Reproducible builds via Docker #799

capossele · 2023-08-15T13:02:00Z

This PR extends the cargo risczero tool with a new build command.

It is based on a docker image built with this Dockerfile and pushed to the risczero Docker hub account.

Installation

To install the tool just run:

cargo install --path risc0/cargo-risczero/

Example usage

# Build the factors example
cargo risczero build --manifest-path examples/factors/methods/guest/Cargo.toml

This will generate an output similar to:

ELFs ready at:
./elfs/multiply/multiply - ImageID: 9ee1612b0a7e270f8df248e47dc85d9908ff4c1e7df42f398a65ca878b57a23d

Packages that generate multiple ELFs are supported as well:

# Build the chess example
cargo risczero build --manifest-path risc0/zkvm/methods/guest/Cargo.toml

This will generate an output similar to:

ELFs ready at:
./elfs/risc0_zkvm_methods_guest/hello_commit - ImageID: eb12f9b97d8759327f651afeb09ae9a5713e7dbc428284d453b8cf56e8dadd5a
./elfs/risc0_zkvm_methods_guest/multi_test - ImageID: 761900e766a4ae1d8edcb2b49dc9aee54b94e42c9b0d6421cfb112314c4e3efc
./elfs/risc0_zkvm_methods_guest/slice_io - ImageID: 3f2ad1a2d500ab4ab927eebe241d872d3f598065b8987b182410cf01f350f74c

Resolves #755
Resolves #116

flaub · 2023-08-15T13:32:15Z

This is great! I think cargo risczero build (drop the -guest) is a better name.

github-actions · 2023-08-15T14:20:59Z

Benchmark for Linux-cuda

    <details open>
      <summary>Click to hide benchmark</summary>
      Benchmarks have changed between the two branches, unable to diff.
    </details>

Benchmark for Linux-default

    <details open>
      <summary>Click to hide benchmark</summary>
      Benchmarks have changed between the two branches, unable to diff.
    </details>

Benchmark for macOS-default

    <details open>
      <summary>Click to hide benchmark</summary>
      Benchmarks have changed between the two branches, unable to diff.
    </details>

Benchmark for macOS-metal

    <details open>
      <summary>Click to hide benchmark</summary>
      Benchmarks have changed between the two branches, unable to diff.
    </details>

flaub · 2023-08-15T14:26:29Z

It'd be great if we had a test for this somehow. I'm also interested in what instructions we need to have in docs for users on various platforms.

capossele · 2023-08-15T14:33:32Z

Yeah I agree. I'm all open for testing suggestions. So far I came up with the following idea:

Build a few guests on Linux and on Mac and check that both result into same values. This requires having Docker installed on both. I'm not sure though if would be feasible to compare a GitHub job (on Linux) result to another job (on Mac)

flaub · 2023-08-15T14:39:55Z

What if we did a manual build locally on any platform, then we checked in the resulting IMAGE_ID. Then in CI, we just run the build and compare?

capossele · 2023-08-15T14:53:20Z

Great idea! that would work! However we must be sure that the Cargo.lock used to do the local build is the same that gets checked in CI. I can try to put together a test for this 👍

mothran · 2023-08-15T16:06:04Z

.github/workflows/reproducible_build.yml

+    branches: [ main, 'release-*' ]
+  pull_request:
+    branches: [ main, 'release-*' ]
+
+  workflow_dispatch:


Generally I think this type of PR checker would live as job inside of the main.yml like: https://github.com/risc0/risc0/blob/main/.github/workflows/main.yml#L195

Any reason here to have it broken off?

mothran · 2023-08-15T16:07:29Z

risc0/cargo-risczero/Cargo.toml

 clap = { version = "4.0", features = ["derive"] }
 const_format = "0.2"
 dirs = "5.0"
+docker-generate = "0.1"


TIL about this crate! Super useful

mothran · 2023-08-15T16:11:02Z

risc0/cargo-risczero/docker/Dockerfile.release

@@ -0,0 +1,11 @@
+# To build run: docker build -f Dockerfile.release -t risczero/risc0-guest-builder:v0.17 .
+FROM ubuntu:20.04@sha256:3246518d9735254519e1b2ff35f95686e4a5011c90c85344c1f38df7bae9dd37


small nit 20.04 will EOL in 2024, vs 22.04 which is is EOL (base support tier) in 2026

