From bc635b0869624eca595ac04a406b1dc7f8b21ecb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?B=C3=A9n=C3=A9dikt=20Tran?= <10796600+picnixz@users.noreply.github.com> Date: Mon, 23 Jun 2025 14:13:06 +0200 Subject: [PATCH] rename undocumented `HACL_CAN_COMPILE_SIMD{128,256}` macros --- Modules/blake2module.c | 66 +++++++++++++++++++------------------- Modules/hmacmodule.c | 21 ++++++------ PCbuild/pythoncore.vcxproj | 8 +++-- configure | 4 +-- configure.ac | 6 ++-- pyconfig.h.in | 12 +++---- 6 files changed, 62 insertions(+), 55 deletions(-) diff --git a/Modules/blake2module.c b/Modules/blake2module.c index 6c4349ac06bb8a..68225a2f4c7edc 100644 --- a/Modules/blake2module.c +++ b/Modules/blake2module.c @@ -43,25 +43,25 @@ // SIMD256 can't be compiled on macOS ARM64, and performance of SIMD128 isn't // great; but when compiling a universal2 binary, autoconf will set -// HACL_CAN_COMPILE_SIMD128 and HACL_CAN_COMPILE_SIMD256 because they *can* be -// compiled on x86_64. If we're on macOS ARM64, disable these preprocessor -// symbols. +// _Py_HACL_CAN_COMPILE_VEC{128,256} because they *can* be compiled on x86_64. +// If we're on macOS ARM64, we however disable these preprocessor symbols. #if defined(__APPLE__) && defined(__arm64__) -# undef HACL_CAN_COMPILE_SIMD128 -# undef HACL_CAN_COMPILE_SIMD256 +# undef _Py_HACL_CAN_COMPILE_VEC128 +# undef _Py_HACL_CAN_COMPILE_VEC256 #endif -// Small mismatch between the variable names Python defines as part of configure -// at the ones HACL* expects to be set in order to enable those headers. -#define HACL_CAN_COMPILE_VEC128 HACL_CAN_COMPILE_SIMD128 -#define HACL_CAN_COMPILE_VEC256 HACL_CAN_COMPILE_SIMD256 +// HACL* expects HACL_CAN_COMPILE_VEC* macros to be set in order to enable +// the corresponding SIMD instructions so we need to "forward" the values +// we just deduced above. +#define HACL_CAN_COMPILE_VEC128 _Py_HACL_CAN_COMPILE_VEC128 +#define HACL_CAN_COMPILE_VEC256 _Py_HACL_CAN_COMPILE_VEC256 #include "_hacl/Hacl_Hash_Blake2s.h" #include "_hacl/Hacl_Hash_Blake2b.h" -#if HACL_CAN_COMPILE_SIMD128 +#if _Py_HACL_CAN_COMPILE_VEC128 #include "_hacl/Hacl_Hash_Blake2s_Simd128.h" #endif -#if HACL_CAN_COMPILE_SIMD256 +#if _Py_HACL_CAN_COMPILE_VEC256 #include "_hacl/Hacl_Hash_Blake2b_Simd256.h" #endif @@ -88,7 +88,7 @@ blake2_get_state(PyObject *module) return (Blake2State *)state; } -#if defined(HACL_CAN_COMPILE_SIMD128) || defined(HACL_CAN_COMPILE_SIMD256) +#if defined(_Py_HACL_CAN_COMPILE_VEC128) || defined(_Py_HACL_CAN_COMPILE_VEC256) static inline Blake2State * blake2_get_state_from_type(PyTypeObject *module) { @@ -181,7 +181,7 @@ blake2module_init_cpu_features(Blake2State *state) #undef ECX_SSE3 #undef EBX_AVX2 -#if HACL_CAN_COMPILE_SIMD128 +#if _Py_HACL_CAN_COMPILE_VEC128 // TODO(picnixz): use py_cpuid_features (gh-125022) to improve detection state->can_run_simd128 = sse && sse2 && sse3 && sse41 && sse42 && cmov; #else @@ -191,7 +191,7 @@ blake2module_init_cpu_features(Blake2State *state) state->can_run_simd128 = false; #endif -#if HACL_CAN_COMPILE_SIMD256 +#if _Py_HACL_CAN_COMPILE_VEC256 // TODO(picnixz): use py_cpuid_features (gh-125022) to improve detection state->can_run_simd256 = state->can_run_simd128 && avx && avx2; #else @@ -332,18 +332,18 @@ is_blake2s(blake2_impl impl) static inline blake2_impl type_to_impl(PyTypeObject *type) { -#if defined(HACL_CAN_COMPILE_SIMD128) || defined(HACL_CAN_COMPILE_SIMD256) +#if defined(_Py_HACL_CAN_COMPILE_VEC128) || defined(_Py_HACL_CAN_COMPILE_VEC256) Blake2State *st = blake2_get_state_from_type(type); #endif if (!strcmp(type->tp_name, blake2b_type_spec.name)) { -#if HACL_CAN_COMPILE_SIMD256 +#if _Py_HACL_CAN_COMPILE_VEC256 return st->can_run_simd256 ? Blake2b_256 : Blake2b; #else return Blake2b; #endif } else if (!strcmp(type->tp_name, blake2s_type_spec.name)) { -#if HACL_CAN_COMPILE_SIMD128 +#if _Py_HACL_CAN_COMPILE_VEC128 return st->can_run_simd128 ? Blake2s_128 : Blake2s; #else return Blake2s; @@ -357,10 +357,10 @@ typedef struct { union { Hacl_Hash_Blake2s_state_t *blake2s_state; Hacl_Hash_Blake2b_state_t *blake2b_state; -#if HACL_CAN_COMPILE_SIMD128 +#if _Py_HACL_CAN_COMPILE_VEC128 Hacl_Hash_Blake2s_Simd128_state_t *blake2s_128_state; #endif -#if HACL_CAN_COMPILE_SIMD256 +#if _Py_HACL_CAN_COMPILE_VEC256 Hacl_Hash_Blake2b_Simd256_state_t *blake2b_256_state; #endif }; @@ -429,13 +429,13 @@ blake2_update_unlocked(Blake2Object *self, uint8_t *buf, Py_ssize_t len) switch (self->impl) { // blake2b_256_state and blake2s_128_state must be if'd since // otherwise this results in an unresolved symbol at link-time. -#if HACL_CAN_COMPILE_SIMD256 +#if _Py_HACL_CAN_COMPILE_VEC256 case Blake2b_256: HACL_UPDATE(Hacl_Hash_Blake2b_Simd256_update, self->blake2b_256_state, buf, len); return; #endif -#if HACL_CAN_COMPILE_SIMD128 +#if _Py_HACL_CAN_COMPILE_VEC128 case Blake2s_128: HACL_UPDATE(Hacl_Hash_Blake2s_Simd128_update, self->blake2s_128_state, buf, len); @@ -555,12 +555,12 @@ py_blake2_new(PyTypeObject *type, PyObject *data, int digest_size, // Ensure that the states are NULL-initialized in case of an error. // See: py_blake2_clear() for more details. switch (self->impl) { -#if HACL_CAN_COMPILE_SIMD256 +#if _Py_HACL_CAN_COMPILE_VEC256 case Blake2b_256: self->blake2b_256_state = NULL; break; #endif -#if HACL_CAN_COMPILE_SIMD128 +#if _Py_HACL_CAN_COMPILE_VEC128 case Blake2s_128: self->blake2s_128_state = NULL; break; @@ -623,12 +623,12 @@ py_blake2_new(PyTypeObject *type, PyObject *data, int digest_size, } while (0) switch (self->impl) { -#if HACL_CAN_COMPILE_SIMD256 +#if _Py_HACL_CAN_COMPILE_VEC256 case Blake2b_256: BLAKE2_MALLOC(Blake2b_Simd256, self->blake2b_256_state); break; #endif -#if HACL_CAN_COMPILE_SIMD128 +#if _Py_HACL_CAN_COMPILE_VEC128 case Blake2s_128: BLAKE2_MALLOC(Blake2s_Simd128, self->blake2s_128_state); break; @@ -756,12 +756,12 @@ blake2_blake2b_copy_unlocked(Blake2Object *self, Blake2Object *cpy) } while (0) switch (self->impl) { -#if HACL_CAN_COMPILE_SIMD256 +#if _Py_HACL_CAN_COMPILE_VEC256 case Blake2b_256: BLAKE2_COPY(Blake2b_Simd256, blake2b_256_state); break; #endif -#if HACL_CAN_COMPILE_SIMD128 +#if _Py_HACL_CAN_COMPILE_VEC128 case Blake2s_128: BLAKE2_COPY(Blake2s_Simd128, blake2s_128_state); break; @@ -838,12 +838,12 @@ static uint8_t blake2_blake2b_compute_digest(Blake2Object *self, uint8_t *digest) { switch (self->impl) { -#if HACL_CAN_COMPILE_SIMD256 +#if _Py_HACL_CAN_COMPILE_VEC256 case Blake2b_256: return