biondizzle/nvfp4-megamoe-kernel
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

auto: pre-test commit

Browse Source
This commit is contained in:
biondizzle
2026-05-22 23:30:43 +00:00
parent 8bafb26162
commit 35738bef08
1 changed files with 133 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

133
tests/diag_tma_shapes3.py Normal file
View File

@@ -0,0 +1,133 @@
"""Diagnostic: print tma_partition output shapes inside @cute.kernel context."""
import torch, cutlass, cutlass.cute as cute, cutlass.utils as utils, cutlass.pipeline as pipeline
from cutlass.cute.nvgpu import cpasync, tcgen05
from cutlass import Float32, BFloat16, Int32, Boolean, const_expr
from cutlass.utils import LayoutEnum
from cutlass.utils.tmem_allocator import find_tmem_tensor_col_offset
import cuda.bindings.driver as cuda
import cutlass.torch as ct
import math
HEAD_DIM = 64
class DiagShapes:
def __init__(self, s_k=256):
self.s_k = s_k
self.n_kv_tiles = s_k // 128
self.qk_acc_dtype = Float32
self.q_dtype = BFloat16; self.o_dtype = BFloat16; self.c_dtype = BFloat16
self.use_2cta_instrs = False; self.epilog_sync_bar_id = 1
self.cluster_shape_mn = (1, 1); self.cta_group = tcgen05.CtaGroup.ONE
self.epilogue_warp_id = (0,1,2,3); self.mma_warp_id = 4; self.tma_warp_id = 5
self.threads_per_cta = 192; self.kv_stage = 2; self.q_stage = 1; self.num_c_stage = 2
def _setup(self, qk_mma, pv_mma):
qk_ik = cute.size(qk_mma.shape_mnk, mode=[2])
self.qk_mma_tiler = (128, 128, qk_ik * 4)
pv_ik = cute.size(pv_mma.shape_mnk, mode=[2])
self.pv_mma_tiler = (128, HEAD_DIM, pv_ik * (128 // pv_ik))
self.mma_tiler = self.qk_mma_tiler
self.cluster_layout_vmnk = cute.tiled_divide(cute.make_layout((1,1,1)), (qk_mma.thr_id.shape,))
self.cta_tile_shape_mnk = (self.qk_mma_tiler[0]//cute.size(qk_mma.thr_id.shape), HEAD_DIM, self.qk_mma_tiler[2])
self.c_layout = LayoutEnum.ROW_MAJOR
self.epi_tile = utils.sm100.compute_epilogue_tile_shape(self.cta_tile_shape_mnk, False, self.c_layout, self.o_dtype)
self.num_ab_stage = 1; self.num_acc_stage = 1
self.q_smem_s = utils.sm100.make_smem_layout_a(qk_mma, self.qk_mma_tiler, self.q_dtype, self.q_stage)
self.k_smem_s = utils.sm100.make_smem_layout_b(qk_mma, self.qk_mma_tiler, self.q_dtype, self.kv_stage)
self.v_smem_s = utils.sm100.make_smem_layout_b(pv_mma, self.pv_mma_tiler, self.q_dtype, self.kv_stage)
self.c_smem_s = utils.sm100.make_smem_layout_epi(self.o_dtype, self.c_layout, self.epi_tile, 2)
self.p_tmem_s = utils.sm100.make_smem_layout_a(pv_mma, self.pv_mma_tiler, self.q_dtype, 1)
@cute.jit
def __call__(self, q, k, v, c, stream):
self.q_dtype = q.element_type; self.o_dtype = c.element_type; self.c_dtype = self.o_dtype
self.a_major = LayoutEnum.from_tensor(q).mma_major_mode()
self.b_major = LayoutEnum.from_tensor(k).mma_major_mode()
v_fmha = cute.make_tensor(v.iterator, cute.make_layout((HEAD_DIM, self.s_k, 1), stride=(1, HEAD_DIM, HEAD_DIM * self.s_k)))
self.v_major = LayoutEnum.from_tensor(v_fmha).mma_major_mode()
self.c_layout = LayoutEnum.from_tensor(c)
qk_mma = utils.sm100.make_trivial_tiled_mma(self.q_dtype, self.q_dtype, self.a_major, self.b_major, self.qk_acc_dtype, self.cta_group, (128,128), tcgen05.OperandSource.SMEM)
pv_mma = utils.sm100.make_trivial_tiled_mma(self.q_dtype, self.q_dtype, cute.nvgpu.OperandMajorMode.K, self.v_major, self.qk_acc_dtype, self.cta_group, (128,HEAD_DIM), tcgen05.OperandSource.TMEM)
self._setup(qk_mma, pv_mma)
q_s = cute.slice_(self.q_smem_s,(None,None,None,0))
k_s = cute.slice_(self.k_smem_s,(None,None,None,0))
v_s = cute.slice_(self.v_smem_s,(None,None,None,0))
tma_q,mQ = cute.nvgpu.make_tiled_tma_atom_A(utils.sm100.cluster_shape_to_tma_atom_A(self.cluster_shape_mn,qk_mma.thr_id),q,q_s,self.qk_mma_tiler,qk_mma,self.cluster_layout_vmnk.shape)
tma_k,mK = cute.nvgpu.make_tiled_tma_atom_B(utils.sm100.cluster_shape_to_tma_atom_B(self.cluster_shape_mn,qk_mma.thr_id),k,k_s,self.qk_mma_tiler,qk_mma,self.cluster_layout_vmnk.shape)