Should we jump to 22.04 right now to extend the lifecycle of these images?

mothran · 2023-08-15T16:12:03Z

risc0/cargo-risczero/docker/Dockerfile.release

+RUN curl --proto '=https' --tlsv1.2 --retry 10 --retry-connrefused -fsSL 'https://sh.rustup.rs' | sh -s -- -y
+ENV PATH="/root/.cargo/bin:${PATH}"
+RUN cargo install cargo-risczero
+RUN cargo risczero install


If we update the remote toolchain, I suspect this could impact the image to change? @flaub is it possible to risczero install <VERSION> so we could optionally lock to a rust release?

Yeah, I think the next evolution of the install command is to have it specify version.

capossele · 2023-08-15T23:00:19Z

The docker container already uses --locked as a manifest-path option to ensure that the Cargo.lock (of the guest package) is up-to-date. We can surely add a check even before starting the docker build to make sure the Cargo.lock file is indeed present

github-actions · 2023-08-15T23:31:17Z

Benchmark for Linux-cuda

    <details open>
      <summary>Click to hide benchmark</summary>
      Benchmarks have changed between the two branches, unable to diff.
    </details>

Benchmark for Linux-default

    <details open>
      <summary>Click to hide benchmark</summary>
      Benchmarks have changed between the two branches, unable to diff.
    </details>

Benchmark for macOS-default `de0c3ba`

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	2.7±0.11ms	2.7±0.18ms	0.00%
fib/100/prove	3.7±0.07s	3.6±0.06s	-2.70%
fib/100/total	3.7±0.06s	3.7±0.06s	0.00%
fib/1000/execute	3.0±0.10ms	2.9±0.08ms	-3.33%
fib/1000/prove	3.7±0.08s	3.7±0.03s	0.00%
fib/1000/total	3.7±0.07s	3.7±0.08s	0.00%
fib/10000/execute	5.2±0.11ms	5.0±0.15ms	-3.85%
fib/10000/prove	15.2±0.11s	15.2±0.10s	0.00%
fib/10000/total	15.3±0.11s	15.2±0.13s	-0.65%

Benchmark for macOS-metal `de0c3ba`

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	2.8±0.12ms	2.8±0.08ms	0.00%
fib/100/prove	853.9±2.77ms	849.6±5.76ms	-0.50%
fib/100/total	875.6±6.51ms	872.0±4.45ms	-0.41%
fib/1000/execute	2.9±0.15ms	2.8±0.08ms	-3.45%
fib/1000/prove	874.4±4.14ms	870.6±3.85ms	-0.43%
fib/1000/total	899.0±6.29ms	890.4±6.89ms	-0.96%
fib/10000/execute	5.1±0.08ms	5.0±0.13ms	-1.96%
fib/10000/prove	3.3±0.01s	3.3±0.01s	0.00%
fib/10000/total	3.3±0.01s	3.3±0.01s	0.00%

heavypackets · 2023-08-15T23:34:37Z

My first thought it that we're using the term "reproducible build" in a non-standard way. For an example of historical context, the organization I've supported since my Docker days goes into the motivations: https://reproducible-builds.org.

A reproducible build implies: identical source code -> identical binary via a deterministic compiler, a guarantee that is portable across disparate environments into perpetuity. This is important for distributing applications that can be rebuilt anywhere with guaranteed security, a primary motivation behind reproducible builds in OSS.

This is not the contract we are implying with our tool. We are concerned about Image ID stability. As I understand, the Image ID ignores parts of the ELF that are unimportant to the program semantics, so binary comparison is unimportant in the zkVM security model. In our model, rustc could produce binaries who's binary hashes do not match but have the same Image ID -- that distinction might not be important in our VM security model, but that is not generalizable to meet the same bar implied by a "reproducible build" of a compiled program.

So, I'd recommend renaming this feature, preferably avoiding anything that implies a deterministic build. Then document it heavily, so people aren't confused what we are/aren't guaranteeing.

mothran · 2023-08-15T23:43:17Z

So, I'd recommend renaming this feature, preferably avoiding anything that implies a deterministic build. Then document it heavily, so people aren't confused what we are/aren't guaranteeing.

