ziglang/zig
261
6.0k
Fork
You've already forked zig
772

Fixes langref explanation of @mulAdd #35285

Closed
MasonRemaley wants to merge 1 commit from MasonRemaley/zig:muladd-langref into master
pull from: MasonRemaley/zig:muladd-langref
merge into: ziglang:master
ziglang:master
ziglang:elfv2
ziglang:spork8
ziglang:restricted
ziglang:0.16.x
ziglang:panic-rewrite
ziglang:parking-futex-lockfree
ziglang:windows-Io-cleanup
ziglang:Io-watch
ziglang:ParseCommandLineOptions
ziglang:windows-async-files
ziglang:poll-ring
ziglang:debug-file-leaks-differently
ziglang:debug-file-leaks
ziglang:ProcessPrng
ziglang:elfv2-dyn
ziglang:jobserver
ziglang:threadtheft
ziglang:io-threaded-no-queue
ziglang:0.15.x
ziglang:Io.net
ziglang:comptime-allocator
ziglang:restricted-function-pointers
ziglang:cli
ziglang:wasm-linker-writer
ziglang:wrangle-writer-buffering
ziglang:sha1-stream
ziglang:async-await-demo
ziglang:fixes
ziglang:0.14.x
ziglang:ast-node-methods
ziglang:macos-debug-info
ziglang:make-vs-configure
ziglang:fuzz-macos
ziglang:sans-aro
ziglang:ArrayList-reserve
ziglang:incr-bug
ziglang:llvm-ir-nosanitize-metadata
ziglang:ci-tarballs
ziglang:ci-scripts
ziglang:threadpool
ziglang:0.12.x
ziglang:new-pkg-hash
ziglang:json-diagnostics
ziglang:more-doctests
ziglang:rework-comptime-mutation
ziglang:0.11.x
ziglang:ci-perf-comment
ziglang:stage2-async
ziglang:0.10.x
ziglang:autofix
ziglang:0.9.x
ziglang:aro
ziglang:hcs
ziglang:0.8.x
ziglang:0.7.x

The current langref for @mulAdd says:

Fused multiply-add, similar to `(a * b) + c`, except only rounds once, and is thus more accurate. 

However, @mulAdd doesn't guarantee that the operations are fused, it gives the compiler the option to fuse them. For example, the LLVM backend emits llvm.fmuladd for @mulAdd which only fuses the operations when FMA is supported and more efficient.

IMO the currently implemented behavior is desirable. For example, I'd like my rendering and physics code that uses @mulAdd to compile for all targets, but get a performance boost if FMA is supported. I wouldn't, for example, want it to try to emulate mulAdd or fail to compile when unsupported.

This PR updates the langref to match the current behavior.

The current langref for `@mulAdd` says: ``` Fused multiply-add, similar to `(a * b) + c`, except only rounds once, and is thus more accurate. ``` However, `@mulAdd` doesn't guarantee that the operations are fused, it gives the compiler the option to fuse them. For example, the LLVM backend emits `llvm.fmuladd` for `@mulAdd` which [only fuses the operations when FMA is supported and more efficient](https://llvm.org/docs/LangRef.html#llvm-fmuladd-intrinsic). IMO the currently implemented behavior is desirable. For example, I'd like my rendering and physics code that uses `@mulAdd` to compile for all targets, but get a performance boost if FMA is supported. I wouldn't, for example, want it to try to emulate `mulAdd` or fail to compile when unsupported. This PR updates the langref to match the current behavior.
@ -4491,1 +4490,3 @@
only rounds once, and is thus more accurate.
Multiply-add, similar to {#syntax#}(a * b) + c{#endsyntax#} except that the operation can be
fused if this is supported and more efficient than the unfused variant on the target. Fused
multiply add only rounds once, and is thus more accurate.
Contributor
Copy link

"multiply add" -> "multiply-add"?

