Latest Webrev: http://cr.openjdk.java.net/~iklam/jdk16/8243208-cleanup-jvmflag-impl.v00/
Goals
- Reduce the complexity of macros related to command-line flags (JDK-8243208)
- Avoid including large globals.hpp in every file. Just include the <mod>_globals.hpp file of the module you need (separate RFE: JDK-8243205)
- Improvement footprint/performance
Non Goals
- The flag declaration is done with large X-macros. Although some people find this objectionably, this proposal does not change that.
Requirements for Command-line Flags
The flags implementation in HotSpot has evolved over 20 years without much documentation.
By reading the JDK 15 code, I can gather the following requirements – which lead to the complex implementation.
(These requirements cannot be implemented elegantly/efficiently using older C++ compilers, leading to the current messy code)
- [REQ1] Flags have 3 types
- PRODUCT - readable/writable in all builds
- DEVELOP - readable in all builds, writable only in debug builds
- NOTPRODUCT - readable/writable only in debug builds
- Using a flag of this type in a product build will result in C compiler error.
- [REQ2] Flags can have optional attributes
- MANAGEABLE, DIAGNOSTIC, EXPERIMENTAL
- MANAGEABLE, DIAGNOSTIC, EXPERIMENTAL
[REQ3] Each flag can be in at most one of the following groups.
C1, C2, JVMCI, ARCH, LP64
(many flags aren't in any of these groups)
- [REQ4] Each flag must have a default value
- Platform-defined default values are specified by XXX_PD macros
- [REQ5] Flag declarations should be concise (default flag attributes shouldn't need to be specified)
- [REQ6] Metadata for flags must be compile-time generated
- Good: stored in .bss section
- Bad: initialized with global constructors
Problems with command-line flags in JDK 15
First, although types, attributes and groups are orthogonal, JDK 15 allows you to only specify a limited number of combinations. For example, all experimental flags must be of the PRODUCT type:
develop(bool, CleanChunkPoolAsync, true, \
"Clean the chunk pool asynchronously") \
experimental(bool, AlwaysSafeConstructors, false, \
"Force safe construction, as if all fields are final.") \
In a way, the [REQ5] conciseness requirement is the root cause of the messy implementation. If we add an extra "kind" parameter to the macros, we can convert the above to the following, which would allow the second flag to be both EXPERIMENTAL and DEVELOP
develop(bool, CleanChunkPoolAsync, true, /*kind=*/ 0\
"Clean the chunk pool asynchronously") \
develop(bool, AlwaysSafeConstructors, false, /*kind=*/ FLAG_KIND_EXPERIMENTAL \
"Force safe construction, as if all fields are final.") \
Similarly, groups could be implemented as something like the following (instead of the messy macros in here).
/*kind=*/ FLAG_KIND_EXPERIMENTAL|FLAG_GROUP_C1
Another way to keep the code concise is to use constructor overloading in the flags definition, something like (simplified)
#ifndef CONSTEXPR
#define CONSTEXPR
#endif
#define FLAG_KIND_EXPERIMENTAL 1
struct Flag {
const char* name; int kind;
CONSTEXPR Flag(const char* n): name(n), kind(0) {}
CONSTEXPR Flag(const char* n, int k): name(n), kind(k) {}
};
Flag flags[] = {
Flag("CleanChunkPoolAsync"),
Flag("AlwaysSafeConstructors", FLAG_KIND_EXPERIMENTAL),
};
However, this runs into problem with [REQ6]. Even the effect of the above code can be completely decided at compilation time, without constexpr, GCC stubbornly insists to initialize the array using global constructors
$ /home/iklam/devkit/latest/bin/gcc -O3 -o - -S test.cpp .... _GLOBAL__sub_I_flags: .LFB7: .cfi_startproc movq $.LC0, flags(%rip) movl $0, flags+8(%rip) movq $.LC1, flags+16(%rip) movl $1, flags+24(%rip) ret
Another option is to use clever macros. However, there's no vararg macro that I can think of that can satisfy all of the above macros
Proposal - Use constexpr!
