[SYCL] Add novalake GPU devices to native bfloat16 target list (#21993)

- [x] Add `nvl_s`, `nvl_u`, `nvl_p` to `GPUArchsWithNBF16` set in
`SYCL.cpp`
- [x] Add `nvl` prefix check to `checkBF` lambda in `SYCL.cpp`
- [x] Add test cases for novalake devices in
`sycl-device-lib-bfloat16.cpp`
- [x] Fix clang-format issue in `GPUArchsWithNBF16` initializer

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: jinge90 <43599496+jinge90@users.noreply.github.com>

Author: Copilot

clang/lib/Driver/ToolChains/SYCL.cpp

@@ -235,8 +235,8 @@ static bool selectBfloatLibs(const llvm::opt::ArgList &Args,
                              const llvm::Triple &Triple, const ToolChain &TC,
                              bool &UseNative) {
   static llvm::SmallSet<StringRef, 8> GPUArchsWithNBF16{
-      "pvc",   "acm_g10", "acm_g11", "acm_g12", "bmg_g21",
-      "lnl_m", "ptl_h",   "ptl_u",   "wcl",     "cri"};
+      "pvc",   "acm_g10", "acm_g11", "acm_g12", "bmg_g21", "lnl_m", "ptl_h",
+      "ptl_u", "wcl",     "cri",     "nvl_s",   "nvl_u",   "nvl_p"};
   bool NeedLibs = false;
 
   // spir64 target is actually JIT compilation, so we defer selection of
@@ -269,7 +269,7 @@ static bool selectBfloatLibs(const llvm::opt::ArgList &Args,
              Device.starts_with("acm") || Device.starts_with("dg2") ||
              Device.starts_with("bmg") || Device.starts_with("lnl") ||
              Device.starts_with("ptl") || Device.starts_with("wcl") ||
-             Device.starts_with("cri");
+             Device.starts_with("cri") || Device.starts_with("nvl");
     };
 
     auto checkSpirvJIT = [](StringRef Target) {

clang/test/Driver/sycl-device-lib-bfloat16.cpp

@@ -61,6 +61,38 @@
 // RUN:   "-device cri" /clang:--sysroot=%S/Inputs/SYCL -- %s 2>&1 \
 // RUN:   | FileCheck %s -check-prefix=BFLOAT16-NATIVE
 
+// Test that a novalake AOT compilation uses the native library.
+// RUN: %clangxx -fsycl -fsycl-targets=spir64_gen -Xsycl-target-backend \
+// RUN:   "-device nvl_s" %s --sysroot=%S/Inputs/SYCL -### 2>&1 \
+// RUN:   | FileCheck %s -check-prefix=BFLOAT16-NATIVE
+// RUN: %clangxx -fsycl -fsycl-targets=intel_gpu_nvl_s %s \
+// RUN:   --sysroot=%S/Inputs/SYCL -### 2>&1 \
+// RUN:   | FileCheck %s -check-prefix=BFLOAT16-NATIVE
+// RUN: %clangxx -fsycl -fsycl-targets=intel_gpu_30_4_0 %s \
+// RUN:   --sysroot=%S/Inputs/SYCL -### 2>&1 \
+// RUN:   | FileCheck %s -check-prefix=BFLOAT16-NATIVE
+// RUN: %clangxx -fsycl -fsycl-targets=spir64_gen -Xsycl-target-backend \
+// RUN:   "-device nvl_u" %s --sysroot=%S/Inputs/SYCL -### 2>&1 \
+// RUN:   | FileCheck %s -check-prefix=BFLOAT16-NATIVE
+// RUN: %clangxx -fsycl -fsycl-targets=intel_gpu_nvl_u %s \
+// RUN:   --sysroot=%S/Inputs/SYCL -### 2>&1 \
+// RUN:   | FileCheck %s -check-prefix=BFLOAT16-NATIVE
+// RUN: %clangxx -fsycl -fsycl-targets=intel_gpu_30_5_0 %s \
+// RUN:   --sysroot=%S/Inputs/SYCL -### 2>&1 \
+// RUN:   | FileCheck %s -check-prefix=BFLOAT16-NATIVE
+// RUN: %clangxx -fsycl -fsycl-targets=spir64_gen -Xsycl-target-backend \
+// RUN:   "-device nvl_p" %s --sysroot=%S/Inputs/SYCL -### 2>&1 \
+// RUN:   | FileCheck %s -check-prefix=BFLOAT16-NATIVE
+// RUN: %clangxx -fsycl -fsycl-targets=intel_gpu_nvl_p %s \
+// RUN:   --sysroot=%S/Inputs/SYCL -### 2>&1 \
+// RUN:   | FileCheck %s -check-prefix=BFLOAT16-NATIVE
+// RUN: %clangxx -fsycl -fsycl-targets=intel_gpu_35_10_0 %s \
+// RUN:   --sysroot=%S/Inputs/SYCL -### 2>&1 \
+// RUN:   | FileCheck %s -check-prefix=BFLOAT16-NATIVE
+// RUN: %clang_cl -### -fsycl -fsycl-targets=spir64_gen -Xsycl-target-backend \
+// RUN:   "-device nvl_s" /clang:--sysroot=%S/Inputs/SYCL -- %s 2>&1 \
+// RUN:   | FileCheck %s -check-prefix=BFLOAT16-NATIVE
+
 // Test that unless all targets support bfloat16, AOT compilation uses the
 // fallback library.
 // RUN: %clangxx -fsycl -fsycl-targets=spir64_gen -Xsycl-target-backend \