[Mlir-commits] [mlir] [mlir][nvvm] Introduce `setmaxregister.sync.aligned` Op (PR #73780)

[llvmlistbot at llvm.org]

================
@@ -132,3 +132,10 @@ func.func @wgmma_f32_m32(%descA : i64, %descB : i64) {
       -> !llvm.struct<(f32, f32, f32, f32, f32, f32, f32, f32)> 
   return 
 }
+// -----
+
+func.func @set_max_register() {
+  // expected-error @+1 {{new register size must be in between 24 to 256}}
+  nvvm.setmaxregister decrease 8
----------------
durga4github wrote:

Shall we add an another check for %8 also ?
for example,
nvvm.setmaxregister decrease 28

otherwise, change looks good to me.

https://github.com/llvm/llvm-project/pull/73780