X86: Use short forward jumps if possible
The optimizing compiler uses 32 bit relative jumps for all forward
jumps, just in case the offset is too large to fit in one byte. Some of
the generated code knows that the jumps will in fact fit.
Use the 'NearLabel' class for the code generator and intrinsics.
Use the jecxz/jrcxz instruction for string intrinsics.
Unfortunately, conditional jumps to basic blocks don't know enough to
use this, as we don't know how much code will be generated.
This saves a whopping 0.24% for core.oat and boot.oat sizes, but
every little bit helps, and it reduces icache footprint slightly.
Change-Id: I633fe3b2e0e810b4ce12fdad8c02135644b63506
Signed-off-by: Mark Mendell <mark.p.mendell@intel.com>
diff --git a/compiler/optimizing/code_generator_x86.cc b/compiler/optimizing/code_generator_x86.cc
index 72c690d..8cddc67 100644
--- a/compiler/optimizing/code_generator_x86.cc
+++ b/compiler/optimizing/code_generator_x86.cc
@@ -1310,7 +1310,7 @@
}
// Convert the jumps into the result.
- Label done_label;
+ NearLabel done_label;
// False case: result = 0.
__ Bind(&false_label);
@@ -1978,7 +1978,7 @@
XmmRegister input = in.AsFpuRegister<XmmRegister>();
Register output = out.AsRegister<Register>();
XmmRegister temp = locations->GetTemp(0).AsFpuRegister<XmmRegister>();
- Label done, nan;
+ NearLabel done, nan;
__ movl(output, Immediate(kPrimIntMax));
// temp = int-to-float(output)
@@ -2003,7 +2003,7 @@
XmmRegister input = in.AsFpuRegister<XmmRegister>();
Register output = out.AsRegister<Register>();
XmmRegister temp = locations->GetTemp(0).AsFpuRegister<XmmRegister>();
- Label done, nan;
+ NearLabel done, nan;
__ movl(output, Immediate(kPrimIntMax));
// temp = int-to-double(output)
@@ -2592,7 +2592,7 @@
PushOntoFPStack(first, 0, 2 * elem_size, /* is_fp */ true, is_wide);
// Loop doing FPREM until we stabilize.
- Label retry;
+ NearLabel retry;
__ Bind(&retry);
__ fprem();
@@ -2706,8 +2706,8 @@
int shift;
CalculateMagicAndShiftForDivRem(imm, false /* is_long */, &magic, &shift);
- Label ndiv;
- Label end;
+ NearLabel ndiv;
+ NearLabel end;
// If numerator is 0, the result is 0, no computation needed.
__ testl(eax, eax);
__ j(kNotEqual, &ndiv);
@@ -3160,7 +3160,7 @@
}
void InstructionCodeGeneratorX86::GenerateShlLong(const Location& loc, Register shifter) {
- Label done;
+ NearLabel done;
__ shld(loc.AsRegisterPairHigh<Register>(), loc.AsRegisterPairLow<Register>(), shifter);
__ shll(loc.AsRegisterPairLow<Register>(), shifter);
__ testl(shifter, Immediate(32));
@@ -3192,7 +3192,7 @@
}
void InstructionCodeGeneratorX86::GenerateShrLong(const Location& loc, Register shifter) {
- Label done;
+ NearLabel done;
__ shrd(loc.AsRegisterPairLow<Register>(), loc.AsRegisterPairHigh<Register>(), shifter);
__ sarl(loc.AsRegisterPairHigh<Register>(), shifter);
__ testl(shifter, Immediate(32));
@@ -3227,7 +3227,7 @@
}
void InstructionCodeGeneratorX86::GenerateUShrLong(const Location& loc, Register shifter) {
- Label done;
+ NearLabel done;
__ shrd(loc.AsRegisterPairLow<Register>(), loc.AsRegisterPairHigh<Register>(), shifter);
__ shrl(loc.AsRegisterPairHigh<Register>(), shifter);
__ testl(shifter, Immediate(32));
@@ -3402,7 +3402,7 @@
Location left = locations->InAt(0);
Location right = locations->InAt(1);
- Label less, greater, done;
+ NearLabel less, greater, done;
switch (compare->InputAt(0)->GetType()) {
case Primitive::kPrimLong: {
Register left_low = left.AsRegisterPairLow<Register>();
@@ -3606,7 +3606,7 @@
Register object,
Register value,
bool value_can_be_null) {
- Label is_null;
+ NearLabel is_null;
if (value_can_be_null) {
__ testl(value, value);
__ j(kEqual, &is_null);
@@ -4839,7 +4839,7 @@
Location cls = locations->InAt(1);
Register out = locations->Out().AsRegister<Register>();
uint32_t class_offset = mirror::Object::ClassOffset().Int32Value();
- Label done, zero;
+ NearLabel done, zero;
SlowPathCodeX86* slow_path = nullptr;
// Return 0 if `obj` is null.