Skip to content

[CHIA-3769] - Update GMP to 6.3.0 and include new patch files #270

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
emlowe wants to merge 4 commits into
base: main
Choose a base branch
from
+428 −2,598
Open

[CHIA-3769] - Update GMP to 6.3.0 and include new patch files #270

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
166f8a2
add patch files for GMP6.3.0 and adjust build steps
emlowe Oct 1, 2025
fbec9a4
Update compat.patch for compatibility changes
altendky Oct 2, 2025
65a521b
a different take on the patching
altendky Oct 2, 2025
0ceeeb9
Merge pull request #273 from Chia-Network/altendky-patch-2
emlowe Nov 5, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 5 additions & 6 deletions pyproject.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,12 +19,11 @@ yum -y install epel-release \
&& echo "epel-release installed" \
&& yum -y install boost-devel lzip \
&& echo "boost-devel and lzip installed" \
&& curl -L https://ftp.gnu.org/gnu/gmp/gmp-6.2.1.tar.lz | tar x --lzip \
&& cp src/lib/gmp-patch-6.2.1/longlong.h gmp-6.2.1/ \
&& cp src/lib/gmp-patch-6.2.1/compat.c gmp-6.2.1/ \
&& cp src/lib/gmp-patch-6.2.1/mpz/inp_raw.c gmp-6.2.1/mpz \
&& cd gmp-6.2.1 && ./configure --enable-fat --enable-cxx \
&& make && make install && cd .. && rm -rf gmp-6.2.1 \
&& curl -L https://ftp.gnu.org/gnu/gmp/gmp-6.3.0.tar.lz | tar x --lzip \
&& cd gmp-6.3.0 \
&& patch < ../src/lib/gmp-6.3.0.diff \
&& ./configure --enable-fat --enable-cxx \
&& make && make install && cd .. && rm -rf gmp-6.3.0 \
&& cmake --version \
&& uname -a \
"""
Expand Down
220 changes: 220 additions & 0 deletions src/lib/gmp-6.2.1.diff
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,220 @@
diff --git a/compat.c b/compat.c
index b4b44ce..3f563dd 100644
--- a/compat.c
+++ b/compat.c
@@ -31,6 +31,12 @@ see https://www.gnu.org/licenses/. */
#include <stdio.h>
#include "gmp-impl.h"

+/* RUNTIMECPUID */
+int bCheckedBMI = 0;
+int bBMI1 = 0;
+int bBMI2 = 0;
+int bCheckedLZCNT = 0;
+int bLZCNT = 0;

/* mpn_divexact_by3 was a function in gmp 3.0.1, but as of gmp 3.1 it's a
macro calling mpn_divexact_by3c. */
diff --git a/longlong.h b/longlong.h
index edbaf56..c0a7468 100644
--- a/longlong.h
+++ b/longlong.h
@@ -1040,6 +1040,86 @@ extern UWtype __MPN(udiv_qrnnd) (UWtype *, UWtype, UWtype, UWtype);
#endif /* 80x86 */

