scott.linder created this revision.
Herald added a project: clang.
Herald added a subscriber: cfe-commits.
The HIP toolchain invokes `llc` without an explicit opt-level, meaning
it always uses the default (-O2). This makes it impossible to use -O1,
for example. The HIP toolchain also coerces -Os/-Oz to -O2 even when
invoking opt, and it coerces -Og to -O2 rather than -O1.
Forward the opt-level to `llc` as well as `opt`, and only coerce levels
where it is required.
Repository:
rG LLVM Github Monorepo
https://reviews.llvm.org/D70987
Files:
clang/lib/Driver/ToolChains/HIP.cpp
clang/test/Driver/hip-toolchain-opt.hip
Index: clang/test/Driver/hip-toolchain-opt.hip
===================================================================
--- /dev/null
+++ clang/test/Driver/hip-toolchain-opt.hip
@@ -0,0 +1,101 @@
+// REQUIRES: clang-driver
+// REQUIRES: x86-registered-target
+// REQUIRES: amdgpu-registered-target
+
+// RUN: %clang -### \
+// RUN: -target x86_64-unknown-linux-gnu \
+// RUN: -x hip --cuda-gpu-arch=gfx900 \
+// RUN: -c -nogpulib \
+// RUN: %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck --check-prefixes=ALL,DEFAULT %s
+
+// RUN: %clang -### -O0 \
+// RUN: -target x86_64-unknown-linux-gnu \
+// RUN: -x hip --cuda-gpu-arch=gfx900 \
+// RUN: -c -nogpulib \
+// RUN: %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck --check-prefixes=ALL,O0 %s
+
+// RUN: %clang -### -O1 \
+// RUN: -target x86_64-unknown-linux-gnu \
+// RUN: -x hip --cuda-gpu-arch=gfx900 \
+// RUN: -c -nogpulib \
+// RUN: %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck --check-prefixes=ALL,O1 %s
+
+// RUN: %clang -### -O2 \
+// RUN: -target x86_64-unknown-linux-gnu \
+// RUN: -x hip --cuda-gpu-arch=gfx900 \
+// RUN: -c -nogpulib \
+// RUN: %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck --check-prefixes=ALL,O2 %s
+
+// RUN: %clang -### -O3 \
+// RUN: -target x86_64-unknown-linux-gnu \
+// RUN: -x hip --cuda-gpu-arch=gfx900 \
+// RUN: -c -nogpulib \
+// RUN: %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck --check-prefixes=ALL,O3 %s
+
+// RUN: %clang -### -Os \
+// RUN: -target x86_64-unknown-linux-gnu \
+// RUN: -x hip --cuda-gpu-arch=gfx900 \
+// RUN: -c -nogpulib \
+// RUN: %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck --check-prefixes=ALL,Os %s
+
+// RUN: %clang -### -Oz \
+// RUN: -target x86_64-unknown-linux-gnu \
+// RUN: -x hip --cuda-gpu-arch=gfx900 \
+// RUN: -c -nogpulib \
+// RUN: %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck --check-prefixes=ALL,Oz %s
+
+// RUN: %clang -### -Og \
+// RUN: -target x86_64-unknown-linux-gnu \
+// RUN: -x hip --cuda-gpu-arch=gfx900 \
+// RUN: -c -nogpulib \
+// RUN: %S/Inputs/hip_multiple_inputs/b.hip \
+// RUN: 2>&1 | FileCheck --check-prefixes=ALL,Og %s
+
+// ALL: "{{.*}}clang{{.*}}" "-cc1" "-triple" "amdgcn-amd-amdhsa"
+// DEFAULT-NOT: "-O{{.}}"
+// O0-SAME: "-O0"
+// O1-SAME: "-O1"
+// O2-SAME: "-O2"
+// O3-SAME: "-O3"
+// Os-SAME: "-Os"
+// Oz-SAME: "-Oz"
+// Og-SAME: "-Og"
+
+// ALL: "{{.*}}opt"
+// DEFAULT-NOT: "-O{{.}}"
+// O0-SAME: "-O0"
+// O1-SAME: "-O1"
+// O2-SAME: "-O2"
+// O3-SAME: "-O3"
+// Os-SAME: "-Os"
+// Oz-SAME: "-Oz"
+// Og-SAME: "-O1"
+// ALL-SAME: "-mtriple=amdgcn-amd-amdhsa"
+
+// ALL: "{{.*}}llc"
+// DEFAULT-NOT: "-O{{.}}"
+// O0-SAME: "-O0"
