aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat (limited to '')
-rw-r--r--Makefile7
-rw-r--r--config-armv8.mk14
-rw-r--r--libsha2.74
-rw-r--r--libsha2_algorithm_output_size.32
-rw-r--r--libsha2_behex_lower.32
-rw-r--r--libsha2_behex_upper.32
-rw-r--r--libsha2_digest.32
-rw-r--r--libsha2_hmac_digest.32
-rw-r--r--libsha2_hmac_init.32
-rw-r--r--libsha2_hmac_marshal.32
-rw-r--r--libsha2_hmac_state_output_size.32
-rw-r--r--libsha2_hmac_unmarshal.32
-rw-r--r--libsha2_hmac_update.32
-rw-r--r--libsha2_init.32
-rw-r--r--libsha2_marshal.32
-rw-r--r--libsha2_state_output_size.32
-rw-r--r--libsha2_sum_fd.32
-rw-r--r--libsha2_unhex.32
-rw-r--r--libsha2_unmarshal.32
-rw-r--r--libsha2_update.32
-rw-r--r--process.c202
21 files changed, 226 insertions, 35 deletions
diff --git a/Makefile b/Makefile
index 3eee714..ed4dfd2 100644
--- a/Makefile
+++ b/Makefile
@@ -7,6 +7,9 @@ CONFIGFILE = config-x86.mk
# enables optimisations on x86 CPU's that have the
# required features.
#
+# Additionally config-arm.mk which uses optimisations
+# for ARMv8 is available.
+#
# config-portable.mk is available for exotic CPU's
# and compiler that do not support the features required
# for the optimisations.
@@ -20,7 +23,7 @@ include mk/$(OS).mk
LIB_MAJOR = 1
-LIB_MINOR = 0
+LIB_MINOR = 1
LIB_VERSION = $(LIB_MAJOR).$(LIB_MINOR)
@@ -99,7 +102,7 @@ libsha2.a: $(OBJ)
$(AR) -s $@
check: test
- ./test
+ $(CHECK_PREFIX) ./test $(CHECK_FLAGS)
install:
mkdir -p -- "$(DESTDIR)$(PREFIX)/lib"
diff --git a/config-armv8.mk b/config-armv8.mk
new file mode 100644
index 0000000..c98fd0e
--- /dev/null
+++ b/config-armv8.mk
@@ -0,0 +1,14 @@
+PREFIX = /usr
+MANPREFIX = $(PREFIX)/share/man
+
+CC = cc -std=c11
+
+CPPFLAGS = -D_DEFAULT_SOURCE -D_BSD_SOURCE -D_XOPEN_SOURCE=700
+CFLAGS = -Wall -O3 -march=armv8-a+crypto
+LDFLAGS = -s
+
+# You can add -DALLOCA_LIMIT=# to CPPFLAGS, where # is a size_t
+# value, to put a limit on how large allocation the library is
+# allowed to make with alloca(3). For buffers that can have any
+# size this limit will be used if it wants to allocate a larger
+# buffer. Choose 0 to use malloc(3) instead of alloca(3).
diff --git a/libsha2.7 b/libsha2.7
index b78db0a..3da7907 100644
--- a/libsha2.7
+++ b/libsha2.7
@@ -1,8 +1,8 @@
-.TH LIBSHA2 7 2022-07-07 libsha1
+.TH LIBSHA2 7 LIBSHA2
.SH NAME
libsha2 \- SHA-2 hashing library
.SH DESCRIPTION
-.B libsha1
+.B libsha2
is an implementation of the SHA-2 family hashing functions:
SHA-224, SHA-256, SHA-384, SHA-512, SHA-512/224, and SHA-512/256;
with support for state marshalling and HMAC.
