AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Support wave32/wave64 scratch allocations on gfx10

Browse Source 
- Use new buffer resource descriptor layout
- Handle wave32 scratch allocation error from CP
- Make wavefront size a property of scratch allocation requests
- Repurpose wave64-specific amd_queue_t.scratch_workitem_byte_size field
- Clear index_stride field in V# on gfx10, calculated per-dispatch by CP

Change-Id: If2acdf6430772abd4d6a8c792fc8c11260764dda
This commit is contained in:
Jay Cornwall
2018-09-18 15:09:08 -05:00
committed by Chris Freehill
parent 08841faf4c
commit f8d0ccd159
5 changed files with 103 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files
runtime/hsa-runtime/inc/amd_hsa_queue.h
+1 -1
View File
@@ -77,7 +77,7 @@ typedef struct AMD_QUEUE_ALIGN amd_queue_s {
uint32_t scratch_resource_descriptor[4];
uint64_t scratch_backing_memory_location;
uint64_t scratch_backing_memory_byte_size;
uint32_t scratch_workitem_byte_size;
uint32_t scratch_wave64_lane_byte_size;
amd_queue_properties32_t queue_properties;
uint32_t reserved3[2];
hsa_signal_t queue_inactive_signal;