[SYCL] Enable standard optimization pipeline for device code by default

bader · bader · commit 0909cfb7bf61 · 2020-08-04T21:41:30.000+03:00
Add new compiler flag to disable optimizations at compile time:
 -fno-sycl-std-optimizations

 Temporally disable optimizations in some LIT tests.
diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td
@@ -3530,6 +3530,7 @@ def fsycl_esimd : Flag<["-"], "fsycl-explicit-simd">, Group<sycl_Group>, Flags<[
   HelpText<"Enable SYCL explicit SIMD extension">;
 def fno_sycl_esimd : Flag<["-"], "fno-sycl-explicit-simd">, Group<sycl_Group>,
   HelpText<"Disable SYCL explicit SIMD extension">, Flags<[NoArgumentUnused, CoreOption]>;
+defm sycl_std_optimizations : OptOutFFlag<"sycl-std-optimizations", "Enable", "Disable", " standard optimization pipeline for SYCL device compiler">;
 
 //===----------------------------------------------------------------------===//
 // CC1 Options
@@ -4454,8 +4455,6 @@ def fsycl_std_layout_kernel_params: Flag<["-"], "fsycl-std-layout-kernel-params"
 def fsycl_allow_func_ptr : Flag<["-"], "fsycl-allow-func-ptr">,
   HelpText<"Allow function pointers in SYCL device.">;
 def fno_sycl_allow_func_ptr : Flag<["-"], "fno-sycl-allow-func-ptr">;
-def fsycl_enable_optimizations: Flag<["-"], "fsycl-enable-optimizations">,
-  HelpText<"Experimental flag enabling standard optimization in the front-end.">;
 
 } // let Flags = [CC1Option]
 
diff --git a/clang/lib/Driver/ToolChains/Clang.cpp b/clang/lib/Driver/ToolChains/Clang.cpp
@@ -4102,6 +4102,11 @@ void Clang::ConstructJob(Compilation &C, const JobAction &JA,
     if (Args.hasFlag(options::OPT_fsycl_esimd, options::OPT_fno_sycl_esimd,
                      false))
       CmdArgs.push_back("-fsycl-explicit-simd");
+
+    if (!Args.hasFlag(options::OPT_fsycl_std_optimizations,
+                      options::OPT_fno_sycl_std_optimizations, true))
+      CmdArgs.push_back("-fno-sycl-std-optimizations");
+
     // Pass the triple of host when doing SYCL
     auto AuxT = llvm::Triple(llvm::sys::getProcessTriple());
     std::string NormalizedTriple = AuxT.normalize();
diff --git a/clang/lib/Frontend/CompilerInvocation.cpp b/clang/lib/Frontend/CompilerInvocation.cpp
@@ -823,7 +823,7 @@ static bool ParseCodeGenArgs(CodeGenOptions &Opts, ArgList &Args, InputKind IK,
   Opts.DisableLLVMPasses =
       Args.hasArg(OPT_disable_llvm_passes) ||
       (Args.hasArg(OPT_fsycl_is_device) && Triple.isSPIR() &&
-       !Args.hasArg(OPT_fsycl_enable_optimizations) && !IsSyclESIMD);
+       Args.hasArg(OPT_fno_sycl_std_optimizations) && !IsSyclESIMD);
   Opts.DisableLifetimeMarkers = Args.hasArg(OPT_disable_lifetimemarkers);
 
   const llvm::Triple::ArchType DebugEntryValueArchs[] = {
diff --git a/clang/test/CodeGenSYCL/inline_asm.cpp b/clang/test/CodeGenSYCL/inline_asm.cpp
@@ -1,4 +1,4 @@
-// RUN: %clang_cc1 -fsycl -fsycl-is-device -fsycl-enable-optimizations -triple spir64-unknown-unknown-sycldevice -emit-llvm -x c++ %s -o - | FileCheck %s
+// RUN: %clang_cc1 -fsycl -fsycl-is-device -triple spir64-unknown-unknown-sycldevice -emit-llvm -x c++ %s -o - | FileCheck %s
 
 class kernel;
 
diff --git a/clang/test/CodeGenSYCL/remove-ur-inst.cpp b/clang/test/CodeGenSYCL/remove-ur-inst.cpp
@@ -1,5 +1,5 @@
+// RUN: %clang_cc1 -fsycl -fsycl-is-device -fno-sycl-std-optimizations -triple spir64-unknown-unknown-sycldevice -emit-llvm %s -o - | FileCheck %s
 // RUN: %clang_cc1 -fsycl -fsycl-is-device -triple spir64-unknown-unknown-sycldevice -emit-llvm %s -o - | FileCheck %s
-// RUN: %clang_cc1 -fsycl -fsycl-is-device -fsycl-enable-optimizations -triple spir64-unknown-unknown-sycldevice -emit-llvm %s -o - | FileCheck %s
 
