GoAT

Go assembly transpiler for C programming languages.

It help to utilize optimization from C compiler in Go projects. For example, generate SIMD vectorized functions for Go (refer to How to Use AVX512 in Golang).

Install

1. Install prerequisites

sudo apt update
sudo apt install clang libc6-dev-i386

2. Install GoAT

go install github.com/gorse-io/goat@latest

Usage

git clone https://github.com/gorse-io/goat.git
cd goat/example
goat src/avx_mul_to.c -O3 -mavx -mfma -mavx512f -mavx512dq

GoAT transpiles example/src/avx_mul_to.c to two files.

Go function definition file avx_mul_to.go:

//go:build !noasm && amd64
// AUTO-GENERATED BY GOAT -- DO NOT EDIT

package example

import "unsafe"

//go:noescape
func avx_mul_to(a, b, c, n unsafe.Pointer)

Go assembly file avx_mul_to.s:

//go:build !noasm && amd64
// AUTO-GENERATED BY GOAT -- DO NOT EDIT

TEXT ·avx_mul_to(SB), $0-32
	MOVQ a+0(FP), DI
	MOVQ b+8(FP), SI
	MOVQ c+16(FP), DX
	MOVQ n+24(FP), CX
	BYTE $0x55               // pushq	%rbp
	WORD $0x8948; BYTE $0xe5 // movq	%rsp, %rbp
	LONG $0xf8e48348         // andq	$-8, %rsp
	WORD $0x8548; BYTE $0xc9 // testq	%rcx, %rcx
	JLE  LBB0_12
	LONG $0x3ff98348         // cmpq	$63, %rcx
	JA   LBB0_7
	WORD $0xc031             // xorl	%eax, %eax
	JMP  LBB0_3

LBB0_7:
	LONG $0x8a0c8d4c         // leaq	(%rdx,%rcx,4), %r9
	LONG $0x8f048d48         // leaq	(%rdi,%rcx,4), %rax
	WORD $0x3948; BYTE $0xd0 // cmpq	%rdx, %rax
	LONG $0xc2970f41         // seta	%r10b
	LONG $0x8e048d48         // leaq	(%rsi,%rcx,4), %rax
	WORD $0x3949; BYTE $0xf9 // cmpq	%rdi, %r9
	LONG $0xc3970f41         // seta	%r11b
	WORD $0x3948; BYTE $0xd0 // cmpq	%rdx, %rax
	LONG $0xc0970f41         // seta	%r8b
	WORD $0x3949; BYTE $0xf1 // cmpq	%rsi, %r9
	LONG $0xc1970f41         // seta	%r9b
	WORD $0xc031             // xorl	%eax, %eax
	WORD $0x8445; BYTE $0xda // testb	%r11b, %r10b
	JNE  LBB0_3
	WORD $0x2045; BYTE $0xc8 // andb	%r9b, %r8b
	JNE  LBB0_3
	WORD $0x8948; BYTE $0xc8 // movq	%rcx, %rax
	LONG $0xc0e08348         // andq	$-64, %rax
	WORD $0x3145; BYTE $0xc0 // xorl	%r8d, %r8d

LBB0_10:
	LONG $0x487cb162; WORD $0x0410; BYTE $0x87 // vmovups	(%rdi,%r8,4), %zmm0
	QUAD $0x01874c10487cb162                   // vmovups	64(%rdi,%r8,4), %zmm1
	QUAD $0x02875410487cb162                   // vmovups	128(%rdi,%r8,4), %zmm2
	QUAD $0x03875c10487cb162                   // vmovups	192(%rdi,%r8,4), %zmm3
	LONG $0x487cb162; WORD $0x0459; BYTE $0x86 // vmulps	(%rsi,%r8,4), %zmm0, %zmm0
	QUAD $0x01864c594874b162                   // vmulps	64(%rsi,%r8,4), %zmm1, %zmm1
	QUAD $0x02865459486cb162                   // vmulps	128(%rsi,%r8,4), %zmm2, %zmm2
	QUAD $0x03865c594864b162                   // vmulps	192(%rsi,%r8,4), %zmm3, %zmm3
	LONG $0x487cb162; WORD $0x0411; BYTE $0x82 // vmovups	%zmm0, (%rdx,%r8,4)
	QUAD $0x01824c11487cb162                   // vmovups	%zmm1, 64(%rdx,%r8,4)
	QUAD $0x02825411487cb162                   // vmovups	%zmm2, 128(%rdx,%r8,4)
	QUAD $0x03825c11487cb162                   // vmovups	%zmm3, 192(%rdx,%r8,4)
	LONG $0x40c08349                           // addq	$64, %r8
	WORD $0x394c; BYTE $0xc0                   // cmpq	%r8, %rax
	JNE  LBB0_10
	WORD $0x3948; BYTE $0xc8                   // cmpq	%rcx, %rax
	JE   LBB0_12

