Add tzcnt as an acceptable instruction to the x86-32 validator (it already

Add tzcnt as an acceptable instruction to the x86-32 validator (it already
handled lzcnt).

BUG= http://code.google.com/p/nativeclient/issues/detail?id=2869
Committed: https://src.chromium.org/viewvc/native_client?view=rev&revision=9281

[Roland McGrath, 2012-07-24 20:52:39]

This change has the wrong semantics.  Validation doesn't complain about a
tzcnt instruction, but it stubs it out (I imagine based on the absence of
the CPUID bit that says tzcnt per se is supported).  But that's not what we
want to happen.  tzcnt is also "rep bsf" and should be permitted on any CPU.
CPUs without tzcnt support will execute it as bsf.

[Karl, 2012-07-24 21:15:44]

http://codereview.chromium.org/10809021/diff/14001/src/trusted/validator/x86/...
File src/trusted/validator/x86/ncval_seg_sfi/generator/ncdecode_table.c (right):

http://codereview.chromium.org/10809021/diff/14001/src/trusted/validator/x86/...
src/trusted/validator/x86/ncval_seg_sfi/generator/ncdecode_table.c:2175:
EncodeOpF30F(0xbc, 1, IMM_NONE, NACLi_386, "tzcnt $Gv, $Ev");
Followed Roland's advice and made it work for all processors.

http://codereview.chromium.org/10809021/diff/14001/src/trusted/validator/x86/...
src/trusted/validator/x86/ncval_seg_sfi/generator/ncdecode_table.c:2183:
EncodeOpF30F(0xbd, 1, IMM_NONE, NACLi_386, "lzcnt $Gv, $Ev");
Did the same for lzcnt.

[Karl, 2012-07-26 19:30:48]

Is anyone reviewing this CL?

[Brad Chen, 2012-07-27 17:29:13]

LGTM