"multiply add" -> "multiply-add"?
MasonRemaley marked this conversation as resolved
MasonRemaley force-pushed muladd-langref from f351be4f60
Some checks failed
ci / riscv64-linux-debug (pull_request) Has been skipped
ci / riscv64-linux-release (pull_request) Has been skipped
ci / x86_64-netbsd-debug (pull_request) Successful in 40m2s
ci / x86_64-netbsd-release (pull_request) Successful in 42m30s
ci / x86_64-freebsd-debug (pull_request) Successful in 42m38s
ci / x86_64-freebsd-release (pull_request) Successful in 43m18s
ci / x86_64-windows-release (pull_request) Successful in 44m46s
ci / x86_64-openbsd-release (pull_request) Successful in 47m33s
ci / x86_64-windows-debug (pull_request) Successful in 50m34s
ci / x86_64-linux-debug (pull_request) Successful in 50m50s
ci / x86_64-openbsd-debug (pull_request) Successful in 55m57s
ci / aarch64-macos-release (pull_request) Successful in 58m9s
ci / aarch64-macos-debug (pull_request) Successful in 1h11m9s
ci / aarch64-linux-release (pull_request) Successful in 1h25m40s
ci / s390x-linux-release (pull_request) Successful in 1h33m59s
ci / loongarch64-linux-release (pull_request) Successful in 1h56m15s
ci / s390x-linux-debug (pull_request) Successful in 2h14m5s
ci / x86_64-linux-debug-llvm (pull_request) Successful in 2h17m27s
ci / x86_64-linux-release (pull_request) Successful in 2h18m46s
ci / powerpc64le-linux-release (pull_request) Successful in 2h24m53s
ci / aarch64-linux-debug (pull_request) Successful in 2h36m44s
ci / loongarch64-linux-debug (pull_request) Successful in 2h51m1s
ci / powerpc64le-linux-debug (pull_request) Successful in 3h23m57s
ci / aarch64-freebsd-debug (pull_request) Successful in 4h1m42s
ci / aarch64-freebsd-release (pull_request) Successful in 2h57m23s
ci / aarch64-netbsd-debug (pull_request) Has been cancelled
ci / aarch64-netbsd-release (pull_request) Has been cancelled
to c8c8a06625
All checks were successful
ci / riscv64-linux-debug (pull_request) Has been skipped
ci / riscv64-linux-release (pull_request) Has been skipped
ci / x86_64-netbsd-debug (pull_request) Successful in 40m15s
ci / x86_64-netbsd-release (pull_request) Successful in 41m14s
ci / x86_64-freebsd-release (pull_request) Successful in 43m54s
ci / x86_64-openbsd-release (pull_request) Successful in 44m59s
ci / x86_64-windows-debug (pull_request) Successful in 48m1s
ci / x86_64-windows-release (pull_request) Successful in 49m48s
ci / x86_64-freebsd-debug (pull_request) Successful in 52m38s
ci / x86_64-openbsd-debug (pull_request) Successful in 52m38s
ci / aarch64-macos-release (pull_request) Successful in 1h3m22s
ci / x86_64-linux-debug (pull_request) Successful in 1h8m2s
ci / aarch64-macos-debug (pull_request) Successful in 1h18m6s
ci / aarch64-linux-release (pull_request) Successful in 1h25m27s
ci / s390x-linux-release (pull_request) Successful in 1h42m11s
ci / loongarch64-linux-release (pull_request) Successful in 1h53m17s
ci / powerpc64le-linux-release (pull_request) Successful in 2h10m47s
ci / s390x-linux-debug (pull_request) Successful in 2h24m50s
ci / aarch64-linux-debug (pull_request) Successful in 2h25m47s
ci / x86_64-linux-release (pull_request) Successful in 2h46m35s
ci / x86_64-linux-debug-llvm (pull_request) Successful in 2h47m47s
ci / loongarch64-linux-debug (pull_request) Successful in 2h55m48s
ci / powerpc64le-linux-debug (pull_request) Successful in 4h56m30s
ci / aarch64-freebsd-release (pull_request) Successful in 2h39m45s
ci / aarch64-freebsd-debug (pull_request) Successful in 3h59m28s
ci / aarch64-netbsd-debug (pull_request) Successful in 4h35m59s
ci / aarch64-netbsd-release (pull_request) Successful in 3h29m16s
2026年05月13日 09:52:03 +02:00
Compare
Contributor
Copy link

I can't reproduce your described behavior. The LLVM backend emits not llvm.fmuladd.* intrinsics, but llvm.fma.* intrinsics. This seems to generate calls to the fmaf function of compiler-rt, if the target architecture does not support it natively.

Note that the current fmaf implementation of compiler-rt has a TODO, because it is inexact. See also #30659.

