sami7777/triangles_v5
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
9f0ce13abc9dc633b7dcf61592b594b14f91a07d
triangles_v5/src/simd.c
T
SamiAhmed7777 9a50ab3b2e Expand script.cpp test coverage: libFuzzer harness + EvalScript stress tests + UBSan fix (#27) 
* simd: fix UBSan signed-shift UB in fft64 INNER macro

The INNER macro at src/simd.c:379 combines the low and high halves of
two FFT values with a multiplier:

    ((u32)((l) * (mm)) & 0xFFFFU) + ((u32)((h) * (mm)) << 16)

When (h)*(mm) is a negative s32, the (u32) cast recovers the bit
pattern (large positive number), but then << 16 operates on the
integer-promoted value (typically int on x86_64). UBSan flags this as
signed-shift of negative.

Fix: explicit (u32) cast inside the shift expression forces the shift
operand to unsigned (well-defined per C++20). Outer (u32) cast keeps
the result type consistent. Bit-equivalent at runtime; type-safe for
UBSan.

Mirrors the pattern Krystie applied in b9d06d5 for the same issue in
FFT8/FFT16 macros. Could not reproduce the trip in a standalone
100-trial test (Hash9's specific call pattern from CBlock::GetHash
may not be reproducible in isolation), but the macro is the same UB
class as already-fixed sites — fix by inspection per the
triangles-test-suite-audit skill.

Build verified: ninja test_triangles clean, all 234 test cases +
21720 assertions pass under sanitizers. Tests exercise Hash9 via
TestingSetup, so the FFT path is covered.

* test: add libFuzzer harness + EvalScript stress tests + CI fuzz job

Three pieces, one goal: expand coverage of script.cpp (the
consensus-critical opcode interpreter) beyond what Boost unit tests
catch.

1. libFuzzer harness (src/test/fuzz/)

   Builds against the existing daemon object files (init.cpp.o,
   wallet.cpp.o, noui.cpp.o) so we get the full CWallet vtable
   without writing 100+ lines of fragile method stubs. Link line
   reuses the sanitizer-friendly flags from test-linux-sanitizers.
   Corpus seeded from src/test/data/script_{valid,invalid}.json
   (1055 real Triangles scripts).

   BUILD_FUZZ is OFF by default — gcc default build doesn't have
   libFuzzer, so the flag gates the custom clang++ build cleanly.

2. Stress tests (src/test/script_stress_tests.cpp)

   Six regression-guard tests for EvalScript's hard limits. Anyone
   who removes a bound will get a test failure:
   - deep_dup_stack_hits_opcount_limit   — 250 OP_DUP rejected <1s
   - max_keys_multisig_20_of_20          — 20-of-20 terminates <2s
   - multisig_rejects_21_keys            — nKeysCount > 20 rejected
   - pushdata_over_520_rejected          — MAX_SCRIPT_ELEMENT_SIZE
   - script_size_over_10000_rejected     — MAX_SCRIPT_SIZE
   - disabled_opcodes_rejected           — all 15 disabled opcodes

   EvalScript contract caveat (captured in the test comments): on
   false return, the stack is left dirty — inputs pushed before
   rejection are still there. Tests assert stack.size() <= N for
   N = number of inputs pushed, not the post-opcode expectation.

3. CI fuzz job (.github/workflows/build-all.yml)

   test-fuzz-smoke job, sibling of test-linux-sanitizers. Reuses
   the same runner + apt-get + RocksDB-from-source + Tor-from-source
   steps so CI runtime doesn't double. Builds with clang-15,
   ASan+UBSan+libFuzzer, seeds corpus, runs 5 minutes, fails only
   on crash artifact (not on find_new_units=0 — libFuzzer always
   writes .tmp churn during normal operation).

Verified:
- ninja test_triangles clean
- 234/234 test cases + 21720/21720 assertions pass
- 4/4 ctest suites pass (triangles_unit + chaindb_equivalence +
  snapshotnet + chaindb_runtime)
- Local fuzz run: 8354 corpus files, ~5400 exec/sec, zero crashes
  after several hours (-jobs=2 -workers=2)

* build: explicit <cassert> in allocators.h

clang's stricter include resolution surfaces the missing include even
though gcc tolerates it via some other transitive path. Without this,
PR #27 build with clang fails on assert() in LockedPageManager.

* fix(ci,fuzz): unbreak test-fuzz-smoke workflow + CMake fuzz link deps

Three bugs in PR #27's fuzz smoke integration that CI caught on first run:

1. CMAKE_EXE_LINKER_FLAGS pulled in '-fsanitize=fuzzer $SAN_FLAGS'. CMake's
   compiler-probe linker test (used to verify the toolchain) doesn't define
   LLVMFuzzerTestOneInput, so adding -fsanitize=fuzzer pulls in
   libclang_rt.fuzzer's main() and trips 'multiple definition of `main`'.
   Remove -fsanitize=fuzzer from global flags; fuzz_script already adds it
   per-target via FUZZ_COMMON_FLAGS in src/CMakeLists.txt.

2. Workflow said --target script_fuzz / ./build-fuzz/bin/script_fuzz, but
   the CMake target is fuzz_script (add_custom_target(fuzz_script ...)).
   CI failed with 'unknown target'. Fixed in workflow + comment.

3. fuzz_script link references static libs at ${CMAKE_BINARY_DIR}/lib/
   (libhash9_crypto.a, libleveldb_lib.a, libleveldb_memenv.a, libsecp256k1.a)
   but didn't declare them as DEPENDS. First clean build races the link
   step and fails with 'no such file or directory'. Added the four static
   library targets to DEPENDS so ninja builds them first.

Verified locally: cmake configure clean, fuzz_script link succeeds,
binary runs (./bin/fuzz_script prints libFuzzer banner and reads corpus).

Tested with the same flag set CI uses (SAN_FLAGS with fuzzer-no-link,
BUILD_FUZZ=ON, clang-14).

* fix(ci,fuzz): make test-fuzz-smoke work end-to-end on clean builds

Three more bugs in PR #27's fuzz integration, caught on local repro
after commit 3239425 fixed the easy ones:

1. clang vs gcc warning mismatch (cmake/AddCompilerFlags.cmake):
   gcc treats -Wreserved-user-defined-literal as a warning. clang-15+
   in C++20 mode promotes it to an error and trips on hundreds of
   Bitcoin-derived sites like strprintf("%"PRId64...) in util.cpp /
   kernel.cpp. Conditional -Wno-reserved-user-defined-literal scoped
   to clang only — gcc builds keep the original diagnostic.

2. secp256k1 ASM strictness (.github/workflows/build-all.yml):
   Add -DSECP256K1_ASM=OFF to the fuzz configure. Clang-15+'s
   register allocator is sometimes stricter than clang-14 about the
   x86_64 inline asm in scalar_4x64_impl.h and fails with 'inline
   assembly requires more registers than available'. The fuzz target
   only needs ECC at the C-fallback level — slower but correct.

3. Empty .o glob at link time (src/CMakeLists.txt):
   The fuzz link line referenced CMakeFiles/triangles_common.dir/*.o
   and CMakeFiles/trianglesd.dir/*.o via file(GLOB), which evaluates
   at cmake CONFIGURE time. On a fresh build dir, no .o files exist
   yet → the link line was always empty → undefined references for
   CKey::GetPubKey, typeinfo for CKeyStore, etc.

   Replace the GLOB with a generated bash wrapper script
   (fuzz_objs/link.sh) that does the find at link time, using bash
   arrays to safely handle paths with spaces. The script is invoked
   via ninja with the original link line as its argv; it prepends
   the discovered .o files (excluding script.cpp.o — we have our
   own clang-instrumented copy in fuzz_objs/) and exec's clang++.

Verified locally:
  - cmake configures cleanly under clang-18 with the same flag set CI uses
  - ninja fuzz_script links end-to-end (132 MB ELF, debug info, all
    sanitizer coverage instrumentation intact)
  - ./bin/fuzz_script runs and discovers coverage: 'INITED cov: 3
    ft: 3 corp: 1/1b' from libFuzzer banner
  - 4/4 ctest suites still pass on the gcc build (master chaindb /
    snapshotnet / unit / equivalence)

This should make test-fuzz-smoke pass on the next CI run.

* fix(ci,fuzz): stabilize fuzz smoke job

* fix(ci,fuzz): portable fuzz build, exclude daemon main, stub globals

- Build daemon (init/wallet/noui) as an OBJECT library under BUILD_FUZZ
  so the fuzz job does not pay the daemon executable link cost.
- Exclude init.cpp.o from the fuzz link wrapper (it defines the daemon's
  main(), conflicting with libFuzzer's own).
- Replace hardcoded libboost/librocksdb filenames with -l flags and
  Boost target paths resolved at configure time.
- Add -lubsan to the fuzz link line so libstdc++'s ubsan hooks resolve.
- Add a generated fuzz_stubs.cpp that defines pwalletMain, uiInterface,
  CheckpointsMode, nNodeLifespan, etc. — every global that init.cpp
  used to provide.
- Keep the CI workflow's libtor build step (fuzz links libtor.a).

Local verify: clang-15 + clang-18 + gcc all build fuzz_script; 15s
fuzz run completes 1913 execs with no crash artifacts; gcc test build
passes 4/4 ctest suites unchanged.

* fix(ci,fuzz): drop libi2pd*.a paths from fuzz link line

The fuzz job in build-all.yml runs libtor.sh but NOT libi2pd.sh, so
src/i2p/i2pd-src/libi2pd{,client,lang}.a don't exist on the CI runner.
The previous commit (720d711) hardcoded them into the fuzz link line;
the link failed with
  clang: error: no such file or directory: '.../i2p/i2pd-src/libi2pd*.a'

i2p_embedded.cpp is already compiled into triangles_common, and with
USE_I2P_EMBEDDED=OFF (the CI default) the only i2p surface is the
no-op stub in triangles_common. So the .a references were both wrong
AND redundant.

Keep libtor.a: src/tor/build-libtor.sh IS run in the fuzz job, so the
file exists when the link wrapper invokes clang++.

Verified locally with clang-15 against the same cmake flags the
workflow uses: clean link, 12s fuzz run did 1239 executions, no crash
artifacts, libFuzzer reporting normal coverage growth.

* fix(ci,fuzz): install libgflags-dev for fuzz smoke job

CI failure on PR #27 test-fuzz-smoke: link step aborted with
`/usr/bin/ld: cannot find -lgflags`. The fuzz link line in
src/CMakeLists.txt references -lgflags (transitive dep of RocksDB),
and CI's ubuntu-22.04 runner does NOT ship libgflags-dev.

DNS2 ships libgflags-dev as an automatic dep of build-essential,
which is why local dry-runs didn't catch this.

Verified locally on DNS2:
- Cloned the exact CI cmake invocation (build-fuzz-verify dir)
- cmake -B + cmake --build --target fuzz_script: clean build
- 30s fuzz pass: 1058 inputs, 1935 features covered, no crashes

The libtor build step also depends on gflags transitively; making
it explicit in the apt-get list future-proofs both paths.

* fix(ci,fuzz): link -lrocksdb unconditionally in fuzz target

CI's test-fuzz-smoke job was failing with a torrent of
`undefined reference to rocksdb::Status::ToString[abi:cxx11]()`
errors after the libgflags fix landed. The fuzz target's RocksDB
link arg was:

    $<IF:$<TARGET_EXISTS:RocksDB::rocksdb>,-lrocksdb,${ROCKSDB_LIBRARY}>

That generator expression was wrong on CI's exact code path:

  1. CMake's `find_package(RocksDB CONFIG)` does NOT find the .cmake
     config RocksDB 8.9.1 ships — only the .pc file.
  2. `pkg_check_modules(RocksDB IMPORTED_TARGET)` therefore exposes
     `PkgConfig::RocksDB` (NOT `RocksDB::rocksdb`), so
     $<TARGET_EXISTS:RocksDB::rocksdb> is FALSE.
  3. The fallback ${ROCKSDB_LIBRARY} is set ONLY inside the manual
     `find_library()` probe at top-level CMakeLists.txt:170-190, which
     is skipped when EITHER `RocksDB::rocksdb` or `PkgConfig::RocksDB`
     already exists.

Result on CI: an empty string landed in the link line, so the link
step saw no `-lrocksdb` arg and every RocksDB symbol the fuzz
binary referenced became undefined.

Fix: just use a bare `-lrocksdb` and let the library search path
do the work. `build-rocksdb.sh` installs to /usr/local/lib (CI);
`librocksdb-dev` (apt) installs to /usr/lib/x86_64-linux-gnu (DNS2).
Both paths are in the default search path.

Verified locally on DNS2 with the exact CI cmake invocation + flags:
- build-fuzz-verify2: clean build, exit 0
- 20s fuzz pass: 1548 inputs, 2024 features covered, no crashes

Co-located with the libgflags fix on feat/script-fuzz-and-stress-tests
because both fixes are required for test-fuzz-smoke to turn green.

