--- id: 04e03e046bb46ab34aaeeed79279923b author: Lee Yunjin title: Analysis of AMD64 Assembly Code Generation and Branch Instructions for Go Conditional Statements (if) description: This article compares the process by which the Go compiler translates if-else conditional statements into AMD64 machine branch instructions such as CMPQ, JGE, and JMP, with the branch code generation results of the C compiler (-O2). language: en date: 2026-05-17T12:00:00Z path: /blog/posts/go-if-statement-analysis-3012d30c category: lowlevel --- ## If Statements in the Go Language First, the reason we chose Go is that, among modern languages, Go has the most "beautiful" assembly, and its syntactic efficiency is sometimes overwhelming even when compared with classical languages. Now that we have understood how a simple Go program works from the previous lecture, let us compare Go and Assembly line by line. ### Source Code First, as is the case with Go, modern compilers, including GCC, automatically optimize branches that serve no purpose. GCC and Clang, which are C language compilers, also perform very aggressive optimizations at -O2, the industry standard, so the era in which programmers could hardly trust compilers fully was, in fact, completed in the late 20th century. Therefore, from the compiler's perspective, a condition must be given that is difficult to predict and replace with a different construct; only then does it have meaning. ```go package main import ( "os" "strconv" ) func main() { // If this were handled with something predictable and removable, such as x = 10, // the compiler would optimize it and remove the branch. // Therefore, to inspect such code directly as assembly, in C one would use -O0 or similar, // or otherwise make the compiler use an external value it cannot predict, // but since this section deals with modern programming, we will not use the method of // disabling Go's binary optimizations. if len(os.Args) < 2 { return } x, _ := strconv.Atoi(os.Args[1]) if x < 10 { println("X is smaller than 10") } else { println("X is larger or same as 10") } } ``` In this case, since the compiler cannot predict the input, the branch statement is translated into machine code as is. ### Assembly Language ```asm TEXT main.main(SB) /home/yjlee/introduction-to-golang/learn-golang/if-and-switch/golang-if/main.go main.go:8 0x47a840 493b6610 CMPQ SP, 0x10(R14) main.go:8 0x47a844 7670 JBE 0x47a8b6 main.go:8 0x47a846 55 PUSHQ BP main.go:8 0x47a847 4889e5 MOVQ SP, BP main.go:8 0x47a84a 4883ec10 SUBQ $0x10, SP main.go:15 0x47a84e 48833d12fb0a0002 CMPQ os.Args+8(SB), $0x2 main.go:15 0x47a856 7c58 JL 0x47a8b0 main.go:15 0x47a858 488b0d01fb0a00 MOVQ os.Args(SB), CX main.go:18 0x47a85f 488b4110 MOVQ 0x10(CX), AX main.go:18 0x47a863 488b5918 MOVQ 0x18(CX), BX main.go:18 0x47a867 e834e8ffff CALL strconv.Atoi(SB) main.go:20 0x47a86c 4883f80a CMPQ AX, $0xa main.go:20 0x47a870 7d1d JGE 0x47a88f main.go:21 0x47a872 e809befbff CALL runtime.printlock(SB) main.go:21 0x47a877 488d0519f50100 LEAQ 0x1f519(IP), AX main.go:21 0x47a87e bb15000000 MOVL $0x15, BX main.go:21 0x47a883 e878c6fbff CALL runtime.printstring(SB) main.go:21 0x47a888 e853befbff CALL runtime.printunlock(SB) main.go:21 0x47a88d eb1b JMP 0x47a8aa main.go:23 0x47a88f e8ecbdfbff CALL runtime.printlock(SB) main.go:23 0x47a894 488d05c8040200 LEAQ 0x204c8(IP), AX main.go:23 0x47a89b bb1a000000 MOVL $0x1a, BX main.go:23 0x47a8a0 e85bc6fbff CALL runtime.printstring(SB) main.go:23 0x47a8a5 e836befbff CALL runtime.printunlock(SB) main.go:25 0x47a8aa 4883c410 ADDQ $0x10, SP main.go:25 0x47a8ae 5d POPQ BP main.go:25 0x47a8af c3 RET main.go:16 0x47a8b0 4883c410 ADDQ $0x10, SP main.go:16 0x47a8b4 5d POPQ BP main.go:16 0x47a8b5 c3 RET main.go:8 0x47a8b6 e845f0feff CALL runtime.morestack_noctxt.abi0(SB) main.go:8 0x47a8bb eb83 JMP main.main(SB) ``` Using the go tool, it kindly shows which syntax matches which