DarthAffe/stable-diffusion.cpp
1
0
Fork 0
mirror of https://github.com/leejet/stable-diffusion.cpp.git synced 2026-09-24 20:20:37 +00:00
Code Issues Packages Projects Releases Wiki Activity

perf: eliminate temporary allocations in Philox rounds (#1982)

Browse Source
This commit is contained in:
leejet 2026-09-17 02:09:12 +08:00 committed by GitHub
parent 3161505fe8
commit cc515a01f9
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 7 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
src/core/rng_philox.hpp
View File

@ -19,13 +19,6 @@ private:
float two_pow32_inv = 2.3283064e-10f; float two_pow32_inv = 2.3283064e-10f;
float two_pow32_inv_2pi = 2.3283064e-10f * 6.2831855f; float two_pow32_inv_2pi = 2.3283064e-10f * 6.2831855f;
std::vector<uint32_t> uint32(uint64_t x) {
std::vector<uint32_t> result(2);
result[0] = static_cast<uint32_t>(x & 0xFFFFFFFF);
result[1] = static_cast<uint32_t>(x >> 32);
return result;
}
std::vector<std::vector<uint32_t>> uint32(const std::vector<uint64_t>& x) { std::vector<std::vector<uint32_t>> uint32(const std::vector<uint64_t>& x) {
uint32_t N = (uint32_t)x.size(); uint32_t N = (uint32_t)x.size();
std::vector<std::vector<uint32_t>> result(2, std::vector<uint32_t>(N)); std::vector<std::vector<uint32_t>> result(2, std::vector<uint32_t>(N));
@ -43,13 +36,13 @@ private:
const std::vector<std::vector<uint32_t>>& key) { const std::vector<std::vector<uint32_t>>& key) {
uint32_t N = (uint32_t)counter[0].size(); uint32_t N = (uint32_t)counter[0].size();
for (uint32_t i = 0; i < N; i++) { for (uint32_t i = 0; i < N; i++) {
std::vector<uint32_t> v1 = uint32(static_cast<uint64_t>(counter[0][i]) * static_cast<uint64_t>(philox_m[0])); const uint64_t v1 = static_cast<uint64_t>(counter[0][i]) * static_cast<uint64_t>(philox_m[0]);
std::vector<uint32_t> v2 = uint32(static_cast<uint64_t>(counter[2][i]) * static_cast<uint64_t>(philox_m[1])); const uint64_t v2 = static_cast<uint64_t>(counter[2][i]) * static_cast<uint64_t>(philox_m[1]);
counter[0][i] = v2[1] ^ counter[1][i] ^ key[0][i]; counter[0][i] = static_cast<uint32_t>(v2 >> 32) ^ counter[1][i] ^ key[0][i];
counter[1][i] = v2[0]; counter[1][i] = static_cast<uint32_t>(v2);
counter[2][i] = v1[1] ^ counter[3][i] ^ key[1][i]; counter[2][i] = static_cast<uint32_t>(v1 >> 32) ^ counter[3][i] ^ key[1][i];
counter[3][i] = v1[0]; counter[3][i] = static_cast<uint32_t>(v1);
} }
} }
@ -126,4 +119,4 @@ public:
} }
}; };
#endif // __SD_CORE_RNG_PHILOX_HPP__ #endif // __SD_CORE_RNG_PHILOX_HPP__