[ppc] Bump (more) instruction sizes after 51c454f6b03b468273bb93db2aa9b76e330a7e57.
authorMauricio Faria de Oliveira <mauricfo@linux.vnet.ibm.com>
Tue, 28 Oct 2014 18:07:33 +0000 (16:07 -0200)
committerMauricio Faria de Oliveira <mauricfo@linux.vnet.ibm.com>
Mon, 17 Nov 2014 02:34:13 +0000 (02:34 +0000)
commit3953793f0c064abc2f6964708d290897afc47e62
tree87c6325056ff086cf1adf9a2865d4c4d571f2c0a
parentba62bdf862560a85e2d637a0762d0e4fdd0db8b0
[ppc] Bump (more) instruction sizes after 51c454f6b03b468273bb93db2aa9b76e330a7e57.

This fixes ~30 ppc-specific failures in mono/tests (e.g., appdomain), like this:
wrong maximal instruction length of instruction atomic_add_i4 (expected 20, got 28)
* Assertion: should not be reached at mini-ppc.c:5235

The commit 51c454f6b03b468273bb93db2aa9b76e330a7e57 introduced 2 ppc_sync
instructions in OP_ATOMIC_CAS_I4/8 *and* OP_ATOMIC_ADD_NEW_I4/8 (mini-ppc64.c).

Accordingly, commit 83a34c40c6a0911a0c27afce63f5ea7ec2bb4b26 updated the maxlength
(+8 bytes) of atomic_cas_i4/8 - *but not* atomic_add_new_i4/8 (cpu-ppc64.md).

Recently, commit b784b866dac0c97780357bf81bf0fa3ebeffa4e made atomic_add_new_i4/8
exercised (as 'atomic_add_i4/8') - with an incorrect/non-updated 20-byte maxlength,
which should actually be 28.

Thus:
wrong maximal instruction length of instruction atomic_add_i4 (expected 20, got 28)

Signed-off-by: Mauricio Faria de Oliveira <mauricfo@linux.vnet.ibm.com>
mono/mini/cpu-ppc64.md