github/concrete
1
1
Fork 0
mirror of https://github.com/zama-ai/concrete.git synced 2026-02-09 03:55:04 -05:00
Code Issues Packages Projects Releases Wiki Activity

fix(GPU): remove transfers of uninitialized data to GPU.

Browse Source
This commit is contained in:
Antoniu Pop
2023-01-30 10:30:55 +00:00
committed by Quentin Bourgerie
parent 36f51ba0c2
commit f60bd16e2b
1 changed files with 2 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
compiler/lib/Runtime/wrappers.cpp
View File

@@ -136,8 +136,7 @@ void memref_batched_keyswitch_lwe_cuda_u64(
// TODO: The allocation should be done by the compiler codegen
void *ct0_gpu = alloc_and_memcpy_async_to_gpu(
ct0_aligned, ct0_offset, ct0_batch_size, gpu_idx, stream);
void *out_gpu = alloc_and_memcpy_async_to_gpu(
out_aligned, out_offset, out_batch_size, gpu_idx, stream);
void *out_gpu = cuda_malloc(out_batch_size * sizeof(uint64_t), gpu_idx);
// Run the keyswitch kernel on the GPU
cuda_keyswitch_lwe_ciphertext_vector_64(
stream, gpu_idx, out_gpu, ct0_gpu, ksk_gpu, input_lwe_dim, output_lwe_dim,
@@ -179,9 +178,7 @@ void memref_batched_bootstrap_lwe_cuda_u64(
// TODO: The allocation should be done by the compiler codegen
void *ct0_gpu = alloc_and_memcpy_async_to_gpu(
ct0_aligned, ct0_offset, ct0_batch_size, gpu_idx, stream);
void *out_gpu = alloc_and_memcpy_async_to_gpu(
out_aligned, out_offset, out_batch_size, gpu_idx, stream);
void *out_gpu = cuda_malloc(out_batch_size * sizeof(uint64_t), gpu_idx);
// Construct the glwe accumulator (on CPU)
// TODO: Should be done outside of the bootstrap call, compile time if
// possible. Refactor in progress