gentoo-mirror/gentoo
1
0
Fork 0
mirror of https://github.com/gentoo-mirror/gentoo.git synced 2026-08-24 18:58:08 -07:00
Code Issues Projects Releases Wiki Activity

dev-libs/libgcrypt: backport asm fixes

Browse Source 
* Backport fixes for x86 with less/no optimisation
* Backport HPPA asm fix

Closes: https://bugs.gentoo.org/915060
Closes: https://bugs.gentoo.org/925284
Signed-off-by: Sam James <sam@gentoo.org>
This commit is contained in:
Sam James
2024-03-04 02:37:32 +00:00
parent 36aa5adc70
commit 361b44c3bb
4 changed files with 810 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

110
dev-libs/libgcrypt/files/libgcrypt-1.10.3-hppa.patch Normal file
View File

@@ -0,0 +1,110 @@
https://bugs.gentoo.org/925284
https://git.gnupg.org/cgi-bin/gitweb.cgi?p=libgcrypt.git;a=commit;h=75e9bcccb69a9dea67d90840bd295bbd1749cea3
From 75e9bcccb69a9dea67d90840bd295bbd1749cea3 Mon Sep 17 00:00:00 2001
From: NIIBE Yutaka <gniibe@fsij.org>
Date: Mon, 4 Mar 2024 09:00:59 +0900
Subject: [PATCH] mpi: Fix ECC computation on hppa.
* mpi/ec-inline.h [__hppa] (ADD4_LIMB32, SUB4_LIMB32): New.
* mpi/longlong.h [__hppa] (add_ssaaaa, sub_ddmmss): Add __CLOBBER_CC.
--
Cherry-pick master commit of:
b757f4130af987bdfc769b754b6e9e27882c349c
GnuPG-bug-id: 7022
Signed-off-by: NIIBE Yutaka <gniibe@fsij.org>
---
mpi/ec-inline.h | 40 ++++++++++++++++++++++++++++++++++++++++
mpi/longlong.h | 12 ++++++------
2 files changed, 46 insertions(+), 6 deletions(-)
diff --git a/mpi/ec-inline.h b/mpi/ec-inline.h
index 0ffdf8eb..c24d5352 100644
--- a/mpi/ec-inline.h
+++ b/mpi/ec-inline.h
@@ -921,6 +921,46 @@ LIMB64_HILO(mpi_limb_t hi, mpi_limb_t lo)
#endif /* HAVE_COMPATIBLE_GCC_ARM_PLATFORM_AS */
+#if defined (__hppa) && __GNUC__ >= 4
+#define ADD4_LIMB32(A3, A2, A1, A0, B3, B2, B1, B0, C3, C2, C1, C0) \
+ __asm__ ("add %7,%11,%3\n\t" \
+ "addc %6,%10,%2\n\t" \
+ "addc %5,%9,%1\n\t" \
+ "addc %4,%8,%0" \