---------

Co-authored-by: Sami Ahmed <sami@sami-ahmed.net>
2026-07-11 16:02:35 -07:00

1815 lines
49 KiB
C
Raw Blame History

/* $Id: simd.c 227 2010-06-16 17:28:38Z tp $ */
/*
* SIMD implementation.
*
* ==========================(LICENSE BEGIN)============================
*
* Copyright (c) 2007-2010 Projet RNRT SAPHIR
*
* Permission is hereby granted, free of charge, to any person obtaining
* a copy of this software and associated documentation files (the
* "Software"), to deal in the Software without restriction, including
* without limitation the rights to use, copy, modify, merge, publish,
* distribute, sublicense, and/or sell copies of the Software, and to
* permit persons to whom the Software is furnished to do so, subject to
* the following conditions:
*
* The above copyright notice and this permission notice shall be
* included in all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*
* ===========================(LICENSE END)=============================
*
* @author Thomas Pornin <thomas.pornin@cryptolog.com>
*/
#include <stddef.h>
#include <string.h>
#include <limits.h>
#include "sph_simd.h"
#ifdef __cplusplus
extern "C"{
#endif
#if SPH_SMALL_FOOTPRINT && !defined SPH_SMALL_FOOTPRINT_SIMD
#define SPH_SMALL_FOOTPRINT_SIMD 1
#endif
#ifdef _MSC_VER
#pragma warning (disable: 4146)
#endif
typedef sph_u32 u32;
typedef sph_s32 s32;
#define C32 SPH_C32
#define T32 SPH_T32
#define ROL32 SPH_ROTL32
#define XCAT(x, y) XCAT_(x, y)
#define XCAT_(x, y) x ## y
/*
* The powers of 41 modulo 257. We use exponents from 0 to 255, inclusive.
*/
static const s32 alpha_tab[] = {
1, 41, 139, 45, 46, 87, 226, 14, 60, 147, 116, 130,
190, 80, 196, 69, 2, 82, 21, 90, 92, 174, 195, 28,
120, 37, 232, 3, 123, 160, 135, 138, 4, 164, 42, 180,
184, 91, 133, 56, 240, 74, 207, 6, 246, 63, 13, 19,
8, 71, 84, 103, 111, 182, 9, 112, 223, 148, 157, 12,
235, 126, 26, 38, 16, 142, 168, 206, 222, 107, 18, 224,
189, 39, 57, 24, 213, 252, 52, 76, 32, 27, 79, 155,
187, 214, 36, 191, 121, 78, 114, 48, 169, 247, 104, 152,
64, 54, 158, 53, 117, 171, 72, 125, 242, 156, 228, 96,
81, 237, 208, 47, 128, 108, 59, 106, 234, 85, 144, 250,
227, 55, 199, 192, 162, 217, 159, 94, 256, 216, 118, 212,
211, 170, 31, 243, 197, 110, 141, 127, 67, 177, 61, 188,
255, 175, 236, 167, 165, 83, 62, 229, 137, 220, 25, 254,
134, 97, 122, 119, 253, 93, 215, 77, 73, 166, 124, 201,
17, 183, 50, 251, 11, 194, 244, 238, 249, 186, 173, 154,
146, 75, 248, 145, 34, 109, 100, 245, 22, 131, 231, 219,
241, 115, 89, 51, 35, 150, 239, 33, 68, 218, 200, 233,
44, 5, 205, 181, 225, 230, 178, 102, 70, 43, 221, 66,
136, 179, 143, 209, 88, 10, 153, 105, 193, 203, 99, 204,
140, 86, 185, 132, 15, 101, 29, 161, 176, 20, 49, 210,
129, 149, 198, 151, 23, 172, 113, 7, 30, 202, 58, 65,
95, 40, 98, 163
};
/*
* Ranges:
* REDS1: from -32768..98302 to -383..383
* REDS2: from -2^31..2^31-1 to -32768..98302
*/
#define REDS1(x) (((x) & 0xFF) - ((x) >> 8))
#define REDS2(x) (((x) & 0xFFFF) + ((x) >> 16))
/*
* If, upon entry, the values of q[] are all in the -N..N range (where
* N >= 98302) then the new values of q[] are in the -2N..2N range.
*
* Since alpha_tab[v] <= 256, maximum allowed range is for N = 8388608.
*/
#define FFT_LOOP(rb, hk, as, id) do { \
size_t u, v; \
s32 m = q[(rb)]; \
s32 n = q[(rb) + (hk)]; \
q[(rb)] = m + n; \
q[(rb) + (hk)] = m - n; \
u = v = 0; \
goto id; \
for (; u < (hk); u += 4, v += 4 * (as)) { \
s32 t; \
m = q[(rb) + u + 0]; \
n = q[(rb) + u + 0 + (hk)]; \
t = REDS2(n * alpha_tab[v + 0 * (as)]); \
q[(rb) + u + 0] = m + t; \
q[(rb) + u + 0 + (hk)] = m - t; \
id: \
m = q[(rb) + u + 1]; \
n = q[(rb) + u + 1 + (hk)]; \
t = REDS2(n * alpha_tab[v + 1 * (as)]); \
q[(rb) + u + 1] = m + t; \
q[(rb) + u + 1 + (hk)] = m - t; \
m = q[(rb) + u + 2]; \
n = q[(rb) + u + 2 + (hk)]; \
t = REDS2(n * alpha_tab[v + 2 * (as)]); \
q[(rb) + u + 2] = m + t; \
q[(rb) + u + 2 + (hk)] = m - t; \
m = q[(rb) + u + 3]; \
n = q[(rb) + u + 3 + (hk)]; \
t = REDS2(n * alpha_tab[v + 3 * (as)]); \
q[(rb) + u + 3] = m + t; \
q[(rb) + u + 3 + (hk)] = m - t; \
} \
} while (0)
/*
* Output ranges:
* d0: min= 0 max= 1020
* d1: min= -67 max= 4587
* d2: min=-4335 max= 4335
* d3: min=-4147 max= 507
* d4: min= -510 max= 510
* d5: min= -252 max= 4402
* d6: min=-4335 max= 4335
* d7: min=-4332 max= 322
*
* Use multiplication, not signed left shift, for powers of two below. FFT
* values can be negative; left-shifting a negative signed integer is
* undefined in C, while these bounded multiplications are defined and
* preserve the intended arithmetic.
*/
#define FFT8(xb, xs, d) do { \
s32 x0 = x[(xb)]; \
s32 x1 = x[(xb) + (xs)]; \
s32 x2 = x[(xb) + 2 * (xs)]; \
s32 x3 = x[(xb) + 3 * (xs)]; \
s32 a0 = x0 + x2; \
s32 a1 = x0 + (x2 * 16); \
s32 a2 = x0 - x2; \
s32 a3 = x0 - (x2 * 16); \
s32 b0 = x1 + x3; \
s32 b1 = REDS1((x1 * 4) + (x3 * 64)); \
s32 b2 = (x1 * 16) - (x3 * 16); \
s32 b3 = REDS1((x1 * 64) + (x3 * 4)); \
d ## 0 = a0 + b0; \
d ## 1 = a1 + b1; \
d ## 2 = a2 + b2; \
d ## 3 = a3 + b3; \
d ## 4 = a0 - b0; \
d ## 5 = a1 - b1; \
d ## 6 = a2 - b2; \
d ## 7 = a3 - b3; \
} while (0)
/*
* When k=16, we have alpha=2. Multiplication by alpha^i is then reduced
* to some shifting.
*
* Output: within -591471..591723
*/
#define FFT16(xb, xs, rb) do { \
s32 d1_0, d1_1, d1_2, d1_3, d1_4, d1_5, d1_6, d1_7; \
s32 d2_0, d2_1, d2_2, d2_3, d2_4, d2_5, d2_6, d2_7; \
FFT8(xb, (xs) << 1, d1_); \
FFT8((xb) + (xs), (xs) << 1, d2_); \
q[(rb) + 0] = d1_0 + d2_0; \
q[(rb) + 1] = d1_1 + (d2_1 * 2); \
q[(rb) + 2] = d1_2 + (d2_2 * 4); \
q[(rb) + 3] = d1_3 + (d2_3 * 8); \
q[(rb) + 4] = d1_4 + (d2_4 * 16); \
q[(rb) + 5] = d1_5 + (d2_5 * 32); \
q[(rb) + 6] = d1_6 + (d2_6 * 64); \
q[(rb) + 7] = d1_7 + (d2_7 * 128); \
q[(rb) + 8] = d1_0 - d2_0; \
q[(rb) + 9] = d1_1 - (d2_1 * 2); \
q[(rb) + 10] = d1_2 - (d2_2 * 4); \
q[(rb) + 11] = d1_3 - (d2_3 * 8); \
q[(rb) + 12] = d1_4 - (d2_4 * 16); \
q[(rb) + 13] = d1_5 - (d2_5 * 32); \
q[(rb) + 14] = d1_6 - (d2_6 * 64); \
q[(rb) + 15] = d1_7 - (d2_7 * 128); \
} while (0)
/*
* Output range: |q| <= 1183446
*/
#define FFT32(xb, xs, rb, id) do { \
FFT16(xb, (xs) << 1, rb); \
FFT16((xb) + (xs), (xs) << 1, (rb) + 16); \
FFT_LOOP(rb, 16, 8, id); \
} while (0)
/*
* Output range: |q| <= 2366892
*/
#define FFT64(xb, xs, rb, id) do { \
FFT32(xb, (xs) << 1, rb, XCAT(id, a)); \
FFT32((xb) + (xs), (xs) << 1, (rb) + 32, XCAT(id, b)); \
FFT_LOOP(rb, 32, 4, id); \
} while (0)
#if SPH_SMALL_FOOTPRINT_SIMD
static void
fft32(unsigned char *x, size_t xs, s32 *q)
{
size_t xd;
xd = xs << 1;
FFT16(0, xd, 0);
FFT16(xs, xd, 16);
FFT_LOOP(0, 16, 8, label_);
}
#define FFT128(xb, xs, rb, id) do { \
fft32(x + (xb) + ((xs) * 0), (xs) << 2, &q[(rb) + 0]); \
fft32(x + (xb) + ((xs) * 2), (xs) << 2, &q[(rb) + 32]); \
FFT_LOOP(rb, 32, 4, XCAT(id, aa)); \
fft32(x + (xb) + ((xs) * 1), (xs) << 2, &q[(rb) + 64]); \
fft32(x + (xb) + ((xs) * 3), (xs) << 2, &q[(rb) + 96]); \
FFT_LOOP((rb) + 64, 32, 4, XCAT(id, ab)); \
FFT_LOOP(rb, 64, 2, XCAT(id, a)); \
} while (0)
#else
/*
* Output range: |q| <= 4733784
*/
#define FFT128(xb, xs, rb, id) do { \
FFT64(xb, (xs) << 1, rb, XCAT(id, a)); \
FFT64((xb) + (xs), (xs) << 1, (rb) + 64, XCAT(id, b)); \
FFT_LOOP(rb, 64, 2, id); \
} while (0)
#endif
/*
* For SIMD-384 / SIMD-512, the fully unrolled FFT yields a compression
* function which does not fit in the 32 kB L1 cache of a typical x86
* Intel. We therefore add a function call layer at the FFT64 level.
*/
static void
fft64(unsigned char *x, size_t xs, s32 *q)
{
size_t xd;
xd = xs << 1;
FFT32(0, xd, 0, label_a);
FFT32(xs, xd, 32, label_b);
FFT_LOOP(0, 32, 4, label_);
}
/*
* Output range: |q| <= 9467568
*/
#define FFT256(xb, xs, rb, id) do { \
fft64(x + (xb) + ((xs) * 0), (xs) << 2, &q[(rb) + 0]); \
fft64(x + (xb) + ((xs) * 2), (xs) << 2, &q[(rb) + 64]); \
FFT_LOOP(rb, 64, 2, XCAT(id, aa)); \
