Thanks for using Compiler Explorer
Sponsors
Jakt
C++
Ada
Analysis
Android Java
Android Kotlin
Assembly
C
C3
Carbon
C++ (Circle)
CIRCT
Clean
CMake
CMakeScript
COBOL
C++ for OpenCL
MLIR
Cppx
Cppx-Blue
Cppx-Gold
Cpp2-cppfront
Crystal
C#
CUDA C++
D
Dart
Elixir
Erlang
Fortran
F#
GLSL
Go
Haskell
HLSL
Hook
Hylo
IL
ispc
Java
Julia
Kotlin
LLVM IR
LLVM MIR
Modula-2
Nim
Objective-C
Objective-C++
OCaml
OpenCL C
Pascal
Pony
Python
Racket
Ruby
Rust
Snowball
Scala
Solidity
Spice
Swift
LLVM TableGen
Toit
TypeScript Native
V
Vala
Visual Basic
WASM
Zig
Javascript
GIMPLE
cuda source #1
Output
Compile to binary object
Link to binary
Execute the code
Intel asm syntax
Demangle identifiers
Verbose demangling
Filters
Unused labels
Library functions
Directives
Comments
Horizontal whitespace
Debug intrinsics
Compiler
10.0.0 sm_75 CUDA-10.2
10.0.1 sm_75 CUDA-10.2
11.0.0 sm_75 CUDA-10.2
NVCC 10.0.130
NVCC 10.1.105
NVCC 10.1.168
NVCC 10.1.243
NVCC 10.2.89
NVCC 11.0.2
NVCC 11.0.3
NVCC 11.1.0
NVCC 11.1.1
NVCC 11.2.0
NVCC 11.2.1
NVCC 11.2.2
NVCC 11.3.0
NVCC 11.3.1
NVCC 11.4.0
NVCC 11.4.1
NVCC 11.4.2
NVCC 11.4.3
NVCC 11.4.4
NVCC 11.5.0
NVCC 11.5.1
NVCC 11.5.2
NVCC 11.6.0
NVCC 11.6.1
NVCC 11.6.2
NVCC 11.7.0
NVCC 11.7.1
NVCC 11.8.0
NVCC 12.0.0
NVCC 12.0.1
NVCC 12.1.0
NVCC 12.2.1
NVCC 12.3.1
NVCC 12.4.1
NVCC 12.5.1
NVCC 12.6.1
NVCC 9.1.85
NVCC 9.2.88
NVRTC 11.0.2
NVRTC 11.0.3
NVRTC 11.1.0
NVRTC 11.1.1
NVRTC 11.2.0
NVRTC 11.2.1
NVRTC 11.2.2
NVRTC 11.3.0
NVRTC 11.3.1
NVRTC 11.4.0
NVRTC 11.4.1
NVRTC 11.5.0
NVRTC 11.5.1
NVRTC 11.5.2
NVRTC 11.6.0
NVRTC 11.6.1
NVRTC 11.6.2
NVRTC 11.7.0
NVRTC 11.7.1
NVRTC 11.8.0
NVRTC 12.0.0
NVRTC 12.0.1
NVRTC 12.1.0
clang 7.0.0 sm_70 CUDA-9.1
clang 8.0.0 sm_75 CUDA-10.0
clang 9.0.0 sm_75 CUDA-10.1
clang rocm-4.5.2
clang rocm-5.0.2
clang rocm-5.1.3
clang rocm-5.2.3
clang rocm-5.3.2
clang rocm-5.7.0
clang rocm-6.0.2
clang rocm-6.1.2
clang staging rocm-6.1.2
clang trunk rocm-6.1.2
trunk sm_86 CUDA-11.3
Options
Source code
#define index_type int #define data_type float #define bs 4 __device__ index_type round_up_to_power_of_two (index_type v) { v--; v |= v >> 1; v |= v >> 2; v |= v >> 4; v |= v >> 8; v |= v >> 16; v++; return v; } __global__ void bcsr_spmv_kernel_column_by_column_template ( const index_type * __restrict__ col_ids, const index_type * __restrict__ row_ptr, const data_type * __restrict__ data, const data_type * __restrict__ x, data_type * __restrict__ y) { const index_type idx = blockIdx.x * blockDim.x + threadIdx.x; const index_type lane = idx % 32; const index_type block_row = idx / 32; ///< Warp per block row const index_type first_block = row_ptr[block_row]; const index_type last_block = row_ptr[block_row + 1]; index_type col = first_block * bs + lane / bs; index_type r = lane % bs; extern __shared__ data_type partial_sums[]; data_type local_out = 0.0; for (; col < last_block * bs; col += 32 / bs) { const index_type block = col / bs; const index_type c = col % bs; const data_type value = data[block * bs * bs + c * bs + r]; const data_type x_value = x[col_ids[block] * bs + c]; local_out += x_value * value; } partial_sums[threadIdx.x] = local_out; for (index_type stride = round_up_to_power_of_two((32 / bs) / 2); stride > 0; stride /= 2) { __syncthreads (); if ((lane < stride * bs) && ((threadIdx.x + stride * bs) < 32)) partial_sums[threadIdx.x] += partial_sums[threadIdx.x + stride * bs]; } if (lane < bs) y[block_row * bs + lane] = partial_sums[threadIdx.x]; }
Become a Patron
Sponsor on GitHub
Donate via PayPal
Source on GitHub
Mailing list
Installed libraries
Wiki
Report an issue
How it works
Contact the author
CE on Mastodon
About the author
Statistics
Changelog
Version tree