[SYCL][Matrix] Add a more general query example

SYCL/Matrix/joint_matrix_query_default_tpu_sel.cpp

@@ -0,0 +1,264 @@
+//==---- joint_matrix_query_default_tpu_sel.cpp  - DPC++ joint_matrix-------==//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+// REQUIRES: matrix
+
+// RUN: %clangxx -fsycl %s -o %t.out -DSYCL_EXT_ONEAPI_MATRIX_VERSION=4
+// RUN: %CPU_RUN_PLACEHOLDER %t.out
+// RUN: %GPU_RUN_PLACEHOLDER %t.out
+
+#include <iostream>
+#include <sycl/sycl.hpp>
+
+using namespace sycl;
+using namespace sycl::ext::oneapi::experimental::matrix;
+
+template <typename T, size_t NUM_ROWS, size_t NUM_COLS> struct big_matrix {
+public:
+  T *mat;
+
+public:
+  T *get_data() { return mat; }
+  void set_data(T *data) { mat = data; }
+  big_matrix(T *data) : mat(data) {}
+};
+
+template <typename T>
+void matrix_vnni(unsigned int rows, unsigned int cols, T *src, T *dest,
+                 unsigned int vnniFactor) {
+  for (unsigned int i = 0; i < rows / vnniFactor; i++) {
+    for (unsigned int j = 0; j < cols; j++) {
+      for (unsigned int k = 0; k < vnniFactor; k++) {
+        dest[i * cols * vnniFactor + j * vnniFactor + k] =
+            src[(i * vnniFactor + k) * cols + j];
+      }
+    }
+  }
+}
+static inline void native_cpuid(unsigned int *eax, unsigned int *ebx,
+                                unsigned int *ecx, unsigned int *edx) {
+  asm volatile("cpuid"
+               : "=a"(*eax), "=b"(*ebx), "=c"(*ecx), "=d"(*edx)
+               : "0"(*eax), "2"(*ecx));
+}
+
+template <typename T1, typename T2, size_t M, size_t N, size_t K, size_t TM,
+          size_t TN, size_t TK>
+void matrix_multiply(queue q, big_matrix<T1, M, N> &C, big_matrix<T2, M, K> &A,
+                     big_matrix<T2, K / 4, N * 4> &B) {
+  constexpr size_t SG_SZ = TN;
+  size_t NDRangeM = M / TM;
+  size_t NDRangeN = N / TN;
+  buffer<int8_t, 2> bufA(A.get_data(), range<2>(M, K));
+  buffer<int8_t, 2> bufB(B.get_data(), range<2>(K, N));
+  buffer<int32_t, 2> bufC(C.get_data(), range<2>(M, N));
+
+  q.submit([&](handler &cgh) {
+     auto accC = bufC.get_access<access::mode::read_write>(cgh);
+     auto accA = bufA.get_access<access::mode::read_write>(cgh);
+     auto accB = bufB.get_access<access::mode::read_write>(cgh);
+
+     cgh.parallel_for(
+         nd_range<2>({NDRangeM, NDRangeN * SG_SZ}, {1, 1 * SG_SZ}),
+         [=](nd_item<2> spmd_item) [[intel::reqd_sub_group_size(SG_SZ)]]
+
+         {
+           const auto global_idx = spmd_item.get_global_id(0);
+           const auto global_idy = spmd_item.get_global_id(1);
+           const auto sg_startx = global_idx - spmd_item.get_local_id(0);
+           const auto sg_starty = global_idy - spmd_item.get_local_id(1);
+
+           sub_group sg = spmd_item.get_sub_group();
+
+           joint_matrix<sub_group, T2, use::a, TM, TK, layout::row_major> sub_a;
+           joint_matrix<sub_group, T2, use::b, TK, TN,
+                        sycl::ext::intel::experimental::matrix::layout::packed>
+               sub_b;
+           joint_matrix<sub_group, T1, use::accumulator, TM, TN> sub_c;
+
+           joint_matrix_load(sg, sub_c,