Hacl_Hash_Blake2b_Simd256_digest( self->blake2b_256_state, digest); #endif -#if HACL_CAN_COMPILE_SIMD128 +#if _Py_HACL_CAN_COMPILE_VEC128 case Blake2s_128: return Hacl_Hash_Blake2s_Simd128_digest( self->blake2s_128_state, digest); @@ -921,11 +921,11 @@ static Hacl_Hash_Blake2b_index hacl_get_blake2_info(Blake2Object *self) { switch (self->impl) { -#if HACL_CAN_COMPILE_SIMD256 +#if _Py_HACL_CAN_COMPILE_VEC256 case Blake2b_256: return Hacl_Hash_Blake2b_Simd256_info(self->blake2b_256_state); #endif -#if HACL_CAN_COMPILE_SIMD128 +#if _Py_HACL_CAN_COMPILE_VEC128 case Blake2s_128: return Hacl_Hash_Blake2s_Simd128_info(self->blake2s_128_state); #endif @@ -973,12 +973,12 @@ py_blake2_clear(PyObject *op) } while (0) switch (self->impl) { -#if HACL_CAN_COMPILE_SIMD256 +#if _Py_HACL_CAN_COMPILE_VEC256 case Blake2b_256: BLAKE2_FREE(Blake2b_Simd256, self->blake2b_256_state); break; #endif -#if HACL_CAN_COMPILE_SIMD128 +#if _Py_HACL_CAN_COMPILE_VEC128 case Blake2s_128: BLAKE2_FREE(Blake2s_Simd128, self->blake2s_128_state); break; diff --git a/Modules/hmacmodule.c b/Modules/hmacmodule.c index e7a5ccbb19b45c..95e400231bb65c 100644 --- a/Modules/hmacmodule.c +++ b/Modules/hmacmodule.c @@ -31,14 +31,15 @@ #endif #if defined(__APPLE__) && defined(__arm64__) -# undef HACL_CAN_COMPILE_SIMD128 -# undef HACL_CAN_COMPILE_SIMD256 +# undef _Py_HACL_CAN_COMPILE_VEC128 +# undef _Py_HACL_CAN_COMPILE_VEC256 #endif -// Small mismatch between the variable names Python defines as part of configure -// at the ones HACL* expects to be set in order to enable those headers. -#define HACL_CAN_COMPILE_VEC128 HACL_CAN_COMPILE_SIMD128 -#define HACL_CAN_COMPILE_VEC256 HACL_CAN_COMPILE_SIMD256 +// HACL* expects HACL_CAN_COMPILE_VEC* macros to be set in order to enable +// the corresponding SIMD instructions so we need to "forward" the values +// we just deduced above. +#define HACL_CAN_COMPILE_VEC128 _Py_HACL_CAN_COMPILE_VEC128 +#define HACL_CAN_COMPILE_VEC256 _Py_HACL_CAN_COMPILE_VEC256 #include "_hacl/Hacl_HMAC.h" #include "_hacl/Hacl_Streaming_HMAC.h" // Hacl_Agile_Hash_* identifiers @@ -361,7 +362,7 @@ narrow_hmac_hash_kind(hmacmodule_state *state, HMAC_Hash_Kind kind) { switch (kind) { case Py_hmac_kind_hmac_blake2s_32: { -#if HACL_CAN_COMPILE_SIMD128 +#if _Py_HACL_CAN_COMPILE_VEC128 if (state->can_run_simd128) { return Py_hmac_kind_hmac_vectorized_blake2s_32; } @@ -369,7 +370,7 @@ narrow_hmac_hash_kind(hmacmodule_state *state, HMAC_Hash_Kind kind) return kind; } case Py_hmac_kind_hmac_blake2b_32: { -#if HACL_CAN_COMPILE_SIMD256 +#if _Py_HACL_CAN_COMPILE_VEC256 if (state->can_run_simd256) { return Py_hmac_kind_hmac_vectorized_blake2b_32; } @@ -1601,7 +1602,7 @@ hmacmodule_init_cpu_features(hmacmodule_state *state) #undef ECX_SSE3 #undef EBX_AVX2 -#if HACL_CAN_COMPILE_SIMD128 +#if _Py_HACL_CAN_COMPILE_VEC128 // TODO(picnixz): use py_cpuid_features (gh-125022) to improve detection state->can_run_simd128 = sse && sse2 && sse3 && sse41 && sse42 && cmov; #else @@ -1611,7 +1612,7 @@ hmacmodule_init_cpu_features(hmacmodule_state *state) state->can_run_simd128 = false; #endif -#if HACL_CAN_COMPILE_SIMD256 +#if _Py_HACL_CAN_COMPILE_VEC256 // TODO(picnixz): use py_cpuid_features (gh-125022) to improve detection state->can_run_simd256 = state->can_run_simd128 && avx && avx2; #else diff --git a/PCbuild/pythoncore.vcxproj b/PCbuild/pythoncore.vcxproj index 32a8f2dbad3d5e..b911c9385634d7 100644 --- a/PCbuild/pythoncore.vcxproj +++ b/PCbuild/pythoncore.vcxproj @@ -419,8 +419,12 @@ - HACL_CAN_COMPILE_SIMD128;%(PreprocessorDefinitions) - HACL_CAN_COMPILE_SIMD256;%(PreprocessorDefinitions) + + _Py_HACL_CAN_COMPILE_VEC128;%(PreprocessorDefinitions) + + + _Py_HACL_CAN_COMPILE_VEC256;%(PreprocessorDefinitions) + diff --git a/configure b/configure index 8334e247c3e1bb..76fc113eed0ea3 100755 --- a/configure +++ b/configure @@ -32636,7 +32636,7 @@ then : LIBHACL_SIMD128_FLAGS="-msse -msse2 -msse3 -msse4.1 -msse4.2" -printf "%s\n" "#define HACL_CAN_COMPILE_SIMD128 1" >>confdefs.h +printf "%s\n" "#define _Py_HACL_CAN_COMPILE_VEC128 1" >>confdefs.h # macOS universal2 builds *support* the -msse etc flags because they're @@ -32712,7 +32712,7 @@ then : LIBHACL_SIMD256_FLAGS="-mavx2" -printf "%s\n" "#define HACL_CAN_COMPILE_SIMD256 1" >>confdefs.h +printf "%s\n" "#define _Py_HACL_CAN_COMPILE_VEC256 1" >>confdefs.h # macOS universal2 builds *support* the -mavx2 compiler flag because it's diff --git a/configure.ac b/configure.ac index 82fb2d8c492d48..b6666dea562ef4 100644 --- a/configure.ac +++ b/configure.ac @@ -8028,7 +8028,8 @@ then AX_CHECK_COMPILE_FLAG([-msse -msse2 -msse3 -msse4.1 -msse4.2],[ [LIBHACL_SIMD128_FLAGS="-msse -msse2 -msse3 -msse4.1 -msse4.2"] - AC_DEFINE([HACL_CAN_COMPILE_SIMD128], [1], [HACL* library can compile SIMD128 implementations]) + AC_DEFINE([_Py_HACL_CAN_COMPILE_VEC128], [1], [ + HACL* library can compile SIMD128 implementations]) # macOS universal2 builds *support* the -msse etc flags because they're # available on x86_64. However, performance of the HACL SIMD128 implementation @@ -8059,7 +8060,8 @@ if test "$ac_sys_system" != "Linux-android" -a "$ac_sys_system" != "WASI" || \ then AX_CHECK_COMPILE_FLAG([-mavx2],[ [LIBHACL_SIMD256_FLAGS="-mavx2"] - AC_DEFINE([HACL_CAN_COMPILE_SIMD256], [1], [HACL* library can compile SIMD256 implementations]) + AC_DEFINE([_Py_HACL_CAN_COMPILE_VEC256], [1], [ + HACL* library can compile SIMD256 implementations]) # macOS universal2 builds *support* the -mavx2 compiler flag because it's # available on x86_64; but the HACL SIMD256 build then fails because the diff --git a/pyconfig.h.in b/pyconfig.h.in index 65a2c55217c258..3d393912ffbfa0 100644 --- a/pyconfig.h.in +++ b/pyconfig.h.in @@ -50,12 +50,6 @@ /* Define if getpgrp() must be called as getpgrp(0). */ #undef GETPGRP_HAVE_ARG -/* HACL* library can compile SIMD128 implementations */ -#undef HACL_CAN_COMPILE_SIMD128 - -/* HACL* library can compile SIMD256 implementations */ -#undef HACL_CAN_COMPILE_SIMD256 - /* Define if you have the 'accept' function. */ #undef HAVE_ACCEPT @@ -2026,6 +2020,12 @@ /* Maximum length in bytes of a thread name */ #undef _PYTHREAD_NAME_MAXLEN +/* HACL* library can compile SIMD128 implementations */ +#undef _Py_HACL_CAN_COMPILE_VEC128 + +/* HACL* library can compile SIMD256 implementations */ +#undef _Py_HACL_CAN_COMPILE_VEC256 + /* Define to force use of thread-safe errno, h_errno, and other functions */ #undef _REENTRANT pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy