Fwd: [RFA] Fix for mcore simulator

Fwd: [RFA] Fix for mcore simulator

[Jeff Law via Gdb]

[-- Attachment #1: Type: text/plain, Size: 1395 bytes --]


Wrongly sent to binutils initially...


-------- Forwarded Message --------
Subject: [RFA] Fix for mcore simulator
Date: Wed, 4 Oct 2023 16:09:35 -0600
From: Jeff Law <jeffreyalaw@gmail.com>
To: Binutils <binutils@sourceware.org>


I was looking for cases where a GCC patch under evaluation would cause 
test results to change.  Quite surprisingly the mcore-elf port showed 
test differences.   After a fair amount of digging my conclusion was the 
sequences before/after the patch should have been semantically the same.


Of course if the code is supposed to behave the same, then that points 
to problems elsewhere (assembler, linker, simulator).  Sure enough the 
mcore simulator was mis-handling the sign extension instructions.  The 
simulator implementation of sextb is via paired shift-by-24 operations. 
Similarly the simulator implements sexth via paired shift-by-16 operations.

The temporary holding the value was declared as a "long" thus this 
approach worked fine for hosts with a 32 bit wide long and failed 
miserably for hosts with a 64 bit wide long.

This patch makes the shift count automatically adjust based on the size 
of the temporary.  It includes a simple test for sextb and sexth.  I 
have _not_ done a full audit of the mcore simulator for more 32->64 bit 
issues.

This also fixes 443 execution tests in the GCC testsuite ;-)



OK for the trunk?

Thanks,
Jeff



[-- Attachment #2: mcore-sim.patch --]
[-- Type: text/plain, Size: 1714 bytes --]

diff --git a/sim/mcore/interp.c b/sim/mcore/interp.c
index 53cfdad050b..48d9ff8645a 100644
--- a/sim/mcore/interp.c
+++ b/sim/mcore/interp.c
@@ -641,8 +641,8 @@ step_once (SIM_DESC sd, SIM_CPU *cpu)
 	      {
 		long tmp;
 		tmp = gr[RD];
-		tmp <<= 24;
-		tmp >>= 24;
+		tmp <<= (sizeof (tmp) * 8) - 8;
+		tmp >>= (sizeof (tmp) * 8) - 8;
 		gr[RD] = tmp;
 	      }
 	      break;
@@ -653,8 +653,8 @@ step_once (SIM_DESC sd, SIM_CPU *cpu)
 	      {
 		long tmp;
 		tmp = gr[RD];
-		tmp <<= 16;
-		tmp >>= 16;
+		tmp <<= (sizeof (tmp) * 8) - 16;
+		tmp >>= (sizeof (tmp) * 8) - 16;
 		gr[RD] = tmp;
 	      }
 	      break;
diff --git a/sim/testsuite/mcore/sextb.s b/sim/testsuite/mcore/sextb.s
new file mode 100644
index 00000000000..5500f7abe67
--- /dev/null
+++ b/sim/testsuite/mcore/sextb.s
@@ -0,0 +1,25 @@
+# check that sext.b/sext.h work correctly
+# mach: mcore
+
+.include "testutils.inc"
+
+	start
+	# Construct -120 using bgeni+addi+sext
+	bgeni	r2, 7
+	addi	r2,8
+	sextb	r2
+
+	# Construct -120 using movi+not
+	movi	r7,119
+	not	r7
+
+	# Compare them, they should be equal
+	cmpne	r2,r7
+	jbt	.L1
+	pass
+.L1:
+	fail
+
+
+
+
diff --git a/sim/testsuite/mcore/sexth.s b/sim/testsuite/mcore/sexth.s
new file mode 100644
index 00000000000..97279c49ed4
--- /dev/null
+++ b/sim/testsuite/mcore/sexth.s
@@ -0,0 +1,27 @@
+# check that sext.b/sext.h work correctly
+# mach: mcore
+
+.include "testutils.inc"
+
+	start
+	# Construct -32760 using bgeni+addi+sext
+	bgeni	r2, 15
+	addi	r2,8
+	sexth	r2
+
+	# Construct -32760 using bmask+subi+not
+        bmaski  r7,15
+        subi    r7,8    // 32759 0x7ff7
+	not	r7
+
+
+	# Compare them, they should be equal
+	cmpne	r2,r7
+	jbt	.L1
+	pass
+.L1:
+	fail
+
+
+
+

Re: Fwd: [RFA] Fix for mcore simulator

[Tom Tromey]

>>>>> "Jeff" == Jeff Law via Gdb <gdb@sourceware.org> writes:

Jeff> Wrongly sent to binutils initially...

Thanks, this is ok.

Tom

Re: Fwd: [RFA] Fix for mcore simulator

[Martin Simmons]

>>>>> On Thu, 5 Oct 2023 10:09:44 -0600, Jeff Law via Gdb said:
> 
> Of course if the code is supposed to behave the same, then that points 
> to problems elsewhere (assembler, linker, simulator).  Sure enough the 
> mcore simulator was mis-handling the sign extension instructions.  The 
> simulator implementation of sextb is via paired shift-by-24 operations. 
> Similarly the simulator implements sexth via paired shift-by-16 operations.
> 
> The temporary holding the value was declared as a "long" thus this 
> approach worked fine for hosts with a 32 bit wide long and failed 
> miserably for hosts with a 64 bit wide long.
> 
> This patch makes the shift count automatically adjust based on the size 
> of the temporary.  It includes a simple test for sextb and sexth.  I 
> have _not_ done a full audit of the mcore simulator for more 32->64 bit 
> issues.

The use of long seems bogus to me.  Why not just declare tmp as int32_t?

__Martin

Re: Fwd: [RFA] Fix for mcore simulator

[Jeff Law via Gdb]

On 10/10/23 15:35, Martin Simmons wrote:
>>>>>> On Thu, 5 Oct 2023 10:09:44 -0600, Jeff Law via Gdb said:
>>
>> Of course if the code is supposed to behave the same, then that points
>> to problems elsewhere (assembler, linker, simulator).  Sure enough the
>> mcore simulator was mis-handling the sign extension instructions.  The
>> simulator implementation of sextb is via paired shift-by-24 operations.
>> Similarly the simulator implements sexth via paired shift-by-16 operations.
>>
>> The temporary holding the value was declared as a "long" thus this
>> approach worked fine for hosts with a 32 bit wide long and failed
>> miserably for hosts with a 64 bit wide long.
>>
>> This patch makes the shift count automatically adjust based on the size
>> of the temporary.  It includes a simple test for sextb and sexth.  I
>> have _not_ done a full audit of the mcore simulator for more 32->64 bit
>> issues.
> 
> The use of long seems bogus to me.  Why not just declare tmp as int32_t?
That code likely predates any modernization efforts in gdbsim and 
binutils-gdb as a whole -- it's the old interp style simulator that was 
common in the 90s.

I wouldn't lose any sleep if someone took up that task (modernizing that 
codebase) but my interest in mcore is essentially zero, so it won't be me.

Jeff