1berry/ruby
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Add VM_CALL_ARGS_SPLAT_MUT callinfo flag

Browse Source 
This flag is set when the caller has already created a new array to
handle a splat, such as for `f(*a, b)` and `f(*a, *b)`.  Previously,
if `f` was defined as `def f(*a)`, these calls would create an extra
array on the callee side, instead of using the new array created
by the caller.

This modifies `setup_args_core` to set the flag whenver it would add
a `splatarray true` instruction.  However, when `splatarray true` is
changed to `splatarray false` in the peephole optimizer, to avoid
unnecessary allocations on the caller side, the flag must be removed.
Add `optimize_args_splat_no_copy` and have the peephole optimizer call
that.  This significantly simplifies the related peephole optimizer
code.

On the callee side, in `setup_parameters_complex`, set
`args->rest_dupped` to true if the flag is set.

This takes a similar approach for optimizing regular splats that was
previiously used for keyword splats in
d2c41b1bff1f3102544bb0d03d4e82356d034d33 (via VM_CALL_KW_SPLAT_MUT).
This commit is contained in:
Jeremy Evans 2023-11-23 10:47:24 -08:00
parent ef276858d9
commit 22e488464a
5 changed files with 95 additions and 106 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

193
compile.c
View File

@ -3190,6 +3190,30 @@ ci_argc_set(const rb_iseq_t *iseq, const struct rb_callinfo *ci, int argc)
return nci; return nci;
} }
static bool
optimize_args_splat_no_copy(rb_iseq_t *iseq, INSN *insn, LINK_ELEMENT *niobj,
unsigned int set_flags, unsigned int unset_flags)
{
LINK_ELEMENT *iobj = (LINK_ELEMENT *)insn;
if (!IS_NEXT_INSN_ID(niobj, send)) {
return false;
}
niobj = niobj->next;
const struct rb_callinfo *ci = (const struct rb_callinfo *)OPERAND_AT(niobj, 0);
unsigned int flags = vm_ci_flag(ci);
if ((flags & set_flags) == set_flags && !(flags & unset_flags)) {
RUBY_ASSERT(flags & VM_CALL_ARGS_SPLAT_MUT);
OPERAND_AT(iobj, 0) = Qfalse;
const struct rb_callinfo *nci = vm_ci_new(vm_ci_mid(ci),
flags & ~VM_CALL_ARGS_SPLAT_MUT, vm_ci_argc(ci), vm_ci_kwarg(ci));
RB_OBJ_WRITTEN(iseq, ci, nci);
OPERAND_AT(niobj, 0) = (VALUE)nci;
return true;
}
return false;
}
static int static int
iseq_peephole_optimize(rb_iseq_t *iseq, LINK_ELEMENT *list, const int do_tailcallopt) iseq_peephole_optimize(rb_iseq_t *iseq, LINK_ELEMENT *list, const int do_tailcallopt)
{ {
@ -3879,58 +3903,46 @@ iseq_peephole_optimize(rb_iseq_t *iseq, LINK_ELEMENT *list, const int do_tailcal
* splatarray false * splatarray false
* send * send
*/ */
if (IS_NEXT_INSN_ID(niobj, send)) { if (optimize_args_splat_no_copy(iseq, iobj, niobj,
niobj = niobj->next; VM_CALL_ARGS_SPLAT, VM_CALL_KW_SPLAT|VM_CALL_ARGS_BLOCKARG)) goto optimized_splat;
