MKLDNN backend

Created On: May 10, 2025 | Last Updated On: Jun 26, 2025

MKLDNN is an open-source cross-platform performance library of basic building blocks for deep learning applications.

# The flag below controls whether enable MKLDNN backend in Pytorch.
torch.backends.mkldnn.enabled = True

Users can disable MKLDNN backend by:

torch.backends.mkldnn.enabled = False

Bfloat16 (BF16) on MKLDNN backend

Starting in PyTorch 2.4, there is a set of APIs to control the internal computation precision for float32 operators.

# The flag below controls the internal computation precision for mkldnn matmul. Default ieee is float32.
torch.backends.mkldnn.matmul.fp32_precision = "ieee"

# The flag below controls the internal computation precision for mkldnn conv. Default ieee is float32.
torch.backends.mkldnn.conv.fp32_precision = "ieee"

# The flag below controls the internal computation precision for mkldnn rnn. Default ieee is float32.
torch.backends.mkldnn.rnn.fp32_precision = "ieee"

Note that besides matmuls and convolutions themselves, functions and nn modules that internally uses matmuls or convolutions are also affected. These include torch.nn.Linear, torch.nn._ConvNd, torch.cdist(), torch.tensordot(), torch.nn.functional.affine_grid() and torch.nn.functional.grid_sample(), torch.nn.AdaptiveLogSoftmaxWithLoss, torch.nn.GRU and torch.nn.LSTM.

To get an idea of the precision and speed, see the example code and benchmark data (on SPR) below:

torch.manual_seed(0)
a_full = torch.randn(10240, 10240, dtype=torch.double)
b_full = torch.randn(10240, 10240, dtype=torch.double)
ab_full = a_full @ b_full
mean = ab_full.abs().mean()  # 80.7451

a = a_full.float()
b = b_full.float()

# Do matmul at BF16 mode.
torch.backends.mkldnn.matmul.fp32_precision = 'bf16'
ab_bf16 = a @ b  # expected speedup with BF16 dot-product acceleration
error = (ab_bf16 - ab_full).abs().max()  # 1.3704
relative_error = error / mean  # 0.0170
print(error, relative_error)

# Do matmul FP32 mode.
torch.backends.mkldnn.matmul.fp32_precision = 'ieee'
ab_fp32 = a @ b
error = (ab_fp32 - ab_full).abs().max()  # 0.0003
relative_error = error / mean  # 0.00000317
print(error, relative_error)

From the above example, we can see that with BF16, the speed is ~7x faster on SPR, and that relative error compared to double precision is approximately 2 orders of magnitude larger. If full FP32 precision is needed, users can disable BF16 by:

torch.backends.mkldnn.matmul.fp32_precision = 'ieee'
torch.backends.mkldnn.conv.fp32_precision = 'ieee'
torch.backends.mkldnn.rnn.fp32_precision = 'ieee'

To toggle the BF16 flags off in C++, you can do

at::globalContext().setFloat32Precision("ieee", "mkldnn", "matmul");
at::globalContext().setFloat32Precision("ieee", "mkldnn", "conv");
at::globalContext().setFloat32Precision("ieee", "mkldnn", "rnn");

We can override a generic setting for a specific operator or backend if the fp32_precision is set to ieee.

torch.backends.fp32_precision = "bf16"
torch.backends.mkldnn.fp32_precision = "ieee"
torch.backends.mkldnn.matmul.fp32_precision = "ieee"

For such case, both torch.backends.mkldnn.fp32_precision and torch.backends.mkldnn.matmul.fp32_precision is overridden to bf16.