diff options
author | Simon Marchi <simon.marchi@efficios.com> | 2022-03-16 09:00:27 -0400 |
---|---|---|
committer | Simon Marchi <simon.marchi@polymtl.ca> | 2022-03-16 09:00:27 -0400 |
commit | 978602e83f038106c8254424ca83991d784b37cf (patch) | |
tree | 2713f291ce01c4b0df77e4cd296ca48c090676f4 /bfd/elf64-amdgcn.c | |
parent | f4f95df31b0172bf182e6a516ba23e299323bb2a (diff) | |
download | gdb-978602e83f038106c8254424ca83991d784b37cf.zip gdb-978602e83f038106c8254424ca83991d784b37cf.tar.gz gdb-978602e83f038106c8254424ca83991d784b37cf.tar.bz2 |
bfd: add AMDGCN architecture
Add support for the AMDGCN architecture to BFD.
This is the bare minimum to get
$ ./configure --target=amdgcn-hsa-amdhsa --disable-gas
$ make all-binutils
working later in this series.
The specific AMDGCN models added here are a bit arbitrary, based on
what we intend to initially support in GDB. This list will need to be
updated in the future anyway. The complete up-to-date list of existing
AMDGPU models can be found here:
https://llvm.org/docs/AMDGPUUsage.html#processors
The ELF format for this architecture is documented here:
https://llvm.org/docs/AMDGPUUsage.html#elf-code-object
The flags for the "HSA" OS ABI are properly versioned and documented on
that page. But the NONE, PAL and MESA3D OS ABIs are not well documented
nor versioned. Taking a peek at the LLVM source code, we see that they
encode their flags the same way as HSA v3. For example, for PAL:
https://github.com/llvm/llvm-project/blob/c8b614cd74a92d85936aed5ac7c642af75ffdc29/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp#L601
So at least, we know that all AMDGPU objects (of which AMDGCN objects
are a subset of) at the time of writing encode the specific GPU model in
the EF_AMDGPU_MACH field of e_flags.
bfd/ChangeLog:
* Makefile.am (ALL_MACHINES, ALL_MACHINES_CFILES):
Add cpu-amdgcn.c.
(BFD64_BACKENDS): Add elf64-amdgcn.lo.
(BFD64_BACKENDS_CFILES): Add elf64-amdgcn.c.
* Makefile.in: Re-generate.
* cpu-amdgcn.c: New.
* elf64-amdgcn.c: New.
* archures.c (bfd_architecture): Add bfd_arch_amdgcn and related
mach defines.
(bfd_amdgcn_arch): New.
(bfd_archures_list): Add bfd_amdgcn_arch.
* bfd-in2.h: Re-generate.
* config.bfd: Handle amdgcn* target.
* configure.ac: Handle amdgcn_elf64_le_vec.
* configure: Re-generate.
* elf-bfd.h (elf_target_id): Add AMDGCN_ELF_DATA.
* targets.c (amdgcn_elf64_le_vec): New.
(_bfd_target_vector): Add amdgcn_elf64_le_vec.
include/ChangeLog:
* elf/amdgpu.h: New.
* elf/common.h (ELFOSABI_AMDGPU_HSA): Add.
Change-Id: I969f7b14960797e88891c308749a6e341eece5b2
Diffstat (limited to 'bfd/elf64-amdgcn.c')
-rw-r--r-- | bfd/elf64-amdgcn.c | 80 |
1 files changed, 80 insertions, 0 deletions
diff --git a/bfd/elf64-amdgcn.c b/bfd/elf64-amdgcn.c new file mode 100644 index 0000000..691fa0a --- /dev/null +++ b/bfd/elf64-amdgcn.c @@ -0,0 +1,80 @@ +/* AMDGCN ELF support for BFD. + + Copyright (C) 2019-2022 Free Software Foundation, Inc. + + This file is part of BFD, the Binary File Descriptor library. + + This program is free software; you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. */ + +/* This file handles ELF files that are of the AMDGCN architecture. The + format is documented here: + + https://llvm.org/docs/AMDGPUUsage.html#elf-code-object */ + +#include "sysdep.h" +#include "bfd.h" +#include "libbfd.h" +#include "elf-bfd.h" +#include "elf/amdgpu.h" + +#include <string.h> + +static bool +elf64_amdgcn_object_p (bfd *abfd) +{ + Elf_Internal_Ehdr *hdr = elf_elfheader (abfd); + unsigned int mach; + unsigned char osabi; + unsigned char osabi_version; + + BFD_ASSERT (hdr->e_machine == EM_AMDGPU); + + osabi = hdr->e_ident[EI_OSABI]; + osabi_version = hdr->e_ident[EI_ABIVERSION]; + + /* Objects with OS ABI HSA version 2 encoded the GPU model differently (in a + note), but they are deprecated, so we don't need to support them. Reject + them specifically. + + At the time of writing, all AMDGCN objects encode the specific GPU + model in the EF_AMDGPU_MACH field of e_flags. */ + if (osabi == ELFOSABI_AMDGPU_HSA + && osabi_version < ELFABIVERSION_AMDGPU_HSA_V3) + return false; + + mach = elf_elfheader (abfd)->e_flags & EF_AMDGPU_MACH; + + /* Avoid matching non-AMDGCN AMDGPU objects (e.g. r600). */ + if (mach < EF_AMDGPU_MACH_AMDGCN_MIN) + return false; + + bfd_default_set_arch_mach (abfd, bfd_arch_amdgcn, mach); + return true; +} + + +#define TARGET_LITTLE_SYM amdgcn_elf64_le_vec +#define TARGET_LITTLE_NAME "elf64-amdgcn" +#define ELF_ARCH bfd_arch_amdgcn +#define ELF_TARGET_ID AMDGCN_ELF_DATA +#define ELF_MACHINE_CODE EM_AMDGPU +#define ELF_MAXPAGESIZE 0x10000 /* 64KB */ +#define ELF_COMMONPAGESIZE 0x1000 /* 4KB */ + +#define bfd_elf64_bfd_reloc_type_lookup bfd_default_reloc_type_lookup +#define bfd_elf64_bfd_reloc_name_lookup _bfd_norelocs_bfd_reloc_name_lookup + +#define elf_backend_object_p elf64_amdgcn_object_p + +#include "elf64-target.h" |