I am confused about this because we can emit a MemoryImage object reproduce-ably and deterministically using this system (ideally assuming we squash most of the entropy bugs). The ELF might not be, but you don't need a ELF, you can skip right to the memory image as your primary input on both the ZKVM and Bonsai.

mothran · 2023-08-15T23:53:05Z

So, I'd recommend renaming this feature, preferably avoiding anything that implies a deterministic build. Then document it heavily, so people aren't confused what we are/aren't guaranteeing.

I am confused about this because we can emit a MemoryImage object reproduce-ably and deterministically using this system (ideally assuming we squash most of the entropy bugs). The ELF might not be, but you don't need a ELF, you can skip right to the memory image as your primary input on both the ZKVM and Bonsai.

I just double checked and while the ImageID is derived from most members of the MemoryImage object, I think its possible to have the data on disk change without the ImageID change because of the PageTableInfo but I would wanna double check with @flaub first.

If that is the case then I think my above comment is probably not accurate.

github-actions · 2023-08-16T00:19:00Z

Benchmark for Linux-cuda `b24512e`

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	5.1±0.12ms	5.1±0.11ms	0.00%
fib/100/prove	755.7±4.22ms	753.3±3.45ms	-0.32%
fib/100/total	757.0±2.15ms	754.1±3.86ms	-0.38%
fib/1000/execute	5.6±0.11ms	5.6±0.12ms	0.00%
fib/1000/prove	783.3±3.26ms	782.2±4.28ms	-0.14%
fib/1000/total	794.8±4.04ms	788.6±4.09ms	-0.78%
fib/10000/execute	10.2±0.13ms	10.2±0.11ms	0.00%
fib/10000/prove	3.0±0.01s	3.0±0.00s	0.00%
fib/10000/total	3.0±0.00s	3.0±0.01s	0.00%

Benchmark for Linux-default `b24512e`

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	6.8±0.40ms	6.7±0.43ms	-1.47%
fib/100/prove	5.1±0.03s	5.1±0.04s	0.00%
fib/100/total	5.1±0.03s	5.1±0.02s	0.00%
fib/1000/execute	7.9±0.62ms	7.8±0.54ms	-1.27%
fib/1000/prove	5.1±0.02s	5.1±0.03s	0.00%
fib/1000/total	5.1±0.02s	5.1±0.04s	0.00%
fib/10000/execute	13.1±0.27ms	13.0±0.54ms	-0.76%
fib/10000/prove	21.3±0.10s	21.2±0.11s	-0.47%
fib/10000/total	21.2±0.12s	21.2±0.13s	0.00%

Benchmark for macOS-default `b24512e`

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	2.8±0.14ms	2.8±0.13ms	0.00%
fib/100/prove	3.7±0.07s	3.7±0.06s	0.00%
fib/100/total	3.7±0.07s	3.7±0.06s	0.00%
fib/1000/execute	3.0±0.10ms	2.9±0.06ms	-3.33%
fib/1000/prove	3.7±0.05s	3.7±0.05s	0.00%
fib/1000/total	3.7±0.08s	3.6±0.08s	-2.70%
fib/10000/execute	5.2±0.06ms	5.1±0.11ms	-1.92%
fib/10000/prove	15.3±0.16s	15.3±0.13s	0.00%
fib/10000/total	15.3±0.10s	15.3±0.12s	0.00%

Benchmark for macOS-metal

    <details open>
      <summary>Click to hide benchmark</summary>
      Benchmarks have changed between the two branches, unable to diff.
    </details>

flaub · 2023-08-16T04:20:15Z

The ImageID computation is a function that has a type like:

fn compute_image_id(used_elf_pages, page_size, page_table_addr, pc) -> ImageID

where:

used_elf_pages: the set of pages that are read by the ELF binary (this means that some ELF pages are not accessed and not needed)
page_size: The circuit is currently hard coded to 1024 bytes.
page_table_addr: The linear address where the page table exists in the runtime memory image. This is also hard coded in the circuit.
pc: Program counter, this is the starting address to begin execution.

If any of the parameters change, you get a different ImageID. The part that we're concerned with here for a "reproducible build" is just the used_elf_pages.

