Remove mixnode modifications

Bundle libcpucycles
mixnode feature
2023-03-27 14:21:44 +02:00 · 2023-03-27 14:11:03 +02:00 · 2023-03-27 09:57:28 +02:00 · 2023-03-27 09:13:20 +02:00 · 2023-03-15 11:21:48 +00:00 · 2023-03-14 14:38:34 +00:00
154 changed files with 4892 additions and 1897 deletions
@@ -43,3 +43,4 @@ Cargo.lock
 nym-connect/Cargo.lock
 .parcel-cache
 **/.DS_Store
+cpu-cycles/libcpucycles/build
@@ -302,23 +302,22 @@ dependencies = [

 [[package]]
 name = "bip39"
-version = "2.0.0"
+version = "1.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "93f2635620bf0b9d4576eb7bb9a38a55df78bd1205d26fa994b25911a69f212f"
+checksum = "b9e89470017230c38e52b82b3ee3f530db1856ba1d434e3a67a3456a8a8dec5f"
 dependencies = [
 "bitcoin_hashes",
- "rand 0.8.5",
- "rand_core 0.6.4",
+ "rand 0.6.5",
+ "rand_core 0.4.2",
 "serde",
 "unicode-normalization",
- "zeroize",
 ]

 [[package]]
 name = "bitcoin_hashes"
-version = "0.11.0"
+version = "0.9.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "90064b8dee6815a6470d60bad07bbbaee885c0e12d04177138fa3291a01b7bc4"
+checksum = "7ce18265ec2324ad075345d5814fbeed4f41f0a660055dc78840b74d19b874b1"

 [[package]]
 name = "bitflags"
@@ -505,22 +504,6 @@ dependencies = [
 "zeroize",
 ]

-[[package]]
-name = "chitchat"
-version = "0.5.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fec1bff4fabc59f0b7408478183a796ab0a9bcb0534bb413f3cf66f44ce2aca1"
-dependencies = [
- "anyhow",
- "async-trait",
- "bytes",
- "rand 0.8.5",
- "serde",
- "tokio",
- "tokio-stream",
- "tracing",
-]
-
 [[package]]
 name = "chrono"
 version = "0.4.23"
@@ -2723,9 +2706,9 @@ dependencies = [

 [[package]]
 name = "libc"
-version = "0.2.139"
+version = "0.2.140"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "201de327520df007757c1f0adce6e827fe8562fbc28bfd9c15571c66ca1f5f79"
+checksum = "99227334921fae1a979cf0bfdfcc6b3e5ce376ef57e16fb6fb3ea2ed6095f80c"

 [[package]]
 name = "libgit2-sys"
@@ -2862,6 +2845,12 @@ dependencies = [
 "serde",
 ]

+[[package]]
+name = "maybe-uninit"
+version = "2.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "60302e4db3a61da70c0cb7991976248362f30319e88850c487b9b95bbf059e00"
+
 [[package]]
 name = "memchr"
 version = "2.5.0"
@@ -2927,7 +2916,6 @@ name = "mixnode-common"
 version = "0.1.0"
 dependencies = [
 "bytes",
- "chitchat",
 "futures",
 "humantime-serde",
 "log",
@@ -2943,7 +2931,6 @@ dependencies = [
 "nym-task",
 "rand 0.8.5",
 "serde",
- "serde_json",
 "thiserror",
 "tokio",
 "tokio-util",
@@ -3509,7 +3496,6 @@ dependencies = [
 "anyhow",
 "atty",
 "bs58",
- "chitchat",
 "clap 4.1.4",
 "colored",
 "cupid",
@@ -4185,7 +4171,7 @@ dependencies = [
 "instant",
 "libc",
 "redox_syscall",
- "smallvec",
+ "smallvec 1.10.0",
 "winapi",
 ]

@@ -4198,7 +4184,7 @@ dependencies = [
 "cfg-if",
 "libc",
 "redox_syscall",
- "smallvec",
+ "smallvec 1.10.0",
 "windows-sys 0.45.0",
 ]

@@ -5071,7 +5057,7 @@ dependencies = [
 "pin-project-lite",
 "ref-cast",
 "serde",
- "smallvec",
+ "smallvec 1.10.0",
 "stable-pattern",
 "state",
 "time 0.3.17",
@@ -5566,6 +5552,15 @@ dependencies = [
 "autocfg 1.1.0",
 ]

+[[package]]
+name = "smallvec"
+version = "0.6.14"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b97fcaeba89edba30f044a10c6a3cc39df9c3f17d7cd829dd1446cab35f890e0"
+dependencies = [
+ "maybe-uninit",
+]
+
 [[package]]
 name = "smallvec"
 version = "1.10.0"
@@ -5730,7 +5725,7 @@ dependencies = [
 "percent-encoding",
 "rustls 0.19.1",
 "sha2 0.10.6",
- "smallvec",
+ "smallvec 1.10.0",
 "sqlformat 0.1.8",
 "sqlx-rt 0.5.13",
 "stringprep",
@@ -5778,7 +5773,7 @@ dependencies = [
 "rustls 0.20.8",
 "rustls-pemfile",
 "sha2 0.10.6",
- "smallvec",
+ "smallvec 1.10.0",
 "sqlformat 0.2.1",
 "sqlx-rt 0.6.2",
 "stringprep",
@@ -6203,21 +6198,6 @@ dependencies = [
 "serde_json",
 ]

-[[package]]
-name = "tinyvec"
-version = "1.6.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "87cc5ceb3875bb20c2890005a4e226a4651264a5c75edb2421b52861a0a0cb50"
-dependencies = [
- "tinyvec_macros",
-]
-
-[[package]]
-name = "tinyvec_macros"
-version = "0.1.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1f3ccbac311fea05f86f61904b462b55fb3df8837a366dfc601a0161d0532f20"
-
 [[package]]
 name = "tokio"
 version = "1.25.0"
@@ -6515,7 +6495,7 @@ dependencies = [
 "once_cell",
 "regex",
 "sharded-slab",
- "smallvec",
+ "smallvec 1.10.0",
 "thread_local",
 "tracing",
 "tracing-core",
@@ -6681,11 +6661,11 @@ checksum = "84a22b9f218b40614adcb3f4ff08b703773ad44fa9423e4e0d346d5db86e4ebc"

 [[package]]
 name = "unicode-normalization"
-version = "0.1.22"
+version = "0.1.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5c5713f0fc4b5db668a2ac63cdb7bb4469d8c9fed047b1d0292cc7b0ce2ba921"
+checksum = "09c8070a9942f5e7cfccd93f490fdebd230ee3c3c9f107cb25bad5351ef671cf"
 dependencies = [
- "tinyvec",
+ "smallvec 0.6.14",
 ]

 [[package]]
@@ -105,7 +105,6 @@ license = "Apache-2.0"

 [workspace.dependencies]
 async-trait = "0.1.64"
-bip39 = { version = "2.0.0", features = ["zeroize"] }
 cfg-if = "1.0.0"
 dotenvy = "0.15.6"
 lazy_static = "1.4.0"
@@ -6,7 +6,7 @@ edition = "2021"
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html

 [dependencies]
-bip39 = { workspace = true }
+bip39 = "1.0.1"
 clap = { version = "4.0", features = ["cargo", "derive"] }
 log = "0.4"
 rand = "0.7.3"
@@ -0,0 +1,3 @@
+/target
+**/*.rs.bk
+Cargo.lock
@@ -37,7 +37,7 @@ nym-execute = { path = "../../execute" }
 # at some point it might be possible to make it wasm-compatible
 # perhaps after https://github.com/cosmos/cosmos-rust/pull/97 is resolved (and tendermint-rs is updated)
 async-trait = { workspace = true, optional = true }
-bip39 = { workspace = true, features = ["rand"], optional = true }
+bip39 = { version = "1", features = ["rand"], optional = true }
 nym-config = { path = "../../config", optional = true }
 cosmrs = { git = "https://github.com/neacsu/cosmos-rust", branch = "neacsu/feegrant_support", features = ["rpc", "bip32", "cosmwasm"], optional = true}
 cw3 = { version = "0.13.4", optional = true }
@@ -47,7 +47,7 @@ flate2 = { version = "1.0.20", optional = true }
 sha2 = { version = "0.9.5", optional = true }
 itertools = { version = "0.10", optional = true }
 cosmwasm-std = { version = "1.0.0", optional = true }
-zeroize = { version = "1.5.7", optional = true, features = ["zeroize_derive"] }
+zeroize = { version = "1.5.7", optional = true }

 [dev-dependencies]
 ts-rs = "6.1.2"
@@ -9,7 +9,6 @@ use crate::nyxd::{Fee, NyxdClient, SigningCosmWasmClient};
 use async_trait::async_trait;
 use cosmrs::AccountId;
 use nym_contracts_common::signing::MessageSignature;
-use nym_mixnet_contract_common::gateway::GatewayConfigUpdate;
 use nym_mixnet_contract_common::mixnode::{MixNodeConfigUpdate, MixNodeCostParams};
 use nym_mixnet_contract_common::reward_params::{IntervalRewardingParamsUpdate, Performance};
 use nym_mixnet_contract_common::{
@@ -499,36 +498,6 @@ pub trait MixnetSigningClient {
        .await
    }

-    async fn update_gateway_config(
-        &self,
-        new_config: GatewayConfigUpdate,
-        fee: Option<Fee>,
-    ) -> Result<ExecuteResult, NyxdError> {
-        self.execute_mixnet_contract(
-            fee,
-            MixnetExecuteMsg::UpdateGatewayConfig { new_config },
-            vec![],
-        )
-        .await
-    }
-
-    async fn update_gateway_config_on_behalf(
-        &self,
-        owner: AccountId,
-        new_config: GatewayConfigUpdate,
-        fee: Option<Fee>,
-    ) -> Result<ExecuteResult, NyxdError> {
-        self.execute_mixnet_contract(
-            fee,
-            MixnetExecuteMsg::UpdateGatewayConfigOnBehalf {
-                new_config,
-                owner: owner.to_string(),
-            },
-            vec![],
-        )
-        .await
-    }
-
    // delegation-related:

    async fn delegate_to_mixnode(
@@ -7,7 +7,6 @@ use crate::nyxd::error::NyxdError;
 use crate::nyxd::{Coin, Fee, NyxdClient};
 use async_trait::async_trait;
 use nym_contracts_common::signing::MessageSignature;
-use nym_mixnet_contract_common::gateway::GatewayConfigUpdate;
 use nym_mixnet_contract_common::mixnode::{MixNodeConfigUpdate, MixNodeCostParams};
 use nym_mixnet_contract_common::{Gateway, MixId, MixNode};
 use nym_vesting_contract_common::messages::{
@@ -36,12 +35,6 @@ pub trait VestingSigningClient {
        fee: Option<Fee>,
    ) -> Result<ExecuteResult, NyxdError>;

-    async fn vesting_update_gateway_config(
-        &self,
-        new_config: GatewayConfigUpdate,
-        fee: Option<Fee>,
-    ) -> Result<ExecuteResult, NyxdError>;
-
    async fn update_mixnet_address(
        &self,
        address: &str,
@@ -192,19 +185,6 @@ impl<C: SigningCosmWasmClient + Sync + Send + Clone> VestingSigningClient for Ny
            .await
    }

-    async fn vesting_update_gateway_config(
-        &self,
-        new_config: GatewayConfigUpdate,
-        fee: Option<Fee>,
-    ) -> Result<ExecuteResult, NyxdError> {
-        self.execute_vesting_contract(
-            fee,
-            VestingExecuteMsg::UpdateGatewayConfig { new_config },
-            vec![],
-        )
-        .await
-    }
-
    async fn update_mixnet_address(
        &self,
        address: &str,
@@ -8,7 +8,7 @@ use cosmrs::crypto::PublicKey;
 use cosmrs::tx::SignDoc;
 use cosmrs::{tx, AccountId};
 use nym_config::defaults;
-use zeroize::{Zeroize, ZeroizeOnDrop};
+use zeroize::Zeroize;

 /// Derivation information required to derive a keypair and an address from a mnemonic.
 #[derive(Debug, Clone)]
@@ -41,7 +41,7 @@ impl AccountData {

 type Secp256k1Keypair = (SigningKey, PublicKey);

-#[derive(Debug, Clone, Zeroize, ZeroizeOnDrop)]
+#[derive(Debug, Clone)]
 pub struct DirectSecp256k1HdWallet {
    /// Base secret
    secret: bip39::Mnemonic,
@@ -54,10 +54,30 @@ pub struct DirectSecp256k1HdWallet {
    // that would include the secret key which is a dyn EcdsaSigner and hence not Sync making the wallet
    // not Sync and if used on the signing client in an async trait, it wouldn't be Send
    /// Derivation instructions
-    #[zeroize(skip)]
    accounts: Vec<Secp256k1Derivation>,
 }

+impl Zeroize for DirectSecp256k1HdWallet {
+    fn zeroize(&mut self) {
+        // in ideal world, Mnemonic would have had zeroize defined on it (there's an almost year old PR that introduces it)
+        // and the memory would have been filled with zeroes.
+        //
+        // we really don't want to keep our real mnemonic in memory, so let's do the semi-nasty thing
+        // of overwriting it with a fresh mnemonic that was never used before
+        //
+        // note: this function can only fail on an invalid word count, which clearly is not the case here
+        self.secret = bip39::Mnemonic::generate(self.secret.word_count()).unwrap();
+        self.seed.zeroize();
+        // there's nothing secret about derivation paths
+    }
+}
+
+impl Drop for DirectSecp256k1HdWallet {
+    fn drop(&mut self) {
+        self.zeroize()
+    }
+}
+
 impl DirectSecp256k1HdWallet {
    pub fn builder(prefix: &str) -> DirectSecp256k1HdWalletBuilder {
        DirectSecp256k1HdWalletBuilder::new(prefix)
@@ -6,7 +6,7 @@ edition = "2021"

 [dependencies]
 base64 = "0.13.0"
-bip39 = { workspace = true }
+bip39 = "1.0.1"
 bs58 = "0.4"
 comfy-table = "6.0.0"
 cfg-if = "1.0.0"
@@ -6,11 +6,9 @@ use clap::{Args, Subcommand};
 pub mod rewards;

 pub mod delegate_to_mixnode;
-pub mod pledge_more;
 pub mod query_for_delegations;
 pub mod undelegate_from_mixnode;
 pub mod vesting_delegate_to_mixnode;
-pub mod vesting_pledge_more;
 pub mod vesting_undelegate_from_mixnode;

 #[derive(Debug, Args)]
@@ -34,8 +32,4 @@ pub enum MixnetDelegatorsCommands {
    DelegateVesting(vesting_delegate_to_mixnode::Args),
    /// Undelegate from a mixnode (when originally using locked tokens)
    UndelegateVesting(vesting_undelegate_from_mixnode::Args),
-    /// Pledge more
-    PledgeMore(pledge_more::Args),
-    /// Pledge more with locked tokens
-    PledgeMoreVesting(vesting_pledge_more::Args),
 }
@@ -1,29 +0,0 @@
-// Copyright 2021 - Nym Technologies SA <contact@nymtech.net>
-// SPDX-License-Identifier: Apache-2.0
-
-use crate::context::SigningClient;
-use clap::Parser;
-use log::info;
-use nym_mixnet_contract_common::Coin;
-use validator_client::nyxd::traits::MixnetSigningClient;
-
-#[derive(Debug, Parser)]
-pub struct Args {
-    #[clap(long)]
-    pub amount: u128,
-}
-
-pub async fn pledge_more(args: Args, client: SigningClient) {
-    let denom = client.current_chain_details().mix_denom.base.as_str();
-
-    info!("Starting to pledge more");
-
-    let coin = Coin::new(args.amount, denom);
-
-    let res = client
-        .pledge_more(coin.into(), None)
-        .await
-        .expect("failed to pledge more!");
-
-    info!("pledging more: {:?}", res);
-}
@@ -1,30 +0,0 @@
-// Copyright 2021 - Nym Technologies SA <contact@nymtech.net>
-// SPDX-License-Identifier: Apache-2.0
-
-use clap::Parser;
-use log::info;
-use nym_mixnet_contract_common::Coin;
-use validator_client::nyxd::VestingSigningClient;
-
-use crate::context::SigningClient;
-
-#[derive(Debug, Parser)]
-pub struct Args {
-    #[clap(long)]
-    pub amount: u128,
-}
-
-pub async fn vesting_pledge_more(args: Args, client: SigningClient) {
-    let denom = client.current_chain_details().mix_denom.base.as_str();
-
-    info!("Starting vesting pledge more");
-
-    let coin = Coin::new(args.amount, denom);
-
-    let res = client
-        .vesting_pledge_more(coin.into(), None)
-        .await
-        .expect("failed to pledge more!");
-
-    info!("vesting pledge more: {:?}", res);
-}
@@ -5,7 +5,6 @@ use clap::{Args, Subcommand};

 pub mod bond_gateway;
 pub mod gateway_bonding_sign_payload;
-pub mod settings;
 pub mod unbond_gateway;
 pub mod vesting_bond_gateway;
 pub mod vesting_unbond_gateway;
@@ -19,16 +18,14 @@ pub struct MixnetOperatorsGateway {

 #[derive(Debug, Subcommand)]
 pub enum MixnetOperatorsGatewayCommands {
-    /// Manage your gateway settings stored in the directory
-    Settings(settings::MixnetOperatorsGatewaySettings),
    /// Bond to a gateway
    Bond(bond_gateway::Args),
-    /// Unbond from a gateway
-    Unbond(unbond_gateway::Args),
+    /// Unbound from a gateway
+    Unbound(unbond_gateway::Args),
    /// Bond to a gateway with locked tokens
    VestingBond(vesting_bond_gateway::Args),
-    /// Unbond from a gateway (when originally using locked tokens)
-    VestingUnbond(vesting_unbond_gateway::Args),
+    /// Unbound from a gateway (when originally using locked tokens)
+    VestingUnbound(vesting_unbond_gateway::Args),
    /// Create base58-encoded payload required for producing valid bonding signature.
    CreateGatewayBondingSignPayload(gateway_bonding_sign_payload::Args),
 }
@@ -1,22 +0,0 @@
-// Copyright 2023 - Nym Technologies SA <contact@nymtech.net>
-// SPDX-License-Identifier: Apache-2.0
-
-use clap::{Args, Subcommand};
-
-pub mod update_config;
-pub mod vesting_update_config;
-
-#[derive(Debug, Args)]
-#[clap(args_conflicts_with_subcommands = true, subcommand_required = true)]
-pub struct MixnetOperatorsGatewaySettings {
-    #[clap(subcommand)]
-    pub command: MixnetOperatorsGatewaySettingsCommands,
-}
-
-#[derive(Debug, Subcommand)]
-pub enum MixnetOperatorsGatewaySettingsCommands {
-    /// Update gateway configuration
-    UpdateConfig(update_config::Args),
-    /// Update gateway configuration for a gateway bonded with locked tokens
-    VestingUpdateConfig(vesting_update_config::Args),
-}
@@ -1,60 +0,0 @@
-// Copyright 2023 - Nym Technologies SA <contact@nymtech.net>
-// SPDX-License-Identifier: Apache-2.0
-
-use crate::context::SigningClient;
-use clap::Parser;
-use log::info;
-use nym_mixnet_contract_common::GatewayConfigUpdate;
-use validator_client::nyxd::traits::{MixnetQueryClient, MixnetSigningClient};
-
-#[derive(Debug, Parser)]
-pub struct Args {
-    #[clap(long)]
-    pub host: Option<String>,
-
-    #[clap(long)]
-    pub mix_port: Option<u16>,
-
-    #[clap(long)]
-    pub clients_port: Option<u16>,
-
-    #[clap(long)]
-    pub location: Option<String>,
-
-    #[clap(long)]
-    pub version: Option<String>,
-}
-
-pub async fn update_config(args: Args, client: SigningClient) {
-    info!("Update gateway config!");
-
-    let current_details = match client
-        .get_owned_gateway(client.address())
-        .await
-        .expect("failed to query the chain for gateway details")
-        .gateway
-    {
-        Some(details) => details,
-        None => {
-            log::warn!("this operator does not own a gateway to update");
-            return;
-        }
-    };
-
-    let update = GatewayConfigUpdate {
-        host: args.host.unwrap_or(current_details.gateway.host),
-        mix_port: args.mix_port.unwrap_or(current_details.gateway.mix_port),
-        clients_port: args
-            .clients_port
-            .unwrap_or(current_details.gateway.clients_port),
-        location: args.location.unwrap_or(current_details.gateway.location),
-        version: args.version.unwrap_or(current_details.gateway.version),
-    };
-
-    let res = client
-        .update_gateway_config(update, None)
-        .await
-        .expect("updating gateway config");
-
-    info!("gateway config updated: {:?}", res)
-}
@@ -1,61 +0,0 @@
-// Copyright 2023 - Nym Technologies SA <contact@nymtech.net>
-// SPDX-License-Identifier: Apache-2.0
-
-use crate::context::SigningClient;
-use clap::Parser;
-use log::info;
-use nym_mixnet_contract_common::GatewayConfigUpdate;
-use validator_client::nyxd::traits::MixnetQueryClient;
-use validator_client::nyxd::VestingSigningClient;
-
-#[derive(Debug, Parser)]
-pub struct Args {
-    #[clap(long)]
-    pub host: Option<String>,
-
-    #[clap(long)]
-    pub mix_port: Option<u16>,
-
-    #[clap(long)]
-    pub clients_port: Option<u16>,
-
-    #[clap(long)]
-    pub location: Option<String>,
-
-    #[clap(long)]
-    pub version: Option<String>,
-}
-
-pub async fn vesting_update_config(client: SigningClient, args: Args) {
-    info!("Update vesting gateway config!");
-
-    let current_details = match client
-        .get_owned_gateway(client.address())
-        .await
-        .expect("failed to query the chain for gateway details")
-        .gateway
-    {
-        Some(details) => details,
-        None => {
-            log::warn!("this operator does not own a gateway to update");
-            return;
-        }
-    };
-
-    let update = GatewayConfigUpdate {
-        host: args.host.unwrap_or(current_details.gateway.host),
-        mix_port: args.mix_port.unwrap_or(current_details.gateway.mix_port),
-        clients_port: args
-            .clients_port
-            .unwrap_or(current_details.gateway.clients_port),
-        location: args.location.unwrap_or(current_details.gateway.location),
-        version: args.version.unwrap_or(current_details.gateway.version),
-    };
-
-    let res = client
-        .vesting_update_gateway_config(update, None)
-        .await
-        .expect("updating vesting gateway config");
-
-    info!("gateway config updated: {:?}", res)
-}
@@ -29,12 +29,12 @@ pub enum MixnetOperatorsMixnodeCommands {
    Settings(settings::MixnetOperatorsMixnodeSettings),
    /// Bond to a mixnode
    Bond(bond_mixnode::Args),
-    /// Unbond from a mixnode
-    Unbond(unbond_mixnode::Args),
+    /// Unbound from a mixnode
+    Unbound(unbond_mixnode::Args),
    /// Bond to a mixnode with locked tokens
    BondVesting(vesting_bond_mixnode::Args),
-    /// Unbond from a mixnode (when originally using locked tokens)
-    UnbondVesting(vesting_unbond_mixnode::Args),
+    /// Unbound from a mixnode (when originally using locked tokens)
+    UnboundVesting(vesting_unbond_mixnode::Args),
    /// Create base58-encoded payload required for producing valid bonding signature.
    CreateMixnodeBondingSignPayload(mixnode_bonding_sign_payload::Args),
 }
@@ -21,7 +21,7 @@ pub const TOTAL_DEALINGS: usize = 2 + 2 + 1;
 #[derive(Serialize, Deserialize, Default, Clone, Debug, PartialEq, Eq, Ord, PartialOrd)]
 pub struct InitialReplacementData {
    pub initial_dealers: Vec<Addr>,
-    pub initial_height: u64,
+    pub initial_height: Option<u64>,
 }

 #[derive(
@@ -1,7 +1,6 @@
 // Copyright 2022 - Nym Technologies SA <contact@nymtech.net>
 // SPDX-License-Identifier: Apache-2.0

-use crate::gateway::GatewayConfigUpdate;
 use crate::mixnode::{MixNodeConfigUpdate, MixNodeCostParams};
 use crate::reward_params::{IntervalRewardParams, IntervalRewardingParamsUpdate};
 use crate::rewarding::RewardDistribution;
@@ -43,7 +42,6 @@ pub enum MixnetEventType {
    ReconcilePendingEvents,
    PendingIntervalConfigUpdate,
    IntervalConfigUpdate,
-    GatewayConfigUpdate,
 }

 impl From<MixnetEventType> for String {
@@ -88,7 +86,6 @@ impl ToString for MixnetEventType {
            MixnetEventType::PendingIntervalConfigUpdate => "pending_interval_config_update",
            MixnetEventType::IntervalConfigUpdate => "interval_config_update",
            MixnetEventType::DelegationOnUnbonding => "delegation_on_unbonding_node",
-            MixnetEventType::GatewayConfigUpdate => "gateway_config_update",
        };

        format!("{EVENT_VERSION_PREFIX}{event_name}")
@@ -125,7 +122,6 @@ pub const OLD_REWARDING_VALIDATOR_ADDRESS_KEY: &str = "old_rewarding_validator_a
 pub const NEW_REWARDING_VALIDATOR_ADDRESS_KEY: &str = "new_rewarding_validator_address";

 pub const UPDATED_MIXNODE_CONFIG_KEY: &str = "updated_mixnode_config";
-pub const UPDATED_GATEWAY_CONFIG_KEY: &str = "updated_gateway_config";
 pub const UPDATED_MIXNODE_COST_PARAMS_KEY: &str = "updated_mixnode_cost_params";

 // rewarding
@@ -386,17 +382,6 @@ pub fn new_mixnode_config_update_event(
        .add_attribute(UPDATED_MIXNODE_CONFIG_KEY, update.to_inline_json())
 }

-pub fn new_gateway_config_update_event(
-    owner: &Addr,
-    proxy: &Option<Addr>,
-    update: &GatewayConfigUpdate,
-) -> Event {
-    Event::new(MixnetEventType::GatewayConfigUpdate)
-        .add_attribute(OWNER_KEY, owner)
-        .add_optional_attribute(PROXY_KEY, proxy.as_ref())
-        .add_attribute(UPDATED_GATEWAY_CONFIG_KEY, update.to_inline_json())
-}
-
 pub fn new_mixnode_pending_cost_params_update_event(
    mix_id: MixId,
    owner: &Addr,
@@ -112,26 +112,6 @@ impl Display for GatewayBond {
    }
 }

-#[cfg_attr(feature = "generate-ts", derive(ts_rs::TS))]
-#[cfg_attr(
-    feature = "generate-ts",
-    ts(export_to = "ts-packages/types/src/types/rust/GatewayConfigUpdate.ts")
-)]
-#[derive(Clone, Debug, Deserialize, PartialEq, Eq, Serialize, JsonSchema)]
-pub struct GatewayConfigUpdate {
-    pub host: String,
-    pub mix_port: u16,
-    pub clients_port: u16,
-    pub location: String,
-    pub version: String,
-}
-
-impl GatewayConfigUpdate {
-    pub fn to_inline_json(&self) -> String {
-        serde_json::to_string(self).unwrap_or_else(|_| "serialisation failure".into())
-    }
-}
-
 #[derive(Clone, Debug, Deserialize, PartialEq, Serialize, JsonSchema)]
 pub struct PagedGatewayResponse {
    pub nodes: Vec<GatewayBond>,
@@ -27,8 +27,7 @@ pub use delegation::{
    PagedMixNodeDelegationsResponse,
 };
 pub use gateway::{
-    Gateway, GatewayBond, GatewayBondResponse, GatewayConfigUpdate, GatewayOwnershipResponse,
-    PagedGatewayResponse,
+    Gateway, GatewayBond, GatewayBondResponse, GatewayOwnershipResponse, PagedGatewayResponse,
 };
 pub use interval::{
    CurrentIntervalResponse, EpochState, EpochStatus, Interval, NumberOfPendingEventsResponse,
@@ -3,7 +3,6 @@

 use crate::delegation::OwnerProxySubKey;
 use crate::error::MixnetContractError;
-use crate::gateway::GatewayConfigUpdate;
 use crate::helpers::IntoBaseDecimal;
 use crate::mixnode::{MixNodeConfigUpdate, MixNodeCostParams};
 use crate::reward_params::{
@@ -200,13 +199,6 @@ pub enum ExecuteMsg {
    UnbondGatewayOnBehalf {
        owner: String,
    },
-    UpdateGatewayConfig {
-        new_config: GatewayConfigUpdate,
-    },
-    UpdateGatewayConfigOnBehalf {
-        new_config: GatewayConfigUpdate,
-        owner: String,
-    },

    // delegation-related:
    DelegateToMixnode {
@@ -321,10 +313,6 @@ impl ExecuteMsg {
            }
            ExecuteMsg::UnbondGateway { .. } => "unbonding gateway".into(),
            ExecuteMsg::UnbondGatewayOnBehalf { .. } => "unbonding gateway on behalf".into(),
-            ExecuteMsg::UpdateGatewayConfig { .. } => "updating gateway configuration".into(),
-            ExecuteMsg::UpdateGatewayConfigOnBehalf { .. } => {
-                "updating gateway configuration on behalf".into()
-            }
            ExecuteMsg::DelegateToMixnode { mix_id } => format!("delegating to mixnode {mix_id}"),
            ExecuteMsg::DelegateToMixnodeOnBehalf { mix_id, .. } => {
                format!("delegating to mixnode {mix_id} on behalf")
@@ -17,7 +17,6 @@ pub const VESTING_MIXNODE_BONDING_EVENT_TYPE: &str = "vesting_mixnode_bonding";
 pub const VESTING_PLEDGE_MORE_EVENT_TYPE: &str = "vesting_pledge_more";
 pub const VESTING_MIXNODE_UNBONDING_EVENT_TYPE: &str = "vesting_mixnode_unbonding";
 pub const VESTING_UPDATE_MIXNODE_CONFIG_EVENT_TYPE: &str = "vesting_update_mixnode_config";
-pub const VESTING_UPDATE_GATEWAY_CONFIG_EVENT_TYPE: &str = "vesting_update_gateway_config";
 pub const VESTING_UPDATE_MIXNODE_COST_PARAMS_EVENT_TYPE: &str =
    "vesting_update_mixnode_cost_params";

@@ -122,10 +121,6 @@ pub fn new_vesting_update_mixnode_config_event() -> Event {
    Event::new(VESTING_UPDATE_MIXNODE_CONFIG_EVENT_TYPE)
 }

-pub fn new_vesting_update_gateway_config_event() -> Event {
-    Event::new(VESTING_UPDATE_GATEWAY_CONFIG_EVENT_TYPE)
-}
-
 pub fn new_vesting_update_mixnode_cost_params_event() -> Event {
    Event::new(VESTING_UPDATE_MIXNODE_COST_PARAMS_EVENT_TYPE)
 }
@@ -1,7 +1,6 @@
 use contracts_common::signing::MessageSignature;
 use cosmwasm_std::{Coin, Timestamp};
 use mixnet_contract_common::{
-    gateway::GatewayConfigUpdate,
    mixnode::{MixNodeConfigUpdate, MixNodeCostParams},
    Gateway, IdentityKey, MixId, MixNode,
 };
@@ -141,9 +140,6 @@ pub enum ExecuteMsg {
        owner: String,
        amount: Coin,
    },
-    UpdateGatewayConfig {
-        new_config: GatewayConfigUpdate,
-    },
    TransferOwnership {
        to_address: String,
    },
@@ -183,7 +179,6 @@ impl ExecuteMsg {
            ExecuteMsg::BondGateway { .. } => "VestingExecuteMsg::BondGateway",
            ExecuteMsg::UnbondGateway { .. } => "VestingExecuteMsg::UnbondGateway",
            ExecuteMsg::TrackUnbondGateway { .. } => "VestingExecuteMsg::TrackUnbondGateway",
-            ExecuteMsg::UpdateGatewayConfig { .. } => "VestingExecuteMsg::UpdateGatewayConfig",
            ExecuteMsg::TransferOwnership { .. } => "VestingExecuteMsg::TransferOwnership",
            ExecuteMsg::UpdateStakingAddress { .. } => "VestingExecuteMsg::UpdateStakingAddress",
            ExecuteMsg::UpdateLockedPledgeCap { .. } => "VestingExecuteMsg::UpdateLockedPledgeCap",
@@ -13,20 +13,12 @@ humantime-serde = "1.0"
 log = { workspace = true }
 rand = "0.8"
 serde = { version = "1.0", features = ["derive"] }
-tokio = { version = "1.24.1", features = [
-    "time",
-    "macros",
-    "rt",
-    "net",
-    "io-util",
-] }
+tokio = { version = "1.24.1", features = ["time", "macros", "rt", "net", "io-util"] }
 tokio-util = { version = "0.7.4", features = ["codec"] }
 url = "2.2"
 thiserror = "1.0.37"
-chitchat = "0.5"
-serde_json = "1"

-nym-crypto = { path = "../crypto" }
+nym-crypto =  { path = "../crypto" }
 nym-network-defaults = { path = "../network-defaults" }
 nym-sphinx-acknowledgements = { path = "../nymsphinx/acknowledgements" }
 nym-sphinx-addressing = { path = "../nymsphinx/addressing" }
@@ -35,7 +27,5 @@ nym-sphinx-framing = { path = "../nymsphinx/framing" }
 nym-sphinx-params = { path = "../nymsphinx/params" }
 nym-sphinx-types = { path = "../nymsphinx/types" }
 nym-task = { path = "../task" }
-validator-client = { path = "../client-libs/validator-client", features = [
-    "nyxd-client",
-] }
+validator-client = { path = "../client-libs/validator-client", features = ["nyxd-client"]}
 nym-bin-common = { path = "../bin-common" }
@@ -3,7 +3,6 @@

 use crate::verloc::listener::PacketListener;
 use crate::verloc::sender::{PacketSender, TestedNode};
-use chitchat::Chitchat;
 use futures::stream::FuturesUnordered;
 use futures::StreamExt;
 use log::*;
@@ -17,7 +16,6 @@ use std::net::SocketAddr;
 use std::net::ToSocketAddrs;
 use std::sync::Arc;
 use std::time::Duration;
-use tokio::sync::Mutex;
 use tokio::task::JoinHandle;
 use tokio::time::sleep;
 use url::Url;
@@ -295,7 +293,7 @@ impl VerlocMeasurer {
        MeasurementOutcome::Done
    }

-    pub async fn run(&mut self, chitchat_handle: Arc<Mutex<Chitchat>>) {
+    pub async fn run(&mut self) {
        self.start_listening();

        while !self.shutdown_listener.is_shutdown() {
@@ -347,17 +345,6 @@ impl VerlocMeasurer {
                })
                .collect::<Vec<_>>();

-            let mut chitchat_guard = chitchat_handle.lock().await;
-            let cc_state = chitchat_guard.self_node_state();
-            let tested_nodes_cc = tested_nodes
-                .iter()
-                .map(|node| node.identity.to_string())
-                .collect::<Vec<String>>();
-            cc_state.set(
-                "tested_nodes",
-                serde_json::to_value(tested_nodes_cc).expect("Could not serialize"),
-            );
-
            // on start of each run remove old results
            self.results.reset_results(tested_nodes.len()).await;

@@ -377,7 +377,6 @@ pub const DEFAULT_CLIENT_LISTENING_PORT: u16 = 9000;
 // 'MIXNODE'
 pub const DEFAULT_VERLOC_LISTENING_PORT: u16 = 1790;
 pub const DEFAULT_HTTP_API_LISTENING_PORT: u16 = 8000;
-pub const DEFAULT_GOSSIP_PORT: u16 = 10000;

 // 'CLIENT'
 pub const DEFAULT_WEBSOCKET_LISTENING_PORT: u16 = 1977;
@@ -0,0 +1 @@
+Cargo.lock
@@ -22,7 +22,7 @@ fn verify_dealer(deps: DepsMut<'_>, dealer: &Addr, resharing: bool) -> Result<()
    let state = STATE.load(deps.storage)?;

    let height = if resharing {
-        Some(INITIAL_REPLACEMENT_DATA.load(deps.storage)?.initial_height)
+        INITIAL_REPLACEMENT_DATA.load(deps.storage)?.initial_height
    } else {
        None
    };
@@ -105,7 +105,7 @@ pub(crate) mod tests {
                    deps.as_mut().storage,
                    &InitialReplacementData {
                        initial_dealers: vec![details1.address, details2.address, details3.address],
-                        initial_height: 1,
+                        initial_height: Some(1),
                    },
                )
                .unwrap();
@@ -126,7 +126,7 @@ pub(crate) mod tests {

            INITIAL_REPLACEMENT_DATA
                .update::<_, ContractError>(deps.as_mut().storage, |mut data| {
-                    data.initial_height = 2;
+                    data.initial_height = Some(2);
                    Ok(data)
                })
                .unwrap();
@@ -110,7 +110,7 @@ pub(crate) mod tests {
                deps.as_mut().storage,
                &InitialReplacementData {
                    initial_dealers: vec![],
-                    initial_height: 1,
+                    initial_height: None,
                },
            )
            .unwrap();
@@ -7,7 +7,6 @@ use crate::epoch_state::storage::{CURRENT_EPOCH, INITIAL_REPLACEMENT_DATA, THRES
 use crate::epoch_state::utils::check_epoch_state;
 use crate::error::ContractError;
 use crate::state::STATE;
-use crate::verification_key_shares::storage::verified_dealers;
 use cosmwasm_std::{Addr, Deps, DepsMut, Env, Order, Response, Storage};
 use nym_coconut_dkg_common::types::{Epoch, EpochState, InitialReplacementData};

@@ -20,13 +19,7 @@ fn reset_epoch_state(storage: &mut dyn Storage) -> Result<(), ContractError> {
    for dealer_addr in dealers {
        let details = current_dealers().load(storage, &dealer_addr)?;
        for dealings in DEALINGS_BYTES {
-            let dealing_keys: Vec<_> = dealings
-                .keys(storage, None, None, Order::Ascending)
-                .flatten()
-                .collect();
-            for key in dealing_keys {
-                dealings.remove(storage, &key);
-            }
+            dealings.remove(storage, &details.address);
        }
        current_dealers().remove(storage, &dealer_addr)?;
        past_dealers().save(storage, &dealer_addr, &details)?;
@@ -53,9 +46,15 @@ fn dealers_still_active(
 }

 fn dealers_eq_members(deps: &DepsMut<'_>) -> Result<bool, ContractError> {
-    let verified_dealers = verified_dealers(deps.storage)?;
-    let all_dealers = verified_dealers.len();
-    let dealers_still_active = dealers_still_active(&deps.as_ref(), verified_dealers.into_iter())?;
+    let dealers_still_active = dealers_still_active(
+        &deps.as_ref(),
+        current_dealers()
+            .keys(deps.storage, None, None, Order::Ascending)
+            .flatten(),
+    )?;
+    let all_dealers = current_dealers()
+        .keys(deps.storage, None, None, Order::Ascending)
+        .count();
    let group_members = STATE
        .load(deps.storage)?
        .group_addr
@@ -67,11 +66,7 @@ fn dealers_eq_members(deps: &DepsMut<'_>) -> Result<bool, ContractError> {

 fn replacement_threshold_surpassed(deps: &DepsMut<'_>) -> Result<bool, ContractError> {
    let threshold = THRESHOLD.load(deps.storage)? as usize;
-    let initial_dealers = verified_dealers(deps.storage)?;
-    if initial_dealers.is_empty() {
-        // possibly failed DKG, just reset and start again
-        return Ok(true);
-    }
+    let initial_dealers = INITIAL_REPLACEMENT_DATA.load(deps.storage)?.initial_dealers;
    let initial_dealer_count = initial_dealers.len();
    let replacement_threshold = threshold - (initial_dealers.len() + 2 - 1) / 2 + 1;
    let removed_dealer_count =
@@ -95,23 +90,24 @@ pub(crate) fn advance_epoch_state(deps: DepsMut<'_>, env: Env) -> Result<Respons
    let next_epoch = if let Some(state) = current_epoch.state.next() {
        // We are during DKG process
        let mut new_state = state;
-        if let EpochState::DealingExchange { .. } = state {
+        if let EpochState::DealingExchange { resharing } = state {
            let current_dealers = current_dealers()
                .keys(deps.storage, None, None, Order::Ascending)
                .collect::<Result<Vec<Addr>, _>>()?;
-            let group_members =
-                STATE
-                    .load(deps.storage)?
-                    .group_addr
-                    .list_members(&deps.querier, None, None)?;
-            if current_dealers.len() < group_members.len() {
-                // If not all group members registered yet, we just stay in the same state until
-                // they either register or they get kicked out of the group
+            if current_dealers.is_empty() {
+                // If no dealer registered yet, we just stay in the same state until there's at least one
                new_state = current_epoch.state;
            } else {
                // note: ceiling in integer division can be achieved via q = (x + y - 1) / y;
                let threshold = (2 * current_dealers.len() as u64 + 3 - 1) / 3;
                THRESHOLD.save(deps.storage, &threshold)?;
+                if !resharing {
+                    let replacement_data = InitialReplacementData {
+                        initial_dealers: current_dealers,
+                        initial_height: None,
+                    };
+                    INITIAL_REPLACEMENT_DATA.save(deps.storage, &replacement_data)?;
+                }
            }
        };
        Epoch::new(
@@ -133,23 +129,13 @@ pub(crate) fn advance_epoch_state(deps: DepsMut<'_>, env: Env) -> Result<Respons
        // Dealer set changed, we need to redo DKG...
        let state = if replacement_threshold_surpassed(&deps)? {
            // ... in reset mode
-            INITIAL_REPLACEMENT_DATA.remove(deps.storage);
            EpochState::default()
        } else {
            // ... in reshare mode
-            if INITIAL_REPLACEMENT_DATA.may_load(deps.storage)?.is_some() {
-                INITIAL_REPLACEMENT_DATA.update::<_, ContractError>(deps.storage, |mut data| {
-                    data.initial_height = env.block.height;
-                    Ok(data)
-                })?;
-            } else {
-                let replacement_data = InitialReplacementData {
-                    initial_dealers: verified_dealers(deps.storage)?,
-                    initial_height: env.block.height,
-                };
-                INITIAL_REPLACEMENT_DATA.save(deps.storage, &replacement_data)?;
-            }
-
+            INITIAL_REPLACEMENT_DATA.update::<_, ContractError>(deps.storage, |mut data| {
+                data.initial_height = Some(env.block.height);
+                Ok(data)
+            })?;
            EpochState::PublicKeySubmission { resharing: true }
        };
        reset_epoch_state(deps.storage)?;
@@ -172,8 +158,10 @@ pub(crate) fn try_surpassed_threshold(
    check_epoch_state(deps.storage, EpochState::InProgress)?;

    let threshold = THRESHOLD.load(deps.storage)?;
-    let dealers = verified_dealers(deps.storage)?;
-    if dealers_still_active(&deps.as_ref(), dealers.into_iter())? < threshold as usize {
+    let dealers = current_dealers()
+        .keys(deps.storage, None, None, Order::Ascending)
+        .flatten();
+    if dealers_still_active(&deps.as_ref(), dealers)? < threshold as usize {
        reset_epoch_state(deps.storage)?;
        CURRENT_EPOCH.update::<_, ContractError>(deps.storage, |epoch| {
            Ok(Epoch::new(
@@ -192,9 +180,8 @@ pub(crate) fn try_surpassed_threshold(
 pub(crate) mod tests {
    use super::*;
    use crate::error::ContractError::EarlyEpochStateAdvancement;
-    use crate::support::tests::fixtures::{dealer_details_fixture, vk_share_fixture};
+    use crate::support::tests::fixtures::dealer_details_fixture;
    use crate::support::tests::helpers::{init_contract, GROUP_MEMBERS};
-    use crate::verification_key_shares::storage::vk_shares;
    use cosmwasm_std::testing::mock_env;
    use cosmwasm_std::Addr;
    use cw4::Member;
@@ -217,15 +204,11 @@ pub(crate) mod tests {

            for n in [10, 25, 50, 100] {
                let dealers: Vec<_> = (0..n).map(dealer_details_fixture).collect();
-                let shares: Vec<_> = (0..n).map(|idx| vk_share_fixture(&format!("owner{}", idx), 0)).collect();
                let initial_dealers = dealers.iter().map(|d| d.address.clone()).collect();
                let data = InitialReplacementData {
                    initial_dealers,
-                    initial_height: 1,
+                    initial_height: None,
                };
-                for share in shares {
-                    vk_shares().save(deps.as_mut().storage, (&share.owner, 0), &share).unwrap();
-                }
                for f in [two_thirds, three_fourths, ninty_pc] {
                    let threshold = f(n);
                    THRESHOLD.save(deps.as_mut().storage, &threshold).unwrap();
@@ -264,39 +247,39 @@ pub(crate) mod tests {

            assert!(dealers_eq_members(&deps.as_mut()).unwrap());

-            let share = vk_share_fixture("owner2", 0);
-            let different_share = vk_share_fixture("owner4", 0);
-            vk_shares()
-                .save(deps.as_mut().storage, (&share.owner, 0), &share)
+            let details = dealer_details_fixture(1);
+            let different_details = dealer_details_fixture(2);
+            current_dealers()
+                .save(deps.as_mut().storage, &details.address, &details)
                .unwrap();
            assert!(!dealers_eq_members(&deps.as_mut()).unwrap());

-            vk_shares()
-                .remove(deps.as_mut().storage, (&share.owner, 0))
+            current_dealers()
+                .remove(deps.as_mut().storage, &details.address)
                .unwrap();
            GROUP_MEMBERS.lock().unwrap().push((
                Member {
-                    addr: "owner2".to_string(),
+                    addr: "owner1".to_string(),
                    weight: 10,
                },
                1,
            ));
            assert!(!dealers_eq_members(&deps.as_mut()).unwrap());

-            vk_shares()
+            current_dealers()
                .save(
                    deps.as_mut().storage,
-                    (&different_share.owner, 0),
-                    &different_share,
+                    &different_details.address,
+                    &different_details,
                )
                .unwrap();
            assert!(!dealers_eq_members(&deps.as_mut()).unwrap());

-            vk_shares()
-                .remove(deps.as_mut().storage, (&different_share.owner, 0))
+            current_dealers()
+                .remove(deps.as_mut().storage, &different_details.address)
                .unwrap();
-            vk_shares()
-                .save(deps.as_mut().storage, (&share.owner, 0), &share)
+            current_dealers()
+                .save(deps.as_mut().storage, &details.address, &details)
                .unwrap();
            assert!(dealers_eq_members(&deps.as_mut()).unwrap());
        }
@@ -424,12 +407,6 @@ pub(crate) mod tests {
            );

            // setup dealer details
-            let all_shares: [_; 4] = std::array::from_fn(|i| vk_share_fixture(&format!("owner{}", i + 1), 0));
-            for share in all_shares.iter() {
-                vk_shares()
-                    .save(deps.as_mut().storage, (&share.owner, 0), share)
-                    .unwrap();
-            }
            let all_details: [_; 4] = std::array::from_fn(|i| dealer_details_fixture(i as u64 + 1));
            for details in all_details.iter() {
                current_dealers()
@@ -454,6 +431,12 @@ pub(crate) mod tests {
                    .time
                    .plus_seconds(epoch.time_configuration.dealing_exchange_time_secs)
            );
+            let replacement_data = INITIAL_REPLACEMENT_DATA.load(&deps.storage).unwrap();
+            let expected_replacement_data = InitialReplacementData {
+                initial_dealers: all_details.iter().map(|d| d.address.clone()).collect(),
+                initial_height: None,
+            };
+            assert_eq!(replacement_data, expected_replacement_data);

            env.block.time = env
                .block
@@ -605,14 +588,8 @@ pub(crate) mod tests {
            );
            assert_eq!(curr_epoch, expected_epoch);
            assert!(THRESHOLD.may_load(&deps.storage).unwrap().is_none());
-            let replacement_data = INITIAL_REPLACEMENT_DATA.load(&deps.storage).unwrap();
-            let expected_replacement_data = InitialReplacementData {
-                initial_dealers: all_details.iter().map(|d| d.address.clone()).collect(),
-                initial_height: 12345,
-            };
-            assert_eq!(replacement_data, expected_replacement_data);

-            let all_details: [_; 4] = std::array::from_fn(|i| dealer_details_fixture(i as u64 + 2));
+            let all_details: [_; 2] = std::array::from_fn(|i| dealer_details_fixture(i as u64 + 2));
            for details in all_details.iter() {
                past_dealers().remove(deps.as_mut().storage, &details.address).unwrap();
                current_dealers()
@@ -630,17 +607,6 @@ pub(crate) mod tests {
                advance_epoch_state(deps.as_mut(), env.clone()).unwrap();
            }

-            let all_shares: [_; 4] = std::array::from_fn(|i| {
-                let mut share = vk_share_fixture(&format!("owner{}", i + 1), 1);
-                share.verified = i % 2 == 0;
-                share
-                });
-            for share in all_shares.iter() {
-                vk_shares()
-                    .save(deps.as_mut().storage, (&share.owner, 0), share)
-                    .unwrap();
-            }
-
            // Group changed even more, surpassing threshold, so re-run dkg in reset mode
            *GROUP_MEMBERS.lock().unwrap().last_mut().unwrap() = (
                Member {
@@ -657,7 +623,7 @@ pub(crate) mod tests {
            advance_epoch_state(deps.as_mut(), env.clone()).unwrap();
            let curr_epoch = CURRENT_EPOCH.load(deps.as_mut().storage).unwrap();
            let expected_epoch = Epoch::new(
-                EpochState::PublicKeySubmission { resharing: true },
+                EpochState::PublicKeySubmission { resharing: false },
                prev_epoch.epoch_id + 1,
                prev_epoch.time_configuration,
                env.block.time,
@@ -706,25 +672,12 @@ pub(crate) mod tests {
                }
            );

-
-            let all_shares: [_; 3] = std::array::from_fn(|i| vk_share_fixture(&format!("owner{}", i + 1), 0));
-            for share in all_shares.iter() {
-                vk_shares()
-                    .save(deps.as_mut().storage, (&share.owner, 0), share)
-                    .unwrap();
-            }
            let all_details: [_; 3] = std::array::from_fn(|i| dealer_details_fixture(i as u64 + 1));
            for details in all_details.iter() {
                current_dealers()
                    .save(deps.as_mut().storage, &details.address, details)
                    .unwrap();
            }
-            let all_shares: [_; 3] = std::array::from_fn(|i| vk_share_fixture(&format!("owner{}", i + 1), 0));
-            for share in all_shares.iter() {
-                vk_shares()
-                    .save(deps.as_mut().storage, (&share.owner, share.epoch_id), share)
-                    .unwrap();
-            }

            for times in [
                time_configuration.public_key_submission_time_secs,
@@ -4,9 +4,7 @@
 // SPDX-License-Identifier: Apache-2.0

 use crate::constants::{VK_SHARES_EPOCH_ID_IDX_NAMESPACE, VK_SHARES_PK_NAMESPACE};
-use crate::epoch_state::storage::CURRENT_EPOCH;
-use crate::error::ContractError;
-use cosmwasm_std::{Addr, Order, Storage};
+use cosmwasm_std::Addr;
 use cw_storage_plus::{Index, IndexList, IndexedMap, MultiIndex};
 use nym_coconut_dkg_common::types::EpochId;
 use nym_coconut_dkg_common::verification_key::ContractVKShare;
@@ -37,21 +35,3 @@ pub(crate) fn vk_shares<'a>() -> IndexedMap<'a, VKShareKey<'a>, ContractVKShare,
    };
    IndexedMap::new(VK_SHARES_PK_NAMESPACE, indexes)
 }
-
-pub(crate) fn verified_dealers(storage: &dyn Storage) -> Result<Vec<Addr>, ContractError> {
-    let epoch_id = CURRENT_EPOCH.load(storage)?.epoch_id;
-    Ok(vk_shares()
-        .idx
-        .epoch_id
-        .prefix(epoch_id)
-        .range(storage, None, None, Order::Ascending)
-        .flatten()
-        .filter_map(|(_, share)| {
-            if share.verified {
-                Some(share.owner)
-            } else {
-                None
-            }
-        })
-        .collect())
-}
@@ -318,14 +318,6 @@ pub fn execute(
        ExecuteMsg::UnbondGatewayOnBehalf { owner } => {
            crate::gateways::transactions::try_remove_gateway_on_behalf(deps, info, owner)
        }
-        ExecuteMsg::UpdateGatewayConfig { new_config } => {
-            crate::gateways::transactions::try_update_gateway_config(deps, info, new_config)
-        }
-        ExecuteMsg::UpdateGatewayConfigOnBehalf { new_config, owner } => {
-            crate::gateways::transactions::try_update_gateway_config_on_behalf(
-                deps, info, new_config, owner,
-            )
-        }

        // delegation-related:
        ExecuteMsg::DelegateToMixnode { mix_id } => {
@@ -1,20 +0,0 @@
-// Copyright 2023 - Nym Technologies SA <contact@nymtech.net>
-// SPDX-License-Identifier: Apache-2.0
-
-use super::storage;
-use cosmwasm_std::{Addr, Storage};
-use mixnet_contract_common::{error::MixnetContractError, GatewayBond};
-
-pub(crate) fn must_get_gateway_bond_by_owner(
-    store: &dyn Storage,
-    owner: &Addr,
-) -> Result<GatewayBond, MixnetContractError> {
-    Ok(storage::gateways()
-        .idx
-        .owner
-        .item(store, owner.clone())?
-        .ok_or(MixnetContractError::NoAssociatedGatewayBond {
-            owner: owner.clone(),
-        })?
-        .1)
-}
@@ -1,7 +1,6 @@
 // Copyright 2021 - Nym Technologies SA <contact@nymtech.net>
 // SPDX-License-Identifier: Apache-2.0

-pub mod helpers;
 pub mod queries;
 pub mod signature_helpers;
 pub mod storage;
@@ -1,20 +1,16 @@
 // Copyright 2021-2022 - Nym Technologies SA <contact@nymtech.net>
 // SPDX-License-Identifier: Apache-2.0

-use super::helpers::must_get_gateway_bond_by_owner;
 use super::storage;
 use crate::gateways::signature_helpers::verify_gateway_bonding_signature;
 use crate::mixnet_contract_settings::storage as mixnet_params_storage;
 use crate::signing::storage as signing_storage;
 use crate::support::helpers::{
-    ensure_no_existing_bond, ensure_proxy_match, ensure_sent_by_vesting_contract, validate_pledge,
+    ensure_no_existing_bond, ensure_sent_by_vesting_contract, validate_pledge,
 };
 use cosmwasm_std::{wasm_execute, Addr, BankMsg, Coin, DepsMut, Env, MessageInfo, Response};
 use mixnet_contract_common::error::MixnetContractError;
-use mixnet_contract_common::events::{
-    new_gateway_bonding_event, new_gateway_config_update_event, new_gateway_unbonding_event,
-};
-use mixnet_contract_common::gateway::GatewayConfigUpdate;
+use mixnet_contract_common::events::{new_gateway_bonding_event, new_gateway_unbonding_event};
 use mixnet_contract_common::{Gateway, GatewayBond};
 use nym_contracts_common::signing::MessageSignature;
 use vesting_contract_common::messages::ExecuteMsg as VestingContractExecuteMsg;
@@ -193,56 +189,6 @@ pub(crate) fn _try_remove_gateway(
    )))
 }

-pub(crate) fn try_update_gateway_config(
-    deps: DepsMut<'_>,
-    info: MessageInfo,
-    new_config: GatewayConfigUpdate,
-) -> Result<Response, MixnetContractError> {
-    let owner = info.sender;
-    _try_update_gateway_config(deps, new_config, owner, None)
-}
-
-pub(crate) fn try_update_gateway_config_on_behalf(
-    deps: DepsMut,
-    info: MessageInfo,
-    new_config: GatewayConfigUpdate,
-    owner: String,
-) -> Result<Response, MixnetContractError> {
-    ensure_sent_by_vesting_contract(&info, deps.storage)?;
-
-    let owner = deps.api.addr_validate(&owner)?;
-    let proxy = info.sender;
-    _try_update_gateway_config(deps, new_config, owner, Some(proxy))
-}
-
-pub(crate) fn _try_update_gateway_config(
-    deps: DepsMut,
-    new_config: GatewayConfigUpdate,
-    owner: Addr,
-    proxy: Option<Addr>,
-) -> Result<Response, MixnetContractError> {
-    let existing_bond = must_get_gateway_bond_by_owner(deps.storage, &owner)?;
-    ensure_proxy_match(&proxy, &existing_bond.proxy)?;
-
-    let cfg_update_event = new_gateway_config_update_event(&owner, &proxy, &new_config);
-
-    let mut updated_bond = existing_bond.clone();
-    updated_bond.gateway.host = new_config.host;
-    updated_bond.gateway.mix_port = new_config.mix_port;
-    updated_bond.gateway.clients_port = new_config.clients_port;
-    updated_bond.gateway.location = new_config.location;
-    updated_bond.gateway.version = new_config.version;
-
-    storage::gateways().replace(
-        deps.storage,
-        existing_bond.identity(),
-        Some(&updated_bond),
-        Some(&existing_bond),
-    )?;
-
-    Ok(Response::new().add_event(cfg_update_event))
-}
-
 #[cfg(test)]
 pub mod tests {
    use super::*;
@@ -250,7 +196,6 @@ pub mod tests {
    use crate::gateways::queries;
    use crate::gateways::transactions::{
        try_add_gateway, try_add_gateway_on_behalf, try_remove_gateway_on_behalf,
-        try_update_gateway_config, try_update_gateway_config_on_behalf,
    };
    use crate::interval::pending_events;
    use crate::mixnet_contract_settings::storage::minimum_gateway_pledge;
@@ -262,7 +207,6 @@ pub mod tests {
    use cosmwasm_std::{Addr, BankMsg, Response, Uint128};
    use mixnet_contract_common::error::MixnetContractError;
    use mixnet_contract_common::events::new_gateway_unbonding_event;
-    use mixnet_contract_common::gateway::GatewayConfigUpdate;
    use mixnet_contract_common::ExecuteMsg;

    #[test]
@@ -541,94 +485,4 @@ pub mod tests {
            }
        )
    }
-
-    #[test]
-    fn update_gateway_config() {
-        let mut test = TestSetup::new();
-
-        let owner = "alice";
-        let info = mock_info(owner, &[]);
-        let update = GatewayConfigUpdate {
-            host: "1.1.1.1:1234".to_string(),
-            mix_port: 1234,
-            clients_port: 1235,
-            location: "home".to_string(),
-            version: "v1.2.3".to_string(),
-        };
-
-        // try updating a non existing gateway bond
-        let res = try_update_gateway_config(test.deps_mut(), info.clone(), update.clone());
-        assert_eq!(
-            res,
-            Err(MixnetContractError::NoAssociatedGatewayBond {
-                owner: Addr::unchecked(owner)
-            })
-        );
-
-        test.add_dummy_gateway(owner, None);
-        let vesting_contract = test.vesting_contract();
-
-        // attempted to remove on behalf with invalid proxy (current is `None`)
-        let res = try_update_gateway_config_on_behalf(
-            test.deps_mut(),
-            mock_info(vesting_contract.as_ref(), &[]),
-            update.clone(),
-            owner.to_string(),
-        );
-        assert_eq!(
-            res,
-            Err(MixnetContractError::ProxyMismatch {
-                existing: "None".to_string(),
-                incoming: vesting_contract.into_string()
-            })
-        );
-
-        // "normal" update succeeds
-        let res = try_update_gateway_config(test.deps_mut(), info, update.clone());
-        assert!(res.is_ok());
-
-        // and the config has actually been updated
-        let bond =
-            must_get_gateway_bond_by_owner(test.deps().storage, &Addr::unchecked(owner)).unwrap();
-        assert_eq!(bond.gateway.host, update.host);
-        assert_eq!(bond.gateway.mix_port, update.mix_port);
-        assert_eq!(bond.gateway.clients_port, update.clients_port);
-        assert_eq!(bond.gateway.location, update.location);
-        assert_eq!(bond.gateway.version, update.version);
-    }
-
-    #[test]
-    fn updating_gateway_config_with_illegal_proxy() {
-        let mut test = TestSetup::new();
-
-        let illegal_proxy = Addr::unchecked("not-vesting-contract");
-        let vesting_contract = test.vesting_contract();
-
-        let owner = "alice";
-
-        test.add_dummy_gateway_with_illegal_proxy(owner, None, illegal_proxy.clone());
-        let update = GatewayConfigUpdate {
-            host: "1.1.1.1:1234".to_string(),
-            mix_port: 1234,
-            clients_port: 1235,
-            location: "at home".to_string(),
-            version: "v1.2.3".to_string(),
-        };
-
-        let res = try_update_gateway_config_on_behalf(
-            test.deps_mut(),
-            mock_info(illegal_proxy.as_ref(), &[]),
-            update,
-            owner.to_string(),
-        )
-        .unwrap_err();
-
-        assert_eq!(
-            res,
-            MixnetContractError::SenderIsNotVestingContract {
-                received: illegal_proxy,
-                vesting_contract
-            }
-        )
-    }
 }
@@ -295,7 +295,7 @@ pub(crate) fn try_update_mixnode_config(
 }

 pub(crate) fn try_update_mixnode_config_on_behalf(
-    deps: DepsMut<'_>,
+    deps: DepsMut,
    info: MessageInfo,
    new_config: MixNodeConfigUpdate,
    owner: String,
@@ -308,7 +308,7 @@ pub(crate) fn try_update_mixnode_config_on_behalf(
 }

 pub(crate) fn _try_update_mixnode_config(
-    deps: DepsMut<'_>,
+    deps: DepsMut,
    new_config: MixNodeConfigUpdate,
    owner: Addr,
    proxy: Option<Addr>,
@@ -20,7 +20,7 @@ pub mod test_helpers {
        perform_pending_epoch_actions, perform_pending_interval_actions, try_begin_epoch_transition,
    };
    use crate::interval::{pending_events, storage as interval_storage};
-    use crate::mixnet_contract_settings::storage::{self as mixnet_params_storage};
+    use crate::mixnet_contract_settings::storage as mixnet_params_storage;
    use crate::mixnet_contract_settings::storage::{
        minimum_gateway_pledge, minimum_mixnode_pledge, rewarding_denom,
        rewarding_validator_address,
@@ -14,7 +14,6 @@ use cosmwasm_std::{
    QueryResponse, Response, StdError, StdResult, Timestamp, Uint128,
 };
 use cw_storage_plus::Bound;
-use mixnet_contract_common::gateway::GatewayConfigUpdate;
 use mixnet_contract_common::mixnode::{MixNodeConfigUpdate, MixNodeCostParams};
 use mixnet_contract_common::{Gateway, MixId, MixNode};
 use semver::Version;
@@ -220,9 +219,6 @@ pub fn execute(
        ExecuteMsg::TrackUnbondGateway { owner, amount } => {
            try_track_unbond_gateway(&owner, amount, info, deps)
        }
-        ExecuteMsg::UpdateGatewayConfig { new_config } => {
-            try_update_gateway_config(new_config, info, deps)
-        }
        ExecuteMsg::TransferOwnership { to_address } => {
            try_transfer_ownership(to_address, info, deps)
        }
@@ -304,15 +300,6 @@ pub fn try_update_mixnode_config(
    account.try_update_mixnode_config(new_config, deps.storage)
 }

-pub fn try_update_gateway_config(
-    new_config: GatewayConfigUpdate,
-    info: MessageInfo,
-    deps: DepsMut,
-) -> Result<Response, ContractError> {
-    let account = account_from_address(info.sender.as_str(), deps.storage, deps.api)?;
-    account.try_update_gateway_config(new_config, deps.storage)
-}
-
 pub fn try_update_mixnode_cost_params(
    new_costs: MixNodeCostParams,
    info: MessageInfo,
@@ -2,7 +2,6 @@ use crate::errors::ContractError;
 use contracts_common::signing::MessageSignature;
 use cosmwasm_std::{Coin, Env, Response, Storage};
 use mixnet_contract_common::{
-    gateway::GatewayConfigUpdate,
    mixnode::{MixNodeConfigUpdate, MixNodeCostParams},
    Gateway, MixNode,
 };
@@ -65,10 +64,4 @@ pub trait GatewayBondingAccount {
        amount: Coin,
        storage: &mut dyn Storage,
    ) -> Result<(), ContractError>;
-
-    fn try_update_gateway_config(
-        &self,
-        new_config: GatewayConfigUpdate,
-        storage: &mut dyn Storage,
-    ) -> Result<Response, ContractError>;
 }
@@ -4,12 +4,9 @@ use crate::storage::MIXNET_CONTRACT_ADDRESS;
 use crate::traits::GatewayBondingAccount;
 use contracts_common::signing::MessageSignature;
 use cosmwasm_std::{wasm_execute, Coin, Env, Response, Storage, Uint128};
-use mixnet_contract_common::{
-    gateway::GatewayConfigUpdate, ExecuteMsg as MixnetExecuteMsg, Gateway,
-};
+use mixnet_contract_common::{ExecuteMsg as MixnetExecuteMsg, Gateway};
 use vesting_contract_common::events::{
    new_vesting_gateway_bonding_event, new_vesting_gateway_unbonding_event,
-    new_vesting_update_gateway_config_event,
 };

