9 کامیت‌ها

مولف	SHA1	پیام	تاریخ
Alex Voicu	4966518846	Revert "Revert adoption of CUDA indexing in general - this can only work with later versions of the compiler, just like module based dispatch, and thus must be guarded against usage in earlier (e.g. 1.6) versions." ... This reverts commit 4792475	2017-11-29 21:49:10 +00:00
Alex Voicu	2557000b56	Revert "Revert adoption of CUDA indexing in general - this can only work with later versions of the compiler, just like module based dispatch, and thus must be guarded against usage in earlier (e.g. 1.6) versions." ... This reverts commit d2fd1f5	2017-11-29 21:36:29 +00:00
Alex Voicu	30d90dab38	Re-sync with upstream.	2017-11-20 15:34:50 +00:00
Alex Voicu	0cc921f103	This implements the trivial change needed to move back from the hip{Something}_{x, y, z} macros to the natural CUDA syntax of Something.{x, y, z}. This is contained in lines 384-404 in hip_runtime.h. All of the other changes have to do with changing unit tests to use this syntax. The macros are retained for backwards compatibility.	2017-11-19 01:54:12 +00:00
Alex Voicu	0ad1308efd	Clean up trailing whitespace so as to reduce noise in #246.	2017-11-08 00:08:55 +00:00
Alex Voicu	28f87f7d2e	This switches HIP from its currently convoluted macro + pfe based dispatch mechanism to a more natural one partially based on the existing module API. The basic idea is that HCC will always correctly emit __global__ functions: as empty-bodied stubs, on host, and as kernels, on device. It then becomes trivial to obtain the mangled name on host, at dispatch, from the function's address, and then to use the mangled name to retrieve the kernel. This should address all problems stemming from serialisation, dubious mismatches due to the manufactured functor, macro-isms et al. It also immediately enables support for generalised globals as a consequence of that being available in the module API. Finally, it will make debug much easier, since the actual names of the __global__ functions will automatically be used in traces etc. One detail is that due to how dispatch works now (hipLaunchKernel and hipLaunchKernelGGL are themselves variadic function templates which deduce the function type of the callee), in certain cases it may be necesssary to insert explicit casts to ensure that the variadic argument list selects a viable overload - this can be observed in some unit tests. Eventually we may be able to remove this limitation, but for now it does not appear terribly onerous. The code is not extremely HIPpie, nor is it fully optimised, but rather is intended as a starting point for the HIP team to make its own.	2017-11-01 15:09:59 +00:00
Ben Sander	8dc968f036	Updates so hip compiles on CUDA.	2017-05-27 15:55:07 -05:00
Ben Sander	8bc6ee5932	Add initial HIP_SYNC_NULL_STREAM=0 mode. ... This eliminates host-synchronization for null stream. Instead, the null-stream uses GPU-side events to wait for other streams. Default is OFF pending additional testing. Add enhanced null-stream test. Also refine HIP_TRACE_API.	2017-05-16 19:04:25 -05:00
Ben Sander	61c8633113	split debugging into separate .md file	2017-05-16 19:04:25 -05:00