diff --git a/libsha2_algorithm_output_size.3 b/libsha2_algorithm_output_size.3
index d083a38..32fb95a 100644
--- a/libsha2_algorithm_output_size.3
+++ b/libsha2_algorithm_output_size.3
@@ -1,4 +1,4 @@
-.TH LIBSHA2_ALGORITHM_OUTPUT_SIZE 3 2019-02-09 libsha2
+.TH LIBSHA2_ALGORITHM_OUTPUT_SIZE 3 LIBSHA2
.SH NAME
libsha2_algorithm_output_size \- Get the size of the output for a SHA-2 algorithm
.SH SYNOPSIS
diff --git a/libsha2_behex_lower.3 b/libsha2_behex_lower.3
index 1172397..e0ce17f 100644
--- a/libsha2_behex_lower.3
+++ b/libsha2_behex_lower.3
@@ -1,4 +1,4 @@
-.TH LIBSHA2_BEHEX_LOWER 3 2019-02-09 libsha2
+.TH LIBSHA2_BEHEX_LOWER 3 LIBSHA2
.SH NAME
libsha2_behex_lower \- Convert binary to lower case hexadecimal
.SH SYNOPSIS
diff --git a/libsha2_behex_upper.3 b/libsha2_behex_upper.3
index c65202b..4d24b99 100644
--- a/libsha2_behex_upper.3
+++ b/libsha2_behex_upper.3
@@ -1,4 +1,4 @@
-.TH LIBSHA2_BEHEX_UPPER 3 2019-02-09 libsha2
+.TH LIBSHA2_BEHEX_UPPER 3 LIBSHA2
.SH NAME
libsha2_behex_upper \- Convert binary to upper case hexadecimal
.SH SYNOPSIS
diff --git a/libsha2_digest.3 b/libsha2_digest.3
index 127171d..afd4d10 100644
--- a/libsha2_digest.3
+++ b/libsha2_digest.3
@@ -1,4 +1,4 @@
-.TH LIBSHA2_DIGEST 3 2019-02-09 libsha2
+.TH LIBSHA2_DIGEST 3 LIBSHA2
.SH NAME
libsha2_digest \- Get the result of a SHA-2 hashing
.SH SYNOPSIS
diff --git a/libsha2_hmac_digest.3 b/libsha2_hmac_digest.3
index 1203049..faf6ab2 100644
--- a/libsha2_hmac_digest.3
+++ b/libsha2_hmac_digest.3
@@ -1,4 +1,4 @@
-.TH LIBSHA2_HMAC_DIGEST 3 2019-02-10 libsha2
+.TH LIBSHA2_HMAC_DIGEST 3 LIBSHA2
.SH NAME
libsha2_hmac_digest \- Get the result of a HMAC-SHA-2 hashing
.SH SYNOPSIS
diff --git a/libsha2_hmac_init.3 b/libsha2_hmac_init.3
index 6f09680..bffe76c 100644
--- a/libsha2_hmac_init.3
+++ b/libsha2_hmac_init.3
@@ -1,4 +1,4 @@
-.TH LIBSHA2_HMAC_INIT 3 2019-02-10 libsha2
+.TH LIBSHA2_HMAC_INIT 3 LIBSHA2
.SH NAME
libsha2_hmac_init \- Initialises hashing with an HMAC-SHA-2 algorithm
.SH SYNOPSIS
diff --git a/libsha2_hmac_marshal.3 b/libsha2_hmac_marshal.3
index eca380c..48fe001 100644
--- a/libsha2_hmac_marshal.3
+++ b/libsha2_hmac_marshal.3
@@ -1,4 +1,4 @@
-.TH LIBSHA2_HMAC_MARSHAL 3 2019-02-10 libsha2
+.TH LIBSHA2_HMAC_MARSHAL 3 LIBSHA2
.SH NAME
libsha2_hmac_marshal \- Marshal an HMAC-SHA-2 hashing state
.SH SYNOPSIS
diff --git a/libsha2_hmac_state_output_size.3 b/libsha2_hmac_state_output_size.3
index aa96436..93c2daf 100644
