Home » Intel Manuals » Processors » Intel SL3VS » Manual Viewer

Intel SL3VS Specification Update - Page 65

C66., MASKMOVQ Instruction Interaction with String Operation May Cause Deadlock, MOVD, CVTSI2SS

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

Page 65 highlights

INTEL® CELERON® PROCESSOR SPECIFICATION UPDATE Status: For the steppings affected, see the Summary of Changes at the beginning of this section. C66. MASKMOVQ Instruction Interaction with String Operation May Cause Deadlock Problem: Under the following scenario, combined with a specific alignment of internal events, the processor may enter a deadlock condition: 1. A store operation completes, leaving a write-combining (WC) buffer partially filled. 2. The target of a subsequent MASKMOVQ instruction is split across a cache line. 3. The data in (2) above results in a hit to the data in the WC buffer in (1). Implication: If this erratum occurs, the processor deadlock condition will occur and result in a system hang. Code execution cannot continue without a system RESET. Workaround: It is possible for BIOS code to contain a workaround for this erratum. Status: For the steppings affected, see the Summary of Changes at the beginning of this section. C67. MOVD, CVTSI2SS, or PINSRW Following Zeroing Instruction Can Cause Incorrect Result Problem: An incorrect result may be calculated after the following circumstances occur: 1. A register has been zeroed with either a SUB reg, reg instruction, or an XOR reg, reg instruction. 2. A value is moved with sign extension into the same register's lower 16 bits; or a signed integer multiply is performed to the same register's lower 16 bits. 3. The register is then copied to an MMX™ technology register using the MOVD, or converted to single precision floating-point and moved to an MMX technology register using the CVTSI2SS instruction prior to any other operations on the sign-extended value, or inserted into an MMX™ technology register using the PINSRW instruction. Specifically, the sign may be incorrectly extended into bits 16-31 of the MMX technology register. In the case of the PINSRW instruction, a non-zero value could be loaded into the MMX™ technology register. This erratum only affects the MMX™ technology register. This erratum only occurs when the following three steps occur in the order shown. This erratum may occur with up to 63 (39 for Pre-CPUID 0x6BX) intervening instructions that do not modify the sign-extended value between steps 2 and 3. 1. XOR EAX, EAX or SUB EAX, EAX 2. MOVSX AX, BL or MOVSX AX, byte ptr or MOVSX AX, BX or MOVSX AX, word ptr or IMUL BL (AX implicit, opcode F6 /5) or IMUL byte ptr (AX implicit, opcode F6 /5) or IMUL AX, BX (opcode 0F AF /r) 57

