Reduces all vector upsamples down to its scalar components, so probably
not the most efficient thing in the world, but it does what the
spec says it needs to do.
Another possible implementation would be to convert/cast everything as
unsigned if necessary, upsample the input vectors, create the upsampled
value, and then cast back to signed if required.
Signed-off-by: Aaron Watry <awatry@gmail.com>
Reviewed-by: Tom Stellard <thomas.stellard at amd.com>
llvm-svn: 186691
The assembly optimizations were making unsafe assumptions about which address
spaces had which identifiers.
Also, fix vload/vstore with 64-bit pointers. This was broken previously on
Radeon SI.
This version still only has assembly versions of int/uint 2/4/8/16 for global
loads and stores on R600, but it does it in a way that would be very easily
extended to private/local/constant and could also be handled easily on other
architectures.
v2: 1) Leave v[load|store]_impl.ll in generic/lib
2) Remove vload_if.ll and vstore_if.ll interfaces
3) Fix address+offset calculations
3) Remove offset from assembly arg list
llvm-svn: 186416
This commit gets us back to pure CLC and fixes offset calculations.
The next commit will re-enable the assembly implementation for R600,
fix bugs related to 64-bit address spaces, and also fix the
incorrect assumption that address space identifiers are the same in
all architectures.
llvm-svn: 186415
libclc was defining and undefing GENTYPE and several other macros with
common names in its header files. This was preventing applications from
defining macros with identical names as command line arguments to the
compiler, because the definitions in the header files were masking the
macros defined as compiler arguements.
Reviewed-by: Aaron Watry <awatry@gmail.com>
llvm-svn: 185838
The assembly should be generic, but at least currently R600 only supports
32-bit stores of [u]int1/4, and I believe that only global is well-supported.
R600 lowers the 8/16 component stores to multiple 4-component stores.
The unoptimized C versions of the other stuff is left in place.
Patch by: Aaron Watry
llvm-svn: 185009
The assembly should be generic, but at least currently R600 only supports
32-bit loads of int1/4, and I believe that only global is well-supported.
R600 lowers the 8/16 component vectors to multiple 4-bit loads.
The unoptimized C versions of the other stuff is left in place.
Patch by: Aaron Watry
llvm-svn: 185008
Squashed commit of the following:
commit a0df0a0e86c55c1bdc0b9c0f5a739e5adef4b056
Author: Aaron Watry <awatry@gmail.com>
Date: Mon Apr 15 18:42:04 2013 -0500
libclc: Rename clz.ll to clz_if.ll to ensure it gets built.
configure.py treats files that have the same name with the .cl and .ll
extensions as overriding eachother.
E.g. If you have clz.cl and clz.ll both specified to be built in the same
SOURCES file, only the first file listed will actually be built.
Since the contents of clz.ll were an interface that is implemented in
clz_impl.ll, rename clz.ll to clz_if.ll to make sure that the interface is
built.
commit 931b62bed05c58f737de625bd415af09571a6a5a
Author: Aaron Watry <awatry@gmail.com>
Date: Sat Apr 13 12:32:54 2013 -0500
libclc: llvm assembly implementation of clz
Untested... currently crashes in the same manner as add_sat.
commit 6ef0b7b0b6d2e5584086b4b9a9243743b2e0538f
Author: Aaron Watry <awatry@gmail.com>
Date: Sat Mar 23 12:35:27 2013 -0500
libclc: Add stub clz builtin
For scalar int/uint, attempt to use the clz llvm builtin.. for all others
return 0 until an actual implementation is finished.
Patch by: Aaron Watry
llvm-svn: 185004
For any GENTYPE that isn't scalar, we need to implement a mixed
vector/scalar version of clamp/max.
This depends on the min() patches I sent to the list a few minutes ago.
Patch by: Aaron Watry
llvm-svn: 185003
Checks if the current GENTYPE is scalar, and if not, then defines a separate
implementation of the function which casts the second arg to vector before
proceeding.
Patch by: Aaron Watry
llvm-svn: 185002
configure.py allows overloading *.cl with *.ll, but will only ever build
the first file listed in SOURCES of ${file}.cl and ${file}.ll
add_sat, sub_sat, (and the soon to be submitted clz) all define interfaces in
${function_name}.ll which are implemented in ${function_name}_impl.ll.
Renaming the interface files is enough to get them to build again, fixing
CL usage of these functions.
Tested on clover/r600g.
Patch by: Aaron Watry
llvm-svn: 185000
This implementation does a lot of bit shifting and masking. Suffice to say,
this is somewhat suboptimal... but it does look to produce correct results
(after the piglit tests were corrected for sign extension issues).
Someone who knows LLVM better than I could re-write this more efficiently.
Patch by: Aaron Watry
llvm-svn: 184996
Created under a new shared/ directory for functions which are available for
both integer and floating point types.
Patch by: Aaron Watry
llvm-svn: 184994
Aaron Watry