Zseries Assembler S8172a

eserver zSeries (Freeway) Machines:
an Assembler Programmer's View

SHARE 96 (Feb. 2001), Session 8172
John R. Ehrman
ehrman@us.ibm.com or ehrman@vnet.ibm.com
IBM Silicon Valley (nee Santa Teresa) Laboratory

555 Bailey Avenue
San Jose, CA 95141
IBM Corporation 2001
February 28, 2001
Table of Contents
Contents-1
Presentation Topics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Freeway Architecture . . . . . . . . . . . . . . . . . . . . . . . . . .
Freeway Architecture Overview . . . . . . . . . . . . . . . . .
Architecture Overview: z/Arch Program Status Word . . .
Architecture Overview: 64-Bit Addressing . . . . . . . . . . .
Architecture Overview: 64-Bit Virtual Address Translation
Architecture Overview: Registers . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
zSeries Instruction Set . . . . . . . . . . . . . . . . . . . . .

z/Architecture Instruction Set . . . . . . . . . . . . . . .
z/Architecture Instruction Set Overview . . . . . . . .
New Instructions (and Old-Instruction Analogs) . .
New Instructions: Familiar RR-Type Operations . .
New Instructions: Familiar RX-Type Operations . .
New Instructions: Familiar RS-Type Operations . .
New Instructions: Relative-Immediate . . . . . . . . .
New Instructions: Logical Loads . . . . . . . . . . . . .
New Instructions: Arithmetic Operations . . . . . . .
New Instructions: Logical-Immediate Operations .
New Instructions: Byte Reversal Operations . . . . .
New Instructions: Float/Integer Conversions . . . . .
New Instructions: Others . . . . . . . . . . . . . . . . . .
New Instructions: Both ESA/390 and z/Arch Modes
Quadword Alignment . . . . . . . . . . . . . . . . . . . . .
High-Order Half of a 64-Bit Register . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. 8
. 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Freeway Programming Overview
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
..
..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
2
3
4
5
6
7
Table of Contents
Modal Instructions and Addressing Modes . . . . . . . . .
Modal Instructions that Depend on Addressing Mode
Trimodal Addressing and Mode Switching . . . . . . . .
Mode-Switching Branch Instructions . . . . . . . . . . . .
Entry Points and Entry Addresses . . . . . . . . . . . . . .
Contents-2
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
25
26
27
29
30
zSeries Support Features in High Level Assembler Release 4 . . . . . . . . . . . . . . . . . . 31

New Data Types and Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
New Options and Statements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Presentation Topics
IBM eserver zSeries z/Architecture enhancements
Dual-mode architecture
64-bit addressing, registers, and operands
New instructions (many!) and instruction types (many!)

Compatibility with and extensions to ESA/390
Addressing modes
Modal instructions
eserver zSeries support features in HLASM R4
New instructions, constants, statements, and options
The IBM eserver zSeries brand consists of the established IBM e-business logo with the descriptive term server
following it. z/Architecture, zSeries, z/OS, and z/VM are trademarks of IBM.
Freeway Architecture
Architectures: ESA/390 vs. z/Arch
PSW
Addressing
Registers
Freeway Architecture Overview
z/Architecture: superset of ESA/390
64-bit General Purpose Registers, Control Registers

64-bit addressing for real and virtual addresses
All CPUs run uniformly in ESA/390 mode or z/Arch mode
System starts initially in ESA/390 mode, can switch to z/Arch
Low-address (Prefix) area completely reorganized, expanded to 8K bytes

Old/New 16-byte PSWs have new addresses (e.g. SVC Old PSW at X'140')
X'01' bit at 163 (X'A3') set if in z/Arch mode (on z/OS and z/VM)
64-bit I/O allows data addresses above 2GB bar
Distinguish architecture and addressing modes:
ESA/390 architecture mode: 24/31-bit addressing modes,

32-bit address generation, 8-byte PSW
z/Arch architecture mode: 24/31/64-bit addressing modes,
64-bit address generation, 16-byte PSW
Don't expect old programs to run succesfully in 64-bit addressing mode
They might, if bits 0-32 of all GRs are zero!

Old ops (and new ESA/390 ops) never change bits 0-31 of a GR (see slide 22)
Architecture Overview: z/Arch Program Status Word
PSW always 16 bytes long (128 bits)

I 32 bits H I 32 bits H

EB
(must be zeros)
usual stuff
31 33
63
I 64 bits H

Instruction Address (IA)

64
127
LPSWE loads a 16-byte PSW

LPSW loads an 8-byte PSW, extends it to the new 16-byte format
Bit 127 must be zero
PSW bits (Extended,Basic) indicate addressing mode:

EB Meaning
00
24-bit (Basic) addressing mode
01
31-bit (Basic) addressing mode
11
64-bit (Extended) addressing mode
Architecture Overview: 64-Bit Addressing
All 64 bits participate in the address
264 bytes = 18 billion billion = 16 Exabytes per address space
That is, 8 billion 2GB spaces' worth!
With data spaces, z/Arch is capable of 75-bit addressing
A total of 3.7 1022 bytes
There's really no way to picture this...
:
:
:
:
16 billion
GB
:
:
:
:
////////////
2 GB
16 MB
64 bits
32 bits
31 bits
24 bits
Architecture Overview: 64-Bit Virtual Address Translation
31-bit addressing uses the familiar scheme for virtual addresses:

1
11
8
12
/ segment page
byte
I 31 bit address H
In z/Arch 64-bit virtual addressing:

11
11
11
11
8
12
region 1 region 2 region 3 segment page

byte
I 64 bit address H
Number of translations depends on size of address space
Segment and page tables always present, always used
Region tables present and used only if necessary
Architecture Overview: Registers
General registers are 32 or 64 bits long, depending on the architecture
In z/Arch mode, GPRs are always 64 bits long
Old ESA/390 instructions use only the right halves (left halves are invisible)
You can use 64-bit registers without being in 64-bit addressing mode!
I 64bit register H z/Arch mode
I 32bit register H ESA/390 mode
GR0
GR1
:
:
GR15
0
31 32
63
64-bit registers in z/Arch mode numbered 0-63
Right half (old 32-bit part) now numbered 32-63, or {0-31} for ESA/390 mode
Access Registers still 32 bits; Control Registers now 64 bits

zSeries Instruction Set
z/Architecture Instruction Set
Many new instructions
Some work in ESA/390 mode, some only in z/Arch mode

I z/Arch Instruction Set H
All instructions in
22 Instr'ns
141 Instructions
ESA/390 mode can be

usable in
usable only in
used in z/Arch mode

both modes
z/Arch mode
I ESA/390 Instruction Set H

I 163 New zSeries Instructions H
Many old instructions don't affect bits 0-31 of the 64-bit register
Instructions usable in both addressing modes:
Non-modal: no dependence on addressing mode (e.g. 64-bit arithmetic)

Modal: behavior depends on addressing mode (e.g. BAL) (more at slide 26)
Some new instructions mix 64-bit and existing data types

No instruction requires being in 64-bit addressing mode
No vector facility or vector instructions
z/Architecture Instruction Set Overview
10
New instructions and their architecture modes:

Type
Number
z/Arch-Only
ESA/390 and z/Arch
General
140
119
21
Floating Point
12
12
Control
11
10
Total
163
141
22
Instructions for both architecture modes summarized on slide 22

New floating-point instructions support 64-bit integer operands
Details at slide 20
New control instructions (more at slide 21):
One z/Arch-only instruction is non-privileged (EREGG)
One ESA/390-or-z/Arch instruction is privileged (STFL)
New Instructions (and Old-Instruction Analogs)
11
Instructions for 64-bit integers similar to existing ops

Mnemonics for 64-bit operations end in (or contain) G
AG
AGR
2,XYZ
2,7
Adds 64bit operand at XYZ to 64bit register 2

Adds 64bit operand in R7 to 64bit register 2
Mnemonics for ops that mix 32- and 64-bit operands to give a 64-bit
result have GF or GH (and in a few cases, GT or GC)
F = Fullword, H = Halfword, T = Thirty-one bits, C = Character
32-bit operand sign-extended or not, per arithmetic or logical operation
AGF 2,PQR
AGFR 2,7
AGLFR 2,7
Adds signed 32bit operand at PQR to 64bit operand in R2

