Merge remote-tracking branch 'intel_llvm/sycl' into llvmspirv_pulldown

Author: vmaksimo

buildbot/dependency.conf

@@ -1,33 +1,33 @@
 [VERSIONS]
-# https://github.com/intel/llvm/releases/download/2020-WW41/oclcpuexp-2020.11.10.0.05_rel.tar.gz
-ocl_cpu_rt_ver=2020.11.10.0.05
-# https://github.com/intel/llvm/releases/download/2020-WW41/win-oclcpuexp-2020.11.10.0.05_rel.zip
-ocl_cpu_rt_ver_win=2020.11.10.0.05
+# https://github.com/intel/llvm/releases/download/2020-WW45/oclcpuexp-2020.11.11.0.04_rel.tar.gz
+ocl_cpu_rt_ver=2020.11.11.0.04
+# https://github.com/intel/llvm/releases/download/2020-WW45/win-oclcpuexp-2020.11.11.0.04_rel.zip
+ocl_cpu_rt_ver_win=2020.11.11.0.04
 # Same GPU driver supports Level Zero and OpenCL:
-# https://github.com/intel/compute-runtime/releases/tag/20.42.18209
-ocl_gpu_rt_ver=20.42.18209
+# https://github.com/intel/compute-runtime/releases/tag/20.44.18297
+ocl_gpu_rt_ver=20.44.18297
 # Same GPU driver supports Level Zero and OpenCL:
-# https://downloadmirror.intel.com/29879/a08/igfx_win10_100.8778.zip
-ocl_gpu_rt_ver_win=27.20.100.8778
+# https://downloadmirror.intel.com/29988/a08/igfx_win10_100.8935.zip
+ocl_gpu_rt_ver_win=27.20.100.8935
 intel_sycl_ver=build
-# https://github.com/oneapi-src/oneTBB/releases/download/v2021.1-beta08/oneapi-tbb-2021.1-beta08-lin.tgz
-tbb_ver=2021.1.10.2267
-# https://github.com/oneapi-src/oneTBB/releases/download/v2021.1-beta08/oneapi-tbb-2021.1-beta08-win.zip
-tbb_ver_win=2021.1.10.2267
-# https://github.com/intel/llvm/releases/download/2020-WW41/fpgaemu-2020.11.10.0.05_rel.tar.gz
-ocl_fpga_emu_ver=2020.11.10.0.05
-# https://github.com/intel/llvm/releases/download/2020-WW41/win-fpgaemu-2020.11.10.0.05_rel.zip
-ocl_fpga_emu_ver_win=2020.11.10.0.05
-fpga_ver=20200913_000017
-fpga_ver_win=20200913_000017
+# https://github.com/oneapi-src/oneTBB/releases/download/v2021.1-beta10/oneapi-tbb-2021.1-beta10-lin.tgz
+tbb_ver=2021.1.053
+# https://github.com/oneapi-src/oneTBB/releases/download/v2021.1-beta10/oneapi-tbb-2021.1-beta10-win.zip
+tbb_ver_win=2021.1.049
+# https://github.com/intel/llvm/releases/download/2020-WW45/fpgaemu-2020.11.11.0.04_rel.tar.gz
+ocl_fpga_emu_ver=2020.11.11.0.04
+# https://github.com/intel/llvm/releases/download/2020-WW45/win-fpgaemu-2020.11.11.0.04_rel.zip
+ocl_fpga_emu_ver_win=2020.11.11.0.04
+fpga_ver=20201021_000005
+fpga_ver_win=20201022_000005
 
 [DRIVER VERSIONS]
-cpu_driver_lin=2020.11.10.0.05
-cpu_driver_win=2020.11.10.0.05
-gpu_driver_lin=20.42.18209
-gpu_driver_win=27.20.100.8778
-fpga_driver_lin=2020.11.10.0.05
-fpga_driver_win=2020.11.10.0.05
+cpu_driver_lin=2020.11.11.0.04
+cpu_driver_win=2020.11.11.0.04
+gpu_driver_lin=20.44.18297
+gpu_driver_win=27.20.100.8935
+fpga_driver_lin=2020.11.11.0.04
+fpga_driver_win=2020.11.11.0.04
 # NVidia CUDA driver
 # TODO provide URL for CUDA driver
 nvidia_gpu_driver_lin=435.21

clang/include/clang/Basic/Attr.td

@@ -1226,6 +1226,19 @@ def SYCLIntelNumSimdWorkItems : InheritableAttr {
   let PragmaAttributeSupport = 0;
 }
 
