PR #32439: [ROCm] Fix and enable xla_gpu_use_embeded_device_lib and xla_gpu_use_…

Imported from GitHub PR #32439

…inprocess_lld

📝 Summary of Changes
Enable embedded device libs and in-process lld by default.

🎯 Justification
Moves amdgpu backend to be more filesystem layout independent.

🚀 Kind of Contribution
🐛 Bug Fix

📊 Benchmark (for Performance Improvements)
N\A

🧪 Unit Tests:
None

🧪 Execution Tests:
None

Copybara import of the project:

--
46a1003 by Dragan Mladjenovic <Dragan.Mladjenovic@amd.com>:

[ROCm] Fix and enable xla_gpu_use_embeded_device_lib and xla_gpu_use_inprocess_lld

Merging this change closes #32439

FUTURE_COPYBARA_INTEGRATE_REVIEW=#32439 from ROCm:enable_bitcode_lld 46a1003
PiperOrigin-RevId: 824375626

Author: draganmladjenovic

xla/BUILD

@@ -4,7 +4,7 @@ load("//third_party/compute_library:build_defs.bzl", "if_enable_acl")
 # copybara:uncomment load("@rules_python//python:proto.bzl", "py_proto_library")
 load("//xla:package_groups.bzl", "xla_package_groups")
 load("//xla:xla.default.bzl", "xla_bzl_library", "xla_cc_test", "xla_py_proto_library")
-load("//xla/tsl:tsl.bzl", "if_google", "internal_visibility")
+load("//xla/tsl:tsl.bzl", "if_google", "if_oss", "internal_visibility")
 load("//xla/tsl:tsl.default.bzl", "filegroup", "get_compatible_with_portable")
 load(
     "//xla/tsl/platform:build_config.bzl",
@@ -1215,7 +1215,12 @@ cc_library(
         "debug_options_parsers.h",
     ],
     hdrs = ["debug_options_flags.h"],
-    copts = if_enable_acl(["-DXLA_CPU_USE_ACL=1"]),
+    copts = if_enable_acl([
+        "-DXLA_CPU_USE_ACL=1",
+    ]) + if_oss([
+        "-DHAS_SUPPORT_FOR_LLD_AS_A_LIBRARY=1",
+        "-DHAS_SUPPORT_FOR_EMBEDDED_LIB_DEVICE=1",
+    ]),
     visibility = internal_visibility([":friends"]),
     deps =
         [

xla/debug_options_flags.cc

@@ -389,6 +389,14 @@ DebugOptions DefaultDebugOptionsIgnoringFlags() {
   const int64_t kDefaultMinGemmRewriteSize = 100;
   opts.set_xla_gpu_gemm_rewrite_size_threshold(kDefaultMinGemmRewriteSize);
 
+#ifdef HAS_SUPPORT_FOR_EMBEDDED_LIB_DEVICE
+  opts.set_xla_gpu_use_embeded_device_lib(true);
+#endif
+
+#ifdef HAS_SUPPORT_FOR_LLD_AS_A_LIBRARY
+  opts.set_xla_gpu_use_inprocess_lld(true);
+#endif
+
   opts.set_xla_gpu_use_memcpy_local_p2p(false);
 
   opts.set_xla_reduce_window_rewrite_base_length(16);

xla/service/gpu/llvm_gpu_backend/generate_amdgpu_device_lib_data_tool.py

@@ -75,7 +75,7 @@ def main():
 
 namespace {cpp_namespace} {{
   inline const char kRaw_{cpp_identifier}[] = "{data_string}";
-  constexpr llvm::StringRef {cpp_identifier}{{kRaw_{cpp_identifier}, sizeof(kRaw_{cpp_identifier})}};
+  constexpr llvm::StringRef {cpp_identifier}{{kRaw_{cpp_identifier}, sizeof(kRaw_{cpp_identifier}) - 1}};
 }} // namespace {cpp_namespace}
 """)