Section	Page
CONTENTS	3
REVISION HISTORY	7
Date of Revision	7
Version	7
Description	7
PREFACE	8
Nomenclature	8
Specification Update for the Intel® Celeron® Processor	9
GENERAL INFORMATION	9
Intel® Celeron® Processor and Boxed Intel® Celeron® Processor Markings (S.E.P. Package)	9
Intel® Celeron® Processor and Boxed Intel® Celeron® Processor Markings (PPGA Package)	10
Intel® Celeron® Processor and Boxed Intel® Celeron® Processor Markings (FC-PGA/FC-PGA2 Package)	11
IDENTIFICATION INFORMATION	12
SUMMARY OF CHANGES	19
Summary of Errata	28
Summary of Documentation Changes	30
Summary of Specification Clarifications	31
Summary of Specification Changes	32
ERRATA	33
C1.	33
FP Data Operand Pointer May Be Incorrectly Calculated After FP Access Which Wraps 64Kbyte Boundary in 16Bit Code	33
Status: For the steppings affected see the Summary of Changes at the beginning of this section.	33
C2.	33
Differences Exist in Debug Exception Reporting	33
C3.	34
Code Fetch Matching Disabled Debug Register May Cause Debug Exception	34
Status: For the steppings affected see the Summary of Changes at the beginning of this section.	34
C4.	35
FP Inexact-Result Exception Flag May Not Be Set	35
C5.	35
BTM for SMI Will Contain Incorrect FROM EIP	35
C6.	36
I/O Restart in SMM May Fail After Simultaneous MCE	36
C7.	36
Branch Traps Do Not Function If BTMs Are Also Enabled	36
C8.	36
Machine Check Exception Handler May Not Always Execute Successfully	36
C9.	37
LBER May Be Corrupted After Some Events	37
C10.	37
BTMs May Be Corrupted During Simultaneous L1 Cache Line Replacement	37
C11.	38
Potential Early Deassertion of LOCK# During Split-Lock Cycles	38
C12.	38
A20M# May Be Inverted After Returning From SMM and Reset	38
C13.	39
Reporting of Floating-Point Exception May Be Delayed	39
C14.	39
Near CALL to ESP Creates Unexpected EIP Address	39
C15.	40
Built-in Self Test Always Gives Nonzero Result	40
C16.	40
THERMTRIP# May Not Be Asserted as Specified	40
C17.	40
Cache State Corruption in the Presence of Page A/D-bit Setting and Snoop Traffic	40
C18.	41
Snoop Cycle Generates Spurious Machine Check Exception	41
C19.	41
MOVD/MOVQ Instruction Writes to Memory Prematurely	41
C20.	42
Memory Type Undefined for Nonmemory Operations	42
C21.	42
Bus Protocol Conflict With Optimized Chipsets	42
C22.	43
FP Data Operand Pointer May Not Be Zero After Power On or Reset	43
C23.	44
MOVD Following Zeroing Instruction Can Cause Incorrect Result	44
C24.	45
Premature Execution of a Load Operation Prior to Exception Handler Invocation	45
C25.	46
Read Portion of RMW Instruction May Execute Twice	46
C26.	46
Test Pin Must Be High During Power Up	46
C27.	46
Intervening Writeback May Occur During Locked Transaction	46
C28.	47
MC2_STATUS MSR Has Model-Specific Error Code and Machine Check Architecture Error Code Reversed	47
C29.	47
MOV With Debug Register Causes Debug Exception	47
C30.	48
Upper Four PAT Entries Not Usable With Mode B or Mode C Paging	48
C31.	48
Incorrect Memory Type May Be Used When MTRRs Are Disabled	48
C32.	49
Misprediction in Program Flow May Cause Unexpected Instruction Execution	49
C33.	49
Data Breakpoint Exception in a Displacement Relative Near Call May Corrupt EIP	49
C34.	49
System Bus ECC Not Functional With 2:1 Ratio	49
C35.	50
Fault on REP CMPS/SCAS Operation May Cause Incorrect EIP	50
C36.	50
RDMSR or WRMSR To Invalid MSR Address May Not Cause GP Fault	50
C37.	51
SYSENTER/SYSEXIT Instructions Can Implicitly Load “Null Segment Selector” to SS and CS Registers	51
C38.	51
PRELOAD Followed by EXTEST Does Not Load Boundary Scan Data	51
C39.	52
Far Jump to New TSS With D-bit Cleared May Cause System Hang	52
C40.	52
Incorrect Chunk Ordering May Prevent Execution of the Machine Check Exception Handler After BINIT#	52
C41.	53
UC Write May Be Reordered Around a Cacheable Write	53
C42.	53
Resume Flag May Not Be Cleared After Debug Exception	53
C43.	54
Internal Cache Protocol Violation May Cause System Hang	54
C44.	54
GP# Fault on WRMSR to ROB_CR_BKUPTMPDR6	54
C45.	55
Machine Check Exception May Occur Due to Improper Line Eviction in the IFU	55
C46.	55
Lower Bits of SMRAM SMBASE Register Cannot Be Written With an ITP	55
C47.	56
Task Switch May Cause Wrong PTE and PDE Access Bit to be Set	56
C48.	56