+                             accC.get_pointer() + (sg_startx * TM) * N +
+                                 sg_starty / SG_SZ * TN,
+                             N, layout::row_major);
+           for (int k = 0; k < K / TK; k += 1) {
+             joint_matrix_load(
+                 sg, sub_a, accA.get_pointer() + (sg_startx * TM) * K + k * TK,
+                 K);
+             joint_matrix_load(sg, sub_b,
+                               accB.get_pointer() + (k * TK / 4) * (N * 4) +
+                                   sg_starty / SG_SZ * TN * 4,
+                               N * 4);
+             sub_c = joint_matrix_mad(sg, sub_a, sub_b, sub_c);
+           }
+           joint_matrix_store(sg, sub_c,
+                              accC.get_pointer() + (sg_startx * TM) * N +
+                                  sg_starty / SG_SZ * TN,
+                              N, layout::row_major);
+         }); // parallel for
+   }).wait();
+}
+
+bool is_this_amx_device(queue q) {
+  auto dev_type =
+      q.get_device().template get_info<sycl::info::device::device_type>();
+  if (dev_type == info::device_type::cpu) {
+    std::cout << "This is a CPU Device" << std::endl;
+    unsigned eax, ebx, ecx, edx;
+
+    eax = 1; /* processor info and feature bits */
+    native_cpuid(&eax, &ebx, &ecx, &edx);
+    int model = (eax >> 4) & 0xF;
+    int family = (eax >> 8) & 0xF;
+    int extended_model = (eax >> 16) & 0xF;
+    int extended_family = (eax >> 20) & 0xFF;
+
+    // SPR processsor means AMX is present
+    if (model == 15 && family == 6 && extended_model == 8 &&
+        extended_family == 0) {
+      std::cout << "This is a SPR CPU Device" << std::endl;
+      return true;
+    }
+  }
+  return false;
+}
+
+bool is_this_xmx16_device(queue q) {
+  auto dev_type =
+      q.get_device().template get_info<sycl::info::device::device_type>();
+  if (dev_type == info::device_type::gpu) {
+    std::cout << "This is a GPU Device" << std::endl;
+
+    auto dev_name =
+        q.get_device().template get_info<sycl::info::device::name>();
+    size_t found = dev_name.find("Intel(R) Graphics [0x0bd5]"); // PVC
+    if (found != std::string::npos) {
+      std::cout << "This is a PVC Device which contains XMX-16 TPU"
+                << std::endl;
+      return true;
+    }
+  }
+  return false;
+}
+
+bool is_this_xmx8_device(queue q) {
+  auto dev_type =
+      q.get_device().template get_info<sycl::info::device::device_type>();
+  if (dev_type == info::device_type::gpu) {
+    std::cout << "This is a GPU Device" << std::endl;
+    auto dev_name =
+        q.get_device().template get_info<sycl::info::device::name>();
+    size_t found = dev_name.find("Intel(R) Graphics [0x5610]"); // ATS-P
+    if (found != std::string::npos) {
+      std::cout << "This is a ATS-M Device which contains XMX-8 TPU"
+                << std::endl;
+      return true;
+    }
+  }
+  return false;
+}
+
+template <typename T1, typename T2, size_t M, size_t N, size_t K>
+bool query_and_matrix_multiply(big_matrix<T1, M, N> &C, big_matrix<T2, M, K> &A,
+                               big_matrix<T2, K / 4, N * 4> &B) {
+  queue q;
+
+  if (is_this_amx_device(q)) {