unsigned int flag = vm_ci_flag((const struct rb_callinfo *)OPERAND_AT(niobj, 0));
if ((flag & VM_CALL_ARGS_SPLAT) && !(flag & (VM_CALL_KW_SPLAT|VM_CALL_ARGS_BLOCKARG))) { if (IS_NEXT_INSN_ID(niobj, getlocal) || IS_NEXT_INSN_ID(niobj, getinstancevariable)) {
OPERAND_AT(iobj, 0) = Qfalse;
}
}
else if (IS_NEXT_INSN_ID(niobj, getlocal) || IS_NEXT_INSN_ID(niobj, getinstancevariable)) {
niobj = niobj->next; niobj = niobj->next;
if (IS_NEXT_INSN_ID(niobj, send)) { /*
niobj = niobj->next; * Eliminate array allocation for f(1, *a, &lvar) and f(1, *a, &@iv)
unsigned int flag = vm_ci_flag((const struct rb_callinfo *)OPERAND_AT(niobj, 0)); *
* splatarray true
* getlocal / getinstancevariable
* send ARGS_SPLAT|ARGS_BLOCKARG and not KW_SPLAT
* =>
* splatarray false
* getlocal / getinstancevariable
* send
*/
if (optimize_args_splat_no_copy(iseq, iobj, niobj,
VM_CALL_ARGS_SPLAT|VM_CALL_ARGS_BLOCKARG, VM_CALL_KW_SPLAT)) goto optimized_splat;
if ((flag & VM_CALL_ARGS_SPLAT)) { /*
/* * Eliminate array allocation for f(*a, **lvar) and f(*a, **@iv)
* Eliminate array allocation for f(1, *a, &lvar) and f(1, *a, &@iv) *
* * splatarray true
* splatarray true * getlocal / getinstancevariable
* getlocal / getinstancevariable * send ARGS_SPLAT|KW_SPLAT and not ARGS_BLOCKARG
* send ARGS_SPLAT|ARGS_BLOCKARG and not KW_SPLAT * =>
* => * splatarray false
* splatarray false * getlocal / getinstancevariable
* getlocal / getinstancevariable * send
* send */
*/ if (optimize_args_splat_no_copy(iseq, iobj, niobj,
if ((flag & VM_CALL_ARGS_BLOCKARG) && !(flag & VM_CALL_KW_SPLAT)) { VM_CALL_ARGS_SPLAT|VM_CALL_KW_SPLAT, VM_CALL_ARGS_BLOCKARG)) goto optimized_splat;
OPERAND_AT(iobj, 0) = Qfalse;
}
/* if (IS_NEXT_INSN_ID(niobj, getlocal) || IS_NEXT_INSN_ID(niobj, getinstancevariable) ||
* Eliminate array allocation for f(*a, **lvar) and f(*a, **@iv) IS_NEXT_INSN_ID(niobj, getblockparamproxy)) {
*
* splatarray true
* getlocal / getinstancevariable
* send ARGS_SPLAT|KW_SPLAT and not ARGS_BLOCKARG
* =>
* splatarray false
* getlocal / getinstancevariable
* send
*/
else if (!(flag & VM_CALL_ARGS_BLOCKARG) && (flag & VM_CALL_KW_SPLAT)) {
OPERAND_AT(iobj, 0) = Qfalse;
}
}
}
else if (IS_NEXT_INSN_ID(niobj, getlocal) || IS_NEXT_INSN_ID(niobj, getinstancevariable) ||
IS_NEXT_INSN_ID(niobj, getblockparamproxy)) {
niobj = niobj->next; niobj = niobj->next;
/* /*
* Eliminate array allocation for f(*a, **lvar, &lvar) and f(*a, **@iv, &@iv) * Eliminate array allocation for f(*a, **lvar, &{arg,lvar,@iv})
* *
* splatarray true * splatarray true
* getlocal / getinstancevariable * getlocal / getinstancevariable
@ -3942,40 +3954,24 @@ iseq_peephole_optimize(rb_iseq_t *iseq, LINK_ELEMENT *list, const int do_tailcal
* getlocal / getinstancevariable / getblockparamproxy * getlocal / getinstancevariable / getblockparamproxy
* send * send
*/ */
if (IS_NEXT_INSN_ID(niobj, send)) { optimize_args_splat_no_copy(iseq, iobj, niobj,
niobj = niobj->next; VM_CALL_ARGS_SPLAT|VM_CALL_KW_SPLAT|VM_CALL_ARGS_BLOCKARG, 0);
unsigned int flag = vm_ci_flag((const struct rb_callinfo *)OPERAND_AT(niobj, 0));
if ((flag & VM_CALL_ARGS_SPLAT) && (flag & VM_CALL_KW_SPLAT) && (flag & VM_CALL_ARGS_BLOCKARG)) {
OPERAND_AT(iobj, 0) = Qfalse;
}
}
} }
} } else if (IS_NEXT_INSN_ID(niobj, getblockparamproxy)) {
else if (IS_NEXT_INSN_ID(niobj, getblockparamproxy)) { /*
niobj = niobj->next; * Eliminate array allocation for f(1, *a, &arg)
*
if (IS_NEXT_INSN_ID(niobj, send)) { * splatarray true
niobj = niobj->next; * getblockparamproxy
unsigned int flag = vm_ci_flag((const struct rb_callinfo *)OPERAND_AT(niobj, 0)); * send ARGS_SPLAT|ARGS_BLOCKARG and not KW_SPLAT
* =>
/* * splatarray false
* Eliminate array allocation for f(1, *a, &arg) * getblockparamproxy
* * send
* splatarray true */
* getblockparamproxy optimize_args_splat_no_copy(iseq, iobj, niobj,
* send ARGS_SPLAT|ARGS_BLOCKARG and not KW_SPLAT VM_CALL_ARGS_SPLAT|VM_CALL_ARGS_BLOCKARG, VM_CALL_KW_SPLAT);
* => } else if (IS_NEXT_INSN_ID(niobj, duphash)) {
* splatarray false
* getblockparamproxy
* send
*/
if ((flag & VM_CALL_ARGS_BLOCKARG) & (flag & VM_CALL_ARGS_SPLAT) && !(flag & VM_CALL_KW_SPLAT)) {
OPERAND_AT(iobj, 0) = Qfalse;
}
}
}
else if (IS_NEXT_INSN_ID(niobj, duphash)) {
niobj = niobj->next; niobj = niobj->next;
/* /*
@ -3989,21 +3985,13 @@ iseq_peephole_optimize(rb_iseq_t *iseq, LINK_ELEMENT *list, const int do_tailcal
* duphash * duphash
* send * send
*/ */
if (IS_NEXT_INSN_ID(niobj, send)) { if (optimize_args_splat_no_copy(iseq, iobj, niobj->next,
niobj = niobj->next; VM_CALL_ARGS_SPLAT|VM_CALL_KW_SPLAT|VM_CALL_KW_SPLAT_MUT, VM_CALL_ARGS_BLOCKARG)) goto optimized_splat;
unsigned int flag = vm_ci_flag((const struct rb_callinfo *)OPERAND_AT(niobj, 0));
if ((flag & VM_CALL_ARGS_SPLAT) && (flag & VM_CALL_KW_SPLAT) &&
(flag & VM_CALL_KW_SPLAT_MUT) && !(flag & VM_CALL_ARGS_BLOCKARG)) {
OPERAND_AT(iobj, 0) = Qfalse;
}
}
else if (IS_NEXT_INSN_ID(niobj, getlocal) || IS_NEXT_INSN_ID(niobj, getinstancevariable) ||
IS_NEXT_INSN_ID(niobj, getblockparamproxy)) {
niobj = niobj->next;
if (IS_NEXT_INSN_ID(niobj, getlocal) || IS_NEXT_INSN_ID(niobj, getinstancevariable) ||
IS_NEXT_INSN_ID(niobj, getblockparamproxy)) {
/* /*
* Eliminate array allocation for f(*a, kw: 1, &lvar) and f(*a, kw: 1, &@iv) * Eliminate array allocation for f(*a, kw: 1, &{arg,lvar,@iv})
* *
* splatarray true * splatarray true
* duphash * duphash
@ -4015,18 +4003,12 @@ iseq_peephole_optimize(rb_iseq_t *iseq, LINK_ELEMENT *list, const int do_tailcal
* getlocal / getinstancevariable / getblockparamproxy * getlocal / getinstancevariable / getblockparamproxy
* send * send
*/ */
if (IS_NEXT_INSN_ID(niobj, send)) { optimize_args_splat_no_copy(iseq, iobj, niobj->next,
niobj = niobj->next; VM_CALL_ARGS_SPLAT|VM_CALL_KW_SPLAT|VM_CALL_KW_SPLAT_MUT|VM_CALL_ARGS_BLOCKARG, 0);
unsigned int flag = vm_ci_flag((const struct rb_callinfo *)OPERAND_AT(niobj, 0));
if ((flag & VM_CALL_ARGS_SPLAT) && (flag & VM_CALL_KW_SPLAT) &&
(flag & VM_CALL_KW_SPLAT_MUT) && (flag & VM_CALL_ARGS_BLOCKARG)) {
OPERAND_AT(iobj, 0) = Qfalse;
}
}
} }
} }
} }
optimized_splat:
return COMPILE_OK; return COMPILE_OK;
} }
@ -6190,12 +6172,15 @@ setup_args_core(rb_iseq_t *iseq, LINK_ANCHOR *const args, const NODE *argn,
// f(*a) // f(*a)
NO_CHECK(COMPILE(args, "args (splat)", RNODE_SPLAT(argn)->nd_head)); NO_CHECK(COMPILE(args, "args (splat)", RNODE_SPLAT(argn)->nd_head));
ADD_INSN1(args, argn, splatarray, RBOOL(dup_rest)); ADD_INSN1(args, argn, splatarray, RBOOL(dup_rest));
if (flag_ptr) *flag_ptr |= VM_CALL_ARGS_SPLAT; if (flag_ptr) {
*flag_ptr |= VM_CALL_ARGS_SPLAT;
if (dup_rest) *flag_ptr |= VM_CALL_ARGS_SPLAT_MUT;
}
RUBY_ASSERT(flag_ptr == NULL || (*flag_ptr & VM_CALL_KW_SPLAT) == 0); RUBY_ASSERT(flag_ptr == NULL || (*flag_ptr & VM_CALL_KW_SPLAT) == 0);
return 1; return 1;
} }
case NODE_ARGSCAT: { case NODE_ARGSCAT: {
if (flag_ptr) *flag_ptr |= VM_CALL_ARGS_SPLAT; if (flag_ptr) *flag_ptr |= VM_CALL_ARGS_SPLAT | VM_CALL_ARGS_SPLAT_MUT;
int argc = setup_args_core(iseq, args, RNODE_ARGSCAT(argn)->nd_head, 1, NULL, NULL); int argc = setup_args_core(iseq, args, RNODE_ARGSCAT(argn)->nd_head, 1, NULL, NULL);
if (nd_type_p(RNODE_ARGSCAT(argn)->nd_body, NODE_LIST)) { if (nd_type_p(RNODE_ARGSCAT(argn)->nd_body, NODE_LIST)) {
@ -6229,7 +6214,7 @@ setup_args_core(rb_iseq_t *iseq, LINK_ANCHOR *const args, const NODE *argn,
return argc; return argc;
} }
case NODE_ARGSPUSH: { case NODE_ARGSPUSH: {
if (flag_ptr) *flag_ptr |= VM_CALL_ARGS_SPLAT; if (flag_ptr) *flag_ptr |= VM_CALL_ARGS_SPLAT | VM_CALL_ARGS_SPLAT_MUT;
int argc = setup_args_core(iseq, args, RNODE_ARGSPUSH(argn)->nd_head, 1, NULL, NULL); int argc = setup_args_core(iseq, args, RNODE_ARGSPUSH(argn)->nd_head, 1, NULL, NULL);
if (nd_type_p(RNODE_ARGSPUSH(argn)->nd_body, NODE_LIST)) { if (nd_type_p(RNODE_ARGSPUSH(argn)->nd_body, NODE_LIST)) {
@ -9080,7 +9065,7 @@ compile_op_asgn1(rb_iseq_t *iseq, LINK_ANCHOR *const ret, const NODE *const node
} }
ADD_INSN1(ret, node, dupn, INT2FIX(dup_argn)); ADD_INSN1(ret, node, dupn, INT2FIX(dup_argn));
flag |= asgnflag; flag |= asgnflag;
ADD_SEND_R(ret, node, idAREF, argc, NULL, INT2FIX(flag & ~VM_CALL_KW_SPLAT_MUT), keywords); ADD_SEND_R(ret, node, idAREF, argc, NULL, INT2FIX(flag & ~(VM_CALL_ARGS_SPLAT_MUT|VM_CALL_KW_SPLAT_MUT)), keywords);
if (id == idOROP || id == idANDOP) { if (id == idOROP || id == idANDOP) {
/* a[x] ||= y or a[x] &&= y /* a[x] ||= y or a[x] &&= y

1
iseq.c
View File

@ -2315,6 +2315,7 @@ rb_insn_operand_intern(const rb_iseq_t *iseq,
VALUE flags = rb_ary_new(); VALUE flags = rb_ary_new();
# define CALL_FLAG(n) if (vm_ci_flag(ci) & VM_CALL_##n) rb_ary_push(flags, rb_str_new2(#n)) # define CALL_FLAG(n) if (vm_ci_flag(ci) & VM_CALL_##n) rb_ary_push(flags, rb_str_new2(#n))
CALL_FLAG(ARGS_SPLAT); CALL_FLAG(ARGS_SPLAT);
CALL_FLAG(ARGS_SPLAT_MUT);
CALL_FLAG(ARGS_BLOCKARG); CALL_FLAG(ARGS_BLOCKARG);
CALL_FLAG(FCALL); CALL_FLAG(FCALL);
CALL_FLAG(VCALL); CALL_FLAG(VCALL);

2
vm_args.c
View File

@ -510,7 +510,7 @@ setup_parameters_complex(rb_execution_context_t * const ec, const rb_iseq_t * co
args = &args_body; args = &args_body;
given_argc = args->argc = calling->argc; given_argc = args->argc = calling->argc;
args->argv = locals; args->argv = locals;
args->rest_dupped = FALSE; args->rest_dupped = vm_ci_flag(ci) & VM_CALL_ARGS_SPLAT_MUT;
if (kw_flag & VM_CALL_KWARG) { if (kw_flag & VM_CALL_KWARG) {
args->kw_arg = vm_ci_kwarg(ci); args->kw_arg = vm_ci_kwarg(ci);

2
vm_callinfo.h
View File

@ -25,6 +25,7 @@ enum vm_call_flag_bits {
VM_CALL_ZSUPER_bit, // zsuper VM_CALL_ZSUPER_bit, // zsuper
VM_CALL_OPT_SEND_bit, // internal flag VM_CALL_OPT_SEND_bit, // internal flag
VM_CALL_KW_SPLAT_MUT_bit, // kw splat hash can be modified (to avoid allocating a new one) VM_CALL_KW_SPLAT_MUT_bit, // kw splat hash can be modified (to avoid allocating a new one)
VM_CALL_ARGS_SPLAT_MUT_bit, // args splat can be modified (to avoid allocating a new one)
VM_CALL__END VM_CALL__END
}; };
@ -40,6 +41,7 @@ enum vm_call_flag_bits {
#define VM_CALL_ZSUPER (0x01 << VM_CALL_ZSUPER_bit) #define VM_CALL_ZSUPER (0x01 << VM_CALL_ZSUPER_bit)
#define VM_CALL_OPT_SEND (0x01 << VM_CALL_OPT_SEND_bit) #define VM_CALL_OPT_SEND (0x01 << VM_CALL_OPT_SEND_bit)
#define VM_CALL_KW_SPLAT_MUT (0x01 << VM_CALL_KW_SPLAT_MUT_bit) #define VM_CALL_KW_SPLAT_MUT (0x01 << VM_CALL_KW_SPLAT_MUT_bit)
#define VM_CALL_ARGS_SPLAT_MUT (0x01 << VM_CALL_ARGS_SPLAT_MUT_bit)
struct rb_callinfo_kwarg { struct rb_callinfo_kwarg {
int keyword_len; int keyword_len;

3
yjit/src/cruby_bindings.inc.rs
View File

@ -632,7 +632,8 @@ pub const VM_CALL_SUPER_bit: vm_call_flag_bits = 8;
pub const VM_CALL_ZSUPER_bit: vm_call_flag_bits = 9; pub const VM_CALL_ZSUPER_bit: vm_call_flag_bits = 9;
pub const VM_CALL_OPT_SEND_bit: vm_call_flag_bits = 10; pub const VM_CALL_OPT_SEND_bit: vm_call_flag_bits = 10;
pub const VM_CALL_KW_SPLAT_MUT_bit: vm_call_flag_bits = 11; pub const VM_CALL_KW_SPLAT_MUT_bit: vm_call_flag_bits = 11;
pub const VM_CALL__END: vm_call_flag_bits = 12; pub const VM_CALL_ARGS_SPLAT_MUT_bit: vm_call_flag_bits = 12;
pub const VM_CALL__END: vm_call_flag_bits = 13;
pub type vm_call_flag_bits = u32; pub type vm_call_flag_bits = u32;
#[repr(C)] #[repr(C)]
pub struct rb_callinfo { pub struct rb_callinfo {