+ : "=r" (A3), \
+ "=&r" (A2), \
+ "=&r" (A1), \
+ "=&r" (A0) \
+ : "rM" ((mpi_limb_t)(B3)), \
+ "rM" ((mpi_limb_t)(B2)), \
+ "rM" ((mpi_limb_t)(B1)), \
+ "rM" ((mpi_limb_t)(B0)), \
+ "rM" ((mpi_limb_t)(C3)), \
+ "rM" ((mpi_limb_t)(C2)), \
+ "rM" ((mpi_limb_t)(C1)), \
+ "rM" ((mpi_limb_t)(C0)) \
+ : "cc")
+
+#define SUB4_LIMB32(A3, A2, A1, A0, B3, B2, B1, B0, C3, C2, C1, C0) \
+ __asm__ ("sub %7,%11,%3\n\t" \
+ "subb %6,%10,%2\n\t" \
+ "subb %5,%9,%1\n\t" \
+ "subb %4,%8,%0\n\t" \
+ : "=r" (A3), \
+ "=&r" (A2), \
+ "=&r" (A1), \
+ "=&r" (A0) \
+ : "rM" ((mpi_limb_t)(B3)), \
+ "rM" ((mpi_limb_t)(B2)), \
+ "rM" ((mpi_limb_t)(B1)), \
+ "rM" ((mpi_limb_t)(B0)), \
+ "rM" ((mpi_limb_t)(C3)), \
+ "rM" ((mpi_limb_t)(C2)), \
+ "rM" ((mpi_limb_t)(C1)), \
+ "rM" ((mpi_limb_t)(C0)) \
+ : "cc")
+
+#endif /* __hppa */
/* Common 32-bit arch addition/subtraction macros. */
diff --git a/mpi/longlong.h b/mpi/longlong.h
index c299534c..1ab70e7e 100644
--- a/mpi/longlong.h
+++ b/mpi/longlong.h
@@ -394,23 +394,23 @@ extern UDItype __udiv_qrnnd ();
***************************************/
#if defined (__hppa) && W_TYPE_SIZE == 32
# define add_ssaaaa(sh, sl, ah, al, bh, bl) \
- __asm__ (" add %4,%5,%1\n" \
- " addc %2,%3,%0" \
+ __asm__ ("add %4,%5,%1\n\t" \
+ "addc %2,%3,%0" \
: "=r" ((USItype)(sh)), \
"=&r" ((USItype)(sl)) \
: "%rM" ((USItype)(ah)), \
"rM" ((USItype)(bh)), \
"%rM" ((USItype)(al)), \
- "rM" ((USItype)(bl)))
+ "rM" ((USItype)(bl)) __CLOBBER_CC)
# define sub_ddmmss(sh, sl, ah, al, bh, bl) \
- __asm__ (" sub %4,%5,%1\n" \
- " subb %2,%3,%0" \
+ __asm__ ("sub %4,%5,%1\n\t" \
+ "subb %2,%3,%0" \
: "=r" ((USItype)(sh)), \
"=&r" ((USItype)(sl)) \
: "rM" ((USItype)(ah)), \
"rM" ((USItype)(bh)), \
"rM" ((USItype)(al)), \
- "rM" ((USItype)(bl)))
+ "rM" ((USItype)(bl)) __CLOBBER_CC)
# if defined (_PA_RISC1_1)
# define umul_ppmm(wh, wl, u, v) \
do { \
--
2.30.2