 SYCL_EXTERNAL void doesNotReturn() throw() __attribute__((__noreturn__));
 
diff --git a/sycl/test/basic_tests/accessor/accessor.cpp b/sycl/test/basic_tests/accessor/accessor.cpp
@@ -1,4 +1,4 @@
-// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
+// RUN: %clangxx -fsycl -fno-sycl-std-optimizations -fsycl-targets=%sycl_triple %s -o %t.out
 // RUN: env SYCL_DEVICE_TYPE=HOST %t.out
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
diff --git a/sycl/test/basic_tests/boolean.cpp b/sycl/test/basic_tests/boolean.cpp
@@ -1,4 +1,4 @@
-// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
+// RUN: %clangxx -fsycl -fno-sycl-std-optimizations -fsycl-targets=%sycl_triple %s -o %t.out
 // RUN: env SYCL_DEVICE_TYPE=HOST %t.out
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
diff --git a/sycl/test/basic_tests/stream/stream.cpp b/sycl/test/basic_tests/stream/stream.cpp
@@ -1,4 +1,4 @@
-// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
+// RUN: %clangxx -fsycl -fno-sycl-std-optimizations -fsycl-targets=%sycl_triple %s -o %t.out
 // RUN: env SYCL_DEVICE_TYPE=HOST %t.out | FileCheck %s
 // RUN: %CPU_RUN_PLACEHOLDER %t.out %CPU_CHECK_PLACEHOLDER
 // RUN: %GPU_RUN_ON_LINUX_PLACEHOLDER %t.out %GPU_CHECK_ON_LINUX_PLACEHOLDER
diff --git a/sycl/test/check_device_code/fpga_ihs_float.cpp b/sycl/test/check_device_code/fpga_ihs_float.cpp
@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 
 // RUN: %clangxx -I %sycl_include -S -emit-llvm -fsycl -fsycl-device-only %s -o - | FileCheck %s
-// RUN: %clangxx -I %sycl_include -S -emit-llvm -fsycl -fsycl-device-only %s -Xclang -fsycl-enable-optimizations -o - | FileCheck %s
+// RUN: %clangxx -I %sycl_include -S -emit-llvm -fsycl -fno-sycl-std-optimizations -fsycl-device-only %s -o - | FileCheck %s
 
 #include "CL/__spirv/spirv_ops.hpp"
 
diff --git a/sycl/test/fpga_tests/fpga_lsu.cpp b/sycl/test/fpga_tests/fpga_lsu.cpp
@@ -1,4 +1,4 @@
-// RUN: %clangxx -fsycl %s -o %t.out
+// RUN: %clangxx -fsycl -fno-sycl-std-optimizations %s -o %t.out
 // RUNx: %ACC_RUN_PLACEHOLDER %t.out
 //==----------------- fpga_lsu.cpp - SYCL FPGA LSU test --------------------==//
 //
diff --git a/sycl/test/hier_par/hier_par_wgscope.cpp b/sycl/test/hier_par/hier_par_wgscope.cpp
@@ -6,7 +6,7 @@
 //
 //===----------------------------------------------------------------------===//
 
-// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
+// RUN: %clangxx -fsycl -fno-sycl-std-optimizations -fsycl-targets=%sycl_triple %s -o %t.out
 // RUN: env SYCL_DEVICE_TYPE=HOST %t.out
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
diff --git a/sycl/test/regression/group.cpp b/sycl/test/regression/group.cpp
@@ -1,4 +1,4 @@
-// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
+// RUN: %clangxx -fsycl -fno-sycl-std-optimizations -fsycl-targets=%sycl_triple %s -o %t.out
 // RUN: env SYCL_DEVICE_TYPE=HOST %t.out
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
diff --git a/sycl/test/spec_const/spec_const_hw.cpp b/sycl/test/spec_const/spec_const_hw.cpp
@@ -1,6 +1,6 @@
 // UNSUPPORTED: cuda || level_zero
 //
-// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
+// RUN: %clangxx -fsycl -fno-sycl-std-optimizations -fsycl-targets=%sycl_triple %s -o %t.out
 // RUN: env SYCL_DEVICE_TYPE=HOST %t.out
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
diff --git a/sycl/test/sub_group/generic_reduce.cpp b/sycl/test/sub_group/generic_reduce.cpp
@@ -1,8 +1,8 @@
 // UNSUPPORTED: cuda
 // CUDA compilation and runtime do not yet support sub-groups.
 //
-// RUN: %clangxx -fsycl -fsycl-unnamed-lambda -std=c++14 %s -o %t.out
-// RUN: %clangxx -fsycl -fsycl-unnamed-lambda -fsycl-targets=%sycl_triple -std=c++14 -D SG_GPU %s -o %t_gpu.out
+// RUN: %clangxx -fsycl -fno-sycl-std-optimizations -fsycl-unnamed-lambda -std=c++14 %s -o %t.out
+// RUN: %clangxx -fsycl -fno-sycl-std-optimizations -fsycl-unnamed-lambda -fsycl-targets=%sycl_triple -std=c++14 -D SG_GPU %s -o %t_gpu.out
 // RUN: env SYCL_DEVICE_TYPE=HOST %t.out
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t_gpu.out
diff --git a/sycl/test/sub_group/load_store.cpp b/sycl/test/sub_group/load_store.cpp
@@ -1,7 +1,7 @@
 // UNSUPPORTED: cuda
 // CUDA compilation and runtime do not yet support sub-groups.
 //
-// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
+// RUN: %clangxx -fsycl -fno-sycl-std-optimizations -fsycl-targets=%sycl_triple %s -o %t.out
 // RUN: env SYCL_DEVICE_TYPE=HOST %t.out
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
diff --git a/sycl/test/sub_group/scan_fp16.cpp b/sycl/test/sub_group/scan_fp16.cpp
@@ -1,7 +1,7 @@
 // UNSUPPORTED: cuda
 // CUDA compilation and runtime do not yet support sub-groups.
 //
-// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
+// RUN: %clangxx -fsycl -fno-sycl-std-optimizations -fsycl-targets=%sycl_triple %s -o %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 
 //==--------------- scan_fp16.cpp - SYCL sub_group scan test --------*- C++ -*---==//
diff --git a/sycl/test/sub_group/shuffle.cpp b/sycl/test/sub_group/shuffle.cpp
@@ -1,7 +1,7 @@
 // UNSUPPORTED: cuda
 // CUDA compilation and runtime do not yet support sub-groups.
 //
-// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
+// RUN: %clangxx -fsycl -fno-sycl-std-optimizations -fsycl-targets=%sycl_triple %s -o %t.out
 // RUN: env SYCL_DEVICE_TYPE=HOST %t.out
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
diff --git a/sycl/test/sub_group/shuffle_fp16.cpp b/sycl/test/sub_group/shuffle_fp16.cpp
@@ -1,7 +1,7 @@
 // UNSUPPORTED: cuda
 // CUDA compilation and runtime do not yet support sub-groups.
 //
-// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
+// RUN: %clangxx -fsycl -fno-sycl-std-optimizations -fsycl-targets=%sycl_triple %s -o %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 //
 //==------------ shuffle_fp16.cpp - SYCL sub_group shuffle test -----*- C++ -*---==//
diff --git a/sycl/test/sub_group/shuffle_fp64.cpp b/sycl/test/sub_group/shuffle_fp64.cpp
@@ -1,7 +1,7 @@
 // UNSUPPORTED: cuda
 // CUDA compilation and runtime do not yet support sub-groups.
 //
-// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out
+// RUN: %clangxx -fsycl -fno-sycl-std-optimizations -fsycl-targets=%sycl_triple %s -o %t.out
 // RUN: env SYCL_DEVICE_TYPE=HOST %t.out
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out

Original file line number	Diff line number	Diff line change
`@@ -1,4 +1,4 @@`
`1`		`-// RUN: %clang_cc1 -fsycl -fsycl-is-device -fsycl-enable-optimizations -triple spir64-unknown-unknown-sycldevice -emit-llvm -x c++ %s -o - \| FileCheck %s`
	`1`	`+// RUN: %clang_cc1 -fsycl -fsycl-is-device -triple spir64-unknown-unknown-sycldevice -emit-llvm -x c++ %s -o - \| FileCheck %s`
`2`	`2`
`3`	`3`	`class kernel;`
`4`	`4`
Original file line number	Diff line number	Diff line change
`@@ -1,4 +1,4 @@`
`1`		`-// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s -o %t.out`
	`1`	`+// RUN: %clangxx -fsycl -fno-sycl-std-optimizations -fsycl-targets=%sycl_triple %s -o %t.out`
`2`	`2`	`// RUN: env SYCL_DEVICE_TYPE=HOST %t.out`
`3`	`3`	`// RUN: %CPU_RUN_PLACEHOLDER %t.out`
`4`	`4`	`// RUN: %GPU_RUN_PLACEHOLDER %t.out`
Original file line number	Diff line number	Diff line change
`@@ -1,4 +1,4 @@`
`1`		`-// RUN: %clangxx -fsycl %s -o %t.out`
	`1`	`+// RUN: %clangxx -fsycl -fno-sycl-std-optimizations %s -o %t.out`
`2`	`2`	`// RUNx: %ACC_RUN_PLACEHOLDER %t.out`
`3`	`3`	`//==----------------- fpga_lsu.cpp - SYCL FPGA LSU test --------------------==//`
`4`	`4`	`//`