summaryrefslogtreecommitdiff
path: root/kernels/src
diff options
context:
space:
mode:
authorlaurent <laurent.mazare@gmail.com>2023-06-22 13:52:02 +0100
committerlaurent <laurent.mazare@gmail.com>2023-06-22 13:52:02 +0100
commite1eb86db61d789627105ca0929b9fd83b1bf991e (patch)
tree9c22ec9542d9eecf18104e3674b492ec6d677836 /kernels/src
parent83d619800950a5f4314fa6bb885835dbfbf9dfcf (diff)
downloadcandle-e1eb86db61d789627105ca0929b9fd83b1bf991e.tar.gz
candle-e1eb86db61d789627105ca0929b9fd83b1bf991e.tar.bz2
candle-e1eb86db61d789627105ca0929b9fd83b1bf991e.zip
Add some first binary op (add).
Diffstat (limited to 'kernels/src')
-rw-r--r--kernels/src/binary_op_macros.cuh7
1 files changed, 4 insertions, 3 deletions
diff --git a/kernels/src/binary_op_macros.cuh b/kernels/src/binary_op_macros.cuh
index 8d4954f5..60531b4d 100644
--- a/kernels/src/binary_op_macros.cuh
+++ b/kernels/src/binary_op_macros.cuh
@@ -4,13 +4,14 @@
extern "C" __global__ void FN_NAME( \
const size_t numel, \
const size_t num_dims, \
- const size_t *dims, \
- const size_t *lhs_strides, \
- const size_t *rhs_strides, \
+ const size_t *dims_and_strides, \
const TYPENAME *lhs, \
const TYPENAME *rhs, \
TYPENAME *out \
) { \
+ const size_t *dims = dims_and_strides; \
+ const size_t *lhs_strides = dims_and_strides + 1 * num_dims; \
+ const size_t *rhs_strides = dims_and_strides + 2 * num_dims; \
for (unsigned int i = blockIdx.x * blockDim.x + threadIdx.x; i < numel; i += blockDim.x * gridDim.x) { \
unsigned int tmp_i = i; \
unsigned int lhs_i = 0; \