aboutsummaryrefslogtreecommitdiff
path: root/clang/utils
diff options
context:
space:
mode:
Diffstat (limited to 'clang/utils')
-rw-r--r--clang/utils/TableGen/RISCVVEmitter.cpp8
-rw-r--r--clang/utils/perf-training/CMakeLists.txt36
-rw-r--r--clang/utils/perf-training/perf-helper.py65
3 files changed, 84 insertions, 25 deletions
diff --git a/clang/utils/TableGen/RISCVVEmitter.cpp b/clang/utils/TableGen/RISCVVEmitter.cpp
index f73b0aecc..74f29ac 100644
--- a/clang/utils/TableGen/RISCVVEmitter.cpp
+++ b/clang/utils/TableGen/RISCVVEmitter.cpp
@@ -133,28 +133,20 @@ static BasicType ParseBasicType(char c) {
switch (c) {
case 'c':
return BasicType::Int8;
- break;
case 's':
return BasicType::Int16;
- break;
case 'i':
return BasicType::Int32;
- break;
case 'l':
return BasicType::Int64;
- break;
case 'x':
return BasicType::Float16;
- break;
case 'f':
return BasicType::Float32;
- break;
case 'd':
return BasicType::Float64;
- break;
case 'y':
return BasicType::BFloat16;
- break;
default:
return BasicType::Unknown;
}
diff --git a/clang/utils/perf-training/CMakeLists.txt b/clang/utils/perf-training/CMakeLists.txt
index 1d7bb78..2cd4c4c 100644
--- a/clang/utils/perf-training/CMakeLists.txt
+++ b/clang/utils/perf-training/CMakeLists.txt
@@ -6,6 +6,10 @@ set(CLANG_PGO_TRAINING_DATA "${CMAKE_CURRENT_SOURCE_DIR}" CACHE PATH
set(CLANG_PGO_TRAINING_DATA_SOURCE_DIR OFF CACHE STRING "Path to source directory containing cmake project with source files to use for generating pgo data")
set(CLANG_PGO_TRAINING_DEPS "" CACHE STRING "Extra dependencies needed to build the PGO training data.")
+add_custom_target(clear-perf-data
+ COMMAND "${Python3_EXECUTABLE}" ${CMAKE_CURRENT_SOURCE_DIR}/perf-helper.py clean ${CMAKE_CURRENT_BINARY_DIR} perf.data
+ COMMENT "Clearing old perf data")
+
option(CLANG_PGO_TRAINING_USE_LLVM_BUILD "Use LLVM build for generating PGO data" ON)
llvm_canonicalize_cmake_booleans(
@@ -21,7 +25,7 @@ if(LLVM_BUILD_INSTRUMENTED)
add_lit_testsuite(generate-profraw "Generating clang PGO data"
${CMAKE_CURRENT_BINARY_DIR}/pgo-data/
EXCLUDE_FROM_CHECK_ALL
- DEPENDS clear-profraw
+ DEPENDS clear-profraw clang
)
add_custom_target(clear-profraw
@@ -55,6 +59,32 @@ if(LLVM_BUILD_INSTRUMENTED)
USE_TOOLCHAIN EXLUDE_FROM_ALL NO_INSTALL DEPENDS generate-profraw)
add_dependencies(generate-profdata generate-profraw-external)
endif()
+
+ if(NOT LLVM_PROFGEN)
+ find_program(LLVM_PROFGEN llvm-profgen)
+ endif()
+
+ if(NOT LLVM_PROFGEN)
+ message(STATUS "To enable converting CSSPGO samples LLVM_PROFGEN has to point to llvm-profgen")
+ elseif(NOT CLANG_PGO_TRAINING_DATA_SOURCE_DIR)
+ message(STATUS "CLANG_PGO_TRAINING_DATA_SOURCE_DIR must be set to collect CSSPGO samples")
+ else()
+ set(PERF_HELPER "${Python3_EXECUTABLE}" ${CMAKE_CURRENT_SOURCE_DIR}/perf-helper.py)
+ set(CLANG_SPROFDATA ${CMAKE_CURRENT_BINARY_DIR}/clang.sprofdata)
+ add_custom_command(
+ OUTPUT ${CLANG_SPROFDATA}
+ # Execute generate-profraw-external under perf
+ COMMAND ${PERF_HELPER} perf --csspgo -- ${CMAKE_COMMAND} --build ${CMAKE_BINARY_DIR} --target generate-profraw-external
+ # Convert perf profile into profraw
+ COMMAND ${PERF_HELPER} perf2prof ${LLVM_PROFGEN} $<TARGET_FILE:clang> ${CMAKE_CURRENT_BINARY_DIR}
+ # Merge profdata
+ COMMAND ${PERF_HELPER} merge --sample ${LLVM_PROFDATA} ${CLANG_SPROFDATA} ${CMAKE_CURRENT_BINARY_DIR}
+ DEPENDS clang ${CLANG_PGO_TRAINING_DEPS} clear-perf-data generate-profraw-external-clean
+ VERBATIM
+ USES_TERMINAL
+ )
+ add_custom_target(generate-sprofdata DEPENDS ${CLANG_SPROFDATA})
+ endif()
endif()
endif()
@@ -104,8 +134,4 @@ if(CLANG_BOLT AND NOT LLVM_BUILD_INSTRUMENTED)
COMMAND "${Python3_EXECUTABLE}" ${CMAKE_CURRENT_SOURCE_DIR}/perf-helper.py clean ${CMAKE_CURRENT_BINARY_DIR} fdata
COMMENT "Clearing old BOLT fdata")
- add_custom_target(clear-perf-data
- COMMAND "${Python3_EXECUTABLE}" ${CMAKE_CURRENT_SOURCE_DIR}/perf-helper.py clean ${CMAKE_CURRENT_BINARY_DIR} perf.data
- COMMENT "Clearing old perf data")
-
endif()
diff --git a/clang/utils/perf-training/perf-helper.py b/clang/utils/perf-training/perf-helper.py
index ab4491d..1c7904e 100644
--- a/clang/utils/perf-training/perf-helper.py
+++ b/clang/utils/perf-training/perf-helper.py
@@ -45,14 +45,22 @@ def clean(args):
def merge(args):
- if len(args) < 3:
- print(
- "Usage: %s merge <llvm-profdata> <output> <paths>\n" % __file__
- + "\tMerges all profraw files from path into output."