Cross Modifying Code Operations on a Jump Instruction May Cause a General Protection Fault	56
C49.	57
Deadlock May Occur Due To Illegal-Instruction/Page-Miss Combination	57
C50.	57
FLUSH# Assertion Following STPCLK# May Prevent CPU Clocks From Stopping	57
C51.	58
Floating-Point Exception Condition May be Deferred	58
C52.	58
Cache Line Reads May Result in Eviction of Invalid Data	58
C53.	59
FLUSH# Servicing Delayed While Waiting for STARTUP_IPI in 2way MP Systems	59
C54.	59
Double ECC Error on Read May Result in BINIT#	59
C55.	60
MCE Due to L2 Parity Error Gives L1 MCACOD.LL	60
C56.	60
EFLAGS Discrepancy on a Page Fault After a Multiprocessor TLB Shootdown	60
C57.	61
Mixed Cacheability of Lock Variables Is Problematic in MP Systems	61
C58.	61
INT 1 with DR7.GD set does not clear DR7.GD	61
C59.	62
Potential Loss of Data Coherency During MP Data Ownership Transfer	62
C60.	62
Misaligned Locked Access to APIC Space Results In a Hang	62
C61.	63
Memory Ordering Based Synchronization May Cause a Livelock Condition in MP Systems	63
C62.	64
Processor May Assert DRDY# on a Write With No Data	64
C63.	64
Machine Check Exception May Occur Due to Improper Line Eviction in the IFU	64
C65.	64
Snoop Request May Cause DBSY# Hang	64
C66.	65
MASKMOVQ Instruction Interaction with String Operation May Cause Deadlock	65
C67.	65
MOVD, CVTSI2SS, or PINSRW Following Zeroing Instruction Can Cause Incorrect Result	65
C68.	67
Snoop Probe During FLUSH# Could Cause L2 to be Left in Shared State.	67
C69.	67
Livelock May Occur Due to IFU Line Eviction	67
C70.	68
Selector for the LTR/LLDT Register May Get Corrupted	68
C71.	68
INIT Does Not Clear Global Entries in the TLB	68
C72.	68
VM Bit Will be Cleared on a Double Fault Handler	68
C73.	69
Memory Aliasing with Inconsistent A and D bits May Cause Processor Deadlock	69
C74.	69
Processor may Report Invalid TSS Fault Instead of Double Fault During Mode C Paging	69
C75.	69
APIC Failure at CPU Core/System Bus Frequency of 766/66 MHz	69
C76.	70
Machine Check Exception may Occur When Interleaving Code Between Different Memory Types	70
C77.	70
Wrong ESP Register Values During a Fault in VM86 Mode	70
C78.	70
APIC ICR Write May Cause Interrupt Not to be Sent When ICR Delivery Bit Pending	70
C79.	71
The instruction Fetch Unit (IFU) May Fetch Instructions Based Upon Stale CR3 Data After a Write to CR3 Register	71
C80.	71
The Processor Might not Exit Sleep State Properly Upon De-assertion of CPUSLP# Signal	71
C81.	71
During Boundary Scan, BCLK not Sampled High When SLP# is Asserted Low	71
C82.	72
Incorrect Assertion of THERMTRIP# Signal	72
C83.	74
Under Some Complex Conditions, the Instructions in the shadow of a JMP FAR may be Unintentionally Executed and Retired	74
C84.	74
Processor Does not Flag #GP on Non-zero Write to Certain MSRs	74
C111	86
Performance Monitoring Event FP_MMX_TRANS_TO_MMX May Not Count Some Transitions	86
DOCUMENTATION CHANGES	87
C1.	87
SSE and SSE2 Instructions Opcodes	87
C3.	88
Direction Flag (DF) Mistakenly Denoted as a System Flag	88
C4.	88
Fopcode Compatibility Mode	88
C5.	89
FCOS, FPTAN, FSIN, and FSINCOS Trigonometric Domain	89
NOT correct.	89
C6.	89
Incorrect Description of stack	89
C7.	90
EFLAGS Register Correction	90
C8.	90
PSE-36 Paging Mechanism	90
C9.	91
0x33 Opcode	91
Opcode Instruction Description	91
C10.	91
Incorrect Information for SLDT	91
C11.	91
LGDT/LIDT Instruction Information Correction	91
C12.	92
Errors in Instruction Set Reference	92
C13.	95
RSM Instruction Set Summary	95
C14.	95
Correct MOVAPS and MOVAPD Operand Section	95
C15.	96
DAA—Decimal Adjust AL after Addition	96
Operation	96
C16.	97
DAS—Decimal Adjust AL after Subtraction	97
Operation	97
C17.	98
Omission of Dependency Between BTM and LBR	98
C18.	99
I/O Permissions Bitmap Base Addy > 0xDFFF Does not	99
Cause #GP(0) Fault	99
C19.	99
C20.	100
C21.	100
C22.	101
Cache Description	101
C23.	101
Instruction Formats and Encoding	101
C24.	101
Machine-Check Initialization	101
C1.	105
PWRGOOD Inactive Pulse Width	105
C2.	105
Floating-Point Opcode Clarification	105
C3.	106
MTRR Initialization Clarification	106
C4.	106
Non-AGTL+ Output Low Current Clarification	106
SPECIFICATION CHANGES	107
C1.	107
RESET# Pin Definition	107
C2.	107
Tco max revision for 533A, 566 & 600 MHz	107
C3.	108