fft64(x + (xb) + ((xs) * 1), (xs) << 2, &q[(rb) + 128]); \
fft64(x + (xb) + ((xs) * 3), (xs) << 2, &q[(rb) + 192]); \
FFT_LOOP((rb) + 128, 64, 2, XCAT(id, ab)); \
FFT_LOOP(rb, 128, 1, XCAT(id, a)); \
} while (0)
/*
* alpha^(127*i) mod 257
*/
static const unsigned short yoff_s_n[] = {
1, 98, 95, 58, 30, 113, 23, 198, 129, 49, 176, 29,
15, 185, 140, 99, 193, 153, 88, 143, 136, 221, 70, 178,
225, 205, 44, 200, 68, 239, 35, 89, 241, 231, 22, 100,
34, 248, 146, 173, 249, 244, 11, 50, 17, 124, 73, 215,
253, 122, 134, 25, 137, 62, 165, 236, 255, 61, 67, 141,
197, 31, 211, 118, 256, 159, 162, 199, 227, 144, 234, 59,
128, 208, 81, 228, 242, 72, 117, 158, 64, 104, 169, 114,
121, 36, 187, 79, 32, 52, 213, 57, 189, 18, 222, 168,
16, 26, 235, 157, 223, 9, 111, 84, 8, 13, 246, 207,
240, 133, 184, 42, 4, 135, 123, 232, 120, 195, 92, 21,
2, 196, 190, 116, 60, 226, 46, 139
};
/*
* alpha^(127*i) + alpha^(125*i) mod 257
*/
static const unsigned short yoff_s_f[] = {
2, 156, 118, 107, 45, 212, 111, 162, 97, 249, 211, 3,
49, 101, 151, 223, 189, 178, 253, 204, 76, 82, 232, 65,
96, 176, 161, 47, 189, 61, 248, 107, 0, 131, 133, 113,
17, 33, 12, 111, 251, 103, 57, 148, 47, 65, 249, 143,
189, 8, 204, 230, 205, 151, 187, 227, 247, 111, 140, 6,
77, 10, 21, 149, 255, 101, 139, 150, 212, 45, 146, 95,
160, 8, 46, 254, 208, 156, 106, 34, 68, 79, 4, 53,
181, 175, 25, 192, 161, 81, 96, 210, 68, 196, 9, 150,
0, 126, 124, 144, 240, 224, 245, 146, 6, 154, 200, 109,
210, 192, 8, 114, 68, 249, 53, 27, 52, 106, 70, 30,
10, 146, 117, 251, 180, 247, 236, 108
};
/*
* beta^(255*i) mod 257
*/
static const unsigned short yoff_b_n[] = {
1, 163, 98, 40, 95, 65, 58, 202, 30, 7, 113, 172,
23, 151, 198, 149, 129, 210, 49, 20, 176, 161, 29, 101,
15, 132, 185, 86, 140, 204, 99, 203, 193, 105, 153, 10,
88, 209, 143, 179, 136, 66, 221, 43, 70, 102, 178, 230,
225, 181, 205, 5, 44, 233, 200, 218, 68, 33, 239, 150,
35, 51, 89, 115, 241, 219, 231, 131, 22, 245, 100, 109,
34, 145, 248, 75, 146, 154, 173, 186, 249, 238, 244, 194,
11, 251, 50, 183, 17, 201, 124, 166, 73, 77, 215, 93,
253, 119, 122, 97, 134, 254, 25, 220, 137, 229, 62, 83,
165, 167, 236, 175, 255, 188, 61, 177, 67, 127, 141, 110,
197, 243, 31, 170, 211, 212, 118, 216, 256, 94, 159, 217,
162, 192, 199, 55, 227, 250, 144, 85, 234, 106, 59, 108,
128, 47, 208, 237, 81, 96, 228, 156, 242, 125, 72, 171,
117, 53, 158, 54, 64, 152, 104, 247, 169, 48, 114, 78,
121, 191, 36, 214, 187, 155, 79, 27, 32, 76, 52, 252,
213, 24, 57, 39, 189, 224, 18, 107, 222, 206, 168, 142,
16, 38, 26, 126, 235, 12, 157, 148, 223, 112, 9, 182,
111, 103, 84, 71, 8, 19, 13, 63, 246, 6, 207, 74,
240, 56, 133, 91, 184, 180, 42, 164, 4, 138, 135, 160,
123, 3, 232, 37, 120, 28, 195, 174, 92, 90, 21, 82,
2, 69, 196, 80, 190, 130, 116, 147, 60, 14, 226, 87,
46, 45, 139, 41
};
/*
* beta^(255*i) + beta^(253*i) mod 257
*/
static const unsigned short yoff_b_f[] = {
2, 203, 156, 47, 118, 214, 107, 106, 45, 93, 212, 20,
111, 73, 162, 251, 97, 215, 249, 53, 211, 19, 3, 89,
49, 207, 101, 67, 151, 130, 223, 23, 189, 202, 178, 239,
253, 127, 204, 49, 76, 236, 82, 137, 232, 157, 65, 79,
96, 161, 176, 130, 161, 30, 47, 9, 189, 247, 61, 226,
248, 90, 107, 64, 0, 88, 131, 243, 133, 59, 113, 115,
17, 236, 33, 213, 12, 191, 111, 19, 251, 61, 103, 208,
57, 35, 148, 248, 47, 116, 65, 119, 249, 178, 143, 40,
189, 129, 8, 163, 204, 227, 230, 196, 205, 122, 151, 45,
187, 19, 227, 72, 247, 125, 111, 121, 140, 220, 6, 107,
77, 69, 10, 101, 21, 65, 149, 171, 255, 54, 101, 210,
139, 43, 150, 151, 212, 164, 45, 237, 146, 184, 95, 6,
160, 42, 8, 204, 46, 238, 254, 168, 208, 50, 156, 190,
106, 127, 34, 234, 68, 55, 79, 18, 4, 130, 53, 208,
181, 21, 175, 120, 25, 100, 192, 178, 161, 96, 81, 127,
96, 227, 210, 248, 68, 10, 196, 31, 9, 167, 150, 193,
0, 169, 126, 14, 124, 198, 144, 142, 240, 21, 224, 44,
245, 66, 146, 238, 6, 196, 154, 49, 200, 222, 109, 9,
210, 141, 192, 138, 8, 79, 114, 217, 68, 128, 249, 94,
53, 30, 27, 61, 52, 135, 106, 212, 70, 238, 30, 185,
10, 132, 146, 136, 117, 37, 251, 150, 180, 188, 247, 156,
236, 192, 108, 86
};
// Compute the FFT "step" combining low and high halves of two q[]
// values with a multiplier. Casts to u32 happen BEFORE the left shift
// to avoid signed-overflow UB when (h)*(mm) is negative — UBSan flags
// signed-shift of negative values even though the intent is modular
// arithmetic. The u32 cast then <<16 is well-defined in C++20.
//
// Why (h)*(mm) can be negative: FFT values are signed; alpha_tab entries
// are signed too. The product can overflow into negative s32 territory
// before the modular reduction step. We compute the multiplication in
// signed, cast to u32 to recover the bit pattern, then shift.
#define INNER(l, h, mm) (((u32)((l) * (mm)) & 0xFFFFU) \
+ ((u32)((u32)((h) * (mm)) << 16)))
#define W_SMALL(sb, o1, o2, mm) \
(INNER(q[8 * (sb) + 2 * 0 + o1], q[8 * (sb) + 2 * 0 + o2], mm), \
INNER(q[8 * (sb) + 2 * 1 + o1], q[8 * (sb) + 2 * 1 + o2], mm), \
INNER(q[8 * (sb) + 2 * 2 + o1], q[8 * (sb) + 2 * 2 + o2], mm), \
INNER(q[8 * (sb) + 2 * 3 + o1], q[8 * (sb) + 2 * 3 + o2], mm)
#define WS_0_0 W_SMALL( 4, 0, 1, 185)
#define WS_0_1 W_SMALL( 6, 0, 1, 185)
#define WS_0_2 W_SMALL( 0, 0, 1, 185)
#define WS_0_3 W_SMALL( 2, 0, 1, 185)
#define WS_0_4 W_SMALL( 7, 0, 1, 185)
#define WS_0_5 W_SMALL( 5, 0, 1, 185)
#define WS_0_6 W_SMALL( 3, 0, 1, 185)
#define WS_0_7 W_SMALL( 1, 0, 1, 185)
#define WS_1_0 W_SMALL(15, 0, 1, 185)
#define WS_1_1 W_SMALL(11, 0, 1, 185)
#define WS_1_2 W_SMALL(12, 0, 1, 185)
#define WS_1_3 W_SMALL( 8, 0, 1, 185)
#define WS_1_4 W_SMALL( 9, 0, 1, 185)
#define WS_1_5 W_SMALL(13, 0, 1, 185)
#define WS_1_6 W_SMALL(10, 0, 1, 185)
#define WS_1_7 W_SMALL(14, 0, 1, 185)
#define WS_2_0 W_SMALL(17, -128, -64, 233)
#define WS_2_1 W_SMALL(18, -128, -64, 233)
#define WS_2_2 W_SMALL(23, -128, -64, 233)
#define WS_2_3 W_SMALL(20, -128, -64, 233)
#define WS_2_4 W_SMALL(22, -128, -64, 233)
#define WS_2_5 W_SMALL(21, -128, -64, 233)
#define WS_2_6 W_SMALL(16, -128, -64, 233)
#define WS_2_7 W_SMALL(19, -128, -64, 233)
#define WS_3_0 W_SMALL(30, -191, -127, 233)
#define WS_3_1 W_SMALL(24, -191, -127, 233)
#define WS_3_2 W_SMALL(25, -191, -127, 233)
#define WS_3_3 W_SMALL(31, -191, -127, 233)
#define WS_3_4 W_SMALL(27, -191, -127, 233)
#define WS_3_5 W_SMALL(29, -191, -127, 233)
#define WS_3_6 W_SMALL(28, -191, -127, 233)
#define WS_3_7 W_SMALL(26, -191, -127, 233)
#define W_BIG(sb, o1, o2, mm) \
(INNER(q[16 * (sb) + 2 * 0 + o1], q[16 * (sb) + 2 * 0 + o2], mm), \
INNER(q[16 * (sb) + 2 * 1 + o1], q[16 * (sb) + 2 * 1 + o2], mm), \
INNER(q[16 * (sb) + 2 * 2 + o1], q[16 * (sb) + 2 * 2 + o2], mm), \
INNER(q[16 * (sb) + 2 * 3 + o1], q[16 * (sb) + 2 * 3 + o2], mm), \
INNER(q[16 * (sb) + 2 * 4 + o1], q[16 * (sb) + 2 * 4 + o2], mm), \
INNER(q[16 * (sb) + 2 * 5 + o1], q[16 * (sb) + 2 * 5 + o2], mm), \
INNER(q[16 * (sb) + 2 * 6 + o1], q[16 * (sb) + 2 * 6 + o2], mm), \
INNER(q[16 * (sb) + 2 * 7 + o1], q[16 * (sb) + 2 * 7 + o2], mm)
#define WB_0_0 W_BIG( 4, 0, 1, 185)
#define WB_0_1 W_BIG( 6, 0, 1, 185)
#define WB_0_2 W_BIG( 0, 0, 1, 185)
#define WB_0_3 W_BIG( 2, 0, 1, 185)
#define WB_0_4 W_BIG( 7, 0, 1, 185)
#define WB_0_5 W_BIG( 5, 0, 1, 185)
#define WB_0_6 W_BIG( 3, 0, 1, 185)
#define WB_0_7 W_BIG( 1, 0, 1, 185)
#define WB_1_0 W_BIG(15, 0, 1, 185)
#define WB_1_1 W_BIG(11, 0, 1, 185)
#define WB_1_2 W_BIG(12, 0, 1, 185)
#define WB_1_3 W_BIG( 8, 0, 1, 185)
#define WB_1_4 W_BIG( 9, 0, 1, 185)
#define WB_1_5 W_BIG(13, 0, 1, 185)
#define WB_1_6 W_BIG(10, 0, 1, 185)
#define WB_1_7 W_BIG(14, 0, 1, 185)
#define WB_2_0 W_BIG(17, -256, -128, 233)
#define WB_2_1 W_BIG(18, -256, -128, 233)
#define WB_2_2 W_BIG(23, -256, -128, 233)
#define WB_2_3 W_BIG(20, -256, -128, 233)
#define WB_2_4 W_BIG(22, -256, -128, 233)
#define WB_2_5 W_BIG(21, -256, -128, 233)
#define WB_2_6 W_BIG(16, -256, -128, 233)
#define WB_2_7 W_BIG(19, -256, -128, 233)
#define WB_3_0 W_BIG(30, -383, -255, 233)
#define WB_3_1 W_BIG(24, -383, -255, 233)
#define WB_3_2 W_BIG(25, -383, -255, 233)
#define WB_3_3 W_BIG(31, -383, -255, 233)
#define WB_3_4 W_BIG(27, -383, -255, 233)
#define WB_3_5 W_BIG(29, -383, -255, 233)
#define WB_3_6 W_BIG(28, -383, -255, 233)
#define WB_3_7 W_BIG(26, -383, -255, 233)
#define IF(x, y, z) ((((y) ^ (z)) & (x)) ^ (z))
#define MAJ(x, y, z) (((x) & (y)) | (((x) | (y)) & (z)))
#define PP4_0_0 1
#define PP4_0_1 0
#define PP4_0_2 3
#define PP4_0_3 2
#define PP4_1_0 2
#define PP4_1_1 3
#define PP4_1_2 0
#define PP4_1_3 1
#define PP4_2_0 3
#define PP4_2_1 2
#define PP4_2_2 1
#define PP4_2_3 0
#define PP8_0_0 1
#define PP8_0_1 0
#define PP8_0_2 3
#define PP8_0_3 2
#define PP8_0_4 5
#define PP8_0_5 4
#define PP8_0_6 7
#define PP8_0_7 6
#define PP8_1_0 6
#define PP8_1_1 7
#define PP8_1_2 4
#define PP8_1_3 5
#define PP8_1_4 2
#define PP8_1_5 3
#define PP8_1_6 0
#define PP8_1_7 1
#define PP8_2_0 2
#define PP8_2_1 3
#define PP8_2_2 0
#define PP8_2_3 1
#define PP8_2_4 6
#define PP8_2_5 7
#define PP8_2_6 4
#define PP8_2_7 5
#define PP8_3_0 3
#define PP8_3_1 2
#define PP8_3_2 1
#define PP8_3_3 0
#define PP8_3_4 7
#define PP8_3_5 6
#define PP8_3_6 5
#define PP8_3_7 4
#define PP8_4_0 5
#define PP8_4_1 4
#define PP8_4_2 7
#define PP8_4_3 6
#define PP8_4_4 1
#define PP8_4_5 0
#define PP8_4_6 3
#define PP8_4_7 2
#define PP8_5_0 7
#define PP8_5_1 6
#define PP8_5_2 5
#define PP8_5_3 4
#define PP8_5_4 3
#define PP8_5_5 2
#define PP8_5_6 1
#define PP8_5_7 0
#define PP8_6_0 4
#define PP8_6_1 5
#define PP8_6_2 6
#define PP8_6_3 7
#define PP8_6_4 0
#define PP8_6_5 1
#define PP8_6_6 2
#define PP8_6_7 3
#if SPH_SIMD_NOCOPY
#define DECL_STATE_SMALL
#define READ_STATE_SMALL(sc)
#define WRITE_STATE_SMALL(sc)
#define DECL_STATE_BIG
#define READ_STATE_BIG(sc)
#define WRITE_STATE_BIG(sc)
#else
#define DECL_STATE_SMALL \
u32 A0, A1, A2, A3, B0, B1, B2, B3, C0, C1, C2, C3, D0, D1, D2, D3;
#define READ_STATE_SMALL(sc) do { \
A0 = (sc)->state[ 0]; \
A1 = (sc)->state[ 1]; \
A2 = (sc)->state[ 2]; \
A3 = (sc)->state[ 3]; \
B0 = (sc)->state[ 4]; \
B1 = (sc)->state[ 5]; \
B2 = (sc)->state[ 6]; \
B3 = (sc)->state[ 7]; \
C0 = (sc)->state[ 8]; \
C1 = (sc)->state[ 9]; \
C2 = (sc)->state[10]; \
C3 = (sc)->state[11]; \
D0 = (sc)->state[12]; \
D1 = (sc)->state[13]; \
D2 = (sc)->state[14]; \
D3 = (sc)->state[15]; \
} while (0)
#define WRITE_STATE_SMALL(sc) do { \
(sc)->state[ 0] = A0; \
(sc)->state[ 1] = A1; \
(sc)->state[ 2] = A2; \
(sc)->state[ 3] = A3; \
(sc)->state[ 4] = B0; \
(sc)->state[ 5] = B1; \
(sc)->state[ 6] = B2; \
(sc)->state[ 7] = B3; \
(sc)->state[ 8] = C0; \
(sc)->state[ 9] = C1; \
(sc)->state[10] = C2; \
(sc)->state[11] = C3; \
(sc)->state[12] = D0; \
(sc)->state[13] = D1; \
(sc)->state[14] = D2; \
(sc)->state[15] = D3; \
} while (0)
#define DECL_STATE_BIG \
u32 A0, A1, A2, A3, A4, A5, A6, A7; \
u32 B0, B1, B2, B3, B4, B5, B6, B7; \
u32 C0, C1, C2, C3, C4, C5, C6, C7; \
u32 D0, D1, D2, D3, D4, D5, D6, D7;
#define READ_STATE_BIG(sc) do { \
A0 = (sc)->state[ 0]; \
A1 = (sc)->state[ 1]; \
A2 = (sc)->state[ 2]; \
A3 = (sc)->state[ 3]; \
A4 = (sc)->state[ 4]; \
A5 = (sc)->state[ 5]; \
A6 = (sc)->state[ 6]; \
A7 = (sc)->state[ 7]; \
B0 = (sc)->state[ 8]; \
B1 = (sc)->state[ 9]; \
B2 = (sc)->state[10]; \
B3 = (sc)->state[11]; \
B4 = (sc)->state[12]; \
B5 = (sc)->state[13]; \
B6 = (sc)->state[14]; \
B7 = (sc)->state[15]; \
C0 = (sc)->state[16]; \
C1 = (sc)->state[17]; \
C2 = (sc)->state[18]; \
C3 = (sc)->state[19]; \
C4 = (sc)->state[20]; \
C5 = (sc)->state[21]; \
C6 = (sc)->state[22]; \
C7 = (sc)->state[23]; \
D0 = (sc)->state[24]; \
D1 = (sc)->state[25]; \
D2 = (sc)->state[26]; \
D3 = (sc)->state[27]; \
D4 = (sc)->state[28]; \
D5 = (sc)->state[29]; \
D6 = (sc)->state[30]; \
D7 = (sc)->state[31]; \
} while (0)
#define WRITE_STATE_BIG(sc) do { \
(sc)->state[ 0] = A0; \
(sc)->state[ 1] = A1; \
(sc)->state[ 2] = A2; \
(sc)->state[ 3] = A3; \
(sc)->state[ 4] = A4; \
(sc)->state[ 5] = A5; \
(sc)->state[ 6] = A6; \
(sc)->state[ 7] = A7; \
(sc)->state[ 8] = B0; \
(sc)->state[ 9] = B1; \
(sc)->state[10] = B2; \
(sc)->state[11] = B3; \
(sc)->state[12] = B4; \
(sc)->state[13] = B5; \
(sc)->state[14] = B6; \
(sc)->state[15] = B7; \
(sc)->state[16] = C0; \
(sc)->state[17] = C1; \
(sc)->state[18] = C2; \
(sc)->state[19] = C3; \
(sc)->state[20] = C4; \
(sc)->state[21] = C5; \
(sc)->state[22] = C6; \
(sc)->state[23] = C7; \
(sc)->state[24] = D0; \
(sc)->state[25] = D1; \
(sc)->state[26] = D2; \
(sc)->state[27] = D3; \
(sc)->state[28] = D4; \
(sc)->state[29] = D5; \
(sc)->state[30] = D6; \
(sc)->state[31] = D7; \
} while (0)
#endif
#define STEP_ELT(n, w, fun, s, ppb) do { \
u32 tt = T32(D ## n + (w) + fun(A ## n, B ## n, C ## n)); \
A ## n = T32(ROL32(tt, s) + XCAT(tA, XCAT(ppb, n))); \
D ## n = C ## n; \
C ## n = B ## n; \
B ## n = tA ## n; \
} while (0)
#define STEP_SMALL(w0, w1, w2, w3, fun, r, s, pp4b) do { \
u32 tA0 = ROL32(A0, r); \
u32 tA1 = ROL32(A1, r); \
u32 tA2 = ROL32(A2, r); \
u32 tA3 = ROL32(A3, r); \
STEP_ELT(0, w0, fun, s, pp4b); \
STEP_ELT(1, w1, fun, s, pp4b); \
STEP_ELT(2, w2, fun, s, pp4b); \
STEP_ELT(3, w3, fun, s, pp4b); \
} while (0)
#define STEP_BIG(w0, w1, w2, w3, w4, w5, w6, w7, fun, r, s, pp8b) do { \
u32 tA0 = ROL32(A0, r); \
u32 tA1 = ROL32(A1, r); \
u32 tA2 = ROL32(A2, r); \
u32 tA3 = ROL32(A3, r); \
u32 tA4 = ROL32(A4, r); \
u32 tA5 = ROL32(A5, r); \
u32 tA6 = ROL32(A6, r); \
u32 tA7 = ROL32(A7, r); \
STEP_ELT(0, w0, fun, s, pp8b); \
STEP_ELT(1, w1, fun, s, pp8b); \
STEP_ELT(2, w2, fun, s, pp8b); \
STEP_ELT(3, w3, fun, s, pp8b); \
STEP_ELT(4, w4, fun, s, pp8b); \
STEP_ELT(5, w5, fun, s, pp8b); \
STEP_ELT(6, w6, fun, s, pp8b); \
STEP_ELT(7, w7, fun, s, pp8b); \
} while (0)
#define M3_0_0 0_
#define M3_1_0 1_
#define M3_2_0 2_
#define M3_3_0 0_
#define M3_4_0 1_
#define M3_5_0 2_
#define M3_6_0 0_
#define M3_7_0 1_
#define M3_0_1 1_
#define M3_1_1 2_
#define M3_2_1 0_
#define M3_3_1 1_
#define M3_4_1 2_
#define M3_5_1 0_
#define M3_6_1 1_
#define M3_7_1 2_
#define M3_0_2 2_
#define M3_1_2 0_
#define M3_2_2 1_
#define M3_3_2 2_
#define M3_4_2 0_
#define M3_5_2 1_
#define M3_6_2 2_
#define M3_7_2 0_
#define STEP_SMALL_(w, fun, r, s, pp4b) STEP_SMALL w, fun, r, s, pp4b)
#define ONE_ROUND_SMALL(ri, isp, p0, p1, p2, p3) do { \
STEP_SMALL_(WS_ ## ri ## 0, \
IF, p0, p1, XCAT(PP4_, M3_0_ ## isp)); \
STEP_SMALL_(WS_ ## ri ## 1, \
IF, p1, p2, XCAT(PP4_, M3_1_ ## isp)); \
STEP_SMALL_(WS_ ## ri ## 2, \
IF, p2, p3, XCAT(PP4_, M3_2_ ## isp)); \
STEP_SMALL_(WS_ ## ri ## 3, \
IF, p3, p0, XCAT(PP4_, M3_3_ ## isp)); \
STEP_SMALL_(WS_ ## ri ## 4, \
MAJ, p0, p1, XCAT(PP4_, M3_4_ ## isp)); \
STEP_SMALL_(WS_ ## ri ## 5, \
MAJ, p1, p2, XCAT(PP4_, M3_5_ ## isp)); \
STEP_SMALL_(WS_ ## ri ## 6, \
MAJ, p2, p3, XCAT(PP4_, M3_6_ ## isp)); \
STEP_SMALL_(WS_ ## ri ## 7, \
MAJ, p3, p0, XCAT(PP4_, M3_7_ ## isp)); \
} while (0)
#define M7_0_0 0_
#define M7_1_0 1_
#define M7_2_0 2_
#define M7_3_0 3_
#define M7_4_0 4_
#define M7_5_0 5_
#define M7_6_0 6_
#define M7_7_0 0_
#define M7_0_1 1_
#define M7_1_1 2_
#define M7_2_1 3_
#define M7_3_1 4_
#define M7_4_1 5_
#define M7_5_1 6_
#define M7_6_1 0_
#define M7_7_1 1_
#define M7_0_2 2_
#define M7_1_2 3_
#define M7_2_2 4_
#define M7_3_2 5_
#define M7_4_2 6_
#define M7_5_2 0_
#define M7_6_2 1_
#define M7_7_2 2_
#define M7_0_3 3_
#define M7_1_3 4_
#define M7_2_3 5_
#define M7_3_3 6_
#define M7_4_3 0_
#define M7_5_3 1_
#define M7_6_3 2_
#define M7_7_3 3_
#define STEP_BIG_(w, fun, r, s, pp8b) STEP_BIG w, fun, r, s, pp8b)
#define ONE_ROUND_BIG(ri, isp, p0, p1, p2, p3) do { \
STEP_BIG_(WB_ ## ri ## 0, \
IF, p0, p1, XCAT(PP8_, M7_0_ ## isp)); \
STEP_BIG_(WB_ ## ri ## 1, \
IF, p1, p2, XCAT(PP8_, M7_1_ ## isp)); \
