Report

ALU Architecture and ISA Extensions Lecture notes from MKP, H. H. Lee and S. Yalamanchili Reading • Sections 3.2-3.5 (only those elements covered in class) • Sections 3.6-3.8 • Appendix B.5 • Goal: Understand the ISA view of the core microarchitecture Organization of functional units and register files into basic data paths (2) Overview • Instruction Set Architectures have a purpose Applications dictate what we need • We only have a fixed number of bits Impact on accuracy • More is not better We cannot afford everything we want • Basic Arithmetic Logic Unit (ALU) Design Addition/subtraction, multiplication, division (3) Reminder: ISA byte addressed memory Register File (Programmer Visible State) Memory Interface stack 0x00 0x01 0x02 0x03 Processor Internal Buses 0x1F Dynamic Data Data segment (static) Text Segment Programmer Invisible State Program Counter Instruction register Kernel registers Reserved 0xFFFFFFFF Arithmetic Logic Unit (ALU) Memory Map Who sees what? (4) Arithmetic for Computers • Operations on integers Addition and subtraction Multiplication and division Dealing with overflow • Operation on floating-point real numbers Representation and operations • Let us first look at integers (5) Integer Addition(3.2) • Example: 7 + 6 Overflow if result out of range Adding +ve and –ve operands, no overflow Adding two +ve operands Overflow if result sign is 1 Adding two –ve operands Overflow if result sign is 0 (6) Integer Subtraction • Add negation of second operand • Example: 7 – 6 = 7 + (–6) +7: –6: +1: 0000 0000 … 0000 0111 1111 1111 … 1111 1010 0000 0000 … 0000 0001 2’s complement representation • Overflow if result out of range Subtracting two +ve or two –ve operands, no overflow Subtracting +ve from –ve operand o Overflow if result sign is 0 Subtracting –ve from +ve operand o Overflow if result sign is 1 (7) ISA Impact • Some languages (e.g., C) ignore overflow Use MIPS addu, addui, subu instructions • Other languages (e.g., Ada, Fortran) require raising an exception Use MIPS add, addi, sub instructions On overflow, invoke exception handler o o o Save PC in exception program counter (EPC) register Jump to predefined handler address mfc0 (move from coprocessor register) instruction can retrieve EPC value, to return after corrective action (more later) • ALU Design leads to many solutions. We look at one simple example (8) Integer ALU (arithmetic logic unit)(B.5) • Build a 1 bit ALU, and use 32 of them (bit-slice) operation a op a b res result b (9) Single Bit ALU Implements only AND and OR operations Operation 0 A Result 1 B (10) Adding Functionality • We can add additional operators (to a point) • How about addition? cout = ab + acin + bcin sum = a b cin CarryIn a Sum b CarryOut • Review full adders from digital design (11) Building a 32-bit ALU CarryIn a0 b0 Operation Operation CarryIn ALU0 Result0 CarryOut CarryIn a1 a 0 b1 CarryIn ALU1 Result1 CarryOut 1 Result a2 2 b b2 CarryIn ALU2 Result2 CarryOut CarryOut a31 b31 CarryIn ALU31 Result31 (12) Subtraction (a – b) ? • Two's complement approach: just negate b and add 1. • How do we negate? sub Binvert CarryIn a0 CarryIn ALU0 b0 Operation Result0 CarryOut • A clever solution: Binvert a1 ALU1 b1 Operation a a2 0 1 0 Result1 CarryOut CarryIn b CarryIn CarryIn ALU2 b2 Result2 CarryOut Result 2 1 a31 CarryOut b31 CarryIn ALU31 Result31 (13) Tailoring the ALU to the MIPS • Need to support the set-on-less-than instruction(slt) remember: slt is an arithmetic instruction produces a 1 if rs < rt and 0 otherwise use subtraction: (a-b) < 0 implies a < b • Need to support test for equality (beq $t5, $t6, $t7) use subtraction: (a-b) = 0 implies a = b (14) What Result31 is when (a-b)<0? Binvert CarryIn a0 b0 CarryIn ALU0 Less CarryOut a1 b1 0 CarryIn ALU1 Less CarryOut Operation