[AMDGPU][HIP] Improve opt-level handling

Summary:
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.

Subscribers: cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D70987
This commit is contained in:
Scott Linder 2019-12-03 17:18:21 -05:00
parent 3c6b5d3674
commit d96ea47c75
5 changed files with 156 additions and 32 deletions

View File

@ -62,6 +62,34 @@ static const char *getOutputFileName(Compilation &C, StringRef Base,
}
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 @@ const char *AMDGCN::Linker::constructOptCommand(
// 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 @@ const char *AMDGCN::Linker::constructLlcCommand(
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, /*IsLlc=*/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;

View File

@ -18,8 +18,9 @@
// CHECK-SAME: "-o" [[OPT_803_BC:".*-gfx803-optimized.*bc"]]
// CHECK: [[LLC: ".*llc"]] [[OPT_803_BC]]
// CHECK-SAME: "-mtriple=amdgcn-amd-amdhsa" "-filetype=obj"
// CHECK-SAME: "-mtriple=amdgcn-amd-amdhsa"
// CHECK-SAME: {{.*}} "-mcpu=gfx803"
// CHECK-SAME: "-filetype=obj"
// CHECK-SAME: "-amdgpu-function-calls=0" "-o" {{".*-gfx803-.*o"}}
// CHECK: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa"
@ -33,6 +34,7 @@
// CHECK-SAME: "-o" [[OPT_900_BC:".*-gfx900-optimized.*bc"]]
// CHECK: [[LLC]] [[OPT_900_BC]]
// CHECK-SAME: "-mtriple=amdgcn-amd-amdhsa" "-filetype=obj"
// CHECK-SAME: "-mtriple=amdgcn-amd-amdhsa"
// CHECK-SAME: {{.*}} "-mcpu=gfx900"
// CHECk-SAME: "-filetype=obj"
// CHECK-SAME: "-amdgpu-function-calls=0" "-o" {{".*-gfx900-.*o"}}

View File

@ -34,8 +34,9 @@
// CHECK-SAME: "-o" [[OPT_BC_DEV_A_803:".*-gfx803-optimized.*bc"]]
// CHECK: [[LLC: ".*llc"]] [[OPT_BC_DEV_A_803]] "-mtriple=amdgcn-amd-amdhsa"
// CHECK-SAME: "-mcpu=gfx803"
// CHECK-SAME: "-filetype=obj"
// CHECK-SAME: "-mcpu=gfx803" "-o" [[OBJ_DEV_A_803:".*-gfx803-.*o"]]
// CHECK-SAME: "-o" [[OBJ_DEV_A_803:".*-gfx803-.*o"]]
// CHECK: [[LLD: ".*lld"]] "-flavor" "gnu" "-shared"
// CHECK-SAME: "-o" "[[IMG_DEV_A_803:.*out]]" [[OBJ_DEV_A_803]]
@ -62,8 +63,9 @@
// CHECK-SAME: "-o" [[OPT_BC_DEV_A_900:".*-gfx900-optimized.*bc"]]
// CHECK: [[LLC: ".*llc"]] [[OPT_BC_DEV_A_900]] "-mtriple=amdgcn-amd-amdhsa"
// CHECK-SAME: "-mcpu=gfx900"
// CHECK-SAME: "-filetype=obj"
// CHECK-SAME: "-mcpu=gfx900" "-o" [[OBJ_DEV_A_900:".*-gfx900-.*o"]]
// CHECK-SAME: "-o" [[OBJ_DEV_A_900:".*-gfx900-.*o"]]
// CHECK: [[LLD: ".*lld"]] "-flavor" "gnu" "-shared"
// CHECK-SAME: "-o" "[[IMG_DEV_A_900:.*out]]" [[OBJ_DEV_A_900]]
@ -106,8 +108,9 @@
// CHECK-SAME: "-o" [[OPT_BC_DEV_B_803:".*-gfx803-optimized.*bc"]]
// CHECK: [[LLC: ".*llc"]] [[OPT_BC_DEV_B_803]] "-mtriple=amdgcn-amd-amdhsa"
// CHECK-SAME: "-mcpu=gfx803"
// CHECK-SAME: "-filetype=obj"
// CHECK-SAME: "-mcpu=gfx803" "-o" [[OBJ_DEV_B_803:".*-gfx803-.*o"]]
// CHECK-SAME: "-o" [[OBJ_DEV_B_803:".*-gfx803-.*o"]]
// CHECK: [[LLD: ".*lld"]] "-flavor" "gnu" "-shared"
// CHECK-SAME: "-o" "[[IMG_DEV_B_803:.*out]]" [[OBJ_DEV_B_803]]
@ -134,8 +137,9 @@
// CHECK-SAME: "-o" [[OPT_BC_DEV_B_900:".*-gfx900-optimized.*bc"]]
// CHECK: [[LLC: ".*llc"]] [[OPT_BC_DEV_B_900]] "-mtriple=amdgcn-amd-amdhsa"
// CHECk-SAME: "-mcpu=gfx900"
// CHECK-SAME: "-filetype=obj"
// CHECK-SAME: "-mcpu=gfx900" "-o" [[OBJ_DEV_B_900:".*-gfx900-.*o"]]
// CHECK-SAME: "-o" [[OBJ_DEV_B_900:".*-gfx900-.*o"]]
// CHECK: [[LLD: ".*lld"]] "-flavor" "gnu" "-shared"
// CHECK-SAME: "-o" "[[IMG_DEV_B_900:.*out]]" [[OBJ_DEV_B_900]]

View File

@ -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"

View File

@ -40,13 +40,14 @@
// CHECK-SAME: "-o" [[OPT_BC_DEV1:".*-gfx803-optimized.*bc"]]
// CHECK: [[LLC: ".*llc"]] [[OPT_BC_DEV1]] "-mtriple=amdgcn-amd-amdhsa"
// CHECK-SAME: "-mcpu=gfx803"
// CHECK-SAME: "-filetype=obj"
// CHECK-SAME: "-mcpu=gfx803" "-o" [[OBJ_DEV1:".*-gfx803-.*o"]]
// CHECK-SAME: "-o" [[OBJ_DEV1:".*-gfx803-.*o"]]
// CHECK: [[LLD: ".*lld"]] "-flavor" "gnu" "-shared"
// CHECK-SAME: "-o" "[[IMG_DEV1:.*out]]" [[OBJ_DEV1]]
// CHECK: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa"
// CHECK: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa"
// CHECK-SAME: "-aux-triple" "x86_64-unknown-linux-gnu"
// CHECK-SAME: "-emit-llvm-bc"
// CHECK-SAME: {{.*}} "-main-file-name" "a.cu" {{.*}} "-target-cpu" "gfx900"
@ -72,8 +73,9 @@
// CHECK-SAME: "-o" [[OPT_BC_DEV2:".*-gfx900-optimized.*bc"]]
// CHECK: [[LLC]] [[OPT_BC_DEV2]] "-mtriple=amdgcn-amd-amdhsa"
// CHECk-SAME: "-mcpu=gfx900"
// CHECK-SAME: "-filetype=obj"
// CHECK-SAME: "-mcpu=gfx900" "-o" [[OBJ_DEV2:".*-gfx900-.*o"]]
// CHECK-SAME: "-o" [[OBJ_DEV2:".*-gfx900-.*o"]]
// CHECK: [[LLD]] "-flavor" "gnu" "-shared"
// CHECK-SAME: "-o" "[[IMG_DEV2:.*out]]" [[OBJ_DEV2]]