gador/nixpkgs
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

glibc: add NVCC include guards to fix ARM CUDA compilation

Browse Source
This commit is contained in:
Connor Baker
2023-11-06 00:53:11 +00:00
parent f190372f1f
commit 2b47345ff5
2 changed files with 50 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
pkgs/development/libraries/glibc/0001-aarch64-math-vector.h-add-NVCC-include-guard.patch
+37
View File
@@ -0,0 +1,37 @@
From 44d0a3a9bd8c6fe59f6ccb44206a50a900bfcf4a Mon Sep 17 00:00:00 2001
From: Connor Baker <connor.baker@tweag.io>
Date: Tue, 31 Oct 2023 14:30:24 +0000
Subject: [PATCH] aarch64/math-vector.h: add NVCC include guard
---
sysdeps/aarch64/fpu/bits/math-vector.h | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/sysdeps/aarch64/fpu/bits/math-vector.h b/sysdeps/aarch64/fpu/bits/math-vector.h
index 7c200599c1..583a426494 100644
--- a/sysdeps/aarch64/fpu/bits/math-vector.h
+++ b/sysdeps/aarch64/fpu/bits/math-vector.h
@@ -25,17 +25,17 @@
/* Get default empty definitions for simd declarations. */
#include <bits/libm-simd-decl-stubs.h>
-#if __GNUC_PREREQ(9, 0)
+#if __GNUC_PREREQ(9, 0) && !defined(__CUDACC__)
# define __ADVSIMD_VEC_MATH_SUPPORTED
typedef __Float32x4_t __f32x4_t;
typedef __Float64x2_t __f64x2_t;
-#elif __glibc_clang_prereq(8, 0)
+#elif __glibc_clang_prereq(8, 0) && !defined(__CUDACC__)
# define __ADVSIMD_VEC_MATH_SUPPORTED
typedef __attribute__ ((__neon_vector_type__ (4))) float __f32x4_t;
typedef __attribute__ ((__neon_vector_type__ (2))) double __f64x2_t;
#endif
-#if __GNUC_PREREQ(10, 0) || __glibc_clang_prereq(11, 0)
+#if (__GNUC_PREREQ(10, 0) || __glibc_clang_prereq(11, 0)) && !defined(__CUDACC__)
# define __SVE_VEC_MATH_SUPPORTED
typedef __SVFloat32_t __sv_f32_t;
typedef __SVFloat64_t __sv_f64_t;
--
2.42.0
pkgs/development/libraries/glibc/common.nix
+13 -1
View File
@@ -97,6 +97,18 @@ stdenv.mkDerivation ({
*/
./reenable_DT_HASH.patch
]
/* NVCC does not support ARM intrinsics. Since <math.h> is pulled in by almost
every HPC piece of software, without this patch CUDA compilation on ARM
is effectively broken. See
https://forums.developer.nvidia.com/t/nvcc-fails-to-build-with-arm-neon-instructions-cpp-vs-cu/248355/2.
*/
++ (
let
isAarch64 = stdenv.buildPlatform.isAarch64 || stdenv.hostPlatform.isAarch64;
isLinux = stdenv.buildPlatform.isLinux || stdenv.hostPlatform.isLinux;
in
lib.optional (isAarch64 && isLinux) ./0001-aarch64-math-vector.h-add-NVCC-include-guard.patch
)
++ lib.optional stdenv.hostPlatform.isMusl ./fix-rpc-types-musl-conflicts.patch
++ lib.optional stdenv.buildPlatform.isDarwin ./darwin-cross-build.patch;
@@ -276,7 +288,7 @@ stdenv.mkDerivation ({
license = licenses.lgpl2Plus;
maintainers = with maintainers; [ eelco ma27 ];
maintainers = with maintainers; [ eelco ma27 connorbaker ];
platforms = platforms.linux;
} // (args.meta or {});
})