This way, the masks fit together visually.
#define P p26
/* Convert 32-bit words into field-element pieces. */
#define P p26
/* Convert 32-bit words into field-element pieces. */
-#define P26W0(x) ((x##0)&0x03ffffff)
-#define P26W1(x) ((((x##1)&0x000fffff) << 6) | (((x##0) >> 26)&0x0000003f))
-#define P26W2(x) ((((x##2)&0x00003fff) << 12) | (((x##1) >> 20)&0x00000fff))
-#define P26W3(x) ((((x##3)&0x000000ff) << 18) | (((x##2) >> 14)&0x0003ffff))
+#define P26W0(x) (((x##0) << 0)&0x03ffffff)
+#define P26W1(x) ((((x##1) << 6)&0x03ffffc0) | (((x##0) >> 26)&0x0000003f))
+#define P26W2(x) ((((x##2) << 12)&0x03ffffff) | (((x##1) >> 20)&0x00000fff))
+#define P26W3(x) ((((x##3) << 18)&0x03fc0000) | (((x##2) >> 14)&0x0003ffff))
#define P26W4(x) (((x##3) >> 8)&0x00ffffff)
/* Propagate carries in parallel. If 0 <= u_i < 2^26 c_i, then we shall have
#define P26W4(x) (((x##3) >> 8)&0x00ffffff)
/* Propagate carries in parallel. If 0 <= u_i < 2^26 c_i, then we shall have