STEP_BIG_(WB_ ## ri ## 2, \
IF, p2, p3, XCAT(PP8_, M7_2_ ## isp)); \
STEP_BIG_(WB_ ## ri ## 3, \
IF, p3, p0, XCAT(PP8_, M7_3_ ## isp)); \
STEP_BIG_(WB_ ## ri ## 4, \
MAJ, p0, p1, XCAT(PP8_, M7_4_ ## isp)); \
STEP_BIG_(WB_ ## ri ## 5, \
MAJ, p1, p2, XCAT(PP8_, M7_5_ ## isp)); \
STEP_BIG_(WB_ ## ri ## 6, \
MAJ, p2, p3, XCAT(PP8_, M7_6_ ## isp)); \
STEP_BIG_(WB_ ## ri ## 7, \
MAJ, p3, p0, XCAT(PP8_, M7_7_ ## isp)); \
} while (0)
#if SPH_SMALL_FOOTPRINT_SIMD
#define A0 state[ 0]
#define A1 state[ 1]
#define A2 state[ 2]
#define A3 state[ 3]
#define B0 state[ 4]
#define B1 state[ 5]
#define B2 state[ 6]
#define B3 state[ 7]
#define C0 state[ 8]
#define C1 state[ 9]
#define C2 state[10]
#define C3 state[11]
#define D0 state[12]
#define D1 state[13]
#define D2 state[14]
#define D3 state[15]
#define STEP2_ELT(n, w, fun, s, ppb) do { \
u32 tt = T32(D ## n + (w) + fun(A ## n, B ## n, C ## n)); \
A ## n = T32(ROL32(tt, s) + tA[(ppb) ^ n]); \
D ## n = C ## n; \
C ## n = B ## n; \
B ## n = tA[n]; \
} while (0)
#define STEP2_SMALL(w0, w1, w2, w3, fun, r, s, pp4b) do { \
u32 tA[4]; \
tA[0] = ROL32(A0, r); \
tA[1] = ROL32(A1, r); \
tA[2] = ROL32(A2, r); \
tA[3] = ROL32(A3, r); \
STEP2_ELT(0, w0, fun, s, pp4b); \
STEP2_ELT(1, w1, fun, s, pp4b); \
STEP2_ELT(2, w2, fun, s, pp4b); \
STEP2_ELT(3, w3, fun, s, pp4b); \
} while (0)
static void
one_round_small(u32 *state, u32 *w, int isp, int p0, int p1, int p2, int p3)
{
static const int pp4k[] = { 1, 2, 3, 1, 2, 3, 1, 2, 3, 1, 2 };
STEP2_SMALL(w[ 0], w[ 1], w[ 2], w[ 3], IF, p0, p1, pp4k[isp + 0]);
STEP2_SMALL(w[ 4], w[ 5], w[ 6], w[ 7], IF, p1, p2, pp4k[isp + 1]);
STEP2_SMALL(w[ 8], w[ 9], w[10], w[11], IF, p2, p3, pp4k[isp + 2]);
STEP2_SMALL(w[12], w[13], w[14], w[15], IF, p3, p0, pp4k[isp + 3]);
STEP2_SMALL(w[16], w[17], w[18], w[19], MAJ, p0, p1, pp4k[isp + 4]);
STEP2_SMALL(w[20], w[21], w[22], w[23], MAJ, p1, p2, pp4k[isp + 5]);
STEP2_SMALL(w[24], w[25], w[26], w[27], MAJ, p2, p3, pp4k[isp + 6]);
STEP2_SMALL(w[28], w[29], w[30], w[31], MAJ, p3, p0, pp4k[isp + 7]);
}
static void
compress_small(sph_simd_small_context *sc, int last)
{
unsigned char *x;
s32 q[128];
int i;
u32 w[32];
u32 state[16];
size_t u;
static const size_t wsp[32] = {
4 << 3, 6 << 3, 0 << 3, 2 << 3,
7 << 3, 5 << 3, 3 << 3, 1 << 3,
15 << 3, 11 << 3, 12 << 3, 8 << 3,
9 << 3, 13 << 3, 10 << 3, 14 << 3,
17 << 3, 18 << 3, 23 << 3, 20 << 3,
22 << 3, 21 << 3, 16 << 3, 19 << 3,
30 << 3, 24 << 3, 25 << 3, 31 << 3,
27 << 3, 29 << 3, 28 << 3, 26 << 3
};
x = sc->buf;
FFT128(0, 1, 0, ll);
if (last) {
for (i = 0; i < 128; i ++) {
s32 tq;
tq = q[i] + yoff_s_f[i];
tq = REDS2(tq);
tq = REDS1(tq);
tq = REDS1(tq);
q[i] = (tq <= 128 ? tq : tq - 257);
}
} else {
for (i = 0; i < 128; i ++) {
s32 tq;
tq = q[i] + yoff_s_n[i];
tq = REDS2(tq);
tq = REDS1(tq);
tq = REDS1(tq);
q[i] = (tq <= 128 ? tq : tq - 257);
}
}
for (i = 0; i < 16; i += 4) {
state[i + 0] = sc->state[i + 0]
^ sph_dec32le_aligned(x + 4 * (i + 0));
state[i + 1] = sc->state[i + 1]
^ sph_dec32le_aligned(x + 4 * (i + 1));
state[i + 2] = sc->state[i + 2]
^ sph_dec32le_aligned(x + 4 * (i + 2));
state[i + 3] = sc->state[i + 3]
^ sph_dec32le_aligned(x + 4 * (i + 3));
}
#define WSREAD(sb, o1, o2, mm) do { \
for (u = 0; u < 32; u += 4) { \
size_t v = wsp[(u >> 2) + (sb)]; \
w[u + 0] = INNER(q[v + 2 * 0 + (o1)], \
q[v + 2 * 0 + (o2)], mm); \
w[u + 1] = INNER(q[v + 2 * 1 + (o1)], \
q[v + 2 * 1 + (o2)], mm); \
w[u + 2] = INNER(q[v + 2 * 2 + (o1)], \
q[v + 2 * 2 + (o2)], mm); \
w[u + 3] = INNER(q[v + 2 * 3 + (o1)], \
q[v + 2 * 3 + (o2)], mm); \
} \
} while (0)
WSREAD( 0, 0, 1, 185);
one_round_small(state, w, 0, 3, 23, 17, 27);
WSREAD( 8, 0, 1, 185);
one_round_small(state, w, 2, 28, 19, 22, 7);
WSREAD(16, -128, -64, 233);
one_round_small(state, w, 1, 29, 9, 15, 5);
WSREAD(24, -191, -127, 233);
one_round_small(state, w, 0, 4, 13, 10, 25);
#undef WSREAD
STEP_SMALL(sc->state[ 0], sc->state[ 1], sc->state[ 2], sc->state[ 3],
IF, 4, 13, PP4_2_);
STEP_SMALL(sc->state[ 4], sc->state[ 5], sc->state[ 6], sc->state[ 7],
IF, 13, 10, PP4_0_);
STEP_SMALL(sc->state[ 8], sc->state[ 9], sc->state[10], sc->state[11],
IF, 10, 25, PP4_1_);
STEP_SMALL(sc->state[12], sc->state[13], sc->state[14], sc->state[15],
IF, 25, 4, PP4_2_);
memcpy(sc->state, state, sizeof state);
}
#undef A0
#undef A1
#undef A2
#undef A3
#undef B0
#undef B1
#undef B2
#undef B3
#undef C0
#undef C1
#undef C2
#undef C3
#undef D0
#undef D1
#undef D2
#undef D3
#else
#if SPH_SIMD_NOCOPY
#define A0 (sc->state[ 0])
#define A1 (sc->state[ 1])
#define A2 (sc->state[ 2])
#define A3 (sc->state[ 3])
#define B0 (sc->state[ 4])
#define B1 (sc->state[ 5])
#define B2 (sc->state[ 6])
#define B3 (sc->state[ 7])
#define C0 (sc->state[ 8])
#define C1 (sc->state[ 9])
#define C2 (sc->state[10])
#define C3 (sc->state[11])
#define D0 (sc->state[12])
#define D1 (sc->state[13])
#define D2 (sc->state[14])
#define D3 (sc->state[15])
#endif
static void
compress_small(sph_simd_small_context *sc, int last)
{
unsigned char *x;
s32 q[128];
int i;
DECL_STATE_SMALL
#if SPH_SIMD_NOCOPY
sph_u32 saved[16];
#endif
#if SPH_SIMD_NOCOPY
memcpy(saved, sc->state, sizeof saved);
#endif
x = sc->buf;
FFT128(0, 1, 0, ll);
if (last) {
for (i = 0; i < 128; i ++) {
s32 tq;
tq = q[i] + yoff_s_f[i];
tq = REDS2(tq);
tq = REDS1(tq);
tq = REDS1(tq);
q[i] = (tq <= 128 ? tq : tq - 257);
}
} else {
for (i = 0; i < 128; i ++) {
s32 tq;
tq = q[i] + yoff_s_n[i];
tq = REDS2(tq);
tq = REDS1(tq);
tq = REDS1(tq);
q[i] = (tq <= 128 ? tq : tq - 257);
}
}
READ_STATE_SMALL(sc);
A0 ^= sph_dec32le_aligned(x + 0);
A1 ^= sph_dec32le_aligned(x + 4);
A2 ^= sph_dec32le_aligned(x + 8);
A3 ^= sph_dec32le_aligned(x + 12);
B0 ^= sph_dec32le_aligned(x + 16);
B1 ^= sph_dec32le_aligned(x + 20);
B2 ^= sph_dec32le_aligned(x + 24);
B3 ^= sph_dec32le_aligned(x + 28);
C0 ^= sph_dec32le_aligned(x + 32);
C1 ^= sph_dec32le_aligned(x + 36);
C2 ^= sph_dec32le_aligned(x + 40);
C3 ^= sph_dec32le_aligned(x + 44);
D0 ^= sph_dec32le_aligned(x + 48);
D1 ^= sph_dec32le_aligned(x + 52);
D2 ^= sph_dec32le_aligned(x + 56);
D3 ^= sph_dec32le_aligned(x + 60);
ONE_ROUND_SMALL(0_, 0, 3, 23, 17, 27);
ONE_ROUND_SMALL(1_, 2, 28, 19, 22, 7);
ONE_ROUND_SMALL(2_, 1, 29, 9, 15, 5);
ONE_ROUND_SMALL(3_, 0, 4, 13, 10, 25);
#if SPH_SIMD_NOCOPY
STEP_SMALL(saved[ 0], saved[ 1], saved[ 2], saved[ 3],
IF, 4, 13, PP4_2_);
STEP_SMALL(saved[ 4], saved[ 5], saved[ 6], saved[ 7],
IF, 13, 10, PP4_0_);
STEP_SMALL(saved[ 8], saved[ 9], saved[10], saved[11],
IF, 10, 25, PP4_1_);
STEP_SMALL(saved[12], saved[13], saved[14], saved[15],
IF, 25, 4, PP4_2_);
#else
STEP_SMALL(sc->state[ 0], sc->state[ 1], sc->state[ 2], sc->state[ 3],
IF, 4, 13, PP4_2_);
STEP_SMALL(sc->state[ 4], sc->state[ 5], sc->state[ 6], sc->state[ 7],
IF, 13, 10, PP4_0_);
STEP_SMALL(sc->state[ 8], sc->state[ 9], sc->state[10], sc->state[11],
IF, 10, 25, PP4_1_);
STEP_SMALL(sc->state[12], sc->state[13], sc->state[14], sc->state[15],
IF, 25, 4, PP4_2_);
WRITE_STATE_SMALL(sc);
#endif
}
#if SPH_SIMD_NOCOPY
#undef A0
#undef A1
#undef A2
#undef A3
#undef B0
#undef B1
#undef B2
#undef B3
#undef C0
#undef C1
#undef C2
#undef C3
#undef D0
#undef D1
#undef D2
#undef D3
#endif
#endif
#if SPH_SMALL_FOOTPRINT_SIMD
#define A0 state[ 0]
#define A1 state[ 1]
#define A2 state[ 2]
#define A3 state[ 3]
#define A4 state[ 4]
#define A5 state[ 5]
#define A6 state[ 6]
#define A7 state[ 7]
#define B0 state[ 8]
#define B1 state[ 9]
#define B2 state[10]
#define B3 state[11]
#define B4 state[12]
#define B5 state[13]
#define B6 state[14]
#define B7 state[15]
#define C0 state[16]
#define C1 state[17]
#define C2 state[18]
#define C3 state[19]
#define C4 state[20]
#define C5 state[21]
#define C6 state[22]
#define C7 state[23]
#define D0 state[24]
#define D1 state[25]
#define D2 state[26]
#define D3 state[27]
