https://bugs.gentoo.org/915060
https://git.gnupg.org/cgi-bin/gitweb.cgi?p=libgcrypt.git;a=commit;h=07f0563d325c6589ca1560525d3b22291feec227
From 07f0563d325c6589ca1560525d3b22291feec227 Mon Sep 17 00:00:00 2001
From: Jussi Kivilinna <jussi.kivilinna@iki.fi>
Date: Tue, 19 Dec 2023 20:23:47 +0200
Subject: [PATCH] mpi/ec-inline: refactor i386 assembly to reduce register
usage
* mpi/ec-inline.h [__i386__] (ADD2_LIMB32_CARRY_OUT)
(ADD2_LIMB32_CARRY_IN_OUT, ADD2_LIB32_CARRY_IN, SUB2_LIMB32_CARRY_OUT)
(SUB2_LIMB32_CARRY_IN_OUT, SUB2_LIB32_CARRY_IN, ADD8_LIMB32)
(ADD10_LIMB32, ADD14_LIMB32, SUB8_LIMB32, SUB10_LIMB32)
(SUB14_LIMB32): New.
[__i386__] (ADD4_LIMB32, ADD6_LIMB32, SUB4_LIMB32, SUB6_LIMB32): Rewrite
to use new *_CARRY_* macros.
[BYTES_PER_MPI_LIMB == 4] (ADD4_LIMB64): Use ADD8_LIMB32 if available.
[BYTES_PER_MPI_LIMB == 4] (ADD5_LIMB64): Use ADD10_LIMB32 if available.
[BYTES_PER_MPI_LIMB == 4] (ADD7_LIMB64): Use ADD14_LIMB32 if available.
[BYTES_PER_MPI_LIMB == 4] (SUB4_LIMB64): Use SUB8_LIMB32 if available.
[BYTES_PER_MPI_LIMB == 4] (SUB5_LIMB64): Use SUB10_LIMB32 if available.
[BYTES_PER_MPI_LIMB == 4] (SUB7_LIMB64): Use SUB14_LIMB32 if available.
--
Cherry pick of master commit:
956f1ed4ec6ead59dc56f574f943f1fe25dac723
This commit reduces number register operands and register pressure
at i386 assembly used in `ec-nist.c` NIST-P192, P224, P256, and P384.
Performance stays relatively same, with P192 being ~2% slower
than before and P384 being ~5% faster.
GnuPG-bug-id: T6892
Signed-off-by: Jussi Kivilinna <jussi.kivilinna@iki.fi>
---
mpi/ec-inline.h | 327 +++++++++++++++++++++++++++++++++---------------
1 file changed, 229 insertions(+), 98 deletions(-)
diff --git a/mpi/ec-inline.h b/mpi/ec-inline.h
index a07826e3..0ffdf8eb 100644
--- a/mpi/ec-inline.h
+++ b/mpi/ec-inline.h
@@ -641,116 +641,192 @@ LIMB64_HILO(mpi_limb_t hi, mpi_limb_t lo)
/* i386 addition/subtraction helpers. */
#if defined (__i386__) && defined(HAVE_CPU_ARCH_X86) && __GNUC__ >= 4
-#define ADD4_LIMB32(a3, a2, a1, a0, b3, b2, b1, b0, c3, c2, c1, c0) \
- __asm__ ("addl %11, %3\n" \
- "adcl %10, %2\n" \
- "adcl %9, %1\n" \
- "adcl %8, %0\n" \
- : "=r" (a3), \
- "=&r" (a2), \
+#define ADD2_LIMB32_CARRY_OUT(carry, a1, a0, b1, b0, c1, c0) \
+ __asm__ ("addl %7, %2\n" \
+ "adcl %6, %1\n" \
+ "sbbl %0, %0\n" \
+ : "=r" (carry), \
"=&r" (a1), \
"=&r" (a0) \
- : "0" ((mpi_limb_t)(b3)), \
- "1" ((mpi_limb_t)(b2)), \
- "2" ((mpi_limb_t)(b1)), \
- "3" ((mpi_limb_t)(b0)), \
- "g" ((mpi_limb_t)(c3)), \
- "g" ((mpi_limb_t)(c2)), \
- "g" ((mpi_limb_t)(c1)), \
- "g" ((mpi_limb_t)(c0)) \
+ : "0" ((mpi_limb_t)(0)), \
+ "1" ((mpi_limb_t)(b1)), \
+ "2" ((mpi_limb_t)(b0)), \
+ "re" ((mpi_limb_t)(c1)), \
+ "re" ((mpi_limb_t)(c0)) \
: "cc")
+#define ADD2_LIMB32_CARRY_IN_OUT(a1, a0, b1, b0, c1, c0, carry) \
+ __asm__ ("addl $1, %0\n" \
+ "adcl %7, %2\n" \
+ "adcl %6, %1\n" \
+ "sbbl %0, %0\n" \
+ : "=r" (carry), \
+ "=&r" (a1), \
+ "=&r" (a0) \
+ : "0" ((mpi_limb_t)(carry)), \
+ "1" ((mpi_limb_t)(b1)), \
+ "2" ((mpi_limb_t)(b0)), \
+ "re" ((mpi_limb_t)(c1)), \
+ "re" ((mpi_limb_t)(c0)) \
+ : "cc")
+
+#define ADD2_LIMB32_CARRY_IN(a1, a0, b1, b0, c1, c0, carry) \
+ __asm__ ("addl $1, %2\n" \
+ "adcl %7, %1\n" \
+ "adcl %6, %0\n" \
+ : "=r" (a1), \
+ "=&r" (a0), \
+ "=&g" (carry) \
+ : "0" ((mpi_limb_t)(b1)), \
+ "1" ((mpi_limb_t)(b0)), \
+ "2" ((mpi_limb_t)(carry)), \
+ "re" ((mpi_limb_t)(c1)), \
+ "re" ((mpi_limb_t)(c0)) \
+ : "cc")
+
+#define ADD4_LIMB32(a3, a2, a1, a0, b3, b2, b1, b0, c3, c2, c1, c0) do { \
+ mpi_limb_t __carry4_32; \
+ ADD2_LIMB32_CARRY_OUT(__carry4_32, a1, a0, b1, b0, c1, c0); \
+ ADD2_LIMB32_CARRY_IN(a3, a2, b3, b2, c3, c2, __carry4_32); \
+ } while (0)
+
#define ADD6_LIMB32(a5, a4, a3, a2, a1, a0, b5, b4, b3, b2, b1, b0, \
c5, c4, c3, c2, c1, c0) do { \
mpi_limb_t __carry6_32; \
- __asm__ ("addl %10, %3\n" \
- "adcl %9, %2\n" \
- "adcl %8, %1\n" \
- "sbbl %0, %0\n" \
- : "=r" (__carry6_32), \
- "=&r" (a2), \
- "=&r" (a1), \
- "=&r" (a0) \
- : "0" ((mpi_limb_t)(0)), \
- "1" ((mpi_limb_t)(b2)), \
- "2" ((mpi_limb_t)(b1)), \
- "3" ((mpi_limb_t)(b0)), \
- "g" ((mpi_limb_t)(c2)), \
- "g" ((mpi_limb_t)(c1)), \
- "g" ((mpi_limb_t)(c0)) \
- : "cc"); \