 use super::Account;
@@ -81,22 +78,4 @@ impl GatewayBondingAccount for Account {
        self.remove_gateway_pledge(storage)?;
        Ok(())
    }
-
-    fn try_update_gateway_config(
-        &self,
-        new_config: GatewayConfigUpdate,
-        storage: &mut dyn Storage,
-    ) -> Result<Response, ContractError> {
-        let msg = MixnetExecuteMsg::UpdateGatewayConfigOnBehalf {
-            new_config,
-            owner: self.owner_address().into_string(),
-        };
-
-        let update_gateway_config_msg =
-            wasm_execute(MIXNET_CONTRACT_ADDRESS.load(storage)?, &msg, vec![])?;
-
-        Ok(Response::new()
-            .add_message(update_gateway_config_msg)
-            .add_event(new_vesting_update_gateway_config_event()))
-    }
 }
@@ -0,0 +1,12 @@
+[package]
+name = "cpu-cycles"
+version = "0.1.0"
+edition = "2021"
+build = "build.rs"
+links = "cpucycles"
+
+[dependencies]
+libc = "0.2.140"
+
+[build-dependencies]
+cfg-if = "1"
@@ -0,0 +1,65 @@
+use std::{env, path::PathBuf, process::Command};
+
+fn main() {
+    let out_dir = env::var("OUT_DIR").unwrap();
+    let out_path = PathBuf::from(&out_dir);
+    let source_path = PathBuf::from("libcpucycles")
+        .canonicalize()
+        .expect("cannot canonicalize path");
+
+    cfg_if::cfg_if! {
+        if #[cfg(not(any(target_arch = "x86", target_arch = "x86_64", target_arch = "mips", target_arch = "powerpc", target_arch = "powerpc64", target_arch = "arm")))] {
+            panic!("Unsupported architecture - {}!", env::var("CARGO_CFG_TARGET_ARCH").unwrap(), )
+        }
+    };
+
+    let mut compile_o_command = Command::new("./configure");
+    let compile_o_command = compile_o_command
+        .current_dir(&source_path)
+        .arg(format!("--prefix={out_dir}"));
+
+    match compile_o_command.output() {
+        Ok(output) => {
+            if !output.status.success() {
+                panic!("{:?}", unsafe {
+                    std::str::from_utf8_unchecked(&output.stderr)
+                })
+            }
+        }
+        Err(e) => panic!("{e}"),
+    }
+
+    let mut compile_o_command = Command::new("make");
+    let compile_o_command = compile_o_command.current_dir(&source_path).arg("install");
+
+    match compile_o_command.output() {
+        Ok(output) => {
+            if !output.status.success() {
+                panic!("{:?}", unsafe {
+                    std::str::from_utf8_unchecked(&output.stderr)
+                })
+            }
+        }
+        Err(e) => panic!("{e}"),
+    }
+
+    println!(
+        "cargo:rustc-link-search=native={}",
+        out_path.join("lib").to_str().unwrap()
+    );
+    println!("cargo:rustc-link-lib=static=cpucycles");
+
+    let mut compile_o_command = Command::new("make");
+    let compile_o_command = compile_o_command.current_dir(source_path).arg("clean");
+
+    match compile_o_command.output() {
+        Ok(output) => {
+            if !output.status.success() {
+                panic!("{:?}", unsafe {
+                    std::str::from_utf8_unchecked(&output.stderr)
+                })
+            }
+        }
+        Err(e) => panic!("{e}"),
+    }
+}
@@ -0,0 +1,8 @@
+default:
+	cd build && $(MAKE)
+
+install:
+	cd build && $(MAKE) install
+
+clean:
+	cd build && $(MAKE) clean
@@ -0,0 +1,69 @@
+#!/usr/bin/env python3
+
+import os
+import datetime
+import markdown
+
+def load(fn):
+  with open(fn) as f:
+    return f.read()
+
+style = load('autogen/html-style')
+sitetitle = load('autogen/html-title')
+
+files = []
+
+with open('autogen/html-files') as f:
+  for line in f:
+    line = line.strip()
+    line = line.split(':')
+    if len(line) != 3: continue
+    files += [line]
+
+for md,html,pagetitle in files:
+  fnmd = 'doc/%s.md' % md
+  fnhtml = 'doc/html/%s.html' % html
+  output = ''
+
+  x = load(fnmd)
+  x = markdown.markdown(x,extensions=['markdown.extensions.extra','markdown.extensions.tables'])
+  mtime = datetime.datetime.utcfromtimestamp(os.path.getmtime(fnmd)).strftime('%Y.%m.%d')
+
+  output += '<html>\n<head>\n'
+  output += style
+  output += '<title>\n'
+  output += pagetitle
+  output += '</title>\n'
+  output += '</head>\n'
+  output += '<body>\n'
+
+  output += '<div class=headline>\n'
+  output += sitetitle
+  output += '</div>\n'
+
+  output += '<div class=nav>\n'
+  for submd,subhtml,subpagetitle in files:
+    if subhtml == html:
+      output += '<div class="navt here">'
+      output += pagetitle+'\n'
+    else:
+      output += '<div class="navt away">'
+      output += '<a href=%s.html>%s</a>\n' % (subhtml,subpagetitle)
+    output += '</div>'
+  output += '</div>\n'
+
+  output += '<div class=main>\n'
+  output += x
+  output += '<hr><font size=1><b>Version:</b>\n'
+  output += 'This is version %s of the "%s" web page.\n' % (mtime,pagetitle)
+  output += '</font>\n'
+  output += '</div>\n'
+
+  output += '</body>\n'
+  output += '</html>\n'
+
+  if not os.path.exists(fnhtml) or output != load(fnhtml):
+    with open(fnhtml+'.new','w') as f:
+      f.write(output)
+    os.chmod(fnhtml+'.new',0o444)
+    os.rename(fnhtml+'.new',fnhtml)
@@ -0,0 +1,7 @@
+readme:index:Intro
+download:download:Download
+install:install:Install
+api:api:API
+counters:counters:Counters
+selection:selection:Selection
+security:security:Security
@@ -0,0 +1,32 @@
+<meta http-equiv="content-type" content="text/html; charset=utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<style type="text/css">
+html{overflow-y:scroll}
+body{font-family:sans-serif}
+p,ul,ol,blockquote,pre{font-size:0.9em;line-height:1.6em}
+li p{font-size:1.0em}
+blockquote p{font-size:1.0em}
+tt{font-size:1.2em}
+code{font-size:1.2em}
+h1{font-size:1.5em}
+h2{font-size:1.3em}
+h3{font-size:1.0em}
+h1 a{text-decoration:none}
+table{border-collapse:collapse}
+th,td{border:1px solid black}
+table a{text-decoration:none}
+table tr{font-size:0.9em;line-height:1.6em}
+.links a:hover{text-decoration:underline}
+.links a:active{text-decoration:underline}
+.links img{width:200px;padding-left:1em}
+.links td{border:0px;padding-top:0.5em;padding-bottom:0.5em}
+.headline{padding:0;font-weight:bold;font-size:1.5em;vertical-align:top;padding-bottom:0.5em;color:#125d0d}
+.navt{display:inline-block;box-sizing:border-box;-moz-box-sizing:border-box;-webkit-box-sizing:border-box;
+min-width:14%;margin:0;padding:0;padding-left:0.5em;padding-right:0.5em;vertical-align:center;
+font-weight:bold;font-size:1.1em;text-align:center;border:1px solid black}
+.here{border-bottom:0px;background-color:#ffffff}
+.away{background-color:#125d0d;}
+.away a{text-decoration:none;display:block;color:#ffffff}
+.away a:hover,.away a:active{text-decoration:underline}
+.main{margin:0;padding-top:0em;padding-bottom:1%;clear:both}
+</style>
@@ -0,0 +1 @@
+libcpucycles
@@ -0,0 +1,3 @@
+#!/bin/sh
+
+pandoc --standalone --to man --metadata title=cpucycles --metadata section=3 < doc/api.md > doc/man/cpucycles.3
@@ -0,0 +1,93 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <time.h>
+#include <sys/time.h>
+#include <sys/types.h>
+#include <cpucycles.h>
+
+#define TIMINGS 63
+static long long t[TIMINGS+1];
+
+static void t_print(void)
+{
+  long long median = 0;
+  long long i,j;
+
+  for (i = 0;i < TIMINGS;++i)
+    t[i] = t[i+1]-t[i];
+  for (j = 0;j < TIMINGS;++j) {
+    long long belowj = 0;
+    long long abovej = 0;
+    for (i = 0;i < TIMINGS;++i) if (t[i] < t[j]) ++belowj;
+    for (i = 0;i < TIMINGS;++i) if (t[i] > t[j]) ++abovej;
+    if (belowj*2 < TIMINGS && abovej*2 < TIMINGS) {
+      median = t[j];
+      break;
+    }
+  }
+  printf(" %lld ",median);
+  for (i = 0;i < TIMINGS;++i)
+    printf("%+lld",t[i]-median);
+  printf("\n");
+  fflush(stdout);
+}
+
+static long long microseconds(void)
+{
+  struct timeval t;
+  long long result;
+  gettimeofday(&t,(struct timezone *) 0);
+  result = t.tv_sec;
+  result *= 1000000;
+  result += t.tv_usec;
+  return result;
+}
+
+static volatile int v;
+
+static void measure_cpucycles(void)
+{
+  long long loops,i,j;
+
+  printf("cpucycles persecond %lld\n",cpucycles_persecond());
+  printf("cpucycles implementation %s\n",cpucycles_implementation());
+
+  for (i = 0;i <= TIMINGS;++i)
+    t[i] = cpucycles();
+  printf("cpucycles median"); t_print();
+
+  for (loops = 1024;loops <= 1048576;loops *= 2) {
+    long long t00,t01,t10,t11;
+    long long m0,m1;
+    double ratiobelow,ratioabove;
+
+    t00 = cpucycles();
+    m0 = microseconds();
+    t01 = cpucycles();
+
+    for (j = 0;j < loops;++j) v = 0;
+
+    t10 = cpucycles();
+    m1 = microseconds();
+    t11 = cpucycles();
+
+    if (t01 < t00) continue;
+    if (t10 < t01) continue;
+    if (t11 < t10) continue;
+    if (m1 <= m0+2) continue;
+
+    ratiobelow = floor((1000000.0*(t10-t01))/(m1+1-m0));
+    ratioabove = ceil((1000000.0*(t11-t00))/(m1-m0-1));
+
+    printf("cpucycles observed persecond %.0lf...%.0lf with %lld loops %lld microseconds\n",ratiobelow,ratioabove,loops,m1-m0);
+  }
+}
+
+int main(int argc,char **argv)
+{
+  cpucycles_tracesetup();
+  printf("cpucycles version %s\n",cpucycles_version());
+  measure_cpucycles();
+  return 0;
+}
@@ -0,0 +1,2 @@
+gcc -Wall -fPIC -fwrapv -O -fvisibility=hidden
+clang -Wall -fPIC -fwrapv -Qunused-arguments -O -fvisibility=hidden
@@ -0,0 +1,309 @@
+#!/usr/bin/env python3
+
+import os
+import shutil
+import sys
+import platform
+import subprocess
+import tempfile
+
+prefix = '/usr/local'
+clean = True
+linktype = 'so'
+
+host = platform.machine()
+host = ''.join(c for c in host if c in '_0123456789abcdefghijklmnopqrstuvwxyz')
+
+if host == 'x86_64': host = 'amd64'
+if host == 'i386': host = 'x86'
+if host == 'i686': host = 'x86'
+
+if host.startswith('armv8') or host.startswith('aarch64'): host = 'arm64'
+if host.startswith('arm'): host = 'arm32'
+
+if host.startswith('riscv64'): host = 'riscv64'
+if host.startswith('riscv'): host = 'riscv32'
+
+if host.startswith('mips64'): host = 'mips64'
+if host.startswith('mips'): host = 'mips32'
+
+if host.startswith('powerpc64') or host.startswith('ppc64'): host = 'ppc64'
+if host.startswith('powerpc') or host.startswith('ppc'): host = 'ppc32'
+
+if host.startswith('sparcv9') or host.startswith('sun4u'): host = 'sparc64'
+if host.startswith('sparc') or host.startswith('sun'): host = 'sparc32'
+
+makefile = ''
+
+for arg in sys.argv[1:]:
+  if arg.startswith('--prefix='):
+    prefix = arg[9:]
+    continue
+  if arg.startswith('--host='):
+    host = arg[7:]
+    host = host.split('-')[0]
+    continue
+  if arg == '--clean':
+    clean = True
+    continue
+  if arg == '--noclean':
+    clean = False
+    continue
+  raise ValueError('unrecognized argument %s' % arg)
+
+echoargs = './configure'
+echoargs += ' --prefix=%s' % prefix
+echoargs += ' --host=%s' % host
+if clean: echoargs += ' --clean'
+if not clean: echoargs += ' --noclean'
+print(echoargs)
+
+if prefix[0] != '/':
+  raise ValueError('prefix %s is not an absolute path' % prefix)
+
+rpath = None
+# XXX: rpath = '%s/lib' % prefix
+
+if clean:
+  shutil.rmtree('build/%s' % host,ignore_errors=True)
+
+def dirlinksym(dir,source,target):
+  with tempfile.TemporaryDirectory(dir=dir) as t:
+    os.symlink(target,'%s/symlink' % t)
+    os.rename('%s/symlink' % t,'%s/%s' % (dir,source))
+
+os.makedirs('build/%s' % host,exist_ok=True)
+os.makedirs('build/%s/package/bin' % host,exist_ok=True)
+os.makedirs('build/%s/package/lib' % host,exist_ok=True)
+os.makedirs('build/%s/package/include' % host,exist_ok=True)
+
+if clean:
+  os.symlink('../..','build/%s/src' % host)
+
+# ----- build scripts
+
+os.makedirs('build/%s/scripts'%host,exist_ok=True)
+dirlinksym('build/%s/scripts'%host,'install','../src/scripts-build/install')
+
+# ----- compilers
+
+def compilerversion(c):
+  try:
+    p = subprocess.Popen(c.split()+['--version'],stdout=subprocess.PIPE,stderr=subprocess.STDOUT,universal_newlines=True)
+    out,err = p.communicate()
+    assert not err
+    assert not p.returncode
+    return out
+  except:
+    pass
+
+firstcompiler = None
+
+with open('compilers/default') as f:
+  for c in f.readlines():
+    c = c.strip()
+    cv = compilerversion(c)
+    if cv == None:
+      print('skipping default compiler %s' % c)
+      continue
+    print('using default compiler %s' % c)
+    firstcompiler = c
+    break
+
+if firstcompiler is None:
+  raise ValueError('did not find a working compiler')
+
+with open('build/%s/scripts/compiledefault' % host,'w') as f:
+  f.write('#!/bin/sh\n')
+  f.write('\n')
+  f.write('dir="$1"; shift\n')
+  f.write('base="$1"; shift\n')
+  f.write('ext="$1"; shift\n')
+  f.write('\n')
+  f.write('cd "$dir" && \\\n')
+  f.write('%s \\\n' % firstcompiler)
+  f.write('  "$@" \\\n')
+  f.write('  -c "$base.$ext"\n')
+os.chmod('build/%s/scripts/compiledefault' % host,0o755)
+
+# ----- libcpucycles
+
+os.makedirs('build/%s/cpucycles' % host,exist_ok=True)
+os.makedirs('build/%s/package/man/man3' % host,exist_ok=True)
+
+dirlinksym('build/%s/cpucycles'%host,'cpucycles.h','../src/cpucycles/cpucycles.h')
+dirlinksym('build/%s/cpucycles'%host,'cpucycles_internal.h','../src/cpucycles/cpucycles_internal.h')
+shutil.copy2('cpucycles/cpucycles.h','build/%s/package/include/cpucycles.h'%host)
+shutil.copy2('doc/man/cpucycles.3','build/%s/package/man/man3/cpucycles.3'%host)
+
+with open('build/%s/cpucycles/compile-ticks' % host,'w') as f:
+  f.write('#!/bin/sh\n')
+  f.write('arch="$1"; shift\n')
+  f.write('x="$1"; shift\n')
+  f.write('for source in try-"$arch"-"$x".c try-default-zero.c\n')
+  f.write('do\n')
+  f.write('  cp "$source" "$arch"-"$x".c\n')
+  f.write('  %s \\\n' % firstcompiler)
+  f.write('    -Dticks=cpucycles_ticks_"$arch"_"$x" \\\n')
+  f.write('    -Dticks_setup=cpucycles_ticks_"$arch"_"$x"_setup \\\n')
+  f.write('    -c "$arch"-"$x".c\n')
+  f.write('  case $? in\n')
+  f.write('    0) break ;;\n')
+  f.write('    111) exit 111 ;;\n')
+  f.write('    *) echo "skipping option that did not compile" ;;\n')
+  f.write('  esac\n')
+  f.write('done\n')
+os.chmod('build/%s/cpucycles/compile-ticks' % host,0o755)
+
+cpucyclesoptions = []
+cpucyclesofiles = []
+
+with open('cpucycles/options') as f:
+  for line in f:
+    line = line.strip()
+    if line == '': continue
+    if line[0] == '#': continue
+    base = line.split()[0]
+    if not os.path.exists('cpucycles/%s.c' % base): continue
+    cpucycles = base.split('-')
+    if len(cpucycles) != 2: continue
+    if cpucycles[0] not in (host,'default'): continue
+    cpucyclesoptions += [cpucycles]
+
+cpucyclesoptions += [['default','zero']] # must be last
+
+for cpucycles in cpucyclesoptions:
+  base = '-'.join(cpucycles)
+  cpucyclesofiles += ['cpucycles/%s.o' % base]
+  dirlinksym('build/%s/cpucycles'%host,'try-%s.c'%base,'../src/cpucycles/%s.c'%base)
+  M = 'cpucycles/%s.o: cpucycles/try-%s.c cpucycles/try-default-zero.c\n' % (base,base)
+  M += '\tcd cpucycles && ./compile-ticks %s %s\n' % tuple(cpucycles)
+  M += '\n'
+  makefile = M + makefile
+
+for fn in sorted(os.listdir('cpucycles')):
+  if not fn.endswith('.c'): continue
+  if '-' in fn: continue
+  base = fn[:-2]
+  cpucyclesofiles += ['cpucycles/%s.o' % base]
+  dirlinksym('build/%s/cpucycles'%host,fn,'../src/cpucycles/%s'%fn)
+  M = 'cpucycles/%s.o: cpucycles/%s.c\n' % (base,base)
+  M += '\tscripts/compiledefault cpucycles %s c\n' % base
+  M += '\n'
+  makefile = M + makefile
+
+with open('build/%s/cpucycles/options.inc' % host,'w') as f:
+  f.write('#define NUMOPTIONS %d\n' % len(cpucyclesoptions))
+  f.write('#define DEFAULTOPTION (NUMOPTIONS-1)\n')
+  f.write('\n')
+  for cpucycles in cpucyclesoptions:
+    f.write('extern long long cpucycles_ticks_%s_%s_setup(void);\n' % (cpucycles[0],cpucycles[1]))
+    f.write('extern long long cpucycles_ticks_%s_%s(void);\n' % (cpucycles[0],cpucycles[1]))
+  f.write('\n')
+  f.write('static struct {\n')
+  f.write('  const char *implementation;\n')
+  f.write('  long long (*ticks_setup)(void);\n')
+  f.write('  long long (*ticks)(void);\n')
+  f.write('} options[NUMOPTIONS] = {\n')
+  for cpucycles in cpucyclesoptions:
+    f.write('{ "%s-%s", cpucycles_ticks_%s_%s_setup, cpucycles_ticks_%s_%s },\n' % (cpucycles[0],cpucycles[1],cpucycles[0],cpucycles[1],cpucycles[0],cpucycles[1]))
+  f.write('} ;\n')
+
+dirlinksym('build/%s/scripts'%host,'staticlib','../src/scripts-build/staticlib')
+
+M = 'package/lib/libcpucycles.a: scripts/staticlib %s\n' % ' '.join(cpucyclesofiles)
+M += '\tscripts/staticlib %s\n' % ' '.join(cpucyclesofiles)
+M += '\n'
+makefile = M + makefile
+
+with open('build/%s/scripts/sharedlib' % host,'w') as f:
+  f.write('#!/bin/sh\n')
+  f.write('\n')
+  f.write('%s -shared \\\n' % firstcompiler)
+  if rpath:
+    f.write('  -Wl,-rpath=%s \\\n' % rpath)
+  f.write('  -Wl,-soname,libcpucycles.so.1 \\\n')
+  f.write('  -o package/lib/libcpucycles.so.1 \\\n')
+  f.write('  "$@"\n')
+  f.write('chmod 644 package/lib/libcpucycles.so.1\n')
+os.chmod('build/%s/scripts/sharedlib' % host,0o755)
+
+M = 'package/lib/libcpucycles.so.1: scripts/sharedlib %s\n' % ' '.join(cpucyclesofiles)
+M += '\tscripts/sharedlib %s\n' % ' '.join(cpucyclesofiles)
+M += '\n'
+makefile = M + makefile
+
+M = 'package/lib/libcpucycles.so: package/lib/libcpucycles.so.1\n'
+M += '\trm -f package/lib/libcpucycles.so\n'
+M += '\tln -s libcpucycles.so.1 package/lib/libcpucycles.so\n'
+M += '\n'
+makefile = M + makefile
+
+# ----- command
+
+os.makedirs('build/%s/command'%host)
+for c in sorted(os.listdir('command')):
+  dirlinksym('build/%s/command'%host,c,'../src/command/%s'%c)
+dirlinksym('build/%s/command'%host,'bin','../package/bin')
+dirlinksym('build/%s/command'%host,'lib','../package/lib')
+dirlinksym('build/%s/command'%host,'include','../package/include')
+
+with open('build/%s/command/link' % host,'w') as f:
+  f.write('#!/bin/sh\n')
+  f.write('target="$1"; shift\n')
+  f.write('%s \\\n' % firstcompiler)
+  f.write('  -o "$target" "$@"\n')
+os.chmod('build/%s/command/link' % host,0o755)
+
+commands = []
+
+for fn in sorted(os.listdir('command')):
+  if not fn.endswith('.c'): continue
+
+  libs = ['libcpucycles']
+
+  base = fn[:-2]
+  M = 'command/%s.o: command/%s.c\n' % (base,base)
+  M += '\tscripts/compiledefault command %s c -I include\n' % base
+  M += '\n'
+  makefile = M + makefile
+  M = 'package/bin/%s: command/%s.o%s\n' % (base,base,''.join(' package/lib/%s.%s' % (x,linktype) for x in libs))
+  M += '\tcd command && ./link bin/%s %s.o%s -lm -lrt\n' % (base,base,''.join(' lib/%s.%s' % (x,linktype) for x in libs))
+  M += '\n'
+  makefile = M + makefile
+  commands += ['package/bin/%s' % base]
+
+M = 'commands: %s\n' % ' '.join(commands)
+M += '\n'
+makefile = M + makefile
+
+# ----- make install
+
+M = 'install: scripts/install default\n'
+M += '\tscripts/install %s\n' % prefix
+M += '\n'
+makefile = M + makefile
+
+# ----- make default
+
+M = 'default: package/lib/libcpucycles.a package/lib/libcpucycles.so package/lib/libcpucycles.so.1 \\\n'
+M += 'commands\n'
+M += '\n'
+makefile = M + makefile
+
+with open('build/%s/Makefile' % host,'w') as f:
+  f.write(makefile)
+
+# ----- build/0, build/Makefile
+
+dirlinksym('build','0',host)
+
+with open('build/Makefile','w') as f:
+  f.write('default:\n')
+  f.write('\tcd %s && $(MAKE)\n' % host)
+  f.write('\n')
+  f.write('install:\n')
+  f.write('\tcd %s && $(MAKE) install\n' % host)
+  f.write('\n')
+  f.write('clean:\n')
+  f.write('\trm -r %s\n' % host)
@@ -0,0 +1,53 @@
+// version 20230105
+// public domain
+// djb
+// adapted from supercop/cpucycles/amd64rdpmc.c
+
+#include <stdlib.h>
+#include <unistd.h>
+#include <sys/types.h>
+#include <sys/mman.h>
+#include <sys/syscall.h>
+#include <linux/perf_event.h>
+#include "cpucycles_internal.h"
+
+static struct perf_event_attr attr;
+static int fdperf = -1;
+static struct perf_event_mmap_page *buf = 0;
+
+long long ticks(void)
+{
+  long long result;
+  unsigned int seq;
+  long long index;
+  long long offset;
+
+  do {
+    seq = buf->lock;
+    asm volatile("" ::: "memory");
+    index = buf->index;
+    offset = buf->offset;
+    asm volatile("rdpmc;shlq $32,%%rdx;orq %%rdx,%%rax"
+      : "=a"(result) : "c"(index-1) : "%rdx");
+    asm volatile("" ::: "memory");
+  } while (buf->lock != seq);
+
+  result += offset;
+  result &= 0xffffffffffff;
+  return result;
+}
+
+long long ticks_setup(void)
+{
+  if (fdperf == -1) {
+    attr.type = PERF_TYPE_HARDWARE;
+    attr.config = PERF_COUNT_HW_CPU_CYCLES;
+    attr.exclude_kernel = 1;
+    fdperf = syscall(__NR_perf_event_open,&attr,0,-1,-1,0);
+    if (fdperf == -1) return cpucycles_SKIP;
+    buf = mmap(NULL,sysconf(_SC_PAGESIZE),PROT_READ,MAP_SHARED,fdperf,0);
+  }
+
+  if (!cpucycles_works(ticks)) return cpucycles_SKIP;
+  return cpucycles_CYCLECOUNTER;
+}
@@ -0,0 +1,22 @@
+// version 20230105
+// public domain
+// djb
+
+#ifdef _MSC_VER
+#include <intrin.h>
+#else
+#include <x86intrin.h>
+#endif
+
+#include "cpucycles_internal.h"
+
+long long ticks(void)
+{
+  return __rdtsc();
+}
+
+long long ticks_setup(void)
+{
+  if (!cpucycles_works(ticks)) return cpucycles_SKIP;
+  return cpucycles_MAYBECYCLECOUNTER;
+}
@@ -0,0 +1,20 @@
+// version 20230105
+// public domain
+// djb
+// adapted from supercop/cpucycles/amd64tscfreq.c
+
+#include "cpucycles_internal.h"
+
+long long ticks(void)
+{
+  unsigned long long result;
+  asm volatile(".byte 15;.byte 49;shlq $32,%%rdx;orq %%rdx,%%rax"
+    : "=a"(result) :: "%rdx");
+  return result;
+}
+
+long long ticks_setup(void)
+{
+  if (!cpucycles_works(ticks)) return cpucycles_SKIP;
+  return cpucycles_MAYBECYCLECOUNTER;
+}
@@ -0,0 +1,27 @@
+// version 20230105
+// public domain
+// djb
+// adapted from supercop/cpucycles/cortex.c
+
+#include "cpucycles_internal.h"
+
+long long ticks(void)
+{
+  unsigned int result;
+  asm volatile("mrc p15, 0, %0, c9, c13, 0" : "=r"(result));
+  return (unsigned long long) result;
+}
+
+static long enable(void)
+{
+  asm volatile("mcr p15, 0, %0, c9, c12, 0" :: "r"(17));
+  asm volatile("mcr p15, 0, %0, c9, c12, 1" :: "r"(0x8000000f));
+  asm volatile("mcr p15, 0, %0, c9, c12, 3" :: "r"(0x8000000f));
+}
+
+long long ticks_setup(void)
+{
+  if (!cpucycles_works(enable)) return cpucycles_SKIP;
+  if (!cpucycles_works(ticks)) return cpucycles_SKIP;
+  return cpucycles_EXTEND32;
+}
@@ -0,0 +1,19 @@
+// version 20230105
+// public domain
+// djb
+// adapted from supercop/cpucycles/armv8.c
+
+#include "cpucycles_internal.h"
+
+long long ticks(void)
+{
+  long long result;
+  asm volatile("mrs %0, PMCCNTR_EL0" : "=r" (result));
+  return result;
+}
+
+long long ticks_setup(void)
+{
+  if (!cpucycles_works(ticks)) return cpucycles_SKIP;
+  return cpucycles_CYCLECOUNTER;
+}
@@ -0,0 +1,19 @@
+// version 20230105
+// public domain
+// djb
+// adapted from supercop/cpucycles/vct.c
+
+#include "cpucycles_internal.h"
+
+long long ticks(void)
+{
+  long long result;
+  asm volatile("mrs %0, CNTVCT_EL0" : "=r" (result));
+  return result;
+}
+
+long long ticks_setup(void)
+{
+  if (!cpucycles_works(ticks)) return cpucycles_SKIP;
+  return cpucycles_FINDMULTIPLIER;
+}
@@ -0,0 +1,25 @@
+// version 20230115
+// public domain
+// djb
+
+// 20230115 djb: cpucycles_version()
+// 20230114 djb: improve punctuation
+
+#ifndef cpucycles_h
+#define cpucycles_h
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+extern long long (*cpucycles)(void) __attribute__((visibility("default")));
+extern const char *cpucycles_implementation(void) __attribute__((visibility("default")));
+extern const char *cpucycles_version(void) __attribute__((visibility("default")));
+extern long long cpucycles_persecond(void) __attribute__((visibility("default")));
+extern void cpucycles_tracesetup(void) __attribute__((visibility("default")));
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
@@ -0,0 +1,20 @@
+// version 20230105
+// public domain
+// djb
+
+#ifndef cpucycles_internal_h
+#define cpucycles_internal_h
+
+extern long long cpucycles_init(void);
+extern long long cpucycles_microseconds(void);
+extern int cpucycles_works(long long (*)(void));
+
+// return values from ticks_setup():
+#define cpucycles_SKIP (0)
+#define cpucycles_CYCLECOUNTER (-1)
+#define cpucycles_MAYBECYCLECOUNTER (-2)
+#define cpucycles_FINDMULTIPLIER (-3)
+#define cpucycles_EXTEND32 (-32)
+// and positive values mean known ticks/second
+
+#endif
@@ -0,0 +1,15 @@
+// version 20230105
+// public domain
+// djb
+
+#include "cpucycles_internal.h"
+
+long long ticks_setup(void)
+{
+  return 1000000;
+}
+
+long long ticks(void)
+{
+  return cpucycles_microseconds();
+}
@@ -0,0 +1,17 @@
+// version 20230105
+// public domain
+// djb
+
+#include <mach/mach_time.h>
+#include "cpucycles_internal.h"
+
+long long ticks(void)
+{
+  return mach_absolute_time();
+}
+
+long long ticks_setup(void)
+{
+  if (!cpucycles_works(ticks)) return cpucycles_SKIP;
+  return cpucycles_FINDMULTIPLIER;
+}
@@ -0,0 +1,23 @@
+// version 20230105
+// public domain
+// djb
+// adapted from supercop/cpucycles/monotonic.c
+
+#include <time.h>
+#include <sys/time.h>
+
+long long ticks_setup(void)
+{
+  return 1000000000;
+}
+
+long long ticks(void)
+{
+  struct timespec t;
+  long long result;
+  clock_gettime(CLOCK_MONOTONIC,&t);
+  result = t.tv_sec;
+  result *= 1000000000;
+  result += t.tv_nsec;
+  return result;
+}
@@ -0,0 +1,101 @@
+// version 20230106
+// public domain
+// djb
+// adapted from supercop/cpucycles/perfevent.c
+
+// 20230106 djb: read() into int64_t instead of long long
+// 20230106 djb: add comment on RUNNING/ENABLED
+
+/*
+This code intentionally avoids dividing by the
+PERF_FORMAT_TOTAL_TIME_RUNNING/ENABLED ratio.
+
+The motivation for that ratio is as follows:
+
+* A typical CPU has a limited number of performance-monitoring
+  counters active at once. For example, there are 8 "programmable"
+  counters on Intel Skylake.
+
+* "perf stat" allows the user to enable more counters. The OS kernel
+  periodically (e.g., every millisecond) changes the limited number of
+  active hardware counters to a new subset of the enabled counters, and
+  "perf stat" reports PERF_FORMAT_TOTAL_TIME_RUNNING/ENABLED for each
+  counter, the fraction of time spent with that counter running.
+
+For long-running programs, dividing the hardware counter by
+RUNNING/ENABLED usually produces a reasonable estimate of what the count
+would have been without competition from other counters.
+
+A fixable problem with this multiplexing of counters is that the kernel
+appears to simply cycle through counters, so unlucky programs can
+trigger moiré effects. The fix is to select random subsets of counters.
+
+A more fundamental problem is that cpucycles() has to be usable for
+timing short subroutines, including subroutines so short that the OS has
+no opportunity to change from one selection of counters to another. Say
+RUNNING is 0; should cpucycles() then divide by 0?
+
+If a caller runs cpucycles(), X(), cpucycles(), X(), etc., and the cycle
+counter happens to be enabled for only 80% of the runs of X(), then
+simply computing the median difference of adjacent cycle counts, with no
+scaling, will filter out the zeros and correctly compute the cost of X.
+Averages won't (without scaling), but averages have other problems, such
+as being heavily influenced by interrupts. (Omitting kernel time from
+perf results does not remove the influence of interrupts on caches.)
+
+Given the importance of cycle counting, it is better to have cycle
+counters always running. For example, on Skylake, Intel provides the 8
+"programmable" counters on top of a separate cycle counter ("fixed
+counter 1"), so there is no good reason for the kernel to waste a
+"programmable" counter on a cycle counter, there is no good reason to
+turn the cycle counter off, and there is no good reason for RUNNING to
+be below ENABLED for the cycle counter.
+
+Of course, applications that use just one performance counter at a time
+don't have to worry about kernels getting this wrong, and don't have to
+worry about the possibility of getting noisy or invalid results on CPUs
+that have heavier constraints on the number of simultaneous counters.
+*/
+
+#include <stdlib.h>
+#include <stdint.h>
+#include <unistd.h>
+#include <string.h>
+#include <sys/ioctl.h>
+#include <sys/types.h>
+#include <sys/syscall.h>
+#include <linux/perf_event.h>
+#include "cpucycles_internal.h"
+
+static int fddev = -1;
+
+long long ticks(void)
+{
+  int64_t result;
+
+  if (read(fddev,&result,sizeof result) < sizeof result) return 0;
+  return result;
+}
+
+long long ticks_setup(void)
+{
+  if (fddev == -1) {
+    static struct perf_event_attr attr;
+
+    memset(&attr,0,sizeof attr);
+    attr.type = PERF_TYPE_HARDWARE;
+    attr.size = sizeof(struct perf_event_attr);
+    attr.config = PERF_COUNT_HW_CPU_CYCLES;
+    attr.disabled = 1;
+    attr.exclude_kernel = 1;
+    attr.exclude_hv = 1;
+
+    fddev = syscall(__NR_perf_event_open,&attr,0,-1,-1,0);
+    if (fddev == -1) return cpucycles_SKIP;
+
+    ioctl(fddev,PERF_EVENT_IOC_RESET,0);
+    ioctl(fddev,PERF_EVENT_IOC_ENABLE,0);
+  }
+
+  return cpucycles_MAYBECYCLECOUNTER;
+}
@@ -0,0 +1,15 @@
+// version 20230105
+// public domain
+// djb
+
+#include "cpucycles_internal.h"
+
+long long ticks_setup(void)
+{
+  return cpucycles_SKIP;
+}
+
+long long ticks(void)
+{
+  return 0;
+}
@@ -0,0 +1,33 @@
+// version 20230105
+// public domain
+// djb
+// adapted from supercop/cpucycles/mips.c
+
+// mips32 release 2 instruction rdhwr
+// 7c02103b: read hwr#2 (cycle count) into $2
+// 7c02183b: read hwr#3 (cycle-count multiplier) into $2
+
+#include "cpucycles_internal.h"
+
+static unsigned int multiplier = 0;
+
+static long long multiplier_set(void)
+{
+  asm volatile(".long 0x7c02183b; move %0,$2" : "=r"(multiplier) : : "$2");
+  return multiplier;
+}
+
+long long ticks(void)
+{
+  unsigned int result;
+  asm volatile(".long 0x7c02103b; move %0,$2" : "=r"(result) :: "$2");
+  result *= multiplier;
+  return (unsigned long long) result;
+}
+
+long long ticks_setup(void)
+{
+  if (!cpucycles_works(multiplier_set)) return cpucycles_SKIP;
+  if (!cpucycles_works(ticks)) return cpucycles_SKIP;
+  return cpucycles_EXTEND32;
+}
@@ -0,0 +1,19 @@
+amd64-pmc
+amd64-tsc
+amd64-tscasm
+arm32-cortex
+arm64-pmc
+arm64-vct
+mips64-cc
+ppc32-mftb
+ppc64-mftb
+riscv32-rdcycle
+riscv64-rdcycle
+s390x-stckf
+sparc64-rdtick
+x86-tsc
+x86-tscasm
+default-perfevent
+default-mach
+default-monotonic
+default-gettimeofday
@@ -0,0 +1,30 @@
+// version 20230105
+// public domain
+// djb
+// adapted from supercop/cpucycles/powerpccpuinfo.c
+
+#include "cpucycles_internal.h"
+
+long long ticks(void)
+{
+  unsigned int high, low, newhigh;
+  unsigned long long result;
+
+  do {
+    asm volatile(
+      "mftbu %0; mftb %1; mftbu %2"
+      : "=r" (high), "=r" (low), "=r" (newhigh)
+    );
+  } while (newhigh != high);
+
+  result = high;
+  result <<= 32;
+  result |= low;
+  return result;
+}
+
+long long ticks_setup(void)
+{
+  if (!cpucycles_works(ticks)) return cpucycles_SKIP;
+  return cpucycles_FINDMULTIPLIER;
+}
@@ -0,0 +1,30 @@
+// version 20230105
+// public domain
+// djb
+// adapted from supercop/cpucycles/powerpccpuinfo.c
+
+#include "cpucycles_internal.h"
+
+long long ticks(void)
+{
+  unsigned int high, low, newhigh;
+  unsigned long long result;
+
+  do {
+    asm volatile(
+      "mftbu %0; mftb %1; mftbu %2"
+      : "=r" (high), "=r" (low), "=r" (newhigh)
+    );
+  } while (newhigh != high);
+
+  result = high;
+  result <<= 32;
+  result |= low;
+  return result;
+}
+
+long long ticks_setup(void)
+{
+  if (!cpucycles_works(ticks)) return cpucycles_SKIP;
+  return cpucycles_FINDMULTIPLIER;
+}
@@ -0,0 +1,39 @@
+// version 20230105
+// public domain
+// djb
+
+// adapted from supercop/cpucycles/riscv.c
