Machine Programming – IA32 memory layout and buffer overflowCENG331: Introduction to Computer Systems7th Lecture

Instructor: Erol Sahin

Acknowledgement: Most of the slides are adapted from the ones prepared by R.E. Bryant, D.R. O’Hallaron of Carnegie-Mellon Univ.

Linux Memory Layout Stack

Runtime stack (8MB limit) Heap

Dynamically allocated storage When call malloc, calloc, new

DLLs Dynamically Linked Libraries Library routines (e.g., printf, malloc) Linked into object code when first executed

Data Statically allocated data E.g., arrays & strings declared in code

Text Executable machine instructions Read-only

Upper 2 hex digits of addressRed Hatv. 6.2~1920MBmemorylimit

FF

BF

7F

3F

C0

80

40

00

Stack

DLLs

TextData

Heap

Heap

08

Linux Memory AllocationLinked

BF

7F

3F

80

40

00

Stack

DLLs

TextData

08

Some Heap

BF

7F

3F

80

40

00

Stack

DLLs

TextData

Heap

08

MoreHeap

BF

7F

3F

80

40

00

Stack

DLLs

TextDataHeap

Heap

08

InitiallyBF

7F

3F

80

40

00

Stack

TextData

08

Text & Stack Example

(gdb) break main(gdb) run Breakpoint 1, 0x804856f in main ()(gdb) print $esp $3 = (void *) 0xbffffc78

Main Address 0x804856f should be read 0x0804856f

Stack Address 0xbffffc78

InitiallyBF

7F

3F

80

40

00

Stack

TextData

08

Dynamic Linking Example(gdb) print malloc $1 = {<text variable, no debug info>} 0x8048454 <malloc>(gdb) run Program exited normally.(gdb) print malloc $2 = {void *(unsigned int)} 0x40006240 <malloc> Initially

Code in text segment that invokes dynamic linker Address 0x8048454 should be read 0x08048454

Final Code in DLL region

LinkedBF

7F

3F

80

40

00

Stack

DLLs

TextData

08

Memory Allocation Example

char big_array[1<<24]; /* 16 MB */char huge_array[1<<28]; /* 256 MB */

int beyond;char *p1, *p2, *p3, *p4;

int useless() { return 0; }

int main(){ p1 = malloc(1 <<28); /* 256 MB */ p2 = malloc(1 << 8); /* 256 B */ p3 = malloc(1 <<28); /* 256 MB */ p4 = malloc(1 << 8); /* 256 B */ /* Some print statements ... */}

Example Addresses

$esp 0xbffffc78p3 0x500b5008p1 0x400b4008Final malloc 0x40006240p4 0x1904a640 p2 0x1904a538beyond 0x1904a524big_array 0x1804a520huge_array 0x0804a510main() 0x0804856fuseless() 0x08048560Initial malloc 0x08048454

BF

7F

3F

80

40

00

Stack

DLLs

TextDataHeap

Heap

08

Internet Worm and IM War November, 1988

Internet Worm attacks thousands of Internet hosts. How did it happen?

July, 1999 Microsoft launches MSN Messenger (instant messaging system). Messenger clients can access popular AOL Instant Messaging Service

(AIM) servers

AIMserver

AIMclient

AIMclient

MSNclient

MSNserver

Internet Worm and IM War (cont.) August 1999

Mysteriously, Messenger clients can no longer access AIM servers. Microsoft and AOL begin the IM war:

AOL changes server to disallow Messenger clients Microsoft makes changes to clients to defeat AOL changes. At least 13 such skirmishes.

How did it happen?

The Internet Worm and AOL/Microsoft War were both based on stack buffer overflow exploits!

many Unix functions do not check argument sizes. allows target buffers to overflow.

