Skip to content

perf: optimization of map access #692

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
samber merged 1 commit into from
Sep 26, 2025
Merged

perf: optimization of map access #692

samber merged 1 commit into from
Sep 26, 2025

Conversation

@mabrarov
Copy link
Contributor

@mabrarov mabrarov commented Sep 26, 2025

Changes

Optimization of map access.

Notes

Probably, more optimizations can be done. This PR covers just author's recent experience.

Details

The way original code (loop only) can look like in terms of Go assembler - refer to https://godbolt.org/z/j8eYrT9jh (x86-64, gc 1.24.2):

        LEAQ    main..autotmp_8+56(SP), DI
        PCDATA  $0, $-2
        LEAQ    -32(DI), DI
        NOP
        DUFFZERO        $331
        PCDATA  $0, $-1
        MOVQ    main..dict+168(SP), DX
        MOVQ    (DX), AX
        MOVQ    main.in+176(SP), BX
        LEAQ    main..autotmp_8+56(SP), CX
        PCDATA  $1, $1
        CALL    runtime.mapIterStart(SB)
        JMP     main_MapKeys[go_shape_int,go_shape_int,go_shape_int]_pc293
main_MapKeys[go_shape_int,go_shape_int,go_shape_int]_pc149:
        MOVQ    (DX), CX
        MOVQ    CX, main.k+24(SP)
        MOVQ    main..dict+168(SP), DX
        MOVQ    (DX), AX
        MOVQ    main.in+176(SP), BX
        CALL    runtime.mapaccess1_fast64(SB)
        MOVQ    main.iteratee+184(SP), DX
        MOVQ    (DX), SI
        MOVQ    (AX), AX
        MOVQ    main.k+24(SP), BX
        CALL    SI
        MOVQ    AX, main..autotmp_18+40(SP)
        MOVQ    main..dict+168(SP), DX
        MOVQ    (DX), AX
        MOVQ    main.in+176(SP), BX
        MOVQ    main.k+24(SP), CX
        CALL    runtime.mapaccess1_fast64(SB)
        MOVQ    main..dict+168(SP), DX
        MOVQ    24(DX), DX
        MOVQ    (AX), SI
        MOVQ    SI, main..autotmp_19+32(SP)
        MOVQ    DX, AX
        MOVQ    main.result+48(SP), BX
        MOVQ    main..autotmp_18+40(SP), CX
        CALL    runtime.mapassign_fast64(SB)
        MOVQ    main..autotmp_19+32(SP), DX
        MOVQ    DX, (AX)
        LEAQ    main..autotmp_8+56(SP), AX
        NOP
        CALL    runtime.mapIterNext(SB)
main_MapKeys[go_shape_int,go_shape_int,go_shape_int]_pc293:
        MOVQ    main..autotmp_8+56(SP), DX
        TESTQ   DX, DX
        JNE     main_MapKeys[go_shape_int,go_shape_int,go_shape_int]_pc149

The way new code (loop only) can look like in terms of Go assembler - refer to https://godbolt.org/z/r83o1E6as (x86-64, gc 1.24.2):

        LEAQ    main..autotmp_12+40(SP), DI
        PCDATA  $0, $-2
        LEAQ    -32(DI), DI
        DUFFZERO        $331
        PCDATA  $0, $-1
        LEAQ    type:map[int]int(SB), AX
        MOVQ    main.in+152(SP), BX
        LEAQ    main..autotmp_12+40(SP), CX
        PCDATA  $1, $1
        CALL    runtime.mapIterStart(SB)
        JMP     main_MapKeys[int,int,int]_pc219
main_MapKeys[int,int,int]_pc146:
        MOVQ    (CX), BX
        MOVQ    main..autotmp_12+48(SP), CX
        MOVQ    (CX), AX
        MOVQ    AX, main.v+24(SP)
        MOVQ    main.iteratee+160(SP), DX
        MOVQ    (DX), CX
        CALL    CX
        LEAQ    type:map[int]int(SB), CX
        MOVQ    main.result+32(SP), BX
        MOVQ    AX, DX
        MOVQ    CX, AX
        MOVQ    DX, CX
        CALL    runtime.mapassign_fast64(SB)
        MOVQ    main.v+24(SP), CX
        MOVQ    CX, (AX)
        LEAQ    main..autotmp_12+40(SP), AX
        CALL    runtime.mapIterNext(SB)
main_MapKeys[int,int,int]_pc219:
        MOVQ    main..autotmp_12+40(SP), CX
        TESTQ   CX, CX
        JNE     main_MapKeys[int,int,int]_pc146

The new code doesn't have 2 calls of runtime.mapaccess1_fast64 (CALL runtime.mapaccess1_fast64) comparing to old code. I guess, it means that the new code is more optimal.

@mabrarov
perf: optimization of map access.
d5ecd78 
Signed-off-by: Marat Abrarov <abrarov@gmail.com>
@mabrarov mabrarov changed the title feat: optimization of map access perf: optimization of map access Sep 26, 2025
@samber samber merged commit 856a933 into samber:master Sep 26, 2025
@NathanBaulch NathanBaulch mentioned this pull request Sep 27, 2025
Merged
NathanBaulch pushed a commit to NathanBaulch/lo that referenced this pull request Sep 30, 2025
@mabrarov @NathanBaulch
perf: optimization of map access. (samber#692)
ed698c2 
Signed-off-by: Marat Abrarov <abrarov@gmail.com>
@renovate renovate bot mentioned this pull request Oct 19, 2025
Merged
1 task
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@mabrarov @samber