+// which has code from djb and Romain Dolbeau
+
+#include "cpucycles_internal.h"
+
+#ifndef __riscv_xlen
+#error this code is only for riscv platforms
+#endif
+
+#if __riscv_xlen != 32
+#error this code is only for riscv32 platforms
+#endif
+
+long long ticks(void)
+{
+  unsigned int low, high, newhigh;
+  unsigned long long result;
+
+  asm volatile( "start%=:\n"
+                "rdcycleh %0\n"
+                "rdcycle %1\n"
+                "rdcycleh %2\n"
+                "bne %0, %2, start%=\n"
+                : "=r"(high), "=r"(low), "=r"(newhigh));
+  result = high;
+  result <<= 32;
+  result |= low;
+  return result;
+}
+
+long long ticks_setup(void)
+{
+  if (!cpucycles_works(ticks)) return cpucycles_SKIP;
+  return cpucycles_CYCLECOUNTER;
+}
@@ -0,0 +1,29 @@
+// version 20230105
+// public domain
+// djb
+
+// adapted from supercop/cpucycles/riscv.c
+// which has code from djb and Romain Dolbeau
+
+#include "cpucycles_internal.h"
+
+#ifndef __riscv_xlen
+#error this code is only for riscv platforms
+#endif
+
+#if __riscv_xlen != 64
+#error this code is only for riscv64 platforms
+#endif
+
+long long ticks(void)
+{
+  long long result;
+  asm volatile("rdcycle %0" : "=r" (result));
+  return result;
+}
+
+long long ticks_setup(void)
+{
+  if (!cpucycles_works(ticks)) return cpucycles_SKIP;
+  return cpucycles_CYCLECOUNTER;
+}
@@ -0,0 +1,20 @@
+// version 20230106
+// public domain
+// djb
+
+// adapted from sparc64-rdtick.c
+
+#include "cpucycles_internal.h"
+
+long long ticks(void)
+{
+  long long result;
+  asm volatile("stckf 0(%0)" :: "a"(&result) : "memory","cc");
+  return result;
+}
+
+long long ticks_setup(void)
+{
+  if (!cpucycles_works(ticks)) return cpucycles_SKIP;
+  return 4096000000; // manual says 2^12 per microsecond
+}
@@ -0,0 +1,24 @@
+// version 20230105
+// public domain
+// djb
+
+// adapted from supercop/cpucycles/sparccpuinfo.c
+
+#include "cpucycles_internal.h"
+
+#if defined(__sparcv8) || defined(__sparcv8plus)
+#error this code is only for sparc64 platforms
+#endif
+
+long long ticks(void)
+{
+  long long result;
+  asm volatile("rd %%tick,%0" : "=r" (result));
+  return result;
+}
+
+long long ticks_setup(void)
+{
+  if (!cpucycles_works(ticks)) return cpucycles_SKIP;
+  return cpucycles_CYCLECOUNTER;
+}
@@ -0,0 +1,420 @@
+// version 20230115
+// public domain
+// djb
+// includes some pieces adapted from supercop
+
+// 20230115 djb: cpucycles_version()
+// 20230106 djb: support "cpu MHz static" (ibm z15)
+
+#include <time.h>
+#include <sys/time.h>
+#include <unistd.h>
+#include <stdio.h>
+#include <string.h>
+#include <stdlib.h>
+#include <inttypes.h>
+#include <signal.h>
+#include <setjmp.h>
+#include "cpucycles.h"
+#include "cpucycles_internal.h"
+
+static int tracesetup = 0;
+
+void cpucycles_tracesetup(void)
+{
+  tracesetup = 1;
+}
+
+static jmp_buf crash_jmp;
+
+static void crash(int s)
+{
+  siglongjmp(crash_jmp,0);
+}
+
+int cpucycles_works(long long (*ticks)(void))
+{
+  volatile int result = 0;
+  struct sigaction old_sigill;
+  struct sigaction old_sigfpe;
+  struct sigaction old_sigbus;
+  struct sigaction old_sigsegv;
+  struct sigaction crash_action;
+
+  memset(&crash_action,0,sizeof crash_action);
+  crash_action.sa_handler = crash;
+
+  sigaction(SIGILL,0,&old_sigill);
+  sigaction(SIGFPE,0,&old_sigfpe);
+  sigaction(SIGBUS,0,&old_sigbus);
+  sigaction(SIGSEGV,0,&old_sigsegv);
+
+  if (!sigsetjmp(crash_jmp,1)) {
+    sigaction(SIGILL,&crash_action,0);
+    sigaction(SIGFPE,&crash_action,0);
+    sigaction(SIGBUS,&crash_action,0);
+    sigaction(SIGSEGV,&crash_action,0);
+    ticks();
+    result = 1;
+  }
+
+  sigaction(SIGILL,&old_sigill,0);
+  sigaction(SIGFPE,&old_sigfpe,0);
+  sigaction(SIGBUS,&old_sigbus,0);
+  sigaction(SIGSEGV,&old_sigsegv,0);
+
+  return result;
+}
+
+static double osfreq(void)
+{
+  FILE *f;
+  char *x;
+  double result;
+  int s;
+
+  f = fopen("/etc/cpucyclespersecond", "r");
+  if (f) {
+    s = fscanf(f,"%lf",&result);
+    fclose(f);
+    if (s > 0) return result;
+  }
+
+  f = fopen("/sys/devices/system/cpu/cpu0/cpufreq/scaling_setspeed", "r");
+  if (f) {
+    s = fscanf(f,"%lf",&result);
+    fclose(f);
+    if (s > 0) return 1000.0 * result;
+  }
+
+  f = fopen("/sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq", "r");
+  if (f) {
+    s = fscanf(f,"%lf",&result);
+    fclose(f);
+    if (s > 0) return 1000.0 * result;
+  }
+
+  f = fopen("/sys/devices/system/cpu/cpu0/clock_tick", "r");
+  if (f) {
+    s = fscanf(f,"%lf",&result);
+    fclose(f);
+    if (s > 0) return result;
+  }
+
+  f = fopen("/proc/cpuinfo","r");
+  if (f) {
+    for (;;) {
+      s = fscanf(f,"cpu MHz : %lf",&result);
+      if (s > 0) break;
+      if (s == 0) s = fscanf(f,"%*[^\n]\n");
+      if (s < 0) { result = 0; break; }
+    }
+    fclose(f);
+    if (result) return 1000000.0 * result;
+  }
+
+  f = fopen("/proc/cpuinfo","r");
+  if (f) {
+    for (;;) {
+      s = fscanf(f,"clock : %lf",&result);
+      if (s > 0) break;
+      if (s == 0) s = fscanf(f,"%*[^\n]\n");
+      if (s < 0) { result = 0; break; }
+    }
+    fclose(f);
+    if (result) return 1000000.0 * result;
+  }
+
+  f = fopen("/proc/cpuinfo","r");
+  if (f) {
+    for (;;) {
+      s = fscanf(f,"cpu MHz static : %lf",&result);
+      if (s > 0) break;
+      if (s == 0) s = fscanf(f,"%*[^\n]\n");
+      if (s < 0) { result = 0; break; }
+    }
+    fclose(f);
+    if (result) return 1000000.0 * result;
+  }
+
+  f = popen("sysctl hw.cpufrequency 2>/dev/null","r");
+  if (f) {
+    s = fscanf(f,"hw.cpufrequency: %lf",&result);
+    pclose(f);
+    if (s > 0) if (result > 0) return result;
+  }
+
+  f = popen("/usr/sbin/lsattr -E -l proc0 -a frequency 2>/dev/null","r");
+  if (f) {
+    s = fscanf(f,"frequency %lf",&result);
+    pclose(f);
+    if (s > 0) return result;
+  }
+
+  f = popen("/usr/sbin/psrinfo -v 2>/dev/null","r");
+  if (f) {
+    for (;;) {
+      s = fscanf(f," The %*s processor operates at %lf MHz",&result);
+      if (s > 0) break;
+      if (s == 0) s = fscanf(f,"%*[^\n]\n");
+      if (s < 0) { result = 0; break; }
+    }
+    pclose(f);
+    if (result) return 1000000.0 * result;
+  }
+
+  x = getenv("cpucyclespersecond");
+  if (x) {
+    s = sscanf(x,"%lf",&result);
+    if (s > 0) return result;
+  }
+
+  return 2399987654.0;
+}
+
+static long long persecond = 0;
+static const char *implementation = "none";
+
+long long (*cpucycles)(void) = cpucycles_init;
+
+const char *cpucycles_implementation(void)
+{
+  cpucycles();
+  return implementation;
+}
+
+long long cpucycles_persecond(void)
+{
+  cpucycles();
+  return persecond;
+}
+
+const char *cpucycles_version(void)
+{
+  return "20230115";
+}
+
+// ----- cycle counter scaled from ticks
+
+static double cpucycles_scaled_scaling = 0;
+static long long cpucycles_scaled_offset = 0;
+static long long (*cpucycles_scaled_from)(void) = 0;
+
+static long long cpucycles_scaled(void)
+{
+  return (cpucycles_scaled_from()-cpucycles_scaled_offset)*cpucycles_scaled_scaling;
+}
+
+// ----- cycle counter extended from 32-bit ticks
+
+static long long (*cpucycles_extend32_from)(void) = 0;
+
+static uint32_t cpucycles_extend32_prev_ticks;
+static long long cpucycles_extend32_prev_us;
+static long long cpucycles_extend32_prev_cycles;
+
+static void cpucycles_extend32_setup(void)
+{
+  long long (*ticks)(void) = cpucycles_extend32_from;
+  cpucycles_extend32_prev_ticks = ticks();
+  cpucycles_extend32_prev_us = cpucycles_microseconds();
+  cpucycles_extend32_prev_cycles = 0;
+}
+
+static long long cpucycles_extend32(void)
+{
+  long long (*ticks)(void) = cpucycles_extend32_from;
+
+  uint32_t new_ticks = ticks();
+  unsigned long long delta_ticks = new_ticks-cpucycles_extend32_prev_ticks;
+  long long new_us = cpucycles_microseconds();
+  long long delta_us = new_us-cpucycles_extend32_prev_us;
+
+  // assume that number of cycles cannot increase by 2^32 in 2ms
+
+  if (delta_us < 1000)
+    return cpucycles_extend32_prev_cycles+delta_ticks;
+
+  cpucycles_extend32_prev_ticks = new_ticks;
+  cpucycles_extend32_prev_us = new_us;
+
+  if (delta_us >= 2000) {
+    long long target = (delta_us*0.000001)*persecond;
+    while (delta_ticks+2147483648ULL < target)
+      delta_ticks += 4294967296ULL;
+  }
+
+  return cpucycles_extend32_prev_cycles += delta_ticks;
+}
+
+// ----- estimating cycles per tick
+
+long long cpucycles_microseconds(void)
+{
+  struct timeval t;
+  long long result;
+  gettimeofday(&t,(struct timezone *) 0);
+  result = t.tv_sec;
+  result *= 1000000;
+  result += t.tv_usec;
+  return result;
+}
+
+static double estimate_cyclespertick(long long (*ticks)(void))
+{
+  long long t0,t1,us0,us1;
+
+  t0 = ticks();
+  us0 = cpucycles_microseconds();
+  do {
+    t1 = ticks();
+    us1 = cpucycles_microseconds();
+  } while (us1-us0 < 10000 || t1-t0 < 1000);
+  if (t1 <= t0) return 0;
+  t1 -= t0;
+  us1 -= us0;
+  return (persecond * 0.000001 * (double) us1) / (double) t1;
+}
+
+// ----- selecting an option
+
+#include "options.inc"
+
+#define CALLS 1000
+#define ESTIMATES 3
+
+long long cpucycles_init(void)
+{
+  long long precision[NUMOPTIONS];
+  double scaling[NUMOPTIONS];
+  int only32[NUMOPTIONS];
+  long long bestprecision;
+  long long bestopt;
+  long long opt;
+
+  persecond = osfreq();
+
+  for (opt = 0;opt < NUMOPTIONS;++opt) {
+    long long freq = options[opt].ticks_setup();
+    long long tries;
+
+    precision[opt] = 0;
+    scaling[opt] = 0;
+    only32[opt] = 0;
+
+    if (freq > 0) {
+      scaling[opt] = persecond*1.0/freq;
+    } else if (freq == cpucycles_CYCLECOUNTER) {
+      scaling[opt] = 1.0;
+    } else if (freq == cpucycles_EXTEND32) {
+      only32[opt] = 1;
+      scaling[opt] = 1.0;
+    } else if (freq == cpucycles_MAYBECYCLECOUNTER) {
+      scaling[opt] = 1.0;
+    } else if (freq == cpucycles_FINDMULTIPLIER) {
+      int ok = 0;
+      double denom;
+      long long loop;
+
+      for (denom = 1;denom <= 1024;denom += denom) {
+        double est[ESTIMATES];
+        for (loop = 0;loop < ESTIMATES;++loop)
+          est[loop] = denom*estimate_cyclespertick(options[opt].ticks);
+        scaling[opt] = (double) (long long) est[0];
+        if (scaling[opt] < est[0]-0.5) scaling[opt] += 1;
+        if (scaling[opt] > est[0]+0.5) scaling[opt] -= 1;
+        ok = 1;
+        for (loop = 0;loop < ESTIMATES;++loop) {
+          if (est[loop]-scaling[opt] > 0.1) ok = 0;
+          if (scaling[opt]-est[loop] > 0.1) ok = 0;
+        }
+        if (ok) {
+          scaling[opt] /= denom;
+          break;
+        }
+        scaling[opt] = 0;
+      }
+      if (!ok) continue;
+    } else {
+      continue;
+    }
+
+    for (tries = 0;tries < 10;++tries) {
+      long long t[CALLS+1];
+      long long ok = 1;
+      long long i;
+
+      if (scaling[opt] == 1.0) {
+        for (i = 0;i <= CALLS;++i)
+          t[i] = options[opt].ticks();
+      } else {
+        double scalingopt = scaling[opt];
+        long long offset = options[opt].ticks();
+        for (i = 0;i <= CALLS;++i)
+          t[i] = (options[opt].ticks()-offset)*scalingopt;
+      }
+      for (i = 0;i < CALLS;++i)
+        if (t[i] > t[i+1])
+          ok = 0;
+      if (t[0] == t[CALLS])
+        ok = 0;
+
+      if (ok) {
+        long long smallestdiff = 0;
+        for (i = 0;i < CALLS;++i) {
+          long long diff = t[i+1]-t[i];
+          if (diff <= 0) continue;
+          if (smallestdiff == 0 || diff < smallestdiff)
+            smallestdiff = diff;
+        }
+        precision[opt] = smallestdiff;
+
+        // tilt selection towards more robust counters
+        if (freq != cpucycles_CYCLECOUNTER && freq != cpucycles_EXTEND32)
+          precision[opt] += 100;
+        if (freq > 0)
+          precision[opt] += 100;
+
+        break;
+      }
+
+      // otherwise keep trying
+      // since !ok can be caused by overflow
+      // or by core swap
+    }
+  }
+
+  if (tracesetup) {
+    for (opt = 0;opt < NUMOPTIONS;++opt)
+      printf("cpucycles tracesetup %lld %s precision %lld scaling %lf only32 %d\n"
+        ,opt,options[opt].implementation,precision[opt],scaling[opt],only32[opt]);
+  }
+
+  bestopt = DEFAULTOPTION;
+  bestprecision = 0;
+  for (opt = 0;opt < NUMOPTIONS;++opt)
+    if (precision[opt] > 0)
+      if (!bestprecision || precision[opt] < bestprecision) {
+        bestopt = opt;
+        bestprecision = precision[opt];
+      }
+
+  implementation = options[bestopt].implementation;
+  
+  if (scaling[bestopt] == 1.0) {
+    if (only32[bestopt]) {
+      cpucycles_extend32_from = options[bestopt].ticks;
+      cpucycles_extend32_setup();
+      cpucycles = cpucycles_extend32;
+    } else {
+      cpucycles = options[bestopt].ticks;
+    }
+  } else {
+    cpucycles_scaled_scaling = scaling[bestopt];
+    cpucycles_scaled_from = options[bestopt].ticks;
+    cpucycles_scaled_offset = cpucycles_scaled_from();
+    cpucycles = cpucycles_scaled;
+  }
+
+  return cpucycles();
+}
@@ -0,0 +1,22 @@
+// version 20230105
+// public domain
+// djb
+
+#ifdef _MSC_VER
+#include <intrin.h>
+#else
+#include <x86intrin.h>
+#endif
+
+#include "cpucycles_internal.h"
+
+long long ticks(void)
+{
+  return __rdtsc();
+}
+
+long long ticks_setup(void)
+{
+  if (!cpucycles_works(ticks)) return cpucycles_SKIP;
+  return cpucycles_MAYBECYCLECOUNTER;
+}
@@ -0,0 +1,22 @@
+// version 20230105
+// public domain
+// djb
+
+#include "cpucycles_internal.h"
+
+#ifndef __i386__
+#error this code is only for 32-bit x86 platforms
+#endif
+
+long long ticks(void)
+{
+  long long result;
+  asm volatile(".byte 15;.byte 49" : "=A" (result));
+  return result;
+}
+
+long long ticks_setup(void)
+{
+  if (!cpucycles_works(ticks)) return cpucycles_SKIP;
+  return cpucycles_MAYBECYCLECOUNTER;
+}
@@ -0,0 +1,47 @@
+### NAME
+
+cpucycles - count CPU cycles
+
+### SYNOPSIS
+
+    #include <cpucycles.h>
+
+    long long count = cpucycles();
+    long long persecond = cpucycles_persecond();
+    const char *implementation = cpucycles_implementation();
+    const char *version = cpucycles_version();
+
+Link with `-lcpucycles`. Old systems may also need `-lrt`.
+
+### DESCRIPTION
+
+`cpucycles()` returns an estimate for the number of CPU cycles that have
+occurred since an unspecified time in the past (perhaps system boot,
+perhaps program startup).
+
+Accessing true cycle counters can be difficult on some CPUs and
+operating systems. `cpucycles()` does its best to produce accurate
+results, but selects a low-precision counter if the only other option is
+failure.
+
+`cpucycles_persecond()` returns an estimate for the number of CPU cycles
+per second. This estimate comes from `/etc/cpucyclespersecond` if that
+file exists, otherwise from various OS mechanisms, otherwise from the
+`cpucyclespersecond` environment variable if that is set, otherwise
+2399987654.
+
+`cpucycles_implementation()` returns the name of the counter in use:
+e.g., `"amd64-pmc"`.
+
+`cpucycles_version()` returns the `libcpucycles` version number as a
+string: e.g., `"20230115"`. Results of `cpucycles_implementation()`
+should be interpreted relative to `cpucycles_version()`.
+
+`cpucycles` is actually a function pointer. The first call to
+`cpucycles()` or `cpucycles_persecond()` or `cpucycles_implementation()`
+selects one of the available counters and updates the `cpucycles`
+pointer accordingly. Subsequent calls to `cpucycles()` are thread-safe.
+
+### SEE ALSO
+
+**gettimeofday**(2), **clock_gettime**(2)
@@ -0,0 +1,447 @@
+Currently libcpucycles supports the following cycle counters. Some
+cycle counters are actually other forms of counters that libcpucycles
+scales to imitate a cycle counter. There is
+[separate documentation](selection.html)
+for how libcpucycles makes a choice of cycle counter. See also
+[security considerations](security.html) regarding enabling or disabling
+counters and regarding Turbo Boost.
+
+`amd64-pmc`: Requires a 64-bit Intel/AMD platform. Requires the Linux
+perf_event interface. Accesses a cycle counter through RDPMC. Requires
+`/proc/sys/kernel/perf_event_paranoid` to be at most 2 for user-level
+RDPMC access. This counter runs at the clock frequency of the CPU core.
+
+`amd64-tsc`, `amd64-tscasm`: Requires a 64-bit Intel/AMD platform.
+Requires RDTSC to be enabled, which it is by default. Uses RDTSC to
+access the CPU's time-stamp counter. On current CPUs, this is an
+off-core clock rather than a cycle counter, but it is typically a very
+fast off-core clock, making it adequate for seeing cycle counts if
+overclocking and underclocking are disabled. The difference between
+`tsc` and `tscasm` is that `tsc` uses the compiler's `__rdtsc()` while
+`tscasm` uses inline assembly.
+
+`arm32-cortex`: Requires a 32-bit ARMv7-A platform. Uses
+`mrc p15, 0, %0, c9, c13, 0` to read the cycle counter. Requires user
+access to the cycle counter, which is not enabled by default but can be
+enabled under Linux via
+[a kernel module](https://github.com/thoughtpolice/enable_arm_pmu).
+This counter is natively 32 bits, but libcpucycles watches how the
+counter and `gettimeofday` increase to compute a 64-bit extension of the
+counter.
+
+`arm64-pmc`: Requires a 64-bit ARMv8-A platform. Uses
+`mrs %0, PMCCNTR_EL0` to read the cycle counter. Requires user access
+to the cycle counter, which is not enabled by default but can be enabled
+under Linux via
+[a kernel module](https://github.com/rdolbeau/enable_arm_pmu).
+
+`arm64-vct`: Requires a 64-bit ARMv8-A platform. Uses
+`mrs %0, CNTVCT_EL0` to read a "virtual count" timer. This is an
+off-core clock, typically running at 24MHz. Results are scaled by
+libcpucycles.
+
+`mips64-cc`: Requires a 64-bit MIPS platform. (Maybe the same code would
+also work as `mips32-cc`, but this has not been tested yet.) Uses RDHWR
+to read the hardware cycle counter (hardware register 2 times a constant
+scale factor in hardware register 3). This counter is natively 32 bits,
+but libcpucycles watches how the counter and `gettimeofday` increase to
+compute a 64-bit extension of the counter.
+
+`ppc32-mftb`: Requires a 32-bit PowerPC platform. Uses `mftb` and
+`mftbu` to read the "time base". This is an off-core clock, typically
+running at 24MHz.
+
+`ppc64-mftb`: Requires a 64-bit PowerPC platform. Uses `mftb` and
+`mftbu` to read the "time base". This is an off-core clock, typically
+running at 24MHz.
+
+`riscv32-rdcycle`: Requires a 32-bit RISC-V platform. Uses `rdcycle`
+and `rdcycleh` to read a cycle counter.
+
+`riscv64-rdcycle`: Requires a 64-bit RISC-V platform. Uses `rdcycle`
+to read a cycle counter.
+
+`s390x-stckf`: Requires a 64-bit z/Architecture platform. Uses `stckf`
+to read the TOD clock, which is documented to run at 4096MHz. On the
+z15, this looks like a doubling of an off-core 2048MHz clock. Results
+are scaled by libcpucycles.
+
+`sparc64-rdtick`: Requires a 64-bit SPARC platform. Uses `rd %tick`
+to read a cycle counter.
+
+`x86-tsc`, `x86-tscasm`: Same as `amd64-tsc` and `amd64-tscasm`, but
+for 32-bit Intel/AMD platforms instead of 64-bit Intel/AMD platforms.
+
+`default-gettimeofday`: Reasonably portable. Resolution is limited to 1
+microsecond. Results are scaled by libcpucycles.
+
+`default-mach`: Requires an OS with `mach_absolute_time()`. Typically
+runs at 24MHz. Results are scaled by libcpucycles.
+
+`default-monotonic`: Requires `CLOCK_MONOTONIC`. Reasonably portable,
+although might fail on older systems where `default-gettimeofday` works.
+Resolution is limited to 1 nanosecond. Can be almost as good as a cycle
+counter, or orders of magnitude worse, depending on the OS and CPU.
+Results are scaled by libcpucycles.
+
+`default-perfevent`: Requires the Linux `perf_event` interface, and a
+CPU where `perf_event` supports `PERF_COUNT_HW_CPU_CYCLES`. Similar
+variations in quality to `default-monotonic`, without the 1-nanosecond
+limitation.
+
+`default-zero`: The horrifying last resort if nothing else works.
+
+## Examples
+
+These are examples of `cpucycles-info` output on various machines. The
+machines named `gcc*` are from the
+[GCC Compile Farm](https://gcc.gnu.org/wiki/CompileFarm).
+
+A `median` line saying, e.g., `47 +47+28+0+2-5+0+2-5...` means that the
+differences between adjacent cycle counts were 47+47, 47+28, 47+0, 47+2,
+47−5, 47+0, 47+2, 47−5, etc., with median difference 47. The first few
+differences are typically larger because of cache effects.
+
+`pi3aplus`,
+Broadcom BCM2837B0:
+```
+cpucycles version 20230105
+cpucycles tracesetup 0 arm64-pmc precision 9 scaling 1.000000 only32 0
+cpucycles tracesetup 1 arm64-vct precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 2 default-perfevent precision 189 scaling 1.000000 only32 0
+cpucycles tracesetup 3 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 4 default-monotonic precision 272 scaling 1.400000 only32 0
+cpucycles tracesetup 5 default-gettimeofday precision 1600 scaling 1400.000000 only32 0
+cpucycles tracesetup 6 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 1400000000
+cpucycles implementation arm64-pmc
+cpucycles median 10 +10+8+3+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0
+cpucycles observed persecond 1032000000...4224666667 with 1024 loops 4 microseconds
+cpucycles observed persecond 1286000000...1756000000 with 2048 loops 7 microseconds
+cpucycles observed persecond 1368266666...1598000000 with 4096 loops 14 microseconds
+cpucycles observed persecond 1366700000...1473428572 with 8192 loops 29 microseconds
+cpucycles observed persecond 1366100000...1417534483 with 16384 loops 59 microseconds
+cpucycles observed persecond 1332739837...1357132232 with 32768 loops 122 microseconds
+cpucycles observed persecond 1354483471...1366945834 with 65536 loops 241 microseconds
+cpucycles observed persecond 1385684989...1392195330 with 131072 loops 472 microseconds
+cpucycles observed persecond 1347223021...1350328528 with 262144 loops 972 microseconds
+cpucycles observed persecond 1375460125...1377069853 with 524288 loops 1905 microseconds
+cpucycles observed persecond 1376527697...1377335961 with 1048576 loops 3808 microseconds
+```
+
+`bblack`,
+TI Sitara XAM3359AZCZ100:
+```
+cpucycles version 20230105
+cpucycles tracesetup 0 arm32-cortex precision 8 scaling 1.000000 only32 1
+cpucycles tracesetup 1 default-perfevent precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 2 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 3 default-monotonic precision 1283 scaling 1.000000 only32 0
+cpucycles tracesetup 4 default-gettimeofday precision 1200 scaling 1000.000000 only32 0
+cpucycles tracesetup 5 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 1000000000
+cpucycles implementation arm32-cortex
+cpucycles median 1260 +1506+62+31+7+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+13+7+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0
+cpucycles observed persecond 622181818...2101888889 with 1024 loops 10 microseconds
+cpucycles observed persecond 806133333...1492615385 with 2048 loops 14 microseconds
+cpucycles observed persecond 879880000...1232565218 with 4096 loops 24 microseconds
+cpucycles observed persecond 939577777...1130581396 with 8192 loops 44 microseconds
+cpucycles observed persecond 956954022...1050047059 with 16384 loops 86 microseconds
+cpucycles observed persecond 982878542...1020685715 with 32768 loops 246 microseconds
+cpucycles observed persecond 988105105...1012217523 with 65536 loops 332 microseconds
+cpucycles observed persecond 993752077...1007159723 with 131072 loops 721 microseconds
+cpucycles observed persecond 995364296...1004009448 with 262144 loops 1377 microseconds
+cpucycles observed persecond 998216306...1001821536 with 524288 loops 2685 microseconds
+cpucycles observed persecond 998991848...1000914196 with 1048576 loops 5397 microseconds
+```
+
+`hiphop`,
+Intel Xeon E3-1220 v3:
+```
+cpucycles version 20230105
+cpucycles tracesetup 0 amd64-pmc precision 40 scaling 1.000000 only32 0
+cpucycles tracesetup 1 amd64-tsc precision 124 scaling 1.000000 only32 0
+cpucycles tracesetup 2 amd64-tscasm precision 124 scaling 1.000000 only32 0
+cpucycles tracesetup 3 default-perfevent precision 160 scaling 1.000000 only32 0
+cpucycles tracesetup 4 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 5 default-monotonic precision 272 scaling 3.100000 only32 0
+cpucycles tracesetup 6 default-gettimeofday precision 3300 scaling 3100.000000 only32 0
+cpucycles tracesetup 7 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 3100000000
+cpucycles implementation amd64-pmc
+cpucycles median 44 +38+23+23+23-4+0-4+0-4+0-4+0+10-4-2+1-4+1-4+1+17+1-4+1-4+1-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4
+cpucycles observed persecond 2066500000...4235000000 with 8192 loops 3 microseconds
+cpucycles observed persecond 2760833333...4200250000 with 16384 loops 5 microseconds
+cpucycles observed persecond 2743416666...3313100000 with 32768 loops 11 microseconds
+cpucycles observed persecond 2986227272...3295000000 with 65536 loops 21 microseconds
+cpucycles observed persecond 3052069767...3206073171 with 131072 loops 42 microseconds
+cpucycles observed persecond 3050395348...3125523810 with 262144 loops 85 microseconds
+cpucycles observed persecond 3085123529...3123059524 with 524288 loops 169 microseconds
+cpucycles observed persecond 3084561764...3103434912 with 1048576 loops 339 microseconds
+```
+
+`nucnuc`,
+Intel Pentium N3700:
+```
+cpucycles version 20230105
+cpucycles tracesetup 0 amd64-pmc precision 26 scaling 1.000000 only32 0
+cpucycles tracesetup 1 amd64-tsc precision 120 scaling 1.000000 only32 0
+cpucycles tracesetup 2 amd64-tscasm precision 120 scaling 1.000000 only32 0
+cpucycles tracesetup 3 default-perfevent precision 427 scaling 1.000000 only32 0
+cpucycles tracesetup 4 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 5 default-monotonic precision 320 scaling 1.600000 only32 0
+cpucycles tracesetup 6 default-gettimeofday precision 1800 scaling 1600.000000 only32 0
+cpucycles tracesetup 7 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 1600000000
+cpucycles implementation amd64-pmc
+cpucycles median 66 +12+12+14+14-1-1+0-1+0-1+0-1+0+1-1+0-1+0-1+0-2+0-1+0-1+0-1+0-2+0-1+0-1+0-1+0-2+0-1+0-1+1-1+0-2-1-1+0-1+0-1+0-2+0-1+2+0-1+0-1+0+0-1
+cpucycles observed persecond 1060500000...2325000000 with 2048 loops 3 microseconds
+cpucycles observed persecond 1387166666...2208250000 with 4096 loops 5 microseconds
+cpucycles observed persecond 1376083333...1705500000 with 8192 loops 11 microseconds
+cpucycles observed persecond 1495727272...1671800000 with 16384 loops 21 microseconds
+cpucycles observed persecond 1563428571...1655100000 with 32768 loops 41 microseconds
+cpucycles observed persecond 1580807228...1626234568 with 65536 loops 82 microseconds
+cpucycles observed persecond 1589539393...1612619632 with 131072 loops 164 microseconds
+cpucycles observed persecond 1598841463...1610230062 with 262144 loops 327 microseconds
+cpucycles observed persecond 1564336810...1569988042 with 524288 loops 670 microseconds
+cpucycles observed persecond 1599759725...1602608098 with 1048576 loops 1310 microseconds
+```
+
+`saber214`,
+AMD FX-8350:
+```
+cpucycles version 20230105
+cpucycles tracesetup 0 amd64-pmc precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 1 amd64-tsc precision 167 scaling 1.000000 only32 0
+cpucycles tracesetup 2 amd64-tscasm precision 168 scaling 1.000000 only32 0
+cpucycles tracesetup 3 default-perfevent precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 4 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 5 default-monotonic precision 376 scaling 4.013452 only32 0
+cpucycles tracesetup 6 default-gettimeofday precision 4213 scaling 4013.452000 only32 0
+cpucycles tracesetup 7 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 4013452000
+cpucycles implementation amd64-tsc
+cpucycles median 77 +87-2+21+7+4+1+0+2-2-7-4+0+1+4-2+3+1-2-2+5-6+2+2+2+2+1-1-1+0-4+0-1-1-1-2+3-1-1+2-2+0+0+2+0+0+2-2-2+1-1-2+2-5+2+0+2+0+1+0+3-2-1-1
+cpucycles observed persecond 2767500000...5759000000 with 4096 loops 3 microseconds
+cpucycles observed persecond 3426000000...4893800000 with 8192 loops 6 microseconds
+cpucycles observed persecond 3724076923...4446363637 with 16384 loops 12 microseconds
+cpucycles observed persecond 3977833333...4363318182 with 32768 loops 23 microseconds
+cpucycles observed persecond 3984854166...4168739131 with 65536 loops 47 microseconds
+cpucycles observed persecond 3981709923...4048193799 with 131072 loops 130 microseconds
+cpucycles observed persecond 3982716417...4026914573 with 262144 loops 200 microseconds
+cpucycles observed persecond 4001637602...4025136987 with 524288 loops 366 microseconds
+cpucycles observed persecond 4007411111...4018600248 with 1048576 loops 809 microseconds
+```
+
+`gcc14`,
+Intel Xeon E5-2620 v3,
+Debian testing (bookworm),
+Linux kernel 6.0.0-6-amd64:
+```
+cpucycles version 20230105
+cpucycles tracesetup 0 amd64-pmc precision 41 scaling 1.000000 only32 0
+cpucycles tracesetup 1 amd64-tsc precision 148 scaling 1.000000 only32 0
+cpucycles tracesetup 2 amd64-tscasm precision 148 scaling 1.000000 only32 0
+cpucycles tracesetup 3 default-perfevent precision 159 scaling 1.000000 only32 0
+cpucycles tracesetup 4 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 5 default-monotonic precision 289 scaling 3.200000 only32 0
+cpucycles tracesetup 6 default-gettimeofday precision 3400 scaling 3200.000000 only32 0
+cpucycles tracesetup 7 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 3200000000
+cpucycles implementation amd64-pmc
+cpucycles median 47 +47+28+0+2-5+0+2-5+16+2-5+0+2-5+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0
+cpucycles observed persecond 1653800000...2819333334 with 8192 loops 4 microseconds
+cpucycles observed persecond 1832111111...2389285715 with 16384 loops 8 microseconds
+cpucycles observed persecond 1936058823...2207200000 with 32768 loops 16 microseconds
+cpucycles observed persecond 2052843750...2196200000 with 65536 loops 31 microseconds
+cpucycles observed persecond 2050750000...2120048388 with 131072 loops 63 microseconds
+cpucycles observed persecond 2081896825...2117048388 with 262144 loops 125 microseconds
+cpucycles observed persecond 2089478087...2107044177 with 524288 loops 250 microseconds
+cpucycles observed persecond 2093343313...2102124249 with 1048576 loops 500 microseconds
+```
+
+`gcc23`,
+Cavium Octeon II V0.1,
+Debian 8.11,
+Linux kernel 4.1.4:
+```
+cpucycles version 20230105
+cpucycles tracesetup 0 mips64-cc precision 24 scaling 1.000000 only32 1
+cpucycles tracesetup 1 default-perfevent precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 2 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 3 default-monotonic precision 46702 scaling 2.399988 only32 0
+cpucycles tracesetup 4 default-gettimeofday precision 45799 scaling 2399.987654 only32 0
+cpucycles tracesetup 5 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 2399987654
+cpucycles implementation mips64-cc
+cpucycles median 2177 +828+17+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0
+cpucycles observed persecond 641900000...1845125000 with 1024 loops 9 microseconds
+cpucycles observed persecond 745357142...1352083334 with 2048 loops 13 microseconds
+cpucycles observed persecond 809826086...1162333334 with 4096 loops 22 microseconds
+cpucycles observed persecond 897717948...1104405406 with 8192 loops 38 microseconds
+cpucycles observed persecond 957467532...1059986667 with 16384 loops 76 microseconds
+cpucycles observed persecond 973102189...1029777778 with 32768 loops 136 microseconds
+cpucycles observed persecond 986518656...1015830828 with 65536 loops 267 microseconds