- __asm__ ("addl $1, %3\n" \
- "adcl %10, %2\n" \
- "adcl %9, %1\n" \
- "adcl %8, %0\n" \
- : "=r" (a5), \
- "=&r" (a4), \
- "=&r" (a3), \
- "=&r" (__carry6_32) \
- : "0" ((mpi_limb_t)(b5)), \
- "1" ((mpi_limb_t)(b4)), \
- "2" ((mpi_limb_t)(b3)), \
- "3" ((mpi_limb_t)(__carry6_32)), \
- "g" ((mpi_limb_t)(c5)), \
- "g" ((mpi_limb_t)(c4)), \
- "g" ((mpi_limb_t)(c3)) \
- : "cc"); \
+ ADD2_LIMB32_CARRY_OUT(__carry6_32, a1, a0, b1, b0, c1, c0); \
+ ADD2_LIMB32_CARRY_IN_OUT(a3, a2, b3, b2, c3, c2, __carry6_32); \
+ ADD2_LIMB32_CARRY_IN(a5, a4, b5, b4, c5, c4, __carry6_32); \
+ } while (0)
+
+#define ADD8_LIMB32(a7, a6, a5, a4, a3, a2, a1, a0, \
+ b7, b6, b5, b4, b3, b2, b1, b0, \
+ c7, c6, c5, c4, c3, c2, c1, c0) do { \
+ mpi_limb_t __carry8_32; \
+ ADD2_LIMB32_CARRY_OUT(__carry8_32, a1, a0, b1, b0, c1, c0); \
+ ADD2_LIMB32_CARRY_IN_OUT(a3, a2, b3, b2, c3, c2, __carry8_32); \
+ ADD2_LIMB32_CARRY_IN_OUT(a5, a4, b5, b4, c5, c4, __carry8_32); \
+ ADD2_LIMB32_CARRY_IN(a7, a6, b7, b6, c7, c6, __carry8_32); \
} while (0)
-#define SUB4_LIMB32(a3, a2, a1, a0, b3, b2, b1, b0, c3, c2, c1, c0) \
- __asm__ ("subl %11, %3\n" \
- "sbbl %10, %2\n" \
- "sbbl %9, %1\n" \
- "sbbl %8, %0\n" \
- : "=r" (a3), \
- "=&r" (a2), \
+#define ADD10_LIMB32(a9, a8, a7, a6, a5, a4, a3, a2, a1, a0, \
+ b9, b8, b7, b6, b5, b4, b3, b2, b1, b0, \
+ c9, c8, c7, c6, c5, c4, c3, c2, c1, c0) do { \
+ mpi_limb_t __carry10_32; \
+ ADD2_LIMB32_CARRY_OUT(__carry10_32, a1, a0, b1, b0, c1, c0); \
+ ADD2_LIMB32_CARRY_IN_OUT(a3, a2, b3, b2, c3, c2, __carry10_32); \
+ ADD2_LIMB32_CARRY_IN_OUT(a5, a4, b5, b4, c5, c4, __carry10_32); \
+ ADD2_LIMB32_CARRY_IN_OUT(a7, a6, b7, b6, c7, c6, __carry10_32); \
+ ADD2_LIMB32_CARRY_IN(a9, a8, b9, b8, c9, c8, __carry10_32); \
+ } while (0)
+
+#define ADD14_LIMB32(a13, a12, a11, a10, a9, a8, a7, \
+ a6, a5, a4, a3, a2, a1, a0, \
+ b13, b12, b11, b10, b9, b8, b7, \
+ b6, b5, b4, b3, b2, b1, b0, \
+ c13, c12, c11, c10, c9, c8, c7, \
+ c6, c5, c4, c3, c2, c1, c0) do { \
+ mpi_limb_t __carry14_32; \
+ ADD2_LIMB32_CARRY_OUT(__carry14_32, a1, a0, b1, b0, c1, c0); \
