Monday, June 02, 2014 International Symposium on Memory Management Vancouver BC, October 2004 Barriers: Friend or Foe? Steve Blackburn Department of Computer.

Monday, April 10, 2023

International Symposium on Memory Management

Vancouver BC, October 2004

Barriers: Friend or Foe?

Steve BlackburnDepartment of Computer Science

Australian National University

Tony HoskingDepartment of Computer Sciences

Purdue University




Read & Write Barrier Costs

Are r/w barrier costs significant?




Read and Write Barriers• Algorithmically powerful mechanisms

– Extend semantics of each read/write• Particularly useful to GC• Untested assumption:

“read/write barriers are expensive”– Curtails creativity in GC algorithm

development– Encourages (unnecessary?) work on

avoidance• Prior work

– [Zorn 1990] (used simulation & traces)– [Blackburn & McKinley 2002] (compilation & inlining)




Our Contributions

• Methodology for measurement• Evaluate mutator overhead

– 5 common w/b, 2 r/b– 9 benchmarks– 3 architectures (AMD, P4, PPC)– Exclude compiler, GC from

measurements




Methodology• Want to remove barrier

– Compare with and without barrier• Add full trace to generational collector

– Remembered objects irrelevant– Can include/exclude barrier

• MMTk, Jikes RVM– Hardware performance counters– Pseudo-adaptive (realistic, deterministic)– Second iteration (avoid compiler overhead)– Best of 5 (least disturbed)




Write Barrier Code

1 public final void writeBarrier(ObjectReference src, Address slot,2 ObjectReference tgt, int mode)3 throws InlinePragma {4 // insert write barrier code here5 slot.store (tgt); 6 }




Write Barrier Code cont.

Java PPC asm x86 asm

Boundary(Slot)

4 if (slot.LT(NURSERY_START)5 && tgt.GE(NURSERY_START))6 remSlots.insert(slot);

1 liu R3,0x6e102 cmplW cr1,R30,R33 bge 1 544 liu R3,0x6e105 cmplW cr1,R31,R36 bge 1 7c

1 cmp edi 0xa02000002 jlge 03 cmp ebx 0xa02000004 jlge 0

Object 4 if (getHeader(src)5 .and(LOGGING_MASK)6 .EQ(UNLOGGED))7 rememberObject(src);

1 lwz R4,-8(R5)2 rlinm R4,R4,0x0,0x1d,0x1d3 cmpiW cr1,R4,0x44 beq 1 78

1 mov ecx -8[edx]2 and ecx 43 cmp ecx 44 jeq 0

Card 4 int card=src.rshl(LOG_CARD_SIZE);5 cardTable.add(card).store((byte) 1);

1 lwz R5,0x1664(JT)2 rlinm R6,R3,0x16,0xa,0x1f3 lil R7,0x14 stbx R7,R5,R6

1 mov ebx [0x290279a]2 shr eax 103 mov [0+ebx+eax<<0] 1

Zone 4 if (slot.xor(tgt).GE(ZONE_SIZE))5 remSlots.insert(slot);

1 xor R3,R30,R312 liu R5,0x403 cmplW cr1,R3,R54 bge 1 74

1 mov edi eax2 mov eax edi3 xor eax ebx4 cmp eax 0x4000005 jlge 0




Experiments: Hardware

• 3 platforms:– 1.9GHz AMD Athlon XP 2600 1GB– 2.6GHz Pentium 4 1GB – 1.6GHz PowerPC 970 768MB

• AMD and Intel performance counters– cycles– instructions retired– L1/L2 cache misses– TLB misses– both mutator and collector, separately




Experiments: Software

• MMTk in Jikes RVM version 2.3.2+CVS – ignore remsets GC configuration (now in

MMTk)– patched to support performance counters– pseudo-adaptive compilation– read barriers

• Debian Linux 2.6.0 kernel + x86 perfctr• Standalone mode




Write Barrier Overheadmean of SPECjvm98 & SPECjbb

0%

1%

2%

3%

4%

5%

6%

Boundary Object Hybrid Zone Card

Overhead

amd

p4

ppc




Write Barrier Code (Again)


Boundary(Slot)


















Boundary(Slot)


















Boundary(Slot)
