428
dev-libs/libgcrypt/files/libgcrypt-1.10.3-x86-refactor.patch Normal file
View File

@@ -0,0 +1,428 @@
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

94
dev-libs/libgcrypt/files/libgcrypt-1.10.3-x86.patch Normal file
View File

@@ -0,0 +1,94 @@
https://bugs.gentoo.org/915060
https://git.gnupg.org/cgi-bin/gitweb.cgi?p=libgcrypt.git;a=commit;h=08b88b4012f7837736b8d29a3689ce3fff2a10c8
From 08b88b4012f7837736b8d29a3689ce3fff2a10c8 Mon Sep 17 00:00:00 2001
From: Jussi Kivilinna <jussi.kivilinna@iki.fi>
Date: Sat, 16 Dec 2023 19:50:23 +0200
Subject: [PATCH] mpi/ec-nist: fix for -Og build failure on i386
* mpi/ec-nist.c (_gcry_mpi_ec_nist256_mod)
(_gcry_mpi_ec_nist384_mod): Load p_mult constant with carry offset
to stack.
--
Cherry pick master commit of:
90097bd2f41c217dc5c666570e5680f432cf92d3
Patch fixes compilation error on i386 with -Og optimization level.
In file included from ../../mpi/ec-nist.c:34:
../../mpi/ec-nist.c: In function '_gcry_mpi_ec_nist256_mod':
../../mpi/ec-inline.h:701:3: error: 'asm' operand has impossible constraints
701 | __asm__ ("subl %11, %3\n" \
| ^~~~~~~
../../mpi/ec-inline.h:894:9: note: in expansion of macro 'SUB4_LIMB32'
894 | SUB4_LIMB32(A1.hi, A1.lo, A0.hi, A0.lo, \
| ^~~~~~~~~~~
../../mpi/ec-inline.h:1009:5: note: in expansion of macro 'SUB2_LIMB64'
1009 | SUB2_LIMB64(A4, A3, B4, B3, C4, C3); \
| ^~~~~~~~~~~
../../mpi/ec-nist.c:474:3: note: in expansion of macro 'SUB5_LIMB64'
474 | SUB5_LIMB64 (s[4], s[3], s[2], s[1], s[0],
| ^~~~~~~~~~~
Appears that in problematic function, too many registers end up being
allocated for addressing and there is not enough register left for
asm input/output (4 registers needed for this block). Problem can be
workaround by reducing needed addressing registers by pushing
`p_mult[carry + ...]` values to stack. On other compiler flag levels
and architectures, compiler should be able to optimize away this
extra copying and have not effect on performance.
GnuPG-bug-id: T6892
Signed-off-by: Jussi Kivilinna <jussi.kivilinna@iki.fi>
---
mpi/ec-nist.c | 23 ++++++++++++++++-------
1 file changed, 16 insertions(+), 7 deletions(-)
diff --git a/mpi/ec-nist.c b/mpi/ec-nist.c
index f792405c..559d02d9 100644
--- a/mpi/ec-nist.c
+++ b/mpi/ec-nist.c
@@ -471,11 +471,15 @@ _gcry_mpi_ec_nist256_mod (gcry_mpi_t w, mpi_ec_t ctx)
carry = LO32_LIMB64(s[4]);
+ /* Load values to stack to ease register pressure on i386. */
+ e[0] = p_mult[carry + 4][0];
+ e[1] = p_mult[carry + 4][1];
+ e[2] = p_mult[carry + 4][2];
+ e[3] = p_mult[carry + 4][3];
+ e[4] = p_mult[carry + 4][4];
SUB5_LIMB64 (s[4], s[3], s[2], s[1], s[0],
s[4], s[3], s[2], s[1], s[0],
- p_mult[carry + 4][4], p_mult[carry + 4][3],
- p_mult[carry + 4][2], p_mult[carry + 4][1],
- p_mult[carry + 4][0]);
+ e[4], e[3], e[2], e[1], e[0]);
/* Add 1*P */
ADD5_LIMB64 (d[4], d[3], d[2], d[1], d[0],
@@ -749,12 +753,17 @@ _gcry_mpi_ec_nist384_mod (gcry_mpi_t w, mpi_ec_t ctx)
carry = LO32_LIMB64(s[6]);
+ /* Load values to stack to ease register pressure on i386. */
+ x[0] = p_mult[carry + 3][0];
+ x[1] = p_mult[carry + 3][1];
+ x[2] = p_mult[carry + 3][2];
+ x[3] = p_mult[carry + 3][3];
+ x[4] = p_mult[carry + 3][4];
+ x[5] = p_mult[carry + 3][5];
+ x[6] = p_mult[carry + 3][6];
SUB7_LIMB64 (s[6], s[5], s[4], s[3], s[2], s[1], s[0],
s[6], s[5], s[4], s[3], s[2], s[1], s[0],
- p_mult[carry + 3][6], p_mult[carry + 3][5],
- p_mult[carry + 3][4], p_mult[carry + 3][3],
- p_mult[carry + 3][2], p_mult[carry + 3][1],
- p_mult[carry + 3][0]);
+ x[6], x[5], x[4], x[3], x[2], x[1], x[0]);
ADD7_LIMB64 (d[6], d[5], d[4], d[3], d[2], d[1], d[0],
s[6], s[5], s[4], s[3], s[2], s[1], s[0],
--
2.30.2

178
dev-libs/libgcrypt/libgcrypt-1.10.3-r2.ebuild Normal file
View File

@@ -0,0 +1,178 @@
# Copyright 1999-2024 Gentoo Authors
# Distributed under the terms of the GNU General Public License v2
EAPI=8
VERIFY_SIG_OPENPGP_KEY_PATH=/usr/share/openpgp-keys/gnupg.asc
inherit autotools flag-o-matic linux-info multilib-minimal toolchain-funcs verify-sig
DESCRIPTION="General purpose crypto library based on the code used in GnuPG"
HOMEPAGE="https://www.gnupg.org/"
SRC_URI="mirror://gnupg/${PN}/${P}.tar.bz2"
SRC_URI+=" verify-sig? ( mirror://gnupg/${PN}/${P}.tar.bz2.sig )"
LICENSE="LGPL-2.1+ GPL-2+ MIT"
SLOT="0/20" # subslot = soname major version
KEYWORDS="~alpha ~amd64 ~arm ~arm64 ~hppa ~ia64 ~loong ~m68k ~mips ~ppc ~ppc64 ~riscv ~s390 ~sparc ~x86 ~amd64-linux ~x86-linux ~arm64-macos ~ppc-macos ~x64-macos ~x64-solaris"
IUSE="+asm cpu_flags_arm_neon cpu_flags_arm_aes cpu_flags_arm_sha1 cpu_flags_arm_sha2 cpu_flags_ppc_altivec cpu_flags_ppc_vsx2 cpu_flags_ppc_vsx3 cpu_flags_x86_aes cpu_flags_x86_avx cpu_flags_x86_avx2 cpu_flags_x86_padlock cpu_flags_x86_sha cpu_flags_x86_sse4_1 doc +getentropy static-libs"
# Build system only has --disable-arm-crypto-support right now
# If changing this, update src_configure logic too.
# ARM CPUs seem to, right now, support all-or-nothing for crypto extensions,
# but this looks like it might change in future. This is just a safety check
# in case people somehow do have a CPU which only supports some. They must
# for now disable them all if that's the case.
REQUIRED_USE="
cpu_flags_arm_aes? ( cpu_flags_arm_sha1 cpu_flags_arm_sha2 )
cpu_flags_arm_sha1? ( cpu_flags_arm_aes cpu_flags_arm_sha2 )
cpu_flags_arm_sha2? ( cpu_flags_arm_aes cpu_flags_arm_sha1 )
cpu_flags_ppc_vsx3? ( cpu_flags_ppc_altivec cpu_flags_ppc_vsx2 )
cpu_flags_ppc_vsx2? ( cpu_flags_ppc_altivec )
"
RDEPEND="
>=dev-libs/libgpg-error-1.25[${MULTILIB_USEDEP}]
getentropy? (
kernel_linux? (
elibc_glibc? ( >=sys-libs/glibc-2.25 )
elibc_musl? ( >=sys-libs/musl-1.1.20 )
)
)
"
DEPEND="${RDEPEND}"
BDEPEND="
doc? ( virtual/texi2dvi )
verify-sig? ( sec-keys/openpgp-keys-gnupg )
"
PATCHES=(
"${FILESDIR}"/${PN}-multilib-syspath.patch
"${FILESDIR}"/${PN}-powerpc-darwin.patch
"${FILESDIR}"/${PN}-1.9.4-no-fgrep-libgcrypt-config.patch
"${FILESDIR}"/${PN}-1.10.3-x86.patch
"${FILESDIR}"/${PN}-1.10.3-x86-refactor.patch
"${FILESDIR}"/${PN}-1.10.3-hppa.patch
)
MULTILIB_CHOST_TOOLS=(
/usr/bin/libgcrypt-config
)
pkg_pretend() {
if [[ ${MERGE_TYPE} == buildonly ]]; then
return
fi
if use kernel_linux && use getentropy; then
unset KV_FULL
get_running_version
if [[ -n ${KV_FULL} ]] && kernel_is -lt 3 17; then
eerror "The getentropy function requires the getrandom syscall."
eerror "This was introduced in Linux 3.17."
eerror "Your system is currently running Linux ${KV_FULL}."
eerror "Disable the 'getentropy' USE flag or upgrade your kernel."
die "Kernel is too old for getentropy"
fi
fi
}
pkg_setup() {
:
}
src_prepare() {
default
eautoreconf
}
multilib_src_configure() {
if [[ ${CHOST} == *86*-solaris* ]] ; then
