archive/jdk
1
0
Fork 0
mirror of https://github.com/openjdk/jdk.git synced 2025-08-26 14:24:46 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

8005419: Improve intrinsics code performance on x86 by using AVX2

Browse source 
Use 256bit vpxor,vptest instructions in String.compareTo() and equals() intrinsics.

Reviewed-by: twisti
This commit is contained in:
Vladimir Kozlov 2013-01-08 11:30:51 -08:00
parent b3fe91a803
commit 38c81fb411
5 changed files with 290 additions and 35 deletions
Download patch file Download diff file Expand all files Collapse all files

20
hotspot/src/cpu/x86/vm/assembler_x86.cpp
View file

@ -2468,6 +2468,26 @@ void Assembler::ptest(XMMRegister dst, XMMRegister src) {
emit_int8((unsigned char)(0xC0 | encode));
}
void Assembler::vptest(XMMRegister dst, Address src) {
assert(VM_Version::supports_avx(), "");
InstructionMark im(this);
bool vector256 = true;
assert(dst != xnoreg, "sanity");
int dst_enc = dst->encoding();
// swap src<->dst for encoding
vex_prefix(src, dst_enc, dst_enc, VEX_SIMD_66, VEX_OPCODE_0F_38, false, vector256);
emit_int8(0x17);
emit_operand(dst, src);
}
void Assembler::vptest(XMMRegister dst, XMMRegister src) {
assert(VM_Version::supports_avx(), "");
bool vector256 = true;
int encode = vex_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_38);
emit_int8(0x17);
emit_int8((unsigned char)(0xC0 | encode));
}
void Assembler::punpcklbw(XMMRegister dst, Address src) {
NOT_LP64(assert(VM_Version::supports_sse2(), ""));
assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes");