3gx

The problem is caused by this: ```c uniform bool write[NBLOCK] = { true }; if (programIndex == 0) { for (int i = 0; i < NHALOL; i++) write[i] =...

further update: this test code generate the following compile error: --- ```c uniform double * uniform fw1d; #define NBLOCK 32 #define KE 10 #define NHALOR 1 #define NHALOL 1 #define...

Have you considered thrust::inclusive_scan_by_key ? [1] [1] https://thrust.github.io/doc/group__segmentedprefixsums.html

Sorting 32 byte strings will take a lot longer than sorting 4 byte ints with radix sort. 8x more bits to sort and 8x more data to move around per...

Your proposed solution sounds okay to me, if char[4] can be treated as unsigned int. Use SortPair in this case, with the value in (key,value) pair being either the corresponding...

From my vantage point there are currently two issues here. The first one, as James pointed out, is lack of LLVM tools to generate SPIR-V from the LLVM. Since ISPC...