+cpucycles observed persecond 993452830...1008166667 with 131072 loops 529 microseconds
+cpucycles observed persecond 996036966...1003403609 with 262144 loops 1054 microseconds
+cpucycles observed persecond 984706378...1001682630 with 524288 loops 2131 microseconds
+cpucycles observed persecond 992585292...1001178580 with 1048576 loops 4296 microseconds
+```
+
+`gcc45`,
+AMD Athlon II X4 640,
+Debian 8.11,
+Linux kernel 3.16.0-11-686-pae:
+```
+cpucycles version 20230105
+cpucycles tracesetup 0 x86-tsc precision 199 scaling 1.000000 only32 0
+cpucycles tracesetup 1 x86-tscasm precision 199 scaling 1.000000 only32 0
+cpucycles tracesetup 2 default-perfevent precision 170 scaling 1.000000 only32 0
+cpucycles tracesetup 3 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 4 default-monotonic precision 941 scaling 3.000000 only32 0
+cpucycles tracesetup 5 default-gettimeofday precision 3200 scaling 3000.000000 only32 0
+cpucycles tracesetup 6 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 3000000000
+cpucycles implementation default-perfevent
+cpucycles median 72 +12+0+0+0+0+0+0+0+5+0+0+0+0+0+0+0+2+0+0+0+0+0+0+0+1+0+0+0+0+0+0+0+2+0+0+0+0+0+0+0+1+0+0+0+0+0+0+0+2+0+0+0+0+0+0+0+1+0+0+0+0+0+0
+cpucycles observed persecond 541500000...1812000000 with 1024 loops 3 microseconds
+cpucycles observed persecond 712333333...1212250000 with 2048 loops 5 microseconds
+cpucycles observed persecond 1193285714...1733600000 with 4096 loops 6 microseconds
+cpucycles observed persecond 1689176470...1804562500 with 8192 loops 33 microseconds
+cpucycles observed persecond 1713074626...1770600000 with 16384 loops 66 microseconds
+cpucycles observed persecond 1765107692...1795140625 with 32768 loops 129 microseconds
+cpucycles observed persecond 1785369649...1800603922 with 65536 loops 256 microseconds
+cpucycles observed persecond 1781377862...1796288462 with 131072 loops 261 microseconds
+cpucycles observed persecond 1772647398...1778247827 with 262144 loops 691 microseconds
+cpucycles observed persecond 1789670493...1794149598 with 524288 loops 870 microseconds
+cpucycles observed persecond 1860276211...1861561332 with 1048576 loops 3156 microseconds
+```
+
+`gcc92`,
+SiFive Freedom U740,
+Ubuntu 22.04,
+Linux kernel 5.15.0-1014-generic:
+```
+cpucycles version 20230105
+cpucycles tracesetup 0 riscv64-rdcycle precision 8 scaling 1.000000 only32 0
+cpucycles tracesetup 1 default-perfevent precision 3024 scaling 1.000000 only32 0
+cpucycles tracesetup 2 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 3 default-monotonic precision 2599 scaling 2.399988 only32 0
+cpucycles tracesetup 4 default-gettimeofday precision 2599 scaling 2399.987654 only32 0
+cpucycles tracesetup 5 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 2399987654
+cpucycles implementation riscv64-rdcycle
+cpucycles median 8 +33+27+1+1+1+1+0+0+0+22+0+0+0+0+0+0+0+628+0+0+0+7+0+0+0+145+0+0+0+0+0+0+0+22+0+0+0+0+0+0+0+158+0+0+0+0+0+0+0+22+0+0+0+0+0+0+0+22+0+0+0+0+0
+cpucycles observed persecond 530250000...1978000000 with 1024 loops 3 microseconds
+cpucycles observed persecond 831000000...1915666667 with 2048 loops 4 microseconds
+cpucycles observed persecond 1055750000...1689500000 with 4096 loops 7 microseconds
+cpucycles observed persecond 1045562500...1305428572 with 8192 loops 15 microseconds
+cpucycles observed persecond 1102700000...1236357143 with 16384 loops 29 microseconds
+cpucycles observed persecond 1176053571...1247444445 with 32768 loops 55 microseconds
+cpucycles observed persecond 1173321428...1209127273 with 65536 loops 111 microseconds
+cpucycles observed persecond 1187805429...1205210046 with 131072 loops 220 microseconds
+cpucycles observed persecond 1192415909...1201157535 with 262144 loops 439 microseconds
+cpucycles observed persecond 1194694760...1199247717 with 524288 loops 877 microseconds
+cpucycles observed persecond 1194656004...1197023034 with 1048576 loops 1781 microseconds
+```
+
+`gcc103`,
+Apple M1 (Icestorm-M1 + Firestorm-M1),
+Debian unstable (bookworm),
+Linux kernel 6.0.0-rc5-asahi-00001-gc62bd3fe430f:
+```
+cpucycles version 20230105
+cpucycles tracesetup 0 arm64-pmc precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 1 arm64-vct precision 186 scaling 86.000000 only32 0
+cpucycles tracesetup 2 default-perfevent precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 3 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 4 default-monotonic precision 285 scaling 2.064000 only32 0
+cpucycles tracesetup 5 default-gettimeofday precision 2264 scaling 2064.000000 only32 0
+cpucycles tracesetup 6 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 2064000000
+cpucycles implementation arm64-vct
+cpucycles median 0 +0+86+0+0+0+0+0+0+0+0+0+0+0+0+86+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+86+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+86+0+0+0+0+0+0+0+0
+cpucycles observed persecond 1784500000...3655000000 with 8192 loops 3 microseconds
+cpucycles observed persecond 1773750000...2393666667 with 16384 loops 7 microseconds
+cpucycles observed persecond 1897733333...2222769231 with 32768 loops 14 microseconds
+cpucycles observed persecond 1951310344...2114962963 with 65536 loops 28 microseconds
+cpucycles observed persecond 2024071428...2107000000 with 131072 loops 55 microseconds
+cpucycles observed persecond 2041531531...2082935780 with 262144 loops 110 microseconds
+cpucycles observed persecond 2051158371...2071461188 with 524288 loops 220 microseconds
+cpucycles observed persecond 2058539682...2068309795 with 1048576 loops 440 microseconds
+```
+
+`gcc112` (`gcc2-power8`),
+IBM POWER8E,
+CentOS 7.9 AltArch,
+Linux kernel 3.10.0-1127.13.1.el7.ppc64le:
+```
+cpucycles version 20230105
+cpucycles tracesetup 0 ppc64-mftb precision 251 scaling 7.207031 only32 0
+cpucycles tracesetup 1 default-perfevent precision 295 scaling 1.000000 only32 0
+cpucycles tracesetup 2 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 3 default-monotonic precision 536 scaling 3.690000 only32 0
+cpucycles tracesetup 4 default-gettimeofday precision 3890 scaling 3690.000000 only32 0
+cpucycles tracesetup 5 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 3690000000
+cpucycles implementation ppc64-mftb
+cpucycles median 195 +2969-8+14+0-8+7-8-7+7+6-7-1+0-1+0+7+7-15+7-1-7+6+0+0-8+0+6+0-8+7+0+7-8-8-7-1+7-8+7+0-8+0+14-8-7+6+0-8+7+7-15+0-1+0-1+14+0-15+14+0-1+7+0
+cpucycles observed persecond 2603750000...5510000000 with 2048 loops 3 microseconds
+cpucycles observed persecond 3430500000...6052250000 with 4096 loops 5 microseconds
+cpucycles observed persecond 3411333333...4457500000 with 8192 loops 11 microseconds
+cpucycles observed persecond 3548695652...4060333334 with 16384 loops 22 microseconds
+cpucycles observed persecond 3624977777...3876534884 with 32768 loops 44 microseconds
+cpucycles observed persecond 3621855555...3745363637 with 65536 loops 89 microseconds
+cpucycles observed persecond 3660157303...3722227273 with 131072 loops 177 microseconds
+cpucycles observed persecond 3680471751...3711622160 with 262144 loops 353 microseconds
+cpucycles observed persecond 3685321074...3700886525 with 524288 loops 706 microseconds
+cpucycles observed persecond 3687745930...3695537208 with 1048576 loops 1412 microseconds
+```
+
+`gcc202`,
+UltraSparc T5,
+Debian unstable (bookworm),
+Linux kernel 5.19.0-2-sparc64-smp:
+```
+cpucycles version 20230105
+cpucycles tracesetup 0 sparc64-rdtick precision 65 scaling 1.000000 only32 0
+cpucycles tracesetup 1 default-perfevent precision 386 scaling 1.000000 only32 0
+cpucycles tracesetup 2 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 3 default-monotonic precision 442 scaling 3.599910 only32 0
+cpucycles tracesetup 4 default-gettimeofday precision 3799 scaling 3599.910000 only32 0
+cpucycles tracesetup 5 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 3599910000
+cpucycles implementation sparc64-rdtick
+cpucycles median 73 +24+0+24+24+24+24+24+24+0+1+24+0+1+24+0+1+24+0+0+1+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+1+0+0+0+0+0+0+0+0+0+0+0+0+0
+cpucycles observed persecond 2751500000...4258250000 with 4096 loops 5 microseconds
+cpucycles observed persecond 3289200000...4206875000 with 8192 loops 9 microseconds
+cpucycles observed persecond 3454789473...3900823530 with 16384 loops 18 microseconds
+cpucycles observed persecond 3452026315...3659888889 with 32768 loops 37 microseconds
+cpucycles observed persecond 3543770270...3650916667 with 65536 loops 73 microseconds
+cpucycles observed persecond 3567299319...3620662069 with 131072 loops 146 microseconds
+cpucycles observed persecond 3591373287...3618220690 with 262144 loops 291 microseconds
+cpucycles observed persecond 3597353344...3610774527 with 524288 loops 582 microseconds
+cpucycles observed persecond 3595899403...3603058071 with 1048576 loops 1172 microseconds
+```
+
+IBM z15:
+```
+cpucycles version 20230106
+cpucycles tracesetup 0 s390x-stckf precision 250 scaling 1.269531 only32 0
+cpucycles tracesetup 1 default-perfevent precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 2 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 3 default-monotonic precision 272 scaling 5.200000 only32 0
+cpucycles tracesetup 4 default-gettimeofday precision 5400 scaling 5200.000000 only32 0
+cpucycles tracesetup 5 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 5200000000
+cpucycles implementation s390x-stckf
+cpucycles median 48 +87+8+0-2+0+0+38-2+0+1-3+1+28+0+3-3+1+0+28+0-2+3+0-2+36+0+0+0+1+0+28+0-2+0+3-2+35+1+0-2+0+3+28+0-2+0+0-2+3+25+3+0-2+0+1+35+1+0+0-2+0+28+0
+cpucycles observed persecond 4948941176...5627733334 with 8192 loops 16 microseconds
+cpucycles observed persecond 4104125000...5515666667 with 16384 loops 7 microseconds
+cpucycles observed persecond 5047076923...5987818182 with 32768 loops 12 microseconds
+cpucycles observed persecond 5044846153...5475708334 with 65536 loops 25 microseconds
+cpucycles observed persecond 5141313725...5357428572 with 131072 loops 50 microseconds
+cpucycles observed persecond 5150892156...5257250000 with 262144 loops 101 microseconds
+cpucycles observed persecond 5183421568...5236549505 with 524288 loops 203 microseconds
+cpucycles observed persecond 5190282555...5216582717 with 1048576 loops 406 microseconds
+```
@@ -0,0 +1,30 @@
+To download and unpack the latest version of libcpucycles:
+
+        wget -m https://cpucycles.cr.yp.to/libcpucycles-latest-version.txt
+        version=$(cat cpucycles.cr.yp.to/libcpucycles-latest-version.txt)
+        wget -m https://cpucycles.cr.yp.to/libcpucycles-$version.tar.gz
+        tar -xzf cpucycles.cr.yp.to/libcpucycles-$version.tar.gz
+        cd libcpucycles-$version
+
+Then [install](install.html).
+
+### Archives and changelog (reverse chronological)
+
+[`libcpucycles-20230115.tar.gz`](libcpucycles-20230115.tar.gz) [browse](libcpucycles-20230115.html)
+
+Update actual `cpucycles_version` behavior to match documentation.
+
+[`libcpucycles-20230110.tar.gz`](libcpucycles-20230110.tar.gz) [browse](libcpucycles-20230110.html)
+
+`doc/api.md`: Document `cpucycles_version()`.
+
+Add `s390x-stckf` counter.
+
+`cpucycles/default-perfevent.c`: Read into `int64_t` instead of `long long`.
+Add comment explaining issues with `PERF_FORMAT_TOTAL_TIME_RUNNING`.
+
+`configure`: Improve `uname` handling.
+
+`doc/api.md`: Update description of default frequency.
+
+[`libcpucycles-20230105.tar.gz`](libcpucycles-20230105.tar.gz) [browse](libcpucycles-20230105.html)
@@ -0,0 +1,91 @@
+<html>
+<head>
+<meta http-equiv="content-type" content="text/html; charset=utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<style type="text/css">
+html{overflow-y:scroll}
+body{font-family:sans-serif}
+p,ul,ol,blockquote,pre{font-size:0.9em;line-height:1.6em}
+li p{font-size:1.0em}
+blockquote p{font-size:1.0em}
+tt{font-size:1.2em}
+code{font-size:1.2em}
+h1{font-size:1.5em}
+h2{font-size:1.3em}
+h3{font-size:1.0em}
+h1 a{text-decoration:none}
+table{border-collapse:collapse}
+th,td{border:1px solid black}
+table a{text-decoration:none}
+table tr{font-size:0.9em;line-height:1.6em}
+.links a:hover{text-decoration:underline}
+.links a:active{text-decoration:underline}
+.links img{width:200px;padding-left:1em}
+.links td{border:0px;padding-top:0.5em;padding-bottom:0.5em}
+.headline{padding:0;font-weight:bold;font-size:1.5em;vertical-align:top;padding-bottom:0.5em;color:#125d0d}
+.navt{display:inline-block;box-sizing:border-box;-moz-box-sizing:border-box;-webkit-box-sizing:border-box;
+min-width:14%;margin:0;padding:0;padding-left:0.5em;padding-right:0.5em;vertical-align:center;
+font-weight:bold;font-size:1.1em;text-align:center;border:1px solid black}
+.here{border-bottom:0px;background-color:#ffffff}
+.away{background-color:#125d0d;}
+.away a{text-decoration:none;display:block;color:#ffffff}
+.away a:hover,.away a:active{text-decoration:underline}
+.main{margin:0;padding-top:0em;padding-bottom:1%;clear:both}
+</style>
+<title>
+API</title>
+</head>
+<body>
+<div class=headline>
+libcpucycles
+</div>
+<div class=nav>
+<div class="navt away"><a href=index.html>Intro</a>
+</div><div class="navt away"><a href=download.html>Download</a>
+</div><div class="navt away"><a href=install.html>Install</a>
+</div><div class="navt here">API
+</div><div class="navt away"><a href=counters.html>Counters</a>
+</div><div class="navt away"><a href=selection.html>Selection</a>
+</div><div class="navt away"><a href=security.html>Security</a>
+</div></div>
+<div class=main>
+<h3>NAME</h3>
+<p>cpucycles - count CPU cycles</p>
+<h3>SYNOPSIS</h3>
+<pre><code>#include &lt;cpucycles.h&gt;
+
+long long count = cpucycles();
+long long persecond = cpucycles_persecond();
+const char *implementation = cpucycles_implementation();
+const char *version = cpucycles_version();
+</code></pre>
+<p>Link with <code>-lcpucycles</code>. Old systems may also need <code>-lrt</code>.</p>
+<h3>DESCRIPTION</h3>
+<p><code>cpucycles()</code> returns an estimate for the number of CPU cycles that have
+occurred since an unspecified time in the past (perhaps system boot,
+perhaps program startup).</p>
+<p>Accessing true cycle counters can be difficult on some CPUs and
+operating systems. <code>cpucycles()</code> does its best to produce accurate
+results, but selects a low-precision counter if the only other option is
+failure.</p>
+<p><code>cpucycles_persecond()</code> returns an estimate for the number of CPU cycles
+per second. This estimate comes from <code>/etc/cpucyclespersecond</code> if that
+file exists, otherwise from various OS mechanisms, otherwise from the
+<code>cpucyclespersecond</code> environment variable if that is set, otherwise
+2399987654.</p>
+<p><code>cpucycles_implementation()</code> returns the name of the counter in use:
+e.g., <code>"amd64-pmc"</code>.</p>
+<p><code>cpucycles_version()</code> returns the <code>libcpucycles</code> version number as a
+string: e.g., <code>"20230115"</code>. Results of <code>cpucycles_implementation()</code>
+should be interpreted relative to <code>cpucycles_version()</code>.</p>
+<p><code>cpucycles</code> is actually a function pointer. The first call to
+<code>cpucycles()</code> or <code>cpucycles_persecond()</code> or <code>cpucycles_implementation()</code>
+selects one of the available counters and updates the <code>cpucycles</code>
+pointer accordingly. Subsequent calls to <code>cpucycles()</code> are thread-safe.</p>
+<h3>SEE ALSO</h3>
+<p><strong>gettimeofday</strong>(2), <strong>clock_gettime</strong>(2)</p><hr><font size=1><b>Version:</b>
+This is version 2023.01.15 of the "API" web page.
+</font>
+</div>
+</body>
+</html>
@@ -0,0 +1,456 @@
+<html>
+<head>
+<meta http-equiv="content-type" content="text/html; charset=utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<style type="text/css">
+html{overflow-y:scroll}
+body{font-family:sans-serif}
+p,ul,ol,blockquote,pre{font-size:0.9em;line-height:1.6em}
+li p{font-size:1.0em}
+blockquote p{font-size:1.0em}
+tt{font-size:1.2em}
+code{font-size:1.2em}
+h1{font-size:1.5em}
+h2{font-size:1.3em}
+h3{font-size:1.0em}
+h1 a{text-decoration:none}
+table{border-collapse:collapse}
+th,td{border:1px solid black}
+table a{text-decoration:none}
+table tr{font-size:0.9em;line-height:1.6em}
+.links a:hover{text-decoration:underline}
+.links a:active{text-decoration:underline}
+.links img{width:200px;padding-left:1em}
+.links td{border:0px;padding-top:0.5em;padding-bottom:0.5em}
+.headline{padding:0;font-weight:bold;font-size:1.5em;vertical-align:top;padding-bottom:0.5em;color:#125d0d}
+.navt{display:inline-block;box-sizing:border-box;-moz-box-sizing:border-box;-webkit-box-sizing:border-box;
+min-width:14%;margin:0;padding:0;padding-left:0.5em;padding-right:0.5em;vertical-align:center;
+font-weight:bold;font-size:1.1em;text-align:center;border:1px solid black}
+.here{border-bottom:0px;background-color:#ffffff}
+.away{background-color:#125d0d;}
+.away a{text-decoration:none;display:block;color:#ffffff}
+.away a:hover,.away a:active{text-decoration:underline}
+.main{margin:0;padding-top:0em;padding-bottom:1%;clear:both}
+</style>
+<title>
+Counters</title>
+</head>
+<body>
+<div class=headline>
+libcpucycles
+</div>
+<div class=nav>
+<div class="navt away"><a href=index.html>Intro</a>
+</div><div class="navt away"><a href=download.html>Download</a>
+</div><div class="navt away"><a href=install.html>Install</a>
+</div><div class="navt away"><a href=api.html>API</a>
+</div><div class="navt here">Counters
+</div><div class="navt away"><a href=selection.html>Selection</a>
+</div><div class="navt away"><a href=security.html>Security</a>
+</div></div>
+<div class=main>
+<p>Currently libcpucycles supports the following cycle counters. Some
+cycle counters are actually other forms of counters that libcpucycles
+scales to imitate a cycle counter. There is
+<a href="selection.html">separate documentation</a>
+for how libcpucycles makes a choice of cycle counter. See also
+<a href="security.html">security considerations</a> regarding enabling or disabling
+counters and regarding Turbo Boost.</p>
+<p><code>amd64-pmc</code>: Requires a 64-bit Intel/AMD platform. Requires the Linux
+perf_event interface. Accesses a cycle counter through RDPMC. Requires
+<code>/proc/sys/kernel/perf_event_paranoid</code> to be at most 2 for user-level
+RDPMC access. This counter runs at the clock frequency of the CPU core.</p>
+<p><code>amd64-tsc</code>, <code>amd64-tscasm</code>: Requires a 64-bit Intel/AMD platform.
+Requires RDTSC to be enabled, which it is by default. Uses RDTSC to
+access the CPU's time-stamp counter. On current CPUs, this is an
+off-core clock rather than a cycle counter, but it is typically a very
+fast off-core clock, making it adequate for seeing cycle counts if
+overclocking and underclocking are disabled. The difference between
+<code>tsc</code> and <code>tscasm</code> is that <code>tsc</code> uses the compiler's <code>__rdtsc()</code> while
+<code>tscasm</code> uses inline assembly.</p>
+<p><code>arm32-cortex</code>: Requires a 32-bit ARMv7-A platform. Uses
+<code>mrc p15, 0, %0, c9, c13, 0</code> to read the cycle counter. Requires user
+access to the cycle counter, which is not enabled by default but can be
+enabled under Linux via
+<a href="https://github.com/thoughtpolice/enable_arm_pmu">a kernel module</a>.
+This counter is natively 32 bits, but libcpucycles watches how the
+counter and <code>gettimeofday</code> increase to compute a 64-bit extension of the
+counter.</p>
+<p><code>arm64-pmc</code>: Requires a 64-bit ARMv8-A platform. Uses
+<code>mrs %0, PMCCNTR_EL0</code> to read the cycle counter. Requires user access
+to the cycle counter, which is not enabled by default but can be enabled
+under Linux via
+<a href="https://github.com/rdolbeau/enable_arm_pmu">a kernel module</a>.</p>
+<p><code>arm64-vct</code>: Requires a 64-bit ARMv8-A platform. Uses
+<code>mrs %0, CNTVCT_EL0</code> to read a "virtual count" timer. This is an
+off-core clock, typically running at 24MHz. Results are scaled by
+libcpucycles.</p>
+<p><code>mips64-cc</code>: Requires a 64-bit MIPS platform. (Maybe the same code would
+also work as <code>mips32-cc</code>, but this has not been tested yet.) Uses RDHWR
+to read the hardware cycle counter (hardware register 2 times a constant
+scale factor in hardware register 3). This counter is natively 32 bits,
+but libcpucycles watches how the counter and <code>gettimeofday</code> increase to
+compute a 64-bit extension of the counter.</p>
+<p><code>ppc32-mftb</code>: Requires a 32-bit PowerPC platform. Uses <code>mftb</code> and
+<code>mftbu</code> to read the "time base". This is an off-core clock, typically
+running at 24MHz.</p>
+<p><code>ppc64-mftb</code>: Requires a 64-bit PowerPC platform. Uses <code>mftb</code> and
+<code>mftbu</code> to read the "time base". This is an off-core clock, typically
+running at 24MHz.</p>
+<p><code>riscv32-rdcycle</code>: Requires a 32-bit RISC-V platform. Uses <code>rdcycle</code>
+and <code>rdcycleh</code> to read a cycle counter.</p>
+<p><code>riscv64-rdcycle</code>: Requires a 64-bit RISC-V platform. Uses <code>rdcycle</code>
+to read a cycle counter.</p>
+<p><code>s390x-stckf</code>: Requires a 64-bit z/Architecture platform. Uses <code>stckf</code>
+to read the TOD clock, which is documented to run at 4096MHz. On the
+z15, this looks like a doubling of an off-core 2048MHz clock. Results
+are scaled by libcpucycles.</p>
+<p><code>sparc64-rdtick</code>: Requires a 64-bit SPARC platform. Uses <code>rd %tick</code>
+to read a cycle counter.</p>
+<p><code>x86-tsc</code>, <code>x86-tscasm</code>: Same as <code>amd64-tsc</code> and <code>amd64-tscasm</code>, but
+for 32-bit Intel/AMD platforms instead of 64-bit Intel/AMD platforms.</p>
+<p><code>default-gettimeofday</code>: Reasonably portable. Resolution is limited to 1
+microsecond. Results are scaled by libcpucycles.</p>
+<p><code>default-mach</code>: Requires an OS with <code>mach_absolute_time()</code>. Typically
+runs at 24MHz. Results are scaled by libcpucycles.</p>
+<p><code>default-monotonic</code>: Requires <code>CLOCK_MONOTONIC</code>. Reasonably portable,
+although might fail on older systems where <code>default-gettimeofday</code> works.
+Resolution is limited to 1 nanosecond. Can be almost as good as a cycle
+counter, or orders of magnitude worse, depending on the OS and CPU.
+Results are scaled by libcpucycles.</p>
+<p><code>default-perfevent</code>: Requires the Linux <code>perf_event</code> interface, and a
+CPU where <code>perf_event</code> supports <code>PERF_COUNT_HW_CPU_CYCLES</code>. Similar
+variations in quality to <code>default-monotonic</code>, without the 1-nanosecond
+limitation.</p>
+<p><code>default-zero</code>: The horrifying last resort if nothing else works.</p>
+<h2>Examples</h2>
+<p>These are examples of <code>cpucycles-info</code> output on various machines. The
+machines named <code>gcc*</code> are from the
+<a href="https://gcc.gnu.org/wiki/CompileFarm">GCC Compile Farm</a>.</p>
+<p>A <code>median</code> line saying, e.g., <code>47 +47+28+0+2-5+0+2-5...</code> means that the
+differences between adjacent cycle counts were 47+47, 47+28, 47+0, 47+2,
+47−5, 47+0, 47+2, 47−5, etc., with median difference 47. The first few
+differences are typically larger because of cache effects.</p>
+<p><code>pi3aplus</code>,
+Broadcom BCM2837B0:</p>
+<pre><code>cpucycles version 20230105
+cpucycles tracesetup 0 arm64-pmc precision 9 scaling 1.000000 only32 0
+cpucycles tracesetup 1 arm64-vct precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 2 default-perfevent precision 189 scaling 1.000000 only32 0
+cpucycles tracesetup 3 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 4 default-monotonic precision 272 scaling 1.400000 only32 0
+cpucycles tracesetup 5 default-gettimeofday precision 1600 scaling 1400.000000 only32 0
+cpucycles tracesetup 6 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 1400000000
+cpucycles implementation arm64-pmc
+cpucycles median 10 +10+8+3+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0
+cpucycles observed persecond 1032000000...4224666667 with 1024 loops 4 microseconds
+cpucycles observed persecond 1286000000...1756000000 with 2048 loops 7 microseconds
+cpucycles observed persecond 1368266666...1598000000 with 4096 loops 14 microseconds
+cpucycles observed persecond 1366700000...1473428572 with 8192 loops 29 microseconds
+cpucycles observed persecond 1366100000...1417534483 with 16384 loops 59 microseconds
+cpucycles observed persecond 1332739837...1357132232 with 32768 loops 122 microseconds
+cpucycles observed persecond 1354483471...1366945834 with 65536 loops 241 microseconds
+cpucycles observed persecond 1385684989...1392195330 with 131072 loops 472 microseconds
+cpucycles observed persecond 1347223021...1350328528 with 262144 loops 972 microseconds
+cpucycles observed persecond 1375460125...1377069853 with 524288 loops 1905 microseconds
+cpucycles observed persecond 1376527697...1377335961 with 1048576 loops 3808 microseconds
+</code></pre>
+<p><code>bblack</code>,
+TI Sitara XAM3359AZCZ100:</p>
+<pre><code>cpucycles version 20230105
+cpucycles tracesetup 0 arm32-cortex precision 8 scaling 1.000000 only32 1
+cpucycles tracesetup 1 default-perfevent precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 2 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 3 default-monotonic precision 1283 scaling 1.000000 only32 0
+cpucycles tracesetup 4 default-gettimeofday precision 1200 scaling 1000.000000 only32 0
+cpucycles tracesetup 5 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 1000000000
+cpucycles implementation arm32-cortex
+cpucycles median 1260 +1506+62+31+7+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+13+7+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0
+cpucycles observed persecond 622181818...2101888889 with 1024 loops 10 microseconds
+cpucycles observed persecond 806133333...1492615385 with 2048 loops 14 microseconds
+cpucycles observed persecond 879880000...1232565218 with 4096 loops 24 microseconds
+cpucycles observed persecond 939577777...1130581396 with 8192 loops 44 microseconds
+cpucycles observed persecond 956954022...1050047059 with 16384 loops 86 microseconds
+cpucycles observed persecond 982878542...1020685715 with 32768 loops 246 microseconds
+cpucycles observed persecond 988105105...1012217523 with 65536 loops 332 microseconds
+cpucycles observed persecond 993752077...1007159723 with 131072 loops 721 microseconds
+cpucycles observed persecond 995364296...1004009448 with 262144 loops 1377 microseconds
+cpucycles observed persecond 998216306...1001821536 with 524288 loops 2685 microseconds
+cpucycles observed persecond 998991848...1000914196 with 1048576 loops 5397 microseconds
+</code></pre>
+<p><code>hiphop</code>,
+Intel Xeon E3-1220 v3:</p>
+<pre><code>cpucycles version 20230105
+cpucycles tracesetup 0 amd64-pmc precision 40 scaling 1.000000 only32 0
+cpucycles tracesetup 1 amd64-tsc precision 124 scaling 1.000000 only32 0
+cpucycles tracesetup 2 amd64-tscasm precision 124 scaling 1.000000 only32 0
+cpucycles tracesetup 3 default-perfevent precision 160 scaling 1.000000 only32 0
+cpucycles tracesetup 4 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 5 default-monotonic precision 272 scaling 3.100000 only32 0
+cpucycles tracesetup 6 default-gettimeofday precision 3300 scaling 3100.000000 only32 0
+cpucycles tracesetup 7 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 3100000000
+cpucycles implementation amd64-pmc
+cpucycles median 44 +38+23+23+23-4+0-4+0-4+0-4+0+10-4-2+1-4+1-4+1+17+1-4+1-4+1-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4+0-4
+cpucycles observed persecond 2066500000...4235000000 with 8192 loops 3 microseconds
+cpucycles observed persecond 2760833333...4200250000 with 16384 loops 5 microseconds
+cpucycles observed persecond 2743416666...3313100000 with 32768 loops 11 microseconds
+cpucycles observed persecond 2986227272...3295000000 with 65536 loops 21 microseconds
+cpucycles observed persecond 3052069767...3206073171 with 131072 loops 42 microseconds
+cpucycles observed persecond 3050395348...3125523810 with 262144 loops 85 microseconds
+cpucycles observed persecond 3085123529...3123059524 with 524288 loops 169 microseconds
+cpucycles observed persecond 3084561764...3103434912 with 1048576 loops 339 microseconds
+</code></pre>
+<p><code>nucnuc</code>,
+Intel Pentium N3700:</p>
+<pre><code>cpucycles version 20230105
+cpucycles tracesetup 0 amd64-pmc precision 26 scaling 1.000000 only32 0
+cpucycles tracesetup 1 amd64-tsc precision 120 scaling 1.000000 only32 0
+cpucycles tracesetup 2 amd64-tscasm precision 120 scaling 1.000000 only32 0
+cpucycles tracesetup 3 default-perfevent precision 427 scaling 1.000000 only32 0
+cpucycles tracesetup 4 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 5 default-monotonic precision 320 scaling 1.600000 only32 0
+cpucycles tracesetup 6 default-gettimeofday precision 1800 scaling 1600.000000 only32 0
+cpucycles tracesetup 7 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 1600000000
+cpucycles implementation amd64-pmc
+cpucycles median 66 +12+12+14+14-1-1+0-1+0-1+0-1+0+1-1+0-1+0-1+0-2+0-1+0-1+0-1+0-2+0-1+0-1+0-1+0-2+0-1+0-1+1-1+0-2-1-1+0-1+0-1+0-2+0-1+2+0-1+0-1+0+0-1
+cpucycles observed persecond 1060500000...2325000000 with 2048 loops 3 microseconds
+cpucycles observed persecond 1387166666...2208250000 with 4096 loops 5 microseconds
+cpucycles observed persecond 1376083333...1705500000 with 8192 loops 11 microseconds
+cpucycles observed persecond 1495727272...1671800000 with 16384 loops 21 microseconds
+cpucycles observed persecond 1563428571...1655100000 with 32768 loops 41 microseconds
+cpucycles observed persecond 1580807228...1626234568 with 65536 loops 82 microseconds
+cpucycles observed persecond 1589539393...1612619632 with 131072 loops 164 microseconds
+cpucycles observed persecond 1598841463...1610230062 with 262144 loops 327 microseconds
+cpucycles observed persecond 1564336810...1569988042 with 524288 loops 670 microseconds
+cpucycles observed persecond 1599759725...1602608098 with 1048576 loops 1310 microseconds
+</code></pre>
+<p><code>saber214</code>,
+AMD FX-8350:</p>
+<pre><code>cpucycles version 20230105
+cpucycles tracesetup 0 amd64-pmc precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 1 amd64-tsc precision 167 scaling 1.000000 only32 0
+cpucycles tracesetup 2 amd64-tscasm precision 168 scaling 1.000000 only32 0
+cpucycles tracesetup 3 default-perfevent precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 4 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 5 default-monotonic precision 376 scaling 4.013452 only32 0
+cpucycles tracesetup 6 default-gettimeofday precision 4213 scaling 4013.452000 only32 0
+cpucycles tracesetup 7 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 4013452000
+cpucycles implementation amd64-tsc
+cpucycles median 77 +87-2+21+7+4+1+0+2-2-7-4+0+1+4-2+3+1-2-2+5-6+2+2+2+2+1-1-1+0-4+0-1-1-1-2+3-1-1+2-2+0+0+2+0+0+2-2-2+1-1-2+2-5+2+0+2+0+1+0+3-2-1-1
+cpucycles observed persecond 2767500000...5759000000 with 4096 loops 3 microseconds
+cpucycles observed persecond 3426000000...4893800000 with 8192 loops 6 microseconds
+cpucycles observed persecond 3724076923...4446363637 with 16384 loops 12 microseconds
+cpucycles observed persecond 3977833333...4363318182 with 32768 loops 23 microseconds
+cpucycles observed persecond 3984854166...4168739131 with 65536 loops 47 microseconds
+cpucycles observed persecond 3981709923...4048193799 with 131072 loops 130 microseconds
+cpucycles observed persecond 3982716417...4026914573 with 262144 loops 200 microseconds
