[llvm] [AMDGPU] Fix maximum physical LDS reporting for gfx6 (PR #223185)

[Matt Arsenault via llvm-commits]

================
@@ -3290,7 +3299,15 @@ TEST(TargetParserTest, testAMDGPUgetLocalMemorySize) {
 
 TEST(TargetParserTest, testAMDGPUgetAddressableLocalMemorySize) {
   // A work-group never allocates past the hardware cap, so the doubled
-  // gfx10/11/12 block is capped back to the addressable size.
+  // gfx6 and gfx10/11/12 blocks are capped back to the addressable size.
+  for (Triple::SubArchType SubArch :
+       {Triple::AMDGPUSubArch600, Triple::AMDGPUSubArch601,
+        Triple::AMDGPUSubArch602}) {
+    SCOPED_TRACE(AMDGPU::getArchNameFromSubArch(SubArch));
----------------
arsenm wrote:

I'm not sure what SCOPED_TRACE does 

https://github.com/llvm/llvm-project/pull/223185