With this approach (using docker) or any other approach we attempt, the goal is to be able to ensure that if multiple users submit the same source code, they get identical used_elf_pages. Thus, to me this satisfies the requirements of a "reproducible build system". So I do think we imply "identical source code -> identical binary via a deterministic compiler". It just so happens that the "deterministic compiler" has to be wrapped in a docker image until rustc and cargo itself can satisfy this requirement.

flaub · 2023-08-16T04:59:34Z

risc0/cargo-risczero/Cargo.toml

 reqwest = { version = "0.11", default-features = false, features = [
  "blocking",
  "json",
  "rustls-tls",
  "gzip",
 ] }
+risc0-zkvm = { workspace = true }


We should be able to directly depend on risc0-binfmt here since all the types you're using from risc0-zkvm are just re-exports.

flaub · 2023-08-16T05:01:40Z

risc0/cargo-risczero/src/commands/build_guest.rs

+    #[ignore] // requires Docker to be installed
+    fn test_reproducible_multiply_method() {
+        let multiply_test = Tester {
+            manifest_path: "examples/factors/methods/guest/Cargo.toml".to_string(),


Something a bit more persistent would be risc0/zkvm/methods/guest/Cargo.toml

We are not committing the Cargo.lock for that method though

We probably should. I'm thinking all guests should have their lock files committed.

Ok added 👍

flaub · 2023-08-16T05:03:12Z

risc0/cargo-risczero/src/commands/build_guest.rs

+    /// Overwrites if an ELF with the same name already exists.
+    fn build(&self) -> Result<()> {
+        Command::new("docker")
+            .args(["build", "--output=elfs/", "--target=binary", "."])


I think the output should land under target/, so maybe target/riscv-guest/riscv32im-risc0-zkvm-elf/docker?

modified 👍

flaub · 2023-08-16T05:03:34Z

.gitignore

 rust-project.json
 target/
 tmp/
+elfs/


I'd prefer the ELFs to land under target/ someplace

heavypackets · 2023-08-16T06:59:47Z

used_elf_pages: the set of pages that are read by the ELF binary (this means that some ELF pages are not accessed
If any of the parameters change, you get a different ImageID. The part that we're concerned with here for a "reproducible build" is just the used_elf_pages.

This highlights my concern. We aren't concerned about the resulting binary on disk being identical through separate builds and we aren't testing binary hashes in our build system. In this scenario, we could have different binaries on disk produce the same ID because parts are ignored -- that would be an example of a violation of reproducible builds as I've seen understood in OSS and in PL/compiler space.

For the Web3 example, the NEAR documentation linked in Slack does claim to build identical binaries: https://docs.near.org/sdk/rust/building/reproducible-builds

To verify your contract user can build it themselves and check that the binaries are identical.

In the NEAR case, two slightly different on disk but identical semantically contracts would not meet their definition of reproducibility.

We are claiming that some derivative of those compiled binaries + other inputs will be identical. That's a different definition, and one that requires its own terminology IMO. Or, at least, very careful explanation how/why our definition is different.

flaub · 2023-08-16T07:48:34Z

We are also coming up with a new Operating System, which is the zkvm. What I claim is that a MemoryImage is equivalent to what an ELF or PE does in other OSes. If we could have rustc emit a MemoryImage directly, that'd be ideal, and then we'd care about comparing MemoryImages bit for bit equivalence. It's possible that multiple ELFs could map to a single MemoryImage, but I don't understand why this matters. Think of it like when we do the conversion, that some bits are redacted. This is similar to the way some reproducible build steps work (they blank out or normalize bits that are entropic, like timestamps).

flaub · 2023-08-16T07:49:52Z

In fact, I think we should also tell users that they should check that their ELF binary is identical after using this docker build. I'm also fine with doing this comparison in our CI tests.

github-actions · 2023-08-16T13:50:47Z

Benchmark for Linux-cuda

    <details open>
      <summary>Click to hide benchmark</summary>
      Benchmarks have changed between the two branches, unable to diff.
    </details>

Benchmark for Linux-default `ff8635a`

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	4.8±0.15ms	4.7±0.16ms	-2.08%
fib/100/prove	2.6±0.21s	1673.8±7.15ms	-35.62%
fib/100/total	2.3±0.14s	1678.8±14.96ms	-27.01%
fib/1000/execute	5.2±0.16ms	5.2±0.09ms	0.00%
fib/1000/prove	3.0±1.63s	1699.8±4.99ms	-43.34%
fib/1000/total	2.2±0.19s	1730.5±96.15ms	-21.34%
fib/10000/execute	8.9±0.14ms	8.9±0.11ms	0.00%
fib/10000/prove	8.9±2.14s	6.4±0.08s	-28.09%
fib/10000/total	7.3±0.35s	6.4±0.21s	-12.33%

Benchmark for macOS-default `ff8635a`

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	2.8±0.12ms	2.7±0.14ms	-3.57%
fib/100/prove	3.7±0.06s	3.7±0.06s	0.00%
fib/100/total	3.7±0.05s	3.7±0.06s	0.00%
fib/1000/execute	2.9±0.07ms	2.9±0.14ms	0.00%
fib/1000/prove	3.7±0.04s	3.7±0.05s	0.00%
fib/1000/total	3.8±0.04s	3.7±0.05s	-2.63%
fib/10000/execute	5.2±0.13ms	5.1±0.07ms	-1.92%
fib/10000/prove	15.3±0.10s	15.3±0.17s	0.00%
fib/10000/total	15.4±0.18s	15.3±0.07s	-0.65%

Benchmark for macOS-metal `ff8635a`

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	2.8±0.06ms	2.7±0.05ms	-3.57%
fib/100/prove	854.3±3.69ms	851.7±4.22ms	-0.30%
fib/100/total	880.8±6.25ms	876.2±5.38ms	-0.52%
fib/1000/execute	3.0±0.07ms	2.9±0.07ms	-3.33%
fib/1000/prove	872.9±4.78ms	871.3±2.85ms	-0.18%
fib/1000/total	898.2±7.78ms	895.6±4.81ms	-0.29%
fib/10000/execute	5.1±0.12ms	4.9±0.05ms	-3.92%
fib/10000/prove	3.3±0.02s	3.3±0.01s	0.00%
fib/10000/total	3.3±0.01s	3.3±0.01s	0.00%

github-actions · 2023-08-16T14:37:28Z

Benchmark for Linux-cuda `60989d8`

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	5.1±0.12ms	5.1±0.11ms	0.00%
fib/100/prove	773.5±4.44ms	719.1±1.78ms	-7.03%
fib/100/total	771.8±2.35ms	727.8±1.92ms	-5.70%
fib/1000/execute	5.6±0.09ms	5.5±0.13ms	-1.79%
fib/1000/prove	801.7±4.59ms	750.0±1.26ms	-6.45%
fib/1000/total	805.2±2.75ms	758.7±2.52ms	-5.77%
fib/10000/execute	10.2±0.14ms	10.1±0.11ms	-0.98%
fib/10000/prove	3.3±0.01s	2.8±0.00s	-15.15%
fib/10000/total	3.3±0.01s	2.8±0.00s	-15.15%

Benchmark for Linux-default

    <details open>
      <summary>Click to hide benchmark</summary>
      Benchmarks have changed between the two branches, unable to diff.
    </details>

Benchmark for macOS-default `60989d8`

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	2.8±0.14ms	2.7±0.16ms	-3.57%
fib/100/prove	3.7±0.05s	3.7±0.07s	0.00%
fib/100/total	3.7±0.06s	3.7±0.08s	0.00%
fib/1000/execute	3.0±0.17ms	2.8±0.11ms	-6.67%
fib/1000/prove	3.7±0.05s	3.7±0.06s	0.00%
fib/1000/total	3.8±0.06s	3.7±0.05s	-2.63%
fib/10000/execute	5.2±0.11ms	5.1±0.16ms	-1.92%
fib/10000/prove	15.3±0.21s	15.2±0.16s	-0.65%
fib/10000/total	15.4±0.18s	15.3±0.08s	-0.65%

