On Nov 24, 2012, at 1:08 PM, Mark Cave-Ayland wrote:
On 20/11/12 03:35, Programmingkid wrote:
The filll word fills a region of memory with a 4 byte value. This word is different from fill because fill is used to fill a region of memory with a 1 byte value.
This word is needed by bootx to boot Mac OS X.
Here are some test results comparing Apple's implementation to mine.
Apple's implementation:
0> 100 buffer: mybuffer ok 0> mybuffer 100 87654321 filll ok 0> mybuffer 100 dump ffbc8800: 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 |.eC!.eC!.eC!.eC!| ffbc8810: 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 |.eC!.eC!.eC!.eC!| ffbc8820: 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 |.eC!.eC!.eC!.eC!| ffbc8830: 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 |.eC!.eC!.eC!.eC!| ffbc8840: 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 |.eC!.eC!.eC!.eC!| ffbc8850: 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 |.eC!.eC!.eC!.eC!| ffbc8860: 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 |.eC!.eC!.eC!.eC!| ffbc8870: 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 |.eC!.eC!.eC!.eC!| ffbc8880: 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 |.eC!.eC!.eC!.eC!| ffbc8890: 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 |.eC!.eC!.eC!.eC!| ffbc88a0: 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 |.eC!.eC!.eC!.eC!| ffbc88b0: 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 |.eC!.eC!.eC!.eC!| ffbc88c0: 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 |.eC!.eC!.eC!.eC!| ffbc88d0: 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 |.eC!.eC!.eC!.eC!| ffbc88e0: 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 |.eC!.eC!.eC!.eC!| ffbc88f0: 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 |.eC!.eC!.eC!.eC!|
My patch:
0> mybuffer 100 87654321 filll ok 0> mybuffer 100 dump fff771e0 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 ?eC!?eC!?eC!?eC! fff771f0 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 ?eC!?eC!?eC!?eC! fff77200 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 ?eC!?eC!?eC!?eC! fff77210 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 ?eC!?eC!?eC!?eC! fff77220 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 ?eC!?eC!?eC!?eC! fff77230 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 ?eC!?eC!?eC!?eC! fff77240 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 ?eC!?eC!?eC!?eC! fff77250 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 ?eC!?eC!?eC!?eC! fff77260 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 ?eC!?eC!?eC!?eC! fff77270 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 ?eC!?eC!?eC!?eC! fff77280 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 ?eC!?eC!?eC!?eC! fff77290 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 ?eC!?eC!?eC!?eC! fff772a0 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 ?eC!?eC!?eC!?eC! fff772b0 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 ?eC!?eC!?eC!?eC! fff772c0 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 ?eC!?eC!?eC!?eC! fff772d0 87 65 43 21 87 65 43 21 87 65 43 21 87 65 43 21 ?eC!?eC!?eC!?eC!
The results are the same. This indicates the patch will work.
After I applied this patch, bootx continued booting a little further than before.
signed-off-by: John Arbuckleprogrammingkidx@gmail.com
Index: kernel/bootstrap.c
--- kernel/bootstrap.c (revision 1067) +++ kernel/bootstrap.c (working copy) @@ -89,7 +89,7 @@ "here", "here!", "dobranch", "do?branch", "unaligned-w@", "unaligned-w!", "unaligned-l@", "unaligned-l!", "ioc@", "iow@", "iol@", "ioc!", "iow!", "iol!", "i", "j", "call", "sys-debug",
- "$include", "$encode-file", "(debug", "(debug-off)"
- "$include", "$encode-file", "(debug", "(debug-off)", "filll"
};
/* Index: kernel/forth.c =================================================================== --- kernel/forth.c (revision 1067) +++ kernel/forth.c (working copy) @@ -1611,6 +1611,24 @@ }
+// filll ( addr len long-number -- )
+static void filll(void) +{
- int * location;
- int length, fillNumber, index;
- fillNumber = POP();
- length = POP();
- location = (int *) POP();
- for(index = 0; index<= length; index++)
- {
location[index] = fillNumber;- }
+}
/*
- unaligned-w@ ( addr -- w )
*/ @@ -1963,4 +1981,5 @@ do_encode_file, /* $encode-file */ do_debug_xt, /* (debug */ do_debug_off, /* (debug-off) */
- filll, /* filll */
};
I can see that the patch works, but in order to commit it I'd like to see the following changes:
- Move the filll C function somewhere under arch/ppc/qemu/ and use bind_func() somewhere in an initialiser to "hook" it into Forth. This is to ensure that we keep the non-IEEE standard words separate from the core project.
I don't see any advantage to doing this. Is there a filll (3 l's) word on any other platform?
Is this how the bind_func call would look like: bind_func("filll", filll); ? Where should I place this call?
- Make the method for filling the memory work better on other architectures. The problem with the above is that you're assuming that sizeof(int) == 4 which may not always be the case (think PPC64). Take a look at lstore() in kernel/forth.c to see how this should be done in a multi-arch friendly way for 32-bit values.
Sounds like a good suggestion.