Skip to content

D
dragonwell11

openanolis / dragonwell11

通知 7
Star 2
Fork 0
D
dragonwell11
提交 6d9460f7 编写于 作者: N never
浏览文件
操作

4809552: Optimize Arrays.fill(...) 

Reviewed-by: kvn
上级 babc6e40
隐藏空白更改
内联 并排
Showing with 940 addition and 11 deletion
hotspot/src/cpu/sparc/vm/stubGenerator_sparc.cpp
浏览文件 @ 6d9460f7
......@@ -1587,6 +1587,185 @@ class StubGenerator: public StubCodeGenerator {
return start;
}
//
// Generate stub for disjoint short fill. If "aligned" is true, the
// "to" address is assumed to be heapword aligned.
//
// Arguments for generated stub:
// to: O0
// value: O1
// count: O2 treated as signed
//
address generate_fill(BasicType t, bool aligned, const char* name) {
__ align(CodeEntryAlignment);
StubCodeMark mark(this, "StubRoutines", name);
address start = __ pc();
const Register to = O0; // source array address
const Register value = O1; // fill value
const Register count = O2; // elements count
// O3 is used as a temp register
assert_clean_int(count, O3); // Make sure 'count' is clean int.
Label L_exit, L_skip_align1, L_skip_align2, L_fill_byte;
Label L_fill_2_bytes, L_fill_4_bytes, L_fill_32_bytes;
int shift = -1;
switch (t) {
case T_BYTE:
shift = 2;
break;
case T_SHORT:
shift = 1;
break;
case T_INT:
shift = 0;
break;
default: ShouldNotReachHere();
}
BLOCK_COMMENT("Entry:");
if (t == T_BYTE) {
// Zero extend value
__ and3(value, 0xff, value);
__ sllx(value, 8, O3);
__ or3(value, O3, value);
}
if (t == T_SHORT) {
// Zero extend value
__ sethi(0xffff0000, O3);
__ andn(value, O3, value);
}
if (t == T_BYTE || t == T_SHORT) {
__ sllx(value, 16, O3);
__ or3(value, O3, value);
}
__ cmp(count, 2<<shift); // Short arrays (< 8 bytes) fill by element
__ brx(Assembler::lessUnsigned, false, Assembler::pn, L_fill_4_bytes); // use unsigned cmp
__ delayed()->andcc(count, 1<<shift, G0);
if (!aligned && (t == T_BYTE || t == T_SHORT)) {
// align source address at 4 bytes address boundary
if (t == T_BYTE) {
// One byte misalignment happens only for byte arrays
__ andcc(to, 1, G0);
__ br(Assembler::zero, false, Assembler::pt, L_skip_align1);
__ delayed()->nop();
__ stb(value, to, 0);
__ inc(to, 1);
__ dec(count, 1);
__ BIND(L_skip_align1);
}
// Two bytes misalignment happens only for byte and short (char) arrays
__ andcc(to, 2, G0);
__ br(Assembler::zero, false, Assembler::pt, L_skip_align2);
__ delayed()->nop();
__ sth(value, to, 0);
__ inc(to, 2);
__ dec(count, 1 << (shift - 1));
__ BIND(L_skip_align2);
}
#ifdef _LP64
if (!aligned) {
#endif
// align to 8 bytes, we know we are 4 byte aligned to start
__ andcc(to, 7, G0);
__ br(Assembler::zero, false, Assembler::pt, L_fill_32_bytes);
__ delayed()->nop();
__ stw(value, to, 0);
__ inc(to, 4);
__ dec(count, 1 << shift);
__ BIND(L_fill_32_bytes);
#ifdef _LP64
}
#endif
Label L_check_fill_8_bytes;
// Fill 32-byte chunks
__ subcc(count, 8 << shift, count);
__ brx(Assembler::less, false, Assembler::pt, L_check_fill_8_bytes);
__ delayed()->nop();
if (t == T_INT) {
// Zero extend value
__ srl(value, 0, value);
}
if (t == T_BYTE || t == T_SHORT || t == T_INT) {
__ sllx(value, 32, O3);
__ or3(value, O3, value);
}
Label L_fill_32_bytes_loop;
__ align(16);
__ BIND(L_fill_32_bytes_loop);
__ stx(value, to, 0);
__ stx(value, to, 8);
__ stx(value, to, 16);
__ stx(value, to, 24);
__ subcc(count, 8 << shift, count);
__ brx(Assembler::greaterEqual, false, Assembler::pt, L_fill_32_bytes_loop);
__ delayed()->add(to, 32, to);
__ BIND(L_check_fill_8_bytes);
__ addcc(count, 8 << shift, count);
__ brx(Assembler::zero, false, Assembler::pn, L_exit);
__ delayed()->subcc(count, 1 << (shift + 1), count);
__ brx(Assembler::less, false, Assembler::pn, L_fill_4_bytes);
__ delayed()->andcc(count, 1<<shift, G0);
//
// length is too short, just fill 8 bytes at a time
//
Label L_fill_8_bytes_loop;
__ BIND(L_fill_8_bytes_loop);
__ stx(value, to, 0);
__ subcc(count, 1 << (shift + 1), count);
__ brx(Assembler::greaterEqual, false, Assembler::pn, L_fill_8_bytes_loop);
__ delayed()->add(to, 8, to);
// fill trailing 4 bytes
__ andcc(count, 1<<shift, G0); // in delay slot of branches
__ BIND(L_fill_4_bytes);
__ brx(Assembler::zero, false, Assembler::pt, L_fill_2_bytes);
if (t == T_BYTE || t == T_SHORT) {
__ delayed()->andcc(count, 1<<(shift-1), G0);
} else {
__ delayed()->nop();
}
__ stw(value, to, 0);
if (t == T_BYTE || t == T_SHORT) {
__ inc(to, 4);
// fill trailing 2 bytes
__ andcc(count, 1<<(shift-1), G0); // in delay slot of branches
__ BIND(L_fill_2_bytes);
__ brx(Assembler::zero, false, Assembler::pt, L_fill_byte);
__ delayed()->andcc(count, 1, count);
__ sth(value, to, 0);
if (t == T_BYTE) {
__ inc(to, 2);
// fill trailing byte
__ andcc(count, 1, count); // in delay slot of branches
__ BIND(L_fill_byte);
__ brx(Assembler::zero, false, Assembler::pt, L_exit);
__ delayed()->nop();
__ stb(value, to, 0);
} else {
__ BIND(L_fill_byte);
}
} else {
__ BIND(L_fill_2_bytes);
}
__ BIND(L_exit);
__ retl();
__ delayed()->mov(G0, O0); // return 0
return start;
}
//
// Generate stub for conjoint short copy. If "aligned" is true, the
// "from" and "to" addresses are assumed to be heapword aligned.
