[SYCL][ESIMD] Fix incorrect intrinsic lowering for signed atomic min/max (#8527)

The correct intrinsic examples are:
@llvm.genx.svm.atomic.imin
@llvm.genx.svm.atomic.imax

Author: turinevgeny

llvm/lib/SYCLLowerIR/ESIMD/LowerESIMD.cpp

@@ -832,10 +832,10 @@ static std::string getESIMDIntrinSuffix(id::FunctionEncoding *FE,
       Suff = ".xor";
       break;
     case 0xb:
-      Suff = ".minsint";
+      Suff = ".imin";
       break;
     case 0xc:
-      Suff = ".maxsint";
+      Suff = ".imax";
       break;
     case 0x10:
       Suff = ".fmax";

sycl/test/esimd/intrins_trans.cpp

@@ -259,6 +259,20 @@ test_mem_intrins(uint64_t addr, const vec<float, 8> &xf,
     // CHECK-LABEL: %{{[a-zA-Z0-9.]+}} = call <8 x float> @llvm.genx.svm.atomic.fcmpwr.v8f32.v8i1.v8i64(<8 x i1> %{{[a-zA-Z0-9.]+}}, <8 x i64> %{{[a-zA-Z0-9.]+}}, <8 x float> %{{[a-zA-Z0-9.]+}}, <8 x float> %{{[a-zA-Z0-9.]+}}, <8 x float> undef)
     use(x);
   }
+  {
+    vec<int, 8> src0 = get8i();
+    auto x = __esimd_svm_atomic1<atomic_op::smin, int, 8>(get8ui64(), src0,
+                                                          get8ui16());
+    // CHECK-LABEL: %{{[a-zA-Z0-9.]+}} = call <8 x i32> @llvm.genx.svm.atomic.imin.v8i32.v8i1.v8i64(<8 x i1> %{{[a-zA-Z0-9.]+}}, <8 x i64> %{{[a-zA-Z0-9.]+}}, <8 x i32> %{{[a-zA-Z0-9.]+}}, <8 x i32> undef)
+    use(x);
+  }
+  {
+    vec<int, 8> src0 = get8i();
+    auto x = __esimd_svm_atomic1<atomic_op::smax, int, 8>(get8ui64(), src0,
+                                                          get8ui16());
+    // CHECK-LABEL: %{{[a-zA-Z0-9.]+}} = call <8 x i32> @llvm.genx.svm.atomic.imax.v8i32.v8i1.v8i64(<8 x i1> %{{[a-zA-Z0-9.]+}}, <8 x i64> %{{[a-zA-Z0-9.]+}}, <8 x i32> %{{[a-zA-Z0-9.]+}}, <8 x i32> undef)
+    use(x);
+  }
   {
     constexpr SurfaceIndex si = 0;
     vec<float, 8> x =