import math

import torch
import torch.nn as nn
import torch.nn.functional as F

# Default ModelNew is PyTorch-only. run_and_check: use check_kernel=False with backend=cuda
# until the candidate file contains a real CUDA kernel (see kernel_static_checker).

class Model(nn.Module):
    """
    Single-layer multi-head cross-attention with RoPE on Q and K, attention
    dropout, and output projection—intended as a fusion target for robotics
    stacks (e.g. proprio / action queries attending to vision or memory).

    Pipeline (conceptually one fused kernel): linear Q from query stream,
    linear K/V from key–value stream → reshape heads → apply RoPE separately
    along query positions (T_q) and key positions (T_k) → scaled dot-product,
    softmax, dropout on weights → aggregate values → output projection.
    """

    def __init__(
        self,
        dim: int,
        num_heads: int,
        dropout: float = 0.0,
        rope_base: float = 10000.0,
    ):
        super().__init__()
        assert dim % num_heads == 0
        self.dim = dim
        self.num_heads = num_heads
        self.head_dim = dim // num_heads
        assert self.head_dim % 2 == 0, "RoPE requires an even head dimension"

        self.q_proj = nn.Linear(dim, dim, bias=False)
        self.k_proj = nn.Linear(dim, dim, bias=False)
        self.v_proj = nn.Linear(dim, dim, bias=False)
        self.out_proj = nn.Linear(dim, dim, bias=False)
        self.dropout = nn.Dropout(dropout)
        self.scale = 1.0 / math.sqrt(self.head_dim)

        inv_freq = 1.0 / (
            rope_base ** (torch.arange(0, self.head_dim, 2).float() / self.head_dim)
        )
        self.register_buffer("inv_freq", inv_freq, persistent=False)

    @staticmethod
    def _rotate_half(x: torch.Tensor) -> torch.Tensor:
        x1, x2 = x.chunk(2, dim=-1)
        return torch.cat([-x2, x1], dim=-1)

    def _rope(self, x: torch.Tensor, seq_len: int) -> torch.Tensor:
        """Apply RoPE along the sequence dimension (last-but-one after head). x: (B, H, L, d)."""
        t = torch.arange(seq_len, device=x.device, dtype=self.inv_freq.dtype)
        freqs = torch.outer(t, self.inv_freq)
        emb = torch.cat([freqs, freqs], dim=-1)
        cos = emb.cos().view(1, 1, seq_len, self.head_dim)
        sin = emb.sin().view(1, 1, seq_len, self.head_dim)
        return x * cos + self._rotate_half(x) * sin

    def forward(self, x_query: torch.Tensor, x_kv: torch.Tensor) -> torch.Tensor:
        """
        Args:
            x_query: (B, T_q, dim) — e.g. robot or decoder tokens
            x_kv:    (B, T_k, dim) — e.g. encoder / memory tokens
        Returns:
            (B, T_q, dim)
        """
        B, T_q, _ = x_query.shape
        _, T_k, _ = x_kv.shape
        h, d = self.num_heads, self.head_dim

        q = self.q_proj(x_query).view(B, T_q, h, d).transpose(1, 2)
        k = self.k_proj(x_kv).view(B, T_k, h, d).transpose(1, 2)
        v = self.v_proj(x_kv).view(B, T_k, h, d).transpose(1, 2)

        q = self._rope(q, T_q)
        k = self._rope(k, T_k)

        attn = torch.matmul(q, k.transpose(-2, -1)) * self.scale
        attn = F.softmax(attn, dim=-1)
        attn = self.dropout(attn)
        out = torch.matmul(attn, v)
        out = out.transpose(1, 2).contiguous().view(B, T_q, self.dim)
        return self.out_proj(out)


class ModelNew(nn.Module):
    """
    KernelBench candidate entry point (`eval_kernel_against_ref` / `run_and_check`).
    Replace this class with a fused CUDA/Triton (etc.) implementation; the default
    delegates to `Model` so you can smoke-test the harness with the same file for
    both `ref_arch_src_path` and `kernel_src_path`.
    """

    def __init__(
        self,
        dim: int,
        num_heads: int,
        dropout: float = 0.0,
        rope_base: float = 10000.0,
    ):
        super().__init__()
        self._impl = Model(dim, num_heads, dropout, rope_base)

    def forward(self, x_query: torch.Tensor, x_kv: torch.Tensor) -> torch.Tensor:
        return self._impl(x_query, x_kv)


batch_size = 4
seq_len_q = 64
seq_len_kv = 128
dim = 256
num_heads = 8
# Use 0.0 so `run_and_check` / multi-trial correctness matches: the harness does not
# re-seed RNG between the reference forward and `ModelNew` forward, so attention
# dropout > 0 would make outputs differ even for identical math.
dropout_p = 0.0


def get_inputs():
    p = popcorn_pri
    bs = p.jitter_int(batch_size)
    tq = p.jitter_int(seq_len_q)
    tk = p.jitter_int(seq_len_kv)
    x_query = torch.randn(bs, tq, dim)
    x_kv = torch.randn(bs, tk, dim)
    return [x_query, x_kv]


def get_init_inputs():
    return [dim, num_heads, dropout_p]

import math
import torch
import torch.nn as nn
from torch.utils.cpp_extension import load_inline

_cpp_src = 'torch::Tensor attn_rope_forward(torch::Tensor q, torch::Tensor k, torch::Tensor v, torch::Tensor inv_freq, int64_t num_heads);'
_cuda_src = r'''
#include <torch/extension.h>
#include <cuda.h>
#include <cuda_runtime.h>
#include <ATen/cuda/CUDAContext.h>
#include <cfloat>

__global__ void rope_kernel(const float* __restrict__ x, float* __restrict__ y,
                            const float* __restrict__ inv_freq,
                            int B, int T, int D, int H, int d) {
    int idx = blockIdx.x * blockDim.x + threadIdx.x;
    int total = B * T * D;
    if (idx >= total) return;
    int j_global = idx % D;
    int j = j_global % d;
    int half = d >> 1;
    int pair = j < half ? j : j - half;
    int base = idx - j;
    float c, s;
    int t = (idx / D) % T;
    sincosf((float)t * inv_freq[pair], &s, &c);
    float a = x[idx];
    float b = (j < half) ? x[base + j + half] : x[base + j - half];
    y[idx] = (j < half) ? (a * c - b * s) : (a * c + b * s);
}

__global__ void attn_kernel(const float* __restrict__ q,
                            const float* __restrict__ k,
                            const float* __restrict__ v,
                            float* __restrict__ out,
                            int B, int Tq, int Tk, int D, int H, int d,
                            float scale) {
    extern __shared__ float scores[];
    int tq = blockIdx.x;
    int h = blockIdx.y;
    int b = blockIdx.z;
    int tx = threadIdx.x;
    int qbase = (b * Tq + tq) * D + h * d;

    if (tx < Tk) {
        int kbase = (b * Tk + tx) * D + h * d;
        float sum = 0.0f;
        for (int j = 0; j < d; ++j) sum += q[qbase + j] * k[kbase + j];
        scores[tx] = sum * scale;
    }
    __syncthreads();

    if (tx == 0) {
        float m = -FLT_MAX;
        for (int i = 0; i < Tk; ++i) m = fmaxf(m, scores[i]);
        float denom = 0.0f;
        for (int i = 0; i < Tk; ++i) {
            float e = __expf(scores[i] - m);
            scores[i] = e;
            denom += e;
        }
        float inv = 1.0f / denom;
        for (int i = 0; i < Tk; ++i) scores[i] *= inv;
    }
    __syncthreads();

    if (tx < d) {
        float acc = 0.0f;
        for (int i = 0; i < Tk; ++i) {
            int vbase = (b * Tk + i) * D + h * d;
            acc += scores[i] * v[vbase + tx];
        }
        out[qbase + tx] = acc;
    }
}

torch::Tensor attn_rope_forward(torch::Tensor q, torch::Tensor k, torch::Tensor v,
                                torch::Tensor inv_freq, int64_t num_heads) {
    int B = q.size(0);
    int Tq = q.size(1);
    int D = q.size(2);
    int Tk = k.size(1);
    int H = (int)num_heads;
    int d = D / H;
    auto qrot = torch::empty_like(q);
    auto krot = torch::empty_like(k);
    auto out = torch::empty_like(q);
    cudaStream_t stream = at::cuda::getCurrentCUDAStream();
    int threads = 256;
    int n_q = B * Tq * D;
    int n_k = B * Tk * D;
    rope_kernel<<<(n_q + threads - 1) / threads, threads, 0, stream>>>(
        q.data_ptr<float>(), qrot.data_ptr<float>(), inv_freq.data_ptr<float>(), B, Tq, D, H, d);
    rope_kernel<<<(n_k + threads - 1) / threads, threads, 0, stream>>>(
        k.data_ptr<float>(), krot.data_ptr<float>(), inv_freq.data_ptr<float>(), B, Tk, D, H, d);
    dim3 grid(Tq, H, B);
    int attn_threads = (Tk <= 128 ? 128 : 256);
    attn_kernel<<<grid, attn_threads, Tk * sizeof(float), stream>>>(
        qrot.data_ptr<float>(), krot.data_ptr<float>(), v.data_ptr<float>(), out.data_ptr<float>(),
        B, Tq, Tk, D, H, d, rsqrtf((float)d));
    return out;
}
'''

_ext = load_inline(name='cross_attn_rope_ext_v2', cpp_sources=_cpp_src, cuda_sources=_cuda_src,
                   functions=['attn_rope_forward'], extra_cuda_cflags=['-O3', '--use_fast_math'], verbose=False)

class ModelNew(nn.Module):
    def __init__(self, dim: int, num_heads: int, dropout: float = 0.0, rope_base: float = 10000.0):
        super().__init__()
        assert dim % num_heads == 0
        self.dim = dim
        self.num_heads = num_heads
        self.head_dim = dim // num_heads
        assert self.head_dim % 2 == 0
        self.q_proj = nn.Linear(dim, dim, bias=False)
        self.k_proj = nn.Linear(dim, dim, bias=False)
        self.v_proj = nn.Linear(dim, dim, bias=False)
        self.out_proj = nn.Linear(dim, dim, bias=False)
        self.dropout = nn.Dropout(dropout)
        self.scale = 1.0 / math.sqrt(self.head_dim)
        inv_freq = 1.0 / (rope_base ** (torch.arange(0, self.head_dim, 2).float() / self.head_dim))
        self.register_buffer('inv_freq', inv_freq, persistent=False)
    def forward(self, x_query: torch.Tensor, x_kv: torch.Tensor) -> torch.Tensor:
        q = self.q_proj(x_query).contiguous()
        k = self.k_proj(x_kv).contiguous()
        v = self.v_proj(x_kv).contiguous()
        out = _ext.attn_rope_forward(q, k, v, self.inv_freq, self.num_heads)
        return self.out_proj(out)