+def SYCLIntelStallEnable : InheritableAttr {
+  let Spellings = [CXX11<"intel","stall_enable">];
+  let LangOpts = [SYCLIsHost, SYCLIsDevice];
+  let Subjects = SubjectList<[Function], ErrorDiag>;
+  let AdditionalMembers = [{
+    static const char *getName() {
+      return "stall_enable";
+    }
+  }];
+  let Documentation = [SYCLIntelStallEnableAttrDocs];
+  let PragmaAttributeSupport = 0;
+}
+
 def SYCLIntelSchedulerTargetFmaxMhz : InheritableAttr {
   let Spellings = [CXX11<"intelfpga","scheduler_target_fmax_mhz">,
                    CXX11<"intel","scheduler_target_fmax_mhz">];

clang/include/clang/Basic/AttrDocs.td

@@ -2229,6 +2229,39 @@ device kernel, the attribute is ignored and it is not propagated to a kernel.
   }];
 }
 
+def SYCLIntelStallEnableAttrDocs : Documentation {
+  let Category = DocCatFunction;
+  let Heading = "intel::stall_enable";
+  let Content = [{
+When applied to a lambda or function call operator (of a function object)
+on device, this requests, to the extent possible, that statically-scheduled
+clusters handle stalls using a stall-enable signal to freeze computation
+within the cluster. This attribute is ignored on the host.
+
+If ``intel::stall_enable`` is applied to a function called from a device
+kernel, the attribute is ignored and it is not propagated to a kernel.
+
+The ``intel::stall_enable`` attribute takes no argument and has an effect
+when applied to a function, and no effect otherwise.
+
+.. code-block:: c++
+
+  class Functor
+  {
+      [[intel::stall_enable]] void operator()(item<1> item)
+      {
+          /* kernel code */
+      }
+  }
+
+  kernel<class kernel_name>(
+  []() [[intel::stall_enable]] {
+       /* kernel code */
+  });
+
+  }];
+}
+
 def ReqdWorkGroupSizeAttrDocs : Documentation {
   let Category = DocCatFunction;
   let Heading = "reqd_work_group_size";

clang/include/clang/Basic/AttributeCommonInfo.h

@@ -165,7 +165,8 @@ class AttributeCommonInfo {
         ParsedAttr == AT_SYCLIntelSchedulerTargetFmaxMhz ||
         ParsedAttr == AT_SYCLIntelMaxWorkGroupSize ||
         ParsedAttr == AT_SYCLIntelMaxGlobalWorkDim ||
-        ParsedAttr == AT_SYCLIntelNoGlobalWorkOffset)
+        ParsedAttr == AT_SYCLIntelNoGlobalWorkOffset ||
+        ParsedAttr == AT_SYCLIntelStallEnable)
       return true;
 
     return false;

clang/include/clang/Basic/DiagnosticSemaKinds.td

@@ -11085,12 +11085,14 @@ def err_ext_int_max_size : Error<"%select{signed|unsigned}0 _ExtInt of bit "
 def err_esimd_glob_cant_init : Error<
   "SYCL explicit SIMD does not permit private global variable to have an initializer">;
 
-def err_sycl_kernel_incorrectly_named : Error<
-  "kernel %select{name is missing"
-  "|needs to have a globally-visible name"
-  "|name is invalid. Unscoped enum requires fixed underlying type"
-  "|name cannot be a type in the \"std\" namespace"
+def err_sycl_kernel_incorrectly_named : Error<"%0 is an invalid kernel name type">;
+def note_invalid_type_in_sycl_kernel : Note<
+  "%select{%1 should be globally-visible"
+  "|unscoped enum %1 requires fixed underlying type"
+  "|type %1 cannot be in the \"std\" namespace"
+  "|unnamed type used in a SYCL kernel name"
   "}0">;
+
 def err_sycl_kernel_not_function_object
     : Error<"kernel parameter must be a lambda or function object">;
 def err_sycl_restrict : Error<

