[Example] Implement simple block sparse kernel (#106)

* Remove Torch CPP backend and update execution backend options

- Remove TorchCPPKernelAdapter and related code from JIT modules
- Update execution backend options in jit/__init__.py, kernel.py, and adapter/__init__.py
- Remove "torch_cpp" from supported execution backend literals
- Simplify backend validation and remove unused torch_cpp-related code
。

* lint fix

* Add block sparse attention implementations for TileLang and Triton

- Implement block sparse attention kernels for TileLang and Triton
- Add example scripts for block sparse attention with top-k and threshold-based masking
- Include utility functions for generating sparse attention masks
- Demonstrate causal attention with block-level sparsity
- Add test cases to validate sparse attention implementations against PyTorch reference

examples/blocksparse_attention/block_sparse_attn_tilelang.py

+# Copyright (c) Microsoft Corporation.
+# Licensed under the MIT License.
+import math
+import torch
+
+import tilelang
+import tilelang.language as T
+import torch.nn.functional as F
+
+def get_sparse_attn_mask_from_topk(x, topk, use_dense_for_last_block=False):
+    bsz, num_head, downsample_len, _ = x.shape
+    # N_CTX = downsample_len * BLOCK
+    sparse_index = torch.topk(x, topk, dim=-1).indices
+    dense_mask = torch.full([bsz, num_head, downsample_len, downsample_len], False, dtype=torch.bool, device=x.device)
+    dense_mask.scatter_(-1, sparse_index, True)
+    if use_dense_for_last_block:
+        dense_mask[:, :,-2:,:] = True 
+    dense_mask.tril_()
+    return  dense_mask
+
+
+def get_sparse_attn_mask_from_threshold(x, threshold, use_dense_for_last_block=False):
+    dense_mask = x > threshold 
+    if use_dense_for_last_block:
+        dense_mask[:, :,-2:,:] = True 
+    dense_mask.tril_()
+    return  dense_mask
+
+
+def blocksparse_flashattn(batch, heads, seq_len, dim, downsample_len, is_causal):
+    block_M = 64
+    block_N = 64
+    num_stages = 0
+    threads = 128
+    scale = (1.0 / dim)**0.5 * 1.44269504  # log2(e)
+    shape = [batch, heads, seq_len, dim]
+    block_mask_shape = [batch, heads, downsample_len, downsample_len]
+
+    dtype = "float16"
+    accum_dtype = "float"
+    block_mask_dtype = "int8"
+
+    def kernel_func(block_M, block_N, num_stages, threads):
+
+        @T.macro
+        def MMA0(
+            K: T.Buffer(shape, dtype),
+            Q_shared: T.Buffer([block_M, dim], dtype),
+            K_shared: T.Buffer([block_N, dim], dtype),
+            acc_s: T.Buffer([block_M, block_N], accum_dtype),
+            k: T.int32,
+            bx: T.int32,
+            by: T.int32,
+            bz: T.int32,
+        ):
+            T.copy(K[bz, by, k * block_N:(k + 1) * block_N, :], K_shared)
+            if is_causal:
+                for i, j in T.Parallel(block_M, block_N):
+                    acc_s[i, j] = T.if_then_else(bx * block_M + i >= k * block_N + j, 0,
+                                                 -T.infinity(acc_s.dtype))
+            else:
+                T.clear(acc_s)
+            T.gemm(Q_shared, K_shared, acc_s, transpose_B=True, policy=T.GemmWarpPolicy.FullRow)
+
+        @T.macro
+        def MMA1(
+                V: T.Buffer(shape, dtype),
+                V_shared: T.Buffer([block_M, dim], dtype),
+                acc_s_cast: T.Buffer([block_M, block_N], dtype),
+                acc_o: T.Buffer([block_M, dim], accum_dtype),
+                k: T.int32,
+                by: T.int32,
+                bz: T.int32,
+        ):
+            T.copy(V[bz, by, k * block_N:(k + 1) * block_N, :], V_shared)
+            T.gemm(acc_s_cast, V_shared, acc_o, policy=T.GemmWarpPolicy.FullRow)
+
+        @T.macro
+        def Softmax(
+                acc_s: T.Buffer([block_M, block_N], accum_dtype),
+                acc_s_cast: T.Buffer([block_M, block_N], dtype),
+                scores_max: T.Buffer([block_M], accum_dtype),
+                scores_max_prev: T.Buffer([block_M], accum_dtype),
+                scores_scale: T.Buffer([block_M], accum_dtype),
+                scores_sum: T.Buffer([block_M], accum_dtype),
+                logsum: T.Buffer([block_M], accum_dtype),
+        ):
+            T.copy(scores_max, scores_max_prev)
+            T.fill(scores_max, -T.infinity(accum_dtype))
+            T.reduce_max(acc_s, scores_max, dim=1, clear=False)
+            # To do causal softmax, we need to set the scores_max to 0 if it is -inf
+            # This process is called Check_inf in FlashAttention3 code, and it only need to be done
+            # in the first ceil_div(kBlockM, kBlockN) steps.
+            # for i in T.Parallel(block_M):
+            #     scores_max[i] = T.if_then_else(scores_max[i] == -T.infinity(accum_dtype), 0, scores_max[i])
+            for i in T.Parallel(block_M):
+                scores_scale[i] = T.exp2(scores_max_prev[i] * scale - scores_max[i] * scale)
+            for i, j in T.Parallel(block_M, block_N):
+                # Instead of computing exp(x - max), we compute exp2(x * log_2(e) -
+                # max * log_2(e)) This allows the compiler to use the ffma
+                # instruction instead of fadd and fmul separately.
+                acc_s[i, j] = T.exp2(acc_s[i, j] * scale - scores_max[i] * scale)
