[flang] [clang] [flang] add fveclib flag (PR #71734)
@@ -81,6 +81,17 @@ class CodeGenOptions : public CodeGenOptionsBase {
RK_WithPattern, // Remark pattern specified via '-Rgroup=regexp'.
};
+ enum class VectorLibrary {
+NoLibrary, // Don't use any vector library.
+Accelerate, // Use the Accelerate framework.
+LIBMVEC,// GLIBC vector math library.
+MASSV, // IBM MASS vector library.
+SVML, // Intel short vector math library.
+SLEEF, // SLEEF SIMD Library for Evaluating Elementary Functions.
+Darwin_libsystem_m, // Use Darwin's libsystem_m vector functions.
+ArmPL // Arm Performance Libraries.
+ };
kiranchandramohan wrote:
Can this class be moved to a file in a new directory
`llvm/include/llvm/Frontend/Driver` and shared with Clang?
https://github.com/llvm/llvm-project/pull/71734
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [clang] [flang] add fveclib flag (PR #71734)
@@ -843,6 +843,44 @@ getOutputStream(CompilerInstance &ci, llvm::StringRef
inFile,
llvm_unreachable("Invalid action!");
}
+static std::unique_ptr
+createTLII(llvm::Triple &targetTriple, const CodeGenOptions &codeGenOpts) {
+ auto tlii = std::make_unique(targetTriple);
+ assert(tlii && "Failed to create TargetLibraryInfo");
+
+ using VecLib = llvm::TargetLibraryInfoImpl::VectorLibrary;
+ VecLib vecLib = VecLib::NoLibrary;
+ switch (codeGenOpts.getVecLib()) {
+ case CodeGenOptions::VectorLibrary::Accelerate:
+vecLib = VecLib::Accelerate;
+break;
+ case CodeGenOptions::VectorLibrary::LIBMVEC:
+vecLib = VecLib::LIBMVEC_X86;
+break;
+ case CodeGenOptions::VectorLibrary::MASSV:
+vecLib = VecLib::MASSV;
+break;
+ case CodeGenOptions::VectorLibrary::SVML:
+vecLib = VecLib::SVML;
+break;
+ case CodeGenOptions::VectorLibrary::SLEEF:
+vecLib = VecLib::SLEEFGNUABI;
+break;
+ case CodeGenOptions::VectorLibrary::Darwin_libsystem_m:
+vecLib = VecLib::DarwinLibSystemM;
+break;
+ case CodeGenOptions::VectorLibrary::ArmPL:
+vecLib = VecLib::ArmPL;
+break;
+ case CodeGenOptions::VectorLibrary::NoLibrary:
+vecLib = VecLib::NoLibrary;
+break;
+ }
+
+ tlii->addVectorizableFunctionsFromVecLib(vecLib, targetTriple);
+ return tlii;
+}
kiranchandramohan wrote:
Can this code be moved to `llvm/lib/Frontend/Driver` and shared with Clang?
https://github.com/llvm/llvm-project/pull/71734
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [llvm] [flang] [flang] add fveclib flag (PR #71734)
https://github.com/kiranchandramohan approved this pull request. Nice work. LGTM. https://github.com/llvm/llvm-project/pull/71734 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[llvm] [flang] [clang] [flang] add fveclib flag (PR #71734)
https://github.com/kiranchandramohan edited https://github.com/llvm/llvm-project/pull/71734 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [llvm] [flang] [flang] add fveclib flag (PR #71734)
@@ -111,7 +111,7 @@ int main(int argc, const char **argv) {
auto Files = llvm::makeIntrusiveRefCnt(FileSystemOptions(),
OFS);
- auto Driver = std::make_unique(
+ auto Driver = std::make_unique(
kiranchandramohan wrote:
Are these clang prefixes required?
https://github.com/llvm/llvm-project/pull/71734
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[llvm] [clang] [flang] [flang] add fveclib flag (PR #71734)
@@ -851,11 +851,10 @@ getOutputStream(CompilerInstance &ci, llvm::StringRef inFile, /// \param [in] act Backend act to run (assembly vs machine-code generation) /// \param [in] llvmModule LLVM module to lower to assembly/machine-code /// \param [out] os Output stream to emit the generated code to kiranchandramohan wrote: Nit: Add the new option here. May be good to have the stream as the last option. https://github.com/llvm/llvm-project/pull/71734 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[llvm] [clang] [flang] [flang] add fveclib flag (PR #71734)
@@ -0,0 +1,13 @@ +! test that -fveclib= is passed to the backend +! -target aarch64 so that ArmPL is available +! RUN: %flang -S -target aarch64-unknown-linux-gnu -mcpu=neoverse-v1 -Ofast -fveclib=ArmPL -o - %s | FileCheck %s + kiranchandramohan wrote: Will this work on all targets? Or only on AArch64? https://github.com/llvm/llvm-project/pull/71734 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[llvm] [flang] [clang] [flang] add fveclib flag (PR #71734)
@@ -0,0 +1,13 @@ +! test that -fveclib= is passed to the backend +! -target aarch64 so that ArmPL is available +! RUN: %flang -S -target aarch64-unknown-linux-gnu -mcpu=neoverse-v1 -Ofast -fveclib=ArmPL -o - %s | FileCheck %s + kiranchandramohan wrote: I was asking whether this test needs to be restricted to only work on a system where Arm target is enabled. https://github.com/llvm/llvm-project/pull/71734 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [clang] [flang][Driver] Support -rpath, -shared, and -static in the frontend (PR #66702)
kiranchandramohan wrote: @banach-space Is this patch OK with you? https://github.com/llvm/llvm-project/pull/66702 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [flang][driver] Rename `flang-new` as `flang` (PR #74377)
kiranchandramohan wrote: I am assuming you have got a go-ahead from the code owner and others involved in the discourse discussion. It is probably good to post a link to this patch in the relevant discourse post for information and confirmation. https://github.com/llvm/llvm-project/pull/74377 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [llvm] [clang] [clang-tools-extra] [flang] GETLOG runtime and extension implementation: get login username (PR #74628)
@@ -10,10 +10,50 @@ // extensions that will eventually be implemented in Fortran. #include "flang/Runtime/extensions.h" +#include "flang/Runtime/character.h" #include "flang/Runtime/command.h" #include "flang/Runtime/descriptor.h" #include "flang/Runtime/io-api.h" +#ifdef _WIN32 +#define WIN32_LEAN_AND_MEAN +#define NOMINMAX +#include + +#include // wcstombs_s +#include // UNLEN=256 +#include // wchar_t cast to LPWSTR +#pragma comment(lib, "Advapi32.lib") // Link Advapi32.lib for GetUserName kiranchandramohan wrote: We usually avoid depending on external libraries. Can this be avoided? https://github.com/llvm/llvm-project/pull/74628 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] 08cb640 - [NFC] Remove an unused decl to avoid warning
Author: Kiran Chandramohan
Date: 2023-12-09T16:27:42Z
New Revision: 08cb64034f17d50a660ec78ce8ea81a025b0ba71
URL:
https://github.com/llvm/llvm-project/commit/08cb64034f17d50a660ec78ce8ea81a025b0ba71
DIFF:
https://github.com/llvm/llvm-project/commit/08cb64034f17d50a660ec78ce8ea81a025b0ba71.diff
LOG: [NFC] Remove an unused decl to avoid warning
Added:
Modified:
clang/lib/Driver/ToolChains/WebAssembly.cpp
Removed:
diff --git a/clang/lib/Driver/ToolChains/WebAssembly.cpp
b/clang/lib/Driver/ToolChains/WebAssembly.cpp
index f04018179a5da..f131b6cf3baff 100644
--- a/clang/lib/Driver/ToolChains/WebAssembly.cpp
+++ b/clang/lib/Driver/ToolChains/WebAssembly.cpp
@@ -143,7 +143,7 @@ void wasm::Linker::ConstructJob(Compilation &C, const
JobAction &JA,
// When optimizing, if wasm-opt is available, run it.
std::string WasmOptPath;
- if (Arg *A = Args.getLastArg(options::OPT_O_Group)) {
+ if (Args.getLastArg(options::OPT_O_Group)) {
WasmOptPath = ToolChain.GetProgramPath("wasm-opt");
if (WasmOptPath == "wasm-opt") {
WasmOptPath = {};
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [llvm] [mlir] [AArch64][SME] Remove immediate argument restriction for svldr and svstr (PR #68565)
https://github.com/kiranchandramohan updated
https://github.com/llvm/llvm-project/pull/68565
>From 83e20904c206980285c4ee9d0227706803147654 Mon Sep 17 00:00:00 2001
From: Samuel Tebbs
Date: Fri, 6 Oct 2023 17:09:36 +0100
Subject: [PATCH 01/13] [AArch64][SME] Remove immediate argument restriction
for svldr and svstr
The svldr_vnum_za and svstr_vnum_za builtins/intrinsics currently
require that the vnum argument be an immediate, since the instructions
take an immediate vector number. However, we emit 0 as the immediate
for the instruction no matter what, and instead modify the base register.
This patch removes that restriction on the argument, so that the
argument can be a non-immediate. If an appropriate immediate was
passed to the builtin then CGBuiltin passes that directly to the LLVM
intrinsic, otherwise it modifies the base register as is existing
behaviour.
---
clang/lib/CodeGen/CGBuiltin.cpp | 45
.../aarch64-sme-intrinsics/acle_sme_ldr.c | 71 ---
.../aarch64-sme-intrinsics/acle_sme_str.c | 51 -
llvm/include/llvm/IR/IntrinsicsAArch64.td | 2 +-
llvm/lib/Target/AArch64/SMEInstrFormats.td| 10 +--
.../CostModel/ARM/unaligned_double_load.ll| 59 +++
.../CodeGen/AArch64/sme-intrinsics-loads.ll | 33 +++--
7 files changed, 165 insertions(+), 106 deletions(-)
create mode 100644 llvm/test/Analysis/CostModel/ARM/unaligned_double_load.ll
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 09309a3937fb613..8444aea8c8ac4b6 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -9815,6 +9815,11 @@ Value *CodeGenFunction::EmitSVEMaskedStore(const
CallExpr *E,
return Store;
}
+Value *CodeGenFunction::EmitTileslice(Value *Offset, Value *Base) {
+ llvm::Value *CastOffset = Builder.CreateIntCast(Offset, Int64Ty, false);
+ return Builder.CreateAdd(Base, CastOffset, "tileslice");
+}
+
Value *CodeGenFunction::EmitSMELd1St1(const SVETypeFlags &TypeFlags,
SmallVectorImpl &Ops,
unsigned IntID) {
@@ -9870,18 +9875,34 @@ Value *CodeGenFunction::EmitSMEZero(const SVETypeFlags
&TypeFlags,
Value *CodeGenFunction::EmitSMELdrStr(const SVETypeFlags &TypeFlags,
SmallVectorImpl &Ops,
unsigned IntID) {
- if (Ops.size() == 3) {
-Function *Cntsb = CGM.getIntrinsic(Intrinsic::aarch64_sme_cntsb);
-llvm::Value *CntsbCall = Builder.CreateCall(Cntsb, {}, "svlb");
-
-llvm::Value *VecNum = Ops[2];
-llvm::Value *MulVL = Builder.CreateMul(CntsbCall, VecNum, "mulvl");
-
-Ops[1] = Builder.CreateGEP(Int8Ty, Ops[1], MulVL);
-Ops[0] = Builder.CreateAdd(
-Ops[0], Builder.CreateIntCast(VecNum, Int32Ty, true), "tileslice");
-Ops.erase(&Ops[2]);
- }
+ if (Ops.size() == 2) {
+// Intrinsics without a vecnum also use this function, so just provide 0
+Ops.push_back(Ops[1]);
+Ops[1] = Builder.getInt32(0);
+ } else {
+int Imm = -1;
+if (ConstantInt* C = dyn_cast(Ops[2]))
+ if (C->getZExtValue() <= 15)
+ Imm = C->getZExtValue();
+
+if (Imm != -1) {
+ Ops[2] = Ops[1];
+ Ops[1] = Builder.getInt32(Imm);
+} else {
+ Function *Cntsb = CGM.getIntrinsic(Intrinsic::aarch64_sme_cntsb);
+ llvm::Value *CntsbCall = Builder.CreateCall(Cntsb, {}, "svlb");
+
+ llvm::Value *VecNum = Ops[2];
+ llvm::Value *MulVL = Builder.CreateMul(
+ CntsbCall,
+ VecNum,
+ "mulvl");
+
+ Ops[2] = Builder.CreateGEP(Int8Ty, Ops[1], MulVL);
+ Ops[1] = Builder.getInt32(0);
+ Ops[0] = Builder.CreateIntCast(EmitTileslice(Ops[0], VecNum), Int32Ty,
false);
+}
+ }
Function *F = CGM.getIntrinsic(IntID, {});
return Builder.CreateCall(F, Ops);
}
diff --git a/clang/test/CodeGen/aarch64-sme-intrinsics/acle_sme_ldr.c
b/clang/test/CodeGen/aarch64-sme-intrinsics/acle_sme_ldr.c
index e85c47072f2df80..8e07cf1d11c19b2 100644
--- a/clang/test/CodeGen/aarch64-sme-intrinsics/acle_sme_ldr.c
+++ b/clang/test/CodeGen/aarch64-sme-intrinsics/acle_sme_ldr.c
@@ -6,57 +6,46 @@
#include
-// CHECK-C-LABEL: define dso_local void @test_svldr_vnum_za(
-// CHECK-C-SAME: i32 noundef [[SLICE_BASE:%.*]], ptr noundef [[PTR:%.*]])
local_unnamed_addr #[[ATTR0:[0-9]+]] {
-// CHECK-C-NEXT: entry:
-// CHECK-C-NEXT:tail call void @llvm.aarch64.sme.ldr(i32 [[SLICE_BASE]],
ptr [[PTR]])
-// CHECK-C-NEXT:ret void
-//
-// CHECK-CXX-LABEL: define dso_local void @_Z18test_svldr_vnum_zajPKv(
-// CHECK-CXX-SAME: i32 noundef [[SLICE_BASE:%.*]], ptr noundef [[PTR:%.*]])
local_unnamed_addr #[[ATTR0:[0-9]+]] {
-// CHECK-CXX-NEXT: entry:
-// CHECK-CXX-NEXT:tail call void @llvm.aarch64.sme.ldr(i32 [[SLICE_BASE]],
ptr [[PTR]])
-// CHECK-CXX-NEXT:ret void
+// CHECK-C-LABEL: @test_svldr_vnum_za(
+// CHECK-CXX-LABEL: @_Z18test_svldr_v
[flang] [clang] [Flang] Add code-object-version option (PR #72638)
@@ -85,6 +85,19 @@ class CodeGenOptions : public CodeGenOptionsBase {
RK_WithPattern, // Remark pattern specified via '-Rgroup=regexp'.
};
+ /// \brief Enumeration value for AMDGPU code object version, which is the
+ /// code object version times 100.
+ enum class CodeObjectVersionKind {
+COV_None,
+COV_2 = 200, // Unsupported.
+COV_3 = 300, // Unsupported.
+COV_4 = 400,
+COV_5 = 500,
+ };
kiranchandramohan wrote:
If possible move this to `llvm/include/llvm/Frontend/Driver/TargetOptions.h`.
https://github.com/llvm/llvm-project/pull/72638
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang-tools-extra] [flang] [llvm] [clang] [flang ]GETLOG runtime and extension implementation: get login username (PR #70917)
@@ -37,5 +80,17 @@ void FORTRAN_PROCEDURE_NAME(getarg)(
(void)RTNAME(GetCommandArgument)(
n, &value, nullptr, nullptr, __FILE__, __LINE__);
}
+
+void FORTRAN_PROCEDURE_NAME(getlog)(std::int8_t *arg, std::int64_t length) {
+ std::array str = {};
+
+ int error = getlogin_r(str.data(), str.size());
kiranchandramohan wrote:
```suggestion
int error{getlogin_r(str.data(), str.size())};
```
https://github.com/llvm/llvm-project/pull/70917
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[llvm] [clang-tools-extra] [flang] [clang] [flang ]GETLOG runtime and extension implementation: get login username (PR #70917)
@@ -10,9 +10,52 @@
// extensions that will eventually be implemented in Fortran.
#include "flang/Runtime/extensions.h"
+#include "terminator.h"
#include "flang/Runtime/command.h"
#include "flang/Runtime/descriptor.h"
#include "flang/Runtime/io-api.h"
+#include "flang/Runtime/time-intrinsic.h" // copyBufferAndPad
+#include
+
+#ifdef _WIN32
+#define WIN32_LEAN_AND_MEAN
+#define NOMINMAX
+#include
+
+#include // wcstombs_s
+#include // UNLEN=256
+#include // wchar_t cast to LPWSTR
+#pragma comment(lib, "Advapi32.lib") // Link Advapi32.lib for GetUserName
+#define LOGIN_NAME_MAX UNLEN
+
+inline int getlogin_r(char *buf, size_t bufSize) {
+ wchar_t w_username[UNLEN + 1] = {};
+ DWORD nameLen = UNLEN + 1;
kiranchandramohan wrote:
```suggestion
DWORD nameLen{UNLEN + 1};
```
https://github.com/llvm/llvm-project/pull/70917
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang][Driver] Let the linker fail on multiple definitions of main() (PR #73124)
@@ -1029,7 +1042,7 @@ void tools::addFortranRuntimeLibraryPath(const ToolChain
&TC,
ArgStringList &CmdArgs) {
// Default to the /../lib directory. This works fine on the
// platforms that we have tested so far. We will probably have to re-fine
- // this in the future. In particular, on some platforms, we may need to use
+ // this in the future. In particular, on some platforms, we may need to useq
kiranchandramohan wrote:
Nit: accidental change?
https://github.com/llvm/llvm-project/pull/73124
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[llvm] [mlir] [libcxx] [clang] [lldb] [flang] [Flang][OpenMP] Port openmp threadprivate-use-association.f90 test to HLFIR flow (PR #71043)
https://github.com/kiranchandramohan approved this pull request. LG. Thanks. https://github.com/llvm/llvm-project/pull/71043 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [llvm] [clang-tools-extra] [clang] [flang] Pass to add frame pointer attribute (PR #74598)
kiranchandramohan wrote: > > llc has a --frame-pointer option, so there is likely a way to pass this > > option to the llvm codegen as a global option instead of setting it on all > > function at the MLIR level if the only goal is to propagate it to llvm > > backend. > > I just discussed a bit with @vzakhari and one advantage of setting it per > function is to later be able to merge llvm IR files compiled with different > options. That would explain why clang does it. I would still add nothing if > the option is None since this is the default, but seems OK to me to decorate > the functions as you are doing based on this argument. @jeanPerier just checking whether you are OK with this pass being here or would prefer it to be in LLVM dialect? https://github.com/llvm/llvm-project/pull/74598 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[llvm] [clang] [flang] [clang-tools-extra] [flang] Pass to add frame pointer attribute (PR #74598)
@@ -38,6 +38,7 @@ CODEGENOPT(Underscoring, 1, 1) ENUM_CODEGENOPT(RelocationModel, llvm::Reloc::Model, 3, llvm::Reloc::PIC_) ///< Name of the relocation model to use. ENUM_CODEGENOPT(DebugInfo, llvm::codegenoptions::DebugInfoKind, 4, llvm::codegenoptions::NoDebugInfo) ///< Level of debug info to generate ENUM_CODEGENOPT(VecLib, llvm::driver::VectorLibrary, 3, llvm::driver::VectorLibrary::NoLibrary) ///< Vector functions library to use +ENUM_CODEGENOPT(FramePointer, llvm::FramePointerKind, 2, llvm::FramePointerKind::None) ///https://github.com/llvm/llvm-project/pull/74598 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[llvm] [clang] [flang] [clang-tools-extra] [flang] Pass to add frame pointer attribute (PR #74598)
@@ -245,6 +245,24 @@ static void
parseCodeGenArgs(Fortran::frontend::CodeGenOptions &opts,
opts.AliasAnalysis = opts.OptimizationLevel > 0;
+ if (const llvm::opt::Arg *a =
+ args.getLastArg(clang::driver::options::OPT_mframe_pointer_EQ)) {
+llvm::StringRef s = a->getValue();
+
+if (!(s == "none" || s == "non-leaf" || s == "all")) {
+ const auto debugWarningId = diags.getCustomDiagID(
+ clang::DiagnosticsEngine::Error, "Frame pointer: %0");
kiranchandramohan wrote:
Do we have a test for this?
https://github.com/llvm/llvm-project/pull/74598
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [llvm] [clang-tools-extra] [flang] [flang] Pass to add frame pointer attribute (PR #74598)
@@ -0,0 +1,62 @@ +//===- FunctionAttr.cpp -===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===--===// + +//===--===// +/// \file +/// This is a generic pass for adding attributes to functions. +//===--== kiranchandramohan wrote: Nit: The alignments of the lines seem to be off. https://github.com/llvm/llvm-project/pull/74598 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang-tools-extra] [flang] [clang] [llvm] [flang] Pass to add frame pointer attribute (PR #74598)
@@ -349,4 +349,23 @@ def VScaleAttr : Pass<"vscale-attr", "mlir::func::FuncOp">
{
let constructor = "::fir::createVScaleAttrPass()";
}
+def FunctionAttr : Pass<"function-attr", "mlir::func::FuncOp"> {
+ let summary = "This is a generic pass that adds function attributes that are
expected at the LLVM IR level";
+ let description = [{ This feature introduces a general attribute aimed at
customizing function characteristics.
+ Options include:
+ Add "frame-pointer" attribute to functions: Set an attribute for the
frame
+ pointer on functions, to avoid saving the frame pointer in a register in
+ functions where it is unnecessary.
+ This eliminates the need for instructions to save, establish, and restore
+ frame pointers, while also freeing up an additional register in numerous
functions.
+ However, this approach can make debugging unfeasible on certain machines.
+ }];
+ let options = [
+Option<"framePointerKind", "frame-pointer",
+ "mlir::LLVM::framePointerKind::FramePointerKind",
/*default=*/"mlir::LLVM::framePointerKind::FramePointerKind{}",
+ "frame pointer">,
+ ];
kiranchandramohan wrote:
Stick to 80 characters if possible in the summary, description and options.
https://github.com/llvm/llvm-project/pull/74598
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [flang][driver] Remove Fortain_main static library from linking stages (PR #75816)
kiranchandramohan wrote: Looks reasonable to me. I guess the environment settings that are there in the EnvironmentDefaults apply only when the main program is compiled with some flags. It would be good to check the options in EnvironmentDefaults and see that they are all applicable only when compiled for the main program. https://github.com/llvm/llvm-project/pull/75816 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [clang] [flang][Driver] Let the linker fail on multiple definitions of main() (PR #73124)
@@ -0,0 +1,15 @@ +! UNSUPPORTED: system-windows + +! RUN: %clang -o %t.c-object -c %S/Inputs/main_dupes.c kiranchandramohan wrote: You can use %cc for the system-compiler and REQUIRE it to be present for this test. Reference: https://github.com/llvm/llvm-project/blob/main/flang/test/Runtime/no-cpp-dep.c https://github.com/llvm/llvm-project/pull/73124 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [clang] [flang][Driver] Let the linker fail on multiple definitions of main() (PR #73124)
@@ -122,6 +122,7 @@
# the build directory holding that tool.
tools = [
ToolSubst("%flang", command=FindTool("flang-new"), unresolved="fatal"),
+ToolSubst("%clang", command=FindTool("clang"), unresolved="fatal"),
kiranchandramohan wrote:
Remove this if using the system compiler works.
https://github.com/llvm/llvm-project/pull/73124
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [clang] [Flang][Clang] Add support for frame pointers in Flang Driver (PR #72146)
@@ -49,6 +49,26 @@ class CodeGenOptionsBase {
class CodeGenOptions : public CodeGenOptionsBase {
public:
+ /// The type of frame pointer used
+ enum class FramePointerKind {
+None,// Omit all frame pointers.