clang/include/clang/Driver/Options.td

@@ -1937,7 +1937,9 @@ def fsycl_link : Flag<["-"], "fsycl-link">, Alias<fsycl_link_EQ>,
 def fsycl_unnamed_lambda : Flag<["-"], "fsycl-unnamed-lambda">,
   Flags<[CC1Option, CoreOption]>, HelpText<"Allow unnamed SYCL lambda kernels">;
 def fsycl_help_EQ : Joined<["-"], "fsycl-help=">,
-  Flags<[NoXarchOption, CoreOption]>, HelpText<"Emit help information from the related offline compilation tool">, Values<"all,fpga,gen,x86_64">;
+  Flags<[NoXarchOption, CoreOption]>, HelpText<"Emit help information from the "
+  "related offline compilation tool. Valid values: all, fpga, gen, x86_64.">,
+  Values<"all,fpga,gen,x86_64">;
 def fsycl_help : Flag<["-"], "fsycl-help">, Alias<fsycl_help_EQ>,
   Flags<[NoXarchOption, CoreOption]>, AliasArgs<["all"]>, HelpText<"Emit help information "
   "from all of the offline compilation tools">;

clang/lib/CodeGen/CodeGenFunction.cpp

@@ -676,6 +676,12 @@ void CodeGenFunction::EmitOpenCLKernelMetadata(const FunctionDecl *FD,
     if (A->getEnabled())
       Fn->setMetadata("no_global_work_offset", llvm::MDNode::get(Context, {}));
   }
+
+  if (FD->hasAttr<SYCLIntelStallEnableAttr>()) {
+    llvm::Metadata *AttrMDArgs[] = {
+        llvm::ConstantAsMetadata::get(Builder.getInt32(1))};
+    Fn->setMetadata("stall_enable", llvm::MDNode::get(Context, AttrMDArgs));
+  }
 }
 
 /// Determine whether the function F ends with a return stmt.

clang/lib/Driver/ToolChains/Clang.cpp

@@ -4906,7 +4906,7 @@ void Clang::ConstructJob(Compilation &C, const JobAction &JA,
   }
 
   if (Arg *A = Args.getLastArg(options::OPT_LongDouble_Group)) {
-    if (TC.getTriple().isX86())
+    if (TC.getTriple().isX86() || TC.getTriple().isSPIR())
       A->render(Args, CmdArgs);
     else if ((TC.getArch() == llvm::Triple::ppc || TC.getTriple().isPPC64()) &&
              (A->getOption().getID() != options::OPT_mlong_double_80))

clang/lib/Sema/SemaDeclAttr.cpp

@@ -3029,6 +3029,20 @@ static void handleNumSimdWorkItemsAttr(Sema &S, Decl *D,
                                                                      E);
 }
 
+// Handles stall_enable
+static void handleStallEnableAttr(Sema &S, Decl *D, const ParsedAttr &Attr) {
+  if (D->isInvalidDecl())
+    return;
+
+  unsigned NumArgs = Attr.getNumArgs();
+  if (NumArgs > 0) {
+    S.Diag(Attr.getLoc(), diag::warn_attribute_too_many_arguments) << Attr << 0;
+    return;
+  }
+
+  handleSimpleAttribute<SYCLIntelStallEnableAttr>(S, D, Attr);
+}
+
 // Add scheduler_target_fmax_mhz
 void Sema::addSYCLIntelSchedulerTargetFmaxMhzAttr(
     Decl *D, const AttributeCommonInfo &Attr, Expr *E) {
@@ -8401,6 +8415,9 @@ static void ProcessDeclAttribute(Sema &S, Scope *scope, Decl *D,
   case ParsedAttr::AT_SYCLIntelNoGlobalWorkOffset:
     handleNoGlobalWorkOffsetAttr(S, D, AL);
     break;
+  case ParsedAttr::AT_SYCLIntelStallEnable:
+    handleStallEnableAttr(S, D, AL);
+    break;
   case ParsedAttr::AT_VecTypeHint:
     handleVecTypeHint(S, D, AL);
     break;

clang/lib/Sema/SemaSYCL.cpp

@@ -545,6 +545,20 @@ class MarkDeviceFunction : public RecursiveASTVisitor<MarkDeviceFunction> {
 
