Merge pull request #303 from relic-toolkit/pornin

Pornin
relic-toolkit · Aug 8, 2024 · 6fde832 · 6fde832
2 parents 0646e4f + 7f9f8d9
commit 6fde832
Show file tree

Hide file tree

Showing 18 changed files with 850 additions and 140 deletions.
diff --git a/include/relic_arch.h b/include/relic_arch.h
@@ -95,6 +95,11 @@ ull_t arch_cycles(void);
  */
 uint_t arch_lzcnt(dig_t);
 
+/**
+ * Return the number of trailing zeros in an integer.
+ */
+uint_t arch_tzcnt(dig_t);
+
 #if ARCH == AVR
 
 /**

diff --git a/include/relic_bn.h b/include/relic_bn.h
@@ -1541,6 +1541,18 @@ void bn_rec_glv(bn_t k0, bn_t k1, const bn_t k, const bn_t n, const bn_t v1[],
 void bn_rec_frb(bn_t *ki, int sub, const bn_t k, const bn_t x, const bn_t n,
 	int cof);
 
+/**
+ * Recodes subscalars in the signed aligned column representation..
+ *
+ * @param[out] b			- the recoded subscalars.
+ * @param[in] len			- the length in bytes of the recoding.
+ * @param[in] k				- the subscalars to recode.
+ * @param[in] m				- the number of subscallars to recode.
+ * @param[in] n				- the elliptic curve group order.
+ * @throw ERR_NO_BUFFER		- if the buffer capacity is insufficient.
+ */
+void bn_rec_sac(int8_t *b, size_t *len, bn_t *k, size_t m, bn_t n);
+
 /**
  * Computes the coefficients of the polynomial representing the Lagrange
  * interpolation for a modulus and a given set of roots.

diff --git a/include/relic_core.h b/include/relic_core.h
@@ -497,8 +497,10 @@ typedef struct _ctx_t {
 	/** Function pointer to underlying lznct implementation. */
 #if ARCH == X86
 	unsigned int (*lzcnt_ptr)(dig_t);
+	unsigned int (*tzcnt_ptr)(dig_t);
 #elif ARCH == X64 || ARCH == A64
 	unsigned int (*lzcnt_ptr)(ull_t);
+	unsigned int (*tzcnt_ptr)(ull_t);
 #endif
 } ctx_t;
 

diff --git a/src/arch/relic_arch_a64.c b/src/arch/relic_arch_a64.c
@@ -36,6 +36,7 @@
 #include "relic_core.h"
 
 #include "lzcnt.inc"
+#include "tzcnt.inc"
 
 /**
  * Renames the inline assembly macro to a prettier name.
@@ -177,6 +178,8 @@ void arch_init(void) {
 	if (ctx != NULL) {
 		core_get()->lzcnt_ptr =
 			(has_lzcnt_hard() ? lzcnt64_hard : lzcnt64_soft);
+		core_get()->tzcnt_ptr =
+			(has_lzcnt_hard() ? tzcnt64_hard : tzcnt64_soft);
 	}
 
 #if TIMER == CYCLE
@@ -199,6 +202,7 @@ void arch_clean(void) {
 	ctx_t *ctx = core_get();
 	if (ctx != NULL) {
 		core_get()->lzcnt_ptr = NULL;
+		core_get()->tzcnt_ptr = NULL;
 	}
 }
 
@@ -234,3 +238,7 @@ ull_t arch_cycles(void) {
 uint_t arch_lzcnt(dig_t x) {
 	return core_get()->lzcnt_ptr((ull_t)x) - (8 * sizeof(ull_t) - WSIZE);
 }
+
+uint_t arch_tzcnt(dig_t x) {
+	return core_get()->tzcnt_ptr(x);
+}
diff --git a/src/arch/relic_arch_arm.c b/src/arch/relic_arch_arm.c
@@ -32,6 +32,7 @@
 #include "relic_types.h"
 
 #include "lzcnt.inc"
+#include "tzcnt.inc"
 
 /**
  * Renames the inline assembly macro to a prettier name.
@@ -111,3 +112,11 @@ uint_t arch_lzcnt(uint_t x) {
 	return lzcnt64_gcc_arm(x);
 #endif
 }
+
+uint_t arch_tzcnt(uint_t x) {
+#ifdef WSIZE == 32
+	return tzcnt32_gcc_arm(x);
+#elif WSIZE == 64
+	return tzcnt64_gcc_arm(x);
+#endif
+}
diff --git a/src/arch/relic_arch_avr.c b/src/arch/relic_arch_avr.c
@@ -62,3 +62,15 @@ uint_t arch_lzcnt() {
 	}
 	return 0;
 }
+
+uint_t arch_tzcnt() {
+	static const uint8_t table[16] = {
+		4, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0
+	};
+	if (a >> 4 != 0) {
+		return table[a & 0xF];
+	} else {
+		return table[a >> 4] + 4;
+	}
+	return 0;
+}
diff --git a/src/arch/relic_arch_msp.c b/src/arch/relic_arch_msp.c
@@ -119,3 +119,32 @@ uint_t arch_lzcnt() {
 	return 0;
 #endif
 }
+
+uint_t arch_tzcnt() {
+    static const uint8_t table[16] = {
+    	4, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0
+    };
+#if WSIZE == 8
+	if (a >> 4 != 0) {
+		return table[a & 0xF];
+	} else {
+		return table[a >> 4] + 4;
+	}
+	return 0;
+#elif WSIZE == 16
+	int offset;
+
+	if (a & 0xFF == 0) {
+		offset = 8;
+	} else {
+		offset = 0;
+	}
+	a = a >> offset;
+	if (a >> 4 != 0) {
+		return table[a & 0xF] + offset;
+	} else {
+		return table[a >> 4] + 4 + offset;
+	}
+	return 0;
+#endif
+}
diff --git a/src/arch/relic_arch_none.c b/src/arch/relic_arch_none.c
@@ -83,7 +83,50 @@ uint_t arch_lzcnt(dig_t a) {
 #ifdef _MSC_VER
     return __lzcnt64(a);
 #else
-	return __builtin_clzll(a);
+	return __builtin_clzl(a);
+#endif
+#endif
+}
+
+uint_t arch_tzcnt(dig_t a) {
+#if WSIZE == 8 || WSIZE == 16
+	static const uint8_t table[16] = {
+		4, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0
+	};
+#endif
+#if WSIZE == 8
+	if (a >> 4 != 0) {
+		return table[a & 0xF];
+	} else {
+		return table[a >> 4] + 4;
+	}
+	return 0;
+#elif WSIZE == 16
+	int offset;
+
+	if (a & 0xFF == 0) {
+		offset = 8;
+	} else {
+		offset = 0;
+	}
+	a = a >> offset;
+	if (a >> 4 != 0) {
+		return table[a & 0xF] + offset;
+	} else {
+		return table[a >> 4] + 4 + offset;
+	}
+	return 0;
+#elif WSIZE == 32
+#ifdef _MSC_VER
+    return __tzcnt(a);
+#else
+	return __builtin_ctz(a);
+#endif
+#elif WSIZE == 64
+#ifdef _MSC_VER
+    return __tzcnt64(a);
+#else
+	return __builtin_ctzl(a);
 #endif
 #endif
 }
diff --git a/src/arch/relic_arch_x64.c b/src/arch/relic_arch_x64.c
@@ -36,6 +36,7 @@
 #include "relic_core.h"
 
 #include "lzcnt.inc"
+#include "tzcnt.inc"
 
 /**
  * Renames the inline assembly macro to a prettier name.
@@ -51,13 +52,16 @@ void arch_init(void) {
 	if (ctx != NULL) {
 		core_get()->lzcnt_ptr =
 			(has_lzcnt_hard() ? lzcnt64_hard : lzcnt64_soft);
+		core_get()->tzcnt_ptr =
+			(has_tzcnt_hard() ? tzcnt64_hard : tzcnt64_soft);
 	}
 }
 
 void arch_clean(void) {
 	ctx_t *ctx = core_get();
 	if (ctx != NULL) {
 		core_get()->lzcnt_ptr = NULL;
+		core_get()->tzcnt_ptr = NULL;
 	}
 }
 
@@ -103,3 +107,7 @@ ull_t arch_cycles(void) {
 uint_t arch_lzcnt(dig_t x) {
 	return core_get()->lzcnt_ptr((ull_t)x) - (8 * sizeof(ull_t) - WSIZE);
 }
+
+uint_t arch_tzcnt(dig_t x) {
+	return core_get()->tzcnt_ptr(x);
+}
diff --git a/src/arch/relic_arch_x86.c b/src/arch/relic_arch_x86.c
@@ -36,17 +36,20 @@
 #include "relic_core.h"
 
 #include "lzcnt.inc"
+#include "tzcnt.inc"
 
 /*============================================================================*/
 /* Public definitions                                                         */
 /*============================================================================*/
 
 void arch_init(void) {
 	core_get()->lzcnt_ptr = (has_lzcnt_hard() ? lzcnt32_hard : lzcnt32_soft);
+	core_get()->tzcnt_ptr = (has_tzcnt_hard() ? tzcnt32_hard : tzcnt32_soft);
 }
 
 void arch_clean(void) {
 	core_get()->lzcnt_ptr = NULL;
+	core_get()->tzcnt_ptr = NULL;
 }
 
 ull_t arch_cycles(void) {
@@ -64,3 +67,7 @@ ull_t arch_cycles(void) {
 uint_t arch_lzcnt(dig_t x) {
 	return core_get()->lzcnt_ptr((uint32_t)x) - (8 * sizeof(uint32_t) - WSIZE);
 }
+
+uint_t arch_tzcnt(dig_t x) {
+	return core_get()->tzcnt_ptr(x);
+}