[dotgnu-pnet-commits] [SCM] DotGNU Portable.NET engine, compilers and tools (pnet) branch, master, updated. 9fa68a32bcd0895a2324b5a76ae90368931c7515

[Klaus Treichel]

This is an automated email from the git hooks/post-receive script. It was
generated because a ref change was pushed to the repository containing
the project "DotGNU Portable.NET engine, compilers and tools (pnet)".

The branch, master has been updated
       via  9fa68a32bcd0895a2324b5a76ae90368931c7515 (commit)
      from  60de7fe82155540ca8f4bb1102d462d2b9fe709b (commit)

Those revisions listed above that are new to this repository have
not appeared on any other notification email; so we list those
revisions in full, below.

- Log -----------------------------------------------------------------
http://git.savannah.gnu.org/cgit/pnet.git/commit/?id=9fa68a32bcd0895a2324b5a76ae90368931c7515

commit 9fa68a32bcd0895a2324b5a76ae90368931c7515
Author: Klaus Treichel <address@hidden>
Date:   Wed Apr 7 19:04:59 2010 +0200

    Reenable unrolling operations with longs on arm eabi.

diff --git a/ChangeLog b/ChangeLog
index 10b5247..7a689de 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,6 +1,7 @@
 2010-04-07  Klaus Treichel  <address@hidden>
 
        * engine/cvm.h: Mark the no longer used opcode COP_CCTOR_ONCE unsused.
+       Omit the padding of the CVMWord for arm eabi too.
 
        * engine/cvm_call.c: Remove the code for the COP_CCTOR_ONCE opcode.
 
@@ -23,6 +24,9 @@
        * engine/unroll.c (GetTopTwoWordAndFPRegisters): Function added for
        floatingpoint array stores.
 
+       * engine/unroll_arith.c: Reenable unrolling of operations with longs for
+       arm eabi.
+
        * engine/unroll_conv.c (COP_I2F, COP_IU2F, COP_L2F, COP_LU2F): Add
        unroller support for these cvm opcodes.
 
diff --git a/engine/cvm.h b/engine/cvm.h
index 1552206..dda6b75 100644
--- a/engine/cvm.h
+++ b/engine/cvm.h
@@ -615,8 +615,11 @@ typedef union
        /* Pad this structure to the best alignment on the underlying platform.
           This is usually needed on 64-bit platforms to ensure that stack
           words are always aligned on the best boundary.  We don't do this for
-          i386 because IL_BEST_ALIGNMENT is sometimes 8, and we need it to be 
4 */
-#if !defined(__i386) && !defined(__i386__)
+          i386 because IL_BEST_ALIGNMENT is sometimes 8, and we need it to be 
4.
+          We don't do this for arm eabi too because the size would be 8 there
+          too. The size *MUST* be the size of an ILInt32 on 32bit archs for the
+          unroller to work. */
+#if !defined(__i386) && !defined(__i386__) && !defined(__ARM_EABI__)
        char            padding[IL_BEST_ALIGNMENT];
 #endif
 
diff --git a/engine/unroll_arith.c b/engine/unroll_arith.c
index 7bfad10..304216c 100644
--- a/engine/unroll_arith.c
+++ b/engine/unroll_arith.c
@@ -350,7 +350,7 @@ case COP_INEG:
 }
 break;
 
-#if !defined(CVM_PPC) && !defined(__ARM_EABI__) /* has 8 byte CVMWords */
+#if !defined(CVM_PPC) /* has 8 byte CVMWords */
 
 case COP_LADD:
 {

-----------------------------------------------------------------------

Summary of changes:
 ChangeLog             |    4 ++++
 engine/cvm.h          |    7 +++++--
 engine/unroll_arith.c |    2 +-
 3 files changed, 10 insertions(+), 3 deletions(-)


hooks/post-receive
-- 
DotGNU Portable.NET engine, compilers and tools (pnet)