PATCH : H8300 GDB simulator

Michael Snyder msnyder@redhat.com
Tue Dec 16 21:31:00 GMT 2003


Anil Paranjape wrote:
> Hi Michael,
> 
> Please find following source code in C as a testcase for H8 simulator bug,
> 
> *********************************************************************************
> int main()
> {
> 	unsigned char test[10];
> 	unsigned short *pWord;
> 	pWord = (unsigned short *) &test[1];
> 	memset ( test, 0, sizeof(test));
> 	*pWord = 0x1234;
> 	printf ("Test[0] = 0x%02x\r\n", test[0] );
> 	printf ("Test[1] = 0x%02x\r\n", test[1] );
> 	return ( 0 );
> }
> **********************************************************************************
> 
> Simulator output before patch,
> Test[0] = 0x00
> Test[1] = 0x12
> 
> Simulator output after patch and same as in case of H/W,
> Test[0] = 0x12
> Test[1] = 0x34
> 
> Regards,
> Anil

Hmmm...  Anil, does this apply to all memory-accessing instructions?
Normally we write the simulator tests in assembly language, so that
we can explicitly control what we are testing.  A year from now, the
compiler might generate different code.

And does it apply to all h8/300 variations, including the
"plain vanilla" h8?

> 
> -----Original Message-----
> From: Michael Snyder [mailto:msnyder@redhat.com]
> Sent: Friday, December 05, 2003 1:08 AM
> To: Anil Paranjape
> Cc: gdb-patches@sources.redhat.com
> Subject: Re: PATCH : H8300 GDB simulator
> 
> 
> Anil Paranjape wrote:
> 
>>Hi,
>>
>>The CPU can access word data and longword data in memory, but word or longword data must begin at an even address. If an attempt is made to access word or longword data at an odd address, no address error occurs but the least significant bit of the address is regarded as 0, so the access starts at the preceding address. This is the case in actual H/W.
>>
>>Now if an odd address is assigned to any word or long word memory pointer in H8 simulator and then if data at that address is accessed byte by byte then simulator was neither showing correct data (which is the case in actual H/W) not generating SIGBUS exception (which is the case in SH).
>>
>>Patch below fixes this problem and after patch, simulator shows correct value as is the case in actual H/W.
> 
> 
> This sounds good (though I hope one of the folks with more knowledge of
> the architecture will verify it).  One request -- could you include a
> testcase for the sim testsuite, such that it will fail before your mods,
> and pass after?  Existing tests (for examples) are at
> sim/testsuite/sim/h8300.
> 
> Thanks,
> Michael
> 
> 
> 
>>ChangeLog/sim/h8300
>>
>>2003-12-03  Anil Paranjpe <anilp1@kpitcummins.com>
>>
>>        * compile.c : In all word and double word memory operations, 
>>			    address is forced to be always even.
>>
>>Patch for /sim/h8300/compile.c
>>
>>--- compile.c.orig	Fri Oct 17 18:15:56 2003
>>+++ compile.c	Wed Dec  3 22:04:22 2003
>>@@ -1265,20 +1265,19 @@
>> 
>> #define GET_MEMORY_L(X) \
>>   ((X) < memory_size \
>>-   ? ((h8_get_memory (sd, (X)+0) << 24) | (h8_get_memory (sd, (X)+1) << 16)  \
>>-    | (h8_get_memory (sd, (X)+2) <<  8) | (h8_get_memory (sd, (X)+3) <<  0)) \
>>-   : ((h8_get_eightbit (sd, ((X)+0) & 0xff) << 24) \
>>-    | (h8_get_eightbit (sd, ((X)+1) & 0xff) << 16) \
>>-    | (h8_get_eightbit (sd, ((X)+2) & 0xff) <<  8) \
>>-    | (h8_get_eightbit (sd, ((X)+3) & 0xff) <<  0)))
>>+   ? ((h8_get_memory (sd, ((X) & 0xfffffffe)+0) << 24) | (h8_get_memory (sd, ((X) & 0xfffffffe)+1) << 16)  \
>>+    | (h8_get_memory (sd, ((X) & 0xfffffffe)+2) <<  8) | (h8_get_memory (sd, ((X) & 0xfffffffe)+3) <<  0)) \
>>+   : ((h8_get_eightbit (sd, (((X) & 0xfffffffe)+0) & 0xff) << 24) \
>>+    | (h8_get_eightbit (sd, (((X) & 0xfffffffe)+1) & 0xff) << 16) \
>>+    | (h8_get_eightbit (sd, (((X) & 0xfffffffe)+2) & 0xff) <<  8) \
>>+    | (h8_get_eightbit (sd, (((X) & 0xfffffffe)+3) & 0xff) <<  0)))
>> 
>> #define GET_MEMORY_W(X) \
>>   ((X) < memory_size \
>>-   ? ((h8_get_memory   (sd, (X)+0) << 8) \
>>-    | (h8_get_memory   (sd, (X)+1) << 0)) \
>>-   : ((h8_get_eightbit (sd, ((X)+0) & 0xff) << 8) \
>>-    | (h8_get_eightbit (sd, ((X)+1) & 0xff) << 0)))
>>-
>>+   ? ((h8_get_memory   (sd, ((X) & 0xfffffffe)+0) << 8) \
>>+    | (h8_get_memory   (sd, ((X) & 0xfffffffe)+1) << 0)) \
>>+   : ((h8_get_eightbit (sd, (((X) & 0xfffffffe)+0) & 0xff) << 8) \
>>+    | (h8_get_eightbit (sd, (((X) & 0xfffffffe)+1) & 0xff) << 0)))
>> 
>> #define GET_MEMORY_B(X) \
>>   ((X) < memory_size ? (h8_get_memory   (sd, (X))) \
>>@@ -1286,16 +1285,16 @@
>> 
>> #define SET_MEMORY_L(X, Y)  \
>> {  register unsigned char *_p; register int __y = (Y); \
>>-   _p = ((X) < memory_size ? h8_get_memory_buf   (sd) +  (X) : \
>>-                             h8_get_eightbit_buf (sd) + ((X) & 0xff)); \
>>+    _p = ((X) < memory_size ? h8_get_memory_buf   (sd) +  ((X) & 0xfffffffe) : \
>>+                             h8_get_eightbit_buf (sd) + (((X) & 0xfffffffe) & 0xff)); \
>>    _p[0] = __y >> 24; _p[1] = __y >> 16; \
>>    _p[2] = __y >>  8; _p[3] = __y >>  0; \
>> }
>> 
>> #define SET_MEMORY_W(X, Y) \
>> {  register unsigned char *_p; register int __y = (Y); \
>>-   _p = ((X) < memory_size ? h8_get_memory_buf   (sd) +  (X) : \
>>-                             h8_get_eightbit_buf (sd) + ((X) & 0xff)); \
>>+   _p = ((X) < memory_size ? h8_get_memory_buf   (sd) +  ((X) & 0xfffffffe) : \
>>+                             h8_get_eightbit_buf (sd) + (((X) & 0xfffffffe) & 0xff)); \
>>    _p[0] = __y >> 8; _p[1] = __y; \
>> }
>> 
>>
>>Regards,
>>Anil Paranjpe
>>
>>
> 
> 
> 
> 




More information about the Gdb-patches mailing list