Sameer Sahasrabuddhe

This is a copy of the internal issue SWDEV-167691. In the attached file, the loop condition depends on __ockl_lane_u32() which itself expands to mbcnt_lo and mbcnt_hi. These intrinsics are marked...

When using a **debug build** of hcc, an attempt to compile the attached file produces an assertion. [hc_ring.txt](https://github.com/RadeonOpenCompute/hcc/files/2231470/hc_ring.txt) Note that the file just contains two #include statements, which must be...

I hit the above assertion when creating a debug build for hcc. Here's the actual output: Building OCL object rocdl/opencl/CMakeFiles/opencl_lib.dir/src/devenq/schedule.cl clang-7: /home/ssahasra/projects/hcc/clang/lib/CodeGen/CGRecordLayoutBuilder.cpp:596: void {anonymous}::CGRecordLowering::clipTailPadding(): Assertion `Prior->Kind == MemberInfo::Field && !...

[AMDGPU][SIInsertWaitcnts] drop OldWaitcntInstr only when it is processed

3

When iterating over instructions, the pass keeps track of pending soft waitcnt instructions. These are usually processed when a non-waitcnt instruction is visited, but some visit may return early. Dropping...