+cpucycles observed persecond 4001637602...4025136987 with 524288 loops 366 microseconds
+cpucycles observed persecond 4007411111...4018600248 with 1048576 loops 809 microseconds
+</code></pre>
+<p><code>gcc14</code>,
+Intel Xeon E5-2620 v3,
+Debian testing (bookworm),
+Linux kernel 6.0.0-6-amd64:</p>
+<pre><code>cpucycles version 20230105
+cpucycles tracesetup 0 amd64-pmc precision 41 scaling 1.000000 only32 0
+cpucycles tracesetup 1 amd64-tsc precision 148 scaling 1.000000 only32 0
+cpucycles tracesetup 2 amd64-tscasm precision 148 scaling 1.000000 only32 0
+cpucycles tracesetup 3 default-perfevent precision 159 scaling 1.000000 only32 0
+cpucycles tracesetup 4 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 5 default-monotonic precision 289 scaling 3.200000 only32 0
+cpucycles tracesetup 6 default-gettimeofday precision 3400 scaling 3200.000000 only32 0
+cpucycles tracesetup 7 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 3200000000
+cpucycles implementation amd64-pmc
+cpucycles median 47 +47+28+0+2-5+0+2-5+16+2-5+0+2-5+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0+1-4+0
+cpucycles observed persecond 1653800000...2819333334 with 8192 loops 4 microseconds
+cpucycles observed persecond 1832111111...2389285715 with 16384 loops 8 microseconds
+cpucycles observed persecond 1936058823...2207200000 with 32768 loops 16 microseconds
+cpucycles observed persecond 2052843750...2196200000 with 65536 loops 31 microseconds
+cpucycles observed persecond 2050750000...2120048388 with 131072 loops 63 microseconds
+cpucycles observed persecond 2081896825...2117048388 with 262144 loops 125 microseconds
+cpucycles observed persecond 2089478087...2107044177 with 524288 loops 250 microseconds
+cpucycles observed persecond 2093343313...2102124249 with 1048576 loops 500 microseconds
+</code></pre>
+<p><code>gcc23</code>,
+Cavium Octeon II V0.1,
+Debian 8.11,
+Linux kernel 4.1.4:</p>
+<pre><code>cpucycles version 20230105
+cpucycles tracesetup 0 mips64-cc precision 24 scaling 1.000000 only32 1
+cpucycles tracesetup 1 default-perfevent precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 2 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 3 default-monotonic precision 46702 scaling 2.399988 only32 0
+cpucycles tracesetup 4 default-gettimeofday precision 45799 scaling 2399.987654 only32 0
+cpucycles tracesetup 5 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 2399987654
+cpucycles implementation mips64-cc
+cpucycles median 2177 +828+17+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0
+cpucycles observed persecond 641900000...1845125000 with 1024 loops 9 microseconds
+cpucycles observed persecond 745357142...1352083334 with 2048 loops 13 microseconds
+cpucycles observed persecond 809826086...1162333334 with 4096 loops 22 microseconds
+cpucycles observed persecond 897717948...1104405406 with 8192 loops 38 microseconds
+cpucycles observed persecond 957467532...1059986667 with 16384 loops 76 microseconds
+cpucycles observed persecond 973102189...1029777778 with 32768 loops 136 microseconds
+cpucycles observed persecond 986518656...1015830828 with 65536 loops 267 microseconds
+cpucycles observed persecond 993452830...1008166667 with 131072 loops 529 microseconds
+cpucycles observed persecond 996036966...1003403609 with 262144 loops 1054 microseconds
+cpucycles observed persecond 984706378...1001682630 with 524288 loops 2131 microseconds
+cpucycles observed persecond 992585292...1001178580 with 1048576 loops 4296 microseconds
+</code></pre>
+<p><code>gcc45</code>,
+AMD Athlon II X4 640,
+Debian 8.11,
+Linux kernel 3.16.0-11-686-pae:</p>
+<pre><code>cpucycles version 20230105
+cpucycles tracesetup 0 x86-tsc precision 199 scaling 1.000000 only32 0
+cpucycles tracesetup 1 x86-tscasm precision 199 scaling 1.000000 only32 0
+cpucycles tracesetup 2 default-perfevent precision 170 scaling 1.000000 only32 0
+cpucycles tracesetup 3 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 4 default-monotonic precision 941 scaling 3.000000 only32 0
+cpucycles tracesetup 5 default-gettimeofday precision 3200 scaling 3000.000000 only32 0
+cpucycles tracesetup 6 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 3000000000
+cpucycles implementation default-perfevent
+cpucycles median 72 +12+0+0+0+0+0+0+0+5+0+0+0+0+0+0+0+2+0+0+0+0+0+0+0+1+0+0+0+0+0+0+0+2+0+0+0+0+0+0+0+1+0+0+0+0+0+0+0+2+0+0+0+0+0+0+0+1+0+0+0+0+0+0
+cpucycles observed persecond 541500000...1812000000 with 1024 loops 3 microseconds
+cpucycles observed persecond 712333333...1212250000 with 2048 loops 5 microseconds
+cpucycles observed persecond 1193285714...1733600000 with 4096 loops 6 microseconds
+cpucycles observed persecond 1689176470...1804562500 with 8192 loops 33 microseconds
+cpucycles observed persecond 1713074626...1770600000 with 16384 loops 66 microseconds
+cpucycles observed persecond 1765107692...1795140625 with 32768 loops 129 microseconds
+cpucycles observed persecond 1785369649...1800603922 with 65536 loops 256 microseconds
+cpucycles observed persecond 1781377862...1796288462 with 131072 loops 261 microseconds
+cpucycles observed persecond 1772647398...1778247827 with 262144 loops 691 microseconds
+cpucycles observed persecond 1789670493...1794149598 with 524288 loops 870 microseconds
+cpucycles observed persecond 1860276211...1861561332 with 1048576 loops 3156 microseconds
+</code></pre>
+<p><code>gcc92</code>,
+SiFive Freedom U740,
+Ubuntu 22.04,
+Linux kernel 5.15.0-1014-generic:</p>
+<pre><code>cpucycles version 20230105
+cpucycles tracesetup 0 riscv64-rdcycle precision 8 scaling 1.000000 only32 0
+cpucycles tracesetup 1 default-perfevent precision 3024 scaling 1.000000 only32 0
+cpucycles tracesetup 2 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 3 default-monotonic precision 2599 scaling 2.399988 only32 0
+cpucycles tracesetup 4 default-gettimeofday precision 2599 scaling 2399.987654 only32 0
+cpucycles tracesetup 5 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 2399987654
+cpucycles implementation riscv64-rdcycle
+cpucycles median 8 +33+27+1+1+1+1+0+0+0+22+0+0+0+0+0+0+0+628+0+0+0+7+0+0+0+145+0+0+0+0+0+0+0+22+0+0+0+0+0+0+0+158+0+0+0+0+0+0+0+22+0+0+0+0+0+0+0+22+0+0+0+0+0
+cpucycles observed persecond 530250000...1978000000 with 1024 loops 3 microseconds
+cpucycles observed persecond 831000000...1915666667 with 2048 loops 4 microseconds
+cpucycles observed persecond 1055750000...1689500000 with 4096 loops 7 microseconds
+cpucycles observed persecond 1045562500...1305428572 with 8192 loops 15 microseconds
+cpucycles observed persecond 1102700000...1236357143 with 16384 loops 29 microseconds
+cpucycles observed persecond 1176053571...1247444445 with 32768 loops 55 microseconds
+cpucycles observed persecond 1173321428...1209127273 with 65536 loops 111 microseconds
+cpucycles observed persecond 1187805429...1205210046 with 131072 loops 220 microseconds
+cpucycles observed persecond 1192415909...1201157535 with 262144 loops 439 microseconds
+cpucycles observed persecond 1194694760...1199247717 with 524288 loops 877 microseconds
+cpucycles observed persecond 1194656004...1197023034 with 1048576 loops 1781 microseconds
+</code></pre>
+<p><code>gcc103</code>,
+Apple M1 (Icestorm-M1 + Firestorm-M1),
+Debian unstable (bookworm),
+Linux kernel 6.0.0-rc5-asahi-00001-gc62bd3fe430f:</p>
+<pre><code>cpucycles version 20230105
+cpucycles tracesetup 0 arm64-pmc precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 1 arm64-vct precision 186 scaling 86.000000 only32 0
+cpucycles tracesetup 2 default-perfevent precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 3 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 4 default-monotonic precision 285 scaling 2.064000 only32 0
+cpucycles tracesetup 5 default-gettimeofday precision 2264 scaling 2064.000000 only32 0
+cpucycles tracesetup 6 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 2064000000
+cpucycles implementation arm64-vct
+cpucycles median 0 +0+86+0+0+0+0+0+0+0+0+0+0+0+0+86+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+86+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+86+0+0+0+0+0+0+0+0
+cpucycles observed persecond 1784500000...3655000000 with 8192 loops 3 microseconds
+cpucycles observed persecond 1773750000...2393666667 with 16384 loops 7 microseconds
+cpucycles observed persecond 1897733333...2222769231 with 32768 loops 14 microseconds
+cpucycles observed persecond 1951310344...2114962963 with 65536 loops 28 microseconds
+cpucycles observed persecond 2024071428...2107000000 with 131072 loops 55 microseconds
+cpucycles observed persecond 2041531531...2082935780 with 262144 loops 110 microseconds
+cpucycles observed persecond 2051158371...2071461188 with 524288 loops 220 microseconds
+cpucycles observed persecond 2058539682...2068309795 with 1048576 loops 440 microseconds
+</code></pre>
+<p><code>gcc112</code> (<code>gcc2-power8</code>),
+IBM POWER8E,
+CentOS 7.9 AltArch,
+Linux kernel 3.10.0-1127.13.1.el7.ppc64le:</p>
+<pre><code>cpucycles version 20230105
+cpucycles tracesetup 0 ppc64-mftb precision 251 scaling 7.207031 only32 0
+cpucycles tracesetup 1 default-perfevent precision 295 scaling 1.000000 only32 0
+cpucycles tracesetup 2 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 3 default-monotonic precision 536 scaling 3.690000 only32 0
+cpucycles tracesetup 4 default-gettimeofday precision 3890 scaling 3690.000000 only32 0
+cpucycles tracesetup 5 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 3690000000
+cpucycles implementation ppc64-mftb
+cpucycles median 195 +2969-8+14+0-8+7-8-7+7+6-7-1+0-1+0+7+7-15+7-1-7+6+0+0-8+0+6+0-8+7+0+7-8-8-7-1+7-8+7+0-8+0+14-8-7+6+0-8+7+7-15+0-1+0-1+14+0-15+14+0-1+7+0
+cpucycles observed persecond 2603750000...5510000000 with 2048 loops 3 microseconds
+cpucycles observed persecond 3430500000...6052250000 with 4096 loops 5 microseconds
+cpucycles observed persecond 3411333333...4457500000 with 8192 loops 11 microseconds
+cpucycles observed persecond 3548695652...4060333334 with 16384 loops 22 microseconds
+cpucycles observed persecond 3624977777...3876534884 with 32768 loops 44 microseconds
+cpucycles observed persecond 3621855555...3745363637 with 65536 loops 89 microseconds
+cpucycles observed persecond 3660157303...3722227273 with 131072 loops 177 microseconds
+cpucycles observed persecond 3680471751...3711622160 with 262144 loops 353 microseconds
+cpucycles observed persecond 3685321074...3700886525 with 524288 loops 706 microseconds
+cpucycles observed persecond 3687745930...3695537208 with 1048576 loops 1412 microseconds
+</code></pre>
+<p><code>gcc202</code>,
+UltraSparc T5,
+Debian unstable (bookworm),
+Linux kernel 5.19.0-2-sparc64-smp:</p>
+<pre><code>cpucycles version 20230105
+cpucycles tracesetup 0 sparc64-rdtick precision 65 scaling 1.000000 only32 0
+cpucycles tracesetup 1 default-perfevent precision 386 scaling 1.000000 only32 0
+cpucycles tracesetup 2 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 3 default-monotonic precision 442 scaling 3.599910 only32 0
+cpucycles tracesetup 4 default-gettimeofday precision 3799 scaling 3599.910000 only32 0
+cpucycles tracesetup 5 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 3599910000
+cpucycles implementation sparc64-rdtick
+cpucycles median 73 +24+0+24+24+24+24+24+24+0+1+24+0+1+24+0+1+24+0+0+1+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+0+1+0+0+0+0+0+0+0+0+0+0+0+0+0
+cpucycles observed persecond 2751500000...4258250000 with 4096 loops 5 microseconds
+cpucycles observed persecond 3289200000...4206875000 with 8192 loops 9 microseconds
+cpucycles observed persecond 3454789473...3900823530 with 16384 loops 18 microseconds
+cpucycles observed persecond 3452026315...3659888889 with 32768 loops 37 microseconds
+cpucycles observed persecond 3543770270...3650916667 with 65536 loops 73 microseconds
+cpucycles observed persecond 3567299319...3620662069 with 131072 loops 146 microseconds
+cpucycles observed persecond 3591373287...3618220690 with 262144 loops 291 microseconds
+cpucycles observed persecond 3597353344...3610774527 with 524288 loops 582 microseconds
+cpucycles observed persecond 3595899403...3603058071 with 1048576 loops 1172 microseconds
+</code></pre>
+<p>IBM z15:</p>
+<pre><code>cpucycles version 20230106
+cpucycles tracesetup 0 s390x-stckf precision 250 scaling 1.269531 only32 0
+cpucycles tracesetup 1 default-perfevent precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 2 default-mach precision 0 scaling 0.000000 only32 0
+cpucycles tracesetup 3 default-monotonic precision 272 scaling 5.200000 only32 0
+cpucycles tracesetup 4 default-gettimeofday precision 5400 scaling 5200.000000 only32 0
+cpucycles tracesetup 5 default-zero precision 0 scaling 0.000000 only32 0
+cpucycles persecond 5200000000
+cpucycles implementation s390x-stckf
+cpucycles median 48 +87+8+0-2+0+0+38-2+0+1-3+1+28+0+3-3+1+0+28+0-2+3+0-2+36+0+0+0+1+0+28+0-2+0+3-2+35+1+0-2+0+3+28+0-2+0+0-2+3+25+3+0-2+0+1+35+1+0+0-2+0+28+0
+cpucycles observed persecond 4948941176...5627733334 with 8192 loops 16 microseconds
+cpucycles observed persecond 4104125000...5515666667 with 16384 loops 7 microseconds
+cpucycles observed persecond 5047076923...5987818182 with 32768 loops 12 microseconds
+cpucycles observed persecond 5044846153...5475708334 with 65536 loops 25 microseconds
+cpucycles observed persecond 5141313725...5357428572 with 131072 loops 50 microseconds
+cpucycles observed persecond 5150892156...5257250000 with 262144 loops 101 microseconds
+cpucycles observed persecond 5183421568...5236549505 with 524288 loops 203 microseconds
+cpucycles observed persecond 5190282555...5216582717 with 1048576 loops 406 microseconds
+</code></pre><hr><font size=1><b>Version:</b>
+This is version 2023.01.06 of the "Counters" web page.
+</font>
+</div>
+</body>
+</html>
@@ -0,0 +1,75 @@
+<html>
+<head>
+<meta http-equiv="content-type" content="text/html; charset=utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<style type="text/css">
+html{overflow-y:scroll}
+body{font-family:sans-serif}
+p,ul,ol,blockquote,pre{font-size:0.9em;line-height:1.6em}
+li p{font-size:1.0em}
+blockquote p{font-size:1.0em}
+tt{font-size:1.2em}
+code{font-size:1.2em}
+h1{font-size:1.5em}
+h2{font-size:1.3em}
+h3{font-size:1.0em}
+h1 a{text-decoration:none}
+table{border-collapse:collapse}
+th,td{border:1px solid black}
+table a{text-decoration:none}
+table tr{font-size:0.9em;line-height:1.6em}
+.links a:hover{text-decoration:underline}
+.links a:active{text-decoration:underline}
+.links img{width:200px;padding-left:1em}
+.links td{border:0px;padding-top:0.5em;padding-bottom:0.5em}
+.headline{padding:0;font-weight:bold;font-size:1.5em;vertical-align:top;padding-bottom:0.5em;color:#125d0d}
+.navt{display:inline-block;box-sizing:border-box;-moz-box-sizing:border-box;-webkit-box-sizing:border-box;
+min-width:14%;margin:0;padding:0;padding-left:0.5em;padding-right:0.5em;vertical-align:center;
+font-weight:bold;font-size:1.1em;text-align:center;border:1px solid black}
+.here{border-bottom:0px;background-color:#ffffff}
+.away{background-color:#125d0d;}
+.away a{text-decoration:none;display:block;color:#ffffff}
+.away a:hover,.away a:active{text-decoration:underline}
+.main{margin:0;padding-top:0em;padding-bottom:1%;clear:both}
+</style>
+<title>
+Download</title>
+</head>
+<body>
+<div class=headline>
+libcpucycles
+</div>
+<div class=nav>
+<div class="navt away"><a href=index.html>Intro</a>
+</div><div class="navt here">Download
+</div><div class="navt away"><a href=install.html>Install</a>
+</div><div class="navt away"><a href=api.html>API</a>
+</div><div class="navt away"><a href=counters.html>Counters</a>
+</div><div class="navt away"><a href=selection.html>Selection</a>
+</div><div class="navt away"><a href=security.html>Security</a>
+</div></div>
+<div class=main>
+<p>To download and unpack the latest version of libcpucycles:</p>
+<pre><code>    wget -m https://cpucycles.cr.yp.to/libcpucycles-latest-version.txt
+    version=$(cat cpucycles.cr.yp.to/libcpucycles-latest-version.txt)
+    wget -m https://cpucycles.cr.yp.to/libcpucycles-$version.tar.gz
+    tar -xzf cpucycles.cr.yp.to/libcpucycles-$version.tar.gz
+    cd libcpucycles-$version
+</code></pre>
+<p>Then <a href="install.html">install</a>.</p>
+<h3>Archives and changelog (reverse chronological)</h3>
+<p><a href="libcpucycles-20230115.tar.gz"><code>libcpucycles-20230115.tar.gz</code></a> <a href="libcpucycles-20230115.html">browse</a></p>
+<p>Update actual <code>cpucycles_version</code> behavior to match documentation.</p>
+<p><a href="libcpucycles-20230110.tar.gz"><code>libcpucycles-20230110.tar.gz</code></a> <a href="libcpucycles-20230110.html">browse</a></p>
+<p><code>doc/api.md</code>: Document <code>cpucycles_version()</code>.</p>
+<p>Add <code>s390x-stckf</code> counter.</p>
+<p><code>cpucycles/default-perfevent.c</code>: Read into <code>int64_t</code> instead of <code>long long</code>.
+Add comment explaining issues with <code>PERF_FORMAT_TOTAL_TIME_RUNNING</code>.</p>
+<p><code>configure</code>: Improve <code>uname</code> handling.</p>
+<p><code>doc/api.md</code>: Update description of default frequency.</p>
+<p><a href="libcpucycles-20230105.tar.gz"><code>libcpucycles-20230105.tar.gz</code></a> <a href="libcpucycles-20230105.html">browse</a></p><hr><font size=1><b>Version:</b>
+This is version 2023.01.15 of the "Download" web page.
+</font>
+</div>
+</body>
+</html>
@@ -0,0 +1,88 @@
+<html>
+<head>
+<meta http-equiv="content-type" content="text/html; charset=utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<style type="text/css">
+html{overflow-y:scroll}
+body{font-family:sans-serif}
+p,ul,ol,blockquote,pre{font-size:0.9em;line-height:1.6em}
+li p{font-size:1.0em}
+blockquote p{font-size:1.0em}
+tt{font-size:1.2em}
+code{font-size:1.2em}
+h1{font-size:1.5em}
+h2{font-size:1.3em}
+h3{font-size:1.0em}
+h1 a{text-decoration:none}
+table{border-collapse:collapse}
+th,td{border:1px solid black}
+table a{text-decoration:none}
+table tr{font-size:0.9em;line-height:1.6em}
+.links a:hover{text-decoration:underline}
+.links a:active{text-decoration:underline}
+.links img{width:200px;padding-left:1em}
+.links td{border:0px;padding-top:0.5em;padding-bottom:0.5em}
+.headline{padding:0;font-weight:bold;font-size:1.5em;vertical-align:top;padding-bottom:0.5em;color:#125d0d}
+.navt{display:inline-block;box-sizing:border-box;-moz-box-sizing:border-box;-webkit-box-sizing:border-box;
+min-width:14%;margin:0;padding:0;padding-left:0.5em;padding-right:0.5em;vertical-align:center;
+font-weight:bold;font-size:1.1em;text-align:center;border:1px solid black}
+.here{border-bottom:0px;background-color:#ffffff}
+.away{background-color:#125d0d;}
+.away a{text-decoration:none;display:block;color:#ffffff}
+.away a:hover,.away a:active{text-decoration:underline}
+.main{margin:0;padding-top:0em;padding-bottom:1%;clear:both}
+</style>
+<title>
+Intro</title>
+</head>
+<body>
+<div class=headline>
+libcpucycles
+</div>
+<div class=nav>
+<div class="navt here">Intro
+</div><div class="navt away"><a href=download.html>Download</a>
+</div><div class="navt away"><a href=install.html>Install</a>
+</div><div class="navt away"><a href=api.html>API</a>
+</div><div class="navt away"><a href=counters.html>Counters</a>
+</div><div class="navt away"><a href=selection.html>Selection</a>
+</div><div class="navt away"><a href=security.html>Security</a>
+</div></div>
+<div class=main>
+<p>libcpucycles is a public-domain microlibrary for counting CPU cycles.
+Cycle counts are not as detailed as
+<a href="https://gamozolabs.github.io/metrology/2019/08/19/sushi_roll.html">Falk diagrams</a>
+but are the most precise timers available to typical software; they are
+central tools used in understanding and improving software performance.</p>
+<p>The libcpucycles <a href="api.html">API</a> is simple: include <code>&lt;cpucycles.h&gt;</code>, call
+<code>cpucycles()</code> to receive a <code>long long</code> whenever desired, and link with
+<code>-lcpucycles</code>.</p>
+<p><a href="counters.html">Internally</a>, libcpucycles understands machine-level
+cycle counters for amd64 (both PMC and TSC), arm32, arm64 (both PMC and
+VCT), mips64, ppc32, ppc64, riscv32, riscv64, s390x, sparc64, and x86.
+libcpucycles also understands four OS-level mechanisms, which give
+varying levels of accuracy: <code>mach_absolute_time</code>, <code>perf_event</code>,
+<code>CLOCK_MONOTONIC</code>, and, as a fallback, microsecond-resolution
+<code>gettimeofday</code>.</p>
+<p>When the program first calls <code>cpucycles()</code>, libcpucycles automatically
+benchmarks the available mechanisms and <a href="selection.html">selects</a> the
+mechanism that does the best job. Subsequent <code>cpucycles()</code> calls are
+thread-safe and very fast. An accompanying <code>cpucycles-info</code> program
+prints a summary of cycle-counter accuracy.</p>
+<p>For comparison, there is a simple-sounding <code>__rdtsc()</code> API provided by
+compilers, but this works only on Intel/AMD CPUs and is generally noisier
+than PMC. There is a <code>__builtin_readcyclecounter()</code> that works on more
+CPUs, but this works only with <code>clang</code> and has the same noise problems.
+Both of these mechanisms put the burden on the caller to figure out what
+can be done on other CPUs. Various packages include their own more
+portable abstraction layers for counting cycles (see, e.g., FFTW's
+<a href="https://github.com/FFTW/fftw3/blob/master/kernel/cycle.h"><code>cycle.h</code></a>,
+used to automatically select from among multiple implementations
+provided by FFTW), but this creates per-package effort to keep up with
+the latest cycle counters. The goal of libcpucycles is to provide
+state-of-the-art cycle counting centrally for all packages to use.</p><hr><font size=1><b>Version:</b>
+This is version 2023.01.06 of the "Intro" web page.
+</font>
+</div>
+</body>
+</html>
@@ -0,0 +1,101 @@
+<html>
+<head>
+<meta http-equiv="content-type" content="text/html; charset=utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<style type="text/css">
+html{overflow-y:scroll}
+body{font-family:sans-serif}
+p,ul,ol,blockquote,pre{font-size:0.9em;line-height:1.6em}
+li p{font-size:1.0em}
+blockquote p{font-size:1.0em}
+tt{font-size:1.2em}
+code{font-size:1.2em}
+h1{font-size:1.5em}
+h2{font-size:1.3em}
+h3{font-size:1.0em}
+h1 a{text-decoration:none}
+table{border-collapse:collapse}
+th,td{border:1px solid black}
+table a{text-decoration:none}
+table tr{font-size:0.9em;line-height:1.6em}
+.links a:hover{text-decoration:underline}
+.links a:active{text-decoration:underline}
+.links img{width:200px;padding-left:1em}
+.links td{border:0px;padding-top:0.5em;padding-bottom:0.5em}
+.headline{padding:0;font-weight:bold;font-size:1.5em;vertical-align:top;padding-bottom:0.5em;color:#125d0d}
+.navt{display:inline-block;box-sizing:border-box;-moz-box-sizing:border-box;-webkit-box-sizing:border-box;
+min-width:14%;margin:0;padding:0;padding-left:0.5em;padding-right:0.5em;vertical-align:center;
+font-weight:bold;font-size:1.1em;text-align:center;border:1px solid black}
+.here{border-bottom:0px;background-color:#ffffff}
+.away{background-color:#125d0d;}
+.away a{text-decoration:none;display:block;color:#ffffff}
+.away a:hover,.away a:active{text-decoration:underline}
+.main{margin:0;padding-top:0em;padding-bottom:1%;clear:both}
+</style>
+<title>
+Install</title>
+</head>
+<body>
+<div class=headline>
+libcpucycles
+</div>
+<div class=nav>
+<div class="navt away"><a href=index.html>Intro</a>
+</div><div class="navt away"><a href=download.html>Download</a>
+</div><div class="navt here">Install
+</div><div class="navt away"><a href=api.html>API</a>
+</div><div class="navt away"><a href=counters.html>Counters</a>
+</div><div class="navt away"><a href=selection.html>Selection</a>
+</div><div class="navt away"><a href=security.html>Security</a>
+</div></div>
+<div class=main>
+<p>Prerequisites: <code>python3</code>; <code>gcc</code> and/or <code>clang</code>. Currently tested only
+under Linux, but porting to other systems shouldn't be difficult.</p>
+<p>For sysadmins, to install in <code>/usr/local/{include,lib,bin}</code>:</p>
+<pre><code>    ./configure &amp;&amp; make -j8 install
+</code></pre>
+<p>For developers with an unprivileged account (typically with</p>
+<pre><code>    export LD_LIBRARY_PATH="$HOME/lib"
+    export LIBRARY_PATH="$HOME/lib"
+    export CPATH="$HOME/include"
+    export PATH="$HOME/bin:$PATH"
+</code></pre>
+<p>in <code>$HOME/.profile</code>), to install in <code>$HOME/{include,lib,bin}</code>:</p>
+<pre><code>    ./configure --prefix=$HOME &amp;&amp; make -j8 install
+</code></pre>
+<p>For distributors creating a package: Run</p>
+<pre><code>    ./configure --prefix=/usr &amp;&amp; make -j8
+</code></pre>
+<p>and then follow your usual packaging procedures for the
+<code>build/0/package</code> files:</p>
+<pre><code>    build/0/package/man/man3/cpucycles.3
+    build/0/package/include/cpucycles.h
+    build/0/package/lib/libcpucycles*
+    build/0/package/bin/cpucycles-info
+</code></pre>
+<p>There are some old systems where libcpucycles requires <code>-lrt</code> for
+<code>clock_gettime</code>; currently <code>libcpucycles.so</code> doesn't link to <code>-lrt</code>,
+so it's up to the caller to link to <code>-lrt</code>.</p>
+<p>More options: You can run</p>
+<pre><code>    ./configure --host=amd64
+</code></pre>
+<p>to override <code>./configure</code>'s guess of the architecture that it should
+compile for. The architecture controls which cycle counters to try
+compiling: e.g., <code>amd64</code> tries compiling <code>cpucycles/amd64*</code> and
+<code>cpucycles/default*</code>.</p>
+<p>Inside the <code>build</code> directory, <code>0</code> is symlinked to <code>amd64</code> for
+<code>--host=amd64</code>. Running <code>make clean</code> removes <code>build/amd64</code>. Re-running
+<code>./configure</code> automatically starts with <code>make clean</code>.</p>
+<p>A subsequent <code>./configure --host=arm64</code> will create <code>build/arm64</code> and
+symlink <code>0 -&gt; arm64</code>, without touching an existing <code>build/amd64</code>.
+However, cross-compilers aren't yet selected automatically.</p>
+<p>Compilers tried are listed in <code>compilers/default</code>. Each compiler
+includes <code>-fPIC</code> to create a shared library, <code>-fvisibility=hidden</code> to
+hide non-public symbols in the library, and <code>-fwrapv</code> to switch to a
+slightly less dangerous version of C. The first compiler that seems to
+work is used to compile everything.</p><hr><font size=1><b>Version:</b>
+This is version 2023.01.05 of the "Install" web page.
+</font>
+</div>
+</body>
+</html>
@@ -0,0 +1,122 @@
+<html>
+<head>
+<meta http-equiv="content-type" content="text/html; charset=utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<style type="text/css">
+html{overflow-y:scroll}
+body{font-family:sans-serif}
+p,ul,ol,blockquote,pre{font-size:0.9em;line-height:1.6em}
+li p{font-size:1.0em}
+blockquote p{font-size:1.0em}
+tt{font-size:1.2em}
+code{font-size:1.2em}
+h1{font-size:1.5em}
+h2{font-size:1.3em}
+h3{font-size:1.0em}
+h1 a{text-decoration:none}
+table{border-collapse:collapse}
+th,td{border:1px solid black}
+table a{text-decoration:none}
+table tr{font-size:0.9em;line-height:1.6em}
+.links a:hover{text-decoration:underline}
+.links a:active{text-decoration:underline}
+.links img{width:200px;padding-left:1em}
+.links td{border:0px;padding-top:0.5em;padding-bottom:0.5em}
+.headline{padding:0;font-weight:bold;font-size:1.5em;vertical-align:top;padding-bottom:0.5em;color:#125d0d}
+.navt{display:inline-block;box-sizing:border-box;-moz-box-sizing:border-box;-webkit-box-sizing:border-box;
+min-width:14%;margin:0;padding:0;padding-left:0.5em;padding-right:0.5em;vertical-align:center;
+font-weight:bold;font-size:1.1em;text-align:center;border:1px solid black}
+.here{border-bottom:0px;background-color:#ffffff}
+.away{background-color:#125d0d;}
+.away a{text-decoration:none;display:block;color:#ffffff}
+.away a:hover,.away a:active{text-decoration:underline}
+.main{margin:0;padding-top:0em;padding-bottom:1%;clear:both}
+</style>
+<title>
+Security</title>
+</head>
+<body>
+<div class=headline>
+libcpucycles
+</div>
+<div class=nav>
+<div class="navt away"><a href=index.html>Intro</a>
+</div><div class="navt away"><a href=download.html>Download</a>
+</div><div class="navt away"><a href=install.html>Install</a>
+</div><div class="navt away"><a href=api.html>API</a>
+</div><div class="navt away"><a href=counters.html>Counters</a>
+</div><div class="navt away"><a href=selection.html>Selection</a>
+</div><div class="navt here">Security
+</div></div>
+<div class=main>
+<p>Many security systems have been shown to be breakable by "timing
+attacks". These attacks extract secrets by analyzing timings of the
+legitimate user's operations on secret data. See the June 2022 survey
+page <a href="https://timing.attacks.cr.yp.to">https://timing.attacks.cr.yp.to</a>
+for an overview and further references.</p>
+<p>Sometimes these attacks are used as motivation to disable the attacker's
+access to various timing mechanisms. For example, Firefox rounds its
+<code>performance.now</code> timer to 1-millisecond resolution
+<a href="https://developer.mozilla.org/en-US/docs/Web/API/Performance/now">"to mitigate potential security threats"</a>.</p>
+<p>As another example, reducing <code>/proc/sys/kernel/perf_event_paranoid</code>
+under Linux to 2 (from 3 or higher), so that libcpucycles has access to
+the best available Intel/AMD cycle counter (RDPMC), also means making
+this cycle counter and other performance-monitoring counters available
+to any attacker-controlled software running on the computer. Perhaps
+this helps timing attacks, not to mention the possibility of opening up
+other vulnerabilities via the complicated <code>perf_event</code> interface.</p>
+<p>As yet another example, ARM CPUs disable user access to the main CPU
+cycle counter by default. Installing a kernel module to enable user
+access to the cycle counter could help attacks.</p>
+<p>Given the availability of simple mechanisms to disable RDPMC etc., it is
+easy to recommend using those mechanisms. To avoid creating unnecessary
+tension between those recommendations and the use of libcpucycles,
+applications that use libcpucycles should be structured so that
+high-resolution timers are used only on controlled development and
+benchmarking machines, not on general end-user machines.</p>
+<p>This structure might seem incompatible with using cycle counts to
+automatically select the best of multiple options, as in FFTW. However,
+new infrastructure introduced in <a href="https://lib25519.cr.yp.to">lib25519</a>
+automatically selects options on end-user machines based on cycle counts
+that were <em>collected on benchmarking machines</em>.</p>
+<p>The above text should not be understood as endorsing the idea that
+disabling timers is an <em>effective</em> defense against timing attacks.
+Certainly disabling high-resolution timers is not sufficient for
+security: there are many ways for attackers to amplify timing signals
+and to statistically filter out noise from low-resolution timers.
+Disabling <em>every</em> standard timing mechanism on the machine does not stop
+the attacker from accessing a remote timer or a counter maintained by
+the attacker's software. Perhaps disabling timers sometimes makes the
+difference between a feasible attack and an infeasible attack, but
+evaluating this is extremely difficult.</p>
+<p>Meanwhile there is an auditable methodology available to stop timing
+attacks: constant-time programming, which systematically cuts off data
+flow from secrets to timings.</p>
+<p>For example, secrets affect a CPU's power consumption, and Turbo Boost
+creates data flow from power consumption to timings, as illustrated by
+the <a href="https://www.hertzbleed.com">Hertzbleed attack</a> extracting secret
+keys from the SIKE cryptosystem (before SIKE was broken in other ways),
+and an <a href="https://arxiv.org/abs/2206.07012">independent attack</a>
+extracting secret AES keys. Consequently, the constant-time methodology
+does not allow Turbo Boost.</p>
+<p>This is why <a href="https://timing.attacks.cr.yp.to">https://timing.attacks.cr.yp.to</a>
+recommends turning off Turbo Boost "right now", and explains the
+mechanisms available to do this. One non-security reason that it was
+already normal (although not universal) for manufacturers to provide
+these mechanisms to end users is that Turbo Boost has a reputation for
+causing premature hardware failures. Turbo Boost also provides very
+little speed benefit for modern multithreaded vectorized applications.</p>
+<p>Another reaction to timing attacks is to apply "masking" techniques.
+These techniques <em>seem</em> to make it more difficult for attackers to
+extract secrets from power consumption and other side channels. However,
+as <a href="https://timing.attacks.cr.yp.to">https://timing.attacks.cr.yp.to</a>
+explains, it is "practically impossible for an auditor to obtain any
+real assurance that these techniques are secure". See the December 2022
+paper
+<a href="https://eprint.iacr.org/2022/1713">"Breaking a fifth-order masked implementation of CRYSTALS-Kyber by copy-paste"</a>
+for a newer example of a security failure in a masked implementation.</p><hr><font size=1><b>Version:</b>
+This is version 2023.01.05 of the "Security" web page.
+</font>
+</div>
+</body>
+</html>
@@ -0,0 +1,158 @@
+<html>
+<head>
+<meta http-equiv="content-type" content="text/html; charset=utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<style type="text/css">
+html{overflow-y:scroll}
+body{font-family:sans-serif}
+p,ul,ol,blockquote,pre{font-size:0.9em;line-height:1.6em}
+li p{font-size:1.0em}
+blockquote p{font-size:1.0em}
+tt{font-size:1.2em}
+code{font-size:1.2em}
+h1{font-size:1.5em}
+h2{font-size:1.3em}
+h3{font-size:1.0em}
+h1 a{text-decoration:none}
+table{border-collapse:collapse}
+th,td{border:1px solid black}
+table a{text-decoration:none}
+table tr{font-size:0.9em;line-height:1.6em}
+.links a:hover{text-decoration:underline}
+.links a:active{text-decoration:underline}
+.links img{width:200px;padding-left:1em}
+.links td{border:0px;padding-top:0.5em;padding-bottom:0.5em}
+.headline{padding:0;font-weight:bold;font-size:1.5em;vertical-align:top;padding-bottom:0.5em;color:#125d0d}
+.navt{display:inline-block;box-sizing:border-box;-moz-box-sizing:border-box;-webkit-box-sizing:border-box;
+min-width:14%;margin:0;padding:0;padding-left:0.5em;padding-right:0.5em;vertical-align:center;
+font-weight:bold;font-size:1.1em;text-align:center;border:1px solid black}
+.here{border-bottom:0px;background-color:#ffffff}
+.away{background-color:#125d0d;}
+.away a{text-decoration:none;display:block;color:#ffffff}
+.away a:hover,.away a:active{text-decoration:underline}
+.main{margin:0;padding-top:0em;padding-bottom:1%;clear:both}
+</style>
+<title>
+Selection</title>
+</head>
+<body>
+<div class=headline>
+libcpucycles
+</div>
+<div class=nav>
+<div class="navt away"><a href=index.html>Intro</a>
+</div><div class="navt away"><a href=download.html>Download</a>
+</div><div class="navt away"><a href=install.html>Install</a>
+</div><div class="navt away"><a href=api.html>API</a>
+</div><div class="navt away"><a href=counters.html>Counters</a>
+</div><div class="navt here">Selection
+</div><div class="navt away"><a href=security.html>Security</a>
+</div></div>
+<div class=main>
+<p>Here is how libcpucycles decides which cycle counter to use. The
+underlying principles are as follows:</p>
+<ul>
+<li>
+<p>Failure is not allowed. Using a low-resolution timer such as
+  <code>gettimeofday()</code> to estimate cycle counts is not desirable but is better
+  than providing no information.</p>
+</li>
+<li>
+<p>A counter that does well on some CPUs and OSes can do badly on others.
+  The counter selection in libcpucycles is based not just on rules set
+  at compile time but also on measurements of how well the counters
+  perform when the program first calls <code>cpucycles()</code>.</p>
+</li>
+<li>
+<p>A critical application of cycle counting is collecting cycle counts
+  for multiple options to see which option is faster. It is the caller's
+  responsibility to compute medians of cycle counts for many runs of
+  whatever is being benchmarked: medians filter out occasional
+  cycle-count jumps caused by migration to another core (if the
+  benchmark is not pinned to a single core) or interrupts from other OS
+  activity. libcpucycles does not reject an otherwise attractive counter
+  merely because of occasional jumps.</p>
+</li>
+<li>
+<p>Cycle-counting overhead is not desirable, but does not directly affect
+  comparisons of multiple options measured using the same cycle counter,
+  so it is less important than consistent major errors such as treating
+  2^32 + x cycles as x cycles. (Performance experts seeing a function
+  that takes billions of cycles usually focus on smaller subroutines,
+  but libcpucycles should not break larger measurements.) This is why
+  libcpucycles does not provide direct access to 32-bit cycle counters:
+  it provides wrappers that combine the counters with gettimeofday() to
+  produce 64 bits, even though this incurs some extra overhead.</p>
+</li>
+<li>
+<p>The noise introduced by typical off-core clocks, such as multiplying a
+  24MHz clock by 86 to estimate cycles on a 2.064GHz CPU core, comes in
+  small part from low resolution but much more from changes in CPU
+  frequency: e.g., a 10000-cycle computation might be measured as 20000
+  cycles when the CPU enters a power-saving mode. When libcpucycles has
+  access to what is believed to be an on-core cycle counter, it uses
+  that even when its measurements show some noise. (Choosing an on-core
+  cycle counter does not magically eliminate the change in the relative
+  speed of the CPU and DRAM; the usual advice to warm up the CPU and set
+  constant frequencies if possible still applies.)</p>
+</li>
+</ul>
+<p>When <code>cpucycles()</code> is first called, libcpucycles tries running each
+cycle counter that has been compiled into the library. For example, for
+64-bit ARM CPUs, libcpucycles will try <code>arm64-pmc</code>, <code>arm64-vct</code>,
+<code>default-gettimeofday</code>, <code>default-mach</code>, <code>default-monotonic</code>, and
+<code>default-perfevent</code>, minus any of those that failed to compile.</p>
+<p>Cycle counters that fail at run time with SIGILL (or SIGFPE or SIGBUS or
+SIGSEGV) are eliminated from the list. For example, <code>arm64-pmc</code> will
+fail with SIGILL if the kernel does not allow user access to
+<code>PMCCNTR_EL0</code>. Beware that libcpucycles does not catch SIGILL after its
+initial tests: if the kernel initially allows user access to
+<code>PMCCNTR_EL0</code> but later turns it off then <code>arm64-pmc</code> will crash.</p>
+<p>Independently of these counters, libcpucycles uses various OS mechanisms
+to obtain an <em>estimate</em> of the CPU frequency. This estimate is also
+available to the caller as <code>cpucycles_persecond()</code>.</p>
+<p>The methods that libcpucycles uses to ask the OS for an estimated CPU
+frequency fail on some OS-CPU combinations, in which case libcpucycles
+falls back to a <code>cpucyclespersecond</code> environment variable, or, if that
+variable does not exist, an estimate of 2399987654 cycles per second.
+(This estimate is in a realistic range of CPU speeds, and is close to
+multiples of 24MHz, 25MHz, and 19.2MHz, which are common crystal
+frequencies.) The sysadmin can create <code>/etc/cpucyclespersecond</code> to
+override all of the OS mechanisms.</p>
+<p>For counters that do not ask for scaling, the estimated CPU frequency is
+shown in <code>cpucycles-info</code> as a double-check on the counter results. For
+counters that ask for scaling, libcpucycles uses the estimated CPU
+frequency to compute the scaling, so this is not a double-check. If a
+counter asks for scaling and the estimated CPU frequency does not seem
+close to a multiple of the counter frequency (possibly with a small
+power-of-2 denominator) then libcpucycles will throw the counter away,
+except in the case of fixed-resolution OS counters such as
+<code>gettimeofday</code> and <code>CLOCK_MONOTONIC</code>.</p>
+<p>libcpucycles computes a precision estimate for each counter (times any
+applicable scaling) as follows. Call the counter 1000 times. Check that
+the counter has never decreased, and has increased at least once. (A
+counter where the decrease/increase checks fail is retried 10 times, so
+10000 calls overall, and removed if it fails all 10 times.) The
+precision estimate is then the smallest nonzero difference between
+adjacent counter results, plus a penalty explained below.</p>
+<p>The penalty is 100 cycles for off-core counters (including RDTSC) and
+<code>default-perfevent</code>, and 200 cycles for fixed-resolution OS counters.
+For example, an on-core CPU cycle counter will be selected even if it
+actually has, e.g., a resolution of 8 cycles and 50 cycles of overhead.</p>
+<p>Finally, libcpucycles selects the counter where the precision estimate
+is the smallest number of cycles. Note that an inaccurate estimate of
+CPU frequency can influence the choice between a scaled counter and an
+unscaled counter.</p>
+<p>libcpucycles does <em>not</em> carry out its counter selection (typically tens
+of milliseconds, sometimes even more) as a static initializer; callers
+are presumed to not want to incur the cost of initialization unless and
+until they are actually using <code>cpucycles()</code>. A multithreaded caller thus
+has to place locks around any possibly-first call to <code>cpucycles()</code>, or
+create its own static initializer (an <code>__attribute__((constructor))</code>
+function) with an initial <code>cpucycles()</code> call so that all subsequent
+<code>cpucycles()</code> calls are thread-safe.</p><hr><font size=1><b>Version:</b>
+This is version 2023.01.05 of the "Selection" web page.
+</font>
+</div>
+</body>
+</html>
@@ -0,0 +1,56 @@
+Prerequisites: `python3`; `gcc` and/or `clang`. Currently tested only
+under Linux, but porting to other systems shouldn't be difficult.
+
+For sysadmins, to install in `/usr/local/{include,lib,bin}`:
+
+        ./configure && make -j8 install
+
+For developers with an unprivileged account (typically with
+
+        export LD_LIBRARY_PATH="$HOME/lib"
+        export LIBRARY_PATH="$HOME/lib"
+        export CPATH="$HOME/include"
+        export PATH="$HOME/bin:$PATH"
+
+in `$HOME/.profile`), to install in `$HOME/{include,lib,bin}`:
+
+        ./configure --prefix=$HOME && make -j8 install
+
+For distributors creating a package: Run
+
+        ./configure --prefix=/usr && make -j8
+
+and then follow your usual packaging procedures for the
+`build/0/package` files:
+
+        build/0/package/man/man3/cpucycles.3
+        build/0/package/include/cpucycles.h
+        build/0/package/lib/libcpucycles*
+        build/0/package/bin/cpucycles-info
+
+There are some old systems where libcpucycles requires `-lrt` for
+`clock_gettime`; currently `libcpucycles.so` doesn't link to `-lrt`,
+so it's up to the caller to link to `-lrt`.
+
+More options: You can run
+
+        ./configure --host=amd64
+
+to override `./configure`'s guess of the architecture that it should
+compile for. The architecture controls which cycle counters to try
+compiling: e.g., `amd64` tries compiling `cpucycles/amd64*` and
+`cpucycles/default*`.
+
+Inside the `build` directory, `0` is symlinked to `amd64` for
+`--host=amd64`. Running `make clean` removes `build/amd64`. Re-running
+`./configure` automatically starts with `make clean`.
+
+A subsequent `./configure --host=arm64` will create `build/arm64` and
+symlink `0 -> arm64`, without touching an existing `build/amd64`.
+However, cross-compilers aren't yet selected automatically.
+
+Compilers tried are listed in `compilers/default`. Each compiler
+includes `-fPIC` to create a shared library, `-fvisibility=hidden` to
+hide non-public symbols in the library, and `-fwrapv` to switch to a
+slightly less dangerous version of C. The first compiler that seems to
+work is used to compile everything.
@@ -0,0 +1,57 @@
+.\" Automatically generated by Pandoc 2.9.2.1
+.\"
+.TH "cpucycles" "3" "" "" ""
+.hy
+.SS NAME
+.PP
+cpucycles - count CPU cycles
+.SS SYNOPSIS
+.IP
+.nf
+\f[C]
+#include <cpucycles.h>
+
+long long count = cpucycles();
+long long persecond = cpucycles_persecond();
+const char *implementation = cpucycles_implementation();
+const char *version = cpucycles_version();
+\f[R]
+.fi
+.PP
+Link with \f[C]-lcpucycles\f[R].
+Old systems may also need \f[C]-lrt\f[R].
+.SS DESCRIPTION
+.PP
+\f[C]cpucycles()\f[R] returns an estimate for the number of CPU cycles
+that have occurred since an unspecified time in the past (perhaps system
+boot, perhaps program startup).
+.PP
+Accessing true cycle counters can be difficult on some CPUs and
+operating systems.
+\f[C]cpucycles()\f[R] does its best to produce accurate results, but
+selects a low-precision counter if the only other option is failure.
+.PP
+\f[C]cpucycles_persecond()\f[R] returns an estimate for the number of
+CPU cycles per second.
+This estimate comes from \f[C]/etc/cpucyclespersecond\f[R] if that file
+exists, otherwise from various OS mechanisms, otherwise from the
+\f[C]cpucyclespersecond\f[R] environment variable if that is set,
+otherwise 2399987654.
+.PP
+\f[C]cpucycles_implementation()\f[R] returns the name of the counter in
+use: e.g., \f[C]\[dq]amd64-pmc\[dq]\f[R].
+.PP
+\f[C]cpucycles_version()\f[R] returns the \f[C]libcpucycles\f[R] version
+number as a string: e.g., \f[C]\[dq]20230115\[dq]\f[R].
+Results of \f[C]cpucycles_implementation()\f[R] should be interpreted
+relative to \f[C]cpucycles_version()\f[R].
+.PP
+\f[C]cpucycles\f[R] is actually a function pointer.
+The first call to \f[C]cpucycles()\f[R] or
+\f[C]cpucycles_persecond()\f[R] or \f[C]cpucycles_implementation()\f[R]
+selects one of the available counters and updates the
+\f[C]cpucycles\f[R] pointer accordingly.
+Subsequent calls to \f[C]cpucycles()\f[R] are thread-safe.
+.SS SEE ALSO
+.PP
+\f[B]gettimeofday\f[R](2), \f[B]clock_gettime\f[R](2)
@@ -0,0 +1,36 @@
+libcpucycles is a public-domain microlibrary for counting CPU cycles.
+Cycle counts are not as detailed as
+[Falk diagrams](https://gamozolabs.github.io/metrology/2019/08/19/sushi_roll.html)
+but are the most precise timers available to typical software; they are
+central tools used in understanding and improving software performance.
+
+The libcpucycles [API](api.html) is simple: include `<cpucycles.h>`, call
+`cpucycles()` to receive a `long long` whenever desired, and link with
+`-lcpucycles`.
+
+[Internally](counters.html), libcpucycles understands machine-level
+cycle counters for amd64 (both PMC and TSC), arm32, arm64 (both PMC and
+VCT), mips64, ppc32, ppc64, riscv32, riscv64, s390x, sparc64, and x86.
+libcpucycles also understands four OS-level mechanisms, which give
+varying levels of accuracy: `mach_absolute_time`, `perf_event`,
+`CLOCK_MONOTONIC`, and, as a fallback, microsecond-resolution
+`gettimeofday`.
+
+When the program first calls `cpucycles()`, libcpucycles automatically
+benchmarks the available mechanisms and [selects](selection.html) the
+mechanism that does the best job. Subsequent `cpucycles()` calls are
+thread-safe and very fast. An accompanying `cpucycles-info` program
+prints a summary of cycle-counter accuracy.
+
+For comparison, there is a simple-sounding `__rdtsc()` API provided by
+compilers, but this works only on Intel/AMD CPUs and is generally noisier
+than PMC. There is a `__builtin_readcyclecounter()` that works on more
+CPUs, but this works only with `clang` and has the same noise problems.
+Both of these mechanisms put the burden on the caller to figure out what
+can be done on other CPUs. Various packages include their own more
+portable abstraction layers for counting cycles (see, e.g., FFTW's
+[`cycle.h`](https://github.com/FFTW/fftw3/blob/master/kernel/cycle.h),
+used to automatically select from among multiple implementations
+provided by FFTW), but this creates per-package effort to keep up with
+the latest cycle counters. The goal of libcpucycles is to provide
+state-of-the-art cycle counting centrally for all packages to use.
@@ -0,0 +1,76 @@
+Many security systems have been shown to be breakable by "timing
+attacks". These attacks extract secrets by analyzing timings of the
+legitimate user's operations on secret data. See the June 2022 survey
+page [https://timing.attacks.cr.yp.to](https://timing.attacks.cr.yp.to)
+for an overview and further references.
+
+Sometimes these attacks are used as motivation to disable the attacker's
+access to various timing mechanisms. For example, Firefox rounds its
+`performance.now` timer to 1-millisecond resolution
+["to mitigate potential security threats"](https://developer.mozilla.org/en-US/docs/Web/API/Performance/now).
+
+As another example, reducing `/proc/sys/kernel/perf_event_paranoid`
+under Linux to 2 (from 3 or higher), so that libcpucycles has access to
+the best available Intel/AMD cycle counter (RDPMC), also means making
+this cycle counter and other performance-monitoring counters available
+to any attacker-controlled software running on the computer. Perhaps
+this helps timing attacks, not to mention the possibility of opening up
+other vulnerabilities via the complicated `perf_event` interface.
+
+As yet another example, ARM CPUs disable user access to the main CPU
+cycle counter by default. Installing a kernel module to enable user
+access to the cycle counter could help attacks.
+
+Given the availability of simple mechanisms to disable RDPMC etc., it is
+easy to recommend using those mechanisms. To avoid creating unnecessary
+tension between those recommendations and the use of libcpucycles,
+applications that use libcpucycles should be structured so that
+high-resolution timers are used only on controlled development and
+benchmarking machines, not on general end-user machines.
+
+This structure might seem incompatible with using cycle counts to
+automatically select the best of multiple options, as in FFTW. However,
+new infrastructure introduced in [lib25519](https://lib25519.cr.yp.to)
+automatically selects options on end-user machines based on cycle counts
+that were _collected on benchmarking machines_.
+
+The above text should not be understood as endorsing the idea that
+disabling timers is an _effective_ defense against timing attacks.
+Certainly disabling high-resolution timers is not sufficient for
+security: there are many ways for attackers to amplify timing signals
+and to statistically filter out noise from low-resolution timers.
+Disabling _every_ standard timing mechanism on the machine does not stop
+the attacker from accessing a remote timer or a counter maintained by
+the attacker's software. Perhaps disabling timers sometimes makes the
+difference between a feasible attack and an infeasible attack, but
+evaluating this is extremely difficult.
+
+Meanwhile there is an auditable methodology available to stop timing
+attacks: constant-time programming, which systematically cuts off data
+flow from secrets to timings.
+
+For example, secrets affect a CPU's power consumption, and Turbo Boost
+creates data flow from power consumption to timings, as illustrated by
+the [Hertzbleed attack](https://www.hertzbleed.com) extracting secret
+keys from the SIKE cryptosystem (before SIKE was broken in other ways),
+and an [independent attack](https://arxiv.org/abs/2206.07012)
+extracting secret AES keys. Consequently, the constant-time methodology
+does not allow Turbo Boost.
+
+This is why [https://timing.attacks.cr.yp.to](https://timing.attacks.cr.yp.to)
+recommends turning off Turbo Boost "right now", and explains the
+mechanisms available to do this. One non-security reason that it was
+already normal (although not universal) for manufacturers to provide
+these mechanisms to end users is that Turbo Boost has a reputation for
+causing premature hardware failures. Turbo Boost also provides very
+little speed benefit for modern multithreaded vectorized applications.
+
+Another reaction to timing attacks is to apply "masking" techniques.
+These techniques _seem_ to make it more difficult for attackers to
+extract secrets from power consumption and other side channels. However,
+as [https://timing.attacks.cr.yp.to](https://timing.attacks.cr.yp.to)
+explains, it is "practically impossible for an auditor to obtain any
+real assurance that these techniques are secure". See the December 2022
+paper
+["Breaking a fifth-order masked implementation of CRYSTALS-Kyber by copy-paste"](https://eprint.iacr.org/2022/1713)
+for a newer example of a security failure in a masked implementation.
@@ -0,0 +1,104 @@
+Here is how libcpucycles decides which cycle counter to use. The
+underlying principles are as follows:
+
+* Failure is not allowed. Using a low-resolution timer such as
+  `gettimeofday()` to estimate cycle counts is not desirable but is better
+  than providing no information.
+
+* A counter that does well on some CPUs and OSes can do badly on others.
+  The counter selection in libcpucycles is based not just on rules set
+  at compile time but also on measurements of how well the counters
+  perform when the program first calls `cpucycles()`.
+
+* A critical application of cycle counting is collecting cycle counts
+  for multiple options to see which option is faster. It is the caller's
+  responsibility to compute medians of cycle counts for many runs of
+  whatever is being benchmarked: medians filter out occasional
+  cycle-count jumps caused by migration to another core (if the
+  benchmark is not pinned to a single core) or interrupts from other OS
+  activity. libcpucycles does not reject an otherwise attractive counter
+  merely because of occasional jumps.
+
+* Cycle-counting overhead is not desirable, but does not directly affect
+  comparisons of multiple options measured using the same cycle counter,
+  so it is less important than consistent major errors such as treating
+  2^32 + x cycles as x cycles. (Performance experts seeing a function
+  that takes billions of cycles usually focus on smaller subroutines,
+  but libcpucycles should not break larger measurements.) This is why
+  libcpucycles does not provide direct access to 32-bit cycle counters:
+  it provides wrappers that combine the counters with gettimeofday() to
+  produce 64 bits, even though this incurs some extra overhead.
+
+* The noise introduced by typical off-core clocks, such as multiplying a
+  24MHz clock by 86 to estimate cycles on a 2.064GHz CPU core, comes in
+  small part from low resolution but much more from changes in CPU
+  frequency: e.g., a 10000-cycle computation might be measured as 20000
+  cycles when the CPU enters a power-saving mode. When libcpucycles has
+  access to what is believed to be an on-core cycle counter, it uses
+  that even when its measurements show some noise. (Choosing an on-core
+  cycle counter does not magically eliminate the change in the relative
+  speed of the CPU and DRAM; the usual advice to warm up the CPU and set
+  constant frequencies if possible still applies.)
+
+When `cpucycles()` is first called, libcpucycles tries running each
+cycle counter that has been compiled into the library. For example, for
+64-bit ARM CPUs, libcpucycles will try `arm64-pmc`, `arm64-vct`,
+`default-gettimeofday`, `default-mach`, `default-monotonic`, and
+`default-perfevent`, minus any of those that failed to compile.
+
+Cycle counters that fail at run time with SIGILL (or SIGFPE or SIGBUS or
+SIGSEGV) are eliminated from the list. For example, `arm64-pmc` will
+fail with SIGILL if the kernel does not allow user access to
+`PMCCNTR_EL0`. Beware that libcpucycles does not catch SIGILL after its
+initial tests: if the kernel initially allows user access to
+`PMCCNTR_EL0` but later turns it off then `arm64-pmc` will crash.
+
+Independently of these counters, libcpucycles uses various OS mechanisms
+to obtain an _estimate_ of the CPU frequency. This estimate is also
+available to the caller as `cpucycles_persecond()`.
+
+The methods that libcpucycles uses to ask the OS for an estimated CPU
+frequency fail on some OS-CPU combinations, in which case libcpucycles
+falls back to a `cpucyclespersecond` environment variable, or, if that
+variable does not exist, an estimate of 2399987654 cycles per second.
+(This estimate is in a realistic range of CPU speeds, and is close to
+multiples of 24MHz, 25MHz, and 19.2MHz, which are common crystal
+frequencies.) The sysadmin can create `/etc/cpucyclespersecond` to
+override all of the OS mechanisms.
+
+For counters that do not ask for scaling, the estimated CPU frequency is
+shown in `cpucycles-info` as a double-check on the counter results. For
+counters that ask for scaling, libcpucycles uses the estimated CPU
+frequency to compute the scaling, so this is not a double-check. If a
+counter asks for scaling and the estimated CPU frequency does not seem
+close to a multiple of the counter frequency (possibly with a small
+power-of-2 denominator) then libcpucycles will throw the counter away,
+except in the case of fixed-resolution OS counters such as
+`gettimeofday` and `CLOCK_MONOTONIC`.
+
+libcpucycles computes a precision estimate for each counter (times any
+applicable scaling) as follows. Call the counter 1000 times. Check that
+the counter has never decreased, and has increased at least once. (A
+counter where the decrease/increase checks fail is retried 10 times, so
+10000 calls overall, and removed if it fails all 10 times.) The
+precision estimate is then the smallest nonzero difference between
+adjacent counter results, plus a penalty explained below.
+
+The penalty is 100 cycles for off-core counters (including RDTSC) and
+`default-perfevent`, and 200 cycles for fixed-resolution OS counters.
+For example, an on-core CPU cycle counter will be selected even if it
+actually has, e.g., a resolution of 8 cycles and 50 cycles of overhead.
+
+Finally, libcpucycles selects the counter where the precision estimate
+is the smallest number of cycles. Note that an inaccurate estimate of
+CPU frequency can influence the choice between a scaled counter and an
+unscaled counter.
+
+libcpucycles does _not_ carry out its counter selection (typically tens
+of milliseconds, sometimes even more) as a static initializer; callers
+are presumed to not want to incur the cost of initialization unless and
+until they are actually using `cpucycles()`. A multithreaded caller thus
+has to place locks around any possibly-first call to `cpucycles()`, or
+create its own static initializer (an `__attribute__((constructor))`
+function) with an initial `cpucycles()` call so that all subsequent
+`cpucycles()` calls are thread-safe.
@@ -0,0 +1,27 @@
+#!/usr/bin/env python3
+
+import os
+import sys
+import shutil
+import tempfile
+
+prefix = sys.argv[1]
+dirs = 'man/man3','lib','include','bin'
+install = {}
+
+os.umask(0o22)
+
+for target in dirs:
+  install[target] = '%s/%s'%(prefix,target)
+  os.makedirs(install[target],exist_ok=True)
+
+os.umask(0o77)
+
+for target in dirs:
+  with tempfile.TemporaryDirectory(dir=install[target]) as t:
+    for fn in sorted(os.listdir('package/'+target)):
+      try:
+        shutil.copy2('package/%s/%s' % (target,fn),'%s/%s' % (t,fn),follow_symlinks=False)
+      except TypeError: # XXX: old python3; should copy symlinks manually
+        shutil.copy2('package/%s/%s' % (target,fn),'%s/%s' % (t,fn))
+      os.rename('%s/%s' % (t,fn),'%s/%s' % (install[target],fn))
@@ -0,0 +1,6 @@
+#!/bin/sh
+
+rm -f package/lib/libcpucycles.a
+ar cr package/lib/libcpucycles.a "$@"
+ranlib package/lib/libcpucycles.a || :
+chmod 644 package/lib/libcpucycles.a
@@ -0,0 +1 @@
+20230115
@@ -0,0 +1,9 @@
+#[link(name = "cpucycles", kind = "static")]
+extern "C" {
+    pub static mut cpucycles:
+        ::std::option::Option<unsafe extern "C" fn() -> ::std::os::raw::c_longlong>;
+    pub fn cpucycles_implementation() -> *const ::std::os::raw::c_char;
+    pub fn cpucycles_version() -> *const ::std::os::raw::c_char;
+    pub fn cpucycles_persecond() -> ::std::os::raw::c_longlong;
+    pub fn cpucycles_tracesetup();
+}
@@ -0,0 +1,82 @@
+#![allow(non_upper_case_globals)]
+#![allow(non_camel_case_types)]
+#![allow(non_snake_case)]
+
+mod bindings;
+use bindings as c;
+
+use std::fmt;
+use std::{
+    error::Error,
+    ffi::{CStr, CString, IntoStringError},
+};
+
+#[derive(Debug)]
+pub struct CpuCyclesError {
+    message: String,
+}
+
+impl fmt::Display for CpuCyclesError {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(f, "{}", self.message)
+    }
+}
+
+impl Error for CpuCyclesError {}
+
+pub fn cpucycles_tracesetup() {
+    unsafe { c::cpucycles_tracesetup() }
+}
+
+pub fn cpucycles() -> Result<i64, CpuCyclesError> {
+    if let Some(count) = unsafe { c::cpucycles.map(|f| f()) } {
+        Ok(count)
+    } else {
+        Err(CpuCyclesError {
+            message: "Could not execute cpucycles!".to_string(),
+        })
+    }
+}
+
+pub fn cpucycles_persecond() -> Result<i64, CpuCyclesError> {
+    Ok(unsafe { c::cpucycles_persecond() })
+}
+
+pub fn cpucycles_implementation() -> Result<String, IntoStringError> {
+    let implementation = unsafe { CString::from(CStr::from_ptr(c::cpucycles_implementation())) };
+    implementation.into_string()
+}
+
+pub fn cpucycles_version() -> Result<String, IntoStringError> {
+    let version = unsafe { CString::from(CStr::from_ptr(c::cpucycles_version())) };
+    version.into_string()
+}
+
+#[cfg(test)]
+mod test {
+    use crate::*;
+
+    #[test]
+    fn cpucycles_test() {
+        let count = cpucycles();
+        assert!(count.is_ok())
+    }
+
+    #[test]
+    fn cpucycles_persecond_test() {
+        let per_second = cpucycles_persecond();
+        assert!(per_second.is_ok());
+    }
+
+    #[test]
+    fn cpucycles_implementation_test() {
+        let implementation = cpucycles_implementation();
+        assert!(implementation.is_ok());
+    }
+
+    #[test]
+    fn cpucycles_version_test() {
+        let version = cpucycles_version();
+        assert!(version.is_ok());
+    }
+}
@@ -17,7 +17,7 @@ rust-version = "1.56"
 [dependencies]
 anyhow = "1.0.53"
 async-trait = { workspace = true }
-bip39 = { workspace = true }
+bip39 = "1.0.1"
 bs58 = "0.4.0"
 clap = { version = "4.0", features = ["cargo", "derive"] }
 colored = "2.0"
@@ -53,9 +53,6 @@ nym-topology = { path = "../common/topology" }
 validator-client = { path = "../common/client-libs/validator-client" }
 nym-bin-common = { path = "../common/bin-common" }

-# chichat
-chitchat = "0.5"
-
 [dev-dependencies]
 tokio = { version = "1.21.2", features = [
    "rt-multi-thread",
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
durch	f838ddffe2	Remove mixnode modifications	2023-03-27 14:21:44 +02:00
durch	7f7c33d10b	Bundle libcpucycles	2023-03-27 14:11:03 +02:00
durch	e7fdd3d076	mixnode feature	2023-03-27 09:57:28 +02:00
durch	4dc89bd65f	Rename	2023-03-27 09:13:20 +02:00
durch	9aa3b9507d	cpu cycle ffi	2023-03-15 11:21:48 +00:00
durch	61e88f304b	Checkpoint	2023-03-14 14:38:34 +00:00