CS 476 Programming Language Design William Mansky LowLevel

CS 476 – Programming Language Design William Mansky

Low-Level Languages High-Level Language C, Java, OCaml, F#, Python, … interpreter output 1

Low-Level Languages High-Level Language C, Java, OCaml, F#, Python, … compiler Low-Level Language interpreter output 2

Low-Level Languages Java compiler Java Virtual Machine Bytecode JVM output 3

Low-Level Languages C compiler Assembly processor output 4

Low-Level Languages C compiler Compiler IR compiler Assembly/Machine Code processor output 5

Low-Level Languages • An intermediate step between high-level languages (C, Java, OCaml, etc. ) and machine code ― Easier than going all the way to machine code in one step ― Easier to optimize, because they expose hidden costs • Usually C-like, but with more machine-level details: ― May have a limited number of variables available ― Only perform one step of computation in each command ― Jump instructions instead of complex control flow • Examples: assembly (x 86, MIPS, ARM), JVM bytecode, LLVM IR 6

Simple Assembly Language E : : = <ident> | <#> C : : = <ident> = add E, E | <ident> = lt E, E | … 7

Simple Assembly: Memory E : : = <ident> | <#> C : : = … | <ident> = load E | store E, E 8

Simple Assembly: Control Flow r = 1 loop: c = le x, 0 cbr c end r = mul r, x x = sub x, 1 br loop end: 9

Simple Assembly: Control Flow loop: c = le x, 0 cbr c end r = mul r, x x = sub x, 1 br loop end: 10

Simple Assembly: Control Flow c = le x, 0 cbr c end r = mul r, x x = sub x, 1 br loop end: 11

Simple Assembly: Control Flow cbr c end r = mul r, x x = sub x, 1 br loop end: 12

Simple Assembly: Control Flow r = mul r, x x = sub x, 1 br loop end: 13

Simple Assembly: Control Flow br loop end: 14

Simple Assembly: Control Flow r = 1 loop: c = le x, 0 cbr c end r = mul r, x x = sub x, 1 br loop end: 15

Simple Assembly: Control Flow r = 1 loop: c = le x, 0 cbr c end r = mul r, x x = sub x, 1 br loop end: 16

Simple Assembly: Control Flow r = 1 loop: c = le x, 0 cbr c end r = mul r, x x = sub x, 1 br loop end: 17

Simple Assembly: Control Flow r = 1 loop: c = le x, 0 cbr c end r = mul r, x x = sub x, 1 br loop end: 18

Unstructured Control Flow • 19

Unstructured Control Flow • 20

Unstructured Control Flow • 21

Simple Assembly: Control Flow r = 1 loop: c = le x, 0 cbr c, end r = mul r, x x = sub x, 1 br loop end: 22

Simple Assembly: Control Flow 0: r = 1 loop: 1: c = le x, 0 2: cbr c, end 3: r = mul r, x 4: x = sub x, 1 5: br loop end: 6: loop 1 end 6 23

Simple Assembly: Control Flow E : : = <ident> | <#> C : : = …| br <ident> 24

Simple Assembly: Control Flow E : : = <ident> | <#> C : : = …| br <ident> | cbr E, <ident> 25

Simple Assembly: Labels as Values store label 0, p q = add p, 1 store label 1, q q = add q, 1 store label 2, q target = add p, x indirectbr target label 0: … label 1: … label 2: … p label 0 26

Simple Assembly: Labels as Values store label 0, p q = add p, 1 store label 1, q q = add q, 1 store label 2, q target = add p, x indirectbr target label 0: … label 1: … label 2: … p label 0 p + 1 label 1 27

Simple Assembly: Labels as Values store label 0, p q = add p, 1 store label 1, q q = add q, 1 store label 2, q target = add p, x indirectbr target label 0: … label 1: … label 2: … p label 0 p + 1 label 1 p + 2 label 2 28

Simple Assembly: Labels as Values store label 0, p q = add p, 1 store label 1, q q = add q, 1 store label 2, q target = add p, x indirectbr target label 0: … label 1: … label 2: … p label 0 p + 1 label 1 p + 2 label 2 29

Simple Assembly: Control Flow E : : = <ident> | <#> C : : = …| br <ident> | cbr E, <ident> | indirectbr E 30

Low-Level Language Semantics • Expose various issues that are hidden by high-level languages • Unstructured control flow is one of the biggest (also in C with goto) • Other issues: limited variables, memory management, function calling conventions, concurrent memory models, … • With semantics for high- and low-level languages, we can prove that compiled code does the same thing as the source! 31