+ ADD2_LIMB32_CARRY_IN_OUT(a3, a2, b3, b2, c3, c2, __carry14_32); \
+ ADD2_LIMB32_CARRY_IN_OUT(a5, a4, b5, b4, c5, c4, __carry14_32); \
+ ADD2_LIMB32_CARRY_IN_OUT(a7, a6, b7, b6, c7, c6, __carry14_32); \
+ ADD2_LIMB32_CARRY_IN_OUT(a9, a8, b9, b8, c9, c8, __carry14_32); \
+ ADD2_LIMB32_CARRY_IN_OUT(a11, a10, b11, b10, c11, c10, __carry14_32); \
+ ADD2_LIMB32_CARRY_IN(a13, a12, b13, b12, c13, c12, __carry14_32); \
+ } while (0)
+
+#define SUB2_LIMB32_CARRY_OUT(carry, a1, a0, b1, b0, c1, c0) \
+ __asm__ ("subl %7, %2\n" \
+ "sbbl %6, %1\n" \
+ "sbbl %0, %0\n" \
+ : "=r" (carry), \
"=&r" (a1), \
"=&r" (a0) \
- : "0" ((mpi_limb_t)(b3)), \
- "1" ((mpi_limb_t)(b2)), \
- "2" ((mpi_limb_t)(b1)), \
- "3" ((mpi_limb_t)(b0)), \
- "g" ((mpi_limb_t)(c3)), \
- "g" ((mpi_limb_t)(c2)), \
- "g" ((mpi_limb_t)(c1)), \
- "g" ((mpi_limb_t)(c0)) \
+ : "0" ((mpi_limb_t)(0)), \
+ "1" ((mpi_limb_t)(b1)), \
+ "2" ((mpi_limb_t)(b0)), \
+ "re" ((mpi_limb_t)(c1)), \
+ "re" ((mpi_limb_t)(c0)) \
+ : "cc")
+
+#define SUB2_LIMB32_CARRY_IN_OUT(a1, a0, b1, b0, c1, c0, carry) \
+ __asm__ ("addl $1, %0\n" \
+ "sbbl %7, %2\n" \
+ "sbbl %6, %1\n" \
+ "sbbl %0, %0\n" \
+ : "=r" (carry), \
+ "=&r" (a1), \
+ "=&r" (a0) \
+ : "0" ((mpi_limb_t)(carry)), \
+ "1" ((mpi_limb_t)(b1)), \
+ "2" ((mpi_limb_t)(b0)), \
+ "re" ((mpi_limb_t)(c1)), \
+ "re" ((mpi_limb_t)(c0)) \
+ : "cc")
+
+#define SUB2_LIMB32_CARRY_IN(a1, a0, b1, b0, c1, c0, carry) \
+ __asm__ ("addl $1, %2\n" \
+ "sbbl %7, %1\n" \
+ "sbbl %6, %0\n" \
+ : "=r" (a1), \
+ "=&r" (a0), \
+ "=&g" (carry) \
+ : "0" ((mpi_limb_t)(b1)), \
+ "1" ((mpi_limb_t)(b0)), \
+ "2" ((mpi_limb_t)(carry)), \
+ "re" ((mpi_limb_t)(c1)), \
+ "re" ((mpi_limb_t)(c0)) \
: "cc")
+#define SUB4_LIMB32(a3, a2, a1, a0, b3, b2, b1, b0, c3, c2, c1, c0) do { \
+ mpi_limb_t __carry4_32; \
+ SUB2_LIMB32_CARRY_OUT(__carry4_32, a1, a0, b1, b0, c1, c0); \
+ SUB2_LIMB32_CARRY_IN(a3, a2, b3, b2, c3, c2, __carry4_32); \
+ } while (0)
+
#define SUB6_LIMB32(a5, a4, a3, a2, a1, a0, b5, b4, b3, b2, b1, b0, \
c5, c4, c3, c2, c1, c0) do { \
- mpi_limb_t __borrow6_32; \
- __asm__ ("subl %10, %3\n" \
- "sbbl %9, %2\n" \