tma_v,mV = cute.nvgpu.make_tiled_tma_atom_B(utils.sm100.cluster_shape_to_tma_atom_B(self.cluster_shape_mn,pv_mma.thr_id),v_fmha,v_s,self.pv_mma_tiler,pv_mma,self.cluster_layout_vmnk.shape)
self._kernel(qk_mma, pv_mma, tma_q, mQ, tma_k, mK, tma_v, mV, self.cluster_layout_vmnk, self.q_smem_s, self.k_smem_s, self.v_smem_s).launch(grid=(1,1,1),block=[self.threads_per_cta,1,1],stream=stream)
@cute.kernel
def _kernel(self, qk_mma, pv_mma, tma_q, mQ, tma_k, mK, tma_v, mV, cl_vmnk, q_smem_s, k_smem_s, v_smem_s):
tidx,_,_ = cute.arch.thread_idx()
sQ = cute.make_tensor(element_type=self.q_dtype, layout=q_smem_s.outer, byte_alignment=128, swizzle=q_smem_s.inner)
sK = cute.make_tensor(element_type=self.q_dtype, layout=k_smem_s.outer, byte_alignment=128, swizzle=k_smem_s.inner)
sV = cute.make_tensor(element_type=self.q_dtype, layout=v_smem_s.outer, byte_alignment=128, swizzle=v_smem_s.inner)
q_s = cute.slice_(q_smem_s,(None,None,None,0))
k_s = cute.slice_(k_smem_s,(None,None,None,0))
v_s = cute.slice_(v_smem_s,(None,None,None,0))
gQ = cute.local_tile(mQ,cute.slice_(self.qk_mma_tiler,(None,0,None)),(None,None,None))
gK = cute.local_tile(mK,cute.slice_(self.qk_mma_tiler,(0,None,None)),(None,None,None))
gV = cute.local_tile(mV,cute.slice_(self.pv_mma_tiler,(0,None,None)),(None,None,None))
qk_thr = qk_mma.get_slice(0); pv_thr = pv_mma.get_slice(0)
tCgQ = qk_thr.partition_A(gQ); tCgK = qk_thr.partition_B(gK)
tCgV = pv_thr.partition_B(gV)
a_lay = cute.make_layout(cute.slice_(cl_vmnk,(0,0,None,0)).shape)
b_lay = cute.make_layout(cute.slice_(cl_vmnk,(0,None,0,0)).shape)
tAsQ,tAgQ = cpasync.tma_partition(tma_q,0,a_lay,cute.group_modes(q_s,0,3),cute.group_modes(tCgQ,0,3))
tBsK,tBgK = cpasync.tma_partition(tma_k,0,b_lay,cute.group_modes(k_s,0,3),cute.group_modes(tCgK,0,3))
tVsV,tVgV = cpasync.tma_partition(tma_v,0,b_lay,cute.group_modes(v_s,0,3),cute.group_modes(tCgV,0,3))
# Print shapes at trace time
print(f"DIAG tAgQ: shape={cute.shape(tAgQ)} stride={tAgQ.layout.stride}")
print(f"DIAG tBgK: shape={cute.shape(tBgK)} stride={tBgK.layout.stride}")
print(f"DIAG tVgV: shape={cute.shape(tVgV)} stride={tVgV.layout.stride}")
print(f"DIAG tAsQ: shape={cute.shape(tAsQ)} stride={tAsQ.layout.stride}")
print(f"DIAG tBsK: shape={cute.shape(tBsK)} stride={tBsK.layout.stride}")
print(f"DIAG tVsV: shape={cute.shape(tVsV)} stride={tVsV.layout.stride}")
# Try pre-slices and print
tAgQ_s = tAgQ[(None,0,None,0)]
print(f"DIAG tAgQ after (None,0,None,0): shape={cute.shape(tAgQ_s)} stride={tAgQ_s.layout.stride}")
tBgK_s = tBgK[(None,0,None,0)]
print(f"DIAG tBgK after (None,0,None,0): shape={cute.shape(tBgK_s)} stride={tBgK_s.layout.stride}")
tVgV_s = tVgV[(None,0,None,0)]
print(f"DIAG tVgV after (None,0,None,0): shape={cute.shape(tVgV_s)} stride={tVgV_s.layout.stride}")
# Also try old pre-slice
tBgK_old = tBgK[(None,None,0,0)]
print(f"DIAG tBgK after (None,None,0,0): shape={cute.shape(tBgK_old)} stride={tBgK_old.layout.stride}")
tVgV_old = tVgV[(None,0,None,0)]
print(f"DIAG tVgV after (None,0,None,0): shape={cute.shape(tVgV_old)} stride={tVgV_old.layout.stride}")
def test():
torch.manual_seed(42)
n = 256
m, hd = 128, HEAD_DIM
q = torch.randn(m, hd, 1, dtype=torch.bfloat16, device='cuda')
k = torch.randn(n, hd, 1, dtype=torch.bfloat16, device='cuda')
v = torch.randn(n, hd, dtype=torch.bfloat16, device='cuda')
c = torch.zeros(m, hd, 1, dtype=torch.bfloat16, device='cuda')
v_kernel = v.unsqueeze(-1)
mQ = ct.from_dlpack(q).mark_layout_dynamic(leading_dim=ct.get_leading_dim(q))
mK = ct.from_dlpack(k).mark_layout_dynamic(leading_dim=ct.get_leading_dim(k))
mV = ct.from_dlpack(v_kernel).mark_layout_dynamic(leading_dim=ct.get_leading_dim(v_kernel))
stream = cuda.CUstream(torch.cuda.current_stream().cuda_stream)
kernel = DiagShapes(s_k=n)
print('Compiling...', flush=True)
compiled = cute.compile(kernel, mQ, mK, mV, ct.from_dlpack(c).mark_layout_dynamic(leading_dim=ct.get_leading_dim(c)), stream)
print('Compiled. Running...', flush=True)
if __name__ == '__main__':
test()