C++ constexpr makes things much easier:
$ /home/iklam/devkit/latest/bin/gcc -O3 -o - -S -DCONSTEXPR=constexpr test.cpp .... flags: .quad .LC0 // name .long 0 // kind .quad .LC1 // name .long 1 // kind
New way of declaring flags - with optional argument for attributes. Also, the flag declaration macros now all take only 7 arguments: (compared to old version in globals.hpp)
#define RUNTIME_FLAGS(develop, \
develop_pd, \
product, \
product_pd, \
/* REMOVED diagnostic, */\
/* REMOVED diagnostic_pd, */ \
/* REMOVED experimental, */\
notproduct, \
/* REMOVED manageable, */ \
/* REMOVED product_rw, */\
/* REMOVED lp64_product, */\
range, \
constraint) \
.....
develop(bool, CleanChunkPoolAsync, true, /* no attr */ \
"Clean the chunk pool asynchronously") \
\
product(uint, HandshakeTimeout, 0, /* attr= */DIAGNOSTIC, \
"If nonzero set a timeout in milliseconds for handshakes") \
\
product(bool, AlwaysSafeConstructors, false, /* attr= */ EXPERIMENTAL, \
"Force safe construction, as if all fields are final.") ....
The flags metadata is defined using overloaded constructors (around lin 679 of jvmFlag.cpp)
constexpr JVMFlag::JVMFlag(int flag_enum, const char* type, const char* name,
void* addr, int attrs, const char* doc) :
_type(type), _name(name), _addr(addr) NOT_PRODUCT(COMMA _doc(doc)) {}
constexpr JVMFlag::JVMFlag(int flag_enum, const char* type, const char* name,
void* addr, const char* doc) :
JVMFlag(flag_enum, type, name, addr, /*attrs*/0, doc) {}
....
#define DEVELOP_FLAG_INIT( type, name, value, ...) JVMFlag(FLAG_MEMBER_ENUM(name), ...snip..., __VA_ARGS__),
#define DEVELOP_FLAG_INIT_PD(type, name, ...) JVMFlag(FLAG_MEMBER_ENUM(name), ...snip..., __VA_ARGS__),
...snip....
static JVMFlag flagTable[NUM_JVMFlagsEnum + 1] = {
ALL_FLAGS(DEVELOP_FLAG_INIT,
DEVELOP_FLAG_INIT_PD,
PRODUCT_FLAG_INIT,
PRODUCT_FLAG_INIT_PD,
NOTPROD_FLAG_INIT,
IGNORE_RANGE,
IGNORE_CONSTRAINT)
....
// NOTE: ALL_FLAGS() calls RUNTIME_FLAGS()
Handling of Groups:
There's a very small number of groups. They don't seem very useful so I don't know if we will add many new groups in the future. So for the time being, I implemented groups by ordering the flags and counting the size of each group:
#define ENUM_F(type, name, ...) enum_##name,
#define IGNORE_F(...)