String Library Code Implementation of Unix function gets

No way to specify limit on number of characters to read

Similar problems with other Unix functions strcpy: Copies string of arbitrary length scanf, fscanf, sscanf, when given %s conversion specification

/* Get string from stdin */char *gets(char *dest){ int c = getc(); char *p = dest; while (c != EOF && c != '\n') { *p++ = c; c = getc(); } *p = '\0'; return dest;}

Vulnerable Buffer Code

int main(){ printf("Type a string:"); echo(); return 0;}

/* Echo Line */void echo(){ char buf[4]; /* Way too small! */ gets(buf); puts(buf);}

Buffer Overflow Executions

unix>./bufdemoType a string:123123

unix>./bufdemoType a string:12345Segmentation Fault

unix>./bufdemoType a string:12345678Segmentation Fault

Buffer Overflow Stack

echo:pushl %ebp # Save %ebp on stackmovl %esp,%ebpsubl $20,%esp # Allocate space on stackpushl %ebx # Save %ebxaddl $-12,%esp # Allocate space on stackleal -4(%ebp),%ebx # Compute buf as %ebp-4pushl %ebx # Push buf on stackcall gets # Call gets. . .

/* Echo Line */void echo(){ char buf[4]; /* Way too small! */ gets(buf); puts(buf);}

Return AddressSaved %ebp[3][2][1][0]buf

%ebp

StackFramefor main

StackFramefor echo

Buffer Overflow Stack Example

Before call to gets

unix> gdb bufdemo(gdb) break echoBreakpoint 1 at 0x8048583(gdb) runBreakpoint 1, 0x8048583 in echo ()(gdb) print /x *(unsigned *)$ebp$1 = 0xbffff8f8(gdb) print /x *((unsigned *)$ebp + 1)$3 = 0x804864d

8048648: call 804857c <echo> 804864d: mov 0xffffffe8(%ebp),%ebx # Return Point

Return AddressSaved %ebp[3][2][1][0]buf

%ebp

StackFramefor main

StackFramefor echo

0xbffff8d8Return AddressSaved %ebp[3][2][1][0]buf

StackFramefor main

StackFramefor echo

bf ff f8 f808 04 86 4d

xx xx xx xx

Buffer Overflow Example #1

Before Call to gets Input = “123”

No Problem

0xbffff8d8Return AddressSaved %ebp[3][2][1][0]buf

StackFramefor main

StackFramefor echo

bf ff f8 f808 04 86 4d

00 33 32 31

Return AddressSaved %ebp[3][2][1][0]buf

%ebp

StackFramefor main

StackFramefor echo

Buffer Overflow Stack Example #2Input = “12345”

8048592: push %ebx 8048593: call 80483e4 <_init+0x50> # gets 8048598: mov 0xffffffe8(%ebp),%ebx 804859b: mov %ebp,%esp 804859d: pop %ebp # %ebp gets set to invalid value 804859e: ret

echo code:

0xbffff8d8Return AddressSaved %ebp[3][2][1][0]buf

StackFramefor main

StackFramefor echo

bf ff 00 3508 04 86 4d

34 33 32 31

Return AddressSaved %ebp[3][2][1][0]buf

%ebp

StackFramefor main

StackFramefor echo

Saved value of %ebp set to 0xbfff0035

Bad news when later attempt to restore %ebp

Buffer Overflow Stack Example #3

Input = “12345678”

Return AddressSaved %ebp[3][2][1][0]buf

%ebp

StackFramefor main

StackFramefor echo

8048648: call 804857c <echo> 804864d: mov 0xffffffe8(%ebp),%ebx # Return Point

0xbffff8d8Return AddressSaved %ebp[3][2][1][0]buf

StackFramefor main

StackFramefor echo

38 37 36 3508 04 86 00

34 33 32 31

Invalid addressNo longer pointing to

desired return point

%ebp and return address corrupted

Malicious Use of Buffer Overflow

Input string contains byte representation of executable code Overwrite return address with address of buffer When bar() executes ret, will jump to exploit code

void bar() { char buf[64]; gets(buf); ... }

void foo(){ bar(); ...}

Stack after call to gets()

B

returnaddressA

foo stack frame

bar stack frameB

exploitcode

pad

data writtenbygets()

Exploits Based on Buffer Overflows Buffer overflow bugs allow remote machines to execute

arbitrary code on victim machines. Internet worm

Early versions of the finger server (fingerd) used gets() to read the argument sent by the client:

finger droh@cs.cmu.edu Worm attacked fingerd server by sending phony argument:

finger “exploit-code padding new-return-address”

exploit code: executed a root shell on the victim machine with a direct TCP connection to the attacker.