- )
- return 1
- cmd = [args[0], "merge", "-o", args[1]]
- for path in args[2:]:
+ parser = argparse.ArgumentParser(
+ prog="perf-helper merge",
+ description="Merges all profraw files from path(s) into output",
+ )
+ parser.add_argument("profdata", help="Path to llvm-profdata tool")
+ parser.add_argument("output", help="Output filename")
+ parser.add_argument(
+ "paths", nargs="+", help="Folder(s) containing input profraw files"
+ )
+ parser.add_argument("--sample", action="store_true", help="Sample profile")
+ opts = parser.parse_args(args)
+
+ cmd = [opts.profdata, "merge", "-o", opts.output]
+ if opts.sample:
+ cmd += ["--sample"]
+ for path in opts.paths:
cmd.extend(findFilesWithExtension(path, "profraw"))
subprocess.check_call(cmd)
return 0
@@ -73,25 +81,30 @@ def merge_fdata(args):
def perf(args):
parser = argparse.ArgumentParser(
- prog="perf-helper perf", description="perf wrapper for BOLT profile collection"
+ prog="perf-helper perf",
+ description="perf wrapper for BOLT/CSSPGO profile collection",
)
parser.add_argument(
"--lbr", action="store_true", help="Use perf with branch stacks"
)
+ parser.add_argument("--csspgo", action="store_true", help="Enable CSSPGO flags")
parser.add_argument("cmd", nargs=argparse.REMAINDER, help="")
opts = parser.parse_args(args)
cmd = opts.cmd[1:]
+ event = "br_inst_retired.near_taken:uppp" if opts.csspgo else "cycles:u"
perf_args = [
"perf",
"record",
- "--event=cycles:u",
+ f"--event={event}",
"--freq=max",
"--output=%d.perf.data" % os.getpid(),
]
- if opts.lbr:
+ if opts.lbr or opts.csspgo:
perf_args += ["--branch-filter=any,u"]
+ if opts.csspgo:
+ perf_args += ["-g", "--call-graph=fp"]
perf_args.extend(cmd)
start_time = time.time()
@@ -127,6 +140,30 @@ def perf2bolt(args):
return 0
+def perf2prof(args):
+ parser = argparse.ArgumentParser(
+ prog="perf-helper perf2prof",
+ description="perf to CSSPGO prof conversion wrapper",
+ )
+ parser.add_argument("profgen", help="Path to llvm-profgen binary")
+ parser.add_argument("binary", help="Input binary")
+ parser.add_argument("paths", nargs="+", help="Path containing perf.data files")
+ opts = parser.parse_args(args)
+
+ profgen_args = [opts.profgen, f"--binary={opts.binary}"]
+ for path in opts.paths:
+ for filename in findFilesWithExtension(path, "perf.data"):
+ subprocess.run(
+ [
+ *profgen_args,
+ f"--perfdata={filename}",
+ f"--output={filename}.profraw",
+ ],
+ check=True,
+ )
+ return 0
+
+
def dtrace(args):
parser = argparse.ArgumentParser(
prog="perf-helper dtrace",
@@ -660,7 +697,10 @@ def bolt_optimize(args):
process.check_returncode()
if opts.method in ["PERF", "LBR"]:
- perf2bolt([opts.bolt, opts.perf_training_binary_dir, opts.input])
+ args = [opts.bolt, opts.perf_training_binary_dir, opts.input]
+ if opts.method == "LBR":
+ args.extend("--lbr")
+ perf2bolt(args)
merge_fdata([opts.merge_fdata, opts.fdata, opts.perf_training_binary_dir])
@@ -707,6 +747,7 @@ commands = {
"merge-fdata": merge_fdata,
"perf": perf,
"perf2bolt": perf2bolt,
+ "perf2prof": perf2prof,
}