// dev dev-pd pro pro-pd notpro range constraint
enum FlagCounter_LP64 { LP64_RUNTIME_FLAGS( ENUM_F, ENUM_F, ENUM_F, ENUM_F, ENUM_F, IGNORE_F, IGNORE_F) num_flags_LP64 };
enum FlagCounter_JVMCI { JVMCI_ONLY(JVMCI_FLAGS( ENUM_F, ENUM_F, ENUM_F, ENUM_F, ENUM_F, IGNORE_F, IGNORE_F)) num_flags_JVMCI };
enum FlagCounter_C1 { COMPILER1_PRESENT(C1_FLAGS(ENUM_F, ENUM_F, ENUM_F, ENUM_F, ENUM_F, IGNORE_F, IGNORE_F)) num_flags_C1 };
enum FlagCounter_C2 { COMPILER2_PRESENT(C2_FLAGS(ENUM_F, ENUM_F, ENUM_F, ENUM_F, ENUM_F, IGNORE_F, IGNORE_F)) num_flags_C2 };
enum FlagCounter_ARCH { ARCH_FLAGS( ENUM_F, ENUM_F, ENUM_F, IGNORE_F, IGNORE_F) num_flags_ARCH };
const int first_flag_enum_LP64 = 0;
const int first_flag_enum_JVMCI = first_flag_enum_LP64 + num_flags_LP64;
const int first_flag_enum_C1 = first_flag_enum_JVMCI + num_flags_JVMCI;
const int first_flag_enum_C2 = first_flag_enum_C1 + num_flags_C1;
const int first_flag_enum_ARCH = first_flag_enum_C2 + num_flags_C2;
const int first_flag_enum_other = first_flag_enum_ARCH + num_flags_ARCH;
static constexpr inline int flag_group(int flag_enum) {
if (flag_enum < first_flag_enum_JVMCI) return JVMFlag::KIND_LP64_PRODUCT;
if (flag_enum < first_flag_enum_C1) return JVMFlag::KIND_JVMCI;
if (flag_enum < first_flag_enum_C2) return JVMFlag::KIND_C1;
if (flag_enum < first_flag_enum_ARCH) return JVMFlag::KIND_C2;
if (flag_enum < first_flag_enum_other) return JVMFlag::KIND_ARCH;
return 0;
}
Handling of types
Same as before, just fewer cases (old version here)
// Interface macros
#define DECLARE_PRODUCT_FLAG(type, name, value, ...) extern "C" type name;
#define DECLARE_PD_PRODUCT_FLAG(type, name, ...) extern "C" type name;
#ifdef PRODUCT
#define DECLARE_DEVELOPER_FLAG(type, name, value, ...) const type name = value;
#define DECLARE_PD_DEVELOPER_FLAG(type, name, ...) const type name = pd_##name;
#define DECLARE_NOTPRODUCT_FLAG(type, name, value, ...) const type name = value;
#else
#define DECLARE_DEVELOPER_FLAG(type, name, value, ...) extern "C" type name;
#define DECLARE_PD_DEVELOPER_FLAG(type, name, ...) extern "C" type name;
#define DECLARE_NOTPRODUCT_FLAG(type, name, value, ...) extern "C" type name;
#endif // PRODUCT
ALL_FLAGS(DECLARE_DEVELOPER_FLAG,
DECLARE_PD_DEVELOPER_FLAG,
DECLARE_PRODUCT_FLAG,
DECLARE_PD_PRODUCT_FLAG,
DECLARE_NOTPRODUCT_FLAG,
IGNORE_RANGE,
IGNORE_CONSTRAINT)
Range and Constraint Checking
The old code has 2 problems
- Builds list of range/constraint checking objects at VM start-up (lots of code and slow start)
- Range/constraint checking needs linear search (further slows down start-up)
The new design (see here for webrev):
- Builds the checker objects (JVMFlagLimit) at build-time using constexpr.
- JVMFlagLimit objects are indexed by each flag's enum (or NULL if no limit exists), so it's O(1) time.
Implementation of JVMFlagLimit
The range/constraint information for a flag of type T is described by a JVMTypedFlagLimit<T>:
class JVMFlagLimit {
enum {
HAS_RANGE = 1,
HAS_CONSTRAINT = 2
};
short _constraint_func;
char _phase;
char _kind; ...};
template <typename T>
class JVMTypedFlagLimit : public JVMFlagLimit {
const T _min;
const T _max; ...};
Each flag is given a unique enum that starts from 0 to NUM_JVMFlagsEnum-1. We use this enum to find the JVMTypedFlagLimit<T> of this flag from an array:
static constexpr const JVMFlagLimit* const flagLimitTable[1 + NUM_JVMFlagsEnum] = { .... }
const JVMFlagLimit* const* JVMFlagLimit::flagLimits = &flagLimitTable[1]; // excludes dummy
/* E.g., to get the limit of this flag:
product(intx, ContendedPaddingWidth, 128, \
"How many bytes to pad the fields/classes marked @Contended with")\
range(0, 8192) \
constraint(ContendedPaddingWidthConstraintFunc,AfterErgo) \
*/
const JVMTypedFlagLimit<intx>* limit = JVMFlagLimit::flagLimits[Flag_ContendedPaddingWidth_Enum];
// We will see these fields:
// limit->_constraint_func ==> constraint_enum_ContendedPaddingWidthConstraintFunc (more on this below)
// limit->_phase ==> AfterErgo
// limit->_kind ==> HAS_RANGE | HAS_CONSTRAINT
// limit->_min ==> 0
// limit->_max ==> 8192
Most flags have neither range nor constraint. For those flags, we want its flagLimits[Flag_flagname_Enum] to be NULL.
To do this, we first define a JVMTypedFlagLimit<T> variable for each flag (including the ones that don't have range/constraint). It's done by this macro:
// macro body starts here -------------------+
// |
// v
#define FLAG_LIMIT_DEFINE( type, name, ...) ); constexpr JVMTypedFlagLimit<type> limit_##name(0
#define FLAG_LIMIT_DEFINE_DUMMY(type, name, ...) ); constexpr DummyLimit nolimit_##name(0
#define FLAG_LIMIT_PTR( type, name, ...) ), LimitGetter<type>::get_limit(&limit_##name, 0
#define FLAG_LIMIT_PTR_NONE( type, name, ...) ), LimitGetter<type>::no_limit(0
#define APPLY_FLAG_RANGE(...) , __VA_ARGS__
#define APPLY_FLAG_CONSTRAINT(func, phase) , next_two_args_are_constraint, (short)CONSTRAINT_ENUM(func), int(JVMFlagConstraint::phase)
constexpr JVMTypedFlagLimit<int> limit_dummy
(
#ifdef PRODUCT
ALL_FLAGS(FLAG_LIMIT_DEFINE_DUMMY,
FLAG_LIMIT_DEFINE_DUMMY,
FLAG_LIMIT_DEFINE,
FLAG_LIMIT_DEFINE,
FLAG_LIMIT_DEFINE_DUMMY,
APPLY_FLAG_RANGE,
APPLY_FLAG_CONSTRAINT)
#else
ALL_FLAGS(FLAG_LIMIT_DEFINE,
FLAG_LIMIT_DEFINE,
FLAG_LIMIT_DEFINE,
FLAG_LIMIT_DEFINE,
FLAG_LIMIT_DEFINE,
APPLY_FLAG_RANGE,
APPLY_FLAG_CONSTRAINT)
#endif
);
To understand how the macros work, it's best to compile jvmFlagLimit.c with gcc -save-temps. and look at the generated jvmFlagLimit.ii with the macros expanded. Here's an example (prettified manually):
constexpr JVMTypedFlagLimit<int> limit_dummy();
constexpr JVMTypedFlagLimit<bool> limit_UseCompressedOops(0);
....
constexpr JVMTypedFlagLimit<intx> limit_ObjectAlignmentInBytes(0, 8, 256,
next_two_args_are_constraint,
(short)constraint_enum_ObjectAlignmentInBytesConstraintFunc, int(JVMFlagConstraint::AtParse));
....
constexpr JVMTypedFlagLimit<intx> limit_JVMCIThreads(0, 1, max_jint);
We use overloaded constructors to fill out the necessarily fields of the JVMTypedFlagLimit<T> variables. Note that the min/max parameters, as well as the constraint_func/phase parameters, can both be integer values. For disambiguation, we pass in a dummy next_two_args_are_constraint for the constraint_func/phase.
We also need to always pass in an initial dummy 0 parameter so that the macros can safely add a comma before passing the min/max or constraint_func/phase.