#if defined (__amd64__) && W_TYPE_SIZE == 64
+
+#ifndef RUNTIMECPUID
+#define RUNTIMECPUID
+
+extern int bCheckedBMI;
+extern int bBMI1;
+extern int bBMI2;
+
+inline void hasBMI()
+{
+ if(bCheckedBMI)
+ return;
+
+ bCheckedBMI = 1;
+ int info[4] = {0};
+#if defined(_MSC_VER)
+ __cpuid(info, 0x7);
+#elif defined(__GNUC__) || defined(__clang__)
+#if defined(ARCH_X86) && defined(__PIC__)
+ __asm__ __volatile__ (
+ "xchg{l} {%%}ebx, %k1;"
+ "cpuid;"
+ "xchg{l} {%%}ebx, %k1;"
+ : "=a"(info[0]), "=&r"(info[1]), "=c"(info[2]), "=d"(info[3]) : "a"(0x7), "c"(0)
+ );
+#else
+ __asm__ __volatile__ (
+ "cpuid" : "=a"(info[0]), "=b"(info[1]), "=c"(info[2]), "=d"(info[3]) : "a"(0x7), "c"(0)
+ );
+#endif
+#endif
+ bBMI1 = ((info[1] & (1 << 3)) != 0);
+ bBMI2 = ((info[1] & (1 << 8)) != 0);
+}
+
+inline int hasBMI1()
+{
+ hasBMI();
+ return bBMI1;
+}
+
+inline int hasBMI2()
+{
+ hasBMI();
+ return bBMI2;
+}
+
+extern int bCheckedLZCNT;
+extern int bLZCNT;
+
+inline int hasLZCNT()
+{
+ if(bCheckedLZCNT)
+ return bLZCNT;
+
+ bCheckedLZCNT = 1;
+ int info[4] = {0};
+ #if defined(_MSC_VER)
+ __cpuid(info, 0x80000001);
+ #elif defined(__GNUC__) || defined(__clang__)
+ #if defined(ARCH_X86) && defined(__PIC__)
+ __asm__ __volatile__ (
+ "xchg{l} {%%}ebx, %k1;"
+ "cpuid;"
+ "xchg{l} {%%}ebx, %k1;"
+ : "=a"(info[0]), "=&r"(info[1]), "=c"(info[2]), "=d"(info[3]) : "a"(0x80000001), "c"(0)
+ );
+ #else
+ __asm__ __volatile__ (
+ "cpuid" : "=a"(info[0]), "=b"(info[1]), "=c"(info[2]), "=d"(info[3]) : "a"(0x80000001), "c"(0)
+ );
+ #endif
+ #endif
+
+ bLZCNT = ((info[2] & (1 << 5)) != 0);
+ return bLZCNT;
+}
+
+#endif // RUNTIMECPUID
+
#define add_ssaaaa(sh, sl, ah, al, bh, bl) \
__asm__ ("addq %5,%q1\n\tadcq %3,%q0" \
: "=r" (sh), "=&r" (sl) \
@@ -1050,61 +1130,52 @@ extern UWtype __MPN(udiv_qrnnd) (UWtype *, UWtype, UWtype, UWtype);
: "=r" (sh), "=&r" (sl) \
: "0" ((UDItype)(ah)), "rme" ((UDItype)(bh)), \
"1" ((UDItype)(al)), "rme" ((UDItype)(bl)))
-#if X86_ASM_MULX \
- && (HAVE_HOST_CPU_haswell || HAVE_HOST_CPU_broadwell \
- || HAVE_HOST_CPU_skylake || HAVE_HOST_CPU_bd4 || HAVE_HOST_CPU_zen)
#define umul_ppmm(w1, w0, u, v) \
- __asm__ ("mulx\t%3, %q0, %q1" \
+ if(hasBMI2()) { \
+ __asm__ ("mulx\t%3, %q0, %q1" \
: "=r" (w0), "=r" (w1) \
- : "%d" ((UDItype)(u)), "rm" ((UDItype)(v)))
-#else
-#define umul_ppmm(w1, w0, u, v) \
- __asm__ ("mulq\t%3" \
+ : "%d" ((UDItype)(u)), "rm" ((UDItype)(v))); \
+ } else { \
+ __asm__ ("mulq\t%3" \
: "=a" (w0), "=d" (w1) \
- : "%0" ((UDItype)(u)), "rm" ((UDItype)(v)))
-#endif
+ : "%0" ((UDItype)(u)), "rm" ((UDItype)(v))); \
+ }
#define udiv_qrnnd(q, r, n1, n0, dx) /* d renamed to dx avoiding "=d" */\
__asm__ ("divq %4" /* stringification in K&R C */ \
: "=a" (q), "=d" (r) \
: "0" ((UDItype)(n0)), "1" ((UDItype)(n1)), "rm" ((UDItype)(dx)))