+NonLeaf, // Keep non-leaf frame pointers.
+All, // Keep all frame pointers.
+ };
+
+ static llvm::StringRef getFramePointerKindName(FramePointerKind Kind) {
+switch (Kind) {
+case FramePointerKind::None:
+ return "none";
+case FramePointerKind::NonLeaf:
+ return "non-leaf";
+case FramePointerKind::All:
+ return "all";
+}
+
+llvm_unreachable("invalid FramePointerKind");
+ };
+
kiranchandramohan wrote:
Move this code and the same code in
`clang/include/clang/Basic/CodeGenOptions.h` to
llvm/include/llvm/Frontend/Driver.
https://github.com/llvm/llvm-project/pull/72146
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [clang] [flang][Driver] Let the linker fail on multiple definitions of main() (PR #73124)
https://github.com/kiranchandramohan approved this pull request. LG. Thanks for the patch and the changes. https://github.com/llvm/llvm-project/pull/73124 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [clang] [flang] Enable alias tags pass by default (PR #73111)
@@ -142,6 +142,26 @@ void Flang::addCodegenOptions(const ArgList &Args,
if (shouldLoopVersion(Args))
CmdArgs.push_back("-fversion-loops-for-stride");
+ Arg *aliasAnalysis = Args.getLastArg(options::OPT_falias_analysis,
+ options::OPT_fno_alias_analysis);
+ Arg *optLevel =
+ Args.getLastArg(options::OPT_Ofast, options::OPT_O, options::OPT_O4);
kiranchandramohan wrote:
Some discussion regarding `-Os` is here.
https://discourse.llvm.org/t/code-size-optimization-flags-in-flang/69482
https://github.com/llvm/llvm-project/pull/73111
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [clang] [flang] Enable alias tags pass by default (PR #73111)
https://github.com/kiranchandramohan edited https://github.com/llvm/llvm-project/pull/73111 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [flang] Enable alias tags pass by default (PR #73111)
https://github.com/kiranchandramohan approved this pull request. LGTM. https://github.com/llvm/llvm-project/pull/73111 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [flang] Enable alias tags pass by default (PR #73111)
@@ -142,6 +142,26 @@ void Flang::addCodegenOptions(const ArgList &Args,
if (shouldLoopVersion(Args))
CmdArgs.push_back("-fversion-loops-for-stride");
+ Arg *aliasAnalysis = Args.getLastArg(options::OPT_falias_analysis,
+ options::OPT_fno_alias_analysis);
+ Arg *optLevel =
+ Args.getLastArg(options::OPT_Ofast, options::OPT_O, options::OPT_O4);
kiranchandramohan wrote:
> That's fine, but then one has to decide whether -f{no}-alias-analysis
> overrides -O{n} or not? I think that "explicit" request from a user should
> always take precedence. This leads to (pseudo code):
>
> opts.AliasAnalysis = 0;
> if (opt level requiring alias analysis)
> opts.AliasAnalysis = 1;
>
> / / User request takes precedence when it comes to alias analysis.
> if (-falias-analysis or -fno-alias-analysis) then
> "do whatever the user requested"
>Separately, could you check what Clang does and make sure that that would be
>consistent?
@banach-space This is exactly the handling in the front-end driver as given
below (and in `lib/Frontend/CompilerInvocation`). The flang driver is only
deciding whether to forward or not.
```
opts.AliasAnalysis = opts.OptimizationLevel > 0;
if (auto *arg =
args.getLastArg(clang::driver::options::OPT_falias_analysis,
clang::driver::options::OPT_fno_alias_analysis))
opts.AliasAnalysis =
arg->getOption().matches(clang::driver::options::OPT_falias_analysis);
```
https://github.com/llvm/llvm-project/pull/73111
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] Revert "[flang] Enable alias tags pass by default (#73111)" (PR #73821)
https://github.com/kiranchandramohan approved this pull request. Thanks @tblah for the quick response. LG. https://github.com/llvm/llvm-project/pull/73821 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [libcxx] [clang] [libc] [libcxxabi] [lldb] [clang-tools-extra] [lld] [llvm] [flang] Pass Argv0 to getIntriniscDir and getOpenMPHeadersDir (PR #73254)
https://github.com/kiranchandramohan approved this pull request. LG. https://github.com/llvm/llvm-project/pull/73254 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[lldb] [clang-tools-extra] [llvm] [libcxxabi] [libc] [flang] [lld] [clang] [libcxx] [flang] Pass Argv0 to getIntriniscDir and getOpenMPHeadersDir (PR #73254)
https://github.com/kiranchandramohan edited https://github.com/llvm/llvm-project/pull/73254 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang-tools-extra] [libcxx] [lld] [libcxxabi] [libc] [flang] [lldb] [llvm] [clang] [flang] Pass Argv0 to getIntriniscDir and getOpenMPHeadersDir (PR #73254)
@@ -101,6 +101,8 @@ class CompilerInvocation : public CompilerInvocationBase {
bool warnAsErr = false;
+ const char *argv0;
kiranchandramohan wrote:
Nit: This is probably obvious but will be good to add executable name as a
comment above.
https://github.com/llvm/llvm-project/pull/73254
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [clang] [flang] (Re-)Enable alias tags pass by default (PR #74250)
https://github.com/kiranchandramohan approved this pull request. LGTM. Thanks. https://github.com/llvm/llvm-project/pull/74250 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [Clang][AArch64] Add fix vector types to header into SVE (PR #73258)
kiranchandramohan wrote: Just a drive-through comment. The CI is failing for the following two tests. Clang :: CodeGen/aapcs-align.cpp Clang :: CodeGen/aapcs64-align.cpp https://github.com/llvm/llvm-project/pull/73258 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [Clang][AArch64] Add fix vector types to header into SVE (PR #73258)
@@ -0,0 +1,13 @@ +//===--- arm_vector_types.td - ARM Fixed vector types compiler interface ---===// +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// +//===--===// +// +// This file defines the TableGen definitions from which the ARM BF16 header kiranchandramohan wrote: Did you mean BF16 here? The name of the file seems to refer to arm_vector_types in general. https://github.com/llvm/llvm-project/pull/73258 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [Clang][AArch64] Add fix vector types to header into SVE (PR #73258)
@@ -2229,6 +2231,12 @@ void NeonEmitter::runHeader(raw_ostream &OS) {
static void emitNeonTypeDefs(const std::string& types, raw_ostream &OS) {
std::string TypedefTypes(types);
std::vector TDTypeVec = TypeSpec::fromTypeSpecs(TypedefTypes);
+ // arm_sve.h followed by arm_neon.h does not emmit these types
kiranchandramohan wrote:
```suggestion
// arm_sve.h followed by arm_neon.h does not emit these types
```
https://github.com/llvm/llvm-project/pull/73258
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [Clang][AArch64] Add fix vector types to header into SVE (PR #73258)
@@ -2546,6 +2555,53 @@ void NeonEmitter::runFP16(raw_ostream &OS) {
OS << "#endif /* __ARM_FP16_H */\n";
}
+void NeonEmitter::runVectorType(raw_ostream &OS) {
+ OS << "/*=== arm_vector_types - ARM vector type "
+"--===\n"
+" *\n"
+" *\n"
+" * Part of the LLVM Project, under the Apache License v2.0 with LLVM "
+"Exceptions.\n"
+" * See https://llvm.org/LICENSE.txt for license information.\n"
+" * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception\n"
+" *\n"
+"
*===-"
+"--===\n"
+" */\n\n";
+ OS << "#ifndef __ARM_NEON_H\n\n";
+ OS << "#ifndef __ARM_NEON_TYPES_H\n";
+ OS << "#define __ARM_NEON_TYPES_H\n";
+ OS << "#ifdef __aarch64__\n";
+ OS << "typedef uint8_t poly8_t;\n";
+ OS << "typedef uint16_t poly16_t;\n";
+ OS << "typedef uint64_t poly64_t;\n";
+ OS << "typedef __uint128_t poly128_t;\n";
+ OS << "#else\n";
+ OS << "typedef int8_t poly8_t;\n";
+ OS << "typedef int16_t poly16_t;\n";
+ OS << "#endif\n";
+
+ // Needs to declare all the types in case there is arm_sve.h followed by
+ // arm_neon.h.
+ // arm_sve defines __ARM_NEON_TYPES_H so it avoids to declare again the
+ // types in arm_neon.h
kiranchandramohan wrote:
Consider revising this to say what is being done here and then provide the
reason.
https://github.com/llvm/llvm-project/pull/73258
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [Clang][AArch64] Add fix vector types to header into SVE (PR #73258)
@@ -2229,6 +2231,12 @@ void NeonEmitter::runHeader(raw_ostream &OS) {
static void emitNeonTypeDefs(const std::string& types, raw_ostream &OS) {
std::string TypedefTypes(types);
std::vector TDTypeVec = TypeSpec::fromTypeSpecs(TypedefTypes);
+ // arm_sve.h followed by arm_neon.h does not emmit these types
+ // because only arm_sve.h defines __ARM_NEON_TYPES_H
+ // arm_neon.h followed by arm_sve.h emmit these types
kiranchandramohan wrote:
```suggestion
// arm_neon.h followed by arm_sve.h emit these types
```
https://github.com/llvm/llvm-project/pull/73258
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang-tools-extra] [llvm] [clang] [flang] [flang] Pass to add frame pointer attribute (PR #74598)
https://github.com/kiranchandramohan edited https://github.com/llvm/llvm-project/pull/74598 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [llvm] [clang] [clang-tools-extra] [flang] Pass to add frame pointer attribute (PR #74598)
@@ -349,4 +349,24 @@ def VScaleAttr : Pass<"vscale-attr", "mlir::func::FuncOp">
{
let constructor = "::fir::createVScaleAttrPass()";
}
+def FunctionAttr : Pass<"function-attr", "mlir::func::FuncOp"> {
+ let summary = "This is a generic pass that adds function attributes expected
at LLVM IR level";
+ let description = [{ This feature introduces a general attribute aimed at
customizing function characteristics.
+ Options include:
+ Add "frame-pointer" attribute to functions: Set an attribute for the
frame
+ pointer on functions, to avoid saving the frame pointer in a register in
+ functions where it is unnecessary.
+ This eliminates the need for instructions to save, establish, and restore
+ frame pointers, while also freeing up an additional register in numerous
functions.
+ However, this approach can make debugging unfeasible on certain machines.
+ }];
kiranchandramohan wrote:
```suggestion
let description = [{ This feature introduces a general attribute aimed at
customizing function characteristics.