These dummy parameters are evaluated at compile time so they can be safely optimized away.
The next step is to fill out the flagLimitTable[] array:
static constexpr const JVMFlagLimit* const flagLimitTable[1 + NUM_JVMFlagsEnum] = {
// Because FLAG_LIMIT_PTR must start with an "),", we have to place a dummy element here.
LimitGetter<int>::get_limit(NULL, 0
#ifdef PRODUCT
ALL_FLAGS(FLAG_LIMIT_PTR_NONE,
FLAG_LIMIT_PTR_NONE,
FLAG_LIMIT_PTR,
FLAG_LIMIT_PTR,
FLAG_LIMIT_PTR_NONE,
APPLY_FLAG_RANGE,
APPLY_FLAG_CONSTRAINT)
#else
ALL_FLAGS(FLAG_LIMIT_PTR,
FLAG_LIMIT_PTR,
FLAG_LIMIT_PTR,
FLAG_LIMIT_PTR,
FLAG_LIMIT_PTR,
APPLY_FLAG_RANGE,
APPLY_FLAG_CONSTRAINT)
#endif
)
};
For the flags shown in the example above, the following code is generated by the macros:
static constexpr const JVMFlagLimit* const flagLimitTable[1 + NUM_JVMFlagsEnum] = {
LimitGetter<int>::get_limit(NULL, 0),
LimitGetter<bool>::get_limit(&limit_UseCompressedOops, 0),
....
LimitGetter<intx>::get_limit(&limit_ObjectAlignmentInBytes, 0,
8, 256,
next_two_args_are_constraint,
(short)constraint_enum_ObjectAlignmentInBytesConstraintFunc, int(JVMFlagConstraint::AtParse)),
....
LimitGetter<intx>::get_limit(&limit_JVMCIThreads, 0, 1, max_jint),
- If a flag has neither range nor constraint, we will call the LimitGetter<T>::get_limit() function with two parameters, which returns NULL.
- If a flag has range and/or constraint, we will call a LimitGetter<T>::get_limit() function with more than two parameters. These functions would return the same JVMTypedFlagLimit<T> as passed in.
As a result, we will end up with this in the final output of the C++ compiler:
static constexpr const JVMFlagLimit* const flagLimitTable[1 + NUM_JVMFlagsEnum] = {
NULL, // dummy
NULL, // UseCompressedOops has no range/constraint
....
&limit_ObjectAlignmentInBytes
....
&limit_JVMCIThreads
What happens to unreferenced flag limits
All the flag limits are defined with the constexpr keyword, which has internal linkage by default. If a flag has no range/constraint, its flag limit (e.g., limit_UseCompressedOops in the example above) will be unused, and will be eliminated by the C++ compiler from the object file. So we don't waste any space.
Why use enums for constraint_func
This is a small optimization: There are 120 flags that use a constraint function, but there are only 65 total constraint functions. By using a short index, we can:
- Reduce the size of the JVMFlagLimit object (the 2 bytes fits in unused space)
- Reduce the number of pointers relocated when libjvm.so is dynamically loaded (from 120 to 65).
The savings are not a big deal, but since we can do it, why not?
Is constexpr really working?
A good way to check is to build the .o with something like "gcc -save-temps" and look at the .s file. Here's an example of jvmFlagLimit.s. You can see that the content of the flagLimitTable is also completely determined at build time (it's in the "ro" section):
.section .data.rel.ro.local,"aw"
.align 32
.type _ZL14flagLimitTable, @object
.size _ZL14flagLimitTable, 9728
_ZL14flagLimitTable:
.quad 0
.quad 0
.quad 0
.quad _ZL28limit_ObjectAlignmentInBytes
.quad 0
.quad 0
.quad 0
.quad 0
.quad 0
.quad 0
.quad 0
.quad _ZL18limit_JVMCIThreads
.quad _ZL22limit_JVMCIHostThreads
....
Overview
Content Tools
ThemeBuilder