Closed GoogleCodeExporter closed 9 years ago
[deleted comment]
[deleted comment]
Status update: 47/56 cuda sdk test passing. Next in line are atomic operations
and
vote instructions.
Original comment by mailsud...@gmail.com
on 11 Oct 2009 at 10:23
Vote and atomics fixed. The following bugs still exist:
1. For shared memory loads/stores. The offset from the base address should wrap
around to the shared memory size. It is necessary to be able to get at the
offset
before the base is added which can be added to a register earlier in the
program. We
need to have the original registers point to 0 and have the base added
immediately
before the instruction is issued.
Original comment by mailsud...@gmail.com
on 16 Oct 2009 at 9:10
All sdk 2.1 tests currently pass. However, the mandelbrot and particles
examples
produce images that are slightly off of the reference. I am considering these
to be
false positive passing tests at this point and am continuing to debug them.
Next in line after this is to verify the 2.2 and 2.3 sdks and parboil.
Finally, I am
going to add multithreading support to the translator before signing off on it.
Original comment by mailsud...@gmail.com
on 22 Oct 2009 at 11:10
All Parboil and SDK tests pass.
Original comment by gregory....@gatech.edu
on 4 Nov 2009 at 6:18
Original issue reported on code.google.com by
gregory....@gatech.edu
on 22 Sep 2009 at 11:01