......@@ -2855,6 +3034,13 @@ class StubGenerator: public StubCodeGenerator {
StubRoutines::_checkcast_arraycopy = generate_checkcast_copy("checkcast_arraycopy");
StubRoutines::_unsafe_arraycopy = generate_unsafe_copy("unsafe_arraycopy");
StubRoutines::_generic_arraycopy = generate_generic_copy("generic_arraycopy");
StubRoutines::_jbyte_fill = generate_fill(T_BYTE, false, "jbyte_fill");
StubRoutines::_jshort_fill = generate_fill(T_SHORT, false, "jshort_fill");
StubRoutines::_jint_fill = generate_fill(T_INT, false, "jint_fill");
StubRoutines::_arrayof_jbyte_fill = generate_fill(T_BYTE, true, "arrayof_jbyte_fill");
StubRoutines::_arrayof_jshort_fill = generate_fill(T_SHORT, true, "arrayof_jshort_fill");
StubRoutines::_arrayof_jint_fill = generate_fill(T_INT, true, "arrayof_jint_fill");
}
void generate_initial() {
......
hotspot/src/cpu/x86/vm/assembler_x86.cpp
浏览文件 @ 6d9460f7
......@@ -8767,6 +8767,186 @@ void MacroAssembler::char_arrays_equals(bool is_array_equ, Register ary1, Regist
bind(DONE);
}
#ifdef PRODUCT
#define BLOCK_COMMENT(str) /* nothing */
#else
#define BLOCK_COMMENT(str) block_comment(str)
#endif
#define BIND(label) bind(label); BLOCK_COMMENT(#label ":")
void MacroAssembler::generate_fill(BasicType t, bool aligned,
Register to, Register value, Register count,
Register rtmp, XMMRegister xtmp) {
assert_different_registers(to, value, count, rtmp);
Label L_exit, L_skip_align1, L_skip_align2, L_fill_byte;
Label L_fill_2_bytes, L_fill_4_bytes;
int shift = -1;
switch (t) {
case T_BYTE:
shift = 2;
break;
case T_SHORT:
shift = 1;
break;
case T_INT:
shift = 0;
break;
default: ShouldNotReachHere();
}
if (t == T_BYTE) {
andl(value, 0xff);
movl(rtmp, value);
shll(rtmp, 8);
orl(value, rtmp);
}
if (t == T_SHORT) {
andl(value, 0xffff);
}
if (t == T_BYTE || t == T_SHORT) {
movl(rtmp, value);
shll(rtmp, 16);
orl(value, rtmp);
}
cmpl(count, 2<<shift); // Short arrays (< 8 bytes) fill by element
jcc(Assembler::below, L_fill_4_bytes); // use unsigned cmp
if (!UseUnalignedLoadStores && !aligned && (t == T_BYTE || t == T_SHORT)) {
// align source address at 4 bytes address boundary
if (t == T_BYTE) {
// One byte misalignment happens only for byte arrays
testptr(to, 1);
jccb(Assembler::zero, L_skip_align1);
movb(Address(to, 0), value);
increment(to);
decrement(count);
BIND(L_skip_align1);
}
// Two bytes misalignment happens only for byte and short (char) arrays
testptr(to, 2);
jccb(Assembler::zero, L_skip_align2);
movw(Address(to, 0), value);
addptr(to, 2);
subl(count, 1<<(shift-1));
BIND(L_skip_align2);
}
if (UseSSE < 2) {
Label L_fill_32_bytes_loop, L_check_fill_8_bytes, L_fill_8_bytes_loop, L_fill_8_bytes;
// Fill 32-byte chunks
subl(count, 8 << shift);
jcc(Assembler::less, L_check_fill_8_bytes);
align(16);
BIND(L_fill_32_bytes_loop);
for (int i = 0; i < 32; i += 4) {
movl(Address(to, i), value);
}
addptr(to, 32);
subl(count, 8 << shift);
jcc(Assembler::greaterEqual, L_fill_32_bytes_loop);
BIND(L_check_fill_8_bytes);
addl(count, 8 << shift);
jccb(Assembler::zero, L_exit);
jmpb(L_fill_8_bytes);
//
// length is too short, just fill qwords
//
BIND(L_fill_8_bytes_loop);
movl(Address(to, 0), value);
movl(Address(to, 4), value);
addptr(to, 8);
BIND(L_fill_8_bytes);
subl(count, 1 << (shift + 1));
jcc(Assembler::greaterEqual, L_fill_8_bytes_loop);
// fall through to fill 4 bytes
} else {
Label L_fill_32_bytes;
if (!UseUnalignedLoadStores) {
// align to 8 bytes, we know we are 4 byte aligned to start
testptr(to, 4);
jccb(Assembler::zero, L_fill_32_bytes);
movl(Address(to, 0), value);
addptr(to, 4);
subl(count, 1<<shift);
}
BIND(L_fill_32_bytes);
{
assert( UseSSE >= 2, "supported cpu only" );
Label L_fill_32_bytes_loop, L_check_fill_8_bytes, L_fill_8_bytes_loop, L_fill_8_bytes;
