tests pass and it mines now

crates/engine-gpu/src/end_to_end_tests.rs

@@ -0,0 +1,196 @@
+use pow_core::{JobContext, hash_from_nonce};
+use primitive_types::U512;
+use wgpu::util::DeviceExt;
+use rand::Rng;
+use rand::SeedableRng;
+
+pub async fn test_end_to_end_mining(
+    device: &wgpu::Device,
+    queue: &wgpu::Queue,
+) -> Result<(), Box<dyn std::error::Error>> {
+    println!("Running End-to-End Mining Test...");
+
+    // 1. Setup a job context
+    let mut rng = rand_chacha::ChaCha8Rng::seed_from_u64(12345);
+    let mut header = [0u8; 32];
+    rng.fill(&mut header);
+
+    // Use difficulty 1 so target is MAX. Any hash should pass.
+    let difficulty = U512::from(1u64);
+    let ctx = JobContext::new(header, difficulty);
+
+    // 2. Pick a nonce
+    let nonce_val = U512::from(123456789u64);
+
+    // 3. Compute expected hash using CPU (pow-core)
+    let expected_hash = hash_from_nonce(&ctx, nonce_val);
+    println!("CPU Expected Hash: {:x}", expected_hash);
+
+    // 4. Run GPU Mining for this specific nonce
+
+    // Header Buffer
+    let mut header_u32s = [0u32; 8];
+    for i in 0..8 {
+        let chunk = &ctx.header[i * 4..(i + 1) * 4];
+        header_u32s[i] = u32::from_le_bytes([chunk[0], chunk[1], chunk[2], chunk[3]]);
+    }
+    let header_buffer = device.create_buffer_init(&wgpu::util::BufferInitDescriptor {
+        label: Some("Header Buffer"),
+        contents: bytemuck::cast_slice(&header_u32s),
+        usage: wgpu::BufferUsages::STORAGE,
+    });
+
+    // Target Buffer
+    let target_bytes = ctx.target.to_little_endian();
+    let mut target_u32s = [0u32; 16];
+    for i in 0..16 {
+        let chunk = &target_bytes[i * 4..(i + 1) * 4];
+        target_u32s[i] = u32::from_le_bytes([chunk[0], chunk[1], chunk[2], chunk[3]]);
+    }
+    let target_buffer = device.create_buffer_init(&wgpu::util::BufferInitDescriptor {
+        label: Some("Target Buffer"),
+        contents: bytemuck::cast_slice(&target_u32s),
+        usage: wgpu::BufferUsages::STORAGE,
+    });
+
+    // Start Nonce Buffer
+    // We set start_nonce = nonce_val. Thread 0 will check start_nonce + 0 = nonce_val.
+    let start_nonce_bytes = nonce_val.to_little_endian();
+    let mut start_nonce_u32s = [0u32; 16];
+    for i in 0..16 {
+        let chunk = &start_nonce_bytes[i * 4..(i + 1) * 4];
+        start_nonce_u32s[i] = u32::from_le_bytes([chunk[0], chunk[1], chunk[2], chunk[3]]);
+    }
+    let start_nonce_buffer = device.create_buffer_init(&wgpu::util::BufferInitDescriptor {
+        label: Some("Start Nonce Buffer"),
+        contents: bytemuck::cast_slice(&start_nonce_u32s),
+        usage: wgpu::BufferUsages::STORAGE,
+    });
+
+    // Results Buffer
+    let results_size = (1 + 16 + 16) * 4;
+    let results_buffer = device.create_buffer(&wgpu::BufferDescriptor {
+        label: Some("Results Buffer"),
+        size: results_size as u64,
+        usage: wgpu::BufferUsages::STORAGE | wgpu::BufferUsages::COPY_SRC | wgpu::BufferUsages::COPY_DST,
+        mapped_at_creation: false,
+    });
+
+    // Clear results buffer
+    let zeros = vec![0u8; results_size];
+    queue.write_buffer(&results_buffer, 0, &zeros);
+
+    // Load Shader
+    let shader_source = include_str!("mining.wgsl");
+    let shader = device.create_shader_module(wgpu::ShaderModuleDescriptor {
+        label: Some("Mining Shader"),
+        source: wgpu::ShaderSource::Wgsl(shader_source.into()),
+    });
+
+    let pipeline = device.create_compute_pipeline(&wgpu::ComputePipelineDescriptor {
+        label: Some("Mining Pipeline"),
+        layout: None,
+        module: &shader,
+        entry_point: Some("mining_main"),
+        compilation_options: Default::default(),
+        cache: None,
+    });
+
+    let bind_group_layout = pipeline.get_bind_group_layout(0);
+    let bind_group = device.create_bind_group(&wgpu::BindGroupDescriptor {
+        label: Some("Mining Bind Group"),
+        layout: &bind_group_layout,
+        entries: &[
+            wgpu::BindGroupEntry {
+                binding: 0,
+                resource: results_buffer.as_entire_binding(),
+            },
+            wgpu::BindGroupEntry {
+                binding: 1,
+                resource: header_buffer.as_entire_binding(),
+            },
+            wgpu::BindGroupEntry {
+                binding: 2,
+                resource: start_nonce_buffer.as_entire_binding(),
+            },
+            wgpu::BindGroupEntry {
+                binding: 3,
+                resource: target_buffer.as_entire_binding(),
+            },
+        ],
+    });
+
+    // Run Compute Pass
+    let mut encoder = device.create_command_encoder(&wgpu::CommandEncoderDescriptor {
+        label: Some("Mining Encoder"),
+    });
+
+    {
+        let mut cpass = encoder.begin_compute_pass(&wgpu::ComputePassDescriptor {
+            label: Some("Mining Compute Pass"),
+            timestamp_writes: None,
+        });
+        cpass.set_pipeline(&pipeline);
+        cpass.set_bind_group(0, &bind_group, &[]);
+        cpass.dispatch_workgroups(1, 1, 1); // 1 workgroup, 256 threads. Thread 0 will check nonce_val.
+    }
+
+    // Read Results
+    let staging_buffer = device.create_buffer(&wgpu::BufferDescriptor {
+        label: Some("Staging Buffer"),
+        size: results_size as u64,
+        usage: wgpu::BufferUsages::MAP_READ | wgpu::BufferUsages::COPY_DST,
+        mapped_at_creation: false,
+    });
+
+    encoder.copy_buffer_to_buffer(&results_buffer, 0, &staging_buffer, 0, results_size as u64);
+    queue.submit(Some(encoder.finish()));
+
+    let buffer_slice = staging_buffer.slice(..);
+    let (sender, receiver) = futures::channel::oneshot::channel();
+    buffer_slice.map_async(wgpu::MapMode::Read, move |v| sender.send(v).unwrap());
+
+    device.poll(wgpu::PollType::Wait {
+        submission_index: None,
+        timeout: None,
+    }).unwrap();
+
+    if let Ok(Ok(())) = receiver.await {
+        let data = buffer_slice.get_mapped_range();
+        let result_u32s: &[u32] = bytemuck::cast_slice(&data);
+
+        if result_u32s[0] != 0 {
+            println!("GPU found solution!");
+
+            // Parse nonce
+            let mut nonce_bytes = [0u8; 64];
+            for i in 0..16 {
+                let bytes = result_u32s[1 + i].to_le_bytes();
+                nonce_bytes[i * 4..(i + 1) * 4].copy_from_slice(&bytes);
+            }
+            let found_nonce = U512::from_little_endian(&nonce_bytes);
+            println!("GPU Nonce: {}", found_nonce);
+
+            assert_eq!(found_nonce, nonce_val, "Nonce mismatch!");
+
+            // Parse hash
+            let mut hash_bytes = [0u8; 64];
+            for i in 0..16 {
+                let bytes = result_u32s[17 + i].to_le_bytes();
+                hash_bytes[i * 4..(i + 1) * 4].copy_from_slice(&bytes);
+            }
+            let found_hash = U512::from_little_endian(&hash_bytes);
+            println!("GPU Hash: {:x}", found_hash);
+
+            assert_eq!(found_hash, expected_hash, "Hash mismatch!");
+            println!("✅ End-to-End Test Passed!");
+        } else {
+            println!("❌ GPU did not find solution (should have passed with MAX target)");
+            return Err("GPU did not find solution".into());
+        }
+    } else {
+        return Err("Failed to map buffer".into());
+    }
+
+    Ok(())
+}