Match case Limit results 1 per page

INTEL

CELERON® PROCESSOR SPECIFICATION UPDATE

Status:

For the steppings affected, see the

Summary of Changes

at the beginning of this section.

C66.

MASKMOVQ Instruction Interaction with String Operation

May Cause Deadlock

Problem:

Under the following scenario, combined with a specific alignment of internal events, the processor

may enter a deadlock condition:

A store operation completes, leaving a write-combining (WC) buffer partially filled.

The target of a subsequent MASKMOVQ instruction is split across a cache line.

The data in (2) above results in a hit to the data in the WC buffer in (1).

Implication:

If this erratum occurs, the processor deadlock condition will occur and result in a system hang.

Code execution cannot continue without a system RESET.

Workaround:

It is possible for BIOS code to contain a workaround for this erratum.

Status:

For the steppings affected, see the

Summary of Changes

at the beginning of this section.

C67.

MOVD, CVTSI2SS, or PINSRW Following Zeroing Instruction

Can Cause Incorrect Result

Problem:

An incorrect result may be calculated after the following circumstances occur:

A register has been zeroed with either a SUB reg, reg instruction, or an XOR reg, reg instruction.

A value is moved with sign extension into the same register’s lower 16 bits; or a signed integer

multiply is performed to the same register’s lower 16 bits.

The register is then copied to an MMX™ technology register using the MOVD, or converted to

single precision floating-point and moved to an MMX technology register using the CVTSI2SS instruction prior

to any other operations on the sign-extended value, or inserted into an MMX™ technology register using the

PINSRW instruction.

Specifically, the sign may be incorrectly extended into bits 16-31 of the MMX technology register. In the case

of the PINSRW instruction, a non-zero value could be loaded into the MMX™ technology register. This

erratum only affects the MMX™ technology register.

This erratum only occurs when the following three steps occur in the order shown. This erratum may occur

with up to 63 (39 for Pre-CPUID 0x6BX) intervening instructions that do not modify the sign-extended value

between steps 2 and 3.

XOR EAX, EAX

or SUB EAX, EAX

MOVSX AX, BL

or MOVSX AX, byte ptr <memory address> or MOVSX AX, BX

or MOVSX AX, word ptr <memory address> or IMUL BL (AX implicit, opcode F6 /5)

or IMUL byte ptr <memory address> (AX implicit, opcode F6 /5) or IMUL AX, BX (opcode 0F AF /r)