Adds signed 32bit operand in R7 to 64bit operand in R2
Adds logical 32bit operand in R7 to 64bit operand in R2
New Instructions: Familiar RR-Type Operations
12
RR Load, Logical, Add, Subtract, Compare (and Branch on Count)
ESA/390 Ops
New 64-bit Ops
New Mixed-Length Ops
LR, LTR
LGR, LTGR
LGFR, LTGFR
LPR, LNR, LCR
LPGR, LNGR, LCGR
LPGFR, LNGFR, LCGFR
AR, SR
AGR, SGR
AGFR, SGFR
ALR, SLR
ALGR, SLGR
ALGFR, SLGFR
CR, CLR
CGR, CLGR
CGFR, CLGFR
NR, OR, XR
NGR, OGR, XGR
BCTR
BCTGR
LCGFR, LPGFR complement 231 without overflow
New Instructions: Familiar RX-Type Operations
13
RX Load, Logical, Add, Subtract, Compare (and a few others)

ESA/390 Ops
New 64-bit Ops
New Mixed-Length Ops
L, LH, ST
LG, STG
LGF, LGH
A, S, C
AG, SG, CG
AGF, SGF, CGF
AL, SL, CL
ALG, SLG, CLG
ALGF, SLGF, CLGF
N, O, X
NG, OG, XG
MS, MSR
MSG, MSGR
BCT
BCTG
CVD, CVB
CVDG, CVBG
MSGF, MSGFR
One new instruction for mixed 16/64 operations: LGH
No mixed 16/64 analogs of other halfword ops (AH, CH, MH, SH)
STH still stores rightmost 16 bits (32- or 64-bit register)
MSG, MSGF products are 64 bits; MSGF multiplier is 32 bits

New Instructions: Familiar RS-Type Operations
14
RS Load, Shift, etc.

ESA/390 Ops
New 64-bit Ops
Comments
BXH, BXLE
BXHG, BXLEG
SRL, SLL, SRA,

SLA
SRLG, SLLG, SRAG,

SLAG, RLL*, RLLG
Source and result registers

may be different!
LM, STM
LMG, STMG
LMD, LMH, STMH
Load/store register halves

at different locations
CLM, ICM, STCM
CLMH, ICMH, STCMH
Use high-order 4 bytes
Compare and Swap

ESA/390 Ops
New 64-bit Ops
Comments
CS, CDS
CSG, CDSG
CDSG operands must be

quad-aligned (see slide 23)
Instruction also available in ESA/390 mode (see slide 22)
New Instructions: Relative-Immediate
15
ESA/390 Ops
z/Arch Ops
Comments
TMH, TML
TMHH, TMHL
TMLH=TMH, TMLL=TML for symmetry
BRXH, BRXLE BRXHG, BRXLG
Relative indexed branch
BRCT
BRCTG
Relative branch on count
LHI, AHI,
MHI, CHI
LGHI, AGHI,
MGHI, CGHI
16-bit immediate operand,

64-bit result
BRC
BRCL*
Long relative conditional branch
BRAS
BRASL*
Long relative branch and save
LARL*
Long relative load-address
Most R-I instructions use a 16-bit (halfword) immediate operand

BRCL, BRASL, LARL use a 32-bit (fullword) offset (in halfwords)
Long signed immediate field allows offsets of 4 GB

Target operand must be at an even address
LARL is like LA, but doesn't require a base register
* Instruction also available in ESA/390 mode (see slide 22)

New Instructions: Logical Loads
16
Logical Load Instructions
z/Arch Ops
Comments
LLGF, LLGFR
Load 32 bits, clear high-order 32
LLGT, LLGTR
LLGH
LLGC
LLGT, LLGTR: lets you set a 64-bit address from a 31-bit address, without
having to worry about the high-order bit
LLGC: like Insert Character, but no need to clear the target register
New Instructions: Arithmetic Operations
17
z/Arch Ops
Comments
DSG, DSGR
Divide single (all operands 64 bits)
DSGF, DSGFR
Divide 64 by 32, 64-bit results
DL*, DLR*
Divide logical (64 by 32; unsigned operands)
DLG, DLGR
Divide logical (128 by 64; unsigned operands)
ML*, MLR*
64-bit logical product of unsigned 32-bit operands
MLG, MLGR
128-bit logical product of unsigned 64-bit operands
ALC*, ALCR*
Add logical with carry (32-bit operands)
ALCG, ALCGR
Add logical with carry (64-bit operands)
SLB*, SLBR*
Subtract logical with borrow (32-bit operands)
SLBG, SLBGR
Subtract logical with borrow (64-bit operands)
ML, MLR product in low halves of two registers

High-order bit of CC is the carry-borrow indicator
*
New Instructions: Logical-Immediate Operations
18
These have no analogs in ESA/90

All instructions have a 16-bit immediate field
z/Arch Ops
Comments
IIHH, IIHL, IILH, IILL
Inserts 16 bits into part of a 64-bit GR
NIHH, NIHL, NILH, NILL
AND-Immediate with part of a 64-bit GR
OIHH, OIHL, OILH, OILL
OR-Immediate with part of a 64-bit GR
LLIHH, LLIHL, LLILH, LLILL
Load part of a 64-bit GR, remaining bits set

to zero
Last two letters of mnemonic indicate which part of the GR:

HH
High Half's High Half (bits 0-15)
HL
High Half's Low Half (bits 16-31)
LH
Low Half's High Half (bits 32-47)
LL
Low Half's Low Half (bits 48-63)
New Instructions: Byte Reversal Operations
19
Byte-Reversing Load/Store
z/Arch Ops
Comments
LRVH*, STRVH*
Transmit 2 rightmost bytes in reverse order
LRV*, LRVR*, STRV*
LRVG, LRVGR, STRVG
Rightmost bytes of register used for 2-byte and 4-byte ops

For Load operations, remaining bytes in register are unchanged
z/Arch is now an Either-Endian architecture!
New Instructions: Float/Integer Conversions
20
Conversions between 64-bit integers and Hex or IEEE-Binary float

Extensions of the G5/G6 floating point enhancements
ESA/390 supports 32-bit integer operations
Operation
32-bit Integer Ops
New 64-bit Integer Ops
Integer to Hex
CEFR, CDFR, CXFR
CEGR, CDGR, CXGR
Integer to IEEE
CEFBR, CDFBR, CXFBR
CEGBR, CDGBR, CXGBR
Hex to Integer
CFER, CFDR, CFXR
CGER, CGDR, CGXR
IEEE to Integer
CFEBR, CFDBR, CFXBR
CGEBR, CGDBR, CGXBR
Same rounding rules apply to 64-bit integers as to 32-bit
New Instructions: Others
21
z/Arch Ops
Comments
LPQ, STPQ
Load/Store of quad-aligned operands
TAM*
Test addressing mode
SAM24*, SAM31*, SAM64
Set addressing mode (without branching)
EPSW
Extract high-order 64 bits of 128-bit PSW
Control instructions:
Privileged, z/Arch or ESA/390:

STFL (described at slide 22)
Privileged, z/Arch-only:
ESEA, LCTLG, LPSWE, LRAG, LURAG, STCTG, STRAG, STURG, TRACG
Nonprivileged, z/Arch-only:
EREGG (extract 64-bit registers from linkage stack)
New Instructions: Both ESA/390 and z/Arch Modes
22
BRCL, BRASL, LARL
32-bit signed Immediate field contains offset in halfwords
TAM, SAM24, SAM31: Test/set addressing mode

LRVR, LRV, LRVH, STRH, STRVH: Byte-reversing load/store
RLL: Rotate left logical (with different source/result registers)
ML, MLR; DL, DLR: Logical multiply and divide
ALC, ALCR; SLB, SLBR
Add Logical with Carry, Subtract Logical with Borrow

Useful for multi-word or multi-precision arithmetic
EPSW: extracts high-order half of PSW in two 32-bit pieces

STFL: Store Facility List (privileged)
Stores bits in fullword at 200 (X'C8') indicating available hardware features
Bit 0: indicates availability of these ops on ESA/390
Bit 1: indicates z/Architecture mode is installed
Bit 2: indicates z/Architecture mode is active (compare bit at X'A3')
Quadword Alignment
23
Need quadword alignment for LPQ, STPQ, CDSG, some PLO operands
LPQ, STPQ, CDSG also provide quadword consistency
Many operating system components don't (yet) support quad alignment

Suggestion for doubleword-aligned storage allocators:
1. Define a DSECT mapping your quad-aligned area, with doubleword padding
QWORK
Quad1
Quad2
DSect
DS
DS
QPad
DS
LQWORK Equ
2D
2D
D
*QWORK
First quadword data area

Second quadword data area
... and so forth, as required
**** Doubleword padding! Never use this field!
Length of DSect
2. Allocate doubleword-aligned storage for the full size (LQWORK) of QWORK,

set its address in (say) R2 (and save the address for deallocation!)
3. Round the pointer to the next quadword boundary
AHI R2,X'0008'
NILL R2,X'FFF0'
Round up if not already quad aligned

Guarantee quadword alignment
USING QWORK,R2
Use DSECT for quadaligned work area
High-Order Half of a 64-Bit Register
24
Some instructions affect part or all of the high-order word
AND, OR, INSERT IMMEDIATE: NIHH, NIHL, OIHH, OIHL, IIHH, IIHL
LOAD LOGICAL IMMEDIATE: LLIHH, LLIHL
INSERT CHARACTERS UNDER MASK: ICMH
LOAD MULTIPLE HIGH: LMH
Some instructions affect both parts of the register:
LOAD MULTIPLE DISJOINT: LMD

Loads high-order and low-order halves from different addresses
LOAD PAIR FROM QUADWORD: LPQ

Requires quadword alignment of operand (see slide 23)
For some instructions, use of high-order half depends on addressing

mode (see slide 26)
25
Modal Instructions and

Addressing Modes
Modal Instructions that Depend on Addressing Mode
26
Most update and/or depend on the values in registers
High-order 32 bits affected only in 64-bit mode
Branch instructions:
BAL, BALR, BAS, BASR, BRAS, BRASL, BSM, BASSM
Comparisons and Move instructions:
CLCL, CLCLE; MVCL, MVCLE
Load Address instructions:
LA, LAE, LARL
Translate instructions:
TRE, TRT
String operations:
CLST, MVST, SRST, CUSE
And a few others:
CKSM, CUUTF, CUTFU, UPT, CFC
Trimodal Addressing and Mode Switching
27
Branch address generation in z/Arch always gives a 64-bit value
Leftmost 40 bits forced to zero in 24-bit addressing mode
Leftmost 33 bits forced to zero in 31-bit addressing mode
Extended addressing mode (see slide 4)
Bit 31 of PSW (E bit):

Set by LPSW, LPSWE, BASSM, BSM, Stacking PC, PR, RP, interruptions
Bit 32 of PSW (B bit):

Set by BASSM, BSM, Stacking PC if they set the E bit
Some ESA/390 call/return combinations aren't supported in z/Arch:
BASSM/BR, BALR/BSM, BASR/BSM
Linkages are either mode-switching or not
Trimodal Addressing and Mode Switching ...
28
Switching among 24-, 31-, and 64-bit addressing modes

Branch and Set Mode instructions: BASSM, BSM
Set 64-bit mode if low order bit of target address is one!

Bit 63 is left set (so the mode switch can be detected),
but is ignored in branch-address generation (target address always even)
Set 24- or 31-bit mode if low-order bit is zero

Choice between 24 or 31 depends on bit 32 being zero or one (as before)
Set Addressing Mode instructions: SAM24*, SAM31*, SAM64
Set PSW's EB bits (see slide 4) without branching
They also check Instruction-Address bits to ensure next instruction isn't

outside the just-specified addressing range!
Test Addressing Mode: TAM* sets CC per PSW's EB bits (see slide 4)
*
Mode-Switching Branch Instructions
29
I 64bit GPR H
0
3132
63
In all modes, BASSM and BSM change addressing mode based on bits
63 and 32 of GPRR2
Bit
63
32
0
0
1
0
1
x
AMODE changed to
24
31
64
In 24/31-bit mode, bits 0-31 of GPRR1 are left unchanged

In 64-bit mode:
BASSM sets a 64-bit return address in GPRR1 with bit 63 = 1

BSM sets bit 63 of GPRR1 to 1 (if R1 0)
BAL/BALR/BAS/BASR do not set bit 63 of GPRR1 to 1
There are many possible to/from combinations!

Entry Points and Entry Addresses
30
BAL useful only in 24-bit mode

BAS/BASR, BRAS/BRASL call without mode change
Entry in 64-bit mode
Bit 63 = 0: entry was from BAS/BASR, BRAS/BRASL

No mode change; must return in same mode
Bit 63 = 1: entry was from BASSM/BSM

Mode has changed: caller's mode is in linkage register; must return via BSM
Entry points reachable in 64-bit mode from an unknown mode:
Callable by BAS/BASR/BRAS/BRASL or BASSM/BSM
Set bit 63 of entry-point register to zero if it's to be used as a base register
NILL EntryReg,X'FFFE'
USING EntryPoint,EntryReg
To test bit 63 of entry address:
TMLL EntryReg,1
Follow with a relative (not based!) conditional branch
31
zSeries Support Features in

High Level Assembler Release 4
HLASM web site: http://www.ibm.com/software/ad/hlasm/
New Data Types and Constants
32
Binary constants
FD
8-byte doubleword-aligned binary integer

