[llvm] [AMDGPU] Limit DVGPR entry function max VGPRs to block size (PR #226242)

[Diana Picus via llvm-commits]

================
@@ -271,12 +271,49 @@ define amdgpu_kernel void @i32_16_elements_attrib(ptr %out) #2 {
   ret void
 }
 
+define amdgpu_cs void @i32_16_elements_dvgpr(ptr %out) #3 {
----------------
rovka wrote:

Can we also have a test where we manage to promote something even if we're in dynamic VGPR mode?

https://github.com/llvm/llvm-project/pull/226242