Result0 Binvert Operation CarryIn a 0 Result1 1 Result a2 b2 0 b CarryIn ALU2 Less CarryOut Result2 CarryIn ALU31 Less 2 1 Less 3 CarryOut CarryIn a31 b31 0 0 Result31 Set Overflow Unsigned vs. signed support (15) Test for equality Bnegate Operation • Notice control lines: 000 001 010 110 111 = = = = = and or add subtract slt •Note: zero is a 1 when the result is zero! a0 b0 CarryIn ALU0 Less CarryOut Result0 a1 b1 0 CarryIn ALU1 Less CarryOut Result1 a2 b2 0 CarryIn ALU2 Less CarryOut Result2 Zero Note test for overflow! a31 b31 0 CarryIn ALU31 Less Result31 Set Overflow (16) ISA View CPU/Core $0 $1 $31 ALU • Register-to-Register data path • We want this to be as fast as possible (17) Multiplication (3.3) • Long multiplication multiplicand multiplier product 1000 × 1001 1000 0000 0000 1000 1001000 Length of product is the sum of operand lengths (18) A Multiplier • Uses multiple adders Cost/performance tradeoff Can be pipelined Several multiplication performed in parallel (19) MIPS Multiplication • Two 32-bit registers for product HI: most-significant 32 bits LO: least-significant 32-bits • Instructions mult rs, rt / multu rs, rt o 64-bit product in HI/LO mfhi rd / mflo rd o o Move from HI/LO to rd Can test HI value to see if product overflows 32 bits mul rd, rs, rt o Least-significant 32 bits of product – > rd Study Exercise: Check out signed and unsigned multiplication with QtSPIM (20) Division(3.4) quotient dividend 1001 1000 1001010 -1000 divisor 10 101 1010 -1000 10 remainder • Check for 0 divisor • Long division approach If divisor ≤ dividend bits o 1 bit in quotient, subtract Otherwise o 0 bit in quotient, bring down next dividend bit • Restoring division n-bit operands yield n-bit • quotient and remainder Do the subtract, and if remainder goes < 0, add divisor back Signed division Divide using absolute values Adjust sign of quotient and remainder as required (21) Faster Division • Can’t use parallel hardware as in multiplier Subtraction is conditional on sign of remainder • Faster dividers (e.g. SRT division) generate multiple quotient bits per step Still require multiple steps • Customized implementations for high performance, e.g., supercomputers (22) MIPS Division • Use HI/LO registers for result HI: 32-bit remainder LO: 32-bit quotient • Instructions div rs, rt / divu rs, rt No overflow or divide-by-0 checking o Software must perform checks if required Use mfhi, mflo to access result Study Exercise: Check out signed and unsigned division with QtSPIM (23) ISA View CPU/Core $0 $1 $31 Multiply Divide ALU Hi Lo • Additional function units and registers (Hi/Lo) • Additional instructions to move data to/from these registers mfhi, mflo • What other instructions would you add? Cost? (24) Floating Point(3.5) • Representation for non-integral numbers Including very small and very large numbers • Like scientific notation –2.34 × 1056 +0.002 × 10–4 +987.02 × 109 normalized not normalized • In binary ±1.xxxxxxx2 × 2yyyy • Types float and double in C (25) IEEE 754 Floating-point Representation Single Precision (32-bit) 31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 S exponent significand 1bit 23 bits 8 bits (–1)sign x (1+fraction) x 2exponent-127 Double Precision (64-bit) 63 62 61 60 59 58 57 56 55 54 53 52 51 50 49 48 47 46 45 44 43 42 41 40 39 38 37 36 35 34 33 32 S exponent significand 1bit 20 bits 11 bits significand (continued) 32 bits (–1)sign x (1+fraction) x 2exponent-1023 (26) Floating Point Standard • Defined by IEEE Std 754-1985 • Developed in response to divergence of representations Portability issues for scientific code • Now almost universally adopted • Two representations Single precision (32-bit) Double precision (64-bit) (27) FP Adder Hardware • Much more complex than integer adder • Doing it in one clock cycle would take too long Much longer than integer operations Slower clock would penalize all instructions • FP adder usually takes several cycles Can be pipelined Example: FP Addition (28) FP Adder Hardware Step 1 Step 2 Step 3 Step 4 (29) FP Arithmetic Hardware • FP multiplier is of similar complexity to FP adder But uses a multiplier for significands instead of an adder • FP arithmetic hardware usually does Addition, subtraction, multiplication, division, reciprocal, square-root FP integer conversion • Operations usually takes several cycles Can be pipelined (30) ISA Impact • FP hardware is coprocessor 1 Adjunct processor that extends the ISA • Separate FP registers 32 single-precision: $f0, $f1, … $f31 Paired for double-precision: $f0/$f1, $f2/$f3, … o Release 2 of MIPs ISA supports 32 × 64-bit FP reg’s • FP instructions operate only on FP registers Programs generally do not perform integer ops on FP data, or vice versa More registers with minimal code-size impact (31) ISA View: The Co-Processor Co-Processor 1 CPU/Core $0 $1 $0 $1 $31 $31 Multiply Divide ALU Hi FP ALU Lo Co-Processor 0 BadVaddr Status Causes EPC later • Floating point operations access a separate set of 32-bit registers Pairs of 32-bit registers are used for double precision (32) ISA View • Distinct instructions operate on the floating point registers (pg. A-73) Arithmetic instructions o add.d fd, fs, ft, and add.s fd, fs, ft double precision single precision • Data movement to/from floating point coprocessors mcf1 rt, fs and mtc1 rd, fs • Note that the ISA design implementation is extensible via co-processors • FP load and store instructions lwc1, ldc1, swc1, sdc1 o e.g., ldc1 $f8, 32($sp) Example: DP Mean (33) Associativity • Floating point arithmetic is not commutative • Parallel programs may interleave operations in unexpected orders Assumptions of associativity may fail (x+y)+z x+(y+z) -1.50E+38 x -1.50E+38 y 1.50E+38 0.00E+00 z 1.0 1.0 1.50E+38 1.00E+00 0.00E+00 Need to validate parallel programs under varying degrees of parallelism (34) Performance Issues • Latency of instructions Integer instructions can take a single cycle Floating point instructions can take multiple cycles Some (FP Divide) can take hundreds of cycles • What about energy (we will get to that shortly) • What other instructions would you like in hardware? Would some applications change your mind? • How do you decide whether to add new instructions? (35) Multimedia (3.6, 3.7, 3.8) • Lower dynamic range and precision requirements Do not need 32-bits! • Inherent parallelism in the operations (36) Vector Computation • Operate on multiple data elements (vectors) at a time • Flexible definition/use of registers • Registers hold integers, floats (SP), doubles DP) 128-bit Register 1x128 bit integer 2x64-bit double precision 4 x 32-bit single precision 8x16 short integers (37) Processing Vectors • When is this more efficient? Memory vector registers • When is this not efficient? • Think of 3D graphics, linear algebra and media processing (38) Case Study: Intel Streaming SIMD Extensions • 8, 128-bit XMM registers X86-64 adds 8 more registers XMM8-XMM15 • 8, 16, 32, 64 bit integers (SSE2) • 32-bit (SP) and 64-bit (DP) floating point • Signed/unsigned integer operations • IEEE 754 floating point support • Reading Assignment: http://en.wikipedia.org/wiki/Streaming_SIMD_Extensions http://neilkemp.us/src/sse_tutorial/sse_tutorial.html#I (39) Instruction Categories • Floating point instructions Arithmetic, movement Comparison, shuffling Type conversion, bit level register memory register • Integer • Other e.g., cache management • ISA extensions! • Advanced Vector Extensions (AVX) Successor to SSE (40) Arithmetic View • Graphics and media processing operates on vectors of 8-bit and 16-bit data Use 