Same as currently-supported FL8, but aligned
000000 000000000000005F
DC
FD'95'
Nominal value in the range ( 263, + 263 1)

Address constants
AD
8-byte doubleword-aligned address constant
Note: may not necessarily be supported in other products
Expressions evaluated to 32 bits, sign-extended to 64

000008 FFFFFFFFFFFFFFA1
DC
AD(95)
Non-relocatable part must be in range ( 231, 231 1)
New Options and Statements
33
OPTABLE(ZOP) specifically selects z/Arch set

New z/Arch machine instructions also in OPTABLE(UNI)
OPTABLE(ESA) option may help avoid macro conflicts with new mnemonics
(Or, use macros with names longer than 5 characters)
Vector instructions still available in the assembler (OPTABLE(UNI))
But the Freeway architecture can't execute them...
New sub-option: OPTABLE(xxx,LIST)
Displays all mnemonics supported by opcode table xxx
LIST sub-option and ZOP opcode table enabled by APAR PQ44437
AMODE and RMODE operand extensions

AMODE
ANY31, 64
RMODE
31, 64
64 operands may not be supported by other products
New mode values indicated in ESD listing's Flag byte
34
Summary
Summary
35
New architecture
Vastly increased addressing capabilities
Upward compatibility
Many new instructions
64-bit arithmetic without 64-bit addressing
New operations for ESA/390 also
Smooth integration with existing architecture
Many operations work in both architecture modes
A substantial base for future growth

Evolution, not revolution!
References
36
z/Architecture Principles of Operation (SA22-7832)

High Level Assembler for MVS & VM & VSE publications:
Language Reference (SC26-4940)
Programmer's Guide (SC26-4941)

FWYPROG Rev. 08 Jan 01 1240

Fmt. 08 Jan 01, 1252

Zseries Assembler S8172a

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Zseries Assembler S8172a

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Zseries Assembler S8172a

Uploaded by

Copyright:

Available Formats

eserver zSeries (Freeway) Machines:

an Assembler Programmer's View

IBM Silicon Valley (nee Santa Teresa) Laboratory

February 28, 2001

zSeries Instruction Set . . . . . . . . . . . . . . . . . . . . .

Freeway Programming Overview

IBM Corporation 2001

zSeries Support Features in High Level Assembler Release 4 . . . . . . . . . . . . . . . . . . 31

Freeway Programming Overview

IBM Corporation 2001

IBM eserver zSeries z/Architecture enhancements

64-bit addressing, registers, and operands

New instructions (many!) and instruction types (many!)

eserver zSeries support features in HLASM R4

New instructions, constants, statements, and options

Freeway Programming Overview

IBM Corporation 2001

Architectures: ESA/390 vs. z/Arch

Freeway Programming Overview

IBM Corporation 2001

Freeway Architecture Overview

z/Architecture: superset of ESA/390

64-bit General Purpose Registers, Control Registers

Low-address (Prefix) area completely reorganized, expanded to 8K bytes

Distinguish architecture and addressing modes:

ESA/390 architecture mode: 24/31-bit addressing modes,

Don't expect old programs to run succesfully in 64-bit addressing mode

They might, if bits 0-32 of all GRs are zero!

Freeway Programming Overview

IBM Corporation 2001

Architecture Overview: z/Arch Program Status Word

PSW always 16 bytes long (128 bits)

Instruction Address (IA)

LPSWE loads a 16-byte PSW

PSW bits (Extended,Basic) indicate addressing mode:

IBM Corporation 2001

Architecture Overview: 64-Bit Addressing

All 64 bits participate in the address

264 bytes = 18 billion billion = 16 Exabytes per address space

That is, 8 billion 2GB spaces' worth!

With data spaces, z/Arch is capable of 75-bit addressing

A total of 3.7 1022 bytes

There's really no way to picture this...

Freeway Programming Overview

IBM Corporation 2001

Architecture Overview: 64-Bit Virtual Address Translation

31-bit addressing uses the familiar scheme for virtual addresses:

In z/Arch 64-bit virtual addressing:

region 1 region 2 region 3 segment page

Number of translations depends on size of address space

Segment and page tables always present, always used

Region tables present and used only if necessary

Freeway Programming Overview

IBM Corporation 2001

Architecture Overview: Registers

General registers are 32 or 64 bits long, depending on the architecture

In z/Arch mode, GPRs are always 64 bits long