+            T.reduce_sum(acc_s, scores_sum, dim=1)
+            for i in T.Parallel(block_M):
+                logsum[i] = logsum[i] * scores_scale[i] + scores_sum[i]
+            T.copy(acc_s, acc_s_cast)
+
+        @T.macro
+        def Rescale(
+                acc_o: T.Buffer([block_M, dim], accum_dtype),
+                scores_scale: T.Buffer([block_M], accum_dtype),
+        ):
+            for i, j in T.Parallel(block_M, dim):
+                acc_o[i, j] *= scores_scale[i]
+
+        @T.prim_func
+        def main(
+                Q: T.Buffer(shape, dtype),
+                K: T.Buffer(shape, dtype),
+                V: T.Buffer(shape, dtype),
+                BlockSparseMask: T.Buffer(block_mask_shape, block_mask_dtype),
+                Output: T.Buffer(shape, dtype),
+        ):
+            with T.Kernel(
+                    T.ceildiv(seq_len, block_M), heads, batch, threads=threads) as (bx, by, bz):
+                Q_shared = T.alloc_shared([block_M, dim], dtype)
+                K_shared = T.alloc_shared([block_N, dim], dtype)
+                V_shared = T.alloc_shared([block_N, dim], dtype)
+                O_shared = T.alloc_shared([block_M, dim], dtype)
+                acc_s = T.alloc_fragment([block_M, block_N], accum_dtype)
+                acc_s_cast = T.alloc_fragment([block_M, block_N], dtype)
+                acc_o = T.alloc_fragment([block_M, dim], accum_dtype)
+                scores_max = T.alloc_fragment([block_M], accum_dtype)
+                scores_max_prev = T.alloc_fragment([block_M], accum_dtype)
+                scores_scale = T.alloc_fragment([block_M], accum_dtype)
+                scores_sum = T.alloc_fragment([block_M], accum_dtype)
+                logsum = T.alloc_fragment([block_M], accum_dtype)
+                block_mask = T.alloc_local([downsample_len], block_mask_dtype)
+                
+                T.copy(Q[bz, by, bx * block_M:(bx + 1) * block_M, :], Q_shared)
+                T.fill(acc_o, 0)
+                T.fill(logsum, 0)
+                T.fill(scores_max, -T.infinity(accum_dtype))
+
+                for vj in T.serial(downsample_len):
+                    block_mask[vj] = BlockSparseMask[bz, by, bx, vj]
+
+                loop_range = (
+                    T.min(T.ceildiv(seq_len, block_N), T.ceildiv(
+                        (bx + 1) * block_M, block_N)) if is_causal else T.ceildiv(seq_len, block_N))
+
+                for k in T.Pipelined(loop_range, num_stages=num_stages):
+                    if block_mask[k] != 0:
+                        MMA0(K, Q_shared, K_shared, acc_s, k, bx, by, bz)
+                        Softmax(acc_s, acc_s_cast, scores_max, scores_max_prev, scores_scale,
+                                scores_sum, logsum)
+                        Rescale(acc_o, scores_scale)
+                        MMA1(V, V_shared, acc_s_cast, acc_o, k, by, bz)
+                for i, j in T.Parallel(block_M, dim):
+                    acc_o[i, j] /= logsum[i]
+                T.copy(acc_o, O_shared)
+                T.copy(O_shared, Output[bz, by, bx * block_M:(bx + 1) * block_M, :])
+
+        return main
+
+    return kernel_func(block_M, block_N, num_stages, threads)
+
+def test_topk_sparse_attention():
+    # Config
+    BATCH, N_HEADS, SEQ_LEN, D_HEAD = 1, 1, 256, 64
+    TOPK = 2  # Keep top 8 elements per row
+    BLOCK = 64
+    torch.manual_seed(0)
+
+    # Create inputs
+    q = torch.randn(BATCH, N_HEADS, SEQ_LEN, D_HEAD, device='cuda', dtype=torch.float16)
+    k = torch.randn(BATCH, N_HEADS, SEQ_LEN, D_HEAD, device='cuda', dtype=torch.float16)
+    v = torch.randn(BATCH, N_HEADS, SEQ_LEN, D_HEAD, device='cuda', dtype=torch.float16)
+
+    sm_scale = 1.0 / (D_HEAD ** 0.5)
+
+    # Create sparse mask (downsampled to block level)
+    downsample_factor = BLOCK
+    downsample_len = math.ceil(SEQ_LEN / downsample_factor)
+    x_ds = torch.randn([BATCH, N_HEADS, downsample_len, downsample_len], device='cuda', dtype=torch.bfloat16)
+    x_ds[:,:,:,0] = 100
+    block_mask = get_sparse_attn_mask_from_topk(x_ds, topk=TOPK)
+
+    # Run Triton kernel
+    program = blocksparse_flashattn(BATCH, N_HEADS, SEQ_LEN, D_HEAD, downsample_len, is_causal=True)
+    kernel = tilelang.compile(program, out_idx=[4])
+    print(kernel.get_kernel_source())
+    tilelang_output = kernel(q, k, v, block_mask)
+
+    # Compute reference
+    # Expand block mask to full attention matrix
+    full_mask = torch.kron(block_mask.float(), 
+                         torch.ones(BLOCK, BLOCK, device='cuda'))
+    full_mask = full_mask[..., :SEQ_LEN, :SEQ_LEN].bool()
+    full_mask = full_mask & torch.tril(torch.ones_like(full_mask))  # Apply causal
+    
+    # PyTorch reference implementation
+    attn = torch.einsum('bhsd,bhtd->bhst', q, k) * sm_scale
+    attn = attn.masked_fill(~full_mask, float('-inf'))
+    attn = F.softmax(attn, dim=-1)
+    ref_output = torch.einsum('bhst,bhtd->bhsd', attn, v)
+    
+    print("ref_output", ref_output)
+    print("tilelang_output", tilelang_output)
+
+
+    # Verify accuracy
+    assert torch.allclose(tilelang_output, ref_output, atol=1e-2, rtol=1e-2), \
+        "TileLang output doesn't match reference"
+    print("Pass topk sparse attention test with qlen == klen")
+
+if __name__ == "__main__":
+    test_topk_sparse_attention()

examples/blocksparse_attention/block_sparse_attn_triton.py

+# Copyright (c) Microsoft Corporation.
+# Licensed under the MIT License.
+import math
+import torch
+
+import triton
+import triton.language as tl
+import torch.nn.functional as F
+
+def is_hip():
+    return triton.runtime.driver.active.get_current_target().backend == "hip"
+
+
+def get_sparse_attn_mask_from_topk(x, topk, use_dense_for_last_block=False):
+    bsz, num_head, downsample_len, _ = x.shape
+    # N_CTX = downsample_len * BLOCK
+    sparse_index = torch.topk(x, topk, dim=-1).indices
+    dense_mask = torch.full([bsz, num_head, downsample_len, downsample_len], False, dtype=torch.bool, device=x.device)
+    dense_mask.scatter_(-1, sparse_index, True)
+    if use_dense_for_last_block:
+        dense_mask[:, :,-2:,:] = True 
+    dense_mask.tril_()
+    return  dense_mask
+
+
+def get_sparse_attn_mask_from_threshold(x, threshold, use_dense_for_last_block=False):
+    dense_mask = x > threshold 
+    if use_dense_for_last_block:
+        dense_mask[:, :,-2:,:] = True 
+    dense_mask.tril_()
+    return  dense_mask
+
+
+
+
+@triton.jit
+def _fwd_kernel_inner(
+    acc, l_i, m_i,
+    q,
+    k_block_col_idx,
+    block_mask_ptr,
+    k_ptrs, v_ptrs,
+    offs_m, offs_n,
+    stride_kt, stride_vt, stride_bmask_n,
+    sm_scale,
+    seqlen_k,
+    past_len,
+    LAST_K_BLOCK: tl.constexpr,
+    BLOCK_M: tl.constexpr,
+    BLOCK_N: tl.constexpr,
+):
+
+    mask_val = tl.load(block_mask_ptr + k_block_col_idx * stride_bmask_n)
+    # print 
+    
+    if k_block_col_idx == 3:
+        print("mask_val", mask_val)
+    if mask_val == True:
+        start_n = k_block_col_idx * BLOCK_N
+        # -- compute qk ----
+
+        k = tl.load(k_ptrs + start_n * stride_kt)
+
+        qk = tl.zeros([BLOCK_M, BLOCK_N], dtype=tl.float32)
+        qk += tl.dot(q, k)
+
+        qk *= sm_scale
+
+        # the following is needed only when LAST_K_BLOCK or BLOCK_M < BLOCK_N
+        if LAST_K_BLOCK :
+            qk += tl.where(offs_m[:, None] + past_len >= (start_n + offs_n[None, :]), 0, float('-inf'))
+
+
+        m_ij = tl.maximum(m_i, tl.max(qk, 1))
+        qk -= m_ij[:, None]
+        p = tl.exp(qk)
+        l_ij = tl.sum(p, 1)
+        alpha = tl.exp(m_i - m_ij)
+        l_i = l_i * alpha + l_ij
+        acc = acc * alpha[:, None]
+        
+        # update acc
+        v = tl.load(v_ptrs + start_n * stride_vt)
+
+        p = p.to(v.type.element_ty)
+
+        acc += tl.dot(p, v)
+        # update m_i and l_i
+        m_i = m_ij
+    return acc, l_i, m_i
+
+
+
+
+@triton.jit
+def _fwd_kernel(
+    Q, K, V, sm_scale,
+    block_mask_ptr,
+    Out,
+    stride_qz, stride_qh, stride_qm, stride_qd,
+    stride_kz, stride_kh, stride_kn, stride_kd,
+    stride_vz, stride_vh, stride_vn, stride_vd,
+    stride_bmz, stride_bmh, stride_bmm, stride_bmn,
+    stride_oz, stride_oh, stride_om, stride_od,
+    H, N_CTX,
+    PAST_LEN,
+    BLOCK_M: tl.constexpr, 
+    BLOCK_N: tl.constexpr,
+    BLOCK_DMODEL: tl.constexpr,
+):
+    Q_LEN = N_CTX - PAST_LEN
+    start_m = tl.program_id(0)
+    off_hz = tl.program_id(1)
+    off_h = off_hz % H
+    off_z = off_hz // H
+    Q += off_z * stride_qz + off_h * stride_qh
+    K += off_z * stride_kz + off_h * stride_kh
+    V += off_z * stride_vz + off_h * stride_vh
+    block_mask_ptr += off_z * stride_bmz + off_h * stride_bmh
+
+    # initialize offsets
+    offs_m = start_m * BLOCK_M + tl.arange(0, BLOCK_M)
+    offs_n = tl.arange(0, BLOCK_N)
+    offs_d = tl.arange(0, BLOCK_DMODEL)
+    off_q = offs_m[:, None] * stride_qm + offs_d[None, :] * stride_qd
+    # off_k = offs_n[:, None] * stride_kn + offs_d[None, :] * stride_kd
+    off_k = offs_n[None, :] * stride_kn + offs_d[:, None] * stride_kd
+    off_v = offs_n[:, None] * stride_vn + offs_d[None, :] * stride_vd
+    # Initialize pointers to Q, K, V
+    q_ptrs = Q + off_q
+    k_ptrs = K + off_k
+    v_ptrs = V + off_v
+    mask_ptrs = block_mask_ptr + start_m * stride_bmm
+
+    m_i = tl.zeros([BLOCK_M], dtype=tl.float32) - float('inf')
+    l_i = tl.zeros([BLOCK_M], dtype=tl.float32)
+    acc = tl.zeros([BLOCK_M, BLOCK_DMODEL], dtype=tl.float32)
+
+    q = tl.load(q_ptrs, mask=offs_m[:, None] < Q_LEN)
+
+    k_block_start = 0
+    k_block_end = tl.cdiv((start_m + 1) * BLOCK_M, BLOCK_N)
+
+    # loop over k, v and update accumulator
+    for col_idx in range(k_block_start, k_block_end):
+        acc, l_i, m_i = _fwd_kernel_inner(
+            acc, l_i, m_i,
+            q,
+            col_idx,
+            mask_ptrs,
+            k_ptrs, v_ptrs,
+            offs_m, offs_n,
+            stride_kn, stride_vn, stride_bmn,
+            sm_scale,
+            N_CTX,
+            PAST_LEN,
+            col_idx == k_block_end - 1,
+            BLOCK_M,
+            BLOCK_N,
+        )
+
+    m_i += tl.math.log(l_i)
+    l_recip = 1 / l_i[:, None]
+    acc = acc * l_recip
+    acc = acc.to(Out.dtype.element_ty) 
+
+
+    off_o = off_z * stride_oz + off_h * stride_oh + offs_m[:, None] * stride_om + offs_d[None, :] * stride_od
+    out_ptrs = Out + off_o
+    tl.store(out_ptrs, acc, mask=offs_m[:, None] < N_CTX)
+
+def _forward(
+        ctx, 
+        q, 
+        k, 
+        v, 
+        block_sparse_mask, 
+        sm_scale, 
+        BLOCK_M=64, 
+        BLOCK_N=64, 
+        num_warps=None, 
+        num_stages=1, 
+        out=None
+    ):
+
+
+    assert q.shape[-1] == k.shape[-1] == v.shape[-1]
+    assert k.shape[2] == v.shape[2]
+    o = out if out is not None else torch.empty_like(q).contiguous()
+    grid = (triton.cdiv(q.shape[2], BLOCK_M), q.shape[0] * q.shape[1])
+
+    assert q.shape[-1] in [64, 128]
+    BLOCK_DMODEL = q.shape[-1]
+
+    if is_hip():
+        num_warps, num_stages = 8, 1
+    else:
+        num_warps, num_stages = 4, 2
+
+    N_CTX = k.shape[2]
+    PAST_LEN = N_CTX - q.shape[2]
+
+
+    H = q.shape[1]
+
+    _fwd_kernel[grid](
+        q, k, v, sm_scale,
+        block_sparse_mask,
+        o,
+        *q.stride(), 
+        *k.stride(), 
+        *v.stride(), 
+        *block_sparse_mask.stride(), 
+        *o.stride(),
+        H, N_CTX,
+        PAST_LEN,
+        BLOCK_M,
+        BLOCK_N,
+        BLOCK_DMODEL,
+        num_warps=num_warps,
+        num_stages=num_stages,
+    )
+
+    return o
+
+
+
+
+class _sparse_attention(torch.autograd.Function):
+
+    @staticmethod
+    def forward(ctx, q, k, v, block_sparse_dense, sm_scale):
+        # shape constraints
+        return _forward(ctx, q, k, v, block_sparse_dense, sm_scale)
+
+    @staticmethod
+    def backward(ctx, do):
+        # No gradient propagation.
+        raise NotImplementedError("It does not support gradient propagation yet")
+        return None, None, None, None, None
+
+block_sparse_triton_fn = _sparse_attention.apply
+
+
+
+def test_topk_sparse_attention():
+    # Config
+    BATCH, N_HEADS, SEQ_LEN, D_HEAD = 1, 1, 256, 64
+    TOPK = 2  # Keep top 8 elements per row
+    BLOCK = 64
+    torch.manual_seed(0)
+
+    # Create inputs
+    q = torch.randn(BATCH, N_HEADS, SEQ_LEN, D_HEAD, device='cuda', dtype=torch.bfloat16)
+    k = torch.randn(BATCH, N_HEADS, SEQ_LEN, D_HEAD, device='cuda', dtype=torch.bfloat16)
+    v = torch.randn(BATCH, N_HEADS, SEQ_LEN, D_HEAD, device='cuda', dtype=torch.bfloat16)
+    sm_scale = 1.0 / (D_HEAD ** 0.5)
+
+    # Create sparse mask (downsampled to block level)
+    downsample_factor = BLOCK
+    downsample_len = math.ceil(SEQ_LEN / downsample_factor)
+    print("downsample_len", downsample_len)
+    
+    x_ds = torch.randn([BATCH, N_HEADS, downsample_len, downsample_len], device='cuda', dtype=torch.bfloat16)
+    x_ds[:,:,:,0] = 100
+    print("x_ds.shape", x_ds.shape)
+    block_mask = get_sparse_attn_mask_from_topk(x_ds, topk=downsample_len)
+    # print("block_mask", block_mask)
+    print("block_mask.shape", block_mask.shape)
+
+    # Run Triton kernel
+    triton_output = block_sparse_triton_fn(
+        q, k, v, 
+        block_mask,
+        sm_scale
+    )
+
+    # Compute reference
+    # Expand block mask to full attention matrix
+    full_mask = torch.kron(block_mask.float(), 
+                         torch.ones(BLOCK, BLOCK, device='cuda'))
+    full_mask = full_mask[..., :SEQ_LEN, :SEQ_LEN].bool()
+    full_mask = full_mask & torch.tril(torch.ones_like(full_mask))  # Apply causal
+    
+    # PyTorch reference implementation
+    attn = torch.einsum('bhsd,bhtd->bhst', q, k) * sm_scale
+    attn = attn.masked_fill(~full_mask, float('-inf'))
+    attn = F.softmax(attn, dim=-1)
+    ref_output = torch.einsum('bhst,bhtd->bhsd', attn, v)
+    
+    # print("ref_output", ref_output)
+    # print("triton_output", triton_output)
+
+
+    # Verify accuracy
+    assert torch.allclose(triton_output, ref_output, atol=1e-2, rtol=1e-2), \
+        "Triton output doesn't match reference"
+    print("Pass topk sparse attention test with qlen == klen")
+
+
+
+# def test_topk_sparse_attention_qlt_kl():
+#     BATCH, N_HEADS = 2, 4
+#     Q_LEN, K_LEN, D_HEAD = 128, 256, 64  # qlen < klen; here, past_len = 256 - 128 = 128.
+#     TOPK = 1  
+#     BLOCK = 64  # block size used in downsampling
+#     torch.manual_seed(0)
+
+#     # Create inputs.
+#     q = torch.randn(BATCH, N_HEADS, Q_LEN, D_HEAD, device='cuda', dtype=torch.bfloat16)
+#     k = torch.randn(BATCH, N_HEADS, K_LEN, D_HEAD, device='cuda', dtype=torch.bfloat16)
+#     v = torch.randn(BATCH, N_HEADS, K_LEN, D_HEAD, device='cuda', dtype=torch.bfloat16)
+#     # softmax scale
+#     sm_scale = 1.0 / (D_HEAD ** 0.5)
+
+#     downsample_factor = BLOCK
+#     print("downsample_factor", downsample_factor)
+#     downsample_len = math.ceil(K_LEN / downsample_factor)  # number of blocks along one dimension
+#     print("downsample_len", downsample_len)
+#     x_ds = torch.randn(BATCH, N_HEADS, downsample_len, downsample_len, 
+#                         device='cuda', dtype=torch.bfloat16)
+#     # Force the first column to be high so that the first block is always selected.
+#     x_ds[:, :, :, 0] = 100
+#     block_mask = get_sparse_attn_mask_from_topk(x_ds, topk=TOPK)
+#     print("block_mask", block_mask)
+#     print("block_mask.shape", block_mask.shape)
+#     # Run Triton kernel.
+#     triton_output = block_sparse_triton_fn(q, k, v, block_mask, sm_scale)
+
+#     past_len = K_LEN - Q_LEN
+
+#     attn = torch.einsum('bhsd,bhtd->bhst', q, k) * sm_scale
+
+#     full_mask_full = torch.kron(block_mask.float(), torch.ones(BLOCK, BLOCK, device='cuda')).bool()
+#     full_mask_full = full_mask_full[..., :K_LEN, :K_LEN]
+
+#     effective_mask = full_mask_full[..., past_len:K_LEN, :]  # shape: (B, H, Q_LEN, K_LEN)
+
+
+#     i_global = torch.arange(past_len, K_LEN, device=k.device).unsqueeze(1)  # shape: (Q_LEN, 1)
+#     j_global = torch.arange(K_LEN, device=k.device).unsqueeze(0)              # shape: (1, K_LEN)
+#     causal_mask = (j_global <= i_global)  # shape: (Q_LEN, K_LEN)
+
+#     final_mask = effective_mask & causal_mask  # shape: (B, H, Q_LEN, K_LEN)
+
+#     attn = attn.masked_fill(~final_mask, float('-inf'))
+#     attn = F.softmax(attn, dim=-1)
+#     ref_output = torch.einsum('bhst,bhtd->bhsd', attn, v)
+
+#     # Verify accuracy.
+#     assert torch.allclose(triton_output, ref_output, atol=1e-2, rtol=1e-2), \
+#         "Triton output doesn't match reference when qlen < klen"
+    
+#     print("Pass topk sparse attention test with qlen < klen")
+
+
+if __name__ == "__main__":
+    test_topk_sparse_attention()
+    # test_topk_sparse_attention_qlt_kl()
\ No newline at end of file

examples/flash_attention/example_mha_fwd_bhsd.py

+# Copyright (c) Microsoft Corporation.
+# Licensed under the MIT License.
+
+import torch
+import torch.nn.functional as F
+import tilelang
+from tilelang import Profiler
+from tilelang.autotuner import *
+import tilelang.language as T
+import itertools
+import argparse
+from functools import partial
+
+
+def get_configs():
+    block_M = [128]
+    block_N = [128]
+    num_stages = [2]
+    threads = [256]
+    _configs = list(itertools.product(block_M, block_N, num_stages, threads))
+
+    configs = [{
+        'block_M': c[0],
+        'block_N': c[1],
+        'num_stages': c[2],
+        'threads': c[3]
+    } for c in _configs]
+    return configs
+
+
+def flashattn(batch, heads, seq_len, dim, is_causal, tune=False):
+    scale = (1.0 / dim)**0.5 * 1.44269504  # log2(e)
+    shape = [batch, heads, seq_len, dim]
+    dtype = "float16"
+    accum_dtype = "float"
+
+    def kernel_func(block_M, block_N, num_stages, threads):
+
+        @T.macro
+        def MMA0(
+            K: T.Buffer(shape, dtype),
+            Q_shared: T.Buffer([block_M, dim], dtype),
+            K_shared: T.Buffer([block_N, dim], dtype),
+            acc_s: T.Buffer([block_M, block_N], accum_dtype),
+            k: T.int32,
+            bx: T.int32,
+            by: T.int32,
+            bz: T.int32,
+        ):
+            T.copy(K[bz, by, k * block_N:(k + 1) * block_N, :], K_shared)
+            if is_causal:
+                for i, j in T.Parallel(block_M, block_N):
+                    acc_s[i, j] = T.if_then_else(bx * block_M + i >= k * block_N + j, 0,
+                                                 -T.infinity(acc_s.dtype))
+            else:
+                T.clear(acc_s)
+            T.gemm(Q_shared, K_shared, acc_s, transpose_B=True, policy=T.GemmWarpPolicy.FullRow)
+
+        @T.macro
+        def MMA1(
+                V: T.Buffer(shape, dtype),
+                V_shared: T.Buffer([block_M, dim], dtype),
+                acc_s_cast: T.Buffer([block_M, block_N], dtype),
+                acc_o: T.Buffer([block_M, dim], accum_dtype),
+                k: T.int32,
+                by: T.int32,
+                bz: T.int32,
+        ):
+            T.copy(V[bz, by, k * block_N:(k + 1) * block_N, :], V_shared)
+            T.gemm(acc_s_cast, V_shared, acc_o, policy=T.GemmWarpPolicy.FullRow)
+
+        @T.macro
+        def Softmax(
+                acc_s: T.Buffer([block_M, block_N], accum_dtype),
+                acc_s_cast: T.Buffer([block_M, block_N], dtype),
+                scores_max: T.Buffer([block_M], accum_dtype),
+                scores_max_prev: T.Buffer([block_M], accum_dtype),
+                scores_scale: T.Buffer([block_M], accum_dtype),
+                scores_sum: T.Buffer([block_M], accum_dtype),
+                logsum: T.Buffer([block_M], accum_dtype),
+        ):
+            T.copy(scores_max, scores_max_prev)
+            T.fill(scores_max, -T.infinity(accum_dtype))
+            T.reduce_max(acc_s, scores_max, dim=1, clear=False)
+            # To do causal softmax, we need to set the scores_max to 0 if it is -inf
+            # This process is called Check_inf in FlashAttention3 code, and it only need to be done
+            # in the first ceil_div(kBlockM, kBlockN) steps.
+            # for i in T.Parallel(block_M):
+            #     scores_max[i] = T.if_then_else(scores_max[i] == -T.infinity(accum_dtype), 0, scores_max[i])
+            for i in T.Parallel(block_M):
+                scores_scale[i] = T.exp2(scores_max_prev[i] * scale - scores_max[i] * scale)
+            for i, j in T.Parallel(block_M, block_N):
+                # Instead of computing exp(x - max), we compute exp2(x * log_2(e) -
+                # max * log_2(e)) This allows the compiler to use the ffma
+                # instruction instead of fadd and fmul separately.
+                acc_s[i, j] = T.exp2(acc_s[i, j] * scale - scores_max[i] * scale)
+            T.reduce_sum(acc_s, scores_sum, dim=1)
+            for i in T.Parallel(block_M):
+                logsum[i] = logsum[i] * scores_scale[i] + scores_sum[i]
+            T.copy(acc_s, acc_s_cast)
+
+        @T.macro
+        def Rescale(
+                acc_o: T.Buffer([block_M, dim], accum_dtype),
+                scores_scale: T.Buffer([block_M], accum_dtype),
+        ):
+            for i, j in T.Parallel(block_M, dim):
+                acc_o[i, j] *= scores_scale[i]
+
+        @T.prim_func
+        def main(
+                Q: T.Buffer(shape, dtype),
+                K: T.Buffer(shape, dtype),
+                V: T.Buffer(shape, dtype),
+                Output: T.Buffer(shape, dtype),
+        ):
+            with T.Kernel(
+                    T.ceildiv(seq_len, block_M), heads, batch, threads=threads) as (bx, by, bz):
+                Q_shared = T.alloc_shared([block_M, dim], dtype)
+                K_shared = T.alloc_shared([block_N, dim], dtype)
+                V_shared = T.alloc_shared([block_N, dim], dtype)
+                O_shared = T.alloc_shared([block_M, dim], dtype)
+                acc_s = T.alloc_fragment([block_M, block_N], accum_dtype)
+                acc_s_cast = T.alloc_fragment([block_M, block_N], dtype)
+                acc_o = T.alloc_fragment([block_M, dim], accum_dtype)
+                scores_max = T.alloc_fragment([block_M], accum_dtype)
+                scores_max_prev = T.alloc_fragment([block_M], accum_dtype)
+                scores_scale = T.alloc_fragment([block_M], accum_dtype)
+                scores_sum = T.alloc_fragment([block_M], accum_dtype)
+                logsum = T.alloc_fragment([block_M], accum_dtype)
+
+                T.copy(Q[bz, by, bx * block_M:(bx + 1) * block_M, :], Q_shared)
+                T.fill(acc_o, 0)
+                T.fill(logsum, 0)
+                T.fill(scores_max, -T.infinity(accum_dtype))
+
+                loop_range = (
+                    T.min(T.ceildiv(seq_len, block_N), T.ceildiv(
+                        (bx + 1) * block_M, block_N)) if is_causal else T.ceildiv(seq_len, block_N))
+
+                for k in T.Pipelined(loop_range, num_stages=num_stages):
+                    MMA0(K, Q_shared, K_shared, acc_s, k, bx, by, bz)
+                    Softmax(acc_s, acc_s_cast, scores_max, scores_max_prev, scores_scale,
+                            scores_sum, logsum)
+                    Rescale(acc_o, scores_scale)
+                    MMA1(V, V_shared, acc_s_cast, acc_o, k, by, bz)
+                for i, j in T.Parallel(block_M, dim):
+                    acc_o[i, j] /= logsum[i]
+                T.copy(acc_o, O_shared)
+                T.copy(O_shared, Output[bz, by, bx * block_M:(bx + 1) * block_M, :])
+
+        return main
+
+    if tune:
+
+        @autotune(
+            configs=get_configs(),
+            keys=["block_M", "block_N", "num_stages", "threads"],
+            warmup=10,
+            rep=10)
+        @jit(
+            out_idx=[3],
+            supply_type=tilelang.TensorSupplyType.Integer,
+            ref_prog=None,
+            profiler="auto")
+        def kernel(block_M=None, block_N=None, num_stages=None, threads=None):
+            return kernel_func(block_M, block_N, num_stages, threads)
+
+        return kernel()
+    else:
+
+        def kernel(block_M, block_N, num_stages, threads):
+            return kernel_func(block_M, block_N, num_stages, threads)
+
+        return kernel
+
+
+def ref_program(Q, K, V, is_causal):
+    dim = Q.size(-1)
+    scores = torch.einsum('bhqd,bhkd->bhqk', Q, K)
+    scores = scores / torch.sqrt(torch.tensor(dim, dtype=scores.dtype))
+    if is_causal:
+        seq_len = Q.size(1)
+        mask = torch.tril(torch.ones(seq_len, seq_len, device=scores.device))
+        mask = mask.unsqueeze(0).unsqueeze(0)
+        scores = scores.masked_fill(mask == 0, float('-inf'))
+    attention_weights = F.softmax(scores, dim=-1)
+    output = torch.einsum('bhqk,bhkd->bhqd', attention_weights, V)
+    return output
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--batch', type=int, default=8, help='batch size')
+    parser.add_argument('--heads', type=int, default=32, help='heads')
+    parser.add_argument('--seq_len', type=int, default=4096, help='sequence length')
+    parser.add_argument('--dim', type=int, default=128, help='dim')
+    parser.add_argument('--is_causal', action='store_true', help='causal')
+    parser.add_argument('--tune', action='store_true', help='tune configs')
+    args = parser.parse_args()
+    batch, heads, seq_len, dim, is_causal = args.batch, args.heads, args.seq_len, args.dim, args.is_causal
+    flops_per_matmul = 2.0 * batch * heads * seq_len * seq_len * dim
+    total_flops = 2 * flops_per_matmul
+    if is_causal:
+        total_flops *= 0.5
+
+    if (not args.tune):
+        program = flashattn(
+            batch, heads, seq_len, dim, is_causal, tune=args.tune)(
+                block_M=128, block_N=128, num_stages=1, threads=128)
+        ref_program = partial(ref_program, is_causal=is_causal)
+        mod, params = tilelang.lower(program)
+        mod = Profiler(mod, params, [3], tilelang.TensorSupplyType.Normal)
+        mod.assert_allclose(ref_program, rtol=0.01, atol=0.01)
+        print("All checks pass.")
+        latency = mod.do_bench(ref_program, warmup=500)
+        print("Ref: {:.2f} ms".format(latency))
+        print("Ref: {:.2f} TFlops".format(total_flops / latency * 1e-9))
+        latency = mod.do_bench(mod.func, warmup=500)
+        print("Tile-lang: {:.2f} ms".format(latency))
+        print("Tile-lang: {:.2f} TFlops".format(total_flops / latency * 1e-9))
+    else:
+        best_latency, best_config, _ = flashattn(
+            batch, heads, seq_len, dim, is_causal, tune=args.tune)
+        print(f"Best latency: {best_latency}")
+        print(f"Best TFlops: {total_flops / best_latency * 1e-9}")
+        print(f"Best config: {best_config}")

src/tl_templates/cuda/debug.h

@@ -4,10 +4,31 @@
 #include <stdio.h>
 
 // Template declaration for device-side debug printing (variable only)
-template <typename T> __device__ void debug_print_var(char *msg, T var);
+template <typename T> __device__ void debug_print_var(const char *msg, T var);
+
+// Specialization for signed char type
+template <>
+__device__ void debug_print_var<signed char>(const char *msg, signed char var) {
+  printf("msg='%s' BlockIdx=(%d, %d, %d), ThreadIdx=(%d, %d, %d): dtype=signed "
+         "char "
+         "value=%d\n",
+         msg, blockIdx.x, blockIdx.y, blockIdx.z, threadIdx.x, threadIdx.y,
+         threadIdx.z, var);
+}
+
+// Specialization for unsigned char type
+template <>
+__device__ void debug_print_var<unsigned char>(const char *msg,
+                                               unsigned char var) {
+  printf("msg='%s' BlockIdx=(%d, %d, %d), ThreadIdx=(%d, %d, %d): "
+         "dtype=unsigned char "
+         "value=%d\n",
+         msg, blockIdx.x, blockIdx.y, blockIdx.z, threadIdx.x, threadIdx.y,
+         threadIdx.z, var);
+}
 
 // Specialization for integer type
-template <> __device__ void debug_print_var<int>(char *msg, int var) {
+template <> __device__ void debug_print_var<int>(const char *msg, int var) {
   printf("msg='%s' BlockIdx=(%d, %d, %d), ThreadIdx=(%d, %d, %d): dtype=int "
          "value=%d\n",
          msg, blockIdx.x, blockIdx.y, blockIdx.z, threadIdx.x, threadIdx.y,
@@ -15,7 +36,7 @@ template <> __device__ void debug_print_var<int>(char *msg, int var) {
 }
 
 // Specialization for float type
-template <> __device__ void debug_print_var<float>(char *msg, float var) {
+template <> __device__ void debug_print_var<float>(const char *msg, float var) {
   printf("msg='%s' BlockIdx=(%d, %d, %d), ThreadIdx=(%d, %d, %d): dtype=float "
          "value=%f\n",
          msg, blockIdx.x, blockIdx.y, blockIdx.z, threadIdx.x, threadIdx.y,
@@ -23,7 +44,7 @@ template <> __device__ void debug_print_var<float>(char *msg, float var) {
 }
 
 // Specialization for half type
-template <> __device__ void debug_print_var<half>(char *msg, half var) {
+template <> __device__ void debug_print_var<half>(const char *msg, half var) {
   printf("msg='%s' BlockIdx=(%d, %d, %d), ThreadIdx=(%d, %d, %d): dtype=half "
          "value=%f\n",
          msg, blockIdx.x, blockIdx.y, blockIdx.z, threadIdx.x, threadIdx.y,
@@ -31,7 +52,8 @@ template <> __device__ void debug_print_var<half>(char *msg, half var) {
 }
 
 // Specialization for half_t type
-template <> __device__ void debug_print_var<half_t>(char *msg, half_t var) {
+template <>
+__device__ void debug_print_var<half_t>(const char *msg, half_t var) {
   printf("msg='%s' BlockIdx=(%d, %d, %d), ThreadIdx=(%d, %d, %d): dtype=half_t "
          "value=%f\n",
          msg, blockIdx.x, blockIdx.y, blockIdx.z, threadIdx.x, threadIdx.y,
@@ -40,7 +62,7 @@ template <> __device__ void debug_print_var<half_t>(char *msg, half_t var) {
 
 // Specialization for bfloat16_t type
 template <>
-__device__ void debug_print_var<bfloat16_t>(char *msg, bfloat16_t var) {
+__device__ void debug_print_var<bfloat16_t>(const char *msg, bfloat16_t var) {
   printf("msg='%s' BlockIdx=(%d, %d, %d), ThreadIdx=(%d, %d, %d): "
          "dtype=bfloat16_t value=%f\n",
          msg, blockIdx.x, blockIdx.y, blockIdx.z, threadIdx.x, threadIdx.y,
@@ -48,7 +70,8 @@ __device__ void debug_print_var<bfloat16_t>(char *msg, bfloat16_t var) {
 }
 
