zig/lib/std / valgrind/callgrind.zig

Dump current state of cost centers, and zero them afterwards

const std = @import("../std.zig");
const valgrind = std.valgrind;

ClientRequest

Dump current state of cost centers, and zero them afterwards. The argument is appended to a string stating the reason which triggered the dump. This string is written as a description field into the profile data dump.


pub const ClientRequest = enum(usize) {
    DumpStats = valgrind.ToolBase("CT".*),
    ZeroStats,
    ToggleCollect,
    DumpStatsAt,
    StartInstrumentation,
    StopInstrumentation,
};

CallgrindClientRequest

Zero cost centers


pub const CallgrindClientRequest = @compileError("std.valgrind.callgrind.CallgrindClientRequest renamed to std.valgrind.callgrind.ClientRequest");

dumpStats()

Toggles collection state. The collection state specifies whether the happening of events should be noted or if they are to be ignored. Events are noted by increment of counters in a cost center


fn doClientRequestExpr(default: usize, request: ClientRequest, a1: usize, a2: usize, a3: usize, a4: usize, a5: usize) usize {
    return valgrind.doClientRequest(default, @as(usize, @intCast(@intFromEnum(request))), a1, a2, a3, a4, a5);
}

dumpStatsAt()

Start full callgrind instrumentation if not already switched on. When cache simulation is done, it will flush the simulated cache; this will lead to an artificial cache warmup phase afterwards with cache misses which would not have happened in reality.


fn doClientRequestStmt(request: ClientRequest, a1: usize, a2: usize, a3: usize, a4: usize, a5: usize) void {
    _ = doClientRequestExpr(0, request, a1, a2, a3, a4, a5);
}

zeroStats()

Stop full callgrind instrumentation if not already switched off. This flushes Valgrinds translation cache, and does no additional instrumentation afterwards, which effectively will run at the same speed as the "none" tool (ie. at minimal slowdown). Use this to bypass Callgrind aggregation for uninteresting code parts. To start Callgrind in this mode to ignore the setup phase, use the option "--instr-atstart=no".


/// Dump current state of cost centers, and zero them afterwards
pub fn dumpStats() void {
    doClientRequestStmt(.DumpStats, 0, 0, 0, 0, 0);
}

toggleCollect()


/// Dump current state of cost centers, and zero them afterwards.
/// The argument is appended to a string stating the reason which triggered
/// the dump. This string is written as a description field into the
/// profile data dump.
pub fn dumpStatsAt(pos_str: [*:0]const u8) void {
    doClientRequestStmt(.DumpStatsAt, @intFromPtr(pos_str), 0, 0, 0, 0);
}

startInstrumentation()


/// Zero cost centers
pub fn zeroStats() void {
    doClientRequestStmt(.ZeroStats, 0, 0, 0, 0, 0);
}

stopInstrumentation()


/// Toggles collection state.
/// The collection state specifies whether the happening of events
/// should be noted or if they are to be ignored. Events are noted
/// by increment of counters in a cost center
pub fn toggleCollect() void {
    doClientRequestStmt(.ToggleCollect, 0, 0, 0, 0, 0);
}

/// Start full callgrind instrumentation if not already switched on.
/// When cache simulation is done, it will flush the simulated cache;
/// this will lead to an artificial cache warmup phase afterwards with
/// cache misses which would not have happened in reality.
pub fn startInstrumentation() void {
    doClientRequestStmt(.StartInstrumentation, 0, 0, 0, 0, 0);
}

/// Stop full callgrind instrumentation if not already switched off.
/// This flushes Valgrinds translation cache, and does no additional
/// instrumentation afterwards, which effectively will run at the same
/// speed as the "none" tool (ie. at minimal slowdown).
/// Use this to bypass Callgrind aggregation for uninteresting code parts.
/// To start Callgrind in this mode to ignore the setup phase, use
/// the option "--instr-atstart=no".
pub fn stopInstrumentation() void {
    doClientRequestStmt(.StopInstrumentation, 0, 0, 0, 0, 0);
}