static int last_codelet_parameter[STARPU_NMAXWORKERS];
 #define MAX_PARAMETERS 8
 static char 
last_codelet_parameter_description[STARPU_NMAXWORKERS][MAX_PARAMETERS][FXT_MAX_PARAMS*sizeof(unsigned
 long)];
@@ -533,7 +535,7 @@ static void thread_set_state(double time, const char 
*prefix, long unsigned int
diff --git a/src/debug/traces/starpu_fxt.h b/src/debug/traces/starpu_fxt.h
index 0ab5a89..d2de7f4 100644
--- a/src/debug/traces/starpu_fxt.h
+++ b/src/debug/traces/starpu_fxt.h
@@ -38,7 +39,7 @@
 #include <starpu.h>
 #include "../../../include/starpu_fxt.h"
 
-extern char 
_starpu_last_codelet_symbol[STARPU_NMAXWORKERS][4*sizeof(unsigned long)];
+extern char 
_starpu_last_codelet_symbol[STARPU_NMAXWORKERS][(FXT_MAX_PARAMS-4)*sizeof(unsigned
 long)];
 
 void _starpu_fxt_dag_init(char *dag_filename);
 void _starpu_fxt_dag_terminate(void);