Exploits Based on Buffer Overflows Buffer overflow bugs allow remote machines to execute

arbitrary code on victim machines. IM War

AOL exploited existing buffer overflow bug in AIM clients exploit code: returned 4-byte signature (the bytes at some location in

the AIM client) to server. When Microsoft changed code to match signature, AOL changed

signature location.

Date: Wed, 11 Aug 1999 11:30:57 -0700 (PDT) From: Phil Bucking <philbucking@yahoo.com> Subject: AOL exploiting buffer overrun bug in their own software! To: rms@pharlap.com

Mr. Smith,

I am writing you because I have discovered something that I think you might find interesting because you are an Internet security expert with experience in this area. I have also tried to contact AOL but received no response.

I am a developer who has been working on a revolutionary new instant messaging client that should be released later this year. ... It appears that the AIM client has a buffer overrun bug. By itself this might not be the end of the world, as MS surely has had its share. But AOL is now *exploiting their own buffer overrun bug* to help in its efforts to block MS Instant Messenger. .... Since you have significant credibility with the press I hope that you can use this information to help inform people that behind AOL's friendly exterior they are nefariously compromising peoples' security.

Sincerely, Phil Bucking Founder, Bucking Consulting philbucking@yahoo.com

It was later determined that this email originated from within Microsoft!

Code Red Worm History

June 18, 2001. Microsoft announces buffer overflow vulnerability in IIS Internet server

July 19, 2001. over 250,000 machines infected by new virus in 9 hours

White house must change its IP address. Pentagon shut down public WWW servers for day

When We Set Up CS:APP Web Site Received strings of formGET /default.ida?

NNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNN....NNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNN%u9090%u6858%ucbd3%u7801%u9090%u6858%ucbd3%u7801%u9090%u6858%ucbd3%u7801%u9090%u9090%u8190%u00c3%u0003%u8b00%u531b%u53ff%u0078%u0000%u00=a

HTTP/1.0" 400 325 "-" "-"

Code Red Exploit Code Starts 100 threads running Spread self

Generate random IP addresses & send attack string Between 1st & 19th of month

Attack www.whitehouse.gov Send 98,304 packets; sleep for 4-1/2 hours; repeat

– Denial of service attack Between 21st & 27th of month

Deface server’s home page After waiting 2 hours

Code Red Effects Later Version Even More Malicious

Code Red II As of April, 2002, over 18,000 machines infected Still spreading

Paved Way for NIMDA Variety of propagation methods One was to exploit vulnerabilities left behind by Code Red II

Avoiding Overflow Vulnerability

Use Library Routines that Limit String Lengths fgets instead of gets strncpy instead of strcpy Don’t use scanf with %s conversion specification

Use fgets to read the string

/* Echo Line */void echo(){ char buf[4]; /* Way too small! */ fgets(buf, 4, stdin); puts(buf);}

Final Observations

Memory Layout OS/machine dependent (including kernel version) Basic partitioning: stack/data/text/heap/DLL found in most

machines Type Declarations in C

Notation obscure, but very systematic Working with Strange Code

Important to analyze nonstandard cases E.g., what happens when stack corrupted due to buffer

overflow Helps to step through with GDB

Machine Programming – x86-64CENG331: Introduction to Computer Systems7th Lecture

Instructor: Erol Sahin

Acknowledgement: Most of the slides are adapted from the ones prepared by R.E. Bryant, D.R. O’Hallaron of Carnegie-Mellon Univ.

%rax

%rbx

%rcx

%rdx

%rsi

%rdi

%rsp

%rbp

x86-64 Integer Registers

Twice the number of registers Accessible as 8, 16, 32, 64 bits

%eax

%ebx

%ecx

%edx

%esi

%edi

%esp

%ebp

%r8

%r9

%r10

%r11

%r12

%r13

%r14

%r15

%r8d

%r9d

%r10d

%r11d

%r12d

%r13d

%r14d

%r15d

%rax

%rbx

%rcx

%rdx

%rsi

%rdi

%rsp

%rbp

x86-64 Integer Registers

%r8

%r9

%r10

%r11

%r12

%r13

%r14

%r15Callee saved Callee saved

Callee saved

Callee saved

C: Callee saved

Callee saved

Callee saved

Stack pointer

Used for linking

Return value

Argument #4

Argument #1

Argument #3

Argument #2

Argument #6

Argument #5

x86-64 Registers Arguments passed to functions via registers

If more than 6 integral parameters, then pass rest on stack These registers can be used as caller-saved as well

All references to stack frame via stack pointer Eliminates need to update %ebp/%rbp

Other Registers 6+1 callee saved 2 or 3 have special uses

x86-64 Long Swap

Operands passed in registers First (xp) in %rdi, second (yp) in %rsi 64-bit pointers

No stack operations required (except ret) Avoiding stack

Can hold all local information in registers

void swap(long *xp, long *yp) { long t0 = *xp; long t1 = *yp; *xp = t1; *yp = t0;}

swap:movq (%rdi), %rdxmovq (%rsi), %raxmovq %rax, (%rdi)movq %rdx, (%rsi)ret

x86-64 Locals in the Red Zone

Avoiding Stack Pointer Change Can hold all information within small

window beyond stack pointer

/* Swap, using local array */void swap_a(long *xp, long *yp) { volatile long loc[2]; loc[0] = *xp; loc[1] = *yp; *xp = loc[1]; *yp = loc[0];}

swap_a: movq (%rdi), %rax movq %rax, -24(%rsp) movq (%rsi), %rax movq %rax, -16(%rsp) movq -16(%rsp), %rax movq %rax, (%rdi) movq -24(%rsp), %rax movq %rax, (%rsi) ret

rtn Ptr

unused

%rsp

−8loc[1]loc[0]

−16

−24Volatile tells the compiler that the value of the variable may change at any time--without any action being taken by the code the compiler finds nearby. Useful when working with I/O devices and interrupts.

x86-64 NonLeaf without Stack Frame No values held while swap

being invoked

No callee save registers needed

long scount = 0;

/* Swap a[i] & a[i+1] */void swap_ele_se (long a[], int i){ swap(&a[i], &a[i+1]); scount++;}

swap_ele_se: movslq %esi,%rsi # Sign extend i leaq (%rdi,%rsi,8), %rdi # &a[i] leaq 8(%rdi), %rsi # &a[i+1] call swap # swap() incq scount(%rip) # scount++; ret

x86-64 Call using Jump

long scount = 0;

/* Swap a[i] & a[i+1] */void swap_ele(long a[], int i){ swap(&a[i], &a[i+1]);}

swap_ele: movslq %esi,%rsi # Sign extend i leaq (%rdi,%rsi,8), %rdi # &a[i] leaq 8(%rdi), %rsi # &a[i+1] jmp swap # swap()

Will disappearBlackboard?

x86-64 Call using Jump When swap executes ret,

it will return from swap_ele

Possible since swap is a “tail call”(no instructions afterwards)

long scount = 0;

/* Swap a[i] & a[i+1] */void swap_ele(long a[], int i){ swap(&a[i], &a[i+1]);}

swap_ele: movslq %esi,%rsi # Sign extend i leaq (%rdi,%rsi,8), %rdi # &a[i] leaq 8(%rdi), %rsi # &a[i+1] jmp swap # swap()

x86-64 Stack Frame Example

Keeps values of a and i in callee save registers

Must set up stack frame to save these registers

long sum = 0;/* Swap a[i] & a[i+1] */void swap_ele_su (long a[], int i){ swap(&a[i], &a[i+1]); sum += a[i];}

swap_ele_su: movq %rbx, -16(%rsp) movslq %esi,%rbx movq %r12, -8(%rsp) movq %rdi, %r12 leaq (%rdi,%rbx,8), %rdi subq $16, %rsp leaq 8(%rdi), %rsi call swap movq (%r12,%rbx,8), %rax addq %rax, sum(%rip) movq (%rsp), %rbx movq 8(%rsp), %r12 addq $16, %rsp ret

Understanding x86-64 Stack Frameswap_ele_su: movq %rbx, -16(%rsp) # Save %rbx movslq %esi,%rbx # Extend & save i movq %r12, -8(%rsp) # Save %r12 movq %rdi, %r12 # Save a leaq (%rdi,%rbx,8), %rdi # &a[i] subq $16, %rsp # Allocate stack frame leaq 8(%rdi), %rsi # &a[i+1] call swap # swap() movq (%r12,%rbx,8), %rax # a[i] addq %rax, sum(%rip) # sum += a[i] movq (%rsp), %rbx # Restore %rbx movq 8(%rsp), %r12 # Restore %r12 addq $16, %rsp # Deallocate stack frame ret

Understanding x86-64 Stack Frameswap_ele_su: movq %rbx, -16(%rsp) # Save %rbx movslq %esi,%rbx # Extend & save i movq %r12, -8(%rsp) # Save %r12 movq %rdi, %r12 # Save a leaq (%rdi,%rbx,8), %rdi # &a[i] subq $16, %rsp # Allocate stack frame leaq 8(%rdi), %rsi # &a[i+1] call swap # swap() movq (%r12,%rbx,8), %rax # a[i] addq %rax, sum(%rip) # sum += a[i] movq (%rsp), %rbx # Restore %rbx movq 8(%rsp), %r12 # Restore %r12 addq $16, %rsp # Deallocate stack frame ret

rtn addr%r12

%rsp−8

%rbx−16

rtn addr%r12

%rsp+8

%rbx

Interesting Features of Stack Frame Allocate entire frame at once

All stack accesses can be relative to %rsp Do by decrementing stack pointer Can delay allocation, since safe to temporarily use red zone

Simple deallocation Increment stack pointer No base/frame pointer needed

x86-64 Procedure Summary Heavy use of registers

Parameter passing More temporaries since more registers

Minimal use of stack Sometimes none Allocate/deallocate entire block

Many tricky optimizations What kind of stack frame to use Calling with jump Various allocation techniques

IA32 Floating Point (x87) History

8086: first computer to implement IEEE FP separate 8087 FPU (floating point unit)

486: merged FPU and Integer Unit onto one chip Becoming obsolete with x86-64

Summary Hardware to add, multiply, and divide Floating point data registers Various control & status registers

Floating Point Formats single precision (C float): 32 bits double precision (C double): 64 bits extended precision (C long double): 80 bits

Instructiondecoder andsequencer

FPUIntegerUnit

Memory

FPU Data Register Stack (x87) FPU register format (80 bit extended precision)

FPU registers 8 registers %st(0) - %st(7) Logically form stack Top: %st(0) Bottom disappears (drops out) after too many pushs

s exp frac063647879

“Top” %st(0)%st(1)%st(2)%st(3)

Instruction Effect Description

fldz push 0.0 Load zeroflds Addr push Mem[Addr] Load single precision realfmuls Addr %st(0) %st(0)*M[Addr] Multiplyfaddp %st(1) %st(0)+%st(1);pop Add and pop

FPU instructions (x87) Large number of floating point instructions and formats

~50 basic instruction types load, store, add, multiply sin, cos, tan, arctan, and log

Often slower than math lib

Sample instructions:

FP Code Example (x87) Compute inner product of two

vectors Single precision arithmetic Common computation

float ipf (float x[], float y[], int n){ int i; float result = 0.0; for (i = 0; i < n; i++) result += x[i]*y[i]; return result;}

pushl %ebp # setup movl %esp,%ebp pushl %ebx movl 8(%ebp),%ebx # %ebx=&x movl 12(%ebp),%ecx # %ecx=&y movl 16(%ebp),%edx # %edx=n fldz # push +0.0 xorl %eax,%eax # i=0 cmpl %edx,%eax # if i>=n done jge .L3 .L5: flds (%ebx,%eax,4) # push x[i] fmuls (%ecx,%eax,4) # st(0)*=y[i] faddp # st(1)+=st(0); pop incl %eax # i++ cmpl %edx,%eax # if i<n repeat jl .L5 .L3: movl -4(%ebp),%ebx # finish movl %ebp, %esp popl %ebp ret # st(0) = result

Inner Product Stack Trace

1. fldz

0.0 %st(0)

2. flds (%ebx,%eax,4) 0.0 %st(1)

x[0] %st(0)

3. fmuls (%ecx,%eax,4)

0.0 %st(1)x[0]*y[0] %st(0)

4. faddp 0.0+x[0]*y[0] %st(0)

5. flds (%ebx,%eax,4) x[0]*y[0] %st(1)

x[1] %st(0)

6. fmuls (%ecx,%eax,4)

x[0]*y[0] %st(1)x[1]*y[1] %st(0)

7. faddp %st(0)x[0]*y[0]+x[1]*y[1]

Initialization

Iteration 0 Iteration 1

eax = i ebx = *x ecx = *y

Vector Instructions: SSE Family SSE : Streaming SIMD Extensions SIMD (single-instruction, multiple data) vector

instructions New data types, registers, operations Parallel operation on small (length 2-8) vectors of integers or floats Example:

Floating point vector instructions Available with Intel’s SSE (streaming SIMD extensions) family SSE starting with Pentium III: 4-way single precision SSE2 starting with Pentium 4: 2-way double precision All x86-64 have SSE3 (superset of SSE2, SSE)

+ x “4-way”

Intel Architectures (Focus Floating Point)

x86-64 / em64t

x86-32

x86-16

MMX

SSE

SSE2

SSE3

SSE4

8086

286386486PentiumPentium MMX

Pentium III

Pentium 4

Pentium 4E

Pentium 4F

Core 2 Duo

time

ArchitecturesProcessors Features

4-way single precision fp

2-way double precision fp

Our focus: SSE3 used for scalar (non-vector)floating point

SSE3 Registers All caller saved %xmm0 for floating point return value

%xmm0

%xmm1

%xmm2

%xmm3

%xmm4

%xmm5

%xmm6

%xmm7

%xmm8

%xmm9

%xmm10

%xmm11

%xmm12

%xmm13

%xmm14

%xmm15

Argument #1

Argument #2

Argument #3

Argument #4

Argument #5

Argument #6

Argument #7

Argument #8

128 bit = 2 doubles = 4 singles

SSE3 Registers Different data types and associated instructions Integer vectors:

16-way byte 8-way 2 bytes 4-way 4 bytes

Floating point vectors: 4-way single 2-way double

Floating point scalars: single double

128 bit LSB

SSE3 Instructions: Examples Single precision 4-way vector add: addps %xmm0 %xmm1

Single precision scalar add: addss %xmm0 %xmm1

+

%xmm0

%xmm1

+

%xmm0

%xmm1

SSE3 Instruction Names

addps addss

addpd addsd

packed (vector) single slot (scalar)

single precision

double precision this course

SSE3 Basic Instructions Moves

Usual operand form: reg → reg, reg → mem, mem → reg

Arithmetic Single Double Effect

addss addsd D ← D + S

subss subsd D ← D – S

mulss mulsd D ← D x S

divss divsd D ← D / S

maxss maxsd D ← max(D,S)

minss minsd D ← min(D,S)

sqrtss sqrtsd D ← sqrt(S)