Options include:
Add "frame-pointer" attribute to functions: Set an attribute for the frame
pointer on functions, to avoid saving the frame pointer in a register in
functions where it is unnecessary. This eliminates the need for
instructions to save, establish, and restore frame pointers, while also
freeing up an additional register in numerous functions. However, this
approach can make debugging unfeasible on certain machines.
}];
```
https://github.com/llvm/llvm-project/pull/74598
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang-tools-extra] [flang] [llvm] [clang] [flang] Pass to add frame pointer attribute (PR #74598)
https://github.com/kiranchandramohan approved this pull request. LG. Please wait for @banach-space https://github.com/llvm/llvm-project/pull/74598 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [llvm] [clang] [clang-tools-extra] [flang] Pass to add frame pointer attribute (PR #74598)
@@ -349,4 +349,24 @@ def VScaleAttr : Pass<"vscale-attr", "mlir::func::FuncOp">
{
let constructor = "::fir::createVScaleAttrPass()";
}
+def FunctionAttr : Pass<"function-attr", "mlir::func::FuncOp"> {
+ let summary = "This is a generic pass that adds function attributes expected
at LLVM IR level";
kiranchandramohan wrote:
```suggestion
let summary = "Pass that adds function attributes expected at LLVM IR level";
```
https://github.com/llvm/llvm-project/pull/74598
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [clang] [Flang, Clang] Enable and test 'rdynamic' flag (PR #75598)
https://github.com/kiranchandramohan closed https://github.com/llvm/llvm-project/pull/75598 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[llvm] [clang-tools-extra] [clang] [flang] [flang] Pass to add frame pointer attribute (PR #74598)
@@ -245,6 +245,24 @@ static void
parseCodeGenArgs(Fortran::frontend::CodeGenOptions &opts,
opts.AliasAnalysis = opts.OptimizationLevel > 0;
+ if (const llvm::opt::Arg *a =
+ args.getLastArg(clang::driver::options::OPT_mframe_pointer_EQ)) {
+llvm::StringRef s = a->getValue();
+
+if (!(s == "none" || s == "non-leaf" || s == "all")) {
+ const auto debugWarningId = diags.getCustomDiagID(
+ clang::DiagnosticsEngine::Error, "Frame pointer: %0");
kiranchandramohan wrote:
@banach-space the test is in func-attr.f90
https://github.com/llvm/llvm-project/pull/74598/files#diff-1d4cea32c615e677fc20e34ffd2f5c3e0f84b00f5af6a692d787747e17bd9e3d
```
! RUN: not %flang_fc1 -triple aarch64-none-none -mframe-pointer=wrongval
-emit-llvm -o - %s 2>&1| FileCheck %s --check-prefix=CHECK-WRONGVALUEFP
```
https://github.com/llvm/llvm-project/pull/74598
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] Fix fveclib on Darwin (PR #77605)
https://github.com/kiranchandramohan approved this pull request. LG. Please add a test if possible. Assuming the issue was that the Argument was not `rendered` if it is `veclib`. https://github.com/llvm/llvm-project/pull/77605 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [flang][draft] Improve debug info generation. (PR #84202)
kiranchandramohan wrote: Thanks for this patch to improve debug support in Flang. If you are planning to work long-time on enabling full debug support then it would be good to write a design document (in flang/docs), get it agreed upon by everyone and then proceed. I attempted some time ago at https://reviews.llvm.org/D138534. Could you go through the discussions there? You could adopt a similar style document. @jeanPerier @vzakhari and @tblah might have opinions. There were some questions on how to generate debug info. This could be generated during lowering and carried in Ops. Alternatively, one of the motivations for adding `hlfir.declare` was to collect and retain variable information suitable for debug info generation. `fir.declare` also probably helps but it is probably dropped before we get to AddDebugFoundation. The AddDebugFoundation pass was added to retain some location information generation that was lost due to some core MLIR changes. It is not compulsory to retain it if there are better ways to generate debug information. https://github.com/llvm/llvm-project/pull/84202 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [flang] Enable polymorphic lowering by default (PR #83285)
https://github.com/kiranchandramohan approved this pull request. LGTM. https://github.com/llvm/llvm-project/pull/83285 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [llvm] [Frontend] Add leaf constructs and association to OpenMP/ACC directives (PR #83625)
https://github.com/kiranchandramohan approved this pull request. > Add members "leafs" and "association" to .td describing OpenMP/ACC > directives: "leafs" are the leaf constructs for composite/combined > constructs, and "association" is the source language construct to which the > directive applies (e.g. loop, block, etc.) Could you clarify that these are terminology used in the OpenMP standard? Great work @kparzysz. https://github.com/llvm/llvm-project/pull/83625 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [llvm] [Frontend] Add leaf constructs and association to OpenMP/ACC directives (PR #83625)
https://github.com/kiranchandramohan edited https://github.com/llvm/llvm-project/pull/83625 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [llvm] [Frontend] Add leaf constructs and association to OpenMP/ACC directives (PR #83625)
@@ -619,44 +595,22 @@ bool
clang::isOpenMPWorksharingDirective(OpenMPDirectiveKind DKind) {
}
bool clang::isOpenMPTaskLoopDirective(OpenMPDirectiveKind DKind) {
- return DKind == OMPD_taskloop || DKind == OMPD_taskloop_simd ||
- DKind == OMPD_master_taskloop || DKind == OMPD_master_taskloop_simd ||
- DKind == OMPD_parallel_master_taskloop ||
- DKind == OMPD_masked_taskloop || DKind == OMPD_masked_taskloop_simd ||
- DKind == OMPD_parallel_masked_taskloop ||
- DKind == OMPD_parallel_masked_taskloop_simd ||
- DKind == OMPD_parallel_master_taskloop_simd;
+ return DKind == OMPD_taskloop ||
+ llvm::is_contained(getLeafConstructs(DKind), OMPD_taskloop);
}
bool clang::isOpenMPParallelDirective(OpenMPDirectiveKind DKind) {
- return DKind == OMPD_parallel || DKind == OMPD_parallel_for ||
- DKind == OMPD_parallel_for_simd || DKind == OMPD_parallel_sections ||
- DKind == OMPD_target_parallel || DKind == OMPD_target_parallel_for ||
- DKind == OMPD_distribute_parallel_for ||
- DKind == OMPD_distribute_parallel_for_simd ||
- DKind == OMPD_target_parallel_for_simd ||
- DKind == OMPD_teams_distribute_parallel_for ||
- DKind == OMPD_teams_distribute_parallel_for_simd ||
- DKind == OMPD_target_teams_distribute_parallel_for ||
- DKind == OMPD_target_teams_distribute_parallel_for_simd ||
- DKind == OMPD_parallel_master || DKind == OMPD_parallel_masked ||
- DKind == OMPD_parallel_master_taskloop ||
- DKind == OMPD_parallel_master_taskloop_simd ||
- DKind == OMPD_parallel_masked_taskloop ||
- DKind == OMPD_parallel_masked_taskloop_simd ||
- DKind == OMPD_parallel_loop || DKind == OMPD_target_parallel_loop ||
- DKind == OMPD_teams_loop;
+ if (DKind == OMPD_parallel_workshare)
+return false;
kiranchandramohan wrote:
Nit: A comment that workshare is not applicable to C/C++ might be helpful.
https://github.com/llvm/llvm-project/pull/83625
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [flang]Add support for -moutline-atomics and -mno-outline-atomics (PR #78755)
@@ -672,7 +674,7 @@ void Flang::ConstructJob(Compilation &C, const JobAction
&JA,
CmdArgs.push_back(Args.MakeArgString(TripleStr));
if (isa(JA)) {
- CmdArgs.push_back("-E");
+CmdArgs.push_back("-E");
kiranchandramohan wrote:
Nit: accidental change.
https://github.com/llvm/llvm-project/pull/78755
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [flang]Add support for -moutline-atomics and -mno-outline-atomics (PR #78755)
@@ -2790,3 +2790,27 @@ void tools::addHIPRuntimeLibArgs(const ToolChain &TC, Compilation &C, } } } + +void tools::addOutlineAtomicsArgs(const Driver &D, const ToolChain &TC, const llvm::opt::ArgList &Args, llvm::opt::ArgStringList & CmdArgs, const llvm::Triple &Triple) kiranchandramohan wrote: There is probably a clang-format issue here. https://github.com/llvm/llvm-project/pull/78755 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [flang]Add support for -moutline-atomics and -mno-outline-atomics (PR #78755)
@@ -215,11 +215,15 @@ void addMachineOutlinerArgs(const Driver &D, const llvm::opt::ArgList &Args, void addOpenMPDeviceRTL(const Driver &D, const llvm::opt::ArgList &DriverArgs, llvm::opt::ArgStringList &CC1Args, StringRef BitcodeSuffix, const llvm::Triple &Triple); + +void addOutlineAtomicsArgs(const Driver &D, const ToolChain &TC, const llvm::opt::ArgList &Args, llvm::opt::ArgStringList &CmdArgs, const llvm::Triple &Triple); + } // end namespace tools } // end namespace driver } // end namespace clang clang::CodeGenOptions::FramePointerKind getFramePointerKind(const llvm::opt::ArgList &Args, const llvm::Triple &Triple); + kiranchandramohan wrote: Nit: accidental change? https://github.com/llvm/llvm-project/pull/78755 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [clang] [flang]Add support for -moutline-atomics and -mno-outline-atomics (PR #78755)
@@ -402,6 +402,21 @@ static void parseTargetArgs(TargetOptions &opts,
llvm::opt::ArgList &args) {
for (const llvm::opt::Arg *currentArg :
args.filtered(clang::driver::options::OPT_target_feature))
opts.featuresAsWritten.emplace_back(currentArg->getValue());
+
+ llvm::Triple targetTriple{llvm::Triple(opts.triple)};
+ if (const llvm::opt::Arg *A =
+ args.getLastArg(clang::driver::options::OPT_moutline_atomics,
+ clang::driver::options::OPT_mno_outline_atomics)) {
+// Option -moutline-atomics supported for AArch64 target only.
+if (targetTriple.isAArch64()) {
+ if (A->getOption().matches(
+ clang::driver::options::OPT_moutline_atomics)) {
+opts.featuresAsWritten.push_back("+outline-atomics");
+ } else {
+opts.featuresAsWritten.push_back("-outline-atomics");
+ }
+}
+ }
kiranchandramohan wrote:
Is this required? I did not see a corresponding change in Clang frontend driver.
https://github.com/llvm/llvm-project/pull/78755
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [flang]Add support for -moutline-atomics and -mno-outline-atomics (PR #78755)
@@ -215,11 +215,15 @@ void addMachineOutlinerArgs(const Driver &D, const llvm::opt::ArgList &Args, void addOpenMPDeviceRTL(const Driver &D, const llvm::opt::ArgList &DriverArgs, llvm::opt::ArgStringList &CC1Args, StringRef BitcodeSuffix, const llvm::Triple &Triple); + +void addOutlineAtomicsArgs(const Driver &D, const ToolChain &TC, const llvm::opt::ArgList &Args, llvm::opt::ArgStringList &CmdArgs, const llvm::Triple &Triple); + kiranchandramohan wrote: Nit: clang-format https://github.com/llvm/llvm-project/pull/78755 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [flang]Add support for -moutline-atomics and -mno-outline-atomics (PR #78755)
https://github.com/kiranchandramohan edited https://github.com/llvm/llvm-project/pull/78755 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [flang]Add support for -moutline-atomics and -mno-outline-atomics (PR #78755)
@@ -0,0 +1,18 @@ +! RUN: %flang -S -emit-llvm --target=aarch64-none-none -moutline-atomics -o - %s | FileCheck %s --check-prefixes=CHECKON,CHECKALL kiranchandramohan wrote: I think it is better if this is in the integration test directory. https://github.com/llvm/llvm-project/pull/78755 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [clang] [flang]Add support for -moutline-atomics and -mno-outline-atomics (PR #78755)
https://github.com/kiranchandramohan approved this pull request. LG. Please address or reply to comments before submitting. https://github.com/llvm/llvm-project/pull/78755 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang-tools-extra] [clang] [llvm] [flang] [Flang][OpenMP] Avoid default none errors for seq loop indices in par… (PR #76258)
https://github.com/kiranchandramohan updated
https://github.com/llvm/llvm-project/pull/76258
>From 03e611d413cffa6ff7432a8a88db5d6901449f3c Mon Sep 17 00:00:00 2001
From: Kiran Chandramohan
Date: Fri, 22 Dec 2023 18:35:00 +
Subject: [PATCH] [Flang][OpenMP] Avoid default none errors for seq loop
indices in parallel
---
flang/lib/Semantics/resolve-directives.cpp | 6 +-
flang/test/Semantics/OpenMP/resolve05.f90 | 13 +
2 files changed, 18 insertions(+), 1 deletion(-)
diff --git a/flang/lib/Semantics/resolve-directives.cpp
b/flang/lib/Semantics/resolve-directives.cpp
index da6c865ad56a3b..6e782204e6368c 100644
--- a/flang/lib/Semantics/resolve-directives.cpp
+++ b/flang/lib/Semantics/resolve-directives.cpp
@@ -1910,7 +1910,11 @@ void OmpAttributeVisitor::Post(const parser::Name &name)
{
if (Symbol * found{currScope().FindSymbol(name.source)}) {
if (symbol != found) {
name.symbol = found; // adjust the symbol within region
-} else if (GetContext().defaultDSA == Symbol::Flag::OmpNone) {
+} else if (GetContext().defaultDSA == Symbol::Flag::OmpNone &&
+// Exclude indices of sequential loops that are privatised in
+// the scope of the parallel region, and not in this scope.
+// TODO: check whether this should be caught in IsObjectWithDSA
+!symbol->test(Symbol::Flag::OmpPrivate)) {
context_.Say(name.source,
"The DEFAULT(NONE) clause requires that '%s' must be listed in "
"a data-sharing attribute clause"_err_en_US,
diff --git a/flang/test/Semantics/OpenMP/resolve05.f90
b/flang/test/Semantics/OpenMP/resolve05.f90
index 00f4860302183d..c4cebb48ac5c2b 100644
--- a/flang/test/Semantics/OpenMP/resolve05.f90
+++ b/flang/test/Semantics/OpenMP/resolve05.f90
@@ -17,7 +17,20 @@ subroutine default_none()
!$omp end parallel
end subroutine default_none
+! Test that indices of sequential loops are privatised and hence do not error
+! for DEFAULT(NONE)
+subroutine default_none_seq_loop
+ integer :: i
+
+ !$omp parallel do default(none)
+ do i = 1, 10
+ do j = 1, 20
+enddo
+ enddo
+end subroutine
+
program mm
call default_none()
+ call default_none_seq_loop()
!TODO: private, firstprivate, shared
end
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang-tools-extra] [clang] [llvm] [flang] [Flang][OpenMP] Avoid default none errors for seq loop indices in par… (PR #76258)
https://github.com/kiranchandramohan ready_for_review https://github.com/llvm/llvm-project/pull/76258 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang-tools-extra] [libcxx] [compiler-rt] [libc] [lldb] [libcxxabi] [flang] [lld] [clang] [llvm] [Flang][OpenMP] Push genEval calls to individual operations, NFC (PR #77758)
https://github.com/kiranchandramohan edited https://github.com/llvm/llvm-project/pull/77758 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang-tools-extra] [flang] [libcxx] [lld] [libcxxabi] [llvm] [lldb] [libc] [clang] [compiler-rt] [Flang][OpenMP] Push genEval calls to individual operations, NFC (PR #77758)
https://github.com/kiranchandramohan approved this pull request. LG. Please add a pointer to the discussion where it was agreed to pass a reference to the localSymbolTable. https://github.com/llvm/llvm-project/pull/77758 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[compiler-rt] [libcxxabi] [libcxx] [llvm] [libc] [clang-tools-extra] [lldb] [flang] [lld] [clang] [Flang][OpenMP] Push genEval calls to individual operations, NFC (PR #77758)
@@ -110,6 +110,34 @@ static void gatherFuncAndVarSyms(
}
}
+static Fortran::lower::pft::Evaluation *
+getCollapsedEval(Fortran::lower::pft::Evaluation &eval, int collapseValue) {
+ // Return the Evaluation of the innermost collapsed loop, or the current
+ // evaluation, if there is nothing to collapse.
+ if (collapseValue == 0)
+return &eval;
kiranchandramohan wrote:
Nit: Is it better to convert this to an assert (for > 0) and move this code to
the parent function?
https://github.com/llvm/llvm-project/pull/77758
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [flang][driver] Allow explicit specification of -lFortran_main (PR #78152)
@@ -1200,6 +1200,14 @@ static void addFortranMain(const ToolChain &TC, const
ArgList &Args,
// TODO: Find an equivalent of `--whole-archive` for Darwin and AIX.
if (!isWholeArchivePresent(Args) && !TC.getTriple().isMacOSX() &&
!TC.getTriple().isOSAIX()) {
+// Adding -lFortran_main with --whole-archive will create an error if the
+// user specifies -lFortran_main explicitly. Remove the user's
+// -lFortran_main arguments to avoid this (making sure -lFortran_main
+// behaves the same as -lFortranRuntime)
+llvm::erase_if(CmdArgs, [](const char *arg) {
+ return strcmp(arg, "-lFortran_main") == 0;
kiranchandramohan wrote:
Nit: I think you should use `strncmp` here.
https://github.com/llvm/llvm-project/pull/78152
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [clang] [clang-tools-extra] [llvm] [Flang][OpenMP] Avoid default none errors for seq loop indices in par… (PR #76258)
https://github.com/kiranchandramohan closed https://github.com/llvm/llvm-project/pull/76258 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[compiler-rt] [libc] [clang-tools-extra] [libcxx] [libcxxabi] [llvm] [lldb] [clang] [lld] [flang] [Flang][OpenMP] Handle SECTION construct from within SECTIONS (PR #77759)
kiranchandramohan wrote: > Introduce createSectionOp `genSectionOp`? https://github.com/llvm/llvm-project/pull/77759 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[compiler-rt] [libc] [clang-tools-extra] [libcxx] [libcxxabi] [llvm] [lldb] [clang] [lld] [flang] [Flang][OpenMP] Handle SECTION construct from within SECTIONS (PR #77759)
https://github.com/kiranchandramohan approved this pull request. LG. https://github.com/llvm/llvm-project/pull/77759 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [llvm] [clang] [libcxx] [mlir] [Flang][OpenMP] : Add a temporary lowering for workshare directive (PR #78268)
https://github.com/kiranchandramohan updated
https://github.com/llvm/llvm-project/pull/78268
>From 71c3449d872247e3af05de545e907407ac7ac9f9 Mon Sep 17 00:00:00 2001
From: Kiran Chandramohan
Date: Tue, 16 Jan 2024 12:45:23 +
Subject: [PATCH] [Flang][OpenMP] : Add a temporary lowering for workshare
directive
As a temporary solution, lower workshare to the single directive
---
flang/lib/Lower/OpenMP.cpp| 6 +-
flang/test/Lower/OpenMP/workshare.f90 | 16
2 files changed, 21 insertions(+), 1 deletion(-)
create mode 100644 flang/test/Lower/OpenMP/workshare.f90
diff --git a/flang/lib/Lower/OpenMP.cpp b/flang/lib/Lower/OpenMP.cpp
index 4f7c99a6d2b840..24531c9f70685f 100644
--- a/flang/lib/Lower/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP.cpp
@@ -3312,7 +3312,11 @@ genOMP(Fortran::lower::AbstractConverter &converter,
/*outerCombined=*/false);
break;
case llvm::omp::Directive::OMPD_workshare:
-TODO(currentLocation, "Workshare construct");
+// FIXME: Workshare is not a commonly used OpenMP construct, an
+// implementation for this feature will come later. For the codes
+// that use this construct, add a single construct for now.
+genSingleOp(converter, eval, currentLocation, beginClauseList,
+endClauseList);
break;
default: {
// Codegen for combined directives
diff --git a/flang/test/Lower/OpenMP/workshare.f90
b/flang/test/Lower/OpenMP/workshare.f90
new file mode 100644
index 00..7f63c99ab05965
--- /dev/null
+++ b/flang/test/Lower/OpenMP/workshare.f90
@@ -0,0 +1,16 @@
+
+!RUN: %flang_fc1 -emit-hlfir -fopenmp %s -o - | FileCheck %s
+
+!CHECK-LABEL: func @_QPsb
+subroutine sb(arr)
+ integer :: arr(:)
+!CHECK: omp.parallel {
+ !$omp parallel
+!CHECK: omp.single {
+ !$omp workshare
+arr = 0
+ !$omp end workshare
+!CHECK: }
+ !$omp end parallel
+!CHECK: }
+end subroutine
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [compiler-rt] [clang] [llvm] [clang-tools-extra] [Flang][OpenMP] Push genEval closer to leaf lowering functions (PR #77760)
https://github.com/kiranchandramohan approved this pull request. LG. https://github.com/llvm/llvm-project/pull/77760 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[libc] [flang] [compiler-rt] [clang-tools-extra] [llvm] [clang] [flang] use setsid to assign the child to prevent zombie as it will be clean up by init process (PR #77944)
kiranchandramohan wrote: Could you add a test? https://github.com/llvm/llvm-project/pull/77944 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[flang] [compiler-rt] [clang] [llvm] [clang-tools-extra] [Flang][OpenMP] Restructure recursive lowering in `createBodyOfOp` (PR #77761)
@@ -2186,11 +2178,43 @@ static void createBodyOfOp(
const llvm::SmallVector &args = {},
bool outerCombined = false, DataSharingProcessor *dsp = nullptr) {
fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
+
+ auto insertMarker = [](fir::FirOpBuilder &builder) {
+mlir::Value undef = builder.create(builder.getUnknownLoc(),
+ builder.getIndexType());
+return undef.getDefiningOp();
+ };
+
+ // Find the block where the OMP terminator should go. In simple cases
+ // it is the single block in the operation's region. When the region
+ // is more complicated, especially with unstructured control flow, there
+ // may be multiple blocks, and some of them may have non-OMP terminators
+ // resulting from lowering of the code contained within the operation.
+ // By OpenMP rules, there should be a single exit point from the region:
+ // here exit means transfering control to the code following the operation.
+ // STOP statement is allowed and does not count as exit for the purpose of
kiranchandramohan wrote:
Where is the STOP statement accounted for here? The code here looks like find
the block without a terminator and assert that there is only ONE.
https://github.com/llvm/llvm-project/pull/77761
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [compiler-rt] [flang] [llvm] [clang-tools-extra] [Flang][OpenMP] Restructure recursive lowering in `createBodyOfOp` (PR #77761)
@@ -2223,37 +2247,64 @@ static void createBodyOfOp(
mlir::omp::YieldOp>(
firOpBuilder, eval.getNestedEvaluations());
- // Insert the terminator.
- Fortran::lower::genOpenMPTerminator(firOpBuilder, op.getOperation(), loc);
- // Reset the insert point to before the terminator.
- resetBeforeTerminator(firOpBuilder, storeOp, block);
+ // Start with privatization, so that the lowering of the nested
+ // code will use the right symbols.
+ constexpr bool isLoop = std::is_same_v ||
+ std::is_same_v;
+ bool privatize = clauses && !outerCombined;
- // Handle privatization. Do not privatize if this is the outer operation.
- if (clauses && !outerCombined) {
-constexpr bool isLoop = std::is_same_v ||
-std::is_same_v;
+ firOpBuilder.setInsertionPoint(marker);
+ std::optional tempDsp;
+ if (privatize) {
if (!dsp) {
- DataSharingProcessor proc(converter, *clauses, eval);
- proc.processStep1();
- proc.processStep2(op, isLoop);
-} else {
- if (isLoop && args.size() > 0)
-dsp->setLoopIV(converter.getSymbolAddress(*args[0]));
- dsp->processStep2(op, isLoop);
+ tempDsp.emplace(converter, *clauses, eval);
+ tempDsp->processStep1();
}
-
-if (storeOp)
- firOpBuilder.setInsertionPointAfter(storeOp);
}
if constexpr (std::is_same_v) {
threadPrivatizeVars(converter, eval);
-if (clauses)
+if (clauses) {
+ firOpBuilder.setInsertionPoint(marker);
ClauseProcessor(converter, *clauses).processCopyin();
+}
}
- if (genNested)
+ if (genNested) {
+// genFIR(Evaluation&) tries to patch up unterminated blocks, causing
+// a lot of trouble if the terminator generation is delayed past this
+// point. Insert a temporary terminator here, then delete it.
+firOpBuilder.setInsertionPointToEnd(&op.getRegion().back());
+auto *temp = Fortran::lower::genOpenMPTerminator(firOpBuilder,
+ op.getOperation(), loc);
+firOpBuilder.setInsertionPointAfter(marker);
genNestedEvaluations(converter, eval);
+temp->erase();
+ }
+
+ if (auto *exitBlock = findExitBlock(op.getRegion())) {
+firOpBuilder.setInsertionPointToEnd(exitBlock);
+auto *term = Fortran::lower::genOpenMPTerminator(firOpBuilder,
+ op.getOperation(), loc);
+// Only insert lastprivate code when there actually is an exit block.
+// Such a block may not exist if the nested code produced an infinite
+// loop (this may not make sense in production code, but a user could
+// write that and we should handle it).
kiranchandramohan wrote:
Is it guaranteed that it is only when there is an infinite loop that no exit
block exists?
Also who/what inserts the terminator for this case (when there is infinite
loop)?
https://github.com/llvm/llvm-project/pull/77761
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[libc] [clang-tools-extra] [compiler-rt] [flang] [llvm] [clang] [flang] use setsid to assign the child to prevent zombie as it will be clean up by init process (PR #77944)
https://github.com/kiranchandramohan approved this pull request. LG. https://github.com/llvm/llvm-project/pull/77944 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang-tools-extra] [llvm] [compiler-rt] [clang] [flang] [Flang][OpenMP] Restructure recursive lowering in `createBodyOfOp` (PR #77761)
https://github.com/kiranchandramohan approved this pull request. LGTM. https://github.com/llvm/llvm-project/pull/77761 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[llvm] [clang] [compiler-rt] [flang] [clang-tools-extra] [Flang][OpenMP] Restructure recursive lowering in `createBodyOfOp` (PR #77761)
https://github.com/kiranchandramohan edited https://github.com/llvm/llvm-project/pull/77761 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[compiler-rt] [clang] [flang] [clang-tools-extra] [llvm] [Flang][OpenMP] Restructure recursive lowering in `createBodyOfOp` (PR #77761)
@@ -2223,37 +2247,64 @@ static void createBodyOfOp(
mlir::omp::YieldOp>(
firOpBuilder, eval.getNestedEvaluations());
- // Insert the terminator.
- Fortran::lower::genOpenMPTerminator(firOpBuilder, op.getOperation(), loc);
- // Reset the insert point to before the terminator.
- resetBeforeTerminator(firOpBuilder, storeOp, block);
+ // Start with privatization, so that the lowering of the nested
+ // code will use the right symbols.
+ constexpr bool isLoop = std::is_same_v ||
+ std::is_same_v;
+ bool privatize = clauses && !outerCombined;
- // Handle privatization. Do not privatize if this is the outer operation.
- if (clauses && !outerCombined) {
-constexpr bool isLoop = std::is_same_v ||
-std::is_same_v;
+ firOpBuilder.setInsertionPoint(marker);
+ std::optional tempDsp;
+ if (privatize) {
if (!dsp) {
- DataSharingProcessor proc(converter, *clauses, eval);
- proc.processStep1();
- proc.processStep2(op, isLoop);
-} else {
- if (isLoop && args.size() > 0)
-dsp->setLoopIV(converter.getSymbolAddress(*args[0]));
- dsp->processStep2(op, isLoop);
+ tempDsp.emplace(converter, *clauses, eval);
+ tempDsp->processStep1();
}
-
-if (storeOp)
- firOpBuilder.setInsertionPointAfter(storeOp);
}
if constexpr (std::is_same_v) {
threadPrivatizeVars(converter, eval);
-if (clauses)
+if (clauses) {
+ firOpBuilder.setInsertionPoint(marker);
ClauseProcessor(converter, *clauses).processCopyin();
+}
}
- if (genNested)
+ if (genNested) {
+// genFIR(Evaluation&) tries to patch up unterminated blocks, causing
+// a lot of trouble if the terminator generation is delayed past this
+// point. Insert a temporary terminator here, then delete it.
+firOpBuilder.setInsertionPointToEnd(&op.getRegion().back());
+auto *temp = Fortran::lower::genOpenMPTerminator(firOpBuilder,
+ op.getOperation(), loc);
+firOpBuilder.setInsertionPointAfter(marker);
genNestedEvaluations(converter, eval);
+temp->erase();
+ }
+
+ if (auto *exitBlock = findExitBlock(op.getRegion())) {
+firOpBuilder.setInsertionPointToEnd(exitBlock);
+auto *term = Fortran::lower::genOpenMPTerminator(firOpBuilder,
+ op.getOperation(), loc);
+// Only insert lastprivate code when there actually is an exit block.
+// Such a block may not exist if the nested code produced an infinite
+// loop (this may not make sense in production code, but a user could
+// write that and we should handle it).
kiranchandramohan wrote:
OK. Makes sense.
https://github.com/llvm/llvm-project/pull/77761
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[llvm] [flang] [compiler-rt] [clang] [clang-tools-extra] [Flang][OpenMP] Restructure recursive lowering in `createBodyOfOp` (PR #77761)
@@ -2223,37 +,100 @@ static void createBodyOfOp(
mlir::omp::YieldOp>(
firOpBuilder, eval.getNestedEvaluations());
- // Insert the terminator.
- Fortran::lower::genOpenMPTerminator(firOpBuilder, op.getOperation(), loc);
- // Reset the insert point to before the terminator.
- resetBeforeTerminator(firOpBuilder, storeOp, block);
+ // Start with privatization, so that the lowering of the nested
+ // code will use the right symbols.
+ constexpr bool isLoop = std::is_same_v ||
+ std::is_same_v;
+ bool privatize = clauses && !outerCombined;
- // Handle privatization. Do not privatize if this is the outer operation.
- if (clauses && !outerCombined) {
-constexpr bool isLoop = std::is_same_v ||
-std::is_same_v;
+ firOpBuilder.setInsertionPoint(marker);
+ std::optional tempDsp;
+ if (privatize) {
if (!dsp) {
- DataSharingProcessor proc(converter, *clauses, eval);
- proc.processStep1();
- proc.processStep2(op, isLoop);
-} else {
- if (isLoop && args.size() > 0)
-dsp->setLoopIV(converter.getSymbolAddress(*args[0]));
- dsp->processStep2(op, isLoop);
+ tempDsp.emplace(converter, *clauses, eval);
+ tempDsp->processStep1();
}
-
-if (storeOp)
- firOpBuilder.setInsertionPointAfter(storeOp);
}
if constexpr (std::is_same_v) {
threadPrivatizeVars(converter, eval);
-if (clauses)
+if (clauses) {
+ firOpBuilder.setInsertionPoint(marker);
ClauseProcessor(converter, *clauses).processCopyin();
+}
}
- if (genNested)
+ if (genNested) {
+// genFIR(Evaluation&) tries to patch up unterminated blocks, causing
+// a lot of trouble if the terminator generation is delayed past this
+// point. Insert a temporary terminator here, then delete it.
+firOpBuilder.setInsertionPointToEnd(&op.getRegion().back());
+auto *temp = Fortran::lower::genOpenMPTerminator(firOpBuilder,
+ op.getOperation(), loc);
+firOpBuilder.setInsertionPointAfter(marker);
genNestedEvaluations(converter, eval);
+temp->erase();
+ }
+
+ // Get or create a unique exiting block from the given region, or
+ // return nullptr if there is no exiting block.
+ auto getUniqueExit = [&](mlir::Region ®ion) -> mlir::Block * {
+// Find the blocks where the OMP terminator should go. In simple cases
+// it is the single block in the operation's region. When the region
+// is more complicated, especially with unstructured control flow, there
+// may be multiple blocks, and some of them may have non-OMP terminators
+// resulting from lowering of the code contained within the operation.
+// All the remaining blocks are potential exit points from the op's region.
+//
+// Explicit control flow cannot exit any OpenMP region (other than via
+// STOP), and that is enforced by semantic checks prior to lowering. STOP
+// statements are lowered to a function call.
+
+// Collect unterminated blocks.
+llvm::SmallVector exits;
+for (mlir::Block &b : region) {
+ if (b.empty() || !b.back().hasTrait())
+exits.push_back(&b);
+}
+
+if (exits.empty())
+ return nullptr;
+// If there already is a unique exiting block, do not create another one.
+// Additionally, some ops (e.g. omp.sections) require onlt 1 block in
kiranchandramohan wrote:
```suggestion
// Additionally, some ops (e.g. omp.sections) require only 1 block in
```
https://github.com/llvm/llvm-project/pull/77761
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [flang]Add support for -moutline-atomics and -mno-outline-atomics (PR #78755)
@@ -0,0 +1,11 @@ +! Test that flang-new forwards the -moutline-atomics and -mno-outline-atomics. +! RUN: %flang -moutline-atomics --target=aarch64-none-none -### %s -o %t 2>&1 | FileCheck %s +! CHECK: "-target-feature" "+outline-atomics" + +! RUN: %flang -mno-outline-atomics --target=aarch64-none-none -### %s -o %t 2>&1 | FileCheck %s --check-prefix=CHECK-NOOUTLINE +! CHECK-NOOUTLINE: "-target-feature" "-outline-atomics" + +! RUN: %flang -mno-outline-atomics --target=x86-none-none -### %s -o %t 2>&1 | FileCheck %s --check-prefix=CHECK-ERRMSG +! CHECK-ERRMSG: warning: 'x86' does not support '-mno-outline-atomics' kiranchandramohan wrote: Should there be a test to check the behaviour when no outline flags are specified and the target is `aarch64`? https://github.com/llvm/llvm-project/pull/78755 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [flang] [flang]Add support for -moutline-atomics and -mno-outline-atomics (PR #78755)
@@ -354,6 +354,27 @@ void Flang::addTargetOptions(const ArgList &Args,
CmdArgs.push_back(Args.MakeArgString(CPU));
}
+ if (Arg *A = Args.getLastArg(options::OPT_moutline_atomics,
+ options::OPT_mno_outline_atomics)) {
+// Option -moutline-atomics supported for AArch64 target only.
