[GEMM] Gemm universal device operation (#1154)

* Optimize GEMM on MI200/300: 1. Add new blockwise gemm pipeline 2. Add irregular splitk intances * clang format + typo fix * Fix a bug * initial commit * Add more instances to irregular splitk * blkgemm pipeline v1~4 prototype * Sanity Checked. Known issue: 1. Poor performance of splitk 2. Register spill on blkgemmpipeline v3 * Sanity and Performance fix: 1. fix a bug related to sanity in grouped b2c mapping 2. fix a bug related to sanity and performance in splitk offset * Sanity and API update: 1. Remove prefetch stage 2. Fix valid check bug 3, Add first gemm_universal instance into ckProfiler * Add NN instances for gemm universal * 1. Add NT instances for gemm_universal 2. Fix a bug about Kpadding in gemm_universal * Fix a bug regarding padding Odd K number * remove kernel print * Fix KPadding bug... * Update safety check * another try to fix kpadding.. * Sanity checked * new instances.. * clang format+typo fix * remove clang format script's change * Add non-hotloop compile option * 1. Add fp16xfp8 example 2. pull packed convert f8 from pr1150 * Some miscs.. opt and fix * Add pipeline description docs * Split universal gemm instance library to cut profiler compiling time * uncomment cmakefile * Fix a bug caused by blockwise_gemm_pipe_v2 * reduce default splitk to 1 * Add 224x256x64 tile size * update, including: 1. Experiment pipeline 5~7 2. Optimization for pipeline 4 3. Organized instance library * temp save * temp save * Permuted lds layout, sanity and function checked * clang format * Move OOB check from RunRead to RunWrite, for better software pipeline. TODO: agpr spill when NN layout * clangformat * A/B splitpipe scheduler for v3 * Fix two bugs * bug fix * fix a bug in oob check * Example for mixed fp16_fp8 gemm * Clean experimental code blocks * Add mixed precision gemm into profiler * tempsave * optimize m/n major lds layout * Add RRR GEMM mixed precision instances * Optimize f8 matrix transpose * Add test_gemm_universal * A/B spilt schedule for blkpip v5 * Take ds_read2 into iglp scheduling scheme * format * fixed cmake * Add llvm-option into CI cmake flag --------- Co-authored-by: Jing Zhang <jizhan@amd.com>
2026-04-19 22:39:03 +00:00 · 2024-04-14 10:03:18 +08:00
parent 7cdf5a96d2
commit f83e9701e9
107 changed files with 10902 additions and 118 deletions
--- a/include/ck/tensor_description/multi_index_transform.hpp
+++ b/include/ck/tensor_description/multi_index_transform.hpp
@@ -1951,4 +1951,89 @@ struct Modulo
        printf("}");
    }
 };
+
+template <typename LowLengths>
+struct Xor
+{
+    using LowerIndex = MultiIndex<2>;
+    using UpperIndex = MultiIndex<2>;
+
+    using UpLengths = LowLengths;
+
+    UpLengths up_lengths_;
+
+    __host__ __device__ constexpr Xor() : up_lengths_{} {}
+
+    __host__ __device__ constexpr Xor(const LowLengths& low_lengths) : up_lengths_{low_lengths} {}
+
+    __host__ __device__ static constexpr index_t GetNumOfLowerDimension() { return 2; }
+
+    __host__ __device__ static constexpr index_t GetNumOfUpperDimension() { return 2; }
+
+    __host__ __device__ constexpr const auto& GetUpperLengths() const { return up_lengths_; }
+
+    template <typename LowIdx, typename UpIdx>
+    __host__ __device__ constexpr void CalculateLowerIndex(LowIdx& idx_low,
+                                                           const UpIdx& idx_up) const
+    {
+        static_assert(LowIdx::Size() == 2 && UpIdx::Size() == 2,
+                      "wrong! inconsistent # of dimension");
+
+        idx_low(Number<0>{}) = idx_up[Number<0>{}];
+
+        idx_low(Number<1>{}) =
+            idx_up[Number<1>{}] ^ (idx_up[Number<0>{}] % up_lengths_[Number<1>{}]);
+    }
+
+    template <typename LowIdxDiff,
+              typename UpIdxDiff,
+              typename LowIdx,
+              typename UpIdx,
+              index_t Hack>
+    __host__ __device__ void UpdateLowerIndex(LowIdxDiff& idx_diff_low,
+                                              const UpIdxDiff&,
+                                              LowIdx& idx_low,
+                                              const UpIdx& idx_up,
+                                              Number<Hack>) const
+    {
+        static_assert(LowIdxDiff::Size() == 2 && UpIdxDiff::Size() == 2 && LowIdx::Size() == 2 &&
+                          UpIdx::Size() == 2,
+                      "wrong! inconsistent # of dimension");
+
+        const auto idx_low_old = idx_low;
+
+        CalculateLowerIndex(idx_low, idx_up);
+
+        idx_diff_low = idx_low - idx_low_old;
+    }
+
+    __host__ __device__ static constexpr bool IsValidUpperIndexAlwaysMappedToValidLowerIndex()
+    {
+        return true;
+    }
+
+    template <typename UpIdx>
+    __host__ __device__ static constexpr bool
+    IsValidUpperIndexMappedToValidLowerIndex(const UpIdx& /* idx_up */)
+    {
+        return true;
+    }
+
+    __host__ __device__ static constexpr bool IsKnownAtCompileTime()
+    {
+        return is_known_at_compile_time<UpLengths>::value;
+    }
+
+    __host__ __device__ void Print() const
+    {
+        printf("Xor{");
+
+        //
+        printf("up_lengths_: ");
+        print(up_lengths_);
+        printf(", ");
+
+        printf("}");
+    }
+};
 } // namespace ck
--- a/include/ck/tensor_description/multi_index_transform_helper.hpp
+++ b/include/ck/tensor_description/multi_index_transform_helper.hpp
@@ -127,4 +127,10 @@ __host__ __device__ constexpr auto make_modulo_transform(const Modulus& modulus,
 {
    return Modulo<Modulus, UpLength>{modulus, up_length};
 }
+
+template <typename LowLengths>
+__host__ __device__ constexpr auto make_xor_transform(const LowLengths& low_lengths)
+{
+    return Xor<LowLengths>{low_lengths};
+}
 } // namespace ck