dev-util/hipcc
Radeon Open Compute hipcc
-
hipcc-7.1.0~amd64debug test +llvm_slot_20
View
Download
Browse License: Apache-2.0 MIT Overlay: gentoo -
hipcc-7.0.2~amd64debug test +llvm_slot_20
View
Download
Browse License: Apache-2.0 MIT Overlay: gentoo -
hipcc-6.4.3~amd64debug test +llvm_slot_20
View
Download
Browse License: Apache-2.0 MIT Overlay: gentoo
ChangeLog
commit fbd5aa2d66440462b9ae239cded452385383ec3b
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Aug 30 23:39:37 2026 +0200
dev-util/hipcc: drop 7.2.3
Superseded by 7.2.4, the last release of the 7.x series and the rollback
target for the ROCm 10.0 migration. Version retention keeps the last two
plus the last of each previous major series; 7.2.4 and 10.0.0 satisfy both,
so the third version is redundant.
commit e7314ae5263d1e6ac00452f89aa16b3d3cfd07a8
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Aug 30 22:13:22 2026 +0200
dev-util/hipcc: guard the device-libs bitcode sed
The two seds above it were guarded during the 10.0 migration; this one was
not. A silent miss leaves hipcc probing /usr/amdgcn/bitcode instead of
/usr/lib/amdgcn/bitcode, so every device compile fails to find the device
libs -- with the ebuild itself reporting success.
Checked for the double-prefix hazard that made the same pattern a real bug in
dev-libs/rocm-device-libs, where an unanchored global substitution also hit an
already-correct occurrence and produced lib/lib/amdgcn/bitcode. It does not
apply here: hipBin_amd.h has exactly one occurrence (:305) and no
already-prefixed spelling, so s///g is safe. Recorded that in the comment so
the next reader does not have to re-derive it.
src_prepare verified against the therock-10.0 source, and by effect rather
than exit status: one lib/amdgcn/bitcode afterwards, zero lib/lib/amdgcn.
commit 9130556401977404f7ea8bcb23671246e13fb051
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Aug 30 00:51:13 2026 +0200
dev-util/hipcc: add amd-llvm USE flag to select llvm-core/rocm-llvm
This is the single lever that switches the whole ROCm library stack onto AMD's
LLVM fork. rocm.eclass's rocm_use_clang() derives CC/CXX from
`hipconfig --hipclangpath`, so making hipconfig report /usr/lib/rocm-llvm/bin
redirects every library that calls it -- composable-kernel, hipBLASLt, rocBLAS,
rocPRIM, rocRAND, rccl, Tensile, roctracer -- with no per-ebuild change and no
CMAKE_CXX_COMPILER plumbing in nine separate places.
USE=-amd-llvm keeps the previous behaviour (the system clang at the
LLVM_COMPAT slot), so this is opt-in and the base/runtime layer is unaffected.
The path shape differs between the two: Gentoo's LLVM is slotted, so the
default branch inserts $ as a path component
(lib/llvm/<slot>/bin), while AMD's fork installs to a single unslotted
directory, so the amd-llvm branch swaps the component instead
(lib/rocm-llvm/bin).
commit 6d13ab842fcfb654a36f7a4b8688a132489ea1b9
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sat Aug 29 18:08:27 2026 +0200
dev-util/hipcc: build 10.0.0 against llvm:23
LLVM_COMPAT was 22, inherited from the 7.2.4 ebuild it was copied from.
That is correct in isolation -- hipcc builds fine against llvm:22 -- but it
is wrong for a ROCm 10.0 stack, and the failure mode is silent: src_prepare
seds lib/llvm/$/bin into the installed hipcc wrapper, so a 22 here
ships a hipcc that drives clang 22 while dev-libs/rocm-device-libs-10.0.0
requires clang 23 (amdhsa_abi.h and the struct_buffer_load_format builtins,
neither present in 22).
Because it builds green either way, no build-check would have caught this;
it has to be fixed by inspection as part of moving the cohort.
commit 7dbfc474db6b0a67330ea2cae155f278a2028fc4
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Fri Aug 28 08:36:36 2026 +0200
dev-util/hipcc: add 10.0.0
commit 7c384c8cac70105c06770e0ccb7af0fc295f2471
Author: Raukaan Cogbrother <cogbrother@raukaan.local>
Date: Sun May 31 03:29:35 2026 +0200
metadata: normalize tab → 2-space indent in 42 imported files
42 metadata.xml files inherited tab indentation from their origin
overlays (mostly the ROCm cluster, plus a handful of dev-python and
AMD-tooling imports). The overlay convention is 2-space; this brings
them in line with the other 498 files. Maintainer attributions
preserved verbatim.
commit 7efa2280df2a1bab0990f205d0b80a57645467dc
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Fri May 29 14:12:05 2026 +0200
dev-util/hipcc: add 7.2.4
commit 6fb7a0f3df6f86a1dc72f6c97632aea9d7b35d66
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Fri May 8 20:06:46 2026 +0200
dev-util/hipcc: drop stale patch superseded by 7.2.3
commit befa078db38faf2d8262505f90b99d7e49f6b304
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Wed May 6 01:54:25 2026 +0200
dev-util/hipcc: add 7.2.3
commit 8a7082f1ba65632398134de525d263330415ac96
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sat Dec 6 21:51:11 2025 +0100
*/*: drop rocm/hip as newer version is in main tree
Signed-off-by: Ivan S. Titov <iohann.s.titov@gmail.com>
commit cfa997ed467a25f0a98e30e69523013fb12b4136
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sat Oct 18 14:26:09 2025 +0200
added rocm and HIP v. 6.4.4 for now, as main tree lags
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Aug 30 23:39:37 2026 +0200
dev-util/hipcc: drop 7.2.3
Superseded by 7.2.4, the last release of the 7.x series and the rollback
target for the ROCm 10.0 migration. Version retention keeps the last two
plus the last of each previous major series; 7.2.4 and 10.0.0 satisfy both,
so the third version is redundant.