--- a/libsha2_hmac_state_output_size.3
+++ b/libsha2_hmac_state_output_size.3
@@ -1,4 +1,4 @@
-.TH LIBSHA2_HMAC_STATE_OUTPUT_SIZE 3 2019-02-10 libsha2
+.TH LIBSHA2_HMAC_STATE_OUTPUT_SIZE 3 LIBSHA2
.SH NAME
libsha2_hmac_state_output_size \- Get the size of the output for a HMAC-SHA-2 algorithm
.SH SYNOPSIS
diff --git a/libsha2_hmac_unmarshal.3 b/libsha2_hmac_unmarshal.3
index 6cc2967..467595c 100644
--- a/libsha2_hmac_unmarshal.3
+++ b/libsha2_hmac_unmarshal.3
@@ -1,4 +1,4 @@
-.TH LIBSHA2_HMAC_UNMARSHAL 3 2019-02-10 libsha2
+.TH LIBSHA2_HMAC_UNMARSHAL 3 LIBSHA2
.SH NAME
libsha2_hmac_unmarshal \- Unmarshal an HMAC-SHA-2 hashing state
.SH SYNOPSIS
diff --git a/libsha2_hmac_update.3 b/libsha2_hmac_update.3
index bb02a18..7764c6e 100644
--- a/libsha2_hmac_update.3
+++ b/libsha2_hmac_update.3
@@ -1,4 +1,4 @@
-.TH LIBSHA2_HMAC_UPDATE 3 2019-02-10 libsha2
+.TH LIBSHA2_HMAC_UPDATE 3 LIBSHA2
.SH NAME
libsha2_hmac_update \- Feed data into a HMAC-SHA-2 algorithm
.SH SYNOPSIS
diff --git a/libsha2_init.3 b/libsha2_init.3
index 5823464..aca97b0 100644
--- a/libsha2_init.3
+++ b/libsha2_init.3
@@ -1,4 +1,4 @@
-.TH LIBSHA2_INIT 3 2019-02-09 libsha2
+.TH LIBSHA2_INIT 3 LIBSHA2
.SH NAME
libsha2_init \- Initialises hashing with a SHA-2 algorithm
.SH SYNOPSIS
diff --git a/libsha2_marshal.3 b/libsha2_marshal.3
index 4b44533..ecb5031 100644
--- a/libsha2_marshal.3
+++ b/libsha2_marshal.3
@@ -1,4 +1,4 @@
-.TH LIBSHA2_MARSHAL 3 2019-02-09 libsha2
+.TH LIBSHA2_MARSHAL 3 LIBSHA2
.SH NAME
libsha2_marshal \- Marshal a SHA-2 hashing state
.SH SYNOPSIS
diff --git a/libsha2_state_output_size.3 b/libsha2_state_output_size.3
index 591eace..ef3130b 100644
--- a/libsha2_state_output_size.3
+++ b/libsha2_state_output_size.3
@@ -1,4 +1,4 @@
-.TH LIBSHA2_STATE_OUTPUT_SIZE 3 2019-02-09 libsha2
+.TH LIBSHA2_STATE_OUTPUT_SIZE 3 LIBSHA2
.SH NAME
libsha2_state_output_size \- Get the size of the output for a SHA-2 algorithm
.SH SYNOPSIS
diff --git a/libsha2_sum_fd.3 b/libsha2_sum_fd.3
index 8850e01..07e1864 100644
--- a/libsha2_sum_fd.3
+++ b/libsha2_sum_fd.3
@@ -1,4 +1,4 @@
-.TH LIBSHA2_SUM_FD 3 2019-02-09 libsha2
+.TH LIBSHA2_SUM_FD 3 LIBSHA2
.SH NAME
libsha2_sum_fd \- Hash a file with a SHA-2 algorithm
.SH SYNOPSIS
diff --git a/libsha2_unhex.3 b/libsha2_unhex.3
index 7e009a0..5737057 100644
--- a/libsha2_unhex.3
+++ b/libsha2_unhex.3
@@ -1,4 +1,4 @@
-.TH LIBSHA2_UNHEX 3 2019-02-09 libsha2
+.TH LIBSHA2_UNHEX 3 LIBSHA2
.SH NAME