#define D4 state[28]
#define D5 state[29]
#define D6 state[30]
#define D7 state[31]
/*
* Not needed -- already defined for SIMD-224 / SIMD-256
*
#define STEP2_ELT(n, w, fun, s, ppb) do { \
u32 tt = T32(D ## n + (w) + fun(A ## n, B ## n, C ## n)); \
A ## n = T32(ROL32(tt, s) + tA[(ppb) ^ n]); \
D ## n = C ## n; \
C ## n = B ## n; \
B ## n = tA[n]; \
} while (0)
*/
#define STEP2_BIG(w0, w1, w2, w3, w4, w5, w6, w7, fun, r, s, pp8b) do { \
u32 tA[8]; \
tA[0] = ROL32(A0, r); \
tA[1] = ROL32(A1, r); \
tA[2] = ROL32(A2, r); \
tA[3] = ROL32(A3, r); \
tA[4] = ROL32(A4, r); \
tA[5] = ROL32(A5, r); \
tA[6] = ROL32(A6, r); \
tA[7] = ROL32(A7, r); \
STEP2_ELT(0, w0, fun, s, pp8b); \
STEP2_ELT(1, w1, fun, s, pp8b); \
STEP2_ELT(2, w2, fun, s, pp8b); \
STEP2_ELT(3, w3, fun, s, pp8b); \
STEP2_ELT(4, w4, fun, s, pp8b); \
STEP2_ELT(5, w5, fun, s, pp8b); \
STEP2_ELT(6, w6, fun, s, pp8b); \
STEP2_ELT(7, w7, fun, s, pp8b); \
} while (0)
static void
one_round_big(u32 *state, u32 *w, int isp, int p0, int p1, int p2, int p3)
{
static const int pp8k[] = { 1, 6, 2, 3, 5, 7, 4, 1, 6, 2, 3 };
STEP2_BIG(w[ 0], w[ 1], w[ 2], w[ 3], w[ 4], w[ 5], w[ 6], w[ 7],
IF, p0, p1, pp8k[isp + 0]);
STEP2_BIG(w[ 8], w[ 9], w[10], w[11], w[12], w[13], w[14], w[15],
IF, p1, p2, pp8k[isp + 1]);
STEP2_BIG(w[16], w[17], w[18], w[19], w[20], w[21], w[22], w[23],
IF, p2, p3, pp8k[isp + 2]);
STEP2_BIG(w[24], w[25], w[26], w[27], w[28], w[29], w[30], w[31],
IF, p3, p0, pp8k[isp + 3]);
STEP2_BIG(w[32], w[33], w[34], w[35], w[36], w[37], w[38], w[39],
MAJ, p0, p1, pp8k[isp + 4]);
STEP2_BIG(w[40], w[41], w[42], w[43], w[44], w[45], w[46], w[47],
MAJ, p1, p2, pp8k[isp + 5]);
STEP2_BIG(w[48], w[49], w[50], w[51], w[52], w[53], w[54], w[55],
MAJ, p2, p3, pp8k[isp + 6]);
STEP2_BIG(w[56], w[57], w[58], w[59], w[60], w[61], w[62], w[63],
MAJ, p3, p0, pp8k[isp + 7]);
}
static void
compress_big(sph_simd_big_context *sc, int last)
{
unsigned char *x;
s32 q[256];
int i;
u32 w[64];
u32 state[32];
size_t u;
static const size_t wbp[32] = {
4 << 4, 6 << 4, 0 << 4, 2 << 4,
7 << 4, 5 << 4, 3 << 4, 1 << 4,
15 << 4, 11 << 4, 12 << 4, 8 << 4,
9 << 4, 13 << 4, 10 << 4, 14 << 4,
17 << 4, 18 << 4, 23 << 4, 20 << 4,
22 << 4, 21 << 4, 16 << 4, 19 << 4,
30 << 4, 24 << 4, 25 << 4, 31 << 4,
27 << 4, 29 << 4, 28 << 4, 26 << 4
};
x = sc->buf;
FFT256(0, 1, 0, ll);
if (last) {
for (i = 0; i < 256; i ++) {
s32 tq;
tq = q[i] + yoff_b_f[i];
tq = REDS2(tq);
tq = REDS1(tq);
tq = REDS1(tq);
q[i] = (tq <= 128 ? tq : tq - 257);
}
} else {
for (i = 0; i < 256; i ++) {
s32 tq;
tq = q[i] + yoff_b_n[i];
tq = REDS2(tq);
tq = REDS1(tq);
tq = REDS1(tq);
q[i] = (tq <= 128 ? tq : tq - 257);
}
}
for (i = 0; i < 32; i += 8) {
state[i + 0] = sc->state[i + 0]
^ sph_dec32le_aligned(x + 4 * (i + 0));
state[i + 1] = sc->state[i + 1]
^ sph_dec32le_aligned(x + 4 * (i + 1));
state[i + 2] = sc->state[i + 2]
^ sph_dec32le_aligned(x + 4 * (i + 2));
state[i + 3] = sc->state[i + 3]
^ sph_dec32le_aligned(x + 4 * (i + 3));
state[i + 4] = sc->state[i + 4]
^ sph_dec32le_aligned(x + 4 * (i + 4));
state[i + 5] = sc->state[i + 5]
^ sph_dec32le_aligned(x + 4 * (i + 5));
state[i + 6] = sc->state[i + 6]
^ sph_dec32le_aligned(x + 4 * (i + 6));
state[i + 7] = sc->state[i + 7]
^ sph_dec32le_aligned(x + 4 * (i + 7));
}
#define WBREAD(sb, o1, o2, mm) do { \
for (u = 0; u < 64; u += 8) { \
size_t v = wbp[(u >> 3) + (sb)]; \
w[u + 0] = INNER(q[v + 2 * 0 + (o1)], \
q[v + 2 * 0 + (o2)], mm); \
w[u + 1] = INNER(q[v + 2 * 1 + (o1)], \
q[v + 2 * 1 + (o2)], mm); \
w[u + 2] = INNER(q[v + 2 * 2 + (o1)], \
q[v + 2 * 2 + (o2)], mm); \
w[u + 3] = INNER(q[v + 2 * 3 + (o1)], \
q[v + 2 * 3 + (o2)], mm); \
w[u + 4] = INNER(q[v + 2 * 4 + (o1)], \
q[v + 2 * 4 + (o2)], mm); \
w[u + 5] = INNER(q[v + 2 * 5 + (o1)], \
q[v + 2 * 5 + (o2)], mm); \
w[u + 6] = INNER(q[v + 2 * 6 + (o1)], \
q[v + 2 * 6 + (o2)], mm); \
w[u + 7] = INNER(q[v + 2 * 7 + (o1)], \
q[v + 2 * 7 + (o2)], mm); \
} \
} while (0)
WBREAD( 0, 0, 1, 185);
one_round_big(state, w, 0, 3, 23, 17, 27);
WBREAD( 8, 0, 1, 185);
one_round_big(state, w, 1, 28, 19, 22, 7);
WBREAD(16, -256, -128, 233);
one_round_big(state, w, 2, 29, 9, 15, 5);
WBREAD(24, -383, -255, 233);
one_round_big(state, w, 3, 4, 13, 10, 25);
#undef WBREAD
STEP_BIG(
sc->state[ 0], sc->state[ 1], sc->state[ 2], sc->state[ 3],
sc->state[ 4], sc->state[ 5], sc->state[ 6], sc->state[ 7],
IF, 4, 13, PP8_4_);
STEP_BIG(
sc->state[ 8], sc->state[ 9], sc->state[10], sc->state[11],
sc->state[12], sc->state[13], sc->state[14], sc->state[15],
IF, 13, 10, PP8_5_);
STEP_BIG(
sc->state[16], sc->state[17], sc->state[18], sc->state[19],
sc->state[20], sc->state[21], sc->state[22], sc->state[23],
IF, 10, 25, PP8_6_);
STEP_BIG(
sc->state[24], sc->state[25], sc->state[26], sc->state[27],
sc->state[28], sc->state[29], sc->state[30], sc->state[31],
IF, 25, 4, PP8_0_);
memcpy(sc->state, state, sizeof state);
}
#undef A0
#undef A1
#undef A2
#undef A3
#undef A4
#undef A5
#undef A6
#undef A7
#undef B0
#undef B1
#undef B2
#undef B3
#undef B4
#undef B5
#undef B6
#undef B7
#undef C0
#undef C1
#undef C2
#undef C3
#undef C4
#undef C5
#undef C6
#undef C7
#undef D0
#undef D1
#undef D2
#undef D3
#undef D4
#undef D5
#undef D6
#undef D7
#else
#if SPH_SIMD_NOCOPY
#define A0 (sc->state[ 0])
#define A1 (sc->state[ 1])
#define A2 (sc->state[ 2])
#define A3 (sc->state[ 3])
#define A4 (sc->state[ 4])
#define A5 (sc->state[ 5])
#define A6 (sc->state[ 6])
#define A7 (sc->state[ 7])
#define B0 (sc->state[ 8])
#define B1 (sc->state[ 9])
#define B2 (sc->state[10])
#define B3 (sc->state[11])
#define B4 (sc->state[12])
#define B5 (sc->state[13])
#define B6 (sc->state[14])
#define B7 (sc->state[15])
#define C0 (sc->state[16])
#define C1 (sc->state[17])
#define C2 (sc->state[18])
#define C3 (sc->state[19])
#define C4 (sc->state[20])
#define C5 (sc->state[21])
#define C6 (sc->state[22])
#define C7 (sc->state[23])
#define D0 (sc->state[24])
#define D1 (sc->state[25])
#define D2 (sc->state[26])
#define D3 (sc->state[27])
#define D4 (sc->state[28])
#define D5 (sc->state[29])
#define D6 (sc->state[30])
#define D7 (sc->state[31])
#endif
static void
compress_big(sph_simd_big_context *sc, int last)
{
unsigned char *x;
s32 q[256];
int i;
DECL_STATE_BIG
#if SPH_SIMD_NOCOPY
sph_u32 saved[32];
#endif
#if SPH_SIMD_NOCOPY
memcpy(saved, sc->state, sizeof saved);
#endif
x = sc->buf;
FFT256(0, 1, 0, ll);
if (last) {
for (i = 0; i < 256; i ++) {
s32 tq;
tq = q[i] + yoff_b_f[i];
tq = REDS2(tq);
tq = REDS1(tq);
tq = REDS1(tq);
q[i] = (tq <= 128 ? tq : tq - 257);
}
} else {
for (i = 0; i < 256; i ++) {
s32 tq;
tq = q[i] + yoff_b_n[i];
tq = REDS2(tq);
tq = REDS1(tq);
tq = REDS1(tq);
q[i] = (tq <= 128 ? tq : tq - 257);
}
}
READ_STATE_BIG(sc);
A0 ^= sph_dec32le_aligned(x + 0);
A1 ^= sph_dec32le_aligned(x + 4);
A2 ^= sph_dec32le_aligned(x + 8);
A3 ^= sph_dec32le_aligned(x + 12);
A4 ^= sph_dec32le_aligned(x + 16);
A5 ^= sph_dec32le_aligned(x + 20);
A6 ^= sph_dec32le_aligned(x + 24);
A7 ^= sph_dec32le_aligned(x + 28);
B0 ^= sph_dec32le_aligned(x + 32);
B1 ^= sph_dec32le_aligned(x + 36);
B2 ^= sph_dec32le_aligned(x + 40);
B3 ^= sph_dec32le_aligned(x + 44);
B4 ^= sph_dec32le_aligned(x + 48);
B5 ^= sph_dec32le_aligned(x + 52);
B6 ^= sph_dec32le_aligned(x + 56);
B7 ^= sph_dec32le_aligned(x + 60);
C0 ^= sph_dec32le_aligned(x + 64);
C1 ^= sph_dec32le_aligned(x + 68);
C2 ^= sph_dec32le_aligned(x + 72);
C3 ^= sph_dec32le_aligned(x + 76);
C4 ^= sph_dec32le_aligned(x + 80);
C5 ^= sph_dec32le_aligned(x + 84);
C6 ^= sph_dec32le_aligned(x + 88);
C7 ^= sph_dec32le_aligned(x + 92);
D0 ^= sph_dec32le_aligned(x + 96);
D1 ^= sph_dec32le_aligned(x + 100);
D2 ^= sph_dec32le_aligned(x + 104);
D3 ^= sph_dec32le_aligned(x + 108);
D4 ^= sph_dec32le_aligned(x + 112);
D5 ^= sph_dec32le_aligned(x + 116);
D6 ^= sph_dec32le_aligned(x + 120);
D7 ^= sph_dec32le_aligned(x + 124);