+// O1-SAME: "-O1"
+// O2-SAME: "-O2"
+// O3-SAME: "-O3"
+// Os-SAME: "-O2"
+// Oz-SAME: "-O2"
+// Og-SAME: "-O1"
+// ALL-SAME: "-mtriple=amdgcn-amd-amdhsa"
+
+// ALL: "{{.*}}clang{{.*}}" "-cc1" "-triple" "x86_64-unknown-linux-gnu"
+// DEFAULT-NOT: "-O{{.}}"
+// O0-SAME: "-O0"
+// O1-SAME: "-O1"
+// O2-SAME: "-O2"
+// O3-SAME: "-O3"
+// Os-SAME: "-Os"
+// Oz-SAME: "-Oz"
+// Og-SAME: "-Og"
Index: clang/lib/Driver/ToolChains/HIP.cpp
===================================================================
--- clang/lib/Driver/ToolChains/HIP.cpp
+++ clang/lib/Driver/ToolChains/HIP.cpp
@@ -62,6 +62,34 @@
}
return OutputFileName;
}
+
+static void addOptLevelArgs(const llvm::opt::ArgList &Args,
+ llvm::opt::ArgStringList &CmdArgs,
+ bool IsLlc = false) {
+ if (Arg *A = Args.getLastArg(options::OPT_O_Group)) {
+ StringRef OOpt = "3";
+ if (A->getOption().matches(options::OPT_O4) ||
+ A->getOption().matches(options::OPT_Ofast))
+ OOpt = "3";
+ else if (A->getOption().matches(options::OPT_O0))
+ OOpt = "0";
+ else if (A->getOption().matches(options::OPT_O)) {
+ // Clang and opt support -Os/-Oz; llc only supports -O0, -O1, -O2 and -O3
+ // so we map -Os/-Oz to -O2.
+ // Only clang supports -Og, and maps it to -O1.
+ // We map anything else to -O2.
+ OOpt = llvm::StringSwitch<const char *>(A->getValue())
+ .Case("1", "1")
+ .Case("2", "2")
+ .Case("3", "3")
+ .Case("s", IsLlc ? "2" : "s")
+ .Case("z", IsLlc ? "2" : "z")
+ .Case("g", "1")
+ .Default("2");
+ }
+ CmdArgs.push_back(Args.MakeArgString("-O" + OOpt));
+ }
+}
} // namespace
const char *AMDGCN::Linker::constructLLVMLinkCommand(
@@ -93,25 +121,7 @@
// The input to opt is the output from llvm-link.
OptArgs.push_back(InputFileName);
// Pass optimization arg to opt.
- if (Arg *A = Args.getLastArg(options::OPT_O_Group)) {
- StringRef OOpt = "3";
- if (A->getOption().matches(options::OPT_O4) ||
- A->getOption().matches(options::OPT_Ofast))
- OOpt = "3";
- else if (A->getOption().matches(options::OPT_O0))
- OOpt = "0";
- else if (A->getOption().matches(options::OPT_O)) {
- // -Os, -Oz, and -O(anything else) map to -O2
- OOpt = llvm::StringSwitch<const char *>(A->getValue())
- .Case("1", "1")
- .Case("2", "2")
- .Case("3", "3")
- .Case("s", "2")
- .Case("z", "2")
- .Default("2");
- }
- OptArgs.push_back(Args.MakeArgString("-O" + OOpt));
- }
+ addOptLevelArgs(Args, OptArgs);
OptArgs.push_back("-mtriple=amdgcn-amd-amdhsa");
OptArgs.push_back(Args.MakeArgString("-mcpu=" + SubArchName));
@@ -136,10 +146,15 @@
llvm::StringRef OutputFilePrefix, const char *InputFileName,
bool OutputIsAsm) const {
// Construct llc command.
- ArgStringList LlcArgs{
- InputFileName, "-mtriple=amdgcn-amd-amdhsa",
- Args.MakeArgString(Twine("-filetype=") + (OutputIsAsm ? "asm" : "obj")),
- Args.MakeArgString("-mcpu=" + SubArchName)};
+ ArgStringList LlcArgs;
+ // The input to llc is the output from opt.
+ LlcArgs.push_back(InputFileName);
+ // Pass optimization arg to llc.
+ addOptLevelArgs(Args, LlcArgs, true);
+ LlcArgs.push_back("-mtriple=amdgcn-amd-amdhsa");
+ LlcArgs.push_back(Args.MakeArgString("-mcpu=" + SubArchName));
+ LlcArgs.push_back(
+ Args.MakeArgString(Twine("-filetype=") + (OutputIsAsm ? "asm" : "obj")));
// Extract all the -m options
std::vector<llvm::StringRef> Features;
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