libsha2_unhex \- Covert hexadecimal to binary
.SH SYNOPSIS
diff --git a/libsha2_unmarshal.3 b/libsha2_unmarshal.3
index 5df4929..a33eec5 100644
--- a/libsha2_unmarshal.3
+++ b/libsha2_unmarshal.3
@@ -1,4 +1,4 @@
-.TH LIBSHA2_UNMARSHAL 3 2019-02-09 libsha2
+.TH LIBSHA2_UNMARSHAL 3 LIBSHA2
.SH NAME
libsha2_unmarshal \- Unmarshal a SHA-2 hashing state
.SH SYNOPSIS
diff --git a/libsha2_update.3 b/libsha2_update.3
index b7baf39..e64c972 100644
--- a/libsha2_update.3
+++ b/libsha2_update.3
@@ -1,4 +1,4 @@
-.TH LIBSHA2_UPDATE 3 2019-02-09 libsha2
+.TH LIBSHA2_UPDATE 3 LIBSHA2
.SH NAME
libsha2_update \- Feed data into a SHA-2 algorithm
.SH SYNOPSIS
diff --git a/process.c b/process.c
index 6f1451e..1a83046 100644
--- a/process.c
+++ b/process.c
@@ -3,14 +3,25 @@
#include <stdatomic.h>
#if defined(__SSE4_1__) && defined(__SSSE3__) && defined(__SSE2__) && defined(__SHA__)
-# define HAVE_X86_SHA_INTRINSICS
+# define HAVE_X86_SHA256_INTRINSICS
#endif
+#if defined(__arm__) || defined(__aarch32__) || defined(__arm64__) || defined(__aarch64__) || defined(_M_ARM)
+# if defined(__ARM_NEON) && (defined(__ARM_ACLE) || defined(__ARM_FEATURE_CRYPTO))
+# define HAVE_ARM_SHA256_INTRINSICS
+# endif
+#endif
-#ifdef HAVE_X86_SHA_INTRINSICS
+
+#ifdef HAVE_X86_SHA256_INTRINSICS
# include <immintrin.h>
#endif
+#ifdef HAVE_ARM_SHA256_INTRINSICS
+# include <arm_neon.h>
+# include <arm_acle.h>
+#endif
+
/**
* Unified implementation (what can unified without performance impact)
@@ -68,7 +79,7 @@
h[i] = TRUNC(h[i] + work_h[i]);
-#ifdef HAVE_X86_SHA_INTRINSICS
+#ifdef HAVE_X86_SHA256_INTRINSICS
static size_t
process_x86_sha256(struct libsha2_state *restrict state, const unsigned char *restrict data, size_t len)
@@ -290,6 +301,168 @@ out:
#endif
+#ifdef HAVE_ARM_SHA256_INTRINSICS
+
+static size_t
+process_arm_sha256(struct libsha2_state *restrict state, const unsigned char *restrict data, size_t len)
+{
+ static const uint32_t rc[] = {
+ UINT32_C(0x428A2F98), UINT32_C(0x71374491), UINT32_C(0xB5C0FBCF), UINT32_C(0xE9B5DBA5),
+ UINT32_C(0x3956C25B), UINT32_C(0x59F111F1), UINT32_C(0x923F82A4), UINT32_C(0xAB1C5ED5),
+ UINT32_C(0xD807AA98), UINT32_C(0x12835B01), UINT32_C(0x243185BE), UINT32_C(0x550C7DC3),
+ UINT32_C(0x72BE5D74), UINT32_C(0x80DEB1FE), UINT32_C(0x9BDC06A7), UINT32_C(0xC19BF174),
+ UINT32_C(0xE49B69C1), UINT32_C(0xEFBE4786), UINT32_C(0x0FC19DC6), UINT32_C(0x240CA1CC),
+ UINT32_C(0x2DE92C6F), UINT32_C(0x4A7484AA), UINT32_C(0x5CB0A9DC), UINT32_C(0x76F988DA),
+ UINT32_C(0x983E5152), UINT32_C(0xA831C66D), UINT32_C(0xB00327C8), UINT32_C(0xBF597FC7),
+ UINT32_C(0xC6E00BF3), UINT32_C(0xD5A79147), UINT32_C(0x06CA6351), UINT32_C(0x14292967),
+ UINT32_C(0x27B70A85), UINT32_C(0x2E1B2138), UINT32_C(0x4D2C6DFC), UINT32_C(0x53380D13),
+ UINT32_C(0x650A7354), UINT32_C(0x766A0ABB), UINT32_C(0x81C2C92E), UINT32_C(0x92722C85),
+ UINT32_C(0xA2BFE8A1), UINT32_C(0xA81A664B), UINT32_C(0xC24B8B70), UINT32_C(0xC76C51A3),
+ UINT32_C(0xD192E819), UINT32_C(0xD6990624), UINT32_C(0xF40E3585), UINT32_C(0x106AA070),
+ UINT32_C(0x19A4C116), UINT32_C(0x1E376C08), UINT32_C(0x2748774C), UINT32_C(0x34B0BCB5),
+ UINT32_C(0x391C0CB3), UINT32_C(0x4ED8AA4A), UINT32_C(0x5B9CCA4F), UINT32_C(0x682E6FF3),
+ UINT32_C(0x748F82EE), UINT32_C(0x78A5636F), UINT32_C(0x84C87814), UINT32_C(0x8CC70208),
+ UINT32_C(0x90BEFFFA), UINT32_C(0xA4506CEB), UINT32_C(0xBEF9A3F7), UINT32_C(0xC67178F2)
+ };
+
+ uint32x4_t abcd, efgh, abcd_orig, efgh_orig;
+ uint32x4_t msg0, msg1, msg2, msg3, tmp0, tmp1, tmp2;
+ const unsigned char *restrict chunk;
+ size_t off = 0;
+
+ abcd_orig = vld1q_u32(&state->h.b32[0]);
+ efgh_orig = vld1q_u32(&state->h.b32[4]);
+
+ for (; len - off >= state->chunk_size; off += state->chunk_size) {
+ abcd = abcd_orig;
+ efgh = efgh_orig;
+
+ chunk = &data[off];
+ msg0 = vld1q_u32((const uint32_t *)&chunk[0]);
+ msg1 = vld1q_u32((const uint32_t *)&chunk[16]);
+ msg2 = vld1q_u32((const uint32_t *)&chunk[32]);
+ msg3 = vld1q_u32((const uint32_t *)&chunk[48]);
+ msg0 = vreinterpretq_u32_u8(vrev32q_u8(vreinterpretq_u8_u32(msg0)));
+ msg1 = vreinterpretq_u32_u8(vrev32q_u8(vreinterpretq_u8_u32(msg1)));
+ msg2 = vreinterpretq_u32_u8(vrev32q_u8(vreinterpretq_u8_u32(msg2)));
+ msg3 = vreinterpretq_u32_u8(vrev32q_u8(vreinterpretq_u8_u32(msg3)));
+
+ tmp0 = vaddq_u32(msg0, vld1q_u32(&rc[0 * 4]));
+
+ msg0 = vsha256su0q_u32(msg0, msg1);
+ tmp2 = abcd;
+ tmp1 = vaddq_u32(msg1, vld1q_u32(&rc[1 * 4]));
+ abcd = vsha256hq_u32(abcd, efgh, tmp0);
+ efgh = vsha256h2q_u32(efgh, tmp2, tmp0);
+ msg0 = vsha256su1q_u32(msg0, msg2, msg3);
+
+ msg1 = vsha256su0q_u32(msg1, msg2);
+ tmp2 = abcd;
+ tmp0 = vaddq_u32(msg2, vld1q_u32(&rc[2 * 4]));
+ abcd = vsha256hq_u32(abcd, efgh, tmp1);
+ efgh = vsha256h2q_u32(efgh, tmp2, tmp1);
+ msg1 = vsha256su1q_u32(msg1, msg3, msg0);
+
+ msg2 = vsha256su0q_u32(msg2, msg3);
+ tmp2 = abcd;
+ tmp1 = vaddq_u32(msg3, vld1q_u32(&rc[3 * 4]));
+ abcd = vsha256hq_u32(abcd, efgh, tmp0);
+ efgh = vsha256h2q_u32(efgh, tmp2, tmp0);
+ msg2 = vsha256su1q_u32(msg2, msg0, msg1);
+
+ msg3 = vsha256su0q_u32(msg3, msg0);
+ tmp2 = abcd;
+ tmp0 = vaddq_u32(msg0, vld1q_u32(&rc[4 * 4]));
+ abcd = vsha256hq_u32(abcd, efgh, tmp1);
+ efgh = vsha256h2q_u32(efgh, tmp2, tmp1);
+ msg3 = vsha256su1q_u32(msg3, msg1, msg2);
+
+ msg0 = vsha256su0q_u32(msg0, msg1);
+ tmp2 = abcd;
+ tmp1 = vaddq_u32(msg1, vld1q_u32(&rc[5 * 4]));
+ abcd = vsha256hq_u32(abcd, efgh, tmp0);
+ efgh = vsha256h2q_u32(efgh, tmp2, tmp0);
+ msg0 = vsha256su1q_u32(msg0, msg2, msg3);
+
+ msg1 = vsha256su0q_u32(msg1, msg2);
+ tmp2 = abcd;
+ tmp0 = vaddq_u32(msg2, vld1q_u32(&rc[6 * 4]));
+ abcd = vsha256hq_u32(abcd, efgh, tmp1);
+ efgh = vsha256h2q_u32(efgh, tmp2, tmp1);
+ msg1 = vsha256su1q_u32(msg1, msg3, msg0);
+
+ msg2 = vsha256su0q_u32(msg2, msg3);
+ tmp2 = abcd;
+ tmp1 = vaddq_u32(msg3, vld1q_u32(&rc[7 * 4]));
+ abcd = vsha256hq_u32(abcd, efgh, tmp0);
+ efgh = vsha256h2q_u32(efgh, tmp2, tmp0);
+ msg2 = vsha256su1q_u32(msg2, msg0, msg1);
+
+ msg3 = vsha256su0q_u32(msg3, msg0);
+ tmp2 = abcd;
+ tmp0 = vaddq_u32(msg0, vld1q_u32(&rc[8 * 4]));
+ abcd = vsha256hq_u32(abcd, efgh, tmp1);
+ efgh = vsha256h2q_u32(efgh, tmp2, tmp1);
+ msg3 = vsha256su1q_u32(msg3, msg1, msg2);
+
+ msg0 = vsha256su0q_u32(msg0, msg1);
+ tmp2 = abcd;
+ tmp1 = vaddq_u32(msg1, vld1q_u32(&rc[9 * 4]));
+ abcd = vsha256hq_u32(abcd, efgh, tmp0);
+ efgh = vsha256h2q_u32(efgh, tmp2, tmp0);
+ msg0 = vsha256su1q_u32(msg0, msg2, msg3);
+
+ msg1 = vsha256su0q_u32(msg1, msg2);
+ tmp2 = abcd;
+ tmp0 = vaddq_u32(msg2, vld1q_u32(&rc[10 * 4]));
+ abcd = vsha256hq_u32(abcd, efgh, tmp1);
+ efgh = vsha256h2q_u32(efgh, tmp2, tmp1);
+ msg1 = vsha256su1q_u32(msg1, msg3, msg0);
+
+ msg2 = vsha256su0q_u32(msg2, msg3);
+ tmp2 = abcd;
+ tmp1 = vaddq_u32(msg3, vld1q_u32(&rc[11 * 4]));
+ abcd = vsha256hq_u32(abcd, efgh, tmp0);
+ efgh = vsha256h2q_u32(efgh, tmp2, tmp0);
+ msg2 = vsha256su1q_u32(msg2, msg0, msg1);
+
+ msg3 = vsha256su0q_u32(msg3, msg0);
+ tmp2 = abcd;
+ tmp0 = vaddq_u32(msg0, vld1q_u32(&rc[12 * 4]));
+ abcd = vsha256hq_u32(abcd, efgh, tmp1);
+ efgh = vsha256h2q_u32(efgh, tmp2, tmp1);
+ msg3 = vsha256su1q_u32(msg3, msg1, msg2);
+
+ tmp2 = abcd;
+ tmp1 = vaddq_u32(msg1, vld1q_u32(&rc[13 * 4]));
+ abcd = vsha256hq_u32(abcd, efgh, tmp0);
+ efgh = vsha256h2q_u32(efgh, tmp2, tmp0);
+
+ tmp2 = abcd;
+ tmp0 = vaddq_u32(msg2, vld1q_u32(&rc[14 * 4]));