-#if HAVE_HOST_CPU_haswell || HAVE_HOST_CPU_broadwell || HAVE_HOST_CPU_skylake \
- || HAVE_HOST_CPU_k10 || HAVE_HOST_CPU_bd1 || HAVE_HOST_CPU_bd2 \
- || HAVE_HOST_CPU_bd3 || HAVE_HOST_CPU_bd4 || HAVE_HOST_CPU_zen \
- || HAVE_HOST_CPU_bobcat || HAVE_HOST_CPU_jaguar
#define count_leading_zeros(count, x) \
- do { \
- /* This is lzcnt, spelled for older assemblers. Destination and */ \
- /* source must be a 64-bit registers, hence cast and %q. */ \
- __asm__ ("rep;bsr\t%1, %q0" : "=r" (count) : "rm" ((UDItype)(x))); \
- } while (0)
+ if(hasLZCNT()) { \
+ do { \
+ /* This is lzcnt, spelled for older assemblers. Destination and */ \
+ /* source must be a 64-bit registers, hence cast and %q. */ \
+ __asm__ ("rep;bsr\t%1, %q0" : "=r" (count) : "rm" ((UDItype)(x))); \
+ } while (0); \
+ } else { \
+ do { \
+ UDItype __cbtmp; \
+ ASSERT ((x) != 0); \
+ __asm__ ("bsr\t%1,%0" : "=r" (__cbtmp) : "rm" ((UDItype)(x))); \
+ (count) = __cbtmp ^ 63; \
+ } while (0); \
+ }
#define COUNT_LEADING_ZEROS_0 64
-#else
-#define count_leading_zeros(count, x) \
- do { \
- UDItype __cbtmp; \
- ASSERT ((x) != 0); \
- __asm__ ("bsr\t%1,%0" : "=r" (__cbtmp) : "rm" ((UDItype)(x))); \
- (count) = __cbtmp ^ 63; \
- } while (0)
-#endif

-#if HAVE_HOST_CPU_bd2 || HAVE_HOST_CPU_bd3 || HAVE_HOST_CPU_bd4 \
- || HAVE_HOST_CPU_zen || HAVE_HOST_CPU_jaguar
#define count_trailing_zeros(count, x) \
- do { \
- /* This is tzcnt, spelled for older assemblers. Destination and */ \
- /* source must be a 64-bit registers, hence cast and %q. */ \
- __asm__ ("rep;bsf\t%1, %q0" : "=r" (count) : "rm" ((UDItype)(x))); \
- } while (0)
+ if(hasBMI1()) { \
+ do { \
+ /* This is tzcnt, spelled for older assemblers. Destination and */ \
+ /* source must be a 64-bit registers, hence cast and %q. */ \
+ __asm__ ("rep;bsf\t%1, %q0" : "=r" (count) : "rm" ((UDItype)(x))); \
+ } while (0); \
+ } else { \
+ do { \
+ ASSERT ((x) != 0); \
+ __asm__ ("bsf\t%1, %q0" : "=r" (count) : "rm" ((UDItype)(x))); \
+ } while (0); \
+ }
#define COUNT_TRAILING_ZEROS_0 64
-#else
-#define count_trailing_zeros(count, x) \
- do { \
- ASSERT ((x) != 0); \
- __asm__ ("bsf\t%1, %q0" : "=r" (count) : "rm" ((UDItype)(x))); \
- } while (0)
-#endif
#endif /* __amd64__ */

#if defined (__i860__) && W_TYPE_SIZE == 32
diff --git a/mpz/inp_raw.c b/mpz/inp_raw.c
index 378c42b..f88fea9 100644
--- a/mpz/inp_raw.c
+++ b/mpz/inp_raw.c
@@ -88,8 +88,11 @@ mpz_inp_raw (mpz_ptr x, FILE *fp)

abs_csize = ABS (csize);

+ if (UNLIKELY (abs_csize > ~(mp_bitcnt_t) 0 / 8))
+ return 0; /* Bit size overflows */
+
/* round up to a multiple of limbs */
- abs_xsize = BITS_TO_LIMBS (abs_csize*8);
+ abs_xsize = BITS_TO_LIMBS ((mp_bitcnt_t) abs_csize * 8);

if (abs_xsize != 0)
{
Loading
Loading