# ASM code uses GNU ELF syntax, divide in particular, we need to
# allow this via ASFLAGS, since we don't have a flag-o-matic
# function for that, we'll have to abuse cflags for this
append-cflags -Wa,--divide
fi
if [[ ${CHOST} == powerpc* ]] ; then
# ./configure does a lot of automagic, prevent that
# generic ppc32+ppc64 altivec
use cpu_flags_ppc_altivec || local -x gcry_cv_cc_ppc_altivec=no
use cpu_flags_ppc_altivec || local -x gcry_cv_cc_ppc_altivec_cflags=no
# power8 vector extension, aka arch 2.07 ISA, also checked below via ppc-crypto-support
use cpu_flags_ppc_vsx2 || local -x gcry_cv_gcc_inline_asm_ppc_altivec=no
# power9 vector extension, aka arch 3.00 ISA
use cpu_flags_ppc_vsx3 || local -x gcry_cv_gcc_inline_asm_ppc_arch_3_00=no
fi
# Workaround for GCC < 11.3 bug
# https://git.gnupg.org/cgi-bin/gitweb.cgi?p=libgcrypt.git;a=commitdiff;h=0b399721ce9709ae25f9d2050360c5ab2115ae29
# https://dev.gnupg.org/T5581
# https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102124
if use arm64 && tc-is-gcc && (($(gcc-major-version) == 11)) &&
(($(gcc-minor-version) <= 2)) && (($(gcc-micro-version) == 0)) ; then
append-flags -fno-tree-loop-vectorize
fi
append-ldflags $(test-flags-CCLD -Wl,--undefined-version)
local myeconfargs=(
CC_FOR_BUILD="$(tc-getBUILD_CC)"
--enable-noexecstack
$(use_enable cpu_flags_arm_neon neon-support)
# See REQUIRED_USE comment above
$(use_enable cpu_flags_arm_aes arm-crypto-support)
$(use_enable cpu_flags_ppc_vsx2 ppc-crypto-support)
$(use_enable cpu_flags_x86_aes aesni-support)
$(use_enable cpu_flags_x86_avx avx-support)
$(use_enable cpu_flags_x86_avx2 avx2-support)
$(use_enable cpu_flags_x86_padlock padlock-support)
$(use_enable cpu_flags_x86_sha shaext-support)
$(use_enable cpu_flags_x86_sse4_1 sse41-support)
# required for sys-power/suspend[crypt], bug 751568
$(use_enable static-libs static)
# disabled due to various applications requiring privileges
# after libgcrypt drops them (bug #468616)
--without-capabilities
# http://trac.videolan.org/vlc/ticket/620
$([[ ${CHOST} == *86*-darwin* ]] && echo "--disable-asm")
$(use asm || echo "--disable-asm")
GPG_ERROR_CONFIG="${ESYSROOT}/usr/bin/${CHOST}-gpg-error-config"
)
if use kernel_linux; then
# --enable-random=getentropy requires getentropy/getrandom.
# --enable-random=linux enables legacy code that tries getrandom
# and falls back to reading /dev/random.
myeconfargs+=( --enable-random=$(usex getentropy getentropy linux) )
fi
ECONF_SOURCE="${S}" econf "${myeconfargs[@]}" \
$("${S}/configure" --help | grep -o -- '--without-.*-prefix')
}
multilib_src_compile() {
default
multilib_is_native_abi && use doc && VARTEXFONTS="${T}/fonts" emake -C doc gcrypt.pdf
}
multilib_src_test() {
# t-secmem and t-sexp need mlock which requires extra privileges; nspawn
# at least disallows that by default.
local -x GCRYPT_IN_ASAN_TEST=1
default
}
multilib_src_install() {
emake DESTDIR="${D}" install
multilib_is_native_abi && use doc && dodoc doc/gcrypt.pdf
}
multilib_src_install_all() {
default
find "${ED}" -type f -name '*.la' -delete || die
}