Skip to content
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

arch/risc-v: Refactor LLVM CPU type handling in Toolchain.cmake #15475

Merged
merged 1 commit into from
Jan 9, 2025
Merged

arch/risc-v: Refactor LLVM CPU type handling in Toolchain.cmake #15475

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
22 changes: 10 additions & 12 deletions arch/risc-v/src/cmake/Toolchain.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -343,21 +343,19 @@ if(CONFIG_RISCV_TOOLCHAIN STREQUAL GNU_RVG)
# These models can't cover all implementation of RISCV, but it's enough for
# most cases.

set(LLVM_CPUFLAGS)

if(CONFIG_ARCH_RV32)
if(${ARCHCPUEXTFLAGS} STREQUAL imc)
list(APPEND LLVM_CPUFLAGS -mcpu=sifive-e20)
elseif(${ARCHCPUEXTFLAGS} STREQUAL imac)
list(APPEND LLVM_CPUFLAGS -mcpu=sifive-e31)
elseif(${ARCHCPUEXTFLAGS} STREQUAL imafc)
list(APPEND LLVM_CPUFLAGS -mcpu=sifive-e76)
if(${ARCHCPUEXTFLAGS} MATCHES "^imc")
set(LLVM_CPUTYPE "sifive-e20")
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

but why use a specific vendor chip sifive-e20? @no1wudi

Copy link
Contributor Author

@no1wudi no1wudi Jan 10, 2025 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

For the RISC-V platform, the CPU_TYPE here is typically not utilized (pass to compiler); it is merely used to denote a combination of ISAs. For instance, sifive-e20 represents IMC. This is because LLVM's ARCH_TYPE only specifies RISCV32/RISCV64 and does not reflect the specific ISA. In practical application scenarios, such as in the Rust compiler, sifive-e20 is decoded as riscv32imc and then passed to the Rust compiler.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

but why not pass the general riscv32imc directly?

Copy link
Contributor Author

@no1wudi no1wudi Jan 10, 2025 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Because you can still pass it to compiler as -mcpu param in some case, it can works although it may not be the optimal choice, it will not cause any issues. But riscv32imc can not, it's not a valid identifier in LLVM backend:

nuttx/tools/Zig.defs

Lines 42 to 46 in 724797e

# Convert cortex-xxx/sifive-exx to cortex_xxx/sifive_exx
ifneq ($(LLVM_CPUTYPE),)
ZIGFLAGS += -mcpu $(subst -,_,$(LLVM_CPUTYPE))
endif

For MCUs, symbols like sifive-e20 are merely used to indicate the ISA types supported by the RISC-V platform and do not entail any substantive optimizations for the CPU core itself.

For example, certain compilers using LLVM, such as wamrc, typically employ --cpu=generic-rv32 in conjunction with the --cpu-features parameter +m,+c to enable support for riscv32imc. In practice, however, this is equivalent to specifying --cpu=sifive-e20.

elseif(${ARCHCPUEXTFLAGS} MATCHES "^imac")
set(LLVM_CPUTYPE "sifive-e31")
elseif(${ARCHCPUEXTFLAGS} MATCHES "^imafc")
set(LLVM_CPUTYPE "sifive-e76")
endif()
else()
if(${ARCHCPUEXTFLAGS} STREQUAL imac)
list(APPEND LLVM_CPUFLAGS -mcpu=sifive-s51)
elseif(${ARCHCPUEXTFLAGS} STREQUAL imafdc)
list(APPEND LLVM_CPUFLAGS -mcpu=sifive-u54)
if(${ARCHCPUEXTFLAGS} MATCHES "^imac")
set(LLVM_CPUTYPE "sifive-s51")
elseif(${ARCHCPUEXTFLAGS} MATCHES "^imafdc")
set(LLVM_CPUTYPE "sifive-u54")
endif()
endif()

Expand Down
Loading