commit e7314ae5263d1e6ac00452f89aa16b3d3cfd07a8
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Aug 30 22:13:22 2026 +0200
dev-util/hipcc: guard the device-libs bitcode sed
The two seds above it were guarded during the 10.0 migration; this one was
not. A silent miss leaves hipcc probing /usr/amdgcn/bitcode instead of
/usr/lib/amdgcn/bitcode, so every device compile fails to find the device
libs -- with the ebuild itself reporting success.
Checked for the double-prefix hazard that made the same pattern a real bug in
dev-libs/rocm-device-libs, where an unanchored global substitution also hit an
already-correct occurrence and produced lib/lib/amdgcn/bitcode. It does not
apply here: hipBin_amd.h has exactly one occurrence (:305) and no
already-prefixed spelling, so s///g is safe. Recorded that in the comment so
the next reader does not have to re-derive it.
src_prepare verified against the therock-10.0 source, and by effect rather
than exit status: one lib/amdgcn/bitcode afterwards, zero lib/lib/amdgcn.
commit 9130556401977404f7ea8bcb23671246e13fb051
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Aug 30 00:51:13 2026 +0200
dev-util/hipcc: add amd-llvm USE flag to select llvm-core/rocm-llvm
This is the single lever that switches the whole ROCm library stack onto AMD's
LLVM fork. rocm.eclass's rocm_use_clang() derives CC/CXX from
`hipconfig --hipclangpath`, so making hipconfig report /usr/lib/rocm-llvm/bin
redirects every library that calls it -- composable-kernel, hipBLASLt, rocBLAS,
rocPRIM, rocRAND, rccl, Tensile, roctracer -- with no per-ebuild change and no
CMAKE_CXX_COMPILER plumbing in nine separate places.
USE=-amd-llvm keeps the previous behaviour (the system clang at the
LLVM_COMPAT slot), so this is opt-in and the base/runtime layer is unaffected.
The path shape differs between the two: Gentoo's LLVM is slotted, so the
default branch inserts $ as a path component
(lib/llvm/<slot>/bin), while AMD's fork installs to a single unslotted
directory, so the amd-llvm branch swaps the component instead
(lib/rocm-llvm/bin).
commit 6d13ab842fcfb654a36f7a4b8688a132489ea1b9
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sat Aug 29 18:08:27 2026 +0200
dev-util/hipcc: build 10.0.0 against llvm:23
LLVM_COMPAT was 22, inherited from the 7.2.4 ebuild it was copied from.
That is correct in isolation -- hipcc builds fine against llvm:22 -- but it
is wrong for a ROCm 10.0 stack, and the failure mode is silent: src_prepare
seds lib/llvm/$/bin into the installed hipcc wrapper, so a 22 here
ships a hipcc that drives clang 22 while dev-libs/rocm-device-libs-10.0.0
requires clang 23 (amdhsa_abi.h and the struct_buffer_load_format builtins,
neither present in 22).
Because it builds green either way, no build-check would have caught this;
it has to be fixed by inspection as part of moving the cohort.
commit 7dbfc474db6b0a67330ea2cae155f278a2028fc4
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Fri Aug 28 08:36:36 2026 +0200
dev-util/hipcc: add 10.0.0
commit 7c384c8cac70105c06770e0ccb7af0fc295f2471
Author: Raukaan Cogbrother <cogbrother@raukaan.local>
Date: Sun May 31 03:29:35 2026 +0200
metadata: normalize tab → 2-space indent in 42 imported files
42 metadata.xml files inherited tab indentation from their origin
overlays (mostly the ROCm cluster, plus a handful of dev-python and
AMD-tooling imports). The overlay convention is 2-space; this brings
them in line with the other 498 files. Maintainer attributions
preserved verbatim.
commit 7efa2280df2a1bab0990f205d0b80a57645467dc
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Fri May 29 14:12:05 2026 +0200
dev-util/hipcc: add 7.2.4
commit 6fb7a0f3df6f86a1dc72f6c97632aea9d7b35d66
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Fri May 8 20:06:46 2026 +0200
dev-util/hipcc: drop stale patch superseded by 7.2.3
commit befa078db38faf2d8262505f90b99d7e49f6b304
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Wed May 6 01:54:25 2026 +0200
dev-util/hipcc: add 7.2.3
commit 8a7082f1ba65632398134de525d263330415ac96
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sat Dec 6 21:51:11 2025 +0100
*/*: drop rocm/hip as newer version is in main tree
Signed-off-by: Ivan S. Titov <iohann.s.titov@gmail.com>
commit cfa997ed467a25f0a98e30e69523013fb12b4136
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sat Oct 18 14:26:09 2025 +0200
added rocm and HIP v. 6.4.4 for now, as main tree lags