I can't reproduce your described behavior. The LLVM backend [emits](https://zig.godbolt.org/z/bjccnfWda) not [`llvm.fmuladd.*` intrinsics](https://llvm.org/docs/LangRef.html#llvm-fmuladd-intrinsic), but [`llvm.fma.*` intrinsics](https://llvm.org/docs/LangRef.html#llvm-fma-intrinsic). This seems to generate calls to the `fmaf` function of compiler-rt, if the target architecture does not support it natively. Note that the current `fmaf` implementation of compiler-rt has a [TODO](https://codeberg.org/ziglang/zig/src/commit/bdfbf432dd626b495cee0a8aa53e77b4ef718ff2/lib/compiler_rt/fma.zig#L40), because it is inexact. See also #30659.
Author
Owner
Copy link

Good catch, I didn't realize that both llvm.fmuladd.* and llvm.fma.* exist and read the docs on the incorrect one.

Good catch, I didn't realize that both [`llvm.fmuladd.*`](https://llvm.org/docs/LangRef.html#llvm-fmuladd-intrinsic) and [`llvm.fma.*`](https://llvm.org/docs/LangRef.html#llvm-fma-intrinsic) exist and read the docs on the incorrect one.
Author
Owner
Copy link

Related: #35312

Related: #35312
All checks were successful
ci / riscv64-linux-debug (pull_request) Has been skipped
ci / riscv64-linux-release (pull_request) Has been skipped
ci / x86_64-netbsd-debug (pull_request) Successful in 40m15s
Required
Details
ci / x86_64-netbsd-release (pull_request) Successful in 41m14s
Required
Details
ci / x86_64-freebsd-release (pull_request) Successful in 43m54s
Required
Details
ci / x86_64-openbsd-release (pull_request) Successful in 44m59s
Required
Details
ci / x86_64-windows-debug (pull_request) Successful in 48m1s
Required
Details
ci / x86_64-windows-release (pull_request) Successful in 49m48s
Required
Details
ci / x86_64-freebsd-debug (pull_request) Successful in 52m38s
Required
Details
ci / x86_64-openbsd-debug (pull_request) Successful in 52m38s
Required
Details
ci / aarch64-macos-release (pull_request) Successful in 1h3m22s
Required
Details
ci / x86_64-linux-debug (pull_request) Successful in 1h8m2s
Required
Details
ci / aarch64-macos-debug (pull_request) Successful in 1h18m6s
Required
Details
ci / aarch64-linux-release (pull_request) Successful in 1h25m27s
Required
Details
ci / s390x-linux-release (pull_request) Successful in 1h42m11s
Required
Details
ci / loongarch64-linux-release (pull_request) Successful in 1h53m17s
ci / powerpc64le-linux-release (pull_request) Successful in 2h10m47s
Required
Details
ci / s390x-linux-debug (pull_request) Successful in 2h24m50s
Required
Details
ci / aarch64-linux-debug (pull_request) Successful in 2h25m47s
Required
Details
ci / x86_64-linux-release (pull_request) Successful in 2h46m35s
Required
Details
ci / x86_64-linux-debug-llvm (pull_request) Successful in 2h47m47s
Required
Details
ci / loongarch64-linux-debug (pull_request) Successful in 2h55m48s
ci / powerpc64le-linux-debug (pull_request) Successful in 4h56m30s
Required
Details
ci / aarch64-freebsd-release (pull_request) Successful in 2h39m45s
ci / aarch64-freebsd-debug (pull_request) Successful in 3h59m28s
ci / aarch64-netbsd-debug (pull_request) Successful in 4h35m59s
ci / aarch64-netbsd-release (pull_request) Successful in 3h29m16s