- "sbbl %8, %1\n" \
- "sbbl %0, %0\n" \
- : "=r" (__borrow6_32), \
- "=&r" (a2), \
- "=&r" (a1), \
- "=&r" (a0) \
- : "0" ((mpi_limb_t)(0)), \
- "1" ((mpi_limb_t)(b2)), \
- "2" ((mpi_limb_t)(b1)), \
- "3" ((mpi_limb_t)(b0)), \
- "g" ((mpi_limb_t)(c2)), \
- "g" ((mpi_limb_t)(c1)), \
- "g" ((mpi_limb_t)(c0)) \
- : "cc"); \
- __asm__ ("addl $1, %3\n" \
- "sbbl %10, %2\n" \
- "sbbl %9, %1\n" \
- "sbbl %8, %0\n" \
- : "=r" (a5), \
- "=&r" (a4), \
- "=&r" (a3), \
- "=&r" (__borrow6_32) \
- : "0" ((mpi_limb_t)(b5)), \
- "1" ((mpi_limb_t)(b4)), \
- "2" ((mpi_limb_t)(b3)), \
- "3" ((mpi_limb_t)(__borrow6_32)), \
- "g" ((mpi_limb_t)(c5)), \
- "g" ((mpi_limb_t)(c4)), \
- "g" ((mpi_limb_t)(c3)) \
- : "cc"); \
+ mpi_limb_t __carry6_32; \
+ SUB2_LIMB32_CARRY_OUT(__carry6_32, a1, a0, b1, b0, c1, c0); \
+ SUB2_LIMB32_CARRY_IN_OUT(a3, a2, b3, b2, c3, c2, __carry6_32); \
+ SUB2_LIMB32_CARRY_IN(a5, a4, b5, b4, c5, c4, __carry6_32); \
+ } while (0)
+
+#define SUB8_LIMB32(a7, a6, a5, a4, a3, a2, a1, a0, \
+ b7, b6, b5, b4, b3, b2, b1, b0, \
+ c7, c6, c5, c4, c3, c2, c1, c0) do { \
+ mpi_limb_t __carry8_32; \
+ SUB2_LIMB32_CARRY_OUT(__carry8_32, a1, a0, b1, b0, c1, c0); \
+ SUB2_LIMB32_CARRY_IN_OUT(a3, a2, b3, b2, c3, c2, __carry8_32); \
+ SUB2_LIMB32_CARRY_IN_OUT(a5, a4, b5, b4, c5, c4, __carry8_32); \
+ SUB2_LIMB32_CARRY_IN(a7, a6, b7, b6, c7, c6, __carry8_32); \
+ } while (0)
+
+#define SUB10_LIMB32(a9, a8, a7, a6, a5, a4, a3, a2, a1, a0, \
+ b9, b8, b7, b6, b5, b4, b3, b2, b1, b0, \
+ c9, c8, c7, c6, c5, c4, c3, c2, c1, c0) do { \
+ mpi_limb_t __carry10_32; \
+ SUB2_LIMB32_CARRY_OUT(__carry10_32, a1, a0, b1, b0, c1, c0); \
+ SUB2_LIMB32_CARRY_IN_OUT(a3, a2, b3, b2, c3, c2, __carry10_32); \
+ SUB2_LIMB32_CARRY_IN_OUT(a5, a4, b5, b4, c5, c4, __carry10_32); \
+ SUB2_LIMB32_CARRY_IN_OUT(a7, a6, b7, b6, c7, c6, __carry10_32); \
+ SUB2_LIMB32_CARRY_IN(a9, a8, b9, b8, c9, c8, __carry10_32); \
+ } while (0)
+
+#define SUB14_LIMB32(a13, a12, a11, a10, a9, a8, a7, \
+ a6, a5, a4, a3, a2, a1, a0, \
+ b13, b12, b11, b10, b9, b8, b7, \
+ b6, b5, b4, b3, b2, b1, b0, \
+ c13, c12, c11, c10, c9, c8, c7, \