Benchmark for macOS-metal

    <details open>
      <summary>Click to hide benchmark</summary>
      Benchmarks have changed between the two branches, unable to diff.
    </details>

github-actions · 2023-08-16T15:48:14Z

Benchmark for Linux-cuda `8dbd3c5`

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	5.2±0.09ms	5.1±0.10ms	-1.92%
fib/100/prove	765.9±5.85ms	730.3±2.99ms	-4.65%
fib/100/total	764.8±3.30ms	734.8±2.16ms	-3.92%
fib/1000/execute	5.6±0.09ms	5.6±0.11ms	0.00%
fib/1000/prove	793.0±3.66ms	757.4±1.52ms	-4.49%
fib/1000/total	797.3±2.74ms	762.6±2.00ms	-4.35%
fib/10000/execute	10.3±0.09ms	10.2±0.13ms	-0.97%
fib/10000/prove	3.3±0.01s	2.8±0.00s	-15.15%
fib/10000/total	3.3±0.01s	2.8±0.00s	-15.15%

Benchmark for Linux-default

    <details open>
      <summary>Click to hide benchmark</summary>
      Benchmarks have changed between the two branches, unable to diff.
    </details>

Benchmark for macOS-default `8dbd3c5`

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	2.8±0.15ms	2.7±0.13ms	-3.57%
fib/100/prove	3.7±0.05s	3.7±0.07s	0.00%
fib/100/total	3.7±0.08s	3.6±0.07s	-2.70%
fib/1000/execute	3.0±0.09ms	2.9±0.07ms	-3.33%
fib/1000/prove	3.7±0.06s	3.7±0.07s	0.00%
fib/1000/total	3.7±0.06s	3.7±0.06s	0.00%
fib/10000/execute	5.2±0.12ms	5.0±0.08ms	-3.85%
fib/10000/prove	15.3±0.22s	15.3±0.10s	0.00%
fib/10000/total	15.4±0.15s	15.3±0.11s	-0.65%

Benchmark for macOS-metal `8dbd3c5`

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	2.8±0.09ms	2.7±0.12ms	-3.57%
fib/100/prove	854.9±3.72ms	848.9±5.25ms	-0.70%
fib/100/total	875.8±7.12ms	875.1±7.21ms	-0.08%
fib/1000/execute	2.9±0.07ms	2.9±0.09ms	0.00%
fib/1000/prove	872.7±3.25ms	869.2±5.48ms	-0.40%
fib/1000/total	895.1±5.51ms	893.2±7.71ms	-0.21%
fib/10000/execute	5.0±0.04ms	5.0±0.04ms	0.00%
fib/10000/prove	3.3±0.01s	3.3±0.01s	0.00%
fib/10000/total	3.3±0.01s	3.3±0.01s	0.00%

capossele added 9 commits August 11, 2023 16:56

wip

909feff

Merge branch 'main' into capossele/reproducible-builds

9ee587b

improve build-guest command

03dba86

ensure docker is installed

98733eb

remove dash replace

2cbe7fa

add rename step

3bf64c0

add missing -d option

3742556

fix renaming

a6d5c44

add support for multiple elfs

59934cc

capossele requested review from flaub, mothran, carterbrett, heavypackets and SchmErik August 15, 2023 13:02

capossele self-assigned this Aug 15, 2023

rename build-guest to build

5124d8d

capossele added 3 commits August 15, 2023 16:42

add test for reproducible build

87f919f

fix typo

0b26e60

add -n option to echo

5afb181

mothran reviewed Aug 15, 2023

View reviewed changes

refactor if

2277dff

add unit test and Cargo.lock check

edf7df2

isolate verifiable build test

7eb38a7

flaub reviewed Aug 16, 2023

View reviewed changes

capossele added 4 commits August 16, 2023 14:03

change output dir

6604c41

update .dockerignore

8d0e829

update .gitignore

0085011

Merge branch 'main' into capossele/reproducible-builds

fb00ba6

update test

ef21f29

flaub approved these changes Aug 16, 2023

View reviewed changes

remove unused hex dependency

a757311

capossele marked this pull request as ready for review August 16, 2023 15:24

capossele merged commit 81578ba into main Aug 16, 2023
20 checks passed

capossele deleted the capossele/reproducible-builds branch August 16, 2023 16:36

jchilds872 approved these changes Oct 25, 2023

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Reproducible builds via Docker #799