LBB0_3:
	WORD $0x8949; BYTE $0xc0 // movq	%rax, %r8
	WORD $0xf749; BYTE $0xd0 // notq	%r8
	WORD $0x0149; BYTE $0xc8 // addq	%rcx, %r8
	WORD $0x8949; BYTE $0xc9 // movq	%rcx, %r9
	LONG $0x03e18349         // andq	$3, %r9
	JE   LBB0_5

LBB0_4:
	LONG $0x0410fac5; BYTE $0x87 // vmovss	(%rdi,%rax,4), %xmm0
	LONG $0x0459fac5; BYTE $0x86 // vmulss	(%rsi,%rax,4), %xmm0, %xmm0
	LONG $0x0411fac5; BYTE $0x82 // vmovss	%xmm0, (%rdx,%rax,4)
	LONG $0x01c08348             // addq	$1, %rax
	LONG $0xffc18349             // addq	$-1, %r9
	JNE  LBB0_4

LBB0_5:
	LONG $0x03f88349 // cmpq	$3, %r8
	JB   LBB0_12

LBB0_6:
	LONG $0x0410fac5; BYTE $0x87   // vmovss	(%rdi,%rax,4), %xmm0
	LONG $0x0459fac5; BYTE $0x86   // vmulss	(%rsi,%rax,4), %xmm0, %xmm0
	LONG $0x0411fac5; BYTE $0x82   // vmovss	%xmm0, (%rdx,%rax,4)
	LONG $0x4410fac5; WORD $0x0487 // vmovss	4(%rdi,%rax,4), %xmm0
	LONG $0x4459fac5; WORD $0x0486 // vmulss	4(%rsi,%rax,4), %xmm0, %xmm0
	LONG $0x4411fac5; WORD $0x0482 // vmovss	%xmm0, 4(%rdx,%rax,4)
	LONG $0x4410fac5; WORD $0x0887 // vmovss	8(%rdi,%rax,4), %xmm0
	LONG $0x4459fac5; WORD $0x0886 // vmulss	8(%rsi,%rax,4), %xmm0, %xmm0
	LONG $0x4411fac5; WORD $0x0882 // vmovss	%xmm0, 8(%rdx,%rax,4)
	LONG $0x4410fac5; WORD $0x0c87 // vmovss	12(%rdi,%rax,4), %xmm0
	LONG $0x4459fac5; WORD $0x0c86 // vmulss	12(%rsi,%rax,4), %xmm0, %xmm0
	LONG $0x4411fac5; WORD $0x0c82 // vmovss	%xmm0, 12(%rdx,%rax,4)
	LONG $0x04c08348               // addq	$4, %rax
	WORD $0x3948; BYTE $0xc1       // cmpq	%rax, %rcx
	JNE  LBB0_6

LBB0_12:
	WORD $0x8948; BYTE $0xec // movq	%rbp, %rsp
	BYTE $0x5d               // popq	%rbp
	WORD $0xf8c5; BYTE $0x77 // vzeroupper
	BYTE $0xc3               // retq

Finally, the mul_to function can be called by:

func AVXMulTo(a, b, c []float32) {
	if len(a) ! = len(b) || len(a) ! = len(c) {
		panic("floats: slice lengths do not match")
	}
	avx_mul_to(unsafe.Pointer(&a[0]), unsafe.Pointer(&b[0]), unsafe.Pointer(&c[0]), unsafe.Pointer(uintptr(len(a))))
}

Limitations

• Arguments need (for now) to be 64-bit size, meaning either a value or a pointer
• Maximum number of 4 arguments
• Generally no call statements

Acknowledgments

• GoAT is inspired by c2goasm.