+ c6, c5, c4, c3, c2, c1, c0) do { \
+ mpi_limb_t __carry14_32; \
+ SUB2_LIMB32_CARRY_OUT(__carry14_32, a1, a0, b1, b0, c1, c0); \
+ SUB2_LIMB32_CARRY_IN_OUT(a3, a2, b3, b2, c3, c2, __carry14_32); \
+ SUB2_LIMB32_CARRY_IN_OUT(a5, a4, b5, b4, c5, c4, __carry14_32); \
+ SUB2_LIMB32_CARRY_IN_OUT(a7, a6, b7, b6, c7, c6, __carry14_32); \
+ SUB2_LIMB32_CARRY_IN_OUT(a9, a8, b9, b8, c9, c8, __carry14_32); \
+ SUB2_LIMB32_CARRY_IN_OUT(a11, a10, b11, b10, c11, c10, __carry14_32); \
+ SUB2_LIMB32_CARRY_IN(a13, a12, b13, b12, c13, c12, __carry14_32); \
} while (0)
#endif /* __i386__ */
@@ -820,7 +896,6 @@ LIMB64_HILO(mpi_limb_t hi, mpi_limb_t lo)
"Ir" ((mpi_limb_t)(C0)) \
: "cc")
-
#define SUB6_LIMB32(A5, A4, A3, A2, A1, A0, B5, B4, B3, B2, B1, B0, \
C5, C4, C3, C2, C1, C0) do { \
mpi_limb_t __borrow6_32; \
@@ -875,7 +950,13 @@ LIMB64_HILO(mpi_limb_t hi, mpi_limb_t lo)
C2.hi, C2.lo, C1.hi, C1.lo, C0.hi, C0.lo)
#endif
-#if defined(ADD6_LIMB32)
+#if defined(ADD8_LIMB32)
+/* A[0..3] = B[0..3] + C[0..3] */
+#define ADD4_LIMB64(A3, A2, A1, A0, B3, B2, B1, B0, C3, C2, C1, C0) \
+ ADD8_LIMB32(A3.hi, A3.lo, A2.hi, A2.lo, A1.hi, A1.lo, A0.hi, A0.lo, \
+ B3.hi, B3.lo, B2.hi, B2.lo, B1.hi, B1.lo, B0.hi, B0.lo, \
+ C3.hi, C3.lo, C2.hi, C2.lo, C1.hi, C1.lo, C0.hi, C0.lo)
+#elif defined(ADD6_LIMB32)
/* A[0..3] = B[0..3] + C[0..3] */
#define ADD4_LIMB64(A3, A2, A1, A0, B3, B2, B1, B0, C3, C2, C1, C0) do { \
mpi_limb_t __carry4; \
@@ -888,6 +969,28 @@ LIMB64_HILO(mpi_limb_t hi, mpi_limb_t lo)
} while (0)
#endif
+#if defined(ADD10_LIMB32)
+/* A[0..4] = B[0..4] + C[0..4] */
+#define ADD5_LIMB64(A4, A3, A2, A1, A0, B4, B3, B2, B1, B0, \
+ C4, C3, C2, C1, C0) \
+ ADD10_LIMB32(A4.hi, A4.lo, A3.hi, A3.lo, A2.hi, A2.lo, A1.hi, A1.lo, \
+ A0.hi, A0.lo, B4.hi, B4.lo, B3.hi, B3.lo, B2.hi, B2.lo, \
+ B1.hi, B1.lo, B0.hi, B0.lo, C4.hi, C4.lo, C3.hi, C3.lo, \
+ C2.hi, C2.lo, C1.hi, C1.lo, C0.hi, C0.lo)
+#endif
+
+#if defined(ADD14_LIMB32)
+/* A[0..6] = B[0..6] + C[0..6] */
+#define ADD7_LIMB64(A6, A5, A4, A3, A2, A1, A0, B6, B5, B4, B3, B2, B1, B0, \
+ C6, C5, C4, C3, C2, C1, C0) \