Reproducible builds via Docker #799

capossele commented Aug 15, 2023 •

edited

flaub commented Aug 15, 2023

github-actions bot commented Aug 15, 2023

flaub commented Aug 15, 2023

capossele commented Aug 15, 2023

flaub commented Aug 15, 2023

capossele commented Aug 15, 2023

mothran Aug 15, 2023

mothran Aug 15, 2023

mothran Aug 15, 2023

mothran Aug 15, 2023

flaub Aug 15, 2023

capossele commented Aug 15, 2023

github-actions bot commented Aug 15, 2023

heavypackets commented Aug 15, 2023

mothran commented Aug 15, 2023

mothran commented Aug 15, 2023

github-actions bot commented Aug 16, 2023

flaub commented Aug 16, 2023 •

edited

flaub Aug 16, 2023

capossele Aug 16, 2023

flaub Aug 16, 2023

capossele Aug 16, 2023

flaub Aug 16, 2023

capossele Aug 16, 2023

flaub Aug 16, 2023

capossele Aug 16, 2023

flaub Aug 16, 2023

heavypackets commented Aug 16, 2023

flaub commented Aug 16, 2023

flaub commented Aug 16, 2023

github-actions bot commented Aug 16, 2023

github-actions bot commented Aug 16, 2023

github-actions bot commented Aug 16, 2023

		@@ -0,0 +1,11 @@
		# To build run: docker build -f Dockerfile.release -t risczero/risc0-guest-builder:v0.17 .
		FROM ubuntu:20.04@sha256:3246518d9735254519e1b2ff35f95686e4a5011c90c85344c1f38df7bae9dd37

Reproducible builds via Docker #799

Reproducible builds via Docker #799

Conversation

capossele commented Aug 15, 2023 • edited

Installation

Example usage

flaub commented Aug 15, 2023

github-actions bot commented Aug 15, 2023

Benchmark for Linux-cuda

Benchmark for Linux-default

Benchmark for macOS-default

Benchmark for macOS-metal

flaub commented Aug 15, 2023

capossele commented Aug 15, 2023

flaub commented Aug 15, 2023

capossele commented Aug 15, 2023

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

capossele commented Aug 15, 2023

github-actions bot commented Aug 15, 2023

Benchmark for Linux-cuda

Benchmark for Linux-default

Benchmark for macOS-default de0c3ba

Benchmark for macOS-metal de0c3ba

heavypackets commented Aug 15, 2023

mothran commented Aug 15, 2023

mothran commented Aug 15, 2023

github-actions bot commented Aug 16, 2023

Benchmark for Linux-cuda b24512e

Benchmark for Linux-default b24512e

Benchmark for macOS-default b24512e

Benchmark for macOS-metal

flaub commented Aug 16, 2023 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

heavypackets commented Aug 16, 2023

flaub commented Aug 16, 2023

flaub commented Aug 16, 2023

github-actions bot commented Aug 16, 2023

Benchmark for Linux-cuda

Benchmark for Linux-default ff8635a

Benchmark for macOS-default ff8635a

Benchmark for macOS-metal ff8635a

github-actions bot commented Aug 16, 2023

Benchmark for Linux-cuda 60989d8

Benchmark for Linux-default

Benchmark for macOS-default 60989d8

Benchmark for macOS-metal

github-actions bot commented Aug 16, 2023

Benchmark for Linux-cuda 8dbd3c5

Benchmark for Linux-default

Benchmark for macOS-default 8dbd3c5

Benchmark for macOS-metal 8dbd3c5

capossele commented Aug 15, 2023 •

edited

Benchmark for macOS-default `de0c3ba`

Benchmark for macOS-metal `de0c3ba`

Benchmark for Linux-cuda `b24512e`

Benchmark for Linux-default `b24512e`

Benchmark for macOS-default `b24512e`

flaub commented Aug 16, 2023 •

edited

Benchmark for Linux-default `ff8635a`

Benchmark for macOS-default `ff8635a`

Benchmark for macOS-metal `ff8635a`

Benchmark for Linux-cuda `60989d8`

Benchmark for macOS-default `60989d8`

Benchmark for Linux-cuda `8dbd3c5`

Benchmark for macOS-default `8dbd3c5`

Benchmark for macOS-metal `8dbd3c5`