// Fill 32-byte chunks
movdl(xtmp, value);
pshufd(xtmp, xtmp, 0);
subl(count, 8 << shift);
jcc(Assembler::less, L_check_fill_8_bytes);
align(16);
BIND(L_fill_32_bytes_loop);
if (UseUnalignedLoadStores) {
movdqu(Address(to, 0), xtmp);
movdqu(Address(to, 16), xtmp);
} else {
movq(Address(to, 0), xtmp);
movq(Address(to, 8), xtmp);
movq(Address(to, 16), xtmp);
movq(Address(to, 24), xtmp);
}
addptr(to, 32);
subl(count, 8 << shift);
jcc(Assembler::greaterEqual, L_fill_32_bytes_loop);
BIND(L_check_fill_8_bytes);
addl(count, 8 << shift);
jccb(Assembler::zero, L_exit);
jmpb(L_fill_8_bytes);
//
// length is too short, just fill qwords
//
BIND(L_fill_8_bytes_loop);
movq(Address(to, 0), xtmp);
addptr(to, 8);
BIND(L_fill_8_bytes);
subl(count, 1 << (shift + 1));
jcc(Assembler::greaterEqual, L_fill_8_bytes_loop);
}
}
// fill trailing 4 bytes
BIND(L_fill_4_bytes);
testl(count, 1<<shift);
jccb(Assembler::zero, L_fill_2_bytes);
movl(Address(to, 0), value);
if (t == T_BYTE || t == T_SHORT) {
addptr(to, 4);
BIND(L_fill_2_bytes);
// fill trailing 2 bytes
testl(count, 1<<(shift-1));
jccb(Assembler::zero, L_fill_byte);
movw(Address(to, 0), value);
if (t == T_BYTE) {
addptr(to, 2);
BIND(L_fill_byte);
// fill trailing byte
testl(count, 1);
jccb(Assembler::zero, L_exit);
movb(Address(to, 0), value);
} else {
BIND(L_fill_byte);
}
} else {
BIND(L_fill_2_bytes);
}
BIND(L_exit);
}
#undef BIND
#undef BLOCK_COMMENT
Assembler::Condition MacroAssembler::negate_condition(Assembler::Condition cond) {
switch (cond) {
// Note some conditions are synonyms for others
......
hotspot/src/cpu/x86/vm/assembler_x86.hpp
浏览文件 @ 6d9460f7
......@@ -2242,6 +2242,11 @@ public:
Register limit, Register result, Register chr,
XMMRegister vec1, XMMRegister vec2);
// Fill primitive arrays
void generate_fill(BasicType t, bool aligned,
Register to, Register value, Register count,
Register rtmp, XMMRegister xtmp);
#undef VIRTUAL
};
......
hotspot/src/cpu/x86/vm/stubGenerator_x86_32.cpp
浏览文件 @ 6d9460f7
......@@ -1039,6 +1039,33 @@ class StubGenerator: public StubCodeGenerator {
}
address generate_fill(BasicType t, bool aligned, const char *name) {
__ align(CodeEntryAlignment);
StubCodeMark mark(this, "StubRoutines", name);
address start = __ pc();
BLOCK_COMMENT("Entry:");
const Register to = rdi; // source array address
const Register value = rdx; // value
const Register count = rsi; // elements count
__ enter(); // required for proper stackwalking of RuntimeStub frame
__ push(rsi);
__ push(rdi);
__ movptr(to , Address(rsp, 12+ 4));
__ movl(value, Address(rsp, 12+ 8));
__ movl(count, Address(rsp, 12+ 12));
__ generate_fill(t, aligned, to, value, count, rax, xmm0);
__ pop(rdi);
__ pop(rsi);
__ leave(); // required for proper stackwalking of RuntimeStub frame
__ ret(0);
return start;
}
address generate_conjoint_copy(BasicType t, bool aligned,
Address::ScaleFactor sf,
address nooverlap_target,
......@@ -2001,6 +2028,13 @@ class StubGenerator: public StubCodeGenerator {
generate_conjoint_long_copy(entry, &entry_jlong_arraycopy,
"jlong_arraycopy");
StubRoutines::_jbyte_fill = generate_fill(T_BYTE, false, "jbyte_fill");
StubRoutines::_jshort_fill = generate_fill(T_SHORT, false, "jshort_fill");
StubRoutines::_jint_fill = generate_fill(T_INT, false, "jint_fill");
StubRoutines::_arrayof_jbyte_fill = generate_fill(T_BYTE, true, "arrayof_jbyte_fill");
StubRoutines::_arrayof_jshort_fill = generate_fill(T_SHORT, true, "arrayof_jshort_fill");
StubRoutines::_arrayof_jint_fill = generate_fill(T_INT, true, "arrayof_jint_fill");
StubRoutines::_arrayof_jint_disjoint_arraycopy =
StubRoutines::_jint_disjoint_arraycopy;
StubRoutines::_arrayof_oop_disjoint_arraycopy =
......
hotspot/src/cpu/x86/vm/stubGenerator_x86_64.cpp
浏览文件 @ 6d9460f7
......@@ -1625,6 +1625,26 @@ class StubGenerator: public StubCodeGenerator {
return start;
}
address generate_fill(BasicType t, bool aligned, const char *name) {
__ align(CodeEntryAlignment);
StubCodeMark mark(this, "StubRoutines", name);
address start = __ pc();
BLOCK_COMMENT("Entry:");
const Register to = c_rarg0; // source array address
const Register value = c_rarg1; // value
const Register count = c_rarg2; // elements count
__ enter(); // required for proper stackwalking of RuntimeStub frame
__ generate_fill(t, aligned, to, value, count, rax, xmm0);
__ leave(); // required for proper stackwalking of RuntimeStub frame
__ ret(0);
return start;
}
// Arguments:
// aligned - true => Input and output aligned on a HeapWord == 8-byte boundary
// ignored
......@@ -2712,6 +2732,13 @@ class StubGenerator: public StubCodeGenerator {
StubRoutines::_unsafe_arraycopy = generate_unsafe_copy("unsafe_arraycopy");
StubRoutines::_generic_arraycopy = generate_generic_copy("generic_arraycopy");
StubRoutines::_jbyte_fill = generate_fill(T_BYTE, false, "jbyte_fill");
StubRoutines::_jshort_fill = generate_fill(T_SHORT, false, "jshort_fill");
StubRoutines::_jint_fill = generate_fill(T_INT, false, "jint_fill");
StubRoutines::_arrayof_jbyte_fill = generate_fill(T_BYTE, true, "arrayof_jbyte_fill");
StubRoutines::_arrayof_jshort_fill = generate_fill(T_SHORT, true, "arrayof_jshort_fill");
StubRoutines::_arrayof_jint_fill = generate_fill(T_INT, true, "arrayof_jint_fill");
// We don't generate specialized code for HeapWord-aligned source
// arrays, so just use the code we've already generated
StubRoutines::_arrayof_jbyte_disjoint_arraycopy = StubRoutines::_jbyte_disjoint_arraycopy;
......
hotspot/src/share/vm/includeDB_compiler2
浏览文件 @ 6d9460f7
......@@ -624,6 +624,7 @@ loopTransform.cpp divnode.hpp
loopTransform.cpp loopnode.hpp
loopTransform.cpp mulnode.hpp
loopTransform.cpp rootnode.hpp
loopTransform.cpp runtime.hpp
loopTransform.cpp subnode.hpp
loopUnswitch.cpp allocation.inline.hpp
......
hotspot/src/share/vm/opto/addnode.cpp
浏览文件 @ 6d9460f7
......@@ -705,6 +705,9 @@ int AddPNode::unpack_offsets(Node* elements[], int length) {
}
addr = addr->in(AddPNode::Address);
}
if (addr != base) {
return -1;
}
return count;
}
......
hotspot/src/share/vm/opto/c2_globals.hpp
浏览文件 @ 6d9460f7
......@@ -157,6 +157,12 @@
develop(bool, TraceLoopPredicate, false, \
"Trace generation of loop predicates") \
\
product(bool, OptimizeFill, false, \
"convert fill/copy loops into intrinsic") \
\
develop(bool, TraceOptimizeFill, false, \
"print detailed information about fill conversion") \
\
develop(bool, OptoCoalesce, true, \
"Use Conservative Copy Coalescing in the Register Allocator") \
\
......
hotspot/src/share/vm/opto/loopTransform.cpp
浏览文件 @ 6d9460f7
......@@ -2049,11 +2049,18 @@ bool IdealLoopTree::is_range_check_if(IfNode *iff, PhaseIdealLoop *phase, Invari
if (cmp->Opcode() != Op_CmpU ) {
return false;
}
if (cmp->in(2)->Opcode() != Op_LoadRange) {
return false;
Node* range = cmp->in(2);
if (range->Opcode() != Op_LoadRange) {
const TypeInt* tint = phase->_igvn.type(range)->isa_int();
if (!OptimizeFill || tint == NULL || tint->empty() || tint->_lo < 0) {
// Allow predication on positive values that aren't LoadRanges.
// This allows optimization of loops where the length of the
// array is a known value and doesn't need to be loaded back
// from the array.
return false;
}
}
LoadRangeNode* lr = (LoadRangeNode*)cmp->in(2);
if (!invar.is_invariant(lr)) { // loadRange must be invariant
if (!invar.is_invariant(range)) {
return false;
}
Node *iv = _head->as_CountedLoop()->phi();
......@@ -2248,9 +2255,9 @@ bool PhaseIdealLoop::loop_predication_impl(IdealLoopTree *loop) {
const Node* cmp = bol->in(1)->as_Cmp();
Node* idx = cmp->in(1);
assert(!invar.is_invariant(idx), "index is variant");
assert(cmp->in(2)->Opcode() == Op_LoadRange, "must be");
Node* ld_rng = cmp->in(2); // LoadRangeNode
assert(invar.is_invariant(ld_rng), "load range must be invariant");
assert(cmp->in(2)->Opcode() == Op_LoadRange || OptimizeFill, "must be");
Node* rng = cmp->in(2);
assert(invar.is_invariant(rng), "range must be invariant");
int scale = 1;
Node* offset = zero;
bool ok = is_scaled_iv_plus_offset(idx, cl->phi(), &scale, &offset);
......@@ -2271,21 +2278,21 @@ bool PhaseIdealLoop::loop_predication_impl(IdealLoopTree *loop) {
// Perform cloning to keep Invariance state correct since the
// late schedule will place invariant things in the loop.