assembly, one by one. Since what we will learn this time is comparison statements and if branch statements, it is enough to focus on a few lines. ### CMPQ Instruction & JL Instruction The `CMPQ` instruction is an instruction for comparing 4-byte (4-word) data types, and its etymology is **C**o**MP**are**Q**uadword, abbreviated to **CMPQ**. Looking at the memory address `0x47a84e`, the statement `CMPQ os.Args+8(SB), $0x2` is present. In this case, the program compares the number of arguments it received with the hexadecimal `0x2` (that is, simply 2). Afterward, through `JL`, the program performs a jump based on this comparison to check whether the number of arguments is less than 2 (that is, whether the program itself is the only argument). In other words, this is **J**ump, **L**ess than' abbreviated to **JL**. For the preceding comparison, if the number of arguments was less than 2, the program jumps to the address `0x47a8b0`, where a `JGE` is located. However, the register used in this statement is the `AX` register, so one must know what value is stored in the register. ### MOVQ Instruction After this, one must understand how the 'CX' register is actually used to store the starting address of the data, and how the actual data is extracted after the address is read. The range `0x47858`-`0x47863` shows this operation being performed step by step. First, the starting address of the argument array is inserted into the CX register with the `MOVQ os.Args(SB), CX` instruction. At this point, one must understand Go's string type. A Go `string` is a struct, and this struct consists of two 8-byte values, making up 16 bytes. | struct | 8 bytes | 8 bytes | | ------ | -------------- | ------------- | | string | memory address | string length | Visually, it looks like the above: the first 8 bytes store the starting address of the string, and the latter 8 bytes store the length of the string. Therefore, the address of the string is stored in the AX register, and the length of the string is stored in the BX register. ### CALL In the previous post as well, when looking at `runtime` functions, the CALL instruction was attached. This is attached in front of functions used in Go, and it means that a certain function is **called**, so it is literally CALL. Afterward, the CALL function is used to convert the string into an integer, and where the integer is stored at this point is _abstracted within the function and not visible._ ### CMPQ Instruction & JGE Instruction Returning to the earlier address `0x47a86c`, the instruction is **comparing the address of the string with the number `0xa` (decimal 10)!** This means that the program no longer uses that argument, so **it overwrote the location of the string to create the slot for the integer variable `x`.** This is the essence of the aggressive optimization performed in Go and other languages. Next, the JGE instruction appears, which is an abbreviation of **J**ump, **G**reater or **E**quals. Therefore, this statement asks whether the value is greater than or equal to the comparison target. Therefore, it is not the `x < 10` statement as it is; rather, it is `x < 10` with **the comparison direction of the statement reversed!** In machine code, skipping ahead preemptively when the condition does not hold is more intuitive than performing the comparison once when the condition holds and then checking again whether it does not hold, and it **saves one instruction**. Because this optimization is very classical, it frequently appears even in compilers whose optimization level is considerably low, unlike the `strconv.Atoi` example seen above, so it is good to know. Therefore, by applying this idea, one can obtain sources that differ but match 100% at the assembly level. ### Mirror Image Code Example Using the script below, the bash script creates two mirror-image sources with 100% equivalence, and then verifies that, excluding metadata that changes each time, when only `main` is viewed, exactly identical assembly is obtained. ```bash #!