64-bit adder, with partitioned carry chain o Operate on 8×8-bit, 4×16-bit, or 2×32-bit vectors SIMD (single-instruction, multiple-data) • Saturating operations On overflow, result is largest representable value o c.f. 2s-complement modulo arithmetic E.g., clipping in audio, saturation in video 4x16-bit 2x32-bit (41) SSE Example // A 16byte = 128bit vector struct struct Vector4 { float x, y, z, w; }; More complex example (matrix multiply) in Section 3.8 – using AVX // Add two constant vectors and return the resulting vector Vector4 SSE_Add ( const Vector4 &Op_A, const Vector4 &Op_B ) { Vector4 Ret_Vector; __asm { MOV EAX Op_A MOV EBX, Op_B } // Load pointers into CPU regs MOVUPS XMM0, [EAX] MOVUPS XMM1, [EBX] // Move unaligned vectors to SSE regs ADDPS XMM0, XMM1 MOVUPS [Ret_Vector], XMM0 // Add vector elements // Save the return vector } return Ret_Vector; From http://neilkemp.us/src/sse_tutorial/sse_tutorial.html#I (42) Characterizing Parallelism Today serial computing cores (von Neumann model) Instruction Streams Data Streams SISD SIMD MISD MIMD Single instruction multiple data stream computing, e.g., SSE Today’s Multicore • Characterization due to M. Flynn* *M. Flynn, (September 1972). "Some Computer Organizations and Their Effectiveness". IEEE Transactions on Computers, C–21 (9): 948–960t (43) Parallelism Categories From http://en.wikipedia.org/wiki/Flynn%27s_taxonomy (44) Data Parallel vs. Traditional Vector Vector Architecture Vector Register A Vector Register C Vector Register B pipelined functional unit Data Parallel Architecture registers Process each square in parallel – data parallel computation (45) ISA View SIMD Registers CPU/Core $0 $1 XMM0 XMM1 $31 XMM15 Multiply Divide ALU Hi Vector ALU Lo • Separate core data path • Can be viewed as a co-processor with a distinct set of instructions (46) Domain Impact on the ISA: Example Scientific Computing • • • • Floats Double precision Massive data Power constrained Embedded Systems • • • • • Integers Lower precision Streaming data Security support Energy constrained (47) Summary • ISAs support operations required of application domains Note the differences between embedded and supercomputers! Signed, unsigned, FP, SIMD, etc. • Bounded precision effects Software must be careful how hardware used e.g., associativity Need standards to promote portability • Avoid “kitchen sink” designs There is no free lunch Impact on speed and energy we will get to this later (48) Study Guide • Perform 2’s complement addition and subtraction (review) • Add a few more instructions to the simple ALU Add an XOR instruction Add an instruction that returns the max of its inputs Make sure all control signals are accounted for • Convert real numbers to single precision floating point (review) and extract the value from an encoded single precision number (review) • Execute the SPIM programs (class website) that use floating point numbers. Study the memory/register contents via single step execution (49) Study Guide (cont.) • Write a few simple SPIM programs for Multiplication/division of signed and unsigned numbers o o Use numbers that produce >32-bit results Move to/from HI and LO registers ( find the instructions for doing so) Addition/subtraction of floating point numbers • Try to write a simple SPIM program that demonstrates that floating point operations are not associative (this takes some thought and review of the range of floating point numbers) • Look up additional SIMD instruction sets and compare AMD NEON, Altivec, AMD 3D Now (50) Glossary • Co-processor • Data parallelism • Data parallel computation vs. vector computation • Instruction set extensions • Overflow • MIMD • Precision • SIMD • Saturating arithmetic • Signed arithmetic support • Unsigned arithmetic support • Vector processing (51)