+ abcd = vsha256hq_u32(abcd, efgh, tmp1);
+ efgh = vsha256h2q_u32(efgh, tmp2, tmp1);
+
+ tmp2 = abcd;
+ tmp1 = vaddq_u32(msg3, vld1q_u32(&rc[15 * 4]));
+ abcd = vsha256hq_u32(abcd, efgh, tmp0);
+ efgh = vsha256h2q_u32(efgh, tmp2, tmp0);
+
+ tmp2 = abcd;
+ abcd = vsha256hq_u32(abcd, efgh, tmp1);
+ efgh = vsha256h2q_u32(efgh, tmp2, tmp1);
+
+ abcd_orig = vaddq_u32(abcd_orig, abcd);
+ efgh_orig = vaddq_u32(efgh_orig, efgh);
+ }
+
+ vst1q_u32(&state->h.b32[0], abcd_orig);
+ vst1q_u32(&state->h.b32[4], efgh_orig);
+
+ return off;
+}
+
+#endif
size_t
libsha2_process(struct libsha2_state *restrict state, const unsigned char *restrict data, size_t len)
@@ -298,6 +471,10 @@ libsha2_process(struct libsha2_state *restrict state, const unsigned char *restr
size_t off = 0;
if (state->algorithm <= LIBSHA2_256) {
+#if defined(HAVE_ARM_SHA256_INTRINSICS)
+ return process_arm_sha256(state, data, len);
+#else
+# define ROTR(X, N) TRUNC32(((X) >> (N)) | ((X) << (32 - (N))))
uint_least32_t s0, s1;
size_t i, j;
@@ -305,36 +482,33 @@ libsha2_process(struct libsha2_state *restrict state, const unsigned char *restr
# pragma GCC diagnostic push
# pragma GCC diagnostic ignored "-Wmemset-elt-size"
#endif
-#define ROTR(X, N) TRUNC32(((X) >> (N)) | ((X) << (32 - (N))))
-
-#ifdef HAVE_X86_SHA_INTRINSICS
+# ifdef HAVE_X86_SHA256_INTRINSICS
if (have_sha_intrinsics())
return process_x86_sha256(state, data, len);
-#endif
-
+# endif
for (; len - off >= state->chunk_size; off += state->chunk_size) {
chunk = &data[off];
SHA2_IMPLEMENTATION(chunk, 7, 18, 3, 17, 19, 10, 6, 11, 25, 2, 13, 22, uint_least32_t, 4,
TRUNC32, state->k.b32, state->w.b32, state->h.b32, state->work_h.b32);
}
-
-#undef ROTR
-#if defined(__GNUC__)
-# pragma GCC diagnostic pop
+# if defined(__GNUC__)
+# pragma GCC diagnostic pop
+# endif
+# undef ROTR
#endif
} else {
+#define ROTR(X, N) TRUNC64(((X) >> (N)) | ((X) << (64 - (N))))
uint_least64_t s0, s1;
size_t i, j;
-#define ROTR(X, N) TRUNC64(((X) >> (N)) | ((X) << (64 - (N))))
+ /* TODO Add optimisation using ARMv8.2 SHA-512 intrinsics (when I've access to a machine supporting it) */
for (; len - off >= state->chunk_size; off += state->chunk_size) {
chunk = &data[off];
SHA2_IMPLEMENTATION(chunk, 1, 8, 7, 19, 61, 6, 14, 18, 41, 28, 34, 39, uint_least64_t, 8,
TRUNC64, state->k.b64, state->w.b64, state->h.b64, state->work_h.b64);
}
-
#undef ROTR
}
generated by cgit v1.2.3-70-g09d2 (git 2.51.0) at 2026-03-30 14:55:25 +0000