+if (!Triple.isAArch64()) {
+ D.Diag(diag::warn_drv_moutline_atomics_unsupported_opt)
+ << Triple.getArchName() << A->getOption().getName();
+} else {
+ if (A->getOption().matches(options::OPT_moutline_atomics)) {
+CmdArgs.push_back("-target-feature");
+CmdArgs.push_back("+outline-atomics");
+ } else {
+CmdArgs.push_back("-target-feature");
+CmdArgs.push_back("-outline-atomics");
+ }
+}
+ } else if (Triple.isAArch64() &&
+ getToolChain().IsAArch64OutlineAtomicsDefault(Args)) {
+CmdArgs.push_back("-target-feature");
+CmdArgs.push_back("+outline-atomics");
+ }
kiranchandramohan wrote:
Can this code be shared with `Clang` ?
https://github.com/llvm/llvm-project/pull/78755
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] c551ba0 - Run test only if X86 target is available
Author: Kiran Chandramohan Date: 2020-10-26T21:28:59Z New Revision: c551ba0e90bd2b49ef501d591f8362ba44e5484d URL: https://github.com/llvm/llvm-project/commit/c551ba0e90bd2b49ef501d591f8362ba44e5484d DIFF: https://github.com/llvm/llvm-project/commit/c551ba0e90bd2b49ef501d591f8362ba44e5484d.diff LOG: Run test only if X86 target is available This fixes failures in AArch64 buildbots by running the clang/test/CodeGen/X86/att-inline-asm-prefix.c only when the X86 target is available. Added: Modified: clang/test/CodeGen/X86/att-inline-asm-prefix.c Removed: diff --git a/clang/test/CodeGen/X86/att-inline-asm-prefix.c b/clang/test/CodeGen/X86/att-inline-asm-prefix.c index 9bd0d937d86e6..789e1f98a2a15 100644 --- a/clang/test/CodeGen/X86/att-inline-asm-prefix.c +++ b/clang/test/CodeGen/X86/att-inline-asm-prefix.c @@ -1,3 +1,5 @@ +// REQUIRES: x86-registered-target + // RUN:%clang_cc1 %s -ferror-limit 0 -triple=x86_64-pc -target-feature +avx512f -target-feature +avx2 -target-feature +avx512vl -S -o - | FileCheck %s -check-prefix CHECK // This test is to check if the prefix in inline assembly is correctly ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] a784de7 - [flang] Add -ffp-contract option processing
Author: Tom Eccles
Date: 2022-10-31T11:32:31Z
New Revision: a784de783af5096e593c5e214c2c78215fe303f5
URL:
https://github.com/llvm/llvm-project/commit/a784de783af5096e593c5e214c2c78215fe303f5
DIFF:
https://github.com/llvm/llvm-project/commit/a784de783af5096e593c5e214c2c78215fe303f5.diff
LOG: [flang] Add -ffp-contract option processing
Only add the option processing and store the result. No attributes are
added to FIR yet.
Only the "off" and "fast" options are supported. "fast-honor-pragmas" is not
applicable because we do not implement `#pragma clang fp contract()` in Fortran
[1]. "on" is not supported because it is unclear how to fuse only within
individual statements. gfortran also does not implement "on": treating it as an
"off".
Currently the default value is "off" to preserve existing behavior. gfortran
uses "fast" by default and that may be the right thing for flang-new after
further discussion in the future, but that can be changed separately.
gfortran's documentation is available [[
https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html | here ]].
[1]
https://clang.llvm.org/docs/LanguageExtensions.html#extensions-to-specify-floating-point-flags
Reviewed By: vzakhari, awarzynski
Differential Revision: https://reviews.llvm.org/D136080
Added:
flang/include/flang/Frontend/LangOptions.def
flang/include/flang/Frontend/LangOptions.h
flang/lib/Frontend/LangOptions.cpp
flang/test/Driver/flang_fp_opts.f90
Modified:
clang/include/clang/Basic/DiagnosticDriverKinds.td
clang/include/clang/Driver/Options.td
clang/lib/Driver/ToolChains/Flang.cpp
flang/include/flang/Frontend/CompilerInvocation.h
flang/lib/Frontend/CMakeLists.txt
flang/lib/Frontend/CompilerInvocation.cpp
flang/test/Driver/driver-help-hidden.f90
flang/test/Driver/driver-help.f90
flang/test/Driver/flang_f_opts.f90
flang/test/Driver/frontend-forwarding.f90
Removed:
diff --git a/clang/include/clang/Basic/DiagnosticDriverKinds.td
b/clang/include/clang/Basic/DiagnosticDriverKinds.td
index e477d93ba067f..44c277b3c3d8b 100644
--- a/clang/include/clang/Basic/DiagnosticDriverKinds.td
+++ b/clang/include/clang/Basic/DiagnosticDriverKinds.td
@@ -110,6 +110,9 @@ def
warn_drv_unsupported_option_for_offload_arch_req_feature : Warning<
def warn_drv_unsupported_option_for_target : Warning<
"ignoring '%0' option as it is not currently supported for target '%1'">,
InGroup;
+def warn_drv_unsupported_option_for_flang : Warning<
+ "the argument '%0' is not supported for option '%1'. Mapping to '%1%2'">,
+ InGroup;
def err_drv_invalid_thread_model_for_target : Error<
"invalid thread model '%0' in '%1' for this target">;
diff --git a/clang/include/clang/Driver/Options.td
b/clang/include/clang/Driver/Options.td
index 6d2c999873733..aa199e4608741 100644
--- a/clang/include/clang/Driver/Options.td
+++ b/clang/include/clang/Driver/Options.td
@@ -1919,12 +1919,14 @@ def fno_rounding_math : Flag<["-"],
"fno-rounding-math">, Group, Flags<
def ftrapping_math : Flag<["-"], "ftrapping-math">, Group;
def fno_trapping_math : Flag<["-"], "fno-trapping-math">, Group;
def ffp_contract : Joined<["-"], "ffp-contract=">, Group,
- Flags<[CC1Option]>, HelpText<"Form fused FP ops (e.g. FMAs):"
+ Flags<[CC1Option, FC1Option, FlangOption]>,
+ DocBrief<"Form fused FP ops (e.g. FMAs):"
" fast (fuses across statements disregarding pragmas)"
" | on (only fuses in the same statement unless dictated by pragmas)"
" | off (never fuses)"
" | fast-honor-pragmas (fuses across statements unless diectated by
pragmas)."
" Default is 'fast' for CUDA, 'fast-honor-pragmas' for HIP, and 'on'
otherwise.">,
+ HelpText<"Form fused FP ops (e.g. FMAs)">,
Values<"fast,on,off,fast-honor-pragmas">;
defm strict_float_cast_overflow : BoolFOption<"strict-float-cast-overflow",
diff --git a/clang/lib/Driver/ToolChains/Flang.cpp
b/clang/lib/Driver/ToolChains/Flang.cpp
index 964f0e2617047..6c6895da61299 100644
--- a/clang/lib/Driver/ToolChains/Flang.cpp
+++ b/clang/lib/Driver/ToolChains/Flang.cpp
@@ -80,6 +80,31 @@ void Flang::AddPicOptions(const ArgList &Args, ArgStringList
&CmdArgs) const {
}
}
+static void addFloatingPointOptions(const Driver &D, const ArgList &Args,
+ArgStringList &CmdArgs) {
+ StringRef FPContract;
+
+ if (const Arg *A = Args.getLastArg(options::OPT_ffp_contract)) {
+const StringRef Val = A->getValue();
+if (Val == "fast" || Val == "off") {
+ FPContract = Val;
+} else if (Val == "on") {
+ // Warn instead of error because users might have makefiles written for
+ // gfortran (which accepts -ffp-contract=on)
+ D.Diag(diag::warn_drv_unsupported_option_for_flang)
+ << Val << A->getOption().getName() << "off";
+ FPContract = "off";
+} else
+ // Clang's "fast-honor-pragmas" option is not suppo
[clang] cd86a03 - [clang,flang] Add help text for -fsyntax-only
Author: Alexander Malkov Date: 2022-08-19T09:54:29Z New Revision: cd86a03246a2df350c2f694457f1ce946cf65663 URL: https://github.com/llvm/llvm-project/commit/cd86a03246a2df350c2f694457f1ce946cf65663 DIFF: https://github.com/llvm/llvm-project/commit/cd86a03246a2df350c2f694457f1ce946cf65663.diff LOG: [clang,flang] Add help text for -fsyntax-only Fix for the problem with displaying options `-fsyntax-only` in clang and flang-new in help Fix https://github.com/llvm/llvm-project/issues/57033 Before: ``` $ clang -help | grep syntax -objcmt-migrate-property-dot-syntax Enable migration of setter/getter messages to property-dot syntax ``` After: ``` $ clang -help | grep syntax -fsyntax-only Run the preprocessor, parser and semantic analysis stages -objcmt-migrate-property-dot-syntax Enable migration of setter/getter messages to property-dot syntax ``` Reviewed By: vzakhari, awarzynski, MaskRay, alexiprof Differential Revision: https://reviews.llvm.org/D131808 Added: Modified: clang/docs/ClangCommandLineReference.rst clang/docs/CommandGuide/clang.rst clang/docs/UsersManual.rst clang/include/clang/Driver/Options.td flang/test/Driver/driver-help-hidden.f90 flang/test/Driver/driver-help.f90 Removed: diff --git a/clang/docs/ClangCommandLineReference.rst b/clang/docs/ClangCommandLineReference.rst index 265a6d7beb6e9..7f9ef3783f9da 100644 --- a/clang/docs/ClangCommandLineReference.rst +++ b/clang/docs/ClangCommandLineReference.rst @@ -831,6 +831,8 @@ Only run the driver. .. option:: -fsyntax-only +Run the preprocessor, parser and semantic analysis stages + .. option:: -module-file-info Provide information about a particular module file diff --git a/clang/docs/CommandGuide/clang.rst b/clang/docs/CommandGuide/clang.rst index af2b0df98d3dc..5e344ec702777 100644 --- a/clang/docs/CommandGuide/clang.rst +++ b/clang/docs/CommandGuide/clang.rst @@ -75,7 +75,7 @@ Stage Selection Options .. option:: -fsyntax-only - Run the preprocessor, parser and type checking stages. + Run the preprocessor, parser and semantic analysis stages. .. option:: -S diff --git a/clang/docs/UsersManual.rst b/clang/docs/UsersManual.rst index f9ccca65f3889..78ccf4572ffc8 100644 --- a/clang/docs/UsersManual.rst +++ b/clang/docs/UsersManual.rst @@ -3999,7 +3999,7 @@ Execute ``clang-cl /?`` to see a list of supported options: /Zl Don't mention any default libraries in the object file /Zp Set the default maximum struct packing alignment to 1 /Zp Specify the default maximum struct packing alignment - /Zs Syntax-check only + /Zs Run the preprocessor, parser and semantic analysis stages OPTIONS: -###Print (but do not run) the commands to run for this compilation @@ -4130,6 +4130,7 @@ Execute ``clang-cl /?`` to see a list of supported options: behavior. See user manual for available checks -fsplit-lto-unitEnables splitting of the LTO unit. -fstandalone-debug Emit full debug info for all types used by the program + -fsyntax-only Run the preprocessor, parser and semantic analysis stages -fwhole-program-vtables Enables whole-program vtable optimization. Requires -flto -gcodeview-ghashEmit type record hashes in a .debug$H section -gcodeview Generate CodeView debug information diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td index a5dfeec4fd463..c8178c13e82e9 100644 --- a/clang/include/clang/Driver/Options.td +++ b/clang/include/clang/Driver/Options.td @@ -2800,7 +2800,8 @@ def fstrict_overflow : Flag<["-"], "fstrict-overflow">, Group; def fdriver_only : Flag<["-"], "fdriver-only">, Flags<[NoXarchOption, CoreOption]>, Group, HelpText<"Only run the driver.">; def fsyntax_only : Flag<["-"], "fsyntax-only">, - Flags<[NoXarchOption,CoreOption,CC1Option,FC1Option]>, Group; + Flags<[NoXarchOption,CoreOption,CC1Option,FC1Option,FlangOption]>, Group, + HelpText<"Run the preprocessor, parser and semantic analysis stages">; def ftabstop_EQ : Joined<["-"], "ftabstop=">, Group; def ftemplate_depth_EQ : Joined<["-"], "ftemplate-depth=">, Group; def ftemplate_depth_ : Joined<["-"], "ftemplate-depth-">, Group; @@ -6618,7 +6619,7 @@ def _SLASH_Zp : CLJoined<"Zp">, def _SLASH_Zp_flag : CLFlag<"Zp">, HelpText<"Set default maximum struct packing alignment to 1">, Alias, AliasArgs<["1"]>; -def _SLASH_Zs : CLFlag<"Zs">, HelpText<"Syntax-check only">, +def _SLASH_Zs : CLFlag<"Zs">, HelpText<"Run the preprocessor, parser and semantic analysis stages">, Alias; def _SLASH_openmp_ : CLFlag<"openmp-">, HelpText<"Disable OpenMP support">, Alias; diff --git
[libunwind] [mlir][OpenMP] Added translation for `omp.teams` to LLVM IR (PR #68042)
@@ -0,0 +1,136 @@
+// RUN: mlir-translate -mlir-to-llvmir -split-input-file %s | FileCheck %s
+
+llvm.func @foo()
+
+// CHECK-LABEL: @omp_teams_simple
+// CHECK: call void {{.*}} @__kmpc_fork_teams(ptr @{{.+}}, i32 0, ptr
[[wrapperfn:.+]])
kiranchandramohan wrote:
Could you convert the captured variables (eg. wrapperfn) to caps? This is to
distinguish them easily from code.
