Home » AMD Manuals » Processors » AMD AMD-K6-2/400 » Manual Viewer

AMD AMD-K6-2/400 User Guide - Page 47

BranchPrediction Logic, Branch History Table

Add to My Manuals
Save this manual to your list of manuals

Page 47 highlights

23542A/0-September 2000 Preliminary Information AMD-K6™-2E+ Embedded Processor Data Sheet 2.6 Branch-Prediction Logic Sophisticated branch logic that can minimize or hide the impact of changes in program flow is designed into the AMD-K6-2E+ processor. Branches in x86 code fit into two categories: s Unconditional branches always change program flow (that is, the branches are always taken) s Conditional branches may or may not divert program flow (that is, the branches are taken or not-taken). When a conditional branch is not taken, the processor simply continues decoding and executing the next instructions in memory. Branch History Table Typical applications have up to 10% of unconditional branches and another 10% to 20% conditional branches. The AMD-K6-2E+ processor branch logic has been designed to handle this type of program behavior and to minimize its negative effects on instruction execution, such as stalls due to delayed instruction fetching and the draining of the processor pipeline. The branch logic contains an 8192-entry branch history table, a 16-entry by 16-byte branch target cache, a 16-entry return address stack, and a branch execution unit. The AMD-K6-2E+ processor handles unconditional branches without any penalty by redirecting instruction fetching to the target address of the unconditional branch. However, conditional branches require the use of the dynamic branch-prediction mechanism built into the AMD-K6-2E+ processor. A two-level adaptive history algorithm is implemented in an 8192-entry branch history table. This table stores executed branch information, predicts individual branches, and predicts the behavior of groups of branches. To accommodate the large branch history table, the AMD-K6-2E+ processor does not store predicted target addresses. Instead, the branch target addresses are calculated on-the-fly using ALUs during the decode stage. The adders calculate all possible target addresses before the instructions are fully decoded and the processor chooses which addresses are valid. Chapter 2 Internal Architecture 25

