core/crypto/poly1305: Cleanups

- Use `encoding/endian`

core/crypto/_fiat/field_poly1305/field.odin

@@ -1,6 +1,6 @@
 package field_poly1305
 
-import "core:crypto/util"
+import "core:encoding/endian"
 import "core:mem"
 
 fe_relax_cast :: #force_inline proc "contextless" (arg1: ^Tight_Field_Element) -> ^Loose_Field_Element {
@@ -11,7 +11,7 @@ fe_tighten_cast :: #force_inline proc "contextless" (arg1: ^Loose_Field_Element)
 	return transmute(^Tight_Field_Element)(arg1)
 }
 
-fe_from_bytes :: #force_inline proc (out1: ^Tight_Field_Element, arg1: []byte, arg2: byte, sanitize: bool = true) {
+fe_from_bytes :: #force_inline proc (out1: ^Tight_Field_Element, arg1: []byte, arg2: byte) {
 	// fiat-crypto's deserialization routine effectively processes a
 	// single byte at a time, and wants 256-bits of input for a value
 	// that will be 128-bits or 129-bits.
@@ -22,42 +22,29 @@ fe_from_bytes :: #force_inline proc (out1: ^Tight_Field_Element, arg1: []byte, a
 
 	assert(len(arg1) == 16)
 
-	when ODIN_ARCH == .i386 || ODIN_ARCH == .amd64 {
-		// While it may be unwise to do deserialization here on our
-		// own when fiat-crypto provides equivalent functionality,
-		// doing it this way provides a little under 3x performance
-		// improvement when optimization is enabled.
-		src_p := transmute(^[2]u64)(&arg1[0])
-		lo := src_p[0]
-		hi := src_p[1]
+	// While it may be unwise to do deserialization here on our
+	// own when fiat-crypto provides equivalent functionality,
+	// doing it this way provides a little under 3x performance
+	// improvement when optimization is enabled.
+	lo := endian.unchecked_get_u64le(arg1[0:])
+	hi := endian.unchecked_get_u64le(arg1[8:])
 
-		// This is inspired by poly1305-donna, though adjustments were
-		// made since a Tight_Field_Element's limbs are 44-bits, 43-bits,
-		// and 43-bits wide.
-		//
-		// Note: This could be transplated into fe_from_u64s, but that
-		// code is called once per MAC, and is non-criticial path.
-		hibit := u64(arg2) << 41 // arg2 << 128
-		out1[0] = lo & 0xfffffffffff
-		out1[1] = ((lo >> 44) | (hi << 20)) & 0x7ffffffffff
-		out1[2] = ((hi >> 23) & 0x7ffffffffff) | hibit
-	} else {
-		tmp: [32]byte
-		copy_slice(tmp[0:16], arg1[:])
-		tmp[16] = arg2
-
-		_fe_from_bytes(out1, &tmp)
-		if sanitize {
-			// This is used to deserialize `s` which is confidential.
-			mem.zero_explicit(&tmp, size_of(tmp))
-		}
-	}
+	// This is inspired by poly1305-donna, though adjustments were
+	// made since a Tight_Field_Element's limbs are 44-bits, 43-bits,
+	// and 43-bits wide.
+	//
+	// Note: This could be transplated into fe_from_u64s, but that
+	// code is called once per MAC, and is non-criticial path.
+	hibit := u64(arg2) << 41 // arg2 << 128
+	out1[0] = lo & 0xfffffffffff
+	out1[1] = ((lo >> 44) | (hi << 20)) & 0x7ffffffffff
+	out1[2] = ((hi >> 23) & 0x7ffffffffff) | hibit
 }
 
 fe_from_u64s :: proc "contextless" (out1: ^Tight_Field_Element, lo, hi: u64) {
 	tmp: [32]byte
-	util.PUT_U64_LE(tmp[0:8], lo)
-	util.PUT_U64_LE(tmp[8:16], hi)
+	endian.unchecked_put_u64le(tmp[0:], lo)
+	endian.unchecked_put_u64le(tmp[8:], hi)
 
 	_fe_from_bytes(out1, &tmp)
 

core/crypto/poly1305/poly1305.odin

@@ -1,8 +1,8 @@
 package poly1305
 
 import "core:crypto"
-import "core:crypto/util"
 import field "core:crypto/_fiat/field_poly1305"
+import "core:encoding/endian"
 import "core:mem"
 
 KEY_SIZE :: 32
@@ -52,8 +52,8 @@ init :: proc (ctx: ^Context, key: []byte) {
 
 	// r = le_bytes_to_num(key[0..15])
 	// r = clamp(r) (r &= 0xffffffc0ffffffc0ffffffc0fffffff)
-	tmp_lo := util.U64_LE(key[0:8]) & 0x0ffffffc0fffffff
-	tmp_hi := util.U64_LE(key[8:16]) & 0xffffffc0ffffffc
+	tmp_lo := endian.unchecked_get_u64le(key[0:]) & 0x0ffffffc0fffffff
+	tmp_hi := endian.unchecked_get_u64le(key[8:]) & 0xffffffc0ffffffc
 	field.fe_from_u64s(&ctx._r, tmp_lo, tmp_hi)
 
 	// s = le_bytes_to_num(key[16..31])
@@ -151,7 +151,7 @@ _blocks :: proc (ctx: ^Context, msg: []byte, final := false) {
 	data_len := len(data)
 	for data_len >= _BLOCK_SIZE {
 		// n = le_bytes_to_num(msg[((i-1)*16)..*i*16] | [0x01])
-		field.fe_from_bytes(&n, data[:_BLOCK_SIZE], final_byte, false)
+		field.fe_from_bytes(&n, data[:_BLOCK_SIZE], final_byte)
 
 		// a += n
 		field.fe_add(field.fe_relax_cast(&ctx._a), &ctx._a, &n) // _a unreduced