/usr/bin/env bash # 1. Completely reset existing leftover files and directories echo "[1/6] Cleaning up old artifacts..." rm -rf test_dir main_orig main_asm orig.asm asm.asm orig_pure.asm asm_pure.asm mkdir -p test_dir # 2. Write the original source code (main.go) echo "[2/6] Generating main.go..." cat << 'EOF' > main.go package main import ( "os" "strconv" ) func main() { if len(os.Args) < 2 { return } x, _ := strconv.Atoi(os.Args[1]) s1 := "X is smaller than 10" s2 := "X is larger or same as 10" if x < 10 { println(s1) } else { println(s2) } } EOF # 3. Write the mirror-image source code (main_from_asm.go) # Perfectly synchronize the operator structure symmetrically so that the compiler uses the optimization template (JGE) as is echo "[3/6] Generating main_from_asm.go..." cat << 'EOF' > main_from_asm.go package main import ( "os" "strconv" ) func main() { if len(os.Args) < 2 { return } x, _ := strconv.Atoi(os.Args[1]) s1 := "X is smaller than 10" s2 := "X is larger or same as 10" // If the x < 10 structure is maintained with 10 as the reference, the compiler // adopts exactly the same JGE mechanism and block layout as main.go. switch { case x < 10: println(s1) default: println(s2) } } EOF # 4. Build each in the same directory path and file name environment echo "[4/6] Compiling both sources inside 'test_dir'..." cp main.go test_dir/main.go cd test_dir && go build -o ../main_orig main.go && cd .. rm test_dir/main.go cp main_from_asm.go test_dir/main.go cd test_dir && go build -o ../main_asm main.go && cd .. # 5. Extract the pure main.main assembly function using go tool objdump echo "[5/6] Extracting main.main assembly sections..." go tool objdump -s "main\.main" main_orig > orig.asm go tool objdump -s "main\.main" main_asm > asm.asm # Remove the virtual addresses, offsets, and machine-code byte data text, # and filter only the pure instruction set (Opcode & Operands) fields that the CPU will execute awk '{print $4, $5, $6, $7}' orig.asm > orig_pure.asm awk '{print $4, $5, $6, $7}' asm.asm > asm_pure.asm # 6. Verify the two machine instruction structures via diff echo "[6/6] Verifying assembly structural integrity via diff..." echo "------------------------------------------------------------" if diff orig_pure.asm asm_pure.asm > /dev/null; then echo "===> [성공] 두 바이너리의 main.main 기계어 로직이 100% 일치합니다! <===" echo "컴파일러의 최적화 파이프라인 가이드라인을 완벽하게 동기화하여 동일한 어셈블리를 얻었습니다." else echo "===> [실패] 어셈블리 명령어 구조에 차이점이 발견되었습니다. <===" diff -u orig_pure.asm asm_pure.asm fi echo "------------------------------------------------------------" ``` When you actually run the source, you can obtain information like the following. ```text [1/6] Cleaning up old artifacts... [2/6] Generating main.go... [3/6] Generating main_from_asm.go... [4/6] Compiling both sources inside 'test_dir'... [5/6] Extracting main.main assembly sections... [6/6] Verifying assembly structural integrity via diff... ------------------------------------------------------------ ===> [성공] 두 바이너리의 main.main 기계어 로직이 100% 일치합니다! <=== 컴파일러의 최적화 파이프라인 가이드라인을 완벽하게 동기화하여 동일한 어셈블리를 얻었습니다. ------------------------------------------------------------ ``` ### Conclusion Programming languages provide many abstractions, but behind those abstractions lie very interesting and aggressive optimizations. In addition, by exploiting this point in reverse, we were able to create mirror-image code whose source differs but whose assembly is identical. If you are interested in low-level matters and encounter proprietary software written in Go, it does not seem impossible to disassemble the assembly directly, analyze it, and recover the source. ## Next Lecture In the next session, we will look at the `select-case` statement, which offers another kind of interest alongside the If statement.