crates/engine-gpu/src/main.rs

@@ -6,6 +6,7 @@ use qp_plonky2_field::types::{Field, PrimeField64};
 use wgpu::{self, util::DeviceExt};
 
 mod tests;
+mod end_to_end_tests;
 
 fn u8_array_to_u32_array(bytes: &[u8; 96]) -> [u32; 24] {
     let mut result = [0u32; 24];
@@ -121,6 +122,15 @@ async fn run() -> Result<(), Box<dyn std::error::Error>> {
         eprintln!("❌ Double hash tests failed: {}", e);
     }
 
+    if let Err(e) = end_to_end_tests::test_end_to_end_mining(&device, &queue).await {
+        eprintln!("❌ End-to-end mining test failed: {}", e);
+    }
+
+    println!("\nAll tests completed!");
+    if let Err(e) = end_to_end_tests::test_end_to_end_mining(&device, &queue).await {
+        eprintln!("❌ End-to-end mining test failed: {}", e);
+    }
+
     println!("\nAll tests completed!");
 
     // generate_correct_wgsl_constants();

crates/engine-gpu/src/mining.wgsl

@@ -973,18 +973,29 @@ fn mining_main(@builtin(global_invocation_id) global_id: vec3<u32>) {
         input[8u + i] = rev;
     }
 
-    // Hash
-    let hash = double_hash(input);
+    // Hash (Big Endian)
+    let hash_be = double_hash(input);
+
+    // Convert to Little Endian for difficulty check and storage
+    var hash_le: array<u32, 16>;
+    for (var i = 0u; i < 16u; i++) {
+        let val = hash_be[15u - i];
+        // Reverse bytes in u32
+        hash_le[i] = ((val & 0xFFu) << 24u) |
+                     ((val & 0xFF00u) << 8u) |
+                     ((val & 0xFF0000u) >> 8u) |
+                     ((val & 0xFF000000u) >> 24u);
+    }
 
     // Check target
-    if (is_below_target(hash, difficulty_target)) {
+    if (is_below_target(hash_le, difficulty_target)) {
         // Try to claim the solution
         if (atomicExchange(&results[0], 1u) == 0u) {
             // We won! Write nonce and hash
             // results layout: [0]=found, [1..16]=nonce, [17..32]=hash
             for (var i = 0u; i < 16u; i++) {
                 atomicStore(&results[1u + i], current_nonce[i]);
-                atomicStore(&results[17u + i], hash[i]);
+                atomicStore(&results[17u + i], hash_le[i]);
             }
         }
     }