- reduce 3.1 sdk regression test runtimes (as per 2.3 sdk versions)

- add ldu support (just use ld for now)
- get rid of asserts for .const[n] 
- status:
	cuda sdk 3.1 dct8x8, eigenvalues, MersenneTwister, reduction, scan still running (too slow)
	ocl sdk 3.1 oclQuasirandomGenerator still running (too slow)
	ocl sdk 3.1 oclReduction fail comparison (cuda 3.1)
	ocl sdk 3.1 oclMatrixMul exits mid computation
	ocl sdk 3.1 oclTranspose hits "clCreateBuffer - buffer already created for this host variable"
	ocl sdk 3.1 oclSortingNetworks fails w/ message about min workgroup size
	remaining 30 tests pass

[git-p4: depot-paths = "//depot/gpgpu_sim_research/fermi/distribution/": change = 6927]

1 files changed, 7 insertions, 1 deletions

diff --git a/src/cuda-sim/memory.cc b/src/cuda-sim/memory.cc
index 754fe7f..09c9e81 100644
--- a/src/cuda-sim/memory.cc
+++ b/src/cuda-sim/memory.cc
@@ -94,7 +94,13 @@ template<unsigned BSIZE> void memory_space_impl<BSIZE>::read( mem_addr_t addr, s
    unsigned offset = addr & (BSIZE-1);
    unsigned nbytes = length;
    typename map_t::const_iterator i = m_data.find(index);
-   assert( (addr+length) <= (index+1)*BSIZE );
+   if( (addr+length) > (index+1)*BSIZE ) {
+      printf("GPGPU-Sim PTX: ERROR * access to memory \'%s\' is unaligned : addr=0x%x, length=%zu\n",
+             m_name.c_str(), addr, length);
+      printf("GPGPU-Sim PTX: (addr+length)=0x%lx > 0x%x=(index+1)*BSIZE, index=0x%x, BSIZE=0x%x\n",
+             (addr+length),(index+1)*BSIZE, index, BSIZE);
+      throw 1;
+   }
    if( i == m_data.end() ) {
       for( size_t n=0; n < length; n++ ) 
          ((unsigned char*)data)[n] = (unsigned char) 0;