Section	Page
IF YOU HAVE QUESTIONS, WE\	3
Contents	5
List of Figures	9
List of Tables	13
Revision History	17
About this Data Sheet	19
1 AMDK6™2E+ Embedded Processor	23
1.1 AMDK6™2E+ Embedded Processor Features	25
1.2 Process Technology	29
1.3 Super7™ Platform	30
2 Internal Architecture	33
2.1 Microarchitecture Overview	33
2.2 Cache, Instruction Prefetch, and Predecode Bits	38
2.3 Instruction Fetch and Decode	39
2.4 Centralized Scheduler	43
2.5 Execution Units	44
2.6 BranchPrediction Logic	47
3 Software Environment	49
3.1 Registers	49
3.2 ModelSpecific Registers (MSR)	66
3.3 Memory Management Registers	76
3.4 Paging	78
3.5 Descriptors and Gates	81
3.6 Exceptions and Interrupts	84
3.7 Instructions Supported by the AMDK6™2E+ Processor	85
4 Logic Symbol Diagram	113
5 Signal Descriptions	115
5.1 Signal Terminology	115
5.2 A20M# (Address Bit 20 Mask)	116
5.3 A[31:3] (Address Bus)	117
5.4 ADS# (Address Strobe)	118
5.5 ADSC# (Address Strobe Copy)	118
5.6 AHOLD (Address Hold)	119
5.7 AP (Address Parity)	120
5.8 APCHK# (Address Parity Check)	121
5.9 BE[7:0]# (Byte Enables)	122
5.10 BF[2:0] (Bus Frequency)	123
5.11 BOFF# (Backoff)	124
5.12 BRDY# (Burst Ready)	125
5.13 BRDYC# (Burst Ready Copy)	126
5.14 BREQ (Bus Request)	126
5.15 CACHE# (Cacheable Access)	127
5.16 CLK (Clock)	127
5.17 D/C# (Data/Code)	128
5.18 D[63:0] (Data Bus)	129
5.19 DP[7:0] (Data Parity)	130
5.20 EADS# (External Address Strobe)	131
5.21 EWBE# (External Write Buffer Empty)	132
5.22 FERR# (FloatingPoint Error)	133
5.23 FLUSH# (Cache Flush)	134
5.24 HIT# (Inquire Cycle Hit)	135
5.25 HITM# (Inquire Cycle Hit To Modified Line)	135
5.26 HLDA (Hold Acknowledge)	136
5.27 HOLD (Bus Hold Request)	137
5.28 IGNNE# (Ignore Numeric Exception)	138
5.29 INIT (Initialization)	139
5.30 INTR (Maskable Interrupt)	140
5.31 INV (Invalidation Request)	140
5.32 KEN# (Cache Enable)	141
5.33 LOCK# (Bus Lock)	142
5.34 M/IO# (Memory or I/O)	143
5.35 NA# (Next Address)	144
5.36 NMI (NonMaskable Interrupt)	145
5.37 PCD (Page Cache Disable)	146
5.38 PCHK# (Parity Check)	147
5.39 PWT (Page Writethrough)	148
5.40 RESET (Reset)	149
5.41 RSVD (Reserved)	150
5.42 SCYC (Split Cycle)	151
5.43 SMI# (System Management Interrupt)	152
5.44 SMIACT# (System Management Interrupt Active)	153
5.45 STPCLK# (Stop Clock)	154
5.46 TCK (Test Clock)	155
5.47 TDI (Test Data Input)	155
5.48 TDO (Test Data Output)	155
5.49 TMS (Test Mode Select)	156
5.50 TRST# (Test Reset)	156
5.51 VCC2DET (VCC2 Detect)	157
5.52 VCC2H/L# (VCC2 High/Low)	158
5.53 VID[4:0] (Voltage Identification)	159
5.54 W/R# (Write/Read)	160
5.55 WB/WT# (Writeback or Writethrough)	161
5.56 Pin Tables by Type	162
5.57 Bus Cycle Definitions	164
6 AMD PowerNow!™ Technology	165
6.1 Enhanced Power Management Features	165
6.2 Dynamic Core Frequency and Core Voltage Control	172
7 Bus Cycles	175
7.1 Timing Diagrams	175
7.2 Bus States	177
7.3 Memory Reads and Writes	180
7.4 I/O Read and Write	188
7.5 Inquire and Bus Arbitration Cycles	190
7.6 Special Bus Cycles	212
8 Poweron Configuration and Initialization	221
8.1 Signals Sampled During the Falling Transition of RESET	221
8.2 RESET Requirements	222
8.3 State of Processor After RESET	222
8.4 State of Processor After INIT	225
9 Cache Organization	227
9.1 MESI States in the L1 Data Cache and L2 Cache	229
9.2 Predecode Bits	230
9.3 Cache Operation	230
9.4 Cache Disabling and Flushing	233
9.5 L2 Cache Testing	235
9.6 CacheLine Fills	235
9.7 CacheLine Replacements	236
9.8 Write Allocate	237
9.9 Prefetching	242
9.10 Cache States	243
9.11 Cache Coherency	244
9.12 Writethrough and Writeback Coherency States	249
9.13 A20M# Masking of Cache Accesses	249
10 Write Merge Buffer	251
10.1 EWBE# Control	251
10.2 Memory Type Range Registers	253
10.3 Memory-Range Restrictions	255
10.4 Examples	257
11 FloatingPoint and Multimedia Execution Units	259
11.1 FloatingPoint Execution Unit	259
11.2 Multimedia and 3DNow!™ Execution Units	261
11.3 FloatingPoint and MMX™/3DNow!™ Instruction Compatibility	262
12 System Management Mode (SMM)	263
12.1 SMM Operating Mode and Default Register Values	263
12.2 SMM StateSave Area	265
12.3 SMM Revision Identifier	267
12.4 SMM Base Address	268
12.5 Halt Restart Slot	268
12.6 I/O Trap Doubleword	269
12.7 I/O Trap Restart Slot	270
12.8 Exceptions, Interrupts, and Debug in SMM	272
13 Test and Debug	273
13.1 BuiltIn SelfTest (BIST)	273
13.2 ThreeState Test Mode	274
13.3 BoundaryScan Test Access Port (TAP)	275
13.4 Cache Inhibit	285
13.5 L2 Cache and Tag Array Testing	286
13.6 Debug	290
14 Clock Control	297
14.1 Clock Control States	297
14.2 Halt State	300
14.3 Stop Grant State	300
14.4 Stop Grant Inquire State	302
14.5 EPM Stop Grant State	303
14.6 Stop Clock State	305
15 Electrical Data	307
15.1 Operating Ranges	308
15.2 Absolute Ratings	309
15.3 DC Characteristics	309
15.4 Power Dissipation	311
15.5 Power and Grounding	313
16 Signal Switching Characteristics	317
16.1 CLK Switching Characteristics	318
16.2 Clock Switching Characteristics for 100MHz Bus Operation	318
16.3 Clock Switching Characteristics for 66MHz Bus Operation	319
16.4 Valid Delay, Float, Setup, and Hold Timings	320
16.5 Output Delay Timings for 100MHz Bus Operation	320
16.6 Input Setup and Hold Timings for 100MHz Bus Operation	322
16.7 Output Delay Timings for 66MHz Bus Operation	324
16.8 Input Setup and Hold Timings for 66MHz Bus Operation	326
16.9 RESET and Test Signal Timing	328
16.10 Timing Diagrams	331
17 Thermal Design	335
17.1 Package Thermal Specifications	335
17.2 Measuring Case Temperature	339
17.3 Layout and Airflow Considerations	339
18 Pin Designations	343
18.1 Pins Designations for CPGA Package	344
18.2 Pins Designations for OBGA Package	348
19 Package Specifications	353
19.1 321Pin Staggered CPGA Package Specification	353
19.2 349Ball OBGA Package Specification	354
20 Ordering Information	355

