AtHeartEngineer/icicle
1
0
Fork 0
mirror of https://github.com/pseXperiments/icicle.git synced 2026-01-08 23:17:54 -05:00
Code Issues Packages Projects Releases Wiki Activity

move backend-specific-config to open part to avoid installing it

Browse Source
This commit is contained in:
Yuval Shekel
2024-08-01 16:56:34 +03:00
parent 73dd80b63b
commit 908a6c536f
7 changed files with 49 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
examples/c++/best-practice-ntt/example.cpp
View File

@@ -8,6 +8,7 @@
using namespace bn254;
#include "examples_utils.h"
#include "icicle/backend/ntt_config.h"
void initialize_input(const unsigned ntt_size, const unsigned batch_size, scalar_t* elements)
{
@@ -41,7 +42,7 @@ int main(int argc, char* argv[])
auto ntt_init_domain_cfg = default_ntt_init_domain_config();
// set CUDA backend specific flag for init_domain
ConfigExtension backend_cfg_ext;
backend_cfg_ext.set("fast_twiddles", true);
backend_cfg_ext.set(CudaBackendConfig::CUDA_NTT_FAST_TWIDDLES_MODE, true);
ntt_init_domain_cfg.ext = &backend_cfg_ext;
ICICLE_CHECK(bn254_ntt_init_domain(&basic_root, ntt_init_domain_cfg));
@@ -78,7 +79,7 @@ int main(int argc, char* argv[])
config_compute.stream = stream_compute;
// backend specific config extension
ConfigExtension ntt_cfg_ext;
ntt_cfg_ext.set("ntt_algorithm", 2); // mixed-radix
ntt_cfg_ext.set(CudaBackendConfig::CUDA_NTT_ALGORITHM, CudaBackendConfig::NttAlgorithm::MixedRadix);
config_compute.ext = &ntt_cfg_ext;
for (int run = 0; run < 10; run++) {
@@ -115,8 +116,8 @@ int main(int argc, char* argv[])
// Clean-up
for (int i = 0; i < 2; i++) {
ICICLE_CHECK(icicle_free(d_vec[i]));
delete[](h_inp[i]);
delete[](h_out[i]);
delete[] (h_inp[i]);
delete[] (h_out[i]);
}
ICICLE_CHECK(icicle_destroy_stream(stream_compute));
ICICLE_CHECK(icicle_destroy_stream(stream_d2h));

8
examples/c++/ntt/example.cpp
View File

@@ -6,6 +6,7 @@
using namespace bn254;
#include "examples_utils.h"
#include "icicle/backend/ntt_config.h"
void initialize_input(const unsigned ntt_size, const unsigned nof_ntts, scalar_t* elements);
int validate_output(const unsigned ntt_size, const unsigned nof_ntts, scalar_t* elements);
@@ -33,7 +34,8 @@ int main(int argc, char* argv[])
scalar_t basic_root = scalar_t::omega(log_ntt_size /*NTT_LOG_SIZscalar_t*/);
auto ntt_init_domain_cfg = default_ntt_init_domain_config();
ConfigExtension backend_cfg_ext;
backend_cfg_ext.set("fast_twiddles", true); // optionally construct fast_twiddles for CUDA backend
backend_cfg_ext.set(
CudaBackendConfig::CUDA_NTT_FAST_TWIDDLES_MODE, true); // optionally construct fast_twiddles for CUDA backend
ntt_init_domain_cfg.ext = &backend_cfg_ext;
ICICLE_CHECK(bn254_ntt_init_domain(&basic_root, ntt_init_domain_cfg));
@@ -48,7 +50,7 @@ int main(int argc, char* argv[])
// NTT radix-2 alg
std::cout << "\nRunning NTT radix-2 alg with on-host data" << std::endl;
ntt_cfg_ext.set("ntt_algorithm", 1); // radix-2
ntt_cfg_ext.set(CudaBackendConfig::CUDA_NTT_ALGORITHM, CudaBackendConfig::NttAlgorithm::Radix2);
START_TIMER(Radix2);
ICICLE_CHECK(bn254_ntt(input.get(), ntt_size, NTTDir::kForward, config, output.get()));
END_TIMER(Radix2, "Radix2 NTT");
@@ -58,7 +60,7 @@ int main(int argc, char* argv[])
// NTT mixed-radix alg
std::cout << "\nRunning NTT mixed-radix alg with on-host data" << std::endl;
ntt_cfg_ext.set("ntt_algorithm", 2); // mixed-radix
ntt_cfg_ext.set(CudaBackendConfig::CUDA_NTT_ALGORITHM, CudaBackendConfig::NttAlgorithm::MixedRadix);
START_TIMER(MixedRadix);
ICICLE_CHECK(bn254_ntt(input.get(), ntt_size, NTTDir::kForward, config, output.get()));
END_TIMER(MixedRadix, "MixedRadix NTT");