+    using myparams2 = tpu_params<tpu::amx, int8_t, int8_t, int>;
+    constexpr int TM = myparams2::M;
+    constexpr int TN = myparams2::N;
+    constexpr int TK = myparams2::K;
+    std::cout << "AMX query default sizes are: M " << TM << " N " << TN << " K "
+              << TK << std::endl;
+    matrix_multiply<T1, T2, M, N, K, TM, TN, TK>(q, C, A, B);
+    return true;
+  }
+
+  if (is_this_xmx16_device(q)) {
+    using myparams2 = tpu_params<tpu::xmx16, int8_t, int8_t, int>;
+    constexpr int TM = myparams2::M;
+    constexpr int TN = myparams2::N;
+    constexpr int TK = myparams2::K;
+    std::cout << "XMX of PVC query default sizes are: M " << TM << " N " << TN
+              << " K " << TK << std::endl;
+    matrix_multiply<T1, T2, M, N, K, TM, TN, TK>(q, C, A, B);
+    return true;
+  }
+
+  if (is_this_xmx8_device(q)) {
+    using myparams2 = tpu_params<tpu::xmx8, int8_t, int8_t, int>;
+    constexpr int TM = myparams2::M;
+    constexpr int TN = myparams2::N;
+    constexpr int TK = myparams2::K;
+    std::cout << "XMX of ATS-M query default sizes are: M " << TM << " N " << TN
+              << " K " << TK << std::endl;
+    matrix_multiply<T1, T2, M, N, K, TM, TN, TK>(q, C, A, B);
+    return true;
+  }
+  return false;
+}
+
+void matrix_multiply_ref(int8_t *A, int8_t *B, int32_t *C, int M, int N,
+                         int K) {
+  for (int m = 0; m < M; m++)
+    for (int n = 0; n < N; n++) {
+      for (int k = 0; k < K; k++) {
+        C[m * N + n] += A[m * K + k] * B[k * N + n];
+      }
+    }
+}
+
+int main() {
+  static constexpr size_t MATRIX_M = 128;
+  static constexpr size_t MATRIX_N = 128;
+  static constexpr size_t MATRIX_K = 128;
+  static constexpr int vnniFactor = 4;
+  int8_t A[MATRIX_M][MATRIX_K];
+  int8_t B[MATRIX_K][MATRIX_N];
+  int8_t Bvnni[MATRIX_K / 4][MATRIX_N * 4];
+  int32_t C[MATRIX_M][MATRIX_N];
+  int32_t D[MATRIX_M][MATRIX_N];
+
+  for (int i = 0; i < MATRIX_M; i++) {
+    for (int j = 0; j < MATRIX_K; j++) {
+      A[i][j] = i + 2 * j;
+    }
+  }
+  for (int i = 0; i < MATRIX_K; i++) {
+    for (int j = 0; j < MATRIX_N; j++) {
+      B[i][j] = i + j;
+    }
+  }
+  for (int i = 0; i < MATRIX_M; i++) {
+    for (int j = 0; j < MATRIX_N; j++) {
+      C[i][j] = 1;
+      D[i][j] = 1;
+    }
+  }
+
+  big_matrix<int32_t, MATRIX_M, MATRIX_N> MC((int32_t *)&C);
+  big_matrix<int32_t, MATRIX_M, MATRIX_N> MD((int32_t *)&D);
+  big_matrix<int8_t, MATRIX_M, MATRIX_K> MA((int8_t *)&A);
+  matrix_vnni<int8_t>(MATRIX_K, MATRIX_N, (int8_t *)&B, (int8_t *)&Bvnni, 4);
+  big_matrix<int8_t, MATRIX_K / 4, MATRIX_N * 4> MBvnni((int8_t *)&Bvnni);
+
+  bool sel = query_and_matrix_multiply(MC, MA, MBvnni);
+  bool res = true;
+  if (sel) {
+    matrix_multiply_ref((int8_t *)&A, (int8_t *)&B, (int32_t *)&D, MATRIX_M,
+                        MATRIX_N, MATRIX_K);
+
+    for (int i = 0; i < MATRIX_M; i++) {
+      for (int j = 0; j < MATRIX_N; j++) {
+        if (C[i][j] != D[i][j])
+          res = false;
+      }
+    }
+    std::cout << (res ? "passed" : "failed") << std::endl;
+  } else
+    std::cout << "default device selector has no TPU" << std::endl;
+  return !res;
+}