Build breaks on 32-bit systems

[yurivict]

FAILED: src/CMakeFiles/mkldnn.dir/cpu/ncsp_batch_normalization.cpp.o 
/usr/bin/c++  -DMKLDNN_DLL -DMKLDNN_DLL_EXPORTS -D__STDC_CONSTANT_MACROS -D__STDC_LIMIT_MACROS -Dmkldnn_EXPORTS -I/wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/include -I/wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/src -I/wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/src/common -I/wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/src/cpu/xbyak -O2 -pipe -fstack-protector -fno-strict-aliasing -std=c++11 -fvisibility-inlines-hidden  -Wall -Werror -Wno-unknown-pragmas -fvisibility=internal  -fPIC -Wformat -Wformat-security -fstack-protector-all   -Wno-pass-failed -O2 -pipe -fstack-protector -fno-strict-aliasing -D_FORTIFY_SOURCE=2 -fPIC   -std=gnu++11 -MD -MT src/CMakeFiles/mkldnn.dir/cpu/ncsp_batch_normalization.cpp.o -MF src/CMakeFiles/mkldnn.dir/cpu/ncsp_batch_normalization.cpp.o.d -o src/CMakeFiles/mkldnn.dir/cpu/ncsp_batch_normalization.cpp.o -c /wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/src/cpu/ncsp_batch_normalization.cpp
In file included from /wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/src/cpu/ncsp_batch_normalization.cpp:20:
In file included from /wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/src/cpu/cpu_batch_normalization_utils.hpp:21:
In file included from /wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/src/cpu/cpu_batch_normalization_pd.hpp:23:
In file included from /wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/src/common/batch_normalization_pd.hpp:23:
In file included from /wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/src/common/primitive_desc.hpp:24:
In file included from /wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/src/common/type_helpers.hpp:29:
/wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/src/common/math_utils.hpp:88:5: error: comparison of constant 4294967296 with expression of type 'size_t' (aka 'unsigned int') is always false [-Werror,-Wtautological-constant-out-of-range-compare]
    CP(32); CP(16); CP(8); CP(4); CP(2); CP(1);
    ^~~~~~
/wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/src/common/math_utils.hpp:87:30: note: expanded from macro 'CP'
#   define CP(pw) do { if (v >= (1ull << pw)) { v >>= pw; p += pw; } } while(0)
                           ~ ^  ~~~~~~~~~~~~
/wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/src/common/math_utils.hpp:88:5: error: shift count >= width of type [-Werror,-Wshift-count-overflow]
    CP(32); CP(16); CP(8); CP(4); CP(2); CP(1);
    ^  ~~
/wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/src/common/math_utils.hpp:87:51: note: expanded from macro 'CP'
#   define CP(pw) do { if (v >= (1ull << pw)) { v >>= pw; p += pw; } } while(0)
                                                  ^   ~~
In file included from /wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/src/cpu/ncsp_batch_normalization.cpp:22:
In file included from /wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/src/cpu/jit_generator.hpp:34:
/wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/src/cpu/xbyak/xbyak.h:1395:54: error: shift count >= width of type [-Werror,-Wshift-count-overflow]
        static const size_t dummyAddr = (size_t(0x11223344) << 32) | 55667788;
                                                            ^  ~~
/wrkdirs/usr/ports/math/mkl-dnn/work/mkl-dnn-0.16/src/cpu/xbyak/xbyak.h:1589:24: error: shift count >= width of type [-Werror,-Wshift-count-overflow]
                size_t high = disp64 >> 32;
                                     ^  ~~

[okapies]

README says “Intel MKL-DNN supports Intel(R) 64 architecture and compatible architectures.”

[emfomenk]

Hi @yurivict,

Yes, @okapies is absolutely right -- 32bits are not supported. See #171.
We will add a guard to cmake to not confuse people :)