[llvm-branch-commits] [mlir] [mlir][AMDGPU] Update gather_to_lds with explicit-async support (PR #181082)

[Zhuoran Yin via llvm-branch-commits]

================
@@ -1099,14 +1100,21 @@ def AMDGPU_GatherToLDSOp :
     * `$transferType`: type of the data to be transferred by each thread. This is used to determine
       the size of the data to be transferred and the number of threads in the subgroup.
       The transfer type must be a scalar type or a vector type with a single element type.
+    * If `$async` is set, the compiler will not attempt to infer the
+      memory waits needed to ensure that the DMA operation has succeeded
+      before a load that might access the stored-to LDS is performed.
+      Instead, the `rocdl.asyncmark` and `rocdl.wait.asyncmark N`
+      operations must be used to explicitly indicate the desired completion
+      behavior. This enables more precise calculation of these waits at the
+      cost of requiring user management of asynchrony.
 
     The `$dst`, along with its indices, points to the memory location the subgroup of this thread
     will write to.
 
     Note: only supported on gfx9 and gfx10.
   }];
   let assemblyFormat = [{
-    $src `[` $srcIndices `]` `,` $dst `[` $dstIndices `]` attr-dict `:` $transferType `,` type($src) `,` type($dst)
+    (`async` $async^)? $src `[` $srcIndices `]` `,` $dst `[` $dstIndices `]` attr-dict `:` $transferType `,` type($src) `,` type($dst)
----------------
jerryyin wrote:

I almost forget what "rule" is when deciding whether or not everything in rocdl deserve a amdgpu counterpart. From my impression, things such as `amdgpu.sched_barrier` is a completely trivial wrapper of the rocdl version of that. So I was assuming we'd always get one for the ergonomic value - That a client can choose to interact with just amdgpu but not a mixture of amdgpu and rocdl from a more "high level" dialect.

https://github.com/llvm/llvm-project/pull/181082