[rocm-libraries] ROCm/rocm-libraries#5323 (commit 5454e9e)

CK Tile MX GEMM Packing Improvement

## Motivation

Reduce the scale loading size and also has better utilization of MFMA
scale selection.

## Technical Details

Add up the packing of mx scales.

## Test Plan

Use the existing test cases.

## Test Result

<!-- Briefly summarize test outcomes. -->

## Submission Checklist

- [ ] Look over the contributing guidelines at
https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.

include/ck_tile/ops/gemm_quant/pipeline/gemm_abquant_pipeline_ag_bg_cr_eight_waves.hpp

@@ -134,7 +134,12 @@ struct ABQuantGemmPipelineAgBgCrEightWaves : public BaseGemmPipelineAgBgCrCompV3
 
     CK_TILE_HOST_DEVICE static constexpr index_t GetSmemSize()
     {
-        return Policy::template GetSmemSize<Problem>();
+        // We are not storing the original packed type in LDS, so we need to multiply the smem size
+        // by the packed size.
+        constexpr index_t smem_size_a = Policy::template GetSmemSizeA<Problem>() * APackedSize;
+        constexpr index_t smem_size_b = Policy::template GetSmemSizeB<Problem>() * BPackedSize;
+
+        return 2 * (smem_size_a + smem_size_b);
     }
 
     CK_TILE_HOST static std::string Print() { return "ABQuantGemmPipelineAgBgCrEightWaves\n"; }