       if (auto *A = FD->getAttr<SYCLSimdAttr>())
         Attrs.insert(A);
+
+      // Allow the kernel attribute "stall_enable" only on lambda functions
+      // and function objects that are called directly from a kernel
+      // (i.e. the one passed to the single_task or parallel_for functions).
+      // For all other cases, emit a warning and ignore.
+      if (auto *A = FD->getAttr<SYCLIntelStallEnableAttr>()) {
+        if (ParentFD == SYCLKernel) {
+          Attrs.insert(A);
+        } else {
+          SemaRef.Diag(A->getLocation(), diag::warn_attribute_ignored) << A;
+          FD->dropAttr<SYCLIntelStallEnableAttr>();
+        }
+      }
+
       // Propagate the explicit SIMD attribute through call graph - it is used
       // to distinguish ESIMD code in ESIMD LLVM passes.
       if (KernelBody && KernelBody->hasAttr<SYCLSimdAttr>() &&
@@ -2833,6 +2847,7 @@ class SYCLKernelNameTypeVisitor
       public ConstTemplateArgumentVisitor<SYCLKernelNameTypeVisitor> {
   Sema &S;
   SourceLocation KernelInvocationFuncLoc;
+  QualType KernelNameType;
   using InnerTypeVisitor = TypeVisitor<SYCLKernelNameTypeVisitor>;
   using InnerTemplArgVisitor =
       ConstTemplateArgumentVisitor<SYCLKernelNameTypeVisitor>;
@@ -2844,8 +2859,10 @@ class SYCLKernelNameTypeVisitor
   }
 
 public:
-  SYCLKernelNameTypeVisitor(Sema &S, SourceLocation KernelInvocationFuncLoc)
-      : S(S), KernelInvocationFuncLoc(KernelInvocationFuncLoc) {}
+  SYCLKernelNameTypeVisitor(Sema &S, SourceLocation KernelInvocationFuncLoc,
+                            QualType KernelNameType)
+      : S(S), KernelInvocationFuncLoc(KernelInvocationFuncLoc),
+        KernelNameType(KernelNameType) {}
 
   bool isValid() { return !IsInvalid; }
 
@@ -2856,7 +2873,9 @@ class SYCLKernelNameTypeVisitor
     if (!RD) {
       if (T->isNullPtrType()) {
         S.Diag(KernelInvocationFuncLoc, diag::err_sycl_kernel_incorrectly_named)
-            << /* kernel name cannot be a type in the std namespace */ 3;
+            << KernelNameType;
+        S.Diag(KernelInvocationFuncLoc, diag::note_invalid_type_in_sycl_kernel)
+            << /* kernel name cannot be a type in the std namespace */ 2 << T;
         IsInvalid = true;
       }
       return;
@@ -2881,9 +2900,10 @@ class SYCLKernelNameTypeVisitor
     const EnumDecl *ED = T->getDecl();
     if (!ED->isScoped() && !ED->isFixed()) {
       S.Diag(KernelInvocationFuncLoc, diag::err_sycl_kernel_incorrectly_named)
-          << /* Unscoped enum requires fixed underlying type */ 2;
-      S.Diag(ED->getSourceRange().getBegin(), diag::note_entity_declared_at)
-          << ED;
+          << KernelNameType;
+      S.Diag(KernelInvocationFuncLoc, diag::note_invalid_type_in_sycl_kernel)
+          << /* Unscoped enum requires fixed underlying type */ 1
+          << QualType(ED->getTypeForDecl(), 0);
       IsInvalid = true;
     }
   }
@@ -2900,7 +2920,10 @@ class SYCLKernelNameTypeVisitor
       auto *NameSpace = dyn_cast_or_null<NamespaceDecl>(DeclCtx);
       if (NameSpace && NameSpace->isStdNamespace()) {
         S.Diag(KernelInvocationFuncLoc, diag::err_sycl_kernel_incorrectly_named)
-            << /* kernel name cannot be a type in the std namespace */ 3;
+            << KernelNameType;
+        S.Diag(KernelInvocationFuncLoc, diag::note_invalid_type_in_sycl_kernel)
+            << /* kernel name cannot be a type in the std namespace */ 2
+            << QualType(Tag->getTypeForDecl(), 0);
         IsInvalid = true;
         return;
       }
@@ -2909,19 +2932,27 @@ class SYCLKernelNameTypeVisitor
         if (KernelNameIsMissing) {
           S.Diag(KernelInvocationFuncLoc,
                  diag::err_sycl_kernel_incorrectly_named)
-              << /* kernel name is missing */ 0;
+              << KernelNameType;
+          S.Diag(KernelInvocationFuncLoc,
+                 diag::note_invalid_type_in_sycl_kernel)
+              << /* unnamed type used in a SYCL kernel name */ 3;
           IsInvalid = true;
           return;
         }
         if (Tag->isCompleteDefinition()) {
           S.Diag(KernelInvocationFuncLoc,
                  diag::err_sycl_kernel_incorrectly_named)
-              << /* kernel name is not globally-visible */ 1;
+              << KernelNameType;
+          S.Diag(KernelInvocationFuncLoc,
+                 diag::note_invalid_type_in_sycl_kernel)
+              << /* kernel name is not globally-visible */ 0
+              << QualType(Tag->getTypeForDecl(), 0);
           IsInvalid = true;
-        } else
+        } else {
           S.Diag(KernelInvocationFuncLoc, diag::warn_sycl_implicit_decl);
-        S.Diag(Tag->getSourceRange().getBegin(), diag::note_previous_decl)
-            << Tag->getName();
+          S.Diag(Tag->getSourceRange().getBegin(), diag::note_previous_decl)
+              << Tag->getName();
+        }
       }
     }
   }
@@ -2998,7 +3029,8 @@ void Sema::CheckSYCLKernelCall(FunctionDecl *KernelFunc, SourceRange CallLoc,
   SyclKernelArgsSizeChecker ArgsSizeChecker(*this, Args[0]->getExprLoc());
 
   KernelObjVisitor Visitor{*this};
-  SYCLKernelNameTypeVisitor KernelNameTypeVisitor(*this, Args[0]->getExprLoc());
+  SYCLKernelNameTypeVisitor KernelNameTypeVisitor(*this, Args[0]->getExprLoc(),
+                                                  KernelNameType);
 
   DiagnosingSYCLKernel = true;
 
@@ -3204,6 +3236,7 @@ void Sema::MarkDevice(void) {
         case attr::Kind::SYCLIntelSchedulerTargetFmaxMhz:
         case attr::Kind::SYCLIntelMaxGlobalWorkDim:
         case attr::Kind::SYCLIntelNoGlobalWorkOffset:
+        case attr::Kind::SYCLIntelStallEnable:
         case attr::Kind::SYCLSimd: {
           if ((A->getKind() == attr::Kind::SYCLSimd) && KernelBody &&
               !KernelBody->getAttr<SYCLSimdAttr>()) {

clang/test/CodeGenSYCL/stall_enable.cpp

@@ -0,0 +1,26 @@
+// RUN: %clang_cc1 -fsycl -fsycl-is-device -internal-isystem %S/Inputs -triple spir64-unknown-unknown-sycldevice -disable-llvm-passes -emit-llvm -o - %s | FileCheck %s
+
+#include "sycl.hpp"
+
+using namespace cl::sycl;
+queue q;
+
+class Foo {
+public:
+  [[intel::stall_enable]] void operator()() const {}
+};
+
+int main() {
+  q.submit([&](handler &h) {
+    Foo f;
+    h.single_task<class test_kernel1>(f);
+
+    h.single_task<class test_kernel2>(
+        []() [[intel::stall_enable]]{});
+  });
+  return 0;
+}
+
+// CHECK: define spir_kernel void @"{{.*}}test_kernel1"() #0 {{.*}} !stall_enable ![[NUM5:[0-9]+]]
+// CHECK: define spir_kernel void @"{{.*}}test_kernel2"() #0 {{.*}} !stall_enable ![[NUM5]]
+// CHECK: ![[NUM5]] = !{i32 1}

clang/test/Driver/mlong-double-128.c

@@ -2,6 +2,7 @@
 // RUN: %clang -target powerpc64-pc-freebsd12 -c -### %s -mlong-double-128 2>&1 | FileCheck %s
 // RUN: %clang -target powerpc64le-linux-musl -c -### %s -mlong-double-128 2>&1 | FileCheck %s
 // RUN: %clang -target i686-linux-gnu -c -### %s -mlong-double-128 2>&1 | FileCheck %s
+// RUN: %clang -target spir64-unknown-unknown-sycldevice -c -### %s -mlong-double-128 2>&1 | FileCheck %s
 
 // RUN: %clang -target x86_64-linux-musl -c -### %s -mlong-double-128 -mlong-double-80 2>&1 | FileCheck --implicit-check-not=-mlong-double-128 /dev/null
 // RUN: %clang -target x86_64-linux-musl -c -### %s -mlong-double-80 -mlong-double-128 2>&1 | FileCheck %s

clang/test/Driver/mlong-double-64.c

@@ -3,6 +3,7 @@
 // RUN: %clang -target powerpc64le-linux-musl -c -### %s -mlong-double-64 2>&1 | FileCheck %s
 // RUN: %clang -target i686-linux-gnu -c -### %s -mlong-double-64 2>&1 | FileCheck %s
 // RUN: %clang -target x86_64-linux-musl -c -### %s -mlong-double-64 2>&1 | FileCheck %s
+// RUN: %clang -target spir64-unknown-unknown-sycldevice -c -### %s -mlong-double-64 2>&1 | FileCheck %s
 
 // CHECK: "-mlong-double-64"