CUDA C++ Best Practices Guide
Release 12.5

NVIDIA

May 09, 2024

Contents

1 What Is This Document?

2 Who Should Read This Guide?

3 Assess, Parallelize, Optimize, Deploy
.
Assess .
.
.
.
Parallelize .
.
.
Optimize .
.
.
.
Deploy .

3.1
3.2
3.3
3.4

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

4 Recommendations and Best Practices

5 Assessing Your Application

6 Heterogeneous Computing

6.1
6.2

.
Differences between Host and Device .
What Runs on a CUDA-Enabled Device? .

.
.

.
.

.

.

.

.

.

.

.

7.1

Profile .

7.1.1
7.1.2
7.1.3

7 Application Profiling
.
.
.
.
.
.
.
Creating the Profile .
Identifying Hotspots .
.
Understanding Scaling .

.
.
.
.
Strong Scaling and Amdahl’s Law .

.
.
.
.
7.1.3.1
.
7.1.3.2 Weak Scaling and Gustafson’s Law .
Applying Strong and Weak Scaling .
7.1.3.3

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

8 Parallelizing Your Application

9 Getting Started

9.1
9.2
9.3

.
Parallel Libraries .
.
Parallelizing Compilers .
.
Coding to Expose Parallelism .

.
.

.
.

.
.

.

.

.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

10 Getting the Right Answer
.

Verification .

10.1

10.2
10.3

.

.

.

.

.

.

.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
10.1.1 Reference Comparison .
.
.
10.1.2 Unit Testing .
Debugging .
.
.
.
Numerical Accuracy and Precision .
.

.
.
.
.
.
10.3.1 Single vs. Double Precision .
.
10.3.2 Floating Point Math Is not Associative .
.
10.3.3 IEEE 754 Compliance .
.
.
10.3.4 x86 80-bit Computations .

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.

.

.

.

.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.

.
.

.
.
.
.
.
.
.

.
.
.

.
.
.
.
.
.
.
.
.

3

5

7
8
8
8
9

.
.
.
.

11

13

15
. 15
. 16

19
. 19
. 19
. 20
. 20
. 20
. 21
. 21

23

25
. 25
. 25
. 26

27
. 27
. 27
. 27
. 28
. 28
. 28
. 29
. 29
. 29

i

11 Optimizing CUDA Applications

12 Performance Metrics
.

12.1

12.2

.

.

.

.

.

.

.

.
.

.
.

.
.

Timing .

.
.
12.1.1 Using CPU Timers .
.
12.1.2 Using CUDA GPU Timers .
.

.
.
.
.
.
.
.
.
.
.
12.2.1 Theoretical Bandwidth Calculation .
12.2.2 Effective Bandwidth Calculation .
.
.
12.2.3 Throughput Reported by Visual Profiler .

Bandwidth .

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.

.

.

.

.

.

.

.

.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

13 Memory Optimizations

13.1

13.2

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

13.2.2.1
13.2.2.2

13.2.2 L2 Cache .

Data Transfer Between Host and Device .
.

Effects of Misaligned Accesses .
.
Strided Accesses .
.
.
.
.

.
.
.
13.2.1 Coalesced Access to Global Memory .
.

.
.
.
13.1.1 Pinned Memory .
13.1.2 Asynchronous and Overlapping Transfers with Computation .
.
13.1.3 Zero Copy .
.
.
.
13.1.4 Unified Virtual Addressing .
.
.
Device Memory Spaces .
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
13.2.1.1 A Simple Access Pattern .
.
.
.
13.2.1.2 A Sequential but Misaligned Access Pattern .
.
.
.
13.2.1.3
.
.
.
.
13.2.1.4
.
.
.
.
.
.
.
.
.
.
.
.
L2 Cache Access Window .
.
.
.
.
Tuning the Access Window Hit-Ratio .
.
.
.
.
.
.
.
.
.
Shared Memory and Memory Banks .
.
.
13.2.3.1
.
Shared Memory in Matrix Multiplication (C=AB)
.
13.2.3.2
13.2.3.3
.
Shared Memory in Matrix Multiplication (C=AAT) .
13.2.3.4 Asynchronous Copy from Global Memory to Shared Memory .
.
.
.
.
.
.
.
.
.
.

.
.
13.2.5.1 Additional Texture Capabilities .
.
.
.
.
.

.
.
.
Register Pressure .
.
.
.

13.2.6 Constant Memory .
.
13.2.7 Registers .

13.2.4 Local Memory .
.
13.2.5 Texture Memory .

Allocation .
.
NUMA Best Practices .

13.2.3 Shared Memory .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

13.2.7.1

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.

.
.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

13.3
13.4

.

.

.

.

.

.

.

.

14.1 Occupancy .

.
14.1.1 Calculating Occupancy .

14 Execution Configuration Optimizations
.
.
.
.
.
.
.

.
.
14.2 Hiding Register Dependencies .
.
14.3
.
14.4
.
14.5
.
14.6 Multiple contexts .

Thread and Block Heuristics .
Effects of Shared Memory .
.
Concurrent Kernel Execution .
.

.
.

.
.

.

.

.

.

.

.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

15 Instruction Optimization

15.1

.
.
Arithmetic Instructions .
.
15.1.1 Division Modulo Operations .
15.1.2 Loop Counters Signed vs. Unsigned .
.
15.1.3 Reciprocal Square Root
.
.
15.1.4 Other Arithmetic Instructions .

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.

.

.

.

.

.

.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

ii

31

33
. 33
. 33
. 34
. 34
. 35
. 35
. 36

37
. 37
. 38
. 38
. 40
. 41
. 42
. 43
. 44
. 44
. 45
. 46
. 47
. 47
. 48
. 50
. 51
. 51
. 54
. 56
. 58
. 59
. 59
. 60
. 60
. 60
. 60
. 61

63
. 63
. 64
. 64
. 65
. 66
. 67
. 67

69
. 69
. 69
. 69
. 70
. 70

.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.

15.1.5 Exponentiation With Small Fractional Arguments .
.
15.1.6 Math Libraries .
.
.
.
15.1.7 Precision-related Compiler Flags .
.
.

15.2 Memory Instructions .

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

16 Control Flow

16.1
16.2

Branching and Divergence .
.
Branch Predication .

.

.

.

.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

17 Deploying CUDA Applications

18 Understanding the Programming Environment
.
.
.
.

.
CUDA Compute Capability .
18.1
.
Additional Hardware Data .
18.2
18.3 Which Compute Capability Target .
.
18.4

CUDA Runtime .

.
.
.
.

.
.
.
.

.
.
.
.

.
.

.
.

.
.

.
.

.

.

.

.

.

.

.

.

.

.

.

.
.

.
.
.
.

.
.

.
.
.
.

.
.

.
.
.
.

.
.

.
.
.
.

.
.

.
.
.
.

.
.

.
.
.
.

.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.
.
.

.
.

.
.
.
.

.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.
.

.
.
.
.

19.4

19.1
19.2
19.3

19 CUDA Compatibility Developer’s Guide
.
.
.

CUDA Toolkit Versioning .
.
Source Compatibility .
.
Binary Compatibility .

.
.
.
19.3.1 CUDA Binary (cubin) Compatibility .

.
.
.
.
CUDA Compatibility Across Minor Releases .

.
.
.
.
.
19.4.1 Existing CUDA Applications within Minor Versions of CUDA .
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Recommendations for building a minor-version compatible library .
.
Recommendations for taking advantage of minor version compatibility in your
.
.
.
application .

19.4.1.1 Handling New CUDA Features and Driver APIs .
.
19.4.1.2 Using PTX .
.
.
.
.
19.4.1.3 Dynamic Code Generation .
19.4.1.4
19.4.1.5

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

20 Preparing for Deployment

.

.

.

20.1
20.2
20.3
20.4

.
.
Testing for CUDA Availability .
.
Error Handling .
.
.
.
.
Building for Maximum Compatibility .
Distributing the CUDA Runtime and Libraries .
.
.

.
.
.
.
.
20.4.1 CUDA Toolkit Library Redistribution .
20.4.1.1 Which Files to Redistribute .
.
.
20.4.1.2 Where to Install Redistributed CUDA Libraries .

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.

.
.

.
.

.
.

.
.

.
.

.
.

.

.

.

.

.

.

.

.

.

.

21.1

Nvidia-SMI

21 Deployment Infrastructure Tools
.
.
.
.

.
21.1.1 Queryable state .
21.1.2 Modifiable state .
.

.
.
.
.
.
.
.
NVML .
.
.
Cluster Management Tools .
.
Compiler JIT Cache Management Tools .
.
CUDA_VISIBLE_DEVICES .

21.2
21.3
21.4
21.5

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

22 Recommendations and Best Practices

22.1 Overall Performance Optimization Strategies .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

23 nvcc Compiler Switches
.
nvcc .

23.1

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

. 70
. 71
. 73
. 73

75
. 75
. 75

77

79
. 79
. 79
. 80
. 80

83
. 84
. 85
. 85
. 87
. 88
. 88
. 89
. 90
. 91
. 93

. 94

95
. 95
. 96
. 96
. 97
. 98
. 99
. 100

103
. 103
. 103
. 104
. 104
. 104
. 105
. 105

107
. 107

109
. 109

iii

24 Notices

Notice .

24.1
.
24.2 OpenCL .
24.3

.
.
Trademarks .

.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

.
.
.

111
. 111
. 112
. 112

iv

CUDA C++ Best Practices Guide, Release 12.5

CUDA C++ Best Practices Guide

The programming guide to using the CUDA Toolkit to obtain the best performance from NVIDIA GPUs.

Contents

1

CUDA C++ Best Practices Guide, Release 12.5

2

Contents

Chapter 1. What Is This Document?

This Best Practices Guide is a manual to help developers obtain the best performance from NVIDIA®
CUDA® GPUs. It presents established parallelization and optimization techniques and explains coding
metaphors and idioms that can greatly simplify programming for CUDA-capable GPU architectures.

While the contents can be used as a reference manual, you should be aware that some topics are revis-
ited in different contexts as various programming and configuration topics are explored. As a result,
it is recommended that first-time readers proceed through the guide sequentially. This approach will
greatly improve your understanding of effective programming practices and enable you to better use
the guide for reference later.

3

CUDA C++ Best Practices Guide, Release 12.5

4

Chapter 1. What Is This Document?

Chapter 2. Who Should Read This

Guide?

The discussions in this guide all use the C++ programming language, so you should be comfortable
reading C++ code.

This guide refers to and relies on several other documents that you should have at your disposal for
reference, all of which are available at no cost from the CUDA website https://docs.nvidia.com/cuda/.
The following documents are especially important resources:

▶ CUDA Installation Guide

▶ CUDA C++ Programming Guide

▶ CUDA Toolkit Reference Manual

In particular, the optimization section of this guide assumes that you have already successfully down-
loaded and installed the CUDA Toolkit (if not, please refer to the relevant CUDA Installation Guide for
your platform) and that you have a basic familiarity with the CUDA C++ programming language and
environment (if not, please refer to the CUDA C++ Programming Guide).

5

CUDA C++ Best Practices Guide, Release 12.5

6

Chapter 2. Who Should Read This Guide?

Chapter 3. Assess, Parallelize, Optimize,

Deploy

This guide introduces the Assess, Parallelize, Optimize, Deploy(APOD) design cycle for applications with
the goal of helping application developers to rapidly identify the portions of their code that would most
readily benefit from GPU acceleration, rapidly realize that benefit, and begin leveraging the resulting
speedups in production as early as possible.

APOD is a cyclical process: initial speedups can be achieved, tested, and deployed with only minimal
initial investment of time, at which point the cycle can begin again by identifying further optimiza-
tion opportunities, seeing additional speedups, and then deploying the even faster versions of the
application into production.

7

CUDA C++ Best Practices Guide, Release 12.5

3.1. Assess

For an existing project, the first step is to assess the application to locate the parts of the code that
are responsible for the bulk of the execution time. Armed with this knowledge, the developer can
evaluate these bottlenecks for parallelization and start to investigate GPU acceleration.

By understanding the end-user’s requirements and constraints and by applying Amdahl’s and
Gustafson’s laws, the developer can determine the upper bound of performance improvement from
acceleration of the identified portions of the application.

3.2. Parallelize

Having identified the hotspots and having done the basic exercises to set goals and expectations, the
developer needs to parallelize the code. Depending on the original code, this can be as simple as calling
into an existing GPU-optimized library such as cuBLAS, cuFFT, or Thrust, or it could be as simple as
adding a few preprocessor directives as hints to a parallelizing compiler.

On the other hand, some applications’ designs will require some amount of refactoring to expose their
inherent parallelism. As even CPU architectures will require exposing parallelism in order to improve or
simply maintain the performance of sequential applications, the CUDA family of parallel programming
languages (CUDA C++, CUDA Fortran, etc.) aims to make the expression of this parallelism as simple
as possible, while simultaneously enabling operation on CUDA-capable GPUs designed for maximum
parallel throughput.

3.3. Optimize

After each round of application parallelization is complete, the developer can move to optimizing the
implementation to improve performance. Since there are many possible optimizations that can be
considered, having a good understanding of the needs of the application can help to make the pro-
cess as smooth as possible. However, as with APOD as a whole, program optimization is an iterative
process (identify an opportunity for optimization, apply and test the optimization, verify the speedup
achieved, and repeat), meaning that it is not necessary for a programmer to spend large amounts of
In-
time memorizing the bulk of all possible optimization strategies prior to seeing good speedups.
stead, strategies can be applied incrementally as they are learned.

Optimizations can be applied at various levels, from overlapping data transfers with computation all the
way down to fine-tuning floating-point operation sequences. The available profiling tools are invaluable
for guiding this process, as they can help suggest a next-best course of action for the developer’s
optimization efforts and provide references into the relevant portions of the optimization section of
this guide.

8

Chapter 3. Assess, Parallelize, Optimize, Deploy

CUDA C++ Best Practices Guide, Release 12.5

3.4. Deploy

Having completed the GPU acceleration of one or more components of the application it is possible
to compare the outcome with the original expectation. Recall that the initial assess step allowed the
developer to determine an upper bound for the potential speedup attainable by accelerating given
hotspots.

Before tackling other hotspots to improve the total speedup, the developer should consider taking
the partially parallelized implementation and carry it through to production. This is important for a
number of reasons; for example, it allows the user to profit from their investment as early as possible
(the speedup may be partial but is still valuable), and it minimizes risk for the developer and the user
by providing an evolutionary rather than revolutionary set of changes to the application.

3.4. Deploy

9

CUDA C++ Best Practices Guide, Release 12.5

10

Chapter 3. Assess, Parallelize, Optimize, Deploy

Chapter 4. Recommendations and Best

Practices

Throughout this guide, specific recommendations are made regarding the design and implementation
of CUDA C++ code. These recommendations are categorized by priority, which is a blend of the effect
of the recommendation and its scope. Actions that present substantial improvements for most CUDA
applications have the highest priority, while small optimizations that affect only very specific situations
are given a lower priority.

Before implementing lower priority recommendations, it is good practice to make sure all higher pri-
ority recommendations that are relevant have already been applied. This approach will tend to provide
the best results for the time invested and will avoid the trap of premature optimization.

The criteria of benefit and scope for establishing priority will vary depending on the nature of the
program. In this guide, they represent a typical case. Your code might reflect different priority factors.
Regardless of this possibility, it is good practice to verify that no higher-priority recommendations
have been overlooked before undertaking lower-priority items.

Note: Code samples throughout the guide omit error checking for conciseness. Production code
should, however, systematically check the error code returned by each API call and check for failures
in kernel launches by calling cudaGetLastError().

11

CUDA C++ Best Practices Guide, Release 12.5

12

Chapter 4. Recommendations and Best Practices

Chapter 5. Assessing Your Application

From supercomputers to mobile phones, modern processors increasingly rely on parallelism to provide
performance. The core computational unit, which includes control, arithmetic, registers and typically
some cache, is replicated some number of times and connected to memory via a network. As a result,
all modern processors require parallel code in order to achieve good utilization of their computational
power.

While processors are evolving to expose more fine-grained parallelism to the programmer, many ex-
isting applications have evolved either as serial codes or as coarse-grained parallel codes (for example,
where the data is decomposed into regions processed in parallel, with sub-regions shared using MPI).
In order to profit from any modern processor architecture, GPUs included, the first steps are to assess
the application to identify the hotspots, determine whether they can be parallelized, and understand
the relevant workloads both now and in the future.

13

CUDA C++ Best Practices Guide, Release 12.5

14

Chapter 5. Assessing Your Application

Chapter 6. Heterogeneous Computing

CUDA programming involves running code on two different platforms concurrently: a host system with
one or more CPUs and one or more CUDA-enabled NVIDIA GPU devices.

While NVIDIA GPUs are frequently associated with graphics, they are also powerful arithmetic engines
capable of running thousands of lightweight threads in parallel. This capability makes them well suited
to computations that can leverage parallel execution.

However, the device is based on a distinctly different design from the host system, and it’s important
to understand those differences and how they determine the performance of CUDA applications in
order to use CUDA effectively.

6.1. Differences between Host and Device

The primary differences are in threading model and in separate physical memories:

Threading resources

Execution pipelines on host systems can support a limited number of concurrent threads. For ex-
ample, servers that have two 32 core processors can run only 64 threads concurrently (or small
multiple of that if the CPUs support simultaneous multithreading). By comparison, the small-
est executable unit of parallelism on a CUDA device comprises 32 threads (termed a warp of
threads). Modern NVIDIA GPUs can support up to 2048 active threads concurrently per multi-
processor (see Features and Specifications of the CUDA C++ Programming Guide) On GPUs with
80 multiprocessors, this leads to more than 160,000 concurrently active threads.

Threads

Threads on a CPU are generally heavyweight entities. The operating system must swap threads
on and off CPU execution channels to provide multithreading capability. Context switches (when
two threads are swapped) are therefore slow and expensive. By comparison, threads on GPUs
In a typical system, thousands of threads are queued up for work
are extremely lightweight.
(in warps of 32 threads each).
If the GPU must wait on one warp of threads, it simply begins
executing work on another. Because separate registers are allocated to all active threads, no
swapping of registers or other state need occur when switching among GPU threads. Resources
stay allocated to each thread until it completes its execution. In short, CPU cores are designed
to minimize latency for a small number of threads at a time each, whereas GPUs are designed to
handle a large number of concurrent, lightweight threads in order to maximize throughput.

RAM

The host system and the device each have their own distinct attached physical memories1. As
the host and device memories are separated, items in the host memory must occasionally be

1 On Systems on a Chip with integrated GPUs, such as NVIDIA® Tegra®, host and device memory are physically the same, but

there is still a logical distinction between host and device memory. See the Application Note on CUDA for Tegra for details.

15

CUDA C++ Best Practices Guide, Release 12.5

communicated between device memory and host memory as described in What Runs on a CUDA-
Enabled Device?.

These are the primary hardware differences between CPU hosts and GPU devices with respect to par-
allel programming. Other differences are discussed as they arise elsewhere in this document. Applica-
tions composed with these differences in mind can treat the host and device together as a cohesive
heterogeneous system wherein each processing unit is leveraged to do the kind of work it does best:
sequential work on the host and parallel work on the device.

6.2. What Runs on a CUDA-Enabled Device?

The following issues should be considered when determining what parts of an application to run on
the device:

▶ The device is ideally suited for computations that can be run on numerous data elements si-
multaneously in parallel. This typically involves arithmetic on large data sets (such as matrices)
where the same operation can be performed across thousands, if not millions, of elements at
the same time. This is a requirement for good performance on CUDA: the software must use a
large number (generally thousands or tens of thousands) of concurrent threads. The support for
running numerous threads in parallel derives from CUDA’s use of a lightweight threading model
described above.

▶ To use CUDA, data values must be transferred from the host to the device. These transfers are
costly in terms of performance and should be minimized. (See Data Transfer Between Host and
Device.) This cost has several ramifications:

▶ The complexity of operations should justify the cost of moving data to and from the device.
Code that transfers data for brief use by a small number of threads will see little or no per-
formance benefit. The ideal scenario is one in which many threads perform a substantial
amount of work.

For example, transferring two matrices to the device to perform a matrix addition and then
transferring the results back to the host will not realize much performance benefit. The
issue here is the number of operations performed per data element transferred. For the
preceding procedure, assuming matrices of size NxN, there are N2 operations (additions) and
3N2 elements transferred, so the ratio of operations to elements transferred is 1:3 or O(1).
Performance benefits can be more readily achieved when this ratio is higher. For example, a
matrix multiplication of the same matrices requires N3 operations (multiply-add), so the ratio
of operations to elements transferred is O(N), in which case the larger the matrix the greater
the performance benefit. The types of operations are an additional factor, as additions have
different complexity profiles than, for example, trigonometric functions. It is important to
include the overhead of transferring data to and from the device in determining whether
operations should be performed on the host or on the device.

▶ Data should be kept on the device as long as possible. Because transfers should be mini-
mized, programs that run multiple kernels on the same data should favor leaving the data
on the device between kernel calls, rather than transferring intermediate results to the host
and then sending them back to the device for subsequent calculations. So, in the previous
example, had the two matrices to be added already been on the device as a result of some
previous calculation, or if the results of the addition would be used in some subsequent
calculation, the matrix addition should be performed locally on the device. This approach
should be used even if one of the steps in a sequence of calculations could be performed
faster on the host. Even a relatively slow kernel may be advantageous if it avoids one or
more transfers between host and device memory. Data Transfer Between Host and Device

16

Chapter 6. Heterogeneous Computing

CUDA C++ Best Practices Guide, Release 12.5

provides further details, including the measurements of bandwidth between the host and
the device versus within the device proper.

▶ For best performance, there should be some coherence in memory access by adjacent threads
running on the device. Certain memory access patterns enable the hardware to coalesce groups
of reads or writes of multiple data items into one operation. Data that cannot be laid out so as to
enable coalescing, or that doesn’t have enough locality to use the L1 or texture caches effectively,
will tend to see lesser speedups when used in computations on GPUs. A noteworthy exception
to this are completely random memory access patterns. In general, they should be avoided, be-
cause compared to peak capabilities any architecture processes these memory access patterns
at a low efficiency. However, compared to cache based architectures, like CPUs, latency hiding
architectures, like GPUs, tend to cope better with completely random memory access patterns.

