gpu: ocl: add dim_t type for offset calculations

The new `dim_t` type can be used for offset calculations. Some control of this type is enabled via `kernel_ctx::require_32_bit_offset()` for use on small problems on hardware without native 64 bit arithmetic.
oneapi-src · May 12, 2023 · d058bd8 · d058bd8
1 parent 7c23d9e
commit d058bd8
Showing 1 changed file with 2 additions and 0 deletions.
diff --git a/src/gpu/ocl/ocl_types.h b/src/gpu/ocl/ocl_types.h
@@ -37,6 +37,8 @@
 #pragma OPENCL EXTENSION cl_khr_fp64 : enable
 #endif
 
+#define dim_t long // 64 bit per the OpenCL specification
+
 #if DT_F32 == 1
 #define DATA_T float
 #define DATA2_T float2