ld_rng = invar.clone(ld_rng, ctrl);
rng = invar.clone(rng, ctrl);
if (offset && offset != zero) {
assert(invar.is_invariant(offset), "offset must be loop invariant");
offset = invar.clone(offset, ctrl);
}
// Test the lower bound
Node* lower_bound_bol = rc_predicate(ctrl, scale, offset, init, limit, stride, ld_rng, false);
Node* lower_bound_bol = rc_predicate(ctrl, scale, offset, init, limit, stride, rng, false);
IfNode* lower_bound_iff = lower_bound_proj->in(0)->as_If();
_igvn.hash_delete(lower_bound_iff);
lower_bound_iff->set_req(1, lower_bound_bol);
if (TraceLoopPredicate) tty->print_cr("lower bound check if: %d", lower_bound_iff->_idx);
// Test the upper bound
Node* upper_bound_bol = rc_predicate(ctrl, scale, offset, init, limit, stride, ld_rng, true);
Node* upper_bound_bol = rc_predicate(ctrl, scale, offset, init, limit, stride, rng, true);
IfNode* upper_bound_iff = upper_bound_proj->in(0)->as_If();
_igvn.hash_delete(upper_bound_iff);
upper_bound_iff->set_req(1, upper_bound_bol);
......@@ -2366,3 +2373,348 @@ bool IdealLoopTree::loop_predication( PhaseIdealLoop *phase) {
return hoisted;
}
// Process all the loops in the loop tree and replace any fill
// patterns with an intrisc version.
bool PhaseIdealLoop::do_intrinsify_fill() {
bool changed = false;
for (LoopTreeIterator iter(_ltree_root); !iter.done(); iter.next()) {
IdealLoopTree* lpt = iter.current();
changed |= intrinsify_fill(lpt);
}
return changed;
}
// Examine an inner loop looking for a a single store of an invariant
// value in a unit stride loop,
bool PhaseIdealLoop::match_fill_loop(IdealLoopTree* lpt, Node*& store, Node*& store_value,
Node*& shift, Node*& con) {
const char* msg = NULL;
Node* msg_node = NULL;
store_value = NULL;
con = NULL;
shift = NULL;
// Process the loop looking for stores. If there are multiple
// stores or extra control flow give at this point.
CountedLoopNode* head = lpt->_head->as_CountedLoop();
for (uint i = 0; msg == NULL && i < lpt->_body.size(); i++) {
Node* n = lpt->_body.at(i);
if (n->outcnt() == 0) continue; // Ignore dead
if (n->is_Store()) {
if (store != NULL) {
msg = "multiple stores";
break;
}
int opc = n->Opcode();
if (opc == Op_StoreP || opc == Op_StoreN || opc == Op_StoreCM) {
msg = "oop fills not handled";
break;
}
Node* value = n->in(MemNode::ValueIn);
if (!lpt->is_invariant(value)) {
msg = "variant store value";
}
store = n;
store_value = value;
} else if (n->is_If() && n != head->loopexit()) {
msg = "extra control flow";
msg_node = n;
}
}
if (store == NULL) {
// No store in loop
return false;
}
if (msg == NULL && head->stride_con() != 1) {
// could handle negative strides too
if (head->stride_con() < 0) {
msg = "negative stride";
} else {
msg = "non-unit stride";
}
}
if (msg == NULL && !store->in(MemNode::Address)->is_AddP()) {
msg = "can't handle store address";
msg_node = store->in(MemNode::Address);
}
// Make sure there is an appropriate fill routine
BasicType t = store->as_Mem()->memory_type();
const char* fill_name;
if (msg == NULL &&
StubRoutines::select_fill_function(t, false, fill_name) == NULL) {
msg = "unsupported store";
msg_node = store;
}
if (msg != NULL) {
#ifndef PRODUCT
if (TraceOptimizeFill) {
tty->print_cr("not fill intrinsic candidate: %s", msg);
if (msg_node != NULL) msg_node->dump();
}
#endif
return false;
}
// Make sure the address expression can be handled. It should be
// head->phi * elsize + con. head->phi might have a ConvI2L.