Pull request closed

Please reopen this pull request to perform a merge.
Sign in to join this conversation.
No reviewers
Labels
Clear labels
abi/f32
abi/ilp32
abi/sf
accepted
This proposal is planned.
arch/21k
arch/6502
arch/aarch64
arch/alpha
arch/amdgcn
arch/arc
arch/arc32
arch/arc64
arch/arm
arch/avr
arch/bfin
arch/bpf
arch/colossus
arch/cris
arch/csky
arch/dlx
arch/epiphany
arch/fr30
arch/frv
arch/hexagon
arch/hppa
arch/hppa64
arch/ia64
arch/kalimba
arch/kvx
arch/lanai
arch/lm32
arch/loongarch32
arch/loongarch64
arch/m32r
arch/m68k
arch/m88k
arch/mcore
arch/microblaze
arch/mips
arch/mips64
arch/mmix
arch/moxie
arch/mrisc32
arch/msp430
arch/nds32
arch/ns32k
arch/nvptx
arch/or1k
arch/powerpc
arch/powerpc64
arch/propeller
arch/riscv32
arch/riscv64
arch/rl78
arch/rx
arch/s390x
arch/sh
arch/sparc
arch/sparc64
arch/spirv
arch/spu
arch/tricore
arch/v850
arch/vax
arch/vc4
arch/ve
arch/wasm
arch/x86
arch/x86_64
arch/xcore
arch/xtensa
autodoc
The web application for interactive documentation and generation of its assets.
backend/c
The C backend outputs C source code.
backend/llvm
The LLVM backend outputs an LLVM bitcode module.
backend/self-hosted
The self-hosted backends produce machine code directly.
binutils
Zig's included binary utilities: zig ar, zig dlltool, zig lib, zig ranlib, zig objcopy, and zig rc.
breaking
Implementing this issue could cause existing code to no longer compile or have different behavior.
build system
The Zig build system - zig build, std.Build, the build runner, and package management.
debug info
An issue related to debug information (e.g. DWARF) produced by the Zig compiler.
docs
An issue with documentation, e.g. the language reference or standard library doc comments.
error message
This issue points out an error message that is unhelpful and should be improved.
frontend
Tokenization, parsing, AstGen, ZonGen, Sema, Legalize, and Liveness.
fuzzing
An issue related to Zig's integrated fuzz testing.
incremental
Reuse of internal compiler state for faster compilation.
lib/c
This issue relates to Zig's libc implementation and/or vendored libcs.
lib/compiler-rt
This issue relates to Zig's compiler-rt library.
lib/cxx
This issue relates to Zig's vendored libc++ and/or libc++abi.
lib/std
This issue relates to Zig's standard library.
lib/tsan
This issue relates to Zig's vendored libtsan.
lib/ubsan-rt
This issue relates to Zig's ubsan-rt library.
lib/unwind
This issue relates to Zig's vendored libunwind.
linking
Zig's integrated object file and incremental linker.
miscompilation
The compiler reports success but produces semantically incorrect code.
os/android
os/contiki
os/dragonfly
os/driverkit
os/emscripten
os/freebsd
os/fuchsia
os/haiku
os/hermit
os/hurd
os/illumos
os/ios
os/linux
os/maccatalyst
os/macos
os/managarm
os/netbsd
os/ohos
os/openbsd
os/plan9
os/redox
os/rtems
os/serenity
os/tvos
os/uefi
os/visionos
os/wasi
os/watchos
os/windows
proposal
This issue suggests language modifications. If it also has the "accepted" label then it is planned.
release notes
This issue or pull request should be mentioned in the release notes.
testing
This issue is related to testing the compiler, standard library, or other parts of Zig.
zig cc
Zig as a drop-in C-family compiler.
zig fmt
The Zig source code formatter.
zig reduce
The Zig source code reduction tool.
bounty
https://ziglang.org/news/announcing-donor-bounties
bug
Observed behavior contradicts documented or intended behavior.
contributor-friendly
This issue is limited in scope and/or knowledge of project internals.
downstream
An issue with a third-party project that uses this project.
enhancement
Solving this issue will likely involve adding new logic or components to the codebase.
infra
An issue related to project infrastructure, e.g. continuous integration.
optimization
A task to improve performance and/or resource usage.
question
No questions on the issue tracker; use a community space instead.
regression
Something that used to work in a previous version stopped working
upstream
An issue with a third-party project that this project uses.
use case
Describes a real use case that is difficult or impossible, but does not propose a solution.
No labels
abi/f32
abi/ilp32
abi/sf
accepted
arch/21k
arch/6502
arch/aarch64
arch/alpha
arch/amdgcn
arch/arc
arch/arc32
arch/arc64
arch/arm
arch/avr
arch/bfin
arch/bpf
arch/colossus
arch/cris
arch/csky
arch/dlx
arch/epiphany
arch/fr30
arch/frv
arch/hexagon
arch/hppa
arch/hppa64
arch/