AMD Athlon 2600+ 1.9GHz

Write Barrier

-4%

-2%

0%

2%

4%

6%

8%

10%

12%

_201_compress

_202_jess

_205_raytrace

_209_db_213_javac

_222_mpegaudio

_227_mtrt_228_jackpseudojbb

mean

Boundary

Object

Hybrid

Zone

Card




Intel P4 2.6GHzWrite Barrier

-4%

-2%

0%

2%

4%

6%

8%

10%

_201_compress

_202_jess

_205_raytrace

_209_db_213_javac

_222_mpegaudio


mean

Boundary

Object

Hybrid

Zone

Card




G5 PowerPC 970 1.6GHz Write Barrier

-2%

0%

2%

4%

6%

8%

10%

12%

14%

_201_compress

_202_jess

_205_raytrace

_209_db_213_javac

_222_mpegaudio


mean

Boundary

Object

Hybrid

Zone

Card




Performance Counters




Intel P4 2.6GHzWrite Barrier Retired

Instructions

-2%

0%

2%

4%

6%

8%

10%

12%

_201_compress

_202_jess

_205_raytrace

_209_db_213_javac

_222_mpegaudio


mean

Boundary

Object

Hybrid

Zone

Card




Intel P4 2.6GHzWrite Barrier L1 Misses

-50%

-40%

-30%

-20%

-10%

0%

10%

20%

_201_compress

_202_jess

_205_raytrace

_209_db_213_javac

_222_mpegaudio


mean

Boundary

Object

Hybrid

Zone

Card




Intel P4 2.6GHzWrite Barrier L2 Misses

-60%

-40%

-20%

0%

20%

40%

60%

80%

100%

120%

140%

_201_compress

_202_jess

_205_raytrace

_209_db_213_javac

_222_mpegaudio


mean

Boundary

Object

Hybrid

Zone

Card




Intel P4 2.6GHzWrite Barrier DTLB Misses

-15%

-10%

-5%

0%

5%

10%

15%

20%

25%

_201_compress

_202_jess

_205_raytrace

_209_db_213_javac

_222_mpegaudio


mean

Boundary

Object

Hybrid

Zone

Card




Read Barrier Code

1 public final ObjectReference readBarrier(ObjectReference obj,

2 Address slot, int mode)

3 throws InlinePragma {4 ObjectReference value = slot.loadObjectReference();5 return value; // insert read barrier code here6 }




Read Barrier Code cont.


Unconditional

5 return value.and(~3); 1 rlinm R3,R3,0x0,0x0,0x1d

1 and cax -4

Conditional 5 if (value.and(1).NE(1))6 return value;7 else8 return 0;

1 rlinm R4,R3,0x0,0x1f,0x1f2 cmpiW cr1,R4,0x13 bne 1 3c

1 mov edx eax2 and edx 13 cmp edx 14 mov edx 05 cmovne edx eax6 mov eax edx




Read Barrier Overheadmean of SPECjvm98 & SPECjbb

0%

5%

10%

15%

20%

25%

Unconditional Conditonal

Overhead

amd

p4

ppc




AMD Athlon 2600+ 1.9GHz Read Barrier

0%5%

10%15%20%25%30%35%40%

_201_compress

_202_jess

_205_raytrace

_209_db_213_javac

_222_mpegaudio


mean

Unconditional

Conditional




Intel P4 2.6GHzRead Barrier

0%

5%

10%

15%

20%

25%

30%

35%

_201_compress

_202_jess

_205_raytrace

_209_db_213_javac

_222_mpegaudio


mean

Unconditional

Conditional




G5 PowerPC 970 1.6GHz Read Barrier

-4%

-2%

0%

2%

4%

6%

8%

10%

12%

14%

16%

_201_compress

_202_jess

_205_raytrace

_209_db_213_javac

_222_mpegaudio


mean

Unconditional

Conditional




Conclusions• New methodology: available in MMTk

– Specific barrier patches at:http://cs.anu.edu.au/~Steve.Blackburn/pubs/wb-ismm-2004.tgz

• Barrier costs (often) surprisingly low• Barrier costs very architecturally

sensitive– GC developers: think about your target arch.– GC papers: what architecture did they use?– Architects: choices impact OO languages in

surprising ways.

Monday, June 02, 2014 International Symposium on Memory Management Vancouver BC, October 2004 Barriers: Friend or Foe? Steve Blackburn Department of Computer.

Documents

memory management vancouver

international symposium

barrier code

disturbed slide

liu r3

includeexclude barrier

rlinm r4

lwz r4