13
icicle_v3/include/icicle/backend/msm_config.h Normal file
View File

@@ -0,0 +1,13 @@
#pragma once
/********* CPU Backend Configurations *********/
/********* CUDA Backend Configurations *********/
namespace CudaBackendConfig {
// Backend-specific configuration flags as constexpr strings
constexpr const char* CUDA_MSM_IS_BIG_TRIANGLE = "is_big_triangle";
constexpr const char* CUDA_MSM_LARGE_BUCKET_FACTOR = "large_bucket_factor";
constexpr const int CUDA_MSM_LARGE_BUCKET_FACTOR_DEFAULT_VAL = 10;
} // namespace CudaBackendConfig

18
icicle_v3/include/icicle/backend/ntt_config.h Normal file
View File

@@ -0,0 +1,18 @@
#pragma once
/********* CPU Backend Configurations *********/
/********* CUDA Backend Configurations *********/
namespace CudaBackendConfig {
// Enumeration for NTT (Number Theoretic Transform) algorithms
enum NttAlgorithm {
Auto, ///< Automatically choose the algorithm
Radix2, ///< Use Radix-2 algorithm
MixedRadix ///< Use Mixed Radix algorithm
};
// Backend-specific configuration flags as constexpr strings
constexpr const char* CUDA_NTT_FAST_TWIDDLES_MODE = "fast_twiddles"; ///< Enable fast twiddles mode
constexpr const char* CUDA_NTT_ALGORITHM = "ntt_algorithm"; ///< Set NTT algorithm
} // namespace CudaBackendConfig

7
icicle_v3/tests/test_field_api.cpp
View File

@@ -10,6 +10,7 @@
#include "icicle/fields/field_config.h"
#include "icicle/utils/log.h"
#include "icicle/backend/ntt_config.h"
using namespace field_config;
using namespace icicle;
@@ -26,10 +27,6 @@ static int ITERS = 16;
static inline std::string s_main_target;
static inline std::string s_reference_target;
// TODO Yuval: better include the CUDA configs and use instead of redefine
#define CUDA_NTT_FAST_TWIDDLES_MODE "fast_twiddles"
#define CUDA_NTT_ALGORITHM "ntt_algorithm"
template <typename T>
class FieldApiTest : public ::testing::Test
{
@@ -328,7 +325,7 @@ TYPED_TEST(FieldApiTest, ntt)
init_domain_config.stream = stream;
init_domain_config.is_async = false;
ConfigExtension ext;
ext.set(CUDA_NTT_FAST_TWIDDLES_MODE, true);
ext.set(CudaBackendConfig::CUDA_NTT_FAST_TWIDDLES_MODE, true);
init_domain_config.ext = &ext;
auto config = default_ntt_config<scalar_t>();

8
wrappers/rust_v3/icicle-core/src/msm/mod.rs
View File

@@ -49,8 +49,8 @@ pub struct MSMConfig {
}
// backend specific options
pub const LARGE_BUCKET_FACTOR: &str = "large_bucket_factor";
pub const IS_BIG_TRIANGLE: &str = "is_big_triangle";
pub const CUDA_MSM_LARGE_BUCKET_FACTOR: &str = "large_bucket_factor";
pub const CUDA_MSM_IS_BIG_TRIANGLE: &str = "is_big_triangle";
impl Default for MSMConfig {
fn default() -> Self {
@@ -297,7 +297,7 @@ macro_rules! impl_msm_bench {
) => {
use criterion::{criterion_group, criterion_main, Criterion};
use icicle_core::curve::{Affine, Curve, Projective};
use icicle_core::msm::{msm, MSMConfig, LARGE_BUCKET_FACTOR, MSM};
use icicle_core::msm::{msm, MSMConfig, CUDA_MSM_LARGE_BUCKET_FACTOR, MSM};
use icicle_core::traits::{FieldImpl, GenerateRandom};
use icicle_runtime::{
device::Device,
@@ -364,7 +364,7 @@ macro_rules! impl_msm_bench {
cfg.stream_handle = *stream;
cfg.is_async = true;
cfg.ext
.set_int(LARGE_BUCKET_FACTOR, 5);
.set_int(CUDA_MSM_LARGE_BUCKET_FACTOR, 5);
cfg.c = 4;
warmup(&stream).unwrap();

4
wrappers/rust_v3/icicle-core/src/msm/tests.rs
View File

@@ -1,5 +1,5 @@
use crate::curve::{Affine, Curve, Projective};
use crate::msm::{msm, precompute_bases, MSMConfig, LARGE_BUCKET_FACTOR, MSM};
use crate::msm::{msm, precompute_bases, MSMConfig, CUDA_MSM_LARGE_BUCKET_FACTOR, MSM};
use crate::test_utilities;
use crate::traits::{FieldImpl, GenerateRandom, MontgomeryConvertible};
use icicle_runtime::{
@@ -109,7 +109,7 @@ where
cfg.stream_handle = *stream;
cfg.is_async = true;
cfg.ext
.set_int(LARGE_BUCKET_FACTOR, 5);
.set_int(CUDA_MSM_LARGE_BUCKET_FACTOR, 5);
cfg.c = 4;
runtime::warmup(&stream).unwrap();
stream