Single Double Effect

movss movsd D ← S

x86-64 FP Code Example Compute inner product of

two vectors Single precision arithmetic Uses SSE3 instructions

float ipf (float x[], float y[], int n) { int i; float result = 0.0; for (i = 0; i < n; i++) result += x[i]*y[i]; return result;}

ipf:xorps %xmm1, %xmm1 # result = 0.0xorl %ecx, %ecx # i = 0jmp .L8 # goto middle

.L10: # loop:movslq %ecx,%rax # icpy = iincl %ecx # i++movss (%rsi,%rax,4), %xmm0 # t = y[icpy]mulss (%rdi,%rax,4), %xmm0 # t *= x[icpy]addss %xmm0, %xmm1 # result += t

.L8: # middle:cmpl %edx, %ecx # i:njl .L10 # if < goto loopmovaps %xmm1, %xmm0 # return resultret

SSE3 Conversion Instructions Conversions

Same operand forms as moves

Instruction Description

cvtss2sd single → double

cvtsd2ss double → single

cvtsi2ss int → single

cvtsi2sd int → double

cvtsi2ssq quad int → single

cvtsi2sdq quad int → double

cvttss2si single → int (truncation)

cvttsd2si double → int (truncation)

cvttss2siq single → quad int (truncation)

cvttss2siq double → quad int (truncation)

x86-64 FP Code Exampledouble funct(double a, float x, double b, int i){ return a*x - b/i;}

a %xmm0 doublex %xmm1 floatb %xmm2 doublei %edi int

funct:cvtss2sd %xmm1, %xmm1 # %xmm1 = (double) xmulsd %xmm0, %xmm1 # %xmm1 = a*xcvtsi2sd %edi, %xmm0 # %xmm0 = (double) idivsd %xmm0, %xmm2 # %xmm2 = b/imovsd %xmm1, %xmm0 # %xmm0 = a*xsubsd %xmm2, %xmm0 # return a*x - b/i

ret

Constantsdouble cel2fahr(double temp){ return 1.8 * temp + 32.0;}

# Constant declarations.LC2: .long 3435973837 # Low order four bytes of 1.8 .long 1073532108 # High order four bytes of 1.8.LC4: .long 0 # Low order four bytes of 32.0 .long 1077936128 # High order four bytes of 32.0

# Codecel2fahr: mulsd .LC2(%rip), %xmm0 # Multiply by 1.8 addsd .LC4(%rip), %xmm0 # Add 32.0 ret

Here: Constants in decimal format

compiler decision hex more readable

Checking Constant Previous slide: Claim.LC4: .long 0 # Low order four bytes of 32.0 .long 1077936128 # High order four bytes of 32.0

Convert to hex format:.LC4: .long 0x0 # Low order four bytes of 32.0 .long 0x40400000 # High order four bytes of 32.0

Convert to double (blackboard?): Remember: e = 11 exponent bits, bias = 2e-1-1 = 1023

Comments SSE3 floating point

Uses lower ½ (double) or ¼ (single) of vector Finally departure from awkward x87 Assembly very similar to integer code

x87 still supported Even mixing with SSE3 possible Not recommended

For highest floating point performance Vectorization a must (but not in this course) See next slide

Vector Instructions Starting with version 4.1.1, gcc can autovectorize to some

extent -O3 or –ftree-vectorize No speed-up guaranteed Very limited icc as of now much better Fish machines: gcc 3.4

For highest performance vectorize yourself using intrinsics Intrinsics = C interface to vector instructions Learn in 18-645

Future Intel AVX announced: 4-way double, 8-way single