6.2. What Runs on a CUDA-Enabled Device?

17

CUDA C++ Best Practices Guide, Release 12.5

18

Chapter 6. Heterogeneous Computing

Chapter 7. Application Profiling

7.1. Profile

Many codes accomplish a significant portion of the work with a relatively small amount of code. Us-
ing a profiler, the developer can identify such hotspots and start to compile a list of candidates for
parallelization.

7.1.1. Creating the Profile

There are many possible approaches to profiling the code, but in all cases the objective is the same:
to identify the function or functions in which the application is spending most of its execution time.

Note: High Priority: To maximize developer productivity, profile the application to determine hotspots
and bottlenecks.

The most important consideration with any profiling activity is to ensure that the workload is realistic
- i.e., that information gained from the test and decisions based upon that information are relevant
to real data. Using unrealistic workloads can lead to sub-optimal results and wasted effort both by
causing developers to optimize for unrealistic problem sizes and by causing developers to concentrate
on the wrong functions.

There are a number of tools that can be used to generate the profile. The following example is based
on gprof, which is an open-source profiler for Linux platforms from the GNU Binutils collection.

$ gcc -O2 -g -pg myprog.c
$ gprof .∕a.out > profile.txt
Each sample counts as 0.01 seconds.

%
time
33.34
16.67
16.67
16.67
16.67
0.00
0.00
0.00
0.00
0.00

cumulative
seconds

0.02
0.03
0.04
0.05
0.06
0.06
0.06
0.06
0.06
0.06

self
seconds
0.02
0.01
0.01
0.01
0.01
0.00
0.00
0.00
0.00
0.00

calls
7208
240
8
7

236
192
47
45
1

self

total

ms∕call ms∕call
0.00
0.12
1.25
1.43

0.00
0.04
1.25
1.43

0.00
0.00
0.00
0.00
0.00

0.00
0.00
0.00
0.00
50.00

name
genTimeStep
calcStats
calcSummaryData
write
mcount
tzset
tolower
strlen
strchr
main

(continues on next page)

19

CUDA C++ Best Practices Guide, Release 12.5

0.00
0.00
0.00
0.00

0.06
0.06
0.06
0.06

0.00
0.00
0.00
0.00

1
1
1
1

0.00
0.00
0.00
0.00

0.00
10.11
0.00
50.00

memcpy
print
profil
report

(continued from previous page)

7.1.2. Identifying Hotspots

In the example above, we can clearly see that the function genTimeStep() takes one-third of the
total running time of the application. This should be our first candidate function for parallelization.
Understanding Scaling discusses the potential benefit we might expect from such parallelization.

It is worth noting that several of the other functions in the above example also take up a significant por-
tion of the overall running time, such as calcStats() and calcSummaryData(). Parallelizing these
functions as well should increase our speedup potential. However, since APOD is a cyclical process,
we might opt to parallelize these functions in a subsequent APOD pass, thereby limiting the scope of
our work in any given pass to a smaller set of incremental changes.

7.1.3. Understanding Scaling

The amount of performance benefit an application will realize by running on CUDA depends entirely
on the extent to which it can be parallelized. Code that cannot be sufficiently parallelized should run
on the host, unless doing so would result in excessive transfers between the host and the device.

Note: High Priority: To get the maximum benefit from CUDA, focus first on finding ways to parallelize
sequential code.

By understanding how applications can scale it is possible to set expectations and plan an incremental
parallelization strategy. Strong Scaling and Amdahl’s Law describes strong scaling, which allows us to
set an upper bound for the speedup with a fixed problem size. Weak Scaling and Gustafson’s Law de-
scribes weak scaling, where the speedup is attained by growing the problem size. In many applications,
a combination of strong and weak scaling is desirable.

7.1.3.1 Strong Scaling and Amdahl’s Law

Strong scaling is a measure of how, for a fixed overall problem size, the time to solution decreases
as more processors are added to a system. An application that exhibits linear strong scaling has a
speedup equal to the number of processors used.

Strong scaling is usually equated with Amdahl’s Law, which specifies the maximum speedup that can
be expected by parallelizing portions of a serial program. Essentially, it states that the maximum
speedup S of a program is:

S =

1
(1−P )+ P
N

Here P is the fraction of the total serial execution time taken by the portion of code that can be par-
allelized and N is the number of processors over which the parallel portion of the code runs.

The larger N is(that is, the greater the number of processors), the smaller the P/N fraction. It can be
simpler to view N as a very large number, which essentially transforms the equation into S = 1/(1 − P ).

20

Chapter 7. Application Profiling

CUDA C++ Best Practices Guide, Release 12.5

Now, if 3/4 of the running time of a sequential program is parallelized, the maximum speedup over
serial code is 1 / (1 - 3/4) = 4.

In reality, most applications do not exhibit perfectly linear strong scaling, even if they do exhibit some
degree of strong scaling. For most purposes, the key point is that the larger the parallelizable portion
P is, the greater the potential speedup. Conversely, if P is a small number (meaning that the applica-
tion is not substantially parallelizable), increasing the number of processors N does little to improve
performance. Therefore, to get the largest speedup for a fixed problem size, it is worthwhile to spend
effort on increasing P, maximizing the amount of code that can be parallelized.

7.1.3.2 Weak Scaling and Gustafson’s Law

Weak scaling is a measure of how the time to solution changes as more processors are added to a
system with a fixed problem size per processor; i.e., where the overall problem size increases as the
number of processors is increased.

Weak scaling is often equated with Gustafson’s Law, which states that in practice, the problem size
scales with the number of processors. Because of this, the maximum speedup S of a program is:

S = N + (1 − P )(1 − N )

Here P is the fraction of the total serial execution time taken by the portion of code that can be par-
allelized and N is the number of processors over which the parallel portion of the code runs.

Another way of looking at Gustafson’s Law is that it is not the problem size that remains constant
as we scale up the system but rather the execution time. Note that Gustafson’s Law assumes that
the ratio of serial to parallel execution remains constant, reflecting additional cost in setting up and
handling the larger problem.

7.1.3.3 Applying Strong and Weak Scaling

Understanding which type of scaling is most applicable to an application is an important part of esti-
mating speedup. For some applications the problem size will remain constant and hence only strong
scaling is applicable. An example would be modeling how two molecules interact with each other,
where the molecule sizes are fixed.

For other applications, the problem size will grow to fill the available processors. Examples include
modeling fluids or structures as meshes or grids and some Monte Carlo simulations, where increasing
the problem size provides increased accuracy.

Having understood the application profile, the developer should understand how the problem
size would change if the computational performance changes and then apply either Amdahl’s or
Gustafson’s Law to determine an upper bound for the speedup.

7.1. Profile

21

CUDA C++ Best Practices Guide, Release 12.5

22

Chapter 7. Application Profiling

Chapter 8. Parallelizing Your Application

Having identified the hotspots and having done the basic exercises to set goals and expectations, the
developer needs to parallelize the code. Depending on the original code, this can be as simple as calling
into an existing GPU-optimized library such as cuBLAS, cuFFT, or Thrust, or it could be as simple as
adding a few preprocessor directives as hints to a parallelizing compiler.

On the other hand, some applications’ designs will require some amount of refactoring to expose their
inherent parallelism. As even CPU architectures require exposing this parallelism in order to improve or
simply maintain the performance of sequential applications, the CUDA family of parallel programming
languages (CUDA C++, CUDA Fortran, etc.) aims to make the expression of this parallelism as simple
as possible, while simultaneously enabling operation on CUDA-capable GPUs designed for maximum
parallel throughput.

23

CUDA C++ Best Practices Guide, Release 12.5

24

Chapter 8. Parallelizing Your Application

Chapter 9. Getting Started

There are several key strategies for parallelizing sequential code. While the details of how to apply
these strategies to a particular application is a complex and problem-specific topic, the general themes
listed here apply regardless of whether we are parallelizing code to run on for multicore CPUs or for
use on CUDA GPUs.

9.1. Parallel Libraries

The most straightforward approach to parallelizing an application is to leverage existing libraries that
take advantage of parallel architectures on our behalf. The CUDA Toolkit includes a number of such
libraries that have been fine-tuned for NVIDIA CUDA GPUs, such as cuBLAS, cuFFT, and so on.

The key here is that libraries are most useful when they match well with the needs of the application.
Applications already using other BLAS libraries can often quite easily switch to cuBLAS, for example,
whereas applications that do little to no linear algebra will have little use for cuBLAS. The same goes
for other CUDA Toolkit libraries: cuFFT has an interface similar to that of FFTW, etc.

Also of note is the Thrust library, which is a parallel C++ template library similar to the C++ Standard
Template Library. Thrust provides a rich collection of data parallel primitives such as scan, sort, and
reduce, which can be composed together to implement complex algorithms with concise, readable
source code. By describing your computation in terms of these high-level abstractions you provide
Thrust with the freedom to select the most efficient implementation automatically. As a result, Thrust
can be utilized in rapid prototyping of CUDA applications, where programmer productivity matters
most, as well as in production, where robustness and absolute performance are crucial.

9.2. Parallelizing Compilers

Another common approach to parallelization of sequential codes is to make use of parallelizing compil-
ers. Often this means the use of directives-based approaches, where the programmer uses a pragma
or other similar notation to provide hints to the compiler about where parallelism can be found with-
out needing to modify or adapt the underlying code itself. By exposing parallelism to the compiler,
directives allow the compiler to do the detailed work of mapping the computation onto the parallel
architecture.

The OpenACC standard provides a set of compiler directives to specify loops and regions of code in
standard C, C++ and Fortran that should be offloaded from a host CPU to an attached accelerator such
as a CUDA GPU. The details of managing the accelerator device are handled implicitly by an OpenACC-
enabled compiler and runtime.

25

CUDA C++ Best Practices Guide, Release 12.5

See http://www.openacc.org/ for details.

9.3. Coding to Expose Parallelism

For applications that need additional functionality or performance beyond what existing parallel li-
braries or parallelizing compilers can provide, parallel programming languages such as CUDA C++ that
integrate seamlessly with existing sequential code are essential.

Once we have located a hotspot in our application’s profile assessment and determined that custom
code is the best approach, we can use CUDA C++ to expose the parallelism in that portion of our
code as a CUDA kernel. We can then launch this kernel onto the GPU and retrieve the results without
requiring major rewrites to the rest of our application.

This approach is most straightforward when the majority of the total running time of our application
is spent in a few relatively isolated portions of the code. More difficult to parallelize are applications
with a very flat profile - i.e., applications where the time spent is spread out relatively evenly across a
wide portion of the code base. For the latter variety of application, some degree of code refactoring
to expose the inherent parallelism in the application might be necessary, but keep in mind that this
refactoring work will tend to benefit all future architectures, CPU and GPU alike, so it is well worth the
effort should it become necessary.

26

Chapter 9. Getting Started

Chapter 10. Getting the Right Answer

Obtaining the right answer is clearly the principal goal of all computation. On parallel systems, it is
possible to run into difficulties not typically found in traditional serial-oriented programming. These
include threading issues, unexpected values due to the way floating-point values are computed, and
challenges arising from differences in the way CPU and GPU processors operate. This chapter exam-
ines issues that can affect the correctness of returned data and points to appropriate solutions.

10.1. Verification

10.1.1. Reference Comparison

A key aspect of correctness verification for modifications to any existing program is to establish some
mechanism whereby previous known-good reference outputs from representative inputs can be com-
pared to new results. After each change is made, ensure that the results match using whatever criteria
apply to the particular algorithm. Some will expect bitwise identical results, which is not always pos-
sible, especially where floating-point arithmetic is concerned; see Numerical Accuracy and Precision
regarding numerical accuracy. For other algorithms, implementations may be considered correct if
they match the reference within some small epsilon.

Note that the process used for validating numerical results can easily be extended to validate perfor-
mance results as well. We want to ensure that each change we make is correct and that it improves
performance (and by how much). Checking these things frequently as an integral part of our cyclical
APOD process will help ensure that we achieve the desired results as rapidly as possible.

10.1.2. Unit Testing

A useful counterpart to the reference comparisons described above is to structure the code itself
in such a way that is readily verifiable at the unit level. For example, we can write our CUDA kernels
as a collection of many short __device__ functions rather than one large monolithic __global__
function; each device function can be tested independently before hooking them all together.

For example, many kernels have complex addressing logic for accessing memory in addition to their
actual computation.
If we validate our addressing logic separately prior to introducing the bulk of
the computation, then this will simplify any later debugging efforts. (Note that the CUDA compiler
considers any device code that does not contribute to a write to global memory as dead code subject
to elimination, so we must at least write something out to global memory as a result of our addressing
logic in order to successfully apply this strategy.)

27

CUDA C++ Best Practices Guide, Release 12.5

Going a step further, if most functions are defined as __host__ __device__ rather than just __de-
vice__ functions, then these functions can be tested on both the CPU and the GPU, thereby increas-
ing our confidence that the function is correct and that there will not be any unexpected differences
in the results. If there are differences, then those differences will be seen early and can be understood
in the context of a simple function.

As a useful side effect, this strategy will allow us a means to reduce code duplication should we wish
to include both CPU and GPU execution paths in our application: if the bulk of the work of our CUDA
kernels is done in __host__ __device__ functions, we can easily call those functions from both the
host code and the device code without duplication.

10.2. Debugging

CUDA-GDB is a port of the GNU Debugger that runs on Linux and Mac; see: https://developer.nvidia.
com/cuda-gdb.

The NVIDIA Nsight Visual Studio Edition for Microsoft Windows 7, Windows HPC Server 2008, Windows
8.1, and Windows 10 is available as a free plugin for Microsoft Visual Studio; see: https://developer.
nvidia.com/nsight-visual-studio-edition.

Several third-party debuggers support CUDA debugging as well; see: https://developer.nvidia.com/
debugging-solutions for more details.

10.3. Numerical Accuracy and Precision

Incorrect or unexpected results arise principally from issues of floating-point accuracy due to the way
floating-point values are computed and stored. The following sections explain the principal items of
interest. Other peculiarities of floating-point arithmetic are presented in Features and Technical Spec-
ifications of the CUDA C++ Programming Guide as well as in a whitepaper and accompanying webi-
nar on floating-point precision and performance available from https://developer.nvidia.com/content/
precision-performance-floating-point-and-ieee-754-compliance-nvidia-gpus.

10.3.1. Single vs. Double Precision

Devices of compute capability 1.3 and higher provide native support for double-precision floating-
point values (that is, values 64 bits wide). Results obtained using double-precision arithmetic will fre-
quently differ from the same operation performed via single-precision arithmetic due to the greater
precision of the former and due to rounding issues. Therefore, it is important to be sure to compare
values of like precision and to express the results within a certain tolerance rather than expecting
them to be exact.

28

Chapter 10. Getting the Right Answer

CUDA C++ Best Practices Guide, Release 12.5

10.3.2. Floating Point Math Is not Associative

Each floating-point arithmetic operation involves a certain amount of rounding. Consequently, the
order in which arithmetic operations are performed is important.
If A, B, and C are floating-point
values, (A+B)+C is not guaranteed to equal A+(B+C) as it is in symbolic math. When you parallelize
computations, you potentially change the order of operations and therefore the parallel results might
not match sequential results. This limitation is not specific to CUDA, but an inherent part of parallel
computation on floating-point values.

10.3.3. IEEE 754 Compliance

All CUDA compute devices follow the IEEE 754 standard for binary floating-point representation, with
some small exceptions. These exceptions, which are detailed in Features and Technical Specifications
of the CUDA C++ Programming Guide, can lead to results that differ from IEEE 754 values computed
on the host system.

One of the key differences is the fused multiply-add (FMA) instruction, which combines multiply-add
operations into a single instruction execution. Its result will often differ slightly from results obtained
by doing the two operations separately.

10.3.4. x86 80-bit Computations

x86 processors can use an 80-bit double extended precision math when performing floating-point cal-
culations. The results of these calculations can frequently differ from pure 64-bit operations per-
formed on the CUDA device. To get a closer match between values, set the x86 host processor to use
regular double or single precision (64 bits and 32 bits, respectively). This is done with the FLDCW x86
assembly instruction or the equivalent operating system API.

10.3. Numerical Accuracy and Precision

29

CUDA C++ Best Practices Guide, Release 12.5

30

Chapter 10. Getting the Right Answer

Chapter 11. Optimizing CUDA

Applications

After each round of application parallelization is complete, the developer can move to optimizing the
implementation to improve performance. Since there are many possible optimizations that can be
considered, having a good understanding of the needs of the application can help to make the pro-
cess as smooth as possible. However, as with APOD as a whole, program optimization is an iterative
process (identify an opportunity for optimization, apply and test the optimization, verify the speedup
achieved, and repeat), meaning that it is not necessary for a programmer to spend large amounts of
In-
time memorizing the bulk of all possible optimization strategies prior to seeing good speedups.
stead, strategies can be applied incrementally as they are learned.

Optimizations can be applied at various levels, from overlapping data transfers with computation all the
way down to fine-tuning floating-point operation sequences. The available profiling tools are invaluable
for guiding this process, as they can help suggest a next-best course of action for the developer’s
optimization efforts and provide references into the relevant portions of the optimization section of
this guide.

31

CUDA C++ Best Practices Guide, Release 12.5

32

Chapter 11. Optimizing CUDA Applications

Chapter 12. Performance Metrics

When attempting to optimize CUDA code, it pays to know how to measure performance accurately and
to understand the role that bandwidth plays in performance measurement. This chapter discusses
how to correctly measure performance using CPU timers and CUDA events.
It then explores how
bandwidth affects performance metrics and how to mitigate some of the challenges it poses.

12.1. Timing

CUDA calls and kernel executions can be timed using either CPU or GPU timers. This section examines
the functionality, advantages, and pitfalls of both approaches.

12.1.1. Using CPU Timers

Any CPU timer can be used to measure the elapsed time of a CUDA call or kernel execution. The
details of various CPU timing approaches are outside the scope of this document, but developers
should always be aware of the resolution their timing calls provide.

When using CPU timers, it is critical to remember that many CUDA API functions are asynchronous;
that is, they return control back to the calling CPU thread prior to completing their work. All ker-
nel launches are asynchronous, as are memory-copy functions with the Async suffix on their names.
Therefore, to accurately measure the elapsed time for a particular call or sequence of CUDA calls, it
is necessary to synchronize the CPU thread with the GPU by calling cudaDeviceSynchronize() im-
mediately before starting and stopping the CPU timer. cudaDeviceSynchronize()blocks the calling
CPU thread until all CUDA calls previously issued by the thread are completed.

Although it is also possible to synchronize the CPU thread with a particular stream or event on the
GPU, these synchronization functions are not suitable for timing code in streams other than the default
stream. cudaStreamSynchronize() blocks the CPU thread until all CUDA calls previously issued into
the given stream have completed. cudaEventSynchronize() blocks until a given event in a particular
stream has been recorded by the GPU. Because the driver may interleave execution of CUDA calls from
other non-default streams, calls in other streams may be included in the timing.

Because the default stream, stream 0, exhibits serializing behavior for work on the device (an operation
in the default stream can begin only after all preceding calls in any stream have completed; and no
subsequent operation in any stream can begin until it finishes), these functions can be used reliably
for timing in the default stream.

Be aware that CPU-to-GPU synchronization points such as those mentioned in this section imply a
stall in the GPU’s processing pipeline and should thus be used sparingly to minimize their performance

33

CUDA C++ Best Practices Guide, Release 12.5

impact.

12.1.2. Using CUDA GPU Timers

The CUDA event API provides calls that create and destroy events, record events (including a times-
tamp), and convert timestamp differences into a floating-point value in milliseconds. How to time
code using CUDA events illustrates their use.

How to time code using CUDA events

cudaEvent_t start, stop;
float time;

cudaEventCreate(&start);
cudaEventCreate(&stop);

cudaEventRecord( start, 0 );
kernel<<<grid,threads>>> ( d_odata, d_idata, size_x, size_y,

cudaEventRecord( stop, 0 );
cudaEventSynchronize( stop );

NUM_REPS);

cudaEventElapsedTime( &time, start, stop );
cudaEventDestroy( start );
cudaEventDestroy( stop );

Here cudaEventRecord() is used to place the start and stop events into the default stream, stream
0. The device will record a timestamp for the event when it reaches that event in the stream. The
cudaEventElapsedTime() function returns the time elapsed between the recording of the start
and stop events. This value is expressed in milliseconds and has a resolution of approximately half a
microsecond. Like the other calls in this listing, their specific operation, parameters, and return values
are described in the CUDA Toolkit Reference Manual. Note that the timings are measured on the GPU
clock, so the timing resolution is operating-system-independent.

12.2. Bandwidth

Bandwidth - the rate at which data can be transferred - is one of the most important gating factors for
performance. Almost all changes to code should be made in the context of how they affect bandwidth.
As described in Memory Optimizations of this guide, bandwidth can be dramatically affected by the
choice of memory in which data is stored, how the data is laid out and the order in which it is accessed,
as well as other factors.

To measure performance accurately, it is useful to calculate theoretical and effective bandwidth. When
the latter is much lower than the former, design or implementation details are likely to reduce band-
width, and it should be the primary goal of subsequent optimization efforts to increase it.

Note: High Priority: Use the effective bandwidth of your computation as a metric when measuring
performance and optimization benefits.

34

Chapter 12. Performance Metrics

CUDA C++ Best Practices Guide, Release 12.5

12.2.1. Theoretical Bandwidth Calculation

Theoretical bandwidth can be calculated using hardware specifications available in the product liter-
ature. For example, the NVIDIA Tesla V100 uses HBM2 (double data rate) RAM with a memory clock
rate of 877 MHz and a 4096-bit-wide memory interface.

Using these data items, the peak theoretical memory bandwidth of the NVIDIA Tesla V100 is 898 GB/s:
(

)

0.877 × 109 × (4096/8) × 2

÷ 109 = 898GB/s

In this calculation, the memory clock rate is converted in to Hz, multiplied by the interface width (di-
vided by 8, to convert bits to bytes) and multiplied by 2 due to the double data rate. Finally, this product
is divided by 109 to convert the result to GB/s.

Note: Some calculations use 10243 instead of 109 for the final calculation. In such a case, the band-
width would be 836.4 GiB/s. It is important to use the same divisor when calculating theoretical and
effective bandwidth so that the comparison is valid.

Note: On GPUs with GDDR memory with ECC enabled the available DRAM is reduced by 6.25% to
allow for the storage of ECC bits. Fetching ECC bits for each memory transaction also reduced the
effective bandwidth by approximately 20% compared to the same GPU with ECC disabled, though
the exact impact of ECC on bandwidth can be higher and depends on the memory access pattern.
HBM2 memories, on the other hand, provide dedicated ECC resources, allowing overhead-free ECC
protection.2

12.2.2. Effective Bandwidth Calculation

Effective bandwidth is calculated by timing specific program activities and by knowing how data is
accessed by the program. To do so, use this equation:

Effective bandwidth =

(Br + Bw) ÷ 109

÷ time

(

)

Here, the effective bandwidth is in units of GB/s, Br is the number of bytes read per kernel, Bw is the
number of bytes written per kernel, and time is given in seconds.

For example, to compute the effective bandwidth of a 2048 x 2048 matrix copy, the following formula
could be used:

Effective bandwidth =

((

)
20482 × 4 × 2

)

÷ 109

÷ time

The number of elements is multiplied by the size of each element (4 bytes for a float), multiplied by 2
(because of the read and write), divided by 109 (or 1,0243) to obtain GB of memory transferred. This
number is divided by the time in seconds to obtain GB/s.

2 As an exception, scattered writes to HBM2 see some overhead from ECC but much less than the overhead with similar

access patterns on ECC-protected GDDR5 memory.

12.2. Bandwidth

35

CUDA C++ Best Practices Guide, Release 12.5

12.2.3. Throughput Reported by Visual Profiler

For devices with compute capability of 2.0 or greater, the Visual Profiler can be used to collect several
different memory throughput measures. The following throughput metrics can be displayed in the
Details or Detail Graphs view:

▶ Requested Global Load Throughput

▶ Requested Global Store Throughput
▶ Global Load Throughput

▶ Global Store Throughput

▶ DRAM Read Throughput

▶ DRAM Write Throughput

The Requested Global Load Throughput and Requested Global Store Throughput values indicate the
global memory throughput requested by the kernel and therefore correspond to the effective band-
width obtained by the calculation shown under Effective Bandwidth Calculation.

Because the minimum memory transaction size is larger than most word sizes, the actual memory
throughput required for a kernel can include the transfer of data not used by the kernel. For global
memory accesses, this actual throughput is reported by the Global Load Throughput and Global Store
Throughput values.

It’s important to note that both numbers are useful. The actual memory throughput shows how close
the code is to the hardware limit, and a comparison of the effective or requested bandwidth to the
actual bandwidth presents a good estimate of how much bandwidth is wasted by suboptimal coalesc-
ing of memory accesses (see Coalesced Access to Global Memory). For global memory accesses, this
comparison of requested memory bandwidth to actual memory bandwidth is reported by the Global
Memory Load Efficiency and Global Memory Store Efficiency metrics.

36

Chapter 12. Performance Metrics

Chapter 13. Memory Optimizations

Memory optimizations are the most important area for performance. The goal is to maximize the use
of the hardware by maximizing bandwidth. Bandwidth is best served by using as much fast memory
and as little slow-access memory as possible. This chapter discusses the various kinds of memory on
the host and device and how best to set up data items to use the memory effectively.

13.1. Data Transfer Between Host and Device

The peak theoretical bandwidth between the device memory and the GPU is much higher (898 GB/s
on the NVIDIA Tesla V100, for example) than the peak theoretical bandwidth between host memory
and device memory (16 GB/s on the PCIe x16 Gen3). Hence, for best overall application performance,
it is important to minimize data transfer between the host and the device, even if that means running
kernels on the GPU that do not demonstrate any speedup compared with running them on the host
CPU.

Note: High Priority: Minimize data transfer between the host and the device, even if it means running
some kernels on the device that do not show performance gains when compared with running them
on the host CPU.

Intermediate data structures should be created in device memory, operated on by the device, and
destroyed without ever being mapped by the host or copied to host memory.

Also, because of the overhead associated with each transfer, batching many small transfers into one
larger transfer performs significantly better than making each transfer separately, even if doing so
requires packing non-contiguous regions of memory into a contiguous buffer and then unpacking
after the transfer.

Finally, higher bandwidth between the host and the device is achieved when using page-locked (or
pinned) memory, as discussed in the CUDA C++ Programming Guide and the Pinned Memory section
of this document.

37

CUDA C++ Best Practices Guide, Release 12.5

13.1.1. Pinned Memory

Page-locked or pinned memory transfers attain the highest bandwidth between the host and the
device. On PCIe x16 Gen3 cards, for example, pinned memory can attain roughly 12 GB/s transfer
rates.

Pinned memory is allocated using the cudaHostAlloc() functions in the Runtime API. The band-
widthTest CUDA Sample shows how to use these functions as well as how to measure memory trans-
fer performance.

For regions of system memory that have already been pre-allocated, cudaHostRegister() can be
used to pin the memory on-the-fly without the need to allocate a separate buffer and copy the data
into it.

Pinned memory should not be overused. Excessive use can reduce overall system performance be-
cause pinned memory is a scarce resource, but how much is too much is difficult to know in advance.
Furthermore, the pinning of system memory is a heavyweight operation compared to most normal
system memory allocations, so as with all optimizations, test the application and the systems it runs
on for optimal performance parameters.

13.1.2. Asynchronous and Overlapping Transfers with

Computation

Data transfers between the host and the device using cudaMemcpy() are blocking transfers; that
is, control is returned to the host thread only after the data transfer is complete. The cudaMem-
cpyAsync() function is a non-blocking variant of cudaMemcpy() in which control is returned immedi-
ately to the host thread. In contrast with cudaMemcpy(), the asynchronous transfer version requires
pinned host memory (see Pinned Memory), and it contains an additional argument, a stream ID. A
stream is simply a sequence of operations that are performed in order on the device. Operations in
different streams can be interleaved and in some cases overlapped - a property that can be used to
hide data transfers between the host and the device.

Asynchronous transfers enable overlap of data transfers with computation in two different ways. On
all CUDA-enabled devices, it is possible to overlap host computation with asynchronous data transfers
and with device computations. For example, Overlapping computation and data transfers demon-
strates how host computation in the routine cpuFunction() is performed while data is transferred
to the device and a kernel using the device is executed.

Overlapping computation and data transfers

cudaMemcpyAsync(a_d, a_h, size, cudaMemcpyHostToDevice, 0);
kernel<<<grid, block>>>(a_d);
cpuFunction();

The last argument to the cudaMemcpyAsync() function is the stream ID, which in this case uses the
default stream, stream 0. The kernel also uses the default stream, and it will not begin execution until
the memory copy completes; therefore, no explicit synchronization is needed. Because the memory
copy and the kernel both return control to the host immediately, the host function cpuFunction()
overlaps their execution.

38

Chapter 13. Memory Optimizations

CUDA C++ Best Practices Guide, Release 12.5

In Overlapping computation and data transfers, the memory copy and kernel execution occur sequen-
tially. On devices that are capable of concurrent copy and compute, it is possible to overlap kernel ex-
ecution on the device with data transfers between the host and the device. Whether a device has this
capability is indicated by the asyncEngineCount field of the cudaDeviceProp structure (or listed in
the output of the deviceQuery CUDA Sample). On devices that have this capability, the overlap once
again requires pinned host memory, and, in addition, the data transfer and kernel must use different,
non-default streams (streams with non-zero stream IDs). Non-default streams are required for this
overlap because memory copy, memory set functions, and kernel calls that use the default stream
begin only after all preceding calls on the device (in any stream) have completed, and no operation on
the device (in any stream) commences until they are finished.

Concurrent copy and execute illustrates the basic technique.

Concurrent copy and execute

cudaStreamCreate(&stream1);
cudaStreamCreate(&stream2);
cudaMemcpyAsync(a_d, a_h, size, cudaMemcpyHostToDevice, stream1);
kernel<<<grid, block, 0, stream2>>>(otherData_d);

In this code, two streams are created and used in the data transfer and kernel executions as specified
in the last arguments of the cudaMemcpyAsync call and the kernel’s execution configuration.

Concurrent copy and execute demonstrates how to overlap kernel execution with asynchronous data
transfer. This technique could be used when the data dependency is such that the data can be broken
into chunks and transferred in multiple stages, launching multiple kernels to operate on each chunk
as it arrives. Sequential copy and execute and Staged concurrent copy and execute demonstrate this.
They produce equivalent results. The first segment shows the reference sequential implementation,
which transfers and operates on an array of N floats (where N is assumed to be evenly divisible by
nThreads).

Sequential copy and execute

cudaMemcpy(a_d, a_h, N*sizeof(float), dir);
kernel<<<N∕nThreads, nThreads>>>(a_d);

Staged concurrent copy and execute shows how the transfer and kernel execution can be broken up
into nStreams stages. This approach permits some overlapping of the data transfer and execution.

Staged concurrent copy and execute

size=N*sizeof(float)∕nStreams;
for (i=0; i<nStreams; i++) {

offset = i*N∕nStreams;
cudaMemcpyAsync(a_d+offset, a_h+offset, size, dir, stream[i]);
kernel<<<N∕(nThreads*nStreams), nThreads, 0,
stream[i]>>>(a_d+offset);

}

is assumed that N is evenly divisible by
it
(In Staged concurrent copy and execute,
nThreads*nStreams.) Because execution within a stream occurs sequentially, none of the ker-
nels will launch until the data transfers in their respective streams complete. Current GPUs can

13.1. Data Transfer Between Host and Device

39

CUDA C++ Best Practices Guide, Release 12.5

simultaneously process asynchronous data transfers and execute kernels. GPUs with a single copy
engine can perform one asynchronous data transfer and execute kernels whereas GPUs with two copy
engines can simultaneously perform one asynchronous data transfer from the host to the device, one
asynchronous data transfer from the device to the host, and execute kernels. The number of copy
engines on a GPU is given by the asyncEngineCount field of the cudaDeviceProp structure, which
is also listed in the output of the deviceQuery CUDA Sample. (It should be mentioned that it is not
possible to overlap a blocking transfer with an asynchronous transfer, because the blocking transfer
occurs in the default stream, so it will not begin until all previous CUDA calls complete. It will not allow
any other CUDA call to begin until it has completed.) A diagram depicting the timeline of execution
for the two code segments is shown in Figure 1, and nStreams is equal to 4 for Staged concurrent
copy and execute in the bottom half of the figure.

Fig. 1: Timeline comparison for copy and kernel execution

Top Sequential

Bottom

Concurrent

For this example, it is assumed that the data transfer and kernel execution times are comparable. In
such cases, and when the execution time (tE) exceeds the transfer time (tT), a rough estimate for the
overall time is tE + tT/nStreams for the staged version versus tE + tT for the sequential version. If the
transfer time exceeds the execution time, a rough estimate for the overall time is tT + tE/nStreams.

13.1.3. Zero Copy

Zero copy is a feature that was added in version 2.2 of the CUDA Toolkit. It enables GPU threads to
directly access host memory. For this purpose, it requires mapped pinned (non-pageable) memory.
On integrated GPUs (i.e., GPUs with the integrated field of the CUDA device properties structure set
to 1), mapped pinned memory is always a performance gain because it avoids superfluous copies as
integrated GPU and CPU memory are physically the same. On discrete GPUs, mapped pinned memory
is advantageous only in certain cases. Because the data is not cached on the GPU, mapped pinned
memory should be read or written only once, and the global loads and stores that read and write the
memory should be coalesced. Zero copy can be used in place of streams because kernel-originated
data transfers automatically overlap kernel execution without the overhead of setting up and deter-
mining the optimal number of streams.

Note: Low Priority: Use zero-copy operations on integrated GPUs for CUDA Toolkit version 2.2 and
later.

40

Chapter 13. Memory Optimizations

The host code in Zero-copy host code shows how zero copy is typically set up.

CUDA C++ Best Practices Guide, Release 12.5

Zero-copy host code

float *a_h, *a_map;
...
cudaGetDeviceProperties(&prop, 0);
if (!prop.canMapHostMemory)

exit(0);

cudaSetDeviceFlags(cudaDeviceMapHost);
cudaHostAlloc(&a_h, nBytes, cudaHostAllocMapped);
cudaHostGetDevicePointer(&a_map, a_h, 0);
kernel<<<gridSize, blockSize>>>(a_map);

In this code, the canMapHostMemory field of the structure returned by cudaGetDeviceProper-
ties() is used to check that the device supports mapping host memory to the device’s address
space. Page-locked memory mapping is enabled by calling cudaSetDeviceFlags() with cudaDe-
viceMapHost. Note that cudaSetDeviceFlags() must be called prior to setting a device or making
a CUDA call that requires state (that is, essentially, before a context is created). Page-locked mapped
host memory is allocated using cudaHostAlloc(), and the pointer to the mapped device address
space is obtained via the function cudaHostGetDevicePointer().
In the code in Zero-copy host
code, kernel() can reference the mapped pinned host memory using the pointer a_map in exactly
the same was as it would if a_map referred to a location in device memory.

Note: Mapped pinned host memory allows you to overlap CPU-GPU memory transfers with compu-
tation while avoiding the use of CUDA streams. But since any repeated access to such memory areas
causes repeated CPU-GPU transfers, consider creating a second area in device memory to manually
cache the previously read host memory data.

13.1.4. Unified Virtual Addressing

Devices of compute capability 2.0 and later support a special addressing mode called Unified Virtual
Addressing (UVA) on 64-bit Linux and Windows. With UVA, the host memory and the device memories
of all installed supported devices share a single virtual address space.

Prior to UVA, an application had to keep track of which pointers referred to device memory (and for
which device) and which referred to host memory as a separate bit of metadata (or as hard-coded
information in the program) for each pointer. Using UVA, on the other hand, the physical memory
space to which a pointer points can be determined simply by inspecting the value of the pointer using
cudaPointerGetAttributes().

Under UVA, pinned host memory allocated with cudaHostAlloc() will have identical host and de-
vice pointers, so it is not necessary to call cudaHostGetDevicePointer() for such allocations. Host
memory allocations pinned after-the-fact via cudaHostRegister(), however, will continue to have
different device pointers than their host pointers, so cudaHostGetDevicePointer() remains nec-
essary in that case.

UVA is also a necessary precondition for enabling peer-to-peer (P2P) transfer of data directly across
the PCIe bus or NVLink for supported GPUs in supported configurations, bypassing host memory.

See the CUDA C++ Programming Guide for further explanations and software requirements for UVA
and P2P.

13.1. Data Transfer Between Host and Device

41

CUDA C++ Best Practices Guide, Release 12.5

13.2. Device Memory Spaces

CUDA devices use several memory spaces, which have different characteristics that reflect their dis-
tinct usages in CUDA applications. These memory spaces include global, local, shared, texture, and
registers, as shown in Figure 2.

Fig. 2: Memory spaces on a CUDA device

Of these different memory spaces, global memory is the most plentiful; see Features and Technical
Specifications of the CUDA C++ Programming Guide for the amounts of memory available in each
memory space at each compute capability level. Global, local, and texture memory have the greatest
access latency, followed by constant memory, shared memory, and the register file.

The various principal traits of the memory types are shown in Table 1.

42

Chapter 13. Memory Optimizations

CUDA C++ Best Practices Guide, Release 12.5

Table 1: Table 1. Salient Features of Device Memory

Location
on/off
chip

On

Off

On

CachedAc-
cess

Scope

Life-
time

n/a

R/W 1 thread

Thread

Yes†† R/W 1 thread

Thread

n/a

R/W All

Block

threads
in block

Off

†

R/W All

Off

Yes

R

Off

Yes

R

threads +
host

All
threads +
host

All
threads +
host

Host
alloca-
tion

Host
alloca-
tion

Host
alloca-
tion

Memory

Register

Local

Shared

Global

Constant

Texture

† Cached in L1 and L2 by default on devices of com-
pute capability 6.0 and 7.x; cached only in L2 by
default on devices of lower compute capabilities,
though some allow opt-in to caching in L1 as well
via compilation flags.

†† Cached in L1 and L2 by default except on devices
of compute capability 5.x; devices of compute capa-
bility 5.x cache locals only in L2.

In the case of texture access, if a texture reference is bound to a linear array in global memory, then
the device code can write to the underlying array. Texture references that are bound to CUDA arrays
can be written to via surface-write operations by binding a surface to the same underlying CUDA array
storage). Reading from a texture while writing to its underlying global memory array in the same kernel
launch should be avoided because the texture caches are read-only and are not invalidated when the
associated global memory is modified.

13.2.1. Coalesced Access to Global Memory

A very important performance consideration in programming for CUDA-capable GPU architectures is
the coalescing of global memory accesses. Global memory loads and stores by threads of a warp are
coalesced by the device into as few as possible transactions.

Note: High Priority: Ensure global memory accesses are coalesced whenever possible.

The access requirements for coalescing depend on the compute capability of the device and are doc-
umented in the CUDA C++ Programming Guide.

13.2. Device Memory Spaces

43

CUDA C++ Best Practices Guide, Release 12.5

For devices of compute capability 6.0 or higher, the requirements can be summarized quite easily: the
concurrent accesses of the threads of a warp will coalesce into a number of transactions equal to the
number of 32-byte transactions necessary to service all of the threads of the warp.

For certain devices of compute capability 5.2, L1-caching of accesses to global memory can be option-
ally enabled. If L1-caching is enabled on these devices, the number of required transactions is equal
to the number of required 128-byte aligned segments.

Note: On devices of compute capability 6.0 or higher, L1-caching is the default, however the data
access unit is 32-byte regardless of whether global loads are cached in L1 or not.

On devices with GDDR memory, accessing memory in a coalesced way is even more important when
ECC is turned on. Scattered accesses increase ECC memory transfer overhead, especially when writing
data to global memory.

Coalescing concepts are illustrated in the following simple examples. These examples assume com-
pute capability 6.0 or higher and that accesses are for 4-byte words, unless otherwise noted.

13.2.1.1 A Simple Access Pattern

The first and simplest case of coalescing can be achieved by any CUDA-enabled device of compute
capability 6.0 or higher: the k-th thread accesses the k-th word in a 32-byte aligned array. Not all
threads need to participate.

For example, if the threads of a warp access adjacent 4-byte words (e.g., adjacent float values), four
coalesced 32-byte transactions will service that memory access. Such a pattern is shown in Figure 3.

Fig. 3: Coalesced access

This access pattern results in four 32-byte transactions, indicated by the red rectangles.

If from any of the four 32-byte segments only a subset of the words are requested (e.g.
if several
threads had accessed the same word or if some threads did not participate in the access), the full
segment is fetched anyway. Furthermore, if accesses by the threads of the warp had been permuted
within or accross the four segments, still only four 32-byte transactions would have been performed
by a device with compute capability 6.0 or higher.

13.2.1.2 A Sequential but Misaligned Access Pattern

If sequential threads in a warp access memory that is sequential but not aligned with a 32-byte seg-
ment, five 32-byte segments will be requested, as shown in Figure 4.

Memory allocated through the CUDA Runtime API, such as via cudaMalloc(), is guaranteed to be
aligned to at least 256 bytes. Therefore, choosing sensible thread block sizes, such as multiples of the
warp size (i.e., 32 on current GPUs), facilitates memory accesses by warps that are properly aligned.
(Consider what would happen to the memory addresses accessed by the second, third, and subse-
quent thread blocks if the thread block size was not a multiple of warp size, for example.)

44

Chapter 13. Memory Optimizations

CUDA C++ Best Practices Guide, Release 12.5

Fig. 4: Misaligned sequential addresses that fall within five 32-byte segments

13.2.1.3 Effects of Misaligned Accesses

It is easy and informative to explore the ramifications of misaligned accesses using a simple copy
kernel, such as the one in A copy kernel that illustrates misaligned accesses.

A copy kernel that illustrates misaligned accesses

__global__ void offsetCopy(float *odata, float* idata, int offset)
{

int xid = blockIdx.x * blockDim.x + threadIdx.x + offset;
odata[xid] = idata[xid];

}

In A copy kernel that illustrates misaligned accesses, data is copied from the input array idata to the
output array, both of which exist in global memory. The kernel is executed within a loop in host code
that varies the parameter offset from 0 to 32. (e.g. Figure 4 corresponds to this misalignments) The
effective bandwidth for the copy with various offsets on an NVIDIA Tesla V100 (compute capability
7.0) is shown in Figure 5.

Fig. 5: Performance of offsetCopy kernel

13.2. Device Memory Spaces

45

CUDA C++ Best Practices Guide, Release 12.5

For the NVIDIA Tesla V100, global memory accesses with no offset or with offsets that are multiples
of 8 words result in four 32-byte transactions. The achieved bandwidth is approximately 790 GB/s.
Otherwise, five 32-byte segments are loaded per warp, and we would expect approximately 4/5th of
the memory throughput achieved with no offsets.

In this particular example, the offset memory throughput achieved is, however, approximately 9/10th,
because adjacent warps reuse the cache lines their neighbors fetched. So while the impact is still
evident it is not as large as we might have expected. It would have been more so if adjacent warps had
not exhibited such a high degree of reuse of the over-fetched cache lines.

13.2.1.4 Strided Accesses

As seen above, in the case of misaligned sequential accesses, caches help to alleviate the performance
impact. It may be different with non-unit-strided accesses, however, and this is a pattern that occurs
frequently when dealing with multidimensional data or matrices. For this reason, ensuring that as
much as possible of the data in each cache line fetched is actually used is an important part of per-
formance optimization of memory accesses on these devices.

To illustrate the effect of strided access on effective bandwidth, see the kernel strideCopy() in
A kernel to illustrate non-unit stride data copy, which copies data with a stride of stride elements
between threads from idata to odata.

A kernel to illustrate non-unit stride data copy

__global__ void strideCopy(float *odata, float* idata, int stride)
{

int xid = (blockIdx.x*blockDim.x + threadIdx.x)*stride;
odata[xid] = idata[xid];

}

Figure 6 illustrates such a situation; in this case, threads within a warp access words in memory with a
stride of 2. This action leads to a load of eight L2 cache segments per warp on the Tesla V100 (compute
capability 7.0).

Fig. 6: Adjacent threads accessing memory with a stride of 2

A stride of 2 results in a 50% of load/store efficiency since half the elements in the transaction are not
used and represent wasted bandwidth. As the stride increases, the effective bandwidth decreases

46

Chapter 13. Memory Optimizations

CUDA C++ Best Practices Guide, Release 12.5

until the point where 32 32-byte segments are loaded for the 32 threads in a warp, as indicated in
Figure 7.

Fig. 7: Performance of strideCopy kernel

As illustrated in Figure 7, non-unit-stride global memory accesses should be avoided whenever possi-
ble. One method for doing so utilizes shared memory, which is discussed in the next section.

13.2.2. L2 Cache

Starting with CUDA 11.0, devices of compute capability 8.0 and above have the capability to influ-
ence persistence of data in the L2 cache. Because L2 cache is on-chip, it potentially provides higher
bandwidth and lower latency accesses to global memory.

For more details refer to the L2 Access Management section in the CUDA C++ Programming Guide.

13.2.2.1 L2 Cache Access Window

When a CUDA kernel accesses a data region in the global memory repeatedly, such data accesses can
be considered to be persisting. On the other hand, if the data is only accessed once, such data accesses
can be considered to be streaming. A portion of the L2 cache can be set aside for persistent accesses
to a data region in global memory. If this set-aside portion is not used by persistent accesses, then
streaming or normal data accesses can use it.

The L2 cache set-aside size for persisting accesses may be adjusted, within limits:

cudaGetDeviceProperties(&prop, device_id);
cudaDeviceSetLimit(cudaLimitPersistingL2CacheSize, prop.persistingL2CacheMaxSize); ∕*￿
,→Set aside max possible size of L2 cache for persisting accesses *∕

13.2. Device Memory Spaces

47

∕∕￿

∕∕￿

∕∕ (Must￿

∕∕ Hint￿

CUDA C++ Best Practices Guide, Release 12.5

Mapping of user data to L2 set-aside portion can be controlled using an access policy window on a
CUDA stream or CUDA graph kernel node. The example below shows how to use the access policy
window on a CUDA stream.

cudaStreamAttrValue stream_attribute;
,→Stream level attributes data structure
stream_attribute.accessPolicyWindow.base_ptr
,→Global Memory data pointer
stream_attribute.accessPolicyWindow.num_bytes = num_bytes;
,→Number of bytes for persisting accesses.

= reinterpret_cast<void*>(ptr); ∕∕￿

,→be less than cudaDeviceProp::accessPolicyMaxWindowSize)
stream_attribute.accessPolicyWindow.hitRatio
,→for L2 cache hit ratio for persisting accesses in the num_bytes region
stream_attribute.accessPolicyWindow.hitProp
,→of access property on cache hit
stream_attribute.accessPolicyWindow.missProp
,→of access property on cache miss.

= 1.0;

= cudaAccessPropertyPersisting; ∕∕ Type￿

= cudaAccessPropertyStreaming;

∕∕ Type￿

∕∕Set the attributes to a CUDA stream of type cudaStream_t
cudaStreamSetAttribute(stream, cudaStreamAttributeAccessPolicyWindow, &stream_
,→attribute);

The access policy window requires a value for hitRatio and num_bytes. Depending on the value of
the num_bytes parameter and the size of L2 cache, one may need to tune the value of hitRatio to
avoid thrashing of L2 cache lines.

13.2.2.2 Tuning the Access Window Hit-Ratio

The hitRatio parameter can be used to specify the fraction of accesses that receive the hitProp
property. For example, if the hitRatio value is 0.6, 60% of the memory accesses in the global memory
region [ptr..ptr+num_bytes) have the persisting property and 40% of the memory accesses have the
streaming property. To understand the effect of hitRatio and num_bytes, we use a sliding window
micro benchmark.

This microbenchmark uses a 1024 MB region in GPU global memory. First, we set aside 30 MB of the
L2 cache for persisting accesses using cudaDeviceSetLimit(), as discussed above. Then, as shown
in the figure below, we specify that the accesses to the first freqSize * sizeof(int) bytes of the
memory region are persistent. This data will thus use the L2 set-aside portion. In our experiment, we
vary the size of this persistent data region from 10 MB to 60 MB to model various scenarios where
data fits in or exceeds the available L2 set-aside portion of 30 MB. Note that the NVIDIA Tesla A100
GPU has 40 MB of total L2 cache capacity. Accesses to the remaining data of the memory region (i.e.,
streaming data) are considered normal or streaming accesses and will thus use the remaining 10 MB
of the non set-aside L2 portion (unless part of the L2 set-aside portion is unused).

Consider the following kernel code and access window parameters, as the implementation of the slid-
ing window experiment.

__global__ void kernel(int *data_persistent, int *data_streaming, int dataSize, int￿
,→freqSize) {

int tid = blockIdx.x * blockDim.x + threadIdx.x;

∕*Each CUDA thread accesses one element in the persistent data section

and one element in the streaming data section.
Because the size of the persistent memory region (freqSize * sizeof(int) bytes)￿

,→is much

48

(continues on next page)

Chapter 13. Memory Optimizations

CUDA C++ Best Practices Guide, Release 12.5

Fig. 8: Mapping Persistent data accesses to set-aside L2 in sliding window experiment

smaller than the size of the streaming memory region (dataSize * sizeof(int)￿

,→bytes), data

in the persistent region is accessed more frequently*∕

(continued from previous page)

data_persistent[tid % freqSize] = 2 * data_persistent[tid % freqSize];
data_streaming[tid % dataSize] = 2 * data_streaming[tid % dataSize];

}

stream_attribute.accessPolicyWindow.base_ptr
,→persistent);
stream_attribute.accessPolicyWindow.num_bytes = freqSize * sizeof(int);
,→bytes for persisting accesses in range 10-60 MB
stream_attribute.accessPolicyWindow.hitRatio
,→cache hit ratio. Fixed value 1.0

= 1.0;

= reinterpret_cast<void*>(data_

∕∕Number of￿

∕∕Hint for￿

The performance of the above kernel is shown in the chart below. When the persistent data region
fits well into the 30 MB set-aside portion of the L2 cache, a performance increase of as much as 50%
is observed. However, once the size of this persistent data region exceeds the size of the L2 set-aside
cache portion, approximately 10% performance drop is observed due to thrashing of L2 cache lines.

Fig. 9: The performance of the sliding-window benchmark with fixed hit-ratio of 1.0

13.2. Device Memory Spaces

49

CUDA C++ Best Practices Guide, Release 12.5

In order to optimize the performance, when the size of the persistent data is more than the size of the
set-aside L2 cache portion, we tune the num_bytes and hitRatio parameters in the access window
as below.

stream_attribute.accessPolicyWindow.base_ptr
,→persistent);
stream_attribute.accessPolicyWindow.num_bytes = 20*1024*1024;
,→
stream_attribute.accessPolicyWindow.hitRatio
,→((float)freqSize*sizeof(int));

∕∕20 MB

= (20*1024*1024)∕
∕∕Such that up to 20MB of data is resident.

= reinterpret_cast<void*>(data_

￿

We fix the num_bytes in the access window to 20 MB and tune the hitRatio such that a random 20
MB of the total persistent data is resident in the L2 set-aside cache portion. The remaining portion
of this persistent data will be accessed using the streaming property. This helps in reducing cache
thrashing. The results are shown in the chart below, where we see good performance regardless of
whether the persistent data fits in the L2 set-aside or not.

Fig. 10: The performance of the sliding-window benchmark with tuned hit-ratio

13.2.3. Shared Memory

Because it is on-chip, shared memory has much higher bandwidth and lower latency than local and
global memory - provided there are no bank conflicts between the threads, as detailed in the following
section.

50

Chapter 13. Memory Optimizations

CUDA C++ Best Practices Guide, Release 12.5

13.2.3.1 Shared Memory and Memory Banks

To achieve high memory bandwidth for concurrent accesses, shared memory is divided into equally
sized memory modules (banks) that can be accessed simultaneously. Therefore, any memory load or
store of n addresses that spans n distinct memory banks can be serviced simultaneously, yielding an
effective bandwidth that is n times as high as the bandwidth of a single bank.

However, if multiple addresses of a memory request map to the same memory bank, the accesses
are serialized. The hardware splits a memory request that has bank conflicts into as many separate
conflict-free requests as necessary, decreasing the effective bandwidth by a factor equal to the num-
ber of separate memory requests. The one exception here is when multiple threads in a warp address
the same shared memory location, resulting in a broadcast.
In this case, multiple broadcasts from
different banks are coalesced into a single multicast from the requested shared memory locations to
the threads.

To minimize bank conflicts, it is important to understand how memory addresses map to memory
banks and how to optimally schedule memory requests.

On devices of compute capability 5.x or newer, each bank has a bandwidth of 32 bits every clock cycle,
and successive 32-bit words are assigned to successive banks. The warp size is 32 threads and the
number of banks is also 32, so bank conflicts can occur between any threads in the warp. See Compute
Capability 5.x in the CUDA C++ Programming Guide for further details.

13.2.3.2 Shared Memory in Matrix Multiplication (C=AB)

Shared memory enables cooperation between threads in a block. When multiple threads in a block
use the same data from global memory, shared memory can be used to access the data from global
memory only once. Shared memory can also be used to avoid uncoalesced memory accesses by loading
and storing data in a coalesced pattern from global memory and then reordering it in shared memory.
Aside from memory bank conflicts, there is no penalty for non-sequential or unaligned accesses by a
warp in shared memory.

The use of shared memory is illustrated via the simple example of a matrix multiplication C = AB for
the case with A of dimension Mxw, B of dimension wxN, and C of dimension MxN. To keep the kernels
simple, M and N are multiples of 32, since the warp size (w) is 32 for current devices.

A natural decomposition of the problem is to use a block and tile size of wxw threads. Therefore, in
terms of wxw tiles, A is a column matrix, B is a row matrix, and C is their outer product; see Figure 11. A
grid of N/w by M/w blocks is launched, where each thread block calculates the elements of a different
tile in C from a single tile of A and a single tile of B.

To do this, the simpleMultiply kernel (Unoptimized matrix multiplication) calculates the output el-
ements of a tile of matrix C.

Unoptimized matrix multiplication

__global__ void simpleMultiply(float *a, float* b, float *c,

int N)

{

int row = blockIdx.y * blockDim.y + threadIdx.y;
int col = blockIdx.x * blockDim.x + threadIdx.x;
float sum = 0.0f;
for (int i = 0; i < TILE_DIM; i++) {

sum += a[row*TILE_DIM+i] * b[i*N+col];

13.2. Device Memory Spaces

51

(continues on next page)

CUDA C++ Best Practices Guide, Release 12.5

Fig. 11: Block-column matrix multiplied by block-row matrix. Block-column matrix (A) multiplied by
block-row matrix (B) with resulting product matrix (C).

(continued from previous page)

}
c[row*N+col] = sum;

}

In Unoptimized matrix multiplication, a, b, and c are pointers to global memory for the matrices A, B,
and C, respectively; blockDim.x, blockDim.y, and TILE_DIM are all equal to w. Each thread in the
wxw-thread block calculates one element in a tile of C. row and col are the row and column of the
element in C being calculated by a particular thread. The for loop over i multiplies a row of A by a
column of B, which is then written to C.

The effective bandwidth of this kernel is 119.9 GB/s on an NVIDIA Tesla V100. To analyze performance,
it is necessary to consider how warps access global memory in the for loop. Each warp of threads
calculates one row of a tile of C, which depends on a single row of A and an entire tile of B as illustrated
in Figure 12.

Fig. 12: Computing a row of a tile. Computing a row of a tile in C using one row of A and an entire tile
of B.

For each iteration i of the for loop, the threads in a warp read a row of the B tile, which is a sequential
and coalesced access for all compute capabilities.

52

Chapter 13. Memory Optimizations

CUDA C++ Best Practices Guide, Release 12.5

However, for each iteration i, all threads in a warp read the same value from global memory for matrix
A, as the index row*TILE_DIM+i is constant within a warp. Even though such an access requires
only 1 transaction on devices of compute capability 2.0 or higher, there is wasted bandwidth in the
transaction, because only one 4-byte word out of 8 words in a 32-byte cache segment is used. We can
reuse this cache line in subsequent iterations of the loop, and we would eventually utilize all 8 words;
however, when many warps execute on the same multiprocessor simultaneously, as is generally the
case, the cache line may easily be evicted from the cache between iterations i and i+1.

The performance on a device of any compute capability can be improved by reading a tile of A into
shared memory as shown in Using shared memory to improve the global memory load efficiency in
matrix multiplication.

Using shared memory to improve the global memory load efficiency in matrix multiplication

__global__ void coalescedMultiply(float *a, float* b, float *c,

int N)

{

}

__shared__ float aTile[TILE_DIM][TILE_DIM];

int row = blockIdx.y * blockDim.y + threadIdx.y;
int col = blockIdx.x * blockDim.x + threadIdx.x;
float sum = 0.0f;
aTile[threadIdx.y][threadIdx.x] = a[row*TILE_DIM+threadIdx.x];
__syncwarp();
for (int i = 0; i < TILE_DIM; i++) {

sum += aTile[threadIdx.y][i]* b[i*N+col];

}
c[row*N+col] = sum;

In Using shared memory to improve the global memory load efficiency in matrix multiplication, each
element in a tile of A is read from global memory only once, in a fully coalesced fashion (with no
wasted bandwidth), to shared memory. Within each iteration of the for loop, a value in shared mem-
ory is broadcast to all threads in a warp. Instead of a __syncthreads()synchronization barrier call,
a __syncwarp() is sufficient after reading the tile of A into shared memory because only threads
within the warp that write the data into shared memory read this data. This kernel has an effective
bandwidth of 144.4 GB/s on an NVIDIA Tesla V100. This illustrates the use of the shared memory as
a user-managed cache when the hardware L1 cache eviction policy does not match up well with the
needs of the application or when L1 cache is not used for reads from global memory.

A further improvement can be made to how Using shared memory to improve the global memory load
efficiency in matrix multiplication deals with matrix B. In calculating each of the rows of a tile of matrix
C, the entire tile of B is read. The repeated reading of the B tile can be eliminated by reading it into
shared memory once (Improvement by reading additional data into shared memory).

13.2. Device Memory Spaces

53

CUDA C++ Best Practices Guide, Release 12.5

Improvement by reading additional data into shared memory

__global__ void sharedABMultiply(float *a, float* b, float *c,

int N)

{

}

__shared__ float aTile[TILE_DIM][TILE_DIM],
bTile[TILE_DIM][TILE_DIM];

int row = blockIdx.y * blockDim.y + threadIdx.y;
int col = blockIdx.x * blockDim.x + threadIdx.x;
float sum = 0.0f;
aTile[threadIdx.y][threadIdx.x] = a[row*TILE_DIM+threadIdx.x];
bTile[threadIdx.y][threadIdx.x] = b[threadIdx.y*N+col];
__syncthreads();
for (int i = 0; i < TILE_DIM; i++) {

sum += aTile[threadIdx.y][i]* bTile[i][threadIdx.x];

}
c[row*N+col] = sum;

Note that in Improvement by reading additional data into shared memory, a __syncthreads() call is
required after reading the B tile because a warp reads data from shared memory that were written to
shared memory by different warps. The effective bandwidth of this routine is 195.5 GB/s on an NVIDIA
Tesla V100. Note that the performance improvement is not due to improved coalescing in either case,
but to avoiding redundant transfers from global memory.

The results of the various optimizations are summarized in Table 2.

Table 2: Table 2. Performance Improvements Optimizing C =
AB Matrix Multiply :class table-no-stripes

Optimization

No optimization

Coalesced using shared memory to store a tile of A

NVIDIA Tesla V100

119.9 GB/s

144.4 GB/s

Using shared memory to eliminate redundant reads of a tile of B 195.5 GB/s

Note: Medium Priority: Use shared memory to avoid redundant transfers from global memory.

13.2.3.3 Shared Memory in Matrix Multiplication (C=AAT)

A variant of the previous matrix multiplication can be used to illustrate how strided accesses to global
memory, as well as shared memory bank conflicts, are handled. This variant simply uses the transpose
of A in place of B, so C = AAT.

A simple implementation for C = AAT is shown in Unoptimized handling of strided accesses to global
memory

54

Chapter 13. Memory Optimizations

CUDA C++ Best Practices Guide, Release 12.5

Unoptimized handling of strided accesses to global memory

__global__ void simpleMultiply(float *a, float *c, int M)
{

int row = blockIdx.y * blockDim.y + threadIdx.y;
int col = blockIdx.x * blockDim.x + threadIdx.x;
float sum = 0.0f;
for (int i = 0; i < TILE_DIM; i++) {

sum += a[row*TILE_DIM+i] * a[col*TILE_DIM+i];

}
c[row*M+col] = sum;

}

In Unoptimized handling of strided accesses to global memory, the row-th, col-th element of C is ob-
tained by taking the dot product of the row-th and col-th rows of A. The effective bandwidth for this
kernel is 12.8 GB/s on an NVIDIA Tesla V100. These results are substantially lower than the corre-
sponding measurements for the C = AB kernel. The difference is in how threads in a half warp access
elements of A in the second term, a[col*TILE_DIM+i], for each iteration i. For a warp of threads,
col represents sequential columns of the transpose of A, and therefore col*TILE_DIM represents a
strided access of global memory with a stride of w, resulting in plenty of wasted bandwidth.

The way to avoid strided access is to use shared memory as before, except in this case a warp reads a
row of A into a column of a shared memory tile, as shown in An optimized handling of strided accesses
using coalesced reads from global memory.

An optimized handling of strided accesses using coalesced reads from global memory

__global__ void coalescedMultiply(float *a, float *c, int M)
{

__shared__ float aTile[TILE_DIM][TILE_DIM],

transposedTile[TILE_DIM][TILE_DIM];

int row = blockIdx.y * blockDim.y + threadIdx.y;
int col = blockIdx.x * blockDim.x + threadIdx.x;
float sum = 0.0f;
aTile[threadIdx.y][threadIdx.x] = a[row*TILE_DIM+threadIdx.x];
transposedTile[threadIdx.x][threadIdx.y] =

a[(blockIdx.x*blockDim.x + threadIdx.y)*TILE_DIM +
threadIdx.x];

__syncthreads();
for (int i = 0; i < TILE_DIM; i++) {

sum += aTile[threadIdx.y][i]* transposedTile[i][threadIdx.x];

}
c[row*M+col] = sum;

}

An optimized handling of strided accesses using coalesced reads from global memory uses the shared
transposedTile to avoid uncoalesced accesses in the second term in the dot product and the shared
aTile technique from the previous example to avoid uncoalesced accesses in the first term. The
effective bandwidth of this kernel is 140.2 GB/s on an NVIDIA Tesla V100.These results are lower than
those obtained by the final kernel for C = AB. The cause of the difference is shared memory bank
conflicts.

The reads of elements in transposedTile within the for loop are free of conflicts, because threads
of each half warp read across rows of the tile, resulting in unit stride across the banks. However, bank
conflicts occur when copying the tile from global memory into shared memory. To enable the loads
from global memory to be coalesced, data are read from global memory sequentially. However, this

13.2. Device Memory Spaces

55

CUDA C++ Best Practices Guide, Release 12.5

requires writing to shared memory in columns, and because of the use of wxw tiles in shared memory,
this results in a stride between threads of w banks - every thread of the warp hits the same bank (Recall
that w is selected as 32). These many-way bank conflicts are very expensive. The simple remedy is to
pad the shared memory array so that it has an extra column, as in the following line of code.

__shared__ float transposedTile[TILE_DIM][TILE_DIM+1];

This padding eliminates the conflicts entirely, because now the stride between threads is w+1 banks
(i.e., 33 for current devices), which, due to modulo arithmetic used to compute bank indices, is equiva-
lent to a unit stride. After this change, the effective bandwidth is 199.4 GB/s on an NVIDIA Tesla V100,
which is comparable to the results from the last C = AB kernel.

The results of these optimizations are summarized in Table 3.

Table 3: Table 3. Performance Improvements Optimizing C =
AAT Matrix Multiplication

Optimization

No optimization

NVIDIA Tesla V100

12.8 GB/s

Using shared memory to coalesce global reads 140.2 GB/s

Removing bank conflicts

199.4 GB/s

These results should be compared with those in Table 2. As can be seen from these tables, judicious
use of shared memory can dramatically improve performance.

The examples in this section have illustrated three reasons to use shared memory:

▶ To enable coalesced accesses to global memory, especially to avoid large strides (for general

matrices, strides are much larger than 32)

▶ To eliminate (or reduce) redundant loads from global memory

▶ To avoid wasted bandwidth

13.2.3.4 Asynchronous Copy from Global Memory to Shared Memory

CUDA 11.0 introduces an async-copy feature that can be used within device code to explicitly manage
the asynchronous copying of data from global memory to shared memory. This feature enables CUDA
It also avoids an
kernels to overlap copying data from global to shared memory with computation.
intermediary register file access traditionally present between the global memory read and the shared
memory write.

For more details refer to the memcpy_async section in the CUDA C++ Programming Guide.

To understand the performance difference between synchronous copy and asynchronous copy of data
from global memory to shared memory, consider the following micro benchmark CUDA kernels for
demonstrating the synchronous and asynchronous approaches. Asynchronous copies are hardware
accelerated for NVIDIA A100 GPU.

template <typename T>
__global__ void pipeline_kernel_sync(T *global, uint64_t *clock, size_t copy_count) {

extern __shared__ char s[];
T *shared = reinterpret_cast<T *>(s);

uint64_t clock_start = clock64();

(continues on next page)

56

Chapter 13. Memory Optimizations

CUDA C++ Best Practices Guide, Release 12.5

(continued from previous page)

for (size_t i = 0; i < copy_count; ++i) {

shared[blockDim.x * i + threadIdx.x] = global[blockDim.x * i + threadIdx.x];

}

uint64_t clock_end = clock64();

atomicAdd(reinterpret_cast<unsigned long long *>(clock),

clock_end - clock_start);

}

template <typename T>
__global__ void pipeline_kernel_async(T *global, uint64_t *clock, size_t copy_count) {

extern __shared__ char s[];
T *shared = reinterpret_cast<T *>(s);

uint64_t clock_start = clock64();

∕∕pipeline pipe;
for (size_t i = 0; i < copy_count; ++i) {

__pipeline_memcpy_async(&shared[blockDim.x * i + threadIdx.x],

&global[blockDim.x * i + threadIdx.x], sizeof(T));

}
__pipeline_commit();
__pipeline_wait_prior(0);

uint64_t clock_end = clock64();

atomicAdd(reinterpret_cast<unsigned long long *>(clock),

clock_end - clock_start);

}

The synchronous version for the kernel loads an element from global memory to an intermediate regis-
ter and then stores the intermediate register value to shared memory. In the asynchronous version of
the kernel, instructions to load from global memory and store directly into shared memory are issued
as soon as __pipeline_memcpy_async() function is called. The __pipeline_wait_prior(0) will
wait until all the instructions in the pipe object have been executed. Using asynchronous copies does
not use any intermediate register. Not using intermediate registers can help reduce register pressure
and can increase kernel occupancy. Data copied from global memory to shared memory using asyn-
chronous copy instructions can be cached in the L1 cache or the L1 cache can be optionally bypassed.
If individual CUDA threads are copying elements of 16 bytes, the L1 cache can be bypassed. This
difference is illustrated in Figure 13.

Fig. 13: Comparing Synchronous vs Asynchronous Copy from Global Memory to Shared Memory

We evaluate the performance of both kernels using elements of size 4B, 8B and 16B per thread i.e.,
using int, int2 and int4 for the template parameter. We adjust the copy_count in the kernels such
that each thread block copies from 512 bytes up to 48 MB. The performance of the kernels is shown
in Figure 14.

From the performance chart, the following observations can be made for this experiment.

13.2. Device Memory Spaces

57

CUDA C++ Best Practices Guide, Release 12.5

Fig. 14: Comparing Performance of Synchronous vs Asynchronous Copy from Global Memory to Shared
Memory

▶ Best performance with synchronous copy is achieved when the copy_count parameter is a mul-
tiple of 4 for all three element sizes. The compiler can optimize groups of 4 load and store in-
structions. This is evident from the saw tooth curves.

▶ Asynchronous copy achieves better performance in nearly all cases.
▶ The async-copy does not require the copy_count parameter to be a multiple of 4, to maximize

performance through compiler optimizations.

▶ Overall, best performance is achieved when using asynchronous copies with an element of size

8 or 16 bytes.

13.2.4. Local Memory

Local memory is so named because its scope is local to the thread, not because of its physical location.
In fact, local memory is off-chip. Hence, access to local memory is as expensive as access to global
memory. In other words, the term local in the name does not imply faster access.

Local memory is used only to hold automatic variables. This is done by the nvcc compiler when it
determines that there is insufficient register space to hold the variable. Automatic variables that are
likely to be placed in local memory are large structures or arrays that would consume too much register
space and arrays that the compiler determines may be indexed dynamically.

Inspection of the PTX assembly code (obtained by compiling with -ptx or -keep command-line op-
tions to nvcc) reveals whether a variable has been placed in local memory during the first compilation
phases. If it has, it will be declared using the .local mnemonic and accessed using the ld.local
and st.local mnemonics. If it has not, subsequent compilation phases might still decide otherwise,
if they find the variable consumes too much register space for the targeted architecture. There is no

58

Chapter 13. Memory Optimizations

CUDA C++ Best Practices Guide, Release 12.5

way to check this for a specific variable, but the compiler reports total local memory usage per kernel
(lmem) when run with the--ptxas-options=-v option.

13.2.5. Texture Memory

The read-only texture memory space is cached. Therefore, a texture fetch costs one device memory
read only on a cache miss; otherwise, it just costs one read from the texture cache. The texture cache
is optimized for 2D spatial locality, so threads of the same warp that read texture addresses that are
close together will achieve best performance. Texture memory is also designed for streaming fetches
with a constant latency; that is, a cache hit reduces DRAM bandwidth demand, but not fetch latency.

In certain addressing situations, reading device memory through texture fetching can be an advanta-
geous alternative to reading device memory from global or constant memory.

13.2.5.1 Additional Texture Capabilities

If textures are fetched using tex1D(),tex2D(), or tex3D() rather than tex1Dfetch(), the hardware
provides other capabilities that might be useful for some applications such as image processing, as
shown in Table 4.

Table 4: Table 4. Useful Features for tex1D(), tex2D(), and
tex3D() Fetches

Feature

Filtering

Use

Caveat

Fast, low-precision in-
terpolation between
texels

Valid only if the texture
reference returns floating-
point data

Normalized texture coordinates

Resolution-
independent coding

None

Addressing modes

Automatic
of boundary cases1

handling

Can be used only with
normalized texture coordi-
nates

1 The automatic handling of boundary cases
in the bottom row of Table 4 refers to how
a texture coordinate is resolved when it falls
outside the valid addressing range. There are
two options: clamp and wrap.
If x is the co-
ordinate and N is the number of texels for
a one-dimensional texture, then with clamp,
x is replaced by 0 if x < 0 and by 1-1/N if 1
<x. With wrap, x is replaced by frac(x) where
frac(x) = x - floor(x). Floor returns the largest
integer less than or equal to x. So, in clamp
mode where N = 1, an x of 1.3 is clamped to
1.0; whereas in wrap mode, it is converted to
0.3

Within a kernel call, the texture cache is not kept coherent with respect to global memory writes, so
texture fetches from addresses that have been written via global stores in the same kernel call return

13.2. Device Memory Spaces

59

CUDA C++ Best Practices Guide, Release 12.5

undefined data. That is, a thread can safely read a memory location via texture if the location has
been updated by a previous kernel call or memory copy, but not if it has been previously updated by
the same thread or another thread within the same kernel call.

13.2.6. Constant Memory

There is a total of 64 KB constant memory on a device. The constant memory space is cached. As
a result, a read from constant memory costs one memory read from device memory only on a cache
miss; otherwise, it just costs one read from the constant cache. Accesses to different addresses by
threads within a warp are serialized, thus the cost scales linearly with the number of unique addresses
read by all threads within a warp. As such, the constant cache is best when threads in the same warp
accesses only a few distinct locations. If all threads of a warp access the same location, then constant
memory can be as fast as a register access.

13.2.7. Registers

Generally, accessing a register consumes zero extra clock cycles per instruction, but delays may occur
due to register read-after-write dependencies and register memory bank conflicts.

The compiler and hardware thread scheduler will schedule instructions as optimally as possible to
avoid register memory bank conflicts. An application has no direct control over these bank conflicts.
In particular, there is no register-related reason to pack data into vector data types such as float4 or
int4 types.

13.2.7.1 Register Pressure

Register pressure occurs when there are not enough registers available for a given task. Even though
each multiprocessor contains thousands of 32-bit registers (see Features and Technical Specifications
of the CUDA C++ Programming Guide), these are partitioned among concurrent threads. To prevent
the compiler from allocating too many registers, use the -maxrregcount=N compiler command-line
option (see nvcc) or the launch bounds kernel definition qualifier (see Execution Configuration of the
CUDA C++ Programming Guide) to control the maximum number of registers to allocated per thread.

13.3. Allocation

Device memory allocation and de-allocation via cudaMalloc() and cudaFree() are expensive op-
erations. It is recommended to use cudaMallocAsync() and cudaFreeAsync() which are stream
ordered pool allocators to manage device memory.

60

Chapter 13. Memory Optimizations

CUDA C++ Best Practices Guide, Release 12.5

13.4. NUMA Best Practices

Some recent Linux distributions enable automatic NUMA balancing (or “AutoNUMA”) by default.
In
some instances, operations performed by automatic NUMA balancing may degrade the performance
of applications running on NVIDIA GPUs. For optimal performance, users should manually tune the
NUMA characteristics of their application.

The optimal NUMA tuning will depend on the characteristics and desired hardware affinities of each
application and node, but in general applications computing on NVIDIA GPUs are advised to choose a
policy that disables automatic NUMA balancing. For example, on IBM Newell POWER9 nodes (where
the CPUs correspond to NUMA nodes 0 and 8), use:

numactl --membind=0,8

to bind memory allocations to the CPUs.

13.4. NUMA Best Practices

61

CUDA C++ Best Practices Guide, Release 12.5

62

Chapter 13. Memory Optimizations

Chapter 14. Execution Configuration

Optimizations

One of the keys to good performance is to keep the multiprocessors on the device as busy as pos-
sible. A device in which work is poorly balanced across the multiprocessors will deliver suboptimal
performance. Hence, it’s important to design your application to use threads and blocks in a way that
maximizes hardware utilization and to limit practices that impede the free distribution of work. A key
concept in this effort is occupancy, which is explained in the following sections.

Hardware utilization can also be improved in some cases by designing your application so that multiple,
independent kernels can execute at the same time. Multiple kernels executing at the same time is
known as concurrent kernel execution. Concurrent kernel execution is described below.

Another important concept is the management of system resources allocated for a particular task.
How to manage this resource utilization is discussed in the final sections of this chapter.

14.1. Occupancy

Thread instructions are executed sequentially in CUDA, and, as a result, executing other warps when
one warp is paused or stalled is the only way to hide latencies and keep the hardware busy. Some metric
related to the number of active warps on a multiprocessor is therefore important in determining how
effectively the hardware is kept busy. This metric is occupancy.

Occupancy is the ratio of the number of active warps per multiprocessor to the maximum number of
possible active warps. (To determine the latter number, see the deviceQuery CUDA Sample or refer
to Compute Capabilities in the CUDA C++ Programming Guide.) Another way to view occupancy is the
percentage of the hardware’s ability to process warps that is actively in use.

Higher occupancy does not always equate to higher performance-there is a point above which addi-
tional occupancy does not improve performance. However, low occupancy always interferes with the
ability to hide memory latency, resulting in performance degradation.

Per thread resources required by a CUDA kernel might limit the maximum block size in an un-
wanted way.
In order to maintain forward compatibility to future hardware and toolkits and to
ensure that at least one thread block can run on an SM, developers should include the single ar-
gument __launch_bounds__(maxThreadsPerBlock) which specifies the largest block size that
the kernel will be launched with. Failure to do so could lead to “too many resources requested for
launch” errors. Providing the two argument version of __launch_bounds__(maxThreadsPerBlock,
minBlocksPerMultiprocessor) can improve performance in some cases. The right value for min-
BlocksPerMultiprocessor should be determined using a detailed per kernel analysis.

63

CUDA C++ Best Practices Guide, Release 12.5

14.1.1. Calculating Occupancy

One of several factors that determine occupancy is register availability. Register storage enables
threads to keep local variables nearby for low-latency access. However, the set of registers (known
as the register file) is a limited commodity that all threads resident on a multiprocessor must share.
Registers are allocated to an entire block all at once. So, if each thread block uses many registers,
the number of thread blocks that can be resident on a multiprocessor is reduced, thereby lower-
ing the occupancy of the multiprocessor. The maximum number of registers per thread can be
set manually at compilation time per-file using the -maxrregcount option or per-kernel using the
__launch_bounds__ qualifier (see Register Pressure).

For purposes of calculating occupancy, the number of registers used by each thread is one of the
key factors. For example, on devices of compute capability 7.0 each multiprocessor has 65,536 32-
bit registers and can have a maximum of 2048 simultaneous threads resident (64 warps x 32 threads
per warp). This means that in one of these devices, for a multiprocessor to have 100% occupancy,
each thread can use at most 32 registers. However, this approach of determining how register count
affects occupancy does not take into account the register allocation granularity. For example, on a
device of compute capability 7.0, a kernel with 128-thread blocks using 37 registers per thread results
in an occupancy of 75% with 12 active 128-thread blocks per multi-processor, whereas a kernel with
320-thread blocks using the same 37 registers per thread results in an occupancy of 63% because only
four 320-thread blocks can reside on a multiprocessor. Furthermore, register allocations are rounded
up to the nearest 256 registers per warp.

The number of registers available, the maximum number of simultaneous threads resident on each
multiprocessor, and the register allocation granularity vary over different compute capabilities. Be-
cause of these nuances in register allocation and the fact that a multiprocessor’s shared memory is
also partitioned between resident thread blocks, the exact relationship between register usage and oc-
cupancy can be difficult to determine. The --ptxas options=v option of nvcc details the number of
registers used per thread for each kernel. See Hardware Multithreading of the CUDA C++ Programming
Guide for the register allocation formulas for devices of various compute capabilities and Features and
Technical Specifications of the CUDA C++ Programming Guide for the total number of registers avail-
able on those devices. Alternatively, NVIDIA provides an occupancy calculator in the form of an Excel
spreadsheet that enables developers to hone in on the optimal balance and to test different possible
scenarios more easily. This spreadsheet, shown in Figure 15, is called CUDA_Occupancy_Calculator.
xls and is located in the tools subdirectory of the CUDA Toolkit installation.

In addition to the calculator spreadsheet, occupancy can be determined using the NVIDIA Nsight Com-
pute Profiler. Details about occupancy are displayed in the Occupancy section.

An application can also use the Occupancy API from the CUDA Runtime, e.g. cudaOccupancyMax-
ActiveBlocksPerMultiprocessor, to dynamically select launch configurations based on runtime
parameters.

14.2. Hiding Register Dependencies

Note: Medium Priority: To hide latency arising from register dependencies, maintain sufficient num-
bers of active threads per multiprocessor (i.e., sufficient occupancy).

Register dependencies arise when an instruction uses a result stored in a register written by an instruc-
tion before it. The latency of most arithmetic instructions is typically 4 cycles on devices of compute
capability 7.0. So threads must wait approximatly 4 cycles before using an arithmetic result. However,

64

Chapter 14. Execution Configuration Optimizations

CUDA C++ Best Practices Guide, Release 12.5

Fig. 1: Using the CUDA Occupancy Calculator to project GPU multiprocessor occupancy

this latency can be completely hidden by the execution of threads in other warps. See Registers for
details.

14.3. Thread and Block Heuristics

Note: Medium Priority: The number of threads per block should be a multiple of 32 threads, because
this provides optimal computing efficiency and facilitates coalescing.

The dimension and size of blocks per grid and the dimension and size of threads per block are both
important factors. The multidimensional aspect of these parameters allows easier mapping of mul-
tidimensional problems to CUDA and does not play a role in performance. As a result, this section
discusses size but not dimension.

Latency hiding and occupancy depend on the number of active warps per multiprocessor, which is
implicitly determined by the execution parameters along with resource (register and shared memory)
constraints. Choosing execution parameters is a matter of striking a balance between latency hiding
(occupancy) and resource utilization.

Choosing the execution configuration parameters should be done in tandem; however, there are cer-
tain heuristics that apply to each parameter individually. When choosing the first execution configu-
ration parameter-the number of blocks per grid, or grid size - the primary concern is keeping the entire
GPU busy. The number of blocks in a grid should be larger than the number of multiprocessors so that
all multiprocessors have at least one block to execute. Furthermore, there should be multiple active
blocks per multiprocessor so that blocks that aren’t waiting for a __syncthreads() can keep the

14.3. Thread and Block Heuristics

65

CUDA C++ Best Practices Guide, Release 12.5

hardware busy. This recommendation is subject to resource availability; therefore, it should be deter-
mined in the context of the second execution parameter - the number of threads per block, or block
size - as well as shared memory usage. To scale to future devices, the number of blocks per kernel
launch should be in the thousands.

When choosing the block size, it is important to remember that multiple concurrent blocks can reside
on a multiprocessor, so occupancy is not determined by block size alone. In particular, a larger block
size does not imply a higher occupancy.

As mentioned in Occupancy, higher occupancy does not always equate to better performance. For
example, improving occupancy from 66 percent to 100 percent generally does not translate to a similar
increase in performance. A lower occupancy kernel will have more registers available per thread than
a higher occupancy kernel, which may result in less register spilling to local memory; in particular, with
a high degree of exposed instruction-level parallelism (ILP) it is, in some cases, possible to fully cover
latency with a low occupancy.

There are many such factors involved in selecting block size, and inevitably some experimentation is
required. However, a few rules of thumb should be followed:

▶ Threads per block should be a multiple of warp size to avoid wasting computation on under-

populated warps and to facilitate coalescing.

▶ A minimum of 64 threads per block should be used, and only if there are multiple concurrent

blocks per multiprocessor.

▶ Between 128 and 256 threads per block is a good initial range for experimentation with different

block sizes.

▶ Use several smaller thread blocks rather than one large thread block per multiprocessor if la-
tency affects performance. This is particularly beneficial to kernels that frequently call __sync-
threads().

Note that when a thread block allocates more registers than are available on a multiprocessor, the
kernel launch fails, as it will when too much shared memory or too many threads are requested.

14.4. Effects of Shared Memory

Shared memory can be helpful in several situations, such as helping to coalesce or eliminate redundant
access to global memory. However, it also can act as a constraint on occupancy. In many cases, the
amount of shared memory required by a kernel is related to the block size that was chosen, but the
mapping of threads to shared memory elements does not need to be one-to-one. For example, it
may be desirable to use a 64x64 element shared memory array in a kernel, but because the maximum
number of threads per block is 1024, it is not possible to launch a kernel with 64x64 threads per block.
In such cases, kernels with 32x32 or 64x16 threads can be launched with each thread processing
four elements of the shared memory array. The approach of using a single thread to process multiple
elements of a shared memory array can be beneficial even if limits such as threads per block are not
an issue. This is because some operations common to each element can be performed by the thread
once, amortizing the cost over the number of shared memory elements processed by the thread.

A useful technique to determine the sensitivity of performance to occupancy is through experimen-
tation with the amount of dynamically allocated shared memory, as specified in the third parameter
of the execution configuration. By simply increasing this parameter (without modifying the kernel), it
is possible to effectively reduce the occupancy of the kernel and measure its effect on performance.

66

Chapter 14. Execution Configuration Optimizations

CUDA C++ Best Practices Guide, Release 12.5

14.5. Concurrent Kernel Execution

As described in Asynchronous and Overlapping Transfers with Computation, CUDA streams can be
used to overlap kernel execution with data transfers. On devices that are capable of concurrent ker-
nel execution, streams can also be used to execute multiple kernels simultaneously to more fully take
advantage of the device’s multiprocessors. Whether a device has this capability is indicated by the
concurrentKernels field of the cudaDeviceProp structure (or listed in the output of the device-
Query CUDA Sample). Non-default streams (streams other than stream 0) are required for concurrent
execution because kernel calls that use the default stream begin only after all preceding calls on the
device (in any stream) have completed, and no operation on the device (in any stream) commences
until they are finished.

The following example illustrates the basic technique. Because kernel1 and kernel2 are executed
in different, non-default streams, a capable device can execute the kernels at the same time.

cudaStreamCreate(&stream1);
cudaStreamCreate(&stream2);
kernel1<<<grid, block, 0, stream1>>>(data_1);
kernel2<<<grid, block, 0, stream2>>>(data_2);

14.6. Multiple contexts

CUDA work occurs within a process space for a particular GPU known as a context. The context encap-
sulates kernel launches and memory allocations for that GPU as well as supporting constructs such
as the page tables. The context is explicit in the CUDA Driver API but is entirely implicit in the CUDA
Runtime API, which creates and manages contexts automatically.

With the CUDA Driver API, a CUDA application process can potentially create more than one context for
a given GPU. If multiple CUDA application processes access the same GPU concurrently, this almost
always implies multiple contexts, since a context is tied to a particular host process unless Multi-
Process Service is in use.

While multiple contexts (and their associated resources such as global memory allocations) can be
allocated concurrently on a given GPU, only one of these contexts can execute work at any given mo-
ment on that GPU; contexts sharing the same GPU are time-sliced. Creating additional contexts incurs
memory overhead for per-context data and time overhead for context switching. Furthermore, the
need for context switching can reduce utilization when work from several contexts could otherwise
execute concurrently (see also Concurrent Kernel Execution).

Therefore, it is best to avoid multiple contexts per GPU within the same CUDA application. To assist
with this, the CUDA Driver API provides methods to access and manage a special context on each GPU
called the primary context. These are the same contexts used implicitly by the CUDA Runtime when
there is not already a current context for a thread.

∕∕ When initializing the program∕library
CUcontext ctx;
cuDevicePrimaryCtxRetain(&ctx, dev);

∕∕ When the program∕library launches work
cuCtxPushCurrent(ctx);
kernel<<<...>>>(...);
cuCtxPopCurrent(&ctx);

14.5. Concurrent Kernel Execution

(continues on next page)

67

CUDA C++ Best Practices Guide, Release 12.5

∕∕ When the program∕library is finished with the context
cuDevicePrimaryCtxRelease(dev);

(continued from previous page)

Note: NVIDIA-SMI can be used to configure a GPU for exclusive process mode, which limits the num-
ber of contexts per GPU to one. This context can be current to as many threads as desired within
the creating process, and cuDevicePrimaryCtxRetain will fail if a non-primary context that was
created with the CUDA driver API already exists on the device.

68

Chapter 14. Execution Configuration Optimizations

Chapter 15. Instruction Optimization

Awareness of how instructions are executed often permits low-level optimizations that can be useful,
especially in code that is run frequently (the so-called hot spot in a program). Best practices suggest
that this optimization be performed after all higher-level optimizations have been completed.

15.1. Arithmetic Instructions

Single-precision floats provide the best performance, and their use is highly encouraged. The through-
put of individual arithmetic operations is detailed in the CUDA C++ Programming Guide.

15.1.1. Division Modulo Operations

Note: Low Priority: Use shift operations to avoid expensive division and modulo calculations.

Integer division and modulo operations are particularly costly and should be avoided or replaced with
bitwise operations whenever possible: If n is a power of 2, ( i/n ) is equivalent to ( i ≫ log2(n) ) and (
i%n ) is equivalent to ( i& (n − 1) ).

The compiler will perform these conversions if n is literal. (For further information, refer to Perfor-
mance Guidelines in the CUDA C++ Programming Guide).

15.1.2. Loop Counters Signed vs. Unsigned

Note: Low Medium Priority: Use signed integers rather than unsigned integers as loop counters.

In the C language standard, unsigned integer overflow semantics are well defined, whereas signed in-
teger overflow causes undefined results. Therefore, the compiler can optimize more aggressively with
signed arithmetic than it can with unsigned arithmetic. This is of particular note with loop counters:
since it is common for loop counters to have values that are always positive, it may be tempting to
declare the counters as unsigned. For slightly better performance, however, they should instead be
declared as signed.

For example, consider the following code:

69

CUDA C++ Best Practices Guide, Release 12.5

for (i = 0; i < n; i++) {

out[i] = in[offset + stride*i];

}

Here, the sub-expression stride*i could overflow a 32-bit integer, so if i is declared as unsigned, the
overflow semantics prevent the compiler from using some optimizations that might otherwise have
applied, such as strength reduction. If instead i is declared as signed, where the overflow semantics
are undefined, the compiler has more leeway to use these optimizations.

15.1.3. Reciprocal Square Root

The reciprocal square root should always be invoked explicitly as rsqrtf() for single precision and
rsqrt() for double precision. The compiler optimizes 1.0f∕sqrtf(x) into rsqrtf() only when this
does not violate IEEE-754 semantics.

15.1.4. Other Arithmetic Instructions

Note: Low Priority: Avoid automatic conversion of doubles to floats.

The compiler must on occasion insert conversion instructions, introducing additional execution cycles.
This is the case for:

▶ Functions operating on char or short whose operands generally need to be converted to an int
▶ Double-precision floating-point constants (defined without any type suffix) used as input to

single-precision floating-point computations

The latter case can be avoided by using single-precision floating-point constants, defined with an f
suffix such as 3.141592653589793f, 1.0f, 0.5f.

For single-precision code, use of the float type and the single-precision math functions are highly
recommended.

It should also be noted that the CUDA math library’s complementary error function, erfcf(), is par-
ticularly fast with full single-precision accuracy.

15.1.5. Exponentiation With Small Fractional Arguments

For some fractional exponents, exponentiation can be accelerated significantly compared to the use
of pow() by using square roots, cube roots, and their inverses. For those exponentiations where the
exponent is not exactly representable as a floating-point number, such as 1/3, this can also provide
much more accurate results, as use of pow() magnifies the initial representational error.

The formulas in the table below are valid for x >= 0, x != -0, that is, signbit(x) == 0.

70

Chapter 15. Instruction Optimization

CUDA C++ Best Practices Guide, Release 12.5

Table 1: Table 5. Formulae for exponentiation by small fractions

Computation Formula

x1/9

x-1/9

x1/6

x-1/6

x1/4

x-1/4

x1/3

x-1/3

x1/2

x-1/2

x2/3

x-2/3

x3/4

x-3/4

x7/6

x-7/6

x5/4

x-5/4

x4/3

x-4/3

x3/2

x-3/2

r = rcbrt(rcbrt(x))

r = cbrt(rcbrt(x))

r = rcbrt(rsqrt(x))

r = rcbrt(sqrt(x))

r = rsqrt(rsqrt(x))

r = sqrt(rsqrt(x))

r = cbrt(x)

r = rcbrt(x)

r = sqrt(x)

r = rsqrt(x)

r = cbrt(x); r = r*r

r = rcbrt(x); r = r*r

r = sqrt(x); r = r*sqrt(r)

r = rsqrt(x); r = r*sqrt(r)

r = x*rcbrt(rsqrt(x))

r = (1∕x) * rcbrt(sqrt(x))

r = x*rsqrt(rsqrt(x))

r = (1∕x)*sqrt(rsqrt(x))

r = x*cbrt(x)

r = (1∕x)*rcbrt(x)

r = x*sqrt(x)

r = (1∕x)*rsqrt(x)

15.1.6. Math Libraries

Note: Medium Priority: Use the fast math library whenever speed trumps precision.

Two types of runtime math operations are supported. They can be distinguished by their names:
some have names with prepended underscores, whereas others do not (e.g., __functionName() ver-
sus functionName()). Functions following the __functionName() naming convention map directly
to the hardware level. They are faster but provide somewhat lower accuracy (e.g., __sinf(x) and
__expf(x)). Functions following functionName() naming convention are slower but have higher ac-
curacy (e.g., sinf(x) and expf(x)). The throughput of __sinf(x), __cosf(x), and__expf(x) is
much greater than that of sinf(x), cosf(x), and expf(x). The latter become even more expensive
(about an order of magnitude slower) if the magnitude of the argument x needs to be reduced. More-

15.1. Arithmetic Instructions

71

CUDA C++ Best Practices Guide, Release 12.5

over, in such cases, the argument-reduction code uses local memory, which can affect performance
even more because of the high latency of local memory. More details are available in the CUDA C++
Programming Guide.

Note also that whenever sine and cosine of the same argument are computed, the sincos family of
instructions should be used to optimize performance:

▶ __sincosf() for single-precision fast math (see next paragraph)
▶ sincosf() for regular single-precision
▶ sincos() for double precision

The -use_fast_math compiler option of nvcc coerces every functionName() call to the equivalent
__functionName() call. It also disables single-precision denormal support and lowers the precision of
single-precision division in general. This is an aggressive optimization that can both reduce numerical
accuracy and alter special case handling. A more robust approach is to selectively introduce calls
to fast intrinsic functions only if merited by performance gains and where altered behavior can be
tolerated. Note this switch is effective only on single-precision floating point.

Note: Medium Priority: Prefer faster, more specialized math functions over slower, more general ones
when possible.

For small integer powers (e.g., x2 or x3), explicit multiplication is almost certainly faster than the use of
general exponentiation routines such as pow(). While compiler optimization improvements continually
seek to narrow this gap, explicit multiplication (or the use of an equivalent purpose-built inline function
or macro) can have a significant advantage. This advantage is increased when several powers of the
same base are needed (e.g., where both x2 and x5 are calculated in close proximity), as this aids the
compiler in its common sub-expression elimination (CSE) optimization.

For exponentiation using base 2 or 10, use the functions exp2() or expf2() and exp10() or
expf10() rather than the functions pow() or powf(). Both pow() and powf() are heavy-weight
functions in terms of register pressure and instruction count due to the numerous special cases aris-
ing in general exponentiation and the difficulty of achieving good accuracy across the entire ranges of
the base and the exponent. The functions exp2(), exp2f(), exp10(), and exp10f(), on the other
hand, are similar to exp() and expf() in terms of performance, and can be as much as ten times
faster than their pow()/powf() equivalents.

For exponentiation with an exponent of 1/3, use the cbrt() or cbrtf() function rather than the
generic exponentiation functions pow() or powf(), as the former are significantly faster than the
latter. Likewise, for exponentation with an exponent of -1/3, use rcbrt() or rcbrtf().

Replace sin(π*<expr>) with sinpi(<expr>), cos(π*<expr>) with cospi(<expr>), and sin-
cos(π*<expr>) with sincospi(<expr>). This is advantageous with regard to both accuracy and
performance. As a particular example, to evaluate the sine function in degrees instead of radians, use
sinpi(x∕180.0). Similarly, the single-precision functions sinpif(), cospif(), and sincospif()
should replace calls to sinf(), cosf(), and sincosf() when the function argument is of the form
π*<expr>. (The performance advantage sinpi() has over sin() is due to simplified argument re-
duction; the accuracy advantage is because sinpi() multiplies by π only implicitly, effectively using
an infinitely precise mathematical π rather than a single- or double-precision approximation thereof.)

72

Chapter 15. Instruction Optimization

CUDA C++ Best Practices Guide, Release 12.5

15.1.7. Precision-related Compiler Flags

By default, the nvcc compiler generates IEEE-compliant code, but it also provides options to generate
code that somewhat less accurate but faster:

▶ -ftz=true (denormalized numbers are flushed to zero)
▶ -prec-div=false (less precise division)
▶ -prec-sqrt=false (less precise square root)

Another, more aggressive, option is -use_fast_math, which coerces every functionName() call to
the equivalent __functionName() call. This makes the code run faster at the cost of diminished
precision and accuracy. See Math Libraries.

15.2. Memory Instructions

Note: High Priority: Minimize the use of global memory. Prefer shared memory access where possible.

Memory instructions include any instruction that reads from or writes to shared, local, or global mem-
ory. When accessing uncached local or global memory, there are hundreds of clock cycles of memory
latency.

As an example, the assignment operator in the following sample code has a high throughput, but,
crucially, there is a latency of hundreds of clock cycles to read data from global memory:

__shared__ float shared[32];
__device__ float device[32];
shared[threadIdx.x] = device[threadIdx.x];

Much of this global memory latency can be hidden by the thread scheduler if there are sufficient
independent arithmetic instructions that can be issued while waiting for the global memory access to
complete. However, it is best to avoid accessing global memory whenever possible.

15.2. Memory Instructions

73

CUDA C++ Best Practices Guide, Release 12.5

74

Chapter 15. Instruction Optimization

Chapter 16. Control Flow

16.1. Branching and Divergence

Note: High Priority: Avoid different execution paths within the same warp.

Flow control instructions (if, switch, do, for, while) can significantly affect the instruction through-
put by causing threads of the same warp to diverge; that is, to follow different execution paths. If this
happens, the different execution paths must be executed separately; this increases the total number
of instructions executed for this warp.

To obtain best performance in cases where the control flow depends on the thread ID, the controlling
condition should be written so as to minimize the number of divergent warps.

This is possible because the distribution of the warps across the block is deterministic as mentioned
in SIMT Architecture of the CUDA C++ Programming Guide. A trivial example is when the controlling
condition depends only on (threadIdx / WSIZE) where WSIZE is the warp size.

In this case, no warp diverges because the controlling condition is perfectly aligned with the warps.

For branches including just a few instructions, warp divergence generally results in marginal perfor-
mance losses. For example, the compiler may use predication to avoid an actual branch. Instead, all
instructions are scheduled, but a per-thread condition code or predicate controls which threads ex-
ecute the instructions. Threads with a false predicate do not write results, and also do not evaluate
addresses or read operands.

Starting with the Volta architecture, Independent Thread Scheduling allows a warp to remain diverged
outside of the data-dependent conditional block. An explicit __syncwarp() can be used to guarantee
that the warp has reconverged for subsequent instructions.

16.2. Branch Predication

Note: Low Priority: Make it easy for the compiler to use branch predication in lieu of loops or control
statements.

Sometimes, the compiler may unroll loops or optimize out if or switch statements by using branch
predication instead. In these cases, no warp can ever diverge. The programmer can also control loop
unrolling using

75

CUDA C++ Best Practices Guide, Release 12.5

#pragma unroll

For more information on this pragma, refer to the CUDA C++ Programming Guide.

When using branch predication, none of the instructions whose execution depends on the controlling
condition is skipped. Instead, each such instruction is associated with a per-thread condition code or
predicate that is set to true or false according to the controlling condition. Although each of these
instructions is scheduled for execution, only the instructions with a true predicate are actually exe-
cuted. Instructions with a false predicate do not write results, and they also do not evaluate addresses
or read operands.

The compiler replaces a branch instruction with predicated instructions only if the number of instruc-
tions controlled by the branch condition is less than or equal to a certain threshold.

76

Chapter 16. Control Flow

Chapter 17. Deploying CUDA

Applications

Having completed the GPU acceleration of one or more components of the application it is possible
to compare the outcome with the original expectation. Recall that the initial assess step allowed the
developer to determine an upper bound for the potential speedup attainable by accelerating given
hotspots.

Before tackling other hotspots to improve the total speedup, the developer should consider taking
the partially parallelized implementation and carry it through to production. This is important for a
number of reasons; for example, it allows the user to profit from their investment as early as possible
(the speedup may be partial but is still valuable), and it minimizes risk for the developer and the user
by providing an evolutionary rather than revolutionary set of changes to the application.

77

CUDA C++ Best Practices Guide, Release 12.5

78

Chapter 17. Deploying CUDA Applications

Chapter 18. Understanding the

Programming Environment

With each generation of NVIDIA processors, new features are added to the GPU that CUDA can lever-
age. Consequently, it’s important to understand the characteristics of the architecture.

Programmers should be aware of two version numbers. The first is the compute capability, and the
second is the version number of the CUDA Runtime and CUDA Driver APIs.

18.1. CUDA Compute Capability

The compute capability describes the features of the hardware and reflects the set of instructions
supported by the device as well as other specifications, such as the maximum number of threads
per block and the number of registers per multiprocessor. Higher compute capability versions are
supersets of lower (that is, earlier) versions, so they are backward compatible.

The compute capability of the GPU in the device can be queried programmatically as illustrated in the
deviceQuery CUDA Sample. The output for that program is shown in Figure 16. This information is
obtained by calling cudaGetDeviceProperties() and accessing the information in the structure it
returns.

The major and minor revision numbers of the compute capability are shown on the seventh line of
Figure 16. Device 0 of this system has compute capability 7.0.

More details about the compute capabilities of various GPUs are in CUDA-Enabled GPUs and Compute
Capabilities of the CUDA C++ Programming Guide. In particular, developers should note the number
of multiprocessors on the device, the number of registers and the amount of memory available, and
any special capabilities of the device.

18.2. Additional Hardware Data

Certain hardware features are not described by the compute capability. For example, the ability to
overlap kernel execution with asynchronous data transfers between the host and the device is avail-
able on most but not all GPUs irrespective of the compute capability. In such cases, call cudaGetDe-
viceProperties() to determine whether the device is capable of a certain feature. For example, the
asyncEngineCount field of the device property structure indicates whether overlapping kernel exe-
cution and data transfers is possible (and, if so, how many concurrent transfers are possible); likewise,
the canMapHostMemory field indicates whether zero-copy data transfers can be performed.

79

CUDA C++ Best Practices Guide, Release 12.5

Fig. 1: Sample CUDA configuration data reported by deviceQuery

18.3. Which Compute Capability Target

To target specific versions of NVIDIA hardware and CUDA software, use the -arch, -code, and
-gencode options of nvcc. Code that uses the warp shuffle operation, for example, must be com-
piled with -arch=sm_30 (or higher compute capability).

See Building for Maximum Compatibility for further discussion of the flags used for building code for
multiple generations of CUDA-capable device simultaneously.

18.4. CUDA Runtime

The host runtime component of the CUDA software environment can be used only by host functions.
It provides functions to handle the following:

▶ Device management

▶ Context management

▶ Memory management

▶ Code module management
▶ Execution control

▶ Texture reference management

▶ Interoperability with OpenGL and Direct3D

80

Chapter 18. Understanding the Programming Environment

CUDA C++ Best Practices Guide, Release 12.5

As compared to the lower-level CUDA Driver API, the CUDA Runtime greatly eases device management
by providing implicit initialization, context management, and device code module management. The
C++ host code generated by nvcc utilizes the CUDA Runtime, so applications that link to this code
will depend on the CUDA Runtime; similarly, any code that uses the cuBLAS, cuFFT, and other CUDA
Toolkit libraries will also depend on the CUDA Runtime, which is used internally by these libraries.

The functions that make up the CUDA Runtime API are explained in the CUDA Toolkit Reference Man-
ual.

The CUDA Runtime handles kernel loading and setting up kernel parameters and launch configuration
before the kernel is launched. The implicit driver version checking, code initialization, CUDA context
management, CUDA module management (cubin to function mapping), kernel configuration, and pa-
rameter passing are all performed by the CUDA Runtime.

It comprises two principal parts:

▶ A C-style function interface (cuda_runtime_api.h).
▶ C++-style convenience wrappers (cuda_runtime.h) built on top of the C-style functions.

For more information on the Runtime API, refer to CUDA Runtime of the CUDA C++ Programming
Guide.

18.4. CUDA Runtime

81

CUDA C++ Best Practices Guide, Release 12.5

82

Chapter 18. Understanding the Programming Environment

Chapter 19. CUDA Compatibility

Developer’s Guide

CUDA Toolkit is released on a monthly release cadence to deliver new features, performance improve-
ments, and critical bug fixes. CUDA compatibility allows users to update the latest CUDA Toolkit soft-
ware (including the compiler, libraries, and tools) without requiring update to the entire driver stack.

The CUDA software environment consists of three parts:

▶ CUDA Toolkit (libraries, CUDA runtime and developer tools) - SDK for developers to build CUDA

applications.

▶ CUDA driver - User-mode driver component used to run CUDA applications (e.g.

libcuda.so on

Linux systems).

▶ NVIDIA GPU device driver - Kernel-mode driver component for NVIDIA GPUs.

On Linux systems, the CUDA driver and kernel mode components are delivered together in the NVIDIA
display driver package. This is shown in Figure 1.

The
CUDA
com-
piler
(nvcc),
pro-
vides
a
way
to
han-
dle
CUDA
and
non-
CUDA
code
(by
split-
ting
and
steer-
ing
com-
pi-

83

Fig. 1: Components of CUDA

la-
tion),
along
with
the
CUDA
run-
time,
is
part

CUDA C++ Best Practices Guide, Release 12.5

of the CUDA compiler toolchain. The CUDA Runtime API provides developers with high-level C++
interface for simplified management of devices, kernel executions etc., While the CUDA driver API
provides (CUDA Driver API) a low-level programming interface for applications to target NVIDIA
hardware.

Built on top of these technologies are CUDA libraries, some of which are included in the CUDA Toolkit,
while others such as cuDNN may be released independently of the CUDA Toolkit.

19.1. CUDA Toolkit Versioning

Starting with CUDA 11, the toolkit versions are based on an industry-standard semantic versioning
scheme: .X.Y.Z, where:

▶ .X stands for the major version - APIs have changed and binary compatibility is broken.

▶ .Y stands for the minor version - Introduction of new APIs, deprecation of old APIs, and source

compatibility might be broken but binary compatibility is maintained.

▶ .Z stands for the release/patch version - new updates and patches will increment this.

Each component in the toolkit is recommended to be semantically versioned. From CUDA 11.3 NVRTC
is also semantically versioned. We will note some of them later on in the document. The versions of
the components in the toolkit are available in this table.

Compatibility of the CUDA platform is thus intended to address a few scenarios:

1. NVIDIA driver upgrades to systems with GPUs running in production for enterprises or datacen-
ters can be complex and may need advance planning. Delays in rolling out new NVIDIA drivers
could mean that users of such systems may not have access to new features available in CUDA
releases. Not requiring driver updates for new CUDA releases can mean that new versions of the
software can be made available faster to users.

2. Many software libraries and applications built on top of CUDA (e.g. math libraries or deep learning
frameworks) do not have a direct dependency on the CUDA runtime, compiler or driver. In such
cases, users or developers can still benefit from not having to upgrade the entire CUDA Toolkit
or driver to use these libraries or frameworks.

3. Upgrading dependencies is error-prone and time consuming, and in some corner cases, can even
change the semantics of a program. Constantly recompiling with the latest CUDA Toolkit means
forcing upgrades on the end-customers of an application product. Package managers facilitate
this process but unexpected issues can still arise and if a bug is found, it necessitates a repeat
of the above upgrade process.

CUDA supports several compatibility choices:

84

Chapter 19. CUDA Compatibility Developer’s Guide

CUDA C++ Best Practices Guide, Release 12.5

1. First introduced in CUDA 10, the CUDA Forward Compatible Upgrade is designed to allow users
to get access to new CUDA features and run applications built with new CUDA releases on sys-
tems with older installations of the NVIDIA datacenter driver.

2. First introduced in CUDA 11.1, CUDA Enhanced Compatibility provides two benefits:

▶ By leveraging semantic versioning across components in the CUDA Toolkit, an application
can be built for one CUDA minor release (for example 11.1) and work across all future minor
releases within the major family (i.e. 11.x).

▶ The CUDA runtime has relaxed the minimum driver version check and thus no longer requires

a driver upgrade when moving to a new minor release.

3. The CUDA driver ensures backward Binary Compatibility is maintained for compiled CUDA appli-
cations. Applications compiled with CUDA toolkit versions as old as 3.2 will run on newer drivers.

19.2. Source Compatibility

We define source compatibility as a set of guarantees provided by the library, where a well-formed
application built against a specific version of the library (using the SDK) will continue to build and run
without errors when a newer version of the SDK is installed.

Both the CUDA driver and the CUDA runtime are not source compatible across the different SDK re-
leases. APIs can be deprecated and removed. Therefore, an application that compiled successfully on
an older version of the toolkit may require changes in order to compile against a newer version of the
toolkit.

Developers are notified through deprecation and documentation mechanisms of any current or up-
coming changes. This does not mean that application binaries compiled using an older toolkit will
not be supported anymore. Application binaries rely on CUDA Driver API interface and even though
the CUDA Driver API itself may also have changed across toolkit versions, CUDA guarantees Binary
Compatibility of the CUDA Driver API interface.

19.3. Binary Compatibility

We define binary compatibility as a set of guarantees provided by the library, where an application
targeting the said library will continue to work when dynamically linked against a different version of
the library.

The CUDA Driver API has a versioned C-style ABI, which guarantees that applications that were running
against an older driver (for example CUDA 3.2) will still run and function correctly against a modern
driver (for example one shipped with CUDA 11.0). This means that even though an application source
might need to be changed if it has to be recompiled against a newer CUDA Toolkit in order to use the
newer features, replacing the driver components installed in a system with a newer version will always
support existing applications and its functions.

The CUDA Driver API thus is binary-compatible (the OS loader can pick up a newer version and the
application continues to work) but not source-compatible (rebuilding your application against a newer
SDK might require source changes).

19.2. Source Compatibility

85

CUDA C++ Best Practices Guide, Release 12.5

Be-
fore
we
pro-
ceed
fur-
ther
on
this
topic,
it’s
im-
por-
tant
for
de-
vel-
op-
ers
to
un-
der-
stand
the
con-
cept
of
Min-
i-
mum
Driver
Ver-
sion
and
how
that

Fig. 2: CUDA Toolkit and Minimum Driver Versions

may affect them.

Each version of the CUDA Toolkit (and runtime) requires a minimum version of the NVIDIA driver. Appli-
cations compiled against a CUDA Toolkit version will only run on systems with the specified minimum
driver version for that toolkit version. Prior to CUDA 11.0, the minimum driver version for a toolkit was
the same as the driver shipped with that version of the CUDA Toolkit.

So, when an application is built with CUDA 11.0, it can only run on a system with an R450 or later driver.
If such an application is run on a system with the R418 driver installed, CUDA initialization will return
an error as can be seen in the example below.

In this example, the deviceQuery sample is compiled with CUDA 11.1 and is run on a system with R418.
In this scenario, CUDA initialization returns an error due to the minimum driver requirement.

ubuntu@:~∕samples∕1_Utilities∕deviceQuery
$ make
∕usr∕local∕cuda-11.1∕bin∕nvcc -ccbin g++ -I..∕..∕common∕inc
,→arch=compute_35,code=sm_35 -gencode arch=compute_37,code=sm_37 -gencode￿
,→arch=compute_50,code=sm_50 -gencode arch=compute_52,code=sm_52 -gencode￿
,→arch=compute_60,code=sm_60 -gencode arch=compute_61,code=sm_61 -gencode￿
,→arch=compute_70,code=sm_70 -gencode arch=compute_75,code=sm_75 -gencode￿
,→arch=compute_80,code=sm_80 -gencode arch=compute_86,code=sm_86 -gencode￿
,→arch=compute_86,code=compute_86 -o deviceQuery.o -c deviceQuery.cpp
86

-m64

-gencode￿

Chapter 19. CUDA Compatibility Developer’s Guide

(continues on next page)

CUDA C++ Best Practices Guide, Release 12.5

(continued from previous page)

∕usr∕local∕cuda-11.1∕bin∕nvcc -ccbin g++
-gencode arch=compute_35,code=sm_
,→35 -gencode arch=compute_37,code=sm_37 -gencode arch=compute_50,code=sm_50 -gencode￿
,→arch=compute_52,code=sm_52 -gencode arch=compute_60,code=sm_60 -gencode￿
,→arch=compute_61,code=sm_61 -gencode arch=compute_70,code=sm_70 -gencode￿
,→arch=compute_75,code=sm_75 -gencode arch=compute_80,code=sm_80 -gencode￿
,→arch=compute_86,code=sm_86 -gencode arch=compute_86,code=compute_86 -o deviceQuery￿
,→deviceQuery.o

-m64

$ nvidia-smi

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.165.02
|
Driver Version: 418.165.02
|-------------------------------+----------------------+----------------------+
| GPU
Disp.A | Volatile Uncorr. ECC |
| Fan
Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
0 |
0
|
| N∕A
Default |
+-------------------------------+----------------------+----------------------+

| 00000000:00:1E.0 Off |
0MiB ∕ 15079MiB |

Persistence-M| Bus-Id
Pwr:Usage∕Cap|

CUDA Version: 10.1

Name
Temp

Tesla T4

28W ∕

70W |

Perf

42C

0%

On

P0

GPU

+-----------------------------------------------------------------------------+
GPU Memory |
| Processes:
|
|
Usage
|=============================================================================|
|
|
+-----------------------------------------------------------------------------+

No running processes found

Process name

Type

PID

$ samples∕bin∕x86_64∕linux∕release∕deviceQuery
samples∕bin∕x86_64∕linux∕release∕deviceQuery Starting...

CUDA Device Query (Runtime API) version (CUDART static linking)

cudaGetDeviceCount returned 3
-> initialization error
Result = FAIL

Refer to the CUDA Toolkit Release Notes for details for the minimum driver version and the version of
the driver shipped with the toolkit.

19.3.1. CUDA Binary (cubin) Compatibility

A slightly related but important topic is one of application binary compatibility across GPU architec-
tures in CUDA.

CUDA C++ provides a simple path for users familiar with the C++ programming language to easily write
programs for execution by the device. Kernels can be written using the CUDA instruction set architec-
ture, called PTX, which is described in the PTX reference manual. It is however usually more effective
to use a high-level programming language such as C++. In both cases, kernels must be compiled into
binary code by nvcc (called cubins) to execute on the device.

The cubins are architecture-specific. Binary compatibility for cubins is guaranteed from one compute
capability minor revision to the next one, but not from one compute capability minor revision to the

19.3. Binary Compatibility

87

CUDA C++ Best Practices Guide, Release 12.5

previous one or across major compute capability revisions. In other words, a cubin object generated
for compute capability X.y will only execute on devices of compute capability X.z where z￿y.

To execute code on devices of specific compute capability, an application must load binary or PTX code
that is compatible with this compute capability. For portability, that is, to be able to execute code on
future GPU architectures with higher compute capability (for which no binary code can be generated
yet), an application must load PTX code that will be just-in-time compiled by the NVIDIA driver for
these future devices.

More information on cubins, PTX and application compatibility can be found in the CUDA C++ Pro-
gramming Guide.

19.4. CUDA Compatibility Across Minor Releases

By leveraging the semantic versioning, starting with CUDA 11, components in the CUDA Toolkit will
remain binary compatible across the minor versions of the toolkit. In order to maintain binary com-
patibility across minor versions, the CUDA runtime no longer bumps up the minimum driver version
required for every minor release - this only happens when a major release is shipped.

One of the main reasons a new toolchain requires a new minimum driver is to handle the JIT compilation
of PTX code and the JIT linking of binary code.

In this section, we will review the usage patterns that may require new user workflows when taking
advantage of the compatibility features of the CUDA platform.

19.4.1. Existing CUDA Applications within Minor Versions

of CUDA

$ nvidia-smi

Name
Temp

Driver Version: 450.80.02

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 450.80.02
|
|-------------------------------+----------------------+----------------------+
Disp.A | Volatile Uncorr. ECC |
| GPU
| Fan
Compute M. |
MIG M. |
|
|===============================+======================+======================|
|
0 |
0
Default |
| N∕A
N∕A |
|
+-------------------------------+----------------------+----------------------+

| 00000000:00:1E.0 Off |
0MiB ∕ 15109MiB |
|

Persistence-M| Bus-Id
Pwr:Usage∕Cap|
|

Memory-Usage | GPU-Util

CUDA Version: 11.0

70W |
|

Tesla T4

9W ∕

Perf

39C

0%

On

P8

|

GPU

+-----------------------------------------------------------------------------+
| Processes:
|
GPU Memory |
|
|
|
Usage
|=============================================================================|
|
|
+-----------------------------------------------------------------------------+

No running processes found

Process name

CI
ID

GI
ID

Type

PID

88

Chapter 19. CUDA Compatibility Developer’s Guide

CUDA C++ Best Practices Guide, Release 12.5

When our CUDA 11.1 application (i.e. cudart 11.1 is statically linked) is run on the system, we see that
it runs successfully even when the driver reports a 11.0 version - that is, without requiring the driver
or other toolkit components to be updated on the system.

$ samples∕bin∕x86_64∕linux∕release∕deviceQuery
samples∕bin∕x86_64∕linux∕release∕deviceQuery Starting...

CUDA Device Query (Runtime API) version (CUDART static linking)

Detected 1 CUDA Capable device(s)

Device 0: "Tesla T4"

CUDA Driver Version ∕ Runtime Version
CUDA Capability Major∕Minor version number:

11.0 ∕ 11.1
7.5

...<snip>...

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 11.0, CUDA Runtime Version =￿
,→11.1, NumDevs = 1
Result = PASS

By using new CUDA versions, users can benefit from new CUDA programming model APIs, compiler
optimizations and math library features.

The following sections discuss some caveats and considerations.

19.4.1.1 Handling New CUDA Features and Driver APIs

A subset of CUDA APIs don’t need a new driver and they can all be used without any driver depen-
dencies. For example, cuMemMap APIs or any of APIs introduced prior to CUDA 11.0, such as cudaDe-
viceSynchronize, do not require a driver upgrade. To use other CUDA APIs introduced in a minor
release (that require a new driver), one would have to implement fallbacks or fail gracefully. This situa-
tion is not different from what is available today where developers use macros to compile out features
based on CUDA versions. Users should refer to the CUDA headers and documentation for new CUDA
APIs introduced in a release.

When working with a feature exposed in a minor version of the toolkit, the feature might not be avail-
able at runtime if the application is running against an older CUDA driver. Users wishing to take ad-
vantage of such a feature should query its availability with a dynamic check in the code:

static bool hostRegisterFeatureSupported = false;
static bool hostRegisterIsDeviceAddress = false;

static error_t cuFooFunction(int *ptr)
{

int *dptr = null;
if (hostRegisterFeatureSupported) {

cudaHostRegister(ptr, size, flags);
if (hostRegisterIsDeviceAddress) {

qptr = ptr;

}
else {

cudaHostGetDevicePointer(&qptr, ptr, 0);
}

}
else {

19.4. CUDA Compatibility Across Minor Releases

89

(continues on next page)

(continued from previous page)

CUDA C++ Best Practices Guide, Release 12.5

∕∕ cudaMalloc();
∕∕ cudaMemcpy();

}

gemm<<<1,1>>>(dptr);
cudaDeviceSynchronize();

}

int main()
{

∕∕ rest of code here
cudaDeviceGetAttribute(

&hostRegisterFeatureSupported,
cudaDevAttrHostRegisterSupported,
0);

cudaDeviceGetAttribute(

&hostRegisterIsDeviceAddress,
cudaDevAttrCanUseHostPointerForRegisteredMem,
0);

cuFooFunction(∕* malloced pointer *∕);

}

Alternatively the application’s interface might not work at all without a new CUDA driver and then its
best to return an error right away:

#define MIN_VERSION 11010
cudaError_t foo()
{

int version = 0;
cudaGetDriverVersion(&version);
if (version < MIN_VERSION) {

return CUDA_ERROR_INSUFFICIENT_DRIVER;

}
∕∕ proceed as normal

}

A new error code is added to indicate that the functionality is missing from the driver you are running
against: cudaErrorCallRequiresNewerDriver.

19.4.1.2 Using PTX

PTX defines a virtual machine and ISA for general purpose parallel thread execution. PTX programs
are translated at load time to the target hardware instruction set via the JIT Compiler which is part
of the CUDA driver. As PTX is compiled by the CUDA driver, new toolchains will generate PTX that is
not compatible with the older CUDA driver. This is not a problem when PTX is used for future device
compatibility (the most common case), but can lead to issues when used for runtime compilation.

For codes continuing to make use of PTX, in order to support compiling on an older driver, your code
must be first transformed into device code via the static ptxjitcompiler library or NVRTC with the
option of generating code for a specific architecture (e.g. sm_80) rather than a virtual architecture
(e.g. compute_80). For this workflow, a new nvptxcompiler_static library is shipped with the CUDA
Toolkit.

We can see this usage in the following example:

90

Chapter 19. CUDA Compatibility Developer’s Guide

CUDA C++ Best Practices Guide, Release 12.5

char* compilePTXToNVElf()
{

nvPTXCompilerHandle compiler = NULL;
nvPTXCompileResult status;

size_t elfSize, infoSize, errorSize;
char *elf, *infoLog, *errorLog;
int minorVer, majorVer;

const char* compile_options[] = { "--gpu-name=sm_80",

"--device-debug"

};

nvPTXCompilerGetVersion(&majorVer, &minorVer);
nvPTXCompilerCreate(&compiler, (size_t)strlen(ptxCode), ptxCode);
status = nvPTXCompilerCompile(compiler, 2, compile_options);
if (status != NVPTXCOMPILE_SUCCESS) {

nvPTXCompilerGetErrorLogSize(compiler, (void*)&errorSize);

if (errorSize != 0) {

errorLog = (char*)malloc(errorSize+1);
nvPTXCompilerGetErrorLog(compiler, (void*)errorLog);
printf("Error log: %s\n", errorLog);
free(errorLog);

}
exit(1);

}

nvPTXCompilerGetCompiledProgramSize(compiler, &elfSize));
elf = (char*)malloc(elfSize);
nvPTXCompilerGetCompiledProgram(compiler, (void*)elf);
nvPTXCompilerGetInfoLogSize(compiler, (void*)&infoSize);

if (infoSize != 0) {

infoLog = (char*)malloc(infoSize+1);
nvPTXCompilerGetInfoLog(compiler, (void*)infoLog);
printf("Info log: %s\n", infoLog);
free(infoLog);

}

nvPTXCompilerDestroy(&compiler);
return elf;

}

19.4.1.3 Dynamic Code Generation

NVRTC is a runtime compilation library for CUDA C++. It accepts CUDA C++ source code in character
string form and creates handles that can be used to obtain the PTX. The PTX string generated by
NVRTC can be loaded by cuModuleLoadData and cuModuleLoadDataEx.

Dealing with relocatable objects is not yet supported, therefore the cuLink* set of APIs in the CUDA
driver will not work with enhanced compatibility. An upgraded driver matching the CUDA runtime
version is currently required for those APIs.

As mentioned in the PTX section, the compilation of PTX to device code lives along with the CUDA
driver, hence the generated PTX might be newer than what is supported by the driver on the deploy-
ment system. When using NVRTC, it is recommended that the resulting PTX code is first transformed

19.4. CUDA Compatibility Across Minor Releases

91

CUDA C++ Best Practices Guide, Release 12.5

to the final device code via the steps outlined by the PTX user workflow. This ensures your code is
compatible. Alternatively, NVRTC can generate cubins directly starting with CUDA 11.1. Applications
using the new API can load the final device code directly using driver APIs cuModuleLoadData and
cuModuleLoadDataEx.

NVRTC used to support only virtual architectures through the option -arch, since it was only emit-
ting PTX. It will now support actual architectures as well to emit SASS. The interface is augmented to
retrieve either the PTX or cubin if an actual architecture is specified.

The example below shows how an existing example can be adapted to use the new features, guarded
by the USE_CUBIN macro in this case:

#include <nvrtc.h>
#include <cuda.h>
#include <iostream>

void NVRTC_SAFE_CALL(nvrtcResult result) {

if (result != NVRTC_SUCCESS) {

std::cerr << "\nnvrtc error: " << nvrtcGetErrorString(result) << '\n';
std::exit(1);

}

}

void CUDA_SAFE_CALL(CUresult result) {

if (result != CUDA_SUCCESS) {

const char *msg;
cuGetErrorName(result, &msg);
std::cerr << "\ncuda error: " << msg << '\n';
std::exit(1);

}

}

const char *hello = "
extern \"C\" __global__ void hello() {

printf(\"hello world\\n\");

}

int main()
{

\n\
\n\
\n\
\n";

nvrtcProgram prog;
NVRTC_SAFE_CALL(nvrtcCreateProgram(&prog, hello, "hello.cu", 0, NULL, NULL));

#ifdef USE_CUBIN

const char *opts[] = {"-arch=sm_70"};

#else

const char *opts[] = {"-arch=compute_70"};

#endif

nvrtcResult compileResult = nvrtcCompileProgram(prog, 1, opts);
size_t logSize;
NVRTC_SAFE_CALL(nvrtcGetProgramLogSize(prog, &logSize));
char *log = new char[logSize];
NVRTC_SAFE_CALL(nvrtcGetProgramLog(prog, log));
std::cout << log << '\n';
delete[] log;
if (compileResult != NVRTC_SUCCESS)

exit(1);

size_t codeSize;

#ifdef USE_CUBIN

NVRTC_SAFE_CALL(nvrtcGetCUBINSize(prog, &codeSize));

(continues on next page)

92

Chapter 19. CUDA Compatibility Developer’s Guide

CUDA C++ Best Practices Guide, Release 12.5

(continued from previous page)

char *code = new char[codeSize];
NVRTC_SAFE_CALL(nvrtcGetCUBIN(prog, code));

#else

NVRTC_SAFE_CALL(nvrtcGetPTXSize(prog, &codeSize));
char *code = new char[codeSize];
NVRTC_SAFE_CALL(nvrtcGetPTX(prog, code));

#endif

NVRTC_SAFE_CALL(nvrtcDestroyProgram(&prog));
CUdevice cuDevice;
CUcontext context;
CUmodule module;
CUfunction kernel;
CUDA_SAFE_CALL(cuInit(0));
CUDA_SAFE_CALL(cuDeviceGet(&cuDevice, 0));
CUDA_SAFE_CALL(cuCtxCreate(&context, 0, cuDevice));
CUDA_SAFE_CALL(cuModuleLoadDataEx(&module, code, 0, 0, 0));
CUDA_SAFE_CALL(cuModuleGetFunction(&kernel, module, "hello"));
CUDA_SAFE_CALL(cuLaunchKernel(kernel, 1, 1, 1, 1, 1, 1, 0, NULL, NULL, 0));
CUDA_SAFE_CALL(cuCtxSynchronize());
CUDA_SAFE_CALL(cuModuleUnload(module));
CUDA_SAFE_CALL(cuCtxDestroy(context));
delete[] code;

}

19.4.1.4 Recommendations for building a minor-version compatible library

We recommend that the CUDA runtime be statically linked to minimize dependencies. Verify that your
library doesn’t leak dependencies, breakages, namespaces, etc. outside your established ABI contract.

Follow semantic versioning for your library’s soname. Having a semantically versioned ABI means the
interfaces need to be maintained and versioned. The library should follow semantic rules and incre-
ment the version number when a change is made that affects this ABI contract. Missing dependencies
is also a binary compatibility break, hence you should provide fallbacks or guards for functionality that
depends on those interfaces. Increment major versions when there are ABI breaking changes such as
API deprecation and modifications. New APIs can be added in minor versions.

Conditionally use features to remain compatible against older drivers. If no new features are used (or
if they are used conditionally with fallbacks provided) you’ll be able to remain compatible.

Don’t expose ABI structures that can change. A pointer to a structure with a size embedded is a better
solution.

When linking with dynamic libraries from the toolkit, the library must be equal to or newer than what
is needed by any one of the components involved in the linking of your application. For example, if you
link against the CUDA 11.1 dynamic runtime, and use functionality from 11.1, as well as a separate
shared library that was linked against the CUDA 11.2 dynamic runtime that requires 11.2 functionality,
the final link step must include a CUDA 11.2 or newer dynamic runtime.

19.4. CUDA Compatibility Across Minor Releases

93

CUDA C++ Best Practices Guide, Release 12.5

19.4.1.5 Recommendations for taking advantage of minor version compatibility in your application

Certain functionality might not be available so you should query where applicable. This is common for
building applications that are GPU architecture, platform and compiler agnostic. However we now add
“the underlying driver” to that mix.

As with the previous section on library building recommendations, if using the CUDA runtime, we rec-
ommend linking to the CUDA runtime statically when building your application. When using the driver
APIs directly, we recommend using the new driver entry point access API (cuGetProcAddress) doc-
umented here: CUDA Driver API :: CUDA Toolkit Documentation.

When using a shared or static library, follow the release notes of said library to determine if the library
supports minor version compatibility.

94

Chapter 19. CUDA Compatibility Developer’s Guide

Chapter 20. Preparing for Deployment

20.1. Testing for CUDA Availability

When deploying a CUDA application, it is often desirable to ensure that the application will continue
to function properly even if the target machine does not have a CUDA-capable GPU and/or a sufficient
version of the NVIDIA Driver installed. (Developers targeting a single machine with known configura-
tion may choose to skip this section.)

Detecting a CUDA-Capable GPU

When an application will be deployed to target machines of arbitrary/unknown configuration, the ap-
plication should explicitly test for the existence of a CUDA-capable GPU in order to take appropri-
ate action when no such device is available. The cudaGetDeviceCount() function can be used to
query for the number of available devices. Like all CUDA Runtime API functions, this function will fail
gracefully and return cudaErrorNoDevice to the application if there is no CUDA-capable GPU or cu-
daErrorInsufficientDriver if there is not an appropriate version of the NVIDIA Driver installed.
If cudaGetDeviceCount() reports an error, the application should fall back to an alternative code
path.

A system with multiple GPUs may contain GPUs of different hardware versions and capabilities. When
using multiple GPUs from the same application, it is recommended to use GPUs of the same type,
rather than mixing hardware generations. The cudaChooseDevice() function can be used to select
the device that most closely matches a desired set of features.

Detecting Hardware and Software Configuration

When an application depends on the availability of certain hardware or software capabilities to enable
certain functionality, the CUDA API can be queried for details about the configuration of the available
device and for the installed software versions.

The cudaGetDeviceProperties() function reports various features of the available devices, includ-
ing the CUDA Compute Capability of the device (see also the Compute Capabilities section of the CUDA
C++ Programming Guide). See Version Management for details on how to query the available CUDA
software API versions.

95

CUDA C++ Best Practices Guide, Release 12.5

20.2. Error Handling

All CUDA Runtime API calls return an error code of type cudaError_t; the return value will be equal
to cudaSuccess if no errors have occurred. (The exceptions to this are kernel launches, which return
void, and cudaGetErrorString(), which returns a character string describing the cudaError_t
code that was passed into it.) The CUDA Toolkit libraries (cuBLAS, cuFFT, etc.) likewise return their
own sets of error codes.

Since some CUDA API calls and all kernel launches are asynchronous with respect to the host code, er-
rors may be reported to the host asynchronously as well; often this occurs the next time the host and
device synchronize with each other, such as during a call to cudaMemcpy() or to cudaDeviceSyn-
chronize().

Always check the error return values on all CUDA API functions, even for functions that are not ex-
pected to fail, as this will allow the application to detect and recover from errors as soon as possible
should they occur. To check for errors occurring during kernel launches using the <<<...>>> syntax,
which does not return any error code, the return code of cudaGetLastError() should be checked
immediately after the kernel launch. Applications that do not check for CUDA API errors could at times
run to completion without having noticed that the data calculated by the GPU is incomplete, invalid,
or uninitialized.

Note: The CUDA Toolkit Samples provide several helper functions for error checking with the various
CUDA APIs; these helper functions are located in the samples∕common∕inc∕helper_cuda.h file in
the CUDA Toolkit.

20.3. Building for Maximum Compatibility

Each generation of CUDA-capable device has an associated compute capability version that indicates
the feature set supported by the device (see CUDA Compute Capability). One or more compute ca-
pability versions can be specified to the nvcc compiler while building a file; compiling for the native
compute capability for the target GPU(s) of the application is important to ensure that application
kernels achieve the best possible performance and are able to use the features that are available on a
given generation of GPU.

When an application is built for multiple compute capabilities simultaneously (using several instances
of the -gencode flag to nvcc), the binaries for the specified compute capabilities are combined into
the executable, and the CUDA Driver selects the most appropriate binary at runtime according to the
compute capability of the present device. If an appropriate native binary (cubin) is not available, but
the intermediate PTX code (which targets an abstract virtual instruction set and is used for forward-
compatibility) is available, then the kernel will be compiled Just In Time (JIT) (see Compiler JIT Cache
Management Tools) from the PTX to the native cubin for the device. If the PTX is also not available,
then the kernel launch will fail.

Windows

nvcc.exe -ccbin "C:\vs2008\VC\bin"

-Xcompiler "∕EHsc ∕W3 ∕nologo ∕O2 ∕Zi ∕MT"
-gencode=arch=compute_30,code=sm_30
-gencode=arch=compute_35,code=sm_35
-gencode=arch=compute_50,code=sm_50

96

Chapter 20. Preparing for Deployment

(continues on next page)

CUDA C++ Best Practices Guide, Release 12.5

(continued from previous page)

-gencode=arch=compute_60,code=sm_60
-gencode=arch=compute_70,code=sm_70
-gencode=arch=compute_75,code=sm_75
-gencode=arch=compute_75,code=compute_75
--compile -o "Release\mykernel.cu.obj" "mykernel.cu"

Mac/Linux

∕usr∕local∕cuda∕bin∕nvcc

-gencode=arch=compute_30,code=sm_30
-gencode=arch=compute_35,code=sm_35
-gencode=arch=compute_50,code=sm_50
-gencode=arch=compute_60,code=sm_60
-gencode=arch=compute_70,code=sm_70
-gencode=arch=compute_75,code=sm_75
-gencode=arch=compute_75,code=compute_75
-O2 -o mykernel.o -c mykernel.cu

Alternatively, the nvcc command-line option -arch=sm_XX can be used as a shorthand equivalent to
the following more explicit -gencode= command-line options described above:

-gencode=arch=compute_XX,code=sm_XX
-gencode=arch=compute_XX,code=compute_XX

However, while the -arch=sm_XX command-line option does result in inclusion of a PTX back-end
target by default (due to the code=compute_XX target it implies), it can only specify a single target
cubin architecture at a time, and it is not possible to use multiple -arch= options on the same nvcc
command line, which is why the examples above use -gencode= explicitly.

20.4. Distributing the CUDA Runtime and

Libraries

CUDA applications are built against the CUDA Runtime library, which handles device, memory, and ker-
nel management. Unlike the CUDA Driver, the CUDA Runtime guarantees neither forward nor back-
ward binary compatibility across versions. It is therefore best to redistribute the CUDA Runtime library
with the application when using dynamic linking or else to statically link against the CUDA Runtime.
This will ensure that the executable will be able to run even if the user does not have the same CUDA
Toolkit installed that the application was built against.

Note: When statically linking to the CUDA Runtime, multiple versions of the runtime can peacably
coexist in the same application process simultaneously; for example, if an application uses one version
of the CUDA Runtime, and a plugin to that application is statically linked to a different version, that is
perfectly acceptable, as long as the installed NVIDIA Driver is sufficient for both.

20.4. Distributing the CUDA Runtime and Libraries

97

CUDA C++ Best Practices Guide, Release 12.5

Statically-linked CUDA Runtime

The easiest option is to statically link against the CUDA Runtime. This is the default if using nvcc to
link in CUDA 5.5 and later. Static linking makes the executable slightly larger, but it ensures that the
correct version of runtime library functions are included in the application binary without requiring
separate redistribution of the CUDA Runtime library.

Dynamically-linked CUDA Runtime

If static linking against the CUDA Runtime is impractical for some reason, then a dynamically-linked
version of the CUDA Runtime library is also available. (This was the default and only option provided in
CUDA versions 5.0 and earlier.)

To use dynamic linking with the CUDA Runtime when using the nvcc from CUDA 5.5 or later to link the
application, add the --cudart=shared flag to the link command line; otherwise the statically-linked
CUDA Runtime library is used by default.

After the application is dynamically linked against the CUDA Runtime, this version of the runtime library
should be bundled with the application. It can be copied into the same directory as the application
executable or into a subdirectory of that installation path.

Other CUDA Libraries

Although the CUDA Runtime provides the option of static linking, some libraries included in the CUDA
Toolkit are available only in dynamically-linked form. As with the dynamically-linked version of the
CUDA Runtime library, these libraries should be bundled with the application executable when dis-
tributing that application.

20.4.1. CUDA Toolkit Library Redistribution

The CUDA Toolkit’s End-User License Agreement (EULA) allows for redistribution of many of the CUDA
libraries under certain terms and conditions. This allows applications that depend on these libraries
to redistribute the exact versions of the libraries against which they were built and tested, thereby
avoiding any trouble for end users who might have a different version of the CUDA Toolkit (or perhaps
none at all) installed on their machines. Please refer to the EULA for details.

Note: This does not apply to the NVIDIA Driver; the end user must still download and install an NVIDIA
Driver appropriate to their GPU(s) and operating system.

98

Chapter 20. Preparing for Deployment

CUDA C++ Best Practices Guide, Release 12.5

20.4.1.1 Which Files to Redistribute

When redistributing the dynamically-linked versions of one or more CUDA libraries, it is important to
identify the exact files that need to be redistributed. The following examples use the cuBLAS library
from CUDA Toolkit 5.5 as an illustration:

Linux

In a shared library on Linux, there is a string field called the SONAME that indicates the binary compati-
bility level of the library. The SONAME of the library against which the application was built must match
the filename of the library that is redistributed with the application.

For example, in the standard CUDA Toolkit installation, the files libcublas.so and libcublas.so.
5.5 are both symlinks pointing to a specific build of cuBLAS, which is named like libcublas.so.5.
5.x, where x is the build number (e.g., libcublas.so.5.5.17). However, the SONAME of this library
is given as “libcublas.so.5.5”:

$ objdump -p ∕usr∕local∕cuda∕lib64∕libcublas.so | grep SONAME

SONAME

libcublas.so.5.5

Because of this, even if -lcublas (with no version number specified) is used when linking the appli-
cation, the SONAME found at link time implies that “libcublas.so.5.5” is the name of the file that
the dynamic loader will look for when loading the application and therefore must be the name of the
file (or a symlink to the same) that is redistributed with the application.

The ldd tool is useful for identifying the exact filenames of the libraries that the application expects
to find at runtime as well as the path, if any, of the copy of that library that the dynamic loader would
select when loading the application given the current library search path:

$ ldd a.out | grep libcublas

libcublas.so.5.5 => ∕usr∕local∕cuda∕lib64∕libcublas.so.5.5

Mac

In a shared library on Mac OS X, there is a field called the install name that indicates the expected
installation path and filename the library; the CUDA libraries also use this filename to indicate binary
compatibility. The value of this field is propagated into an application built against the library and is
used to locate the library of the correct version at runtime.

For example, if the install name of the cuBLAS library is given as @rpath∕libcublas.5.5.dylib,
then the library is version 5.5 and the copy of this library redistributed with the application must be
named libcublas.5.5.dylib, even though only -lcublas (with no version number specified) is
used at link time. Furthermore, this file should be installed into the @rpath of the application; see
Where to Install Redistributed CUDA Libraries.

To view a library’s install name, use the otool -L command:

$ otool -L a.out
a.out:

@rpath∕libcublas.5.5.dylib (...)

Windows

The binary compatibility version of the CUDA libraries on Windows is indicated as part of the filename.

For example, a 64-bit application linked to cuBLAS 5.5 will look for cublas64_55.dll at runtime, so
this is the file that should be redistributed with that application, even though cublas.lib is the file
that the application is linked against. For 32-bit applications, the file would be cublas32_55.dll.

20.4. Distributing the CUDA Runtime and Libraries

99

CUDA C++ Best Practices Guide, Release 12.5

To verify the exact DLL filename that the application expects to find at runtime, use the dumpbin tool
from the Visual Studio command prompt:

$ dumpbin ∕IMPORTS a.exe
Microsoft (R) COFF∕PE Dumper Version 10.00.40219.01
Copyright (C) Microsoft Corporation.

All rights reserved.

Dump of file a.exe

File Type: EXECUTABLE IMAGE

Section contains the following imports:

...
cublas64_55.dll
...

20.4.1.2 Where to Install Redistributed CUDA Libraries

Once the correct library files are identified for redistribution, they must be configured for installation
into a location where the application will be able to find them.

On Windows, if the CUDA Runtime or other dynamically-linked CUDA Toolkit library is placed in the
same directory as the executable, Windows will locate it automatically. On Linux and Mac, the -rpath
linker option should be used to instruct the executable to search its local path for these libraries before
searching the system paths:

Linux/Mac

nvcc -I $(CUDA_HOME)∕include

-Xlinker "-rpath '$ORIGIN'" --cudart=shared
-o myprogram myprogram.cu

Windows

nvcc.exe -ccbin "C:\vs2008\VC\bin"

-Xcompiler "∕EHsc ∕W3 ∕nologo ∕O2 ∕Zi ∕MT" --cudart=shared
-o "Release\myprogram.exe" "myprogram.cu"

It may be necessary to adjust the value of -ccbin to reflect the location of your Visual Studio

Note:
installation.

To specify an alternate path where the libraries will be distributed, use linker options similar to those
below:

Linux/Mac

nvcc -I $(CUDA_HOME)∕include

-Xlinker "-rpath '$ORIGIN∕lib'" --cudart=shared
-o myprogram myprogram.cu

Windows

100

Chapter 20. Preparing for Deployment

CUDA C++ Best Practices Guide, Release 12.5

nvcc.exe -ccbin "C:\vs2008\VC\bin"

-Xcompiler "∕EHsc ∕W3 ∕nologo ∕O2 ∕Zi ∕MT ∕DELAY" --cudart=shared
-o "Release\myprogram.exe" "myprogram.cu"

For Linux and Mac, the -rpath option is used as before. For Windows, the ∕DELAY option is used; this
requires that the application call SetDllDirectory() before the first call to any CUDA API function
in order to specify the directory containing the CUDA DLLs.

Note: For Windows 8, SetDefaultDLLDirectories() and AddDllDirectory() should be used
instead of SetDllDirectory(). Please see the MSDN documentation for these routines for more
information.

20.4. Distributing the CUDA Runtime and Libraries

101

CUDA C++ Best Practices Guide, Release 12.5

102

Chapter 20. Preparing for Deployment

Chapter 21. Deployment Infrastructure

Tools

21.1. Nvidia-SMI

The NVIDIA System Management Interface (nvidia-smi) is a command line utility that aids in the
management and monitoring of NVIDIA GPU devices. This utility allows administrators to query GPU
device state and, with the appropriate privileges, permits administrators to modify GPU device state.
nvidia-smi is targeted at Tesla and certain Quadro GPUs, though limited support is also available
on other NVIDIA GPUs. nvidia-smi ships with NVIDIA GPU display drivers on Linux, and with 64-bit
Windows Server 2008 R2 and Windows 7. nvidia-smi can output queried information as XML or as
human-readable plain text either to standard output or to a file. See the nvidia-smi documenation for
details. Please note that new versions of nvidia-smi are not guaranteed to be backward-compatible
with previous versions.

21.1.1. Queryable state

ECC error counts

Both correctable single-bit and detectable double-bit errors are reported. Error counts are pro-
vided for both the current boot cycle and the lifetime of the GPU.

GPU utilization

Current utilization rates are reported for both the compute resources of the GPU and the memory
interface.

Active compute process

The list of active processes running on the GPU is reported, along with the corresponding process
name/ID and allocated GPU memory.

Clocks and performance state

Max and current clock rates are reported for several important clock domains, as well as the
current GPU performance state (pstate).

Temperature and fan speed

The current GPU core temperature is reported, along with fan speeds for products with active
cooling.

Power management

The current board power draw and power limits are reported for products that report these mea-
surements.

103

CUDA C++ Best Practices Guide, Release 12.5

Identification

Various dynamic and static information is reported, including board serial numbers, PCI device
IDs, VBIOS/Inforom version numbers and product names.

21.1.2. Modifiable state

ECC mode

Enable and disable ECC reporting.

ECC reset

Clear single-bit and double-bit ECC error counts.

Compute mode

Indicate whether compute processes can run on the GPU and whether they run exclusively or
concurrently with other compute processes.

Persistence mode

Indicate whether the NVIDIA driver stays loaded when no applications are connected to the GPU.
It is best to enable this option in most circumstances.

GPU reset

Reinitialize the GPU hardware and software state via a secondary bus reset.

21.2. NVML

The NVIDIA Management Library (NVML) is a C-based interface that provides direct access to the
queries and commands exposed via nvidia-smi intended as a platform for building 3rd-party system
management applications. The NVML API is shipped with the CUDA Toolkit (since version 8.0) and is
also available standalone on the NVIDIA developer website as part of the GPU Deployment Kit through
a single header file accompanied by PDF documentation, stub libraries, and sample applications; see
https://developer.nvidia.com/gpu-deployment-kit. Each new version of NVML is backward-compatible.

An additional set of Perl and Python bindings are provided for the NVML API. These bindings expose the
same features as the C-based interface and also provide backwards compatibility. The Perl bindings
are provided via CPAN and the Python bindings via PyPI.

All of these products (nvidia-smi, NVML, and the NVML language bindings) are updated with each
new CUDA release and provide roughly the same functionality.

See https://developer.nvidia.com/nvidia-management-library-nvml for additional information.

21.3. Cluster Management Tools

Managing your GPU cluster will help achieve maximum GPU utilization and help you and your users
extract the best possible performance. Many of the industry’s most popular cluster management
tools support CUDA GPUs via NVML. For a listing of some of these tools, see https://developer.nvidia.
com/cluster-management.

104

Chapter 21. Deployment Infrastructure Tools

CUDA C++ Best Practices Guide, Release 12.5

21.4. Compiler JIT Cache Management Tools

Any PTX device code loaded by an application at runtime is compiled further to binary code by the
device driver. This is called just-in-time compilation (JIT). Just-in-time compilation increases application
load time but allows applications to benefit from latest compiler improvements. It is also the only way
for applications to run on devices that did not exist at the time the application was compiled.

When JIT compilation of PTX device code is used, the NVIDIA driver caches the resulting binary code on
disk. Some aspects of this behavior such as cache location and maximum cache size can be controlled
via the use of environment variables; see Just in Time Compilation of the CUDA C++ Programming
Guide.

21.5. CUDA_VISIBLE_DEVICES

It is possible to rearrange the collection of installed CUDA devices that will be visible to and enumerated
by a CUDA application prior to the start of that application by way of the CUDA_VISIBLE_DEVICES
environment variable.

Devices to be made visible to the application should be included as a comma-separated list in terms
of the system-wide list of enumerable devices. For example, to use only devices 0 and 2 from the
system-wide list of devices, set CUDA_VISIBLE_DEVICES=0,2 before launching the application. The
application will then enumerate these devices as device 0 and device 1, respectively.

21.4. Compiler JIT Cache Management Tools

105

CUDA C++ Best Practices Guide, Release 12.5

106

Chapter 21. Deployment Infrastructure Tools

Chapter 22. Recommendations and

Best Practices

This chapter contains a summary of the recommendations for optimization that are explained in this
document.

22.1. Overall Performance Optimization

Strategies

Performance optimization revolves around three basic strategies:

▶ Maximizing parallel execution

▶ Optimizing memory usage to achieve maximum memory bandwidth
▶ Optimizing instruction usage to achieve maximum instruction throughput

Maximizing parallel execution starts with structuring the algorithm in a way that exposes as much par-
allelism as possible. Once the parallelism of the algorithm has been exposed, it needs to be mapped
to the hardware as efficiently as possible. This is done by carefully choosing the execution configu-
ration of each kernel launch. The application should also maximize parallel execution at a higher level
by explicitly exposing concurrent execution on the device through streams, as well as maximizing con-
current execution between the host and the device.

Optimizing memory usage starts with minimizing data transfers between the host and the device
because those transfers have much lower bandwidth than internal device data transfers. Kernel access
to global memory also should be minimized by maximizing the use of shared memory on the device.
Sometimes, the best optimization might even be to avoid any data transfer in the first place by simply
recomputing the data whenever it is needed.

The effective bandwidth can vary by an order of magnitude depending on the access pattern for each
type of memory. The next step in optimizing memory usage is therefore to organize memory accesses
according to the optimal memory access patterns. This optimization is especially important for global
memory accesses, because latency of access costs hundreds of clock cycles. Shared memory ac-
cesses, in counterpoint, are usually worth optimizing only when there exists a high degree of bank
conflicts.

As for optimizing instruction usage, the use of arithmetic instructions that have low throughput should
be avoided. This suggests trading precision for speed when it does not affect the end result, such as
using intrinsics instead of regular functions or single precision instead of double precision. Finally, par-

107

CUDA C++ Best Practices Guide, Release 12.5

ticular attention must be paid to control flow instructions due to the SIMT (single instruction multiple
thread) nature of the device.

108

Chapter 22. Recommendations and Best Practices

Chapter 23. nvcc Compiler Switches

23.1. nvcc

The NVIDIA nvcc compiler driver converts .cu files into C++ for the host system and CUDA assembly
or binary instructions for the device. It supports a number of command-line parameters, of which the
following are especially useful for optimization and related best practices:

▶ -maxrregcount=N specifies the maximum number of registers kernels can use at a per-file level.
See Register Pressure. (See also the__launch_bounds__ qualifier discussed in Execution Con-
figuration of the CUDA C++ Programming Guide to control the number of registers used on a
per-kernel basis.)

▶ --ptxas-options=-v or -Xptxas=-v lists per-kernel register, shared, and constant memory

usage.

▶ -ftz=true (denormalized numbers are flushed to zero)
▶ -prec-div=false (less precise division)
▶ -prec-sqrt=false (less precise square root)
▶ -use_fast_math compiler option of nvcc coerces every functionName() call to the equivalent
__functionName() call. This makes the code run faster at the cost of diminished precision and
accuracy. See Math Libraries.

109

CUDA C++ Best Practices Guide, Release 12.5

110

Chapter 23. nvcc Compiler Switches

Chapter 24. Notices

24.1. Notice

This document is provided for information purposes only and shall not be regarded as a warranty of a
certain functionality, condition, or quality of a product. NVIDIA Corporation (“NVIDIA”) makes no repre-
sentations or warranties, expressed or implied, as to the accuracy or completeness of the information
contained in this document and assumes no responsibility for any errors contained herein. NVIDIA shall
have no liability for the consequences or use of such information or for any infringement of patents
or other rights of third parties that may result from its use. This document is not a commitment to
develop, release, or deliver any Material (defined below), code, or functionality.

NVIDIA reserves the right to make corrections, modifications, enhancements, improvements, and any
other changes to this document, at any time without notice.

Customer should obtain the latest relevant information before placing orders and should verify that
such information is current and complete.

NVIDIA products are sold subject to the NVIDIA standard terms and conditions of sale supplied at the
time of order acknowledgement, unless otherwise agreed in an individual sales agreement signed by
authorized representatives of NVIDIA and customer (“Terms of Sale”). NVIDIA hereby expressly objects
to applying any customer general terms and conditions with regards to the purchase of the NVIDIA
product referenced in this document. No contractual obligations are formed either directly or indirectly
by this document.

NVIDIA products are not designed, authorized, or warranted to be suitable for use in medical, military,
aircraft, space, or life support equipment, nor in applications where failure or malfunction of the NVIDIA
product can reasonably be expected to result in personal injury, death, or property or environmental
damage. NVIDIA accepts no liability for inclusion and/or use of NVIDIA products in such equipment or
applications and therefore such inclusion and/or use is at customer’s own risk.

NVIDIA makes no representation or warranty that products based on this document will be suitable for
any specified use. Testing of all parameters of each product is not necessarily performed by NVIDIA.
It is customer’s sole responsibility to evaluate and determine the applicability of any information con-
tained in this document, ensure the product is suitable and fit for the application planned by customer,
and perform the necessary testing for the application in order to avoid a default of the application or
the product. Weaknesses in customer’s product designs may affect the quality and reliability of the
NVIDIA product and may result in additional or different conditions and/or requirements beyond those
contained in this document. NVIDIA accepts no liability related to any default, damage, costs, or prob-
lem which may be based on or attributable to: (i) the use of the NVIDIA product in any manner that is
contrary to this document or (ii) customer product designs.

No license, either expressed or implied, is granted under any NVIDIA patent right, copyright, or other
NVIDIA intellectual property right under this document.
Information published by NVIDIA regarding
third-party products or services does not constitute a license from NVIDIA to use such products or

111

CUDA C++ Best Practices Guide, Release 12.5

services or a warranty or endorsement thereof. Use of such information may require a license from a
third party under the patents or other intellectual property rights of the third party, or a license from
NVIDIA under the patents or other intellectual property rights of NVIDIA.

Reproduction of information in this document is permissible only if approved in advance by NVIDIA
in writing, reproduced without alteration and in full compliance with all applicable export laws and
regulations, and accompanied by all associated conditions, limitations, and notices.

THIS DOCUMENT AND ALL NVIDIA DESIGN SPECIFICATIONS, REFERENCE BOARDS, FILES, DRAWINGS,
DIAGNOSTICS, LISTS, AND OTHER DOCUMENTS (TOGETHER AND SEPARATELY, “MATERIALS”) ARE
BEING PROVIDED “AS IS.” NVIDIA MAKES NO WARRANTIES, EXPRESSED, IMPLIED, STATUTORY, OR
OTHERWISE WITH RESPECT TO THE MATERIALS, AND EXPRESSLY DISCLAIMS ALL IMPLIED WAR-
RANTIES OF NONINFRINGEMENT, MERCHANTABILITY, AND FITNESS FOR A PARTICULAR PURPOSE.
TO THE EXTENT NOT PROHIBITED BY LAW, IN NO EVENT WILL NVIDIA BE LIABLE FOR ANY DAMAGES,
INCLUDING WITHOUT LIMITATION ANY DIRECT, INDIRECT, SPECIAL, INCIDENTAL, PUNITIVE, OR CON-
SEQUENTIAL DAMAGES, HOWEVER CAUSED AND REGARDLESS OF THE THEORY OF LIABILITY, ARIS-
ING OUT OF ANY USE OF THIS DOCUMENT, EVEN IF NVIDIA HAS BEEN ADVISED OF THE POSSIBILITY
OF SUCH DAMAGES. Notwithstanding any damages that customer might incur for any reason whatso-
ever, NVIDIA’s aggregate and cumulative liability towards customer for the products described herein
shall be limited in accordance with the Terms of Sale for the product.

24.2. OpenCL

OpenCL is a trademark of Apple Inc. used under license to the Khronos Group Inc.

24.3. Trademarks

NVIDIA and the NVIDIA logo are trademarks or registered trademarks of NVIDIA Corporation in the
U.S. and other countries. Other company and product names may be trademarks of the respective
companies with which they are associated.

Copyright

©2007-2024, NVIDIA Corporation & affiliates. All rights reserved

112

Chapter 24. Notices