[kaffe] mipsel and jit3

[Mikolaj Habryn]

On Tue, 2004-10-12 at 10:47, Dalibor Topic wrote:
> Are you in Sydney, by chance?

Indeed so! The city that never sleeps, and occasionally forgets to
bathe.

> Very nice trick, I like it. Makes me wonder how that would work with 
> endianness issues. 

Not at all, I'm afraid. Here's an alternative that leaves the smarts
regarding endianness, multi-size loads and unaligned accesses up to the
compiler to resolve through the deft application of memcpy, which any
reasonable compiler (ie, gcc) will resolve as a builtin function rather
than a library call.

--- kaffe/kaffe/kaffevm/jit3/funcs.c.~1~        2004-07-12 05:03:04.000000000 +1000
+++ kaffe/kaffe/kaffevm/jit3/funcs.c    2004-10-14 18:58:47.000000000 +1000
@@ -36,6 +36,12 @@
                   : CODEPC))

 #if defined(KAFFE_VMDEBUG)
+#define _mjh_OUT(v, var, count) do {           \
+    var _dummy = (v);                            \
+    DBGEXPR(JIT,(void)printCodeLabels(),0); \
+    memcpy(codeblock + CODEPC, &_dummy, count); \
+    CODEPC += count;                           \
+  } while(0)
 /*
  * Print any labels that refer to the current address.
  */
@@ -49,19 +55,18 @@
                kaffe_dprintf("%s:\n", il->name);
        }
 }
-#undef OUT
-#define        OUT(v)  do { DBGEXPR(JIT,(void)printCodeLabels(),0),codeblock[CODEPC] = v; CODEPC++; } while (0)
-#define        BOUT(v) do { DBGEXPR(JIT,(void)printCodeLabels(),0),*(uint8*)&codeblock[CODEPC] = v; CODEPC++; } while (0)
-#define        WOUT(v) do { DBGEXPR(JIT,(void)printCodeLabels(),0),*(uint16*)&codeblock[CODEPC] = v; CODEPC += 2; } while (0)
-#define        LOUT(v) do { DBGEXPR(JIT,(void)printCodeLabels(),0),*(uint32*)&codeblock[CODEPC] = v; CODEPC += 4; } while (0)
-#define        QOUT(v) do { DBGEXPR(JIT,(void)printCodeLabels(),0),*(uint64*)&codeblock[CODEPC] = v; CODEPC += 8; } while (0)
 #else /* !  defined(KAFFE_VMDEBUG) */
-#undef OUT
-#define        OUT(v)  do { codeblock[CODEPC] = v; CODEPC++; } while (0)
-#define        BOUT(v) do { *(uint8*)&codeblock[CODEPC] = v; CODEPC++; } while (0)
-#define        WOUT(v) do { *(uint16*)&codeblock[CODEPC] = v; CODEPC += 2; } while (0)
-#define        LOUT(v) do { *(uint32*)&codeblock[CODEPC] = v; CODEPC += 4; } while (0)
-#define        QOUT(v) do { *(uint64*)&codeblock[CODEPC] = v; CODEPC += 8; } while (0)
+#define _mjh_OUT(v, var, count) do {           \
+    var _dummy = (v); \
+    memcpy(codeblock + CODEPC, &_dummy, count); \
+    CODEPC += count;                           \
+  } while(0)
 #endif /* defined(KAFFE_VMDEBUG) */
+#define        BOUT(v) _mjh_OUT((v), u_int8_t, 1)
+#define        WOUT(v) _mjh_OUT((v), u_int16_t, 2)
+#define        LOUT(v) _mjh_OUT((v), u_int32_t, 4)
+#define        QOUT(v) _mjh_OUT((v), u_int64_t, 8)
+#undef OUT
+#define OUT BOUT

 #include "jit.def"


> Uh. Weird. I've seen stuff like that when I was fixing some mips jit 
> breakage, but didn't want to touch it without access to a mips. :(

Hey, I'm working on a US$80 MIPS/Linux box here. What's your postal
address? I'll buy you one :)

I'll come back to the other points I mentioned in my earlier mail, with
patches. I'm still hoping for a flash of inspiration and a one-line fix
that will magically make the whole thing work perfectly.

m.