Match case Limit results 1 per page

Chapter 2

Internal Architecture

23542A/0—September 2000

AMD-K6™-2E+ Embedded Processor Data Sheet

Preliminary Information

2.6

Branch-Prediction Logic

Sophisticated branch logic that can minimize or hide the impact

of changes in program flow is designed into the AMD-K6-2E+

processor. Branches in x86 code fit into two categories:

■

Unconditional branches

always change program flow (that is,

the branches are always taken)

■

onditional branches

may or may not divert program flow

(that is, the branches are taken or not-taken). When a

conditional branch is not taken, the processor simply

continues decoding and executing the next instructions in

memory.

Typical applications have up to 10% of unconditional branches

and another 10% to 20% conditional branches. The

AMD-K6-2E+ processor branch logic has been designed to

handle this type of program behavior and to minimize its

negative effects on instruction execution, such as stalls due to

delayed instruction fetching and the draining of the processor

pipeline. The branch logic contains an 8192-entry branch

history table, a 16-entry by 16-byte branch target cache, a

16-entry return address stack, and a branch execution unit.

Branch History Table

The AMD-K6-2E+ processor handles unconditional branches

without any penalty by redirecting instruction fetching to the

target address of the unconditional branch. However,

conditional branches require the use of the dynamic

branch-prediction mechanism built into the AMD-K6-2E+

processor.

A two-level adaptive history algorithm is implemented in an

8192-entry branch history table. This table stores executed

branch information, predicts individual branches, and predicts

the behavior of groups of branches.

To accommodate the large branch history table, the

AMD-K6-2E+ processor does not store predicted target

addresses. Instead, the branch target addresses are calculated

on-the-fly using ALUs during the decode stage. The adders

calculate all possible target addresses before the instructions

are fully decoded and the processor chooses which addresses

are valid.

AMD AMD-K6-2/400 User Guide - Page 47

BranchPrediction Logic, Branch History Table

Page 47 highlights

BranchPrediction Logic, Branch History Table