ONE_ROUND_BIG(0_, 0, 3, 23, 17, 27);
ONE_ROUND_BIG(1_, 1, 28, 19, 22, 7);
ONE_ROUND_BIG(2_, 2, 29, 9, 15, 5);
ONE_ROUND_BIG(3_, 3, 4, 13, 10, 25);
#if SPH_SIMD_NOCOPY
STEP_BIG(
saved[ 0], saved[ 1], saved[ 2], saved[ 3],
saved[ 4], saved[ 5], saved[ 6], saved[ 7],
IF, 4, 13, PP8_4_);
STEP_BIG(
saved[ 8], saved[ 9], saved[10], saved[11],
saved[12], saved[13], saved[14], saved[15],
IF, 13, 10, PP8_5_);
STEP_BIG(
saved[16], saved[17], saved[18], saved[19],
saved[20], saved[21], saved[22], saved[23],
IF, 10, 25, PP8_6_);
STEP_BIG(
saved[24], saved[25], saved[26], saved[27],
saved[28], saved[29], saved[30], saved[31],
IF, 25, 4, PP8_0_);
#else
STEP_BIG(
sc->state[ 0], sc->state[ 1], sc->state[ 2], sc->state[ 3],
sc->state[ 4], sc->state[ 5], sc->state[ 6], sc->state[ 7],
IF, 4, 13, PP8_4_);
STEP_BIG(
sc->state[ 8], sc->state[ 9], sc->state[10], sc->state[11],
sc->state[12], sc->state[13], sc->state[14], sc->state[15],
IF, 13, 10, PP8_5_);
STEP_BIG(
sc->state[16], sc->state[17], sc->state[18], sc->state[19],
sc->state[20], sc->state[21], sc->state[22], sc->state[23],
IF, 10, 25, PP8_6_);
STEP_BIG(
sc->state[24], sc->state[25], sc->state[26], sc->state[27],
sc->state[28], sc->state[29], sc->state[30], sc->state[31],
IF, 25, 4, PP8_0_);
WRITE_STATE_BIG(sc);
#endif
}
#if SPH_SIMD_NOCOPY
#undef A0
#undef A1
#undef A2
#undef A3
#undef A4
#undef A5
#undef A6
#undef A7
#undef B0
#undef B1
#undef B2
#undef B3
#undef B4
#undef B5
#undef B6
#undef B7
#undef C0
#undef C1
#undef C2
#undef C3
#undef C4
#undef C5
#undef C6
#undef C7
#undef D0
#undef D1
#undef D2
#undef D3
#undef D4
#undef D5
#undef D6
#undef D7
#endif
#endif
static const u32 IV224[] = {
C32(0x33586E9F), C32(0x12FFF033), C32(0xB2D9F64D), C32(0x6F8FEA53),
C32(0xDE943106), C32(0x2742E439), C32(0x4FBAB5AC), C32(0x62B9FF96),
C32(0x22E7B0AF), C32(0xC862B3A8), C32(0x33E00CDC), C32(0x236B86A6),
C32(0xF64AE77C), C32(0xFA373B76), C32(0x7DC1EE5B), C32(0x7FB29CE8)
};
static const u32 IV256[] = {
C32(0x4D567983), C32(0x07190BA9), C32(0x8474577B), C32(0x39D726E9),
C32(0xAAF3D925), C32(0x3EE20B03), C32(0xAFD5E751), C32(0xC96006D3),
C32(0xC2C2BA14), C32(0x49B3BCB4), C32(0xF67CAF46), C32(0x668626C9),
C32(0xE2EAA8D2), C32(0x1FF47833), C32(0xD0C661A5), C32(0x55693DE1)
};
static const u32 IV384[] = {
C32(0x8A36EEBC), C32(0x94A3BD90), C32(0xD1537B83), C32(0xB25B070B),
C32(0xF463F1B5), C32(0xB6F81E20), C32(0x0055C339), C32(0xB4D144D1),
C32(0x7360CA61), C32(0x18361A03), C32(0x17DCB4B9), C32(0x3414C45A),
C32(0xA699A9D2), C32(0xE39E9664), C32(0x468BFE77), C32(0x51D062F8),
C32(0xB9E3BFE8), C32(0x63BECE2A), C32(0x8FE506B9), C32(0xF8CC4AC2),
C32(0x7AE11542), C32(0xB1AADDA1), C32(0x64B06794), C32(0x28D2F462),
C32(0xE64071EC), C32(0x1DEB91A8), C32(0x8AC8DB23), C32(0x3F782AB5),
C32(0x039B5CB8), C32(0x71DDD962), C32(0xFADE2CEA), C32(0x1416DF71)
};
static const u32 IV512[] = {
C32(0x0BA16B95), C32(0x72F999AD), C32(0x9FECC2AE), C32(0xBA3264FC),
C32(0x5E894929), C32(0x8E9F30E5), C32(0x2F1DAA37), C32(0xF0F2C558),
C32(0xAC506643), C32(0xA90635A5), C32(0xE25B878B), C32(0xAAB7878F),
C32(0x88817F7A), C32(0x0A02892B), C32(0x559A7550), C32(0x598F657E),
C32(0x7EEF60A1), C32(0x6B70E3E8), C32(0x9C1714D1), C32(0xB958E2A8),
C32(0xAB02675E), C32(0xED1C014F), C32(0xCD8D65BB), C32(0xFDB7A257),
C32(0x09254899), C32(0xD699C7BC), C32(0x9019B6DC), C32(0x2B9022E4),
C32(0x8FA14956), C32(0x21BF9BD3), C32(0xB94D0943), C32(0x6FFDDC22)
};
static void
init_small(void *cc, const u32 *iv)
{
sph_simd_small_context *sc;
sc = cc;
memcpy(sc->state, iv, sizeof sc->state);
sc->count_low = sc->count_high = 0;
sc->ptr = 0;
}
static void
init_big(void *cc, const u32 *iv)
{
sph_simd_big_context *sc;
sc = cc;
memcpy(sc->state, iv, sizeof sc->state);
sc->count_low = sc->count_high = 0;
sc->ptr = 0;
}
static void
update_small(void *cc, const void *data, size_t len)
{
sph_simd_small_context *sc;
sc = cc;
while (len > 0) {
size_t clen;
clen = (sizeof sc->buf) - sc->ptr;
if (clen > len)
clen = len;
memcpy(sc->buf + sc->ptr, data, clen);
data = (const unsigned char *)data + clen;
len -= clen;
if ((sc->ptr += clen) == sizeof sc->buf) {
compress_small(sc, 0);
sc->ptr = 0;
sc->count_low = T32(sc->count_low + 1);
if (sc->count_low == 0)
sc->count_high ++;
}
}
}
static void
update_big(void *cc, const void *data, size_t len)
{
sph_simd_big_context *sc;
sc = cc;
while (len > 0) {
size_t clen;
clen = (sizeof sc->buf) - sc->ptr;
if (clen > len)
clen = len;
memcpy(sc->buf + sc->ptr, data, clen);
data = (const unsigned char *)data + clen;
len -= clen;
if ((sc->ptr += clen) == sizeof sc->buf) {
compress_big(sc, 0);
sc->ptr = 0;
sc->count_low = T32(sc->count_low + 1);
if (sc->count_low == 0)
sc->count_high ++;
}
}
}
static void
encode_count_small(unsigned char *dst,
u32 low, u32 high, size_t ptr, unsigned n)
{
low = T32(low << 9);
high = T32(high << 9) + (low >> 23);
low += (ptr << 3) + n;
sph_enc32le(dst, low);
sph_enc32le(dst + 4, high);
}
static void
encode_count_big(unsigned char *dst,
u32 low, u32 high, size_t ptr, unsigned n)
{
low = T32(low << 10);
high = T32(high << 10) + (low >> 22);
low += (ptr << 3) + n;
sph_enc32le(dst, low);
sph_enc32le(dst + 4, high);
}
static void
finalize_small(void *cc, unsigned ub, unsigned n, void *dst, size_t dst_len)
{
sph_simd_small_context *sc;
unsigned char *d;
size_t u;
sc = cc;
if (sc->ptr > 0 || n > 0) {
memset(sc->buf + sc->ptr, 0,
(sizeof sc->buf) - sc->ptr);
sc->buf[sc->ptr] = ub & (0xFF << (8 - n));
compress_small(sc, 0);
}
memset(sc->buf, 0, sizeof sc->buf);
encode_count_small(sc->buf, sc->count_low, sc->count_high, sc->ptr, n);
compress_small(sc, 1);
d = dst;
for (d = dst, u = 0; u < dst_len; u ++)
sph_enc32le(d + (u << 2), sc->state[u]);
}
static void
finalize_big(void *cc, unsigned ub, unsigned n, void *dst, size_t dst_len)
{
sph_simd_big_context *sc;
unsigned char *d;
size_t u;
sc = cc;
if (sc->ptr > 0 || n > 0) {
memset(sc->buf + sc->ptr, 0,
(sizeof sc->buf) - sc->ptr);
sc->buf[sc->ptr] = ub & (0xFF << (8 - n));
compress_big(sc, 0);
}
memset(sc->buf, 0, sizeof sc->buf);
encode_count_big(sc->buf, sc->count_low, sc->count_high, sc->ptr, n);
compress_big(sc, 1);
d = dst;
for (d = dst, u = 0; u < dst_len; u ++)
sph_enc32le(d + (u << 2), sc->state[u]);
}
void
sph_simd224_init(void *cc)
{
init_small(cc, IV224);
}
void
sph_simd224(void *cc, const void *data, size_t len)
{
update_small(cc, data, len);
}
void
sph_simd224_close(void *cc, void *dst)
{
sph_simd224_addbits_and_close(cc, 0, 0, dst);
}
void
sph_simd224_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
{
finalize_small(cc, ub, n, dst, 7);
sph_simd224_init(cc);
}
void
sph_simd256_init(void *cc)
{
init_small(cc, IV256);
}
void
sph_simd256(void *cc, const void *data, size_t len)
{
update_small(cc, data, len);
}
void
sph_simd256_close(void *cc, void *dst)
{
sph_simd256_addbits_and_close(cc, 0, 0, dst);
}
void
sph_simd256_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
{
finalize_small(cc, ub, n, dst, 8);
sph_simd256_init(cc);
}
void
sph_simd384_init(void *cc)
{
init_big(cc, IV384);
}
void
sph_simd384(void *cc, const void *data, size_t len)
{
update_big(cc, data, len);
}
void
sph_simd384_close(void *cc, void *dst)
{
sph_simd384_addbits_and_close(cc, 0, 0, dst);
}
void
sph_simd384_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
{
finalize_big(cc, ub, n, dst, 12);
sph_simd384_init(cc);
}
void
sph_simd512_init(void *cc)
{
init_big(cc, IV512);
}
void
sph_simd512(void *cc, const void *data, size_t len)
{
update_big(cc, data, len);
}
void
sph_simd512_close(void *cc, void *dst)
{
sph_simd512_addbits_and_close(cc, 0, 0, dst);
}
void
sph_simd512_addbits_and_close(void *cc, unsigned ub, unsigned n, void *dst)
{
finalize_big(cc, ub, n, dst, 16);
sph_simd512_init(cc);
}
#ifdef __cplusplus
}
#endif
Reference in New Issue View Git Blame Copy Permalink