Skip to content

Split gotoblas dispatch table by group / operation (2/2) #6088

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鈥檒l occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
topolarity wants to merge 7 commits into
base: develop
Choose a base branch
from
Open

Split gotoblas dispatch table by group / operation (2/2) #6088

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
b52307c
gemm: name the base of the small-matrix kernel offsets
topolarity Oct 1, 2026
1524c8f
build: generate dyn_cores.h from DYNAMIC_CORE
topolarity Oct 1, 2026
5e55689
DYNAMIC_ARCH: number the cores, and add per-group dispatch macros
topolarity Oct 1, 2026
f78159e
DYNAMIC_ARCH: build with -ffunction-sections -fdata-sections
topolarity Oct 1, 2026
297da0b
setparam-ref.c: use designated initializers for gotoblas_t
topolarity Oct 1, 2026
2f8f611
DYNAMIC_ARCH: split the kernels out of gotoblas_t into per-group tables
topolarity Oct 1, 2026
37da176
Remove FUNC_OFFSET and FUNC_BASE
topolarity Oct 1, 2026
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
1 change: 1 addition & 0 deletions .gitignore
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -120,3 +120,4 @@ CMakeCache.txt
CMakeFiles/*
.vscode
**/__pycache__
dyn_cores.h
2 changes: 1 addition & 1 deletion Makefile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -449,7 +449,7 @@ clean ::
ifeq ($(OSNAME), Darwin)
@rm -rf getarch.dSYM getarch_2nd.dSYM
endif
@rm -f Makefile.conf config.h Makefile_kernel.conf config_kernel.h st* *.dylib
@rm -f Makefile.conf config.h Makefile_kernel.conf config_kernel.h dyn_cores.h st* *.dylib
@rm -f cblas.tmp cblas.tmp2
@touch $(NETLIB_LAPACK_DIR)/make.inc
@$(MAKE) -C $(NETLIB_LAPACK_DIR) clean
Expand Down
12 changes: 12 additions & 0 deletions Makefile.system
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -867,6 +867,15 @@ override DYNAMIC_ARCH=
endif
endif

# dyn_cores.h tells common_param.h which cores this DYNAMIC_ARCH build has.
ifeq ($(DYNAMIC_ARCH), 1)
ifndef OPENBLAS_DYN_CORES
export OPENBLAS_DYN_CORES := $(sort $(DYNAMIC_CORE))
HASH := \#
DUMMY := $(shell echo '$(HASH)define OPENBLAS_CORE_LIST(X, arg) $(foreach c,$(OPENBLAS_DYN_CORES),X($(c), arg))' > $(TOPDIR)/dyn_cores.h)
endif
endif

ifeq ($(ARCH), ia64)
NO_BINARY_MODE = 1
BINARY_DEFINED = 1
Expand Down Expand Up @@ -1484,6 +1493,9 @@ endif

ifeq ($(DYNAMIC_ARCH), 1)
CCOMMON_OPT += -DDYNAMIC_ARCH
ifeq ($(C_COMPILER), $(filter $(C_COMPILER),GCC LSB CLANG INTEL))
CCOMMON_OPT += -ffunction-sections -fdata-sections
endif
endif

ifeq ($(DYNAMIC_OLDER), 1)
Expand Down
17 changes: 17 additions & 0 deletions cmake/arch.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -129,9 +129,26 @@ if (DYNAMIC_ARCH)
message (FATAL_ERROR "Your build directory contains a file config_kernel.h, probably from a previous compilation with make. This will conflict with the cmake compilation and cause strange compiler errors - please remove the file before trying again")
endif ()

if (EXISTS ${PROJECT_SOURCE_DIR}/dyn_cores.h)
message (FATAL_ERROR "Your source directory contains a file dyn_cores.h, probably from a previous compilation with make. It would be used instead of the one generated for this build - please remove the file before trying again")
endif ()

if (NOT DYNAMIC_CORE)
message (STATUS "DYNAMIC_ARCH is not supported on this architecture, removing from options")
unset(DYNAMIC_ARCH CACHE)
else ()
# dyn_cores.h tells common_param.h which cores this build has;
# see the equivalent rule in Makefile.system.
set(DYN_CORES_LIST ${DYNAMIC_CORE})
list(REMOVE_DUPLICATES DYN_CORES_LIST)
list(SORT DYN_CORES_LIST)
set(DYN_CORES_H "#define OPENBLAS_CORE_LIST(X, arg)")
foreach (DCORE ${DYN_CORES_LIST})
string(APPEND DYN_CORES_H " X(${DCORE}, arg)")
endforeach ()
file(WRITE ${PROJECT_BINARY_DIR}/dyn_cores.h.tmp "${DYN_CORES_H}\n")
configure_file(${PROJECT_BINARY_DIR}/dyn_cores.h.tmp ${PROJECT_BINARY_DIR}/dyn_cores.h COPYONLY)
file(REMOVE ${PROJECT_BINARY_DIR}/dyn_cores.h.tmp)
endif ()
endif ()

Expand Down
3 changes: 3 additions & 0 deletions cmake/system.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -536,6 +536,9 @@ endif ()
if (DYNAMIC_ARCH)
if (X86 OR X86_64 OR ARM64 OR POWER OR RISCV64 OR LOONGARCH64 OR ZARCH)
set(CCOMMON_OPT "${CCOMMON_OPT} -DDYNAMIC_ARCH")
if (CMAKE_C_COMPILER_ID MATCHES "GNU|Clang|Intel" AND NOT MSVC)
set(CCOMMON_OPT "${CCOMMON_OPT} -ffunction-sections -fdata-sections")
endif ()
if (DYNAMIC_OLDER)
set(CCOMMON_OPT "${CCOMMON_OPT} -DDYNAMIC_OLDER")
endif ()
Expand Down
18 changes: 9 additions & 9 deletions common_b.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -50,17 +50,17 @@
#define BGEMM_KERNEL bgemm_kernel

#else
#define BGEMV_N_K gotoblas->bgemv_n
#define BGEMV_T_K gotoblas->bgemv_t
#define BGEMV_N_K OPENBLAS_DISPATCH(bgemv)->bgemv_n
#define BGEMV_T_K OPENBLAS_DISPATCH(bgemv)->bgemv_t

#define BSCAL_K gotoblas->bscal_k
#define BSCAL_K OPENBLAS_DISPATCH(bscal)->bscal_k

#define BGEMM_ONCOPY gotoblas->bgemm_oncopy
#define BGEMM_OTCOPY gotoblas->bgemm_otcopy
#define BGEMM_INCOPY gotoblas->bgemm_incopy
#define BGEMM_ITCOPY gotoblas->bgemm_itcopy
#define BGEMM_BETA gotoblas->bgemm_beta
#define BGEMM_KERNEL gotoblas->bgemm_kernel
#define BGEMM_ONCOPY OPENBLAS_DISPATCH(bgemm)->bgemm_oncopy
#define BGEMM_OTCOPY OPENBLAS_DISPATCH(bgemm)->bgemm_otcopy
#define BGEMM_INCOPY OPENBLAS_DISPATCH(bgemm)->bgemm_incopy
#define BGEMM_ITCOPY OPENBLAS_DISPATCH(bgemm)->bgemm_itcopy
#define BGEMM_BETA OPENBLAS_DISPATCH(bgemm)->bgemm_beta
#define BGEMM_KERNEL OPENBLAS_DISPATCH(bgemm)->bgemm_kernel

#endif

Expand Down
Loading
Loading