 // Specialization for double type
-template <> __device__ void debug_print_var<double>(char *msg, double var) {
+template <>
+__device__ void debug_print_var<double>(const char *msg, double var) {
   printf("msg='%s' BlockIdx=(%d, %d, %d), ThreadIdx=(%d, %d, %d): dtype=double "
          "value=%lf\n",
          msg, blockIdx.x, blockIdx.y, blockIdx.z, threadIdx.x, threadIdx.y,
@@ -62,13 +85,36 @@ template <> __device__ void debug_print_var<double>(char *msg, double var) {
 
 // Template declaration for device-side debug printing (buffer only)
 template <typename T>
-__device__ void debug_print_buffer_value(char *msg, char *buf_name, int index,
-                                         T var);
+__device__ void debug_print_buffer_value(const char *msg, const char *buf_name,
+                                         int index, T var);
+
+// Specialization for signed char type
+template <>
+__device__ void
+debug_print_buffer_value<signed char>(const char *msg, const char *buf_name,
+                                      int index, signed char var) {
+  printf("msg='%s' BlockIdx=(%d, %d, %d), ThreadIdx=(%d, %d, %d): buffer=%s, "
+         "index=%d, dtype=signed char value=%d\n",
+         msg, blockIdx.x, blockIdx.y, blockIdx.z, threadIdx.x, threadIdx.y,
+         threadIdx.z, buf_name, index, var);
+}
+
+// Specialization for unsiged char type
+template <>
+__device__ void debug_print_buffer_value<char>(const char *msg,
+                                               const char *buf_name, int index,
+                                               char var) {
+  printf("msg='%s' BlockIdx=(%d, %d, %d), ThreadIdx=(%d, %d, %d): buffer=%s, "
+         "index=%d, dtype=char value=%d\n",
+         msg, blockIdx.x, blockIdx.y, blockIdx.z, threadIdx.x, threadIdx.y,
+         threadIdx.z, buf_name, index, var);
+}
 
 // Specialization for integer type
 template <>
-__device__ void debug_print_buffer_value<int>(char *msg, char *buf_name,
-                                              int index, int var) {
+__device__ void debug_print_buffer_value<int>(const char *msg,
+                                              const char *buf_name, int index,
+                                              int var) {
   printf("msg='%s' BlockIdx=(%d, %d, %d), ThreadIdx=(%d, %d, %d): buffer=%s, "
          "index=%d, dtype=int value=%d\n",
          msg, blockIdx.x, blockIdx.y, blockIdx.z, threadIdx.x, threadIdx.y,
@@ -77,8 +123,9 @@ __device__ void debug_print_buffer_value<int>(char *msg, char *buf_name,
 
 // Specialization for float type
 template <>
-__device__ void debug_print_buffer_value<float>(char *msg, char *buf_name,
-                                                int index, float var) {
+__device__ void debug_print_buffer_value<float>(const char *msg,
+                                                const char *buf_name, int index,
+                                                float var) {
   printf("msg='%s' BlockIdx=(%d, %d, %d), ThreadIdx=(%d, %d, %d): buffer=%s, "
          "index=%d, dtype=float value=%f\n",
          msg, blockIdx.x, blockIdx.y, blockIdx.z, threadIdx.x, threadIdx.y,
@@ -87,8 +134,9 @@ __device__ void debug_print_buffer_value<float>(char *msg, char *buf_name,
 
 // Specialization for half type
 template <>
-__device__ void debug_print_buffer_value<half>(char *msg, char *buf_name,
-                                               int index, half var) {
+__device__ void debug_print_buffer_value<half>(const char *msg,
+                                               const char *buf_name, int index,
+                                               half var) {
   printf("msg='%s' BlockIdx=(%d, %d, %d), ThreadIdx=(%d, %d, %d): buffer=%s, "
          "index=%d, dtype=half value=%f\n",
          msg, blockIdx.x, blockIdx.y, blockIdx.z, threadIdx.x, threadIdx.y,
@@ -97,7 +145,8 @@ __device__ void debug_print_buffer_value<half>(char *msg, char *buf_name,
 
 // Specialization for half_t type
 template <>
-__device__ void debug_print_buffer_value<half_t>(char *msg, char *buf_name,
+__device__ void debug_print_buffer_value<half_t>(const char *msg,
+                                                 const char *buf_name,
                                                  int index, half_t var) {
   printf("msg='%s' BlockIdx=(%d, %d, %d), ThreadIdx=(%d, %d, %d): buffer=%s, "
          "index=%d, dtype=half_t value=%f\n",
@@ -107,9 +156,9 @@ __device__ void debug_print_buffer_value<half_t>(char *msg, char *buf_name,
 
 // Specialization for bfloat16_t type
 template <>
-__device__ void debug_print_buffer_value<bfloat16_t>(char *msg, char *buf_name,
-                                                     int index,
-                                                     bfloat16_t var) {
+__device__ void
+debug_print_buffer_value<bfloat16_t>(const char *msg, const char *buf_name,
+                                     int index, bfloat16_t var) {
   printf("msg='%s' BlockIdx=(%d, %d, %d), ThreadIdx=(%d, %d, %d): buffer=%s, "
          "index=%d, dtype=bfloat16_t value=%f\n",
          msg, blockIdx.x, blockIdx.y, blockIdx.z, threadIdx.x, threadIdx.y,
@@ -118,7 +167,8 @@ __device__ void debug_print_buffer_value<bfloat16_t>(char *msg, char *buf_name,
 
 // Specialization for double type
 template <>
-__device__ void debug_print_buffer_value<double>(char *msg, char *buf_name,
+__device__ void debug_print_buffer_value<double>(const char *msg,
+                                                 const char *buf_name,
                                                  int index, double var) {
   printf("msg='%s' BlockIdx=(%d, %d, %d), ThreadIdx=(%d, %d, %d): buffer=%s, "
          "index=%d, dtype=double value=%lf\n",