https://github.com/llvm/llvm-project/pull/68042
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [mlir][OpenMP] Added translation for `omp.teams` to LLVM IR (PR #68042)
https://github.com/kiranchandramohan approved this pull request. LGTM. Will the wrapper function stay or be removed? https://github.com/llvm/llvm-project/pull/68042 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[libunwind] [mlir][OpenMP] Added translation for `omp.teams` to LLVM IR (PR #68042)
https://github.com/kiranchandramohan approved this pull request. LGTM. Will the wrapper function stay or be removed? https://github.com/llvm/llvm-project/pull/68042 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [mlir][OpenMP] Added translation for `omp.teams` to LLVM IR (PR #68042)
@@ -0,0 +1,136 @@
+// RUN: mlir-translate -mlir-to-llvmir -split-input-file %s | FileCheck %s
+
+llvm.func @foo()
+
+// CHECK-LABEL: @omp_teams_simple
+// CHECK: call void {{.*}} @__kmpc_fork_teams(ptr @{{.+}}, i32 0, ptr
[[wrapperfn:.+]])
kiranchandramohan wrote:
Could you convert the captured variables (eg. wrapperfn) to caps? This is to
distinguish them easily from code.
https://github.com/llvm/llvm-project/pull/68042
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang-tools-extra] [OpenMPIRBuilder] Remove wrapper function in `createTask`, `createTeams` (PR #67723)
https://github.com/kiranchandramohan edited https://github.com/llvm/llvm-project/pull/67723 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang-tools-extra] [OpenMPIRBuilder] Remove wrapper function in `createTask`, `createTeams` (PR #67723)
@@ -5748,6 +5758,7 @@ OpenMPIRBuilder::createTeams(const LocationDescription &Loc, BasicBlock *BodyBB = splitBB(Builder, /*CreateBranch=*/true, "teams.entry"); Builder.SetInsertPoint(BodyBB, BodyBB->begin()); } + InsertPointTy OuterAllocaIP(&OuterAllocaBB, OuterAllocaBB.begin()); kiranchandramohan wrote: Can this be defined close to its use? https://github.com/llvm/llvm-project/pull/67723 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [OpenMPIRBuilder] Remove wrapper function in `createTask`, `createTeams` (PR #67723)
@@ -5748,6 +5758,7 @@ OpenMPIRBuilder::createTeams(const LocationDescription &Loc, BasicBlock *BodyBB = splitBB(Builder, /*CreateBranch=*/true, "teams.entry"); Builder.SetInsertPoint(BodyBB, BodyBB->begin()); } + InsertPointTy OuterAllocaIP(&OuterAllocaBB, OuterAllocaBB.begin()); kiranchandramohan wrote: Can this be defined close to its use? https://github.com/llvm/llvm-project/pull/67723 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang-tools-extra] [OpenMPIRBuilder] Remove wrapper function in `createTask`, `createTeams` (PR #67723)
@@ -340,6 +340,44 @@ BasicBlock *llvm::splitBBWithSuffix(IRBuilderBase
&Builder, bool CreateBranch,
return splitBB(Builder, CreateBranch, Old->getName() + Suffix);
}
+// This function creates a fake integer value and a fake use for the integer
+// value. It returns the fake value created. This is useful in modeling the
+// extra arguments to the outlined functions.
+Value *createFakeIntVal(IRBuilder<> &Builder,
+OpenMPIRBuilder::InsertPointTy OuterAllocaIP,
+std::stack &ToBeDeleted,
+OpenMPIRBuilder::InsertPointTy InnerAllocaIP,
+const Twine &Name = "", bool AsPtr = true) {
+ Builder.restoreIP(OuterAllocaIP);
+ Instruction *FakeVal;
+ AllocaInst *FakeValAddr =
+ Builder.CreateAlloca(Builder.getInt32Ty(), nullptr, Name + ".addr");
+ ToBeDeleted.push(FakeValAddr);
+
+ if (AsPtr)
+FakeVal = FakeValAddr;
+ else {
+FakeVal =
+Builder.CreateLoad(Builder.getInt32Ty(), FakeValAddr, Name + ".val");
+ToBeDeleted.push(FakeVal);
kiranchandramohan wrote:
Would this delete twice for the `AsPtr` case?
https://github.com/llvm/llvm-project/pull/67723
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[libunwind] [OpenMPIRBuilder] Remove wrapper function in `createTask`, `createTeams` (PR #67723)
@@ -340,6 +340,44 @@ BasicBlock *llvm::splitBBWithSuffix(IRBuilderBase
&Builder, bool CreateBranch,
return splitBB(Builder, CreateBranch, Old->getName() + Suffix);
}
+// This function creates a fake integer value and a fake use for the integer
+// value. It returns the fake value created. This is useful in modeling the
+// extra arguments to the outlined functions.
+Value *createFakeIntVal(IRBuilder<> &Builder,
+OpenMPIRBuilder::InsertPointTy OuterAllocaIP,
+std::stack &ToBeDeleted,
+OpenMPIRBuilder::InsertPointTy InnerAllocaIP,
+const Twine &Name = "", bool AsPtr = true) {
+ Builder.restoreIP(OuterAllocaIP);
+ Instruction *FakeVal;
+ AllocaInst *FakeValAddr =
+ Builder.CreateAlloca(Builder.getInt32Ty(), nullptr, Name + ".addr");
+ ToBeDeleted.push(FakeValAddr);
+
+ if (AsPtr)
+FakeVal = FakeValAddr;
+ else {
kiranchandramohan wrote:
Nit: braces to match else. Same for tid below.
https://github.com/llvm/llvm-project/pull/67723
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [OpenMPIRBuilder] Remove wrapper function in `createTask`, `createTeams` (PR #67723)
https://github.com/kiranchandramohan commented: Could you expand the summary to describe the changes. What are the changes required to remove the wrapper function (Why was it required in the first place?) Why are the fake vals necessary? https://github.com/llvm/llvm-project/pull/67723 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[libunwind] [OpenMPIRBuilder] Remove wrapper function in `createTask`, `createTeams` (PR #67723)
https://github.com/kiranchandramohan edited https://github.com/llvm/llvm-project/pull/67723 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang] [OpenMPIRBuilder] Remove wrapper function in `createTask`, `createTeams` (PR #67723)
@@ -340,6 +340,44 @@ BasicBlock *llvm::splitBBWithSuffix(IRBuilderBase
&Builder, bool CreateBranch,
return splitBB(Builder, CreateBranch, Old->getName() + Suffix);
}
+// This function creates a fake integer value and a fake use for the integer
+// value. It returns the fake value created. This is useful in modeling the
+// extra arguments to the outlined functions.
+Value *createFakeIntVal(IRBuilder<> &Builder,
+OpenMPIRBuilder::InsertPointTy OuterAllocaIP,
+std::stack &ToBeDeleted,
+OpenMPIRBuilder::InsertPointTy InnerAllocaIP,
+const Twine &Name = "", bool AsPtr = true) {
+ Builder.restoreIP(OuterAllocaIP);
+ Instruction *FakeVal;
+ AllocaInst *FakeValAddr =
+ Builder.CreateAlloca(Builder.getInt32Ty(), nullptr, Name + ".addr");
+ ToBeDeleted.push(FakeValAddr);
+
+ if (AsPtr)
+FakeVal = FakeValAddr;
+ else {
+FakeVal =
+Builder.CreateLoad(Builder.getInt32Ty(), FakeValAddr, Name + ".val");
+ToBeDeleted.push(FakeVal);
kiranchandramohan wrote:
Would this delete twice for the `AsPtr` case?
https://github.com/llvm/llvm-project/pull/67723
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[libunwind] [OpenMPIRBuilder] Remove wrapper function in `createTask`, `createTeams` (PR #67723)
@@ -5748,6 +5758,7 @@ OpenMPIRBuilder::createTeams(const LocationDescription &Loc, BasicBlock *BodyBB = splitBB(Builder, /*CreateBranch=*/true, "teams.entry"); Builder.SetInsertPoint(BodyBB, BodyBB->begin()); } + InsertPointTy OuterAllocaIP(&OuterAllocaBB, OuterAllocaBB.begin()); kiranchandramohan wrote: Can this be defined close to its use? https://github.com/llvm/llvm-project/pull/67723 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
[clang-tools-extra] [OpenMPIRBuilder] Remove wrapper function in `createTask`, `createTeams` (PR #67723)
@@ -340,6 +340,44 @@ BasicBlock *llvm::splitBBWithSuffix(IRBuilderBase
&Builder, bool CreateBranch,
return splitBB(Builder, CreateBranch, Old->getName() + Suffix);
}
+// This function creates a fake integer value and a fake use for the integer
+// value. It returns the fake value created. This is useful in modeling the
+// extra arguments to the outlined functions.
+Value *createFakeIntVal(IRBuilder<> &Builder,
+OpenMPIRBuilder::InsertPointTy OuterAllocaIP,
+std::stack &ToBeDeleted,
+OpenMPIRBuilder::InsertPointTy InnerAllocaIP,
+const Twine &Name = "", bool AsPtr = true) {
+ Builder.restoreIP(OuterAllocaIP);
+ Instruction *FakeVal;
+ AllocaInst *FakeValAddr =
+ Builder.CreateAlloca(Builder.getInt32Ty(), nullptr, Name + ".addr");
+ ToBeDeleted.push(FakeValAddr);
+
+ if (AsPtr)
+FakeVal = FakeValAddr;
+ else {
kiranchandramohan wrote:
Nit: braces to match else. Same for tid below.
https://github.com/llvm/llvm-project/pull/67723
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