+ ADD14_LIMB32(A6.hi, A6.lo, A5.hi, A5.lo, A4.hi, A4.lo, A3.hi, A3.lo, \
+ A2.hi, A2.lo, A1.hi, A1.lo, A0.hi, A0.lo, B6.hi, B6.lo, \
+ B5.hi, B5.lo, B4.hi, B4.lo, B3.hi, B3.lo, B2.hi, B2.lo, \
+ B1.hi, B1.lo, B0.hi, B0.lo, C6.hi, C6.lo, C5.hi, C5.lo, \
+ C4.hi, C4.lo, C3.hi, C3.lo, C2.hi, C2.lo, C1.hi, C1.lo, \
+ C0.hi, C0.lo)
+#endif
+
#if defined(SUB4_LIMB32)
/* A[0..1] = B[0..1] - C[0..1] */
#define SUB2_LIMB64(A1, A0, B1, B0, C1, C0) \
@@ -914,7 +1017,13 @@ LIMB64_HILO(mpi_limb_t hi, mpi_limb_t lo)
C2.hi, C2.lo, C1.hi, C1.lo, C0.hi, C0.lo)
#endif
-#if defined(SUB6_LIMB32)
+#if defined(SUB8_LIMB32)
+/* A[0..3] = B[0..3] - C[0..3] */
+#define SUB4_LIMB64(A3, A2, A1, A0, B3, B2, B1, B0, C3, C2, C1, C0) \
+ SUB8_LIMB32(A3.hi, A3.lo, A2.hi, A2.lo, A1.hi, A1.lo, A0.hi, A0.lo, \
+ B3.hi, B3.lo, B2.hi, B2.lo, B1.hi, B1.lo, B0.hi, B0.lo, \
+ C3.hi, C3.lo, C2.hi, C2.lo, C1.hi, C1.lo, C0.hi, C0.lo)
+#elif defined(SUB6_LIMB32)
/* A[0..3] = B[0..3] - C[0..3] */
#define SUB4_LIMB64(A3, A2, A1, A0, B3, B2, B1, B0, C3, C2, C1, C0) do { \
mpi_limb_t __borrow4; \
@@ -927,6 +1036,28 @@ LIMB64_HILO(mpi_limb_t hi, mpi_limb_t lo)
} while (0)
#endif
+#if defined(SUB10_LIMB32)
+/* A[0..4] = B[0..4] - C[0..4] */
+#define SUB5_LIMB64(A4, A3, A2, A1, A0, B4, B3, B2, B1, B0, \
+ C4, C3, C2, C1, C0) \
+ SUB10_LIMB32(A4.hi, A4.lo, A3.hi, A3.lo, A2.hi, A2.lo, A1.hi, A1.lo, \
+ A0.hi, A0.lo, B4.hi, B4.lo, B3.hi, B3.lo, B2.hi, B2.lo, \
+ B1.hi, B1.lo, B0.hi, B0.lo, C4.hi, C4.lo, C3.hi, C3.lo, \
+ C2.hi, C2.lo, C1.hi, C1.lo, C0.hi, C0.lo)
+#endif
+
+#if defined(SUB14_LIMB32)
+/* A[0..6] = B[0..6] - C[0..6] */
+#define SUB7_LIMB64(A6, A5, A4, A3, A2, A1, A0, B6, B5, B4, B3, B2, B1, B0, \
+ C6, C5, C4, C3, C2, C1, C0) \
+ SUB14_LIMB32(A6.hi, A6.lo, A5.hi, A5.lo, A4.hi, A4.lo, A3.hi, A3.lo, \
+ A2.hi, A2.lo, A1.hi, A1.lo, A0.hi, A0.lo, B6.hi, B6.lo, \
+ B5.hi, B5.lo, B4.hi, B4.lo, B3.hi, B3.lo, B2.hi, B2.lo, \
+ B1.hi, B1.lo, B0.hi, B0.lo, C6.hi, C6.lo, C5.hi, C5.lo, \
+ C4.hi, C4.lo, C3.hi, C3.lo, C2.hi, C2.lo, C1.hi, C1.lo, \
+ C0.hi, C0.lo)
+#endif
+
#endif /* BYTES_PER_MPI_LIMB == 4 */
--
2.30.2