Node* elements[4];
Node* conv = NULL;
int count = store->in(MemNode::Address)->as_AddP()->unpack_offsets(elements, ARRAY_SIZE(elements));
for (int e = 0; e < count; e++) {
Node* n = elements[e];
if (n->is_Con() && con == NULL) {
con = n;
} else if (n->Opcode() == Op_LShiftX && shift == NULL) {
Node* value = n->in(1);
#ifdef _LP64
if (value->Opcode() == Op_ConvI2L) {
conv = value;
value = value->in(1);
}
#endif
if (value != head->phi()) {
msg = "unhandled shift in address";
} else {
shift = n;
assert(type2aelembytes(store->as_Mem()->memory_type(), true) == 1 << shift->in(2)->get_int(), "scale should match");
}
} else if (n->Opcode() == Op_ConvI2L && conv == NULL) {
if (n->in(1) == head->phi()) {
conv = n;
} else {
msg = "unhandled input to ConvI2L";
}
} else if (n == head->phi()) {
// no shift, check below for allowed cases
} else {
msg = "unhandled node in address";
msg_node = n;
}
}
if (count == -1) {
msg = "malformed address expression";
msg_node = store;
}
// byte sized items won't have a shift
if (msg == NULL && shift == NULL && t != T_BYTE && t != T_BOOLEAN) {
msg = "can't find shift";
msg_node = store;
}
if (msg != NULL) {
#ifndef PRODUCT
if (TraceOptimizeFill) {
tty->print_cr("not fill intrinsic: %s", msg);
if (msg_node != NULL) msg_node->dump();
}
#endif
return false;
}
// No make sure all the other nodes in the loop can be handled
VectorSet ok(Thread::current()->resource_area());
// store related values are ok
ok.set(store->_idx);
ok.set(store->in(MemNode::Memory)->_idx);
// Loop structure is ok
ok.set(head->_idx);
ok.set(head->loopexit()->_idx);
ok.set(head->phi()->_idx);
ok.set(head->incr()->_idx);
ok.set(head->loopexit()->cmp_node()->_idx);
ok.set(head->loopexit()->in(1)->_idx);
// Address elements are ok
if (con) ok.set(con->_idx);
if (shift) ok.set(shift->_idx);
if (conv) ok.set(conv->_idx);
for (uint i = 0; msg == NULL && i < lpt->_body.size(); i++) {
Node* n = lpt->_body.at(i);
if (n->outcnt() == 0) continue; // Ignore dead
if (ok.test(n->_idx)) continue;
// Backedge projection is ok
if (n->is_IfTrue() && n->in(0) == head->loopexit()) continue;
if (!n->is_AddP()) {
msg = "unhandled node";
msg_node = n;
break;
}
}
// Make sure no unexpected values are used outside the loop
for (uint i = 0; msg == NULL && i < lpt->_body.size(); i++) {
Node* n = lpt->_body.at(i);
// These values can be replaced with other nodes if they are used
// outside the loop.
if (n == store || n == head->loopexit() || n == head->incr()) continue;
for (SimpleDUIterator iter(n); iter.has_next(); iter.next()) {
Node* use = iter.get();
if (!lpt->_body.contains(use)) {
msg = "node is used outside loop";
// lpt->_body.dump();
msg_node = n;
break;
}
}
}
#ifdef ASSERT
if (TraceOptimizeFill) {
if (msg != NULL) {
tty->print_cr("no fill intrinsic: %s", msg);
if (msg_node != NULL) msg_node->dump();
} else {
tty->print_cr("fill intrinsic for:");
}
store->dump();
if (Verbose) {
lpt->_body.dump();
}
}
#endif
return msg == NULL;
}
bool PhaseIdealLoop::intrinsify_fill(IdealLoopTree* lpt) {
// Only for counted inner loops
if (!lpt->is_counted() || !lpt->is_inner()) {
return false;
}
// Must have constant stride
CountedLoopNode* head = lpt->_head->as_CountedLoop();
if (!head->stride_is_con() || !head->is_normal_loop()) {
return false;
}
// Check that the body only contains a store of a loop invariant
// value that is indexed by the loop phi.
Node* store = NULL;
Node* store_value = NULL;
Node* shift = NULL;
Node* offset = NULL;
if (!match_fill_loop(lpt, store, store_value, shift, offset)) {
return false;
}
// Now replace the whole loop body by a call to a fill routine that
// covers the same region as the loop.
Node* base = store->in(MemNode::Address)->as_AddP()->in(AddPNode::Base);
// Build an expression for the beginning of the copy region
Node* index = head->init_trip();
#ifdef _LP64
index = new (C, 2) ConvI2LNode(index);
_igvn.register_new_node_with_optimizer(index);
#endif
if (shift != NULL) {
// byte arrays don't require a shift but others do.
index = new (C, 3) LShiftXNode(index, shift->in(2));
_igvn.register_new_node_with_optimizer(index);
}
index = new (C, 4) AddPNode(base, base, index);
_igvn.register_new_node_with_optimizer(index);
Node* from = new (C, 4) AddPNode(base, index, offset);
_igvn.register_new_node_with_optimizer(from);
// Compute the number of elements to copy
Node* len = new (C, 3) SubINode(head->limit(), head->init_trip());
_igvn.register_new_node_with_optimizer(len);
BasicType t = store->as_Mem()->memory_type();
bool aligned = false;
if (offset != NULL && head->init_trip()->is_Con()) {
int element_size = type2aelembytes(t);
aligned = (offset->find_intptr_t_type()->get_con() + head->init_trip()->get_int() * element_size) % HeapWordSize == 0;
}
// Build a call to the fill routine
const char* fill_name;
address fill = StubRoutines::select_fill_function(t, aligned, fill_name);
assert(fill != NULL, "what?");
// Convert float/double to int/long for fill routines
if (t == T_FLOAT) {
store_value = new (C, 2) MoveF2INode(store_value);
_igvn.register_new_node_with_optimizer(store_value);
} else if (t == T_DOUBLE) {
store_value = new (C, 2) MoveD2LNode(store_value);
_igvn.register_new_node_with_optimizer(store_value);
}
Node* mem_phi = store->in(MemNode::Memory);
Node* result_ctrl;
Node* result_mem;
const TypeFunc* call_type = OptoRuntime::array_fill_Type();
int size = call_type->domain()->cnt();
CallLeafNode *call = new (C, size) CallLeafNoFPNode(call_type, fill,
fill_name, TypeAryPtr::get_array_body_type(t));
call->init_req(TypeFunc::Parms+0, from);
call->init_req(TypeFunc::Parms+1, store_value);
call->init_req(TypeFunc::Parms+2, len);
call->init_req( TypeFunc::Control, head->init_control());
call->init_req( TypeFunc::I_O , C->top() ) ; // does no i/o
call->init_req( TypeFunc::Memory , mem_phi->in(LoopNode::EntryControl) );
call->init_req( TypeFunc::ReturnAdr, C->start()->proj_out(TypeFunc::ReturnAdr) );
call->init_req( TypeFunc::FramePtr, C->start()->proj_out(TypeFunc::FramePtr) );
_igvn.register_new_node_with_optimizer(call);
result_ctrl = new (C, 1) ProjNode(call,TypeFunc::Control);
_igvn.register_new_node_with_optimizer(result_ctrl);
result_mem = new (C, 1) ProjNode(call,TypeFunc::Memory);
_igvn.register_new_node_with_optimizer(result_mem);
// If this fill is tightly coupled to an allocation and overwrites
// the whole body, allow it to take over the zeroing.
AllocateNode* alloc = AllocateNode::Ideal_allocation(base, this);
if (alloc != NULL && alloc->is_AllocateArray()) {
Node* length = alloc->as_AllocateArray()->Ideal_length();
if (head->limit() == length &&
head->init_trip() == _igvn.intcon(0)) {
if (TraceOptimizeFill) {
tty->print_cr("Eliminated zeroing in allocation");
}
alloc->maybe_set_complete(&_igvn);
} else {
#ifdef ASSERT
if (TraceOptimizeFill) {
tty->print_cr("filling array but bounds don't match");
alloc->dump();
head->init_trip()->dump();
head->limit()->dump();
length->dump();
}
#endif
}
}
// Redirect the old control and memory edges that are outside the loop.
Node* exit = head->loopexit()->proj_out(0);
_igvn.replace_node(exit, result_ctrl);
_igvn.replace_node(store, result_mem);
// Any uses the increment outside of the loop become the loop limit.
_igvn.replace_node(head->incr(), head->limit());
// Disconnect the head from the loop.
for (uint i = 0; i < lpt->_body.size(); i++) {
Node* n = lpt->_body.at(i);
_igvn.replace_node(n, C->top());
}
return true;
}
hotspot/src/share/vm/opto/loopnode.cpp
浏览文件 @ 6d9460f7
......@@ -1673,6 +1673,12 @@ void PhaseIdealLoop::build_and_optimize(bool do_split_ifs, bool do_loop_pred) {
_ltree_root->_child->loop_predication(this);
}
if (OptimizeFill && UseLoopPredicate && C->has_loops() && !C->major_progress()) {
if (do_intrinsify_fill()) {
C->set_major_progress();
}
}
// Perform iteration-splitting on inner loops. Split iterations to avoid
// range checks or one-shot null checks.
......
hotspot/src/share/vm/opto/loopnode.hpp
浏览文件 @ 6d9460f7
......@@ -937,6 +937,12 @@ public:
// same block. Split thru the Region.
void do_split_if( Node *iff );
// Conversion of fill/copy patterns into intrisic versions
bool do_intrinsify_fill();
bool intrinsify_fill(IdealLoopTree* lpt);
bool match_fill_loop(IdealLoopTree* lpt, Node*& store, Node*& store_value,
Node*& shift, Node*& offset);
private:
// Return a type based on condition control flow
const TypeInt* filtered_type( Node *n, Node* n_ctrl);
......
hotspot/src/share/vm/opto/runtime.cpp
浏览文件 @ 6d9460f7
......@@ -645,6 +645,22 @@ const TypeFunc* OptoRuntime::generic_arraycopy_Type() {
}
const TypeFunc* OptoRuntime::array_fill_Type() {
// create input type (domain)
const Type** fields = TypeTuple::fields(3);
fields[TypeFunc::Parms+0] = TypePtr::NOTNULL;
fields[TypeFunc::Parms+1] = TypeInt::INT;
fields[TypeFunc::Parms+2] = TypeInt::INT;
const TypeTuple *domain = TypeTuple::make(TypeFunc::Parms + 3, fields);
// create result type
fields = TypeTuple::fields(1);
fields[TypeFunc::Parms+0] = NULL; // void
const TypeTuple *range = TypeTuple::make(TypeFunc::Parms, fields);
return TypeFunc::make(domain, range);
}
//------------- Interpreter state access for on stack replacement
const TypeFunc* OptoRuntime::osr_end_Type() {
// create input type (domain)
......
hotspot/src/share/vm/opto/runtime.hpp
浏览文件 @ 6d9460f7
......@@ -260,6 +260,8 @@ private:
static const TypeFunc* generic_arraycopy_Type();
static const TypeFunc* slow_arraycopy_Type(); // the full routine
static const TypeFunc* array_fill_Type();
// leaf on stack replacement interpreter accessor types
static const TypeFunc* osr_end_Type();
......
hotspot/src/share/vm/runtime/arguments.cpp
浏览文件 @ 6d9460f7
......@@ -1513,6 +1513,9 @@ void Arguments::set_aggressive_opts_flags() {
if (AggressiveOpts && FLAG_IS_DEFAULT(OptimizeStringConcat)) {
FLAG_SET_DEFAULT(OptimizeStringConcat, true);
}
if (AggressiveOpts && FLAG_IS_DEFAULT(OptimizeFill)) {
FLAG_SET_DEFAULT(OptimizeFill, true);
}
#endif
if (AggressiveOpts) {
......
hotspot/src/share/vm/runtime/stubRoutines.cpp
浏览文件 @ 6d9460f7
......@@ -97,6 +97,15 @@ address StubRoutines::_checkcast_arraycopy = NULL;
address StubRoutines::_unsafe_arraycopy = NULL;
address StubRoutines::_generic_arraycopy = NULL;
address StubRoutines::_jbyte_fill;
address StubRoutines::_jshort_fill;
address StubRoutines::_jint_fill;
address StubRoutines::_arrayof_jbyte_fill;
address StubRoutines::_arrayof_jshort_fill;
address StubRoutines::_arrayof_jint_fill;
double (* StubRoutines::_intrinsic_log )(double) = NULL;
double (* StubRoutines::_intrinsic_log10 )(double) = NULL;
double (* StubRoutines::_intrinsic_exp )(double) = NULL;
......@@ -195,6 +204,46 @@ void StubRoutines::initialize2() {
#undef TEST_ARRAYCOPY
#define TEST_FILL(type) \
if (_##type##_fill != NULL) { \
union { \
double d; \
type body[96]; \
} s; \
\
int v = 32; \
for (int offset = -2; offset <= 2; offset++) { \
for (int i = 0; i < 96; i++) { \
s.body[i] = 1; \
} \
type* start = s.body + 8 + offset; \
for (int aligned = 0; aligned < 2; aligned++) { \
if (aligned) { \
if (((intptr_t)start) % HeapWordSize == 0) { \
((void (*)(type*, int, int))StubRoutines::_arrayof_##type##_fill)(start, v, 80); \
} else { \
continue; \
} \
} else { \
((void (*)(type*, int, int))StubRoutines::_##type##_fill)(start, v, 80); \
} \
for (int i = 0; i < 96; i++) { \
if (i < (8 + offset) || i >= (88 + offset)) { \
assert(s.body[i] == 1, "what?"); \
} else { \
assert(s.body[i] == 32, "what?"); \
} \
} \
} \
} \
} \
TEST_FILL(jbyte);
TEST_FILL(jshort);
TEST_FILL(jint);
#undef TEST_FILL
#define TEST_COPYRTN(type) \
test_arraycopy_func(CAST_FROM_FN_PTR(address, Copy::conjoint_##type##s_atomic), sizeof(type)); \
test_arraycopy_func(CAST_FROM_FN_PTR(address, Copy::arrayof_conjoint_##type##s), (int)MAX2(sizeof(HeapWord), sizeof(type)))
......@@ -315,3 +364,39 @@ JRT_LEAF(void, StubRoutines::arrayof_oop_copy(HeapWord* src, HeapWord* dest, siz
Copy::arrayof_conjoint_oops(src, dest, count);
gen_arraycopy_barrier((oop *) dest, count);
JRT_END
address StubRoutines::select_fill_function(BasicType t, bool aligned, const char* &name) {
#define RETURN_STUB(xxx_fill) { \
name = #xxx_fill; \
return StubRoutines::xxx_fill(); }
switch (t) {
case T_BYTE:
case T_BOOLEAN:
if (!aligned) RETURN_STUB(jbyte_fill);
RETURN_STUB(arrayof_jbyte_fill);
case T_CHAR:
case T_SHORT:
if (!aligned) RETURN_STUB(jshort_fill);
RETURN_STUB(arrayof_jshort_fill);
case T_INT:
case T_FLOAT:
if (!aligned) RETURN_STUB(jint_fill);
RETURN_STUB(arrayof_jint_fill);
case T_DOUBLE:
case T_LONG:
case T_ARRAY:
case T_OBJECT:
case T_NARROWOOP:
case T_ADDRESS:
// Currently unsupported
return NULL;
default:
ShouldNotReachHere();
return NULL;
}
#undef RETURN_STUB
}
hotspot/src/share/vm/runtime/stubRoutines.hpp
浏览文件 @ 6d9460f7
......@@ -148,6 +148,13 @@ class StubRoutines: AllStatic {
static address _unsafe_arraycopy;
static address _generic_arraycopy;
static address _jbyte_fill;
static address _jshort_fill;
static address _jint_fill;
static address _arrayof_jbyte_fill;
static address _arrayof_jshort_fill;
static address _arrayof_jint_fill;
// These are versions of the java.lang.Math methods which perform
// the same operations as the intrinsic version. They are used for
// constant folding in the compiler to ensure equivalence. If the
......@@ -259,6 +266,16 @@ class StubRoutines: AllStatic {
static address unsafe_arraycopy() { return _unsafe_arraycopy; }
static address generic_arraycopy() { return _generic_arraycopy; }
static address jbyte_fill() { return _jbyte_fill; }
static address jshort_fill() { return _jshort_fill; }
static address jint_fill() { return _jint_fill; }
static address arrayof_jbyte_fill() { return _arrayof_jbyte_fill; }
static address arrayof_jshort_fill() { return _arrayof_jshort_fill; }
static address arrayof_jint_fill() { return _arrayof_jint_fill; }
static address select_fill_function(BasicType t, bool aligned, const char* &name);
static double intrinsic_log(double d) {
assert(_intrinsic_log != NULL, "must be defined");
return _intrinsic_log(d);
......
hotspot/src/share/vm/utilities/globalDefinitions.hpp
浏览文件 @ 6d9460f7
......@@ -529,7 +529,7 @@ extern int _type2aelembytes[T_CONFLICT+1]; // maps a BasicType to nof bytes used
#ifdef ASSERT
extern int type2aelembytes(BasicType t, bool allow_address = false); // asserts
#else
inline int type2aelembytes(BasicType t) { return _type2aelembytes[t]; }
inline int type2aelembytes(BasicType t, bool allow_address = false) { return _type2aelembytes[t]; }
#endif
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或