[SYCL][ESIMD] Fix several issues related to building of ESIMD on Wind…

…ows (#6971)
intel · Oct 18, 2022 · 77e92ce · 77e92ce
1 parent be0905f
commit 77e92ce
Show file tree

Hide file tree

Showing 3 changed files with 30 additions and 27 deletions.
diff --git a/sycl/include/sycl/ext/intel/experimental/esimd/detail/math_intrin.hpp b/sycl/include/sycl/ext/intel/experimental/esimd/detail/math_intrin.hpp
@@ -551,11 +551,11 @@ __esimd_dpas_inner(const __ESIMD_DNS::vector_type_t<T0, SZ> *src0,
 
   __ESIMD_DNS::vector_type_t<TmpAccEl, SIMDSize> simdAcc;
 
-  for (uint r = 0; r < repeat_count; r++) {
+  for (unsigned r = 0; r < repeat_count; r++) {
     V = r;
     k = 0;
 
-    for (uint n = 0; n < SIMDSize; n++) {
+    for (unsigned n = 0; n < SIMDSize; n++) {
       if (src0 != nullptr) {
         auto src0El = src0[0][r * SIMDSize + n];
 
@@ -570,13 +570,13 @@ __esimd_dpas_inner(const __ESIMD_DNS::vector_type_t<T0, SZ> *src0,
         simdAcc[n] = 0;
     }
 
-    for (uint s = 0; s < systolic_depth; s++) {
+    for (unsigned s = 0; s < systolic_depth; s++) {
       src1_ops_per_dword = 32 / (ops_per_chan * src1_el_bits);
       // U = s / src1_ops_per_dword;
-      U = s >> uint(log2(src1_ops_per_dword));
+      U = s >> unsigned(log2(src1_ops_per_dword));
 
-      for (uint n = 0; n < SIMDSize; n++) {
-        for (uint d = 0; d < ops_per_chan; d++) {
+      for (unsigned n = 0; n < SIMDSize; n++) {
+        for (unsigned d = 0; d < ops_per_chan; d++) {
           p = d + (s % src1_ops_per_dword) * ops_per_chan;
           uint32_t extension_temp = false;
 
@@ -618,7 +618,7 @@ __esimd_dpas_inner(const __ESIMD_DNS::vector_type_t<T0, SZ> *src0,
 
     } // Systolic phase.
 
-    for (uint n = 0; n < SIMDSize; n++) {
+    for (unsigned n = 0; n < SIMDSize; n++) {
       if constexpr (pvcBfDest) {
         // TODO: make abstraction, support saturation, review rounding algo for
         // corner cases.

diff --git a/sycl/include/sycl/ext/intel/experimental/esimd/math.hpp b/sycl/include/sycl/ext/intel/experimental/esimd/math.hpp
@@ -988,8 +988,8 @@ template <typename RT, typename T0, int SZ,
 __ESIMD_API __ESIMD_NS::simd<RT, SZ> lzd(__ESIMD_NS::simd<T0, SZ> src0,
                                          Sat sat = {}) {
   // Saturation parameter ignored
-  __ESIMD_NS::simd<uint, SZ> Src0 = src0;
-  return __esimd_lzd<uint>(Src0.data());
+  __ESIMD_NS::simd<__ESIMD_NS::uint, SZ> Src0 = src0;
+  return __esimd_lzd<__ESIMD_NS::uint, SZ>(Src0.data());
 }
 
 template <typename RT, typename T0, class Sat = __ESIMD_NS::saturation_off_tag>

diff --git a/sycl/test/esimd/regression/windows_build_test.cpp b/sycl/test/esimd/regression/windows_build_test.cpp
@@ -6,36 +6,39 @@
 //
 //===----------------------------------------------------------------------===//
 // REQUIRES: windows
-// RUN: %clangxx -fsycl -fsycl-device-only -fsyntax-only -Xclang -verify %s -I %sycl_include
+// RUN: %clangxx -fsycl -fsyntax-only -Xclang -verify %s -I %sycl_include
 // expected-no-diagnostics
 
 // The tests validates an ability to build ESIMD code on windows platform
 
-#include <iostream>
 #include <CL/sycl.hpp>
+#include <iostream>
 #include <sycl/ext/intel/esimd.hpp>
-#include <sycl/ext/intel/experimental/esimd/memory.hpp>
 
 class Kernel;
 
-int main()
-{
+int main() {
   sycl::queue q;
   sycl::device dev = q.get_device();
   sycl::context ctx = q.get_context();
-  std::cout << "Device: " << dev.get_info<sycl::info::device::name>() << std::endl;
-
-  int* buffer = (int*)sycl::aligned_alloc_device(128, 1024, q);
-
-  q.parallel_for<Kernel>(
-    1,
-    [=](sycl::item<1> it) SYCL_ESIMD_KERNEL {
-      using namespace sycl::ext::intel::esimd;
-      using namespace sycl::ext::intel::experimental::esimd;
-
-      simd<int, 32> blk;
-      lsc_block_store<int, 32>(buffer, blk);
-    });
+  std::cout << "Device: " << dev.get_info<sycl::info::device::name>()
+            << std::endl;
+
+  int *buffer = (int *)sycl::aligned_alloc_device(128, 1024, q);
+
+  q.parallel_for<Kernel>(1, [=](sycl::item<1> it) SYCL_ESIMD_KERNEL {
+    using namespace sycl::ext::intel::esimd;
+    using namespace sycl::ext::intel::experimental::esimd;
+
+    simd<int, 32> blk;
+    simd<sycl::ext::oneapi::experimental::bfloat16, 16> A;
+    simd<sycl::ext::oneapi::experimental::bfloat16, 256> B;
+    simd<float, 16> C;
+    lzd<uint>(blk);
+    lzd<uint>(35);
+    sycl::ext::intel::esimd::xmx::dpas<8, 1, float>(C, B, A);
+    lsc_block_store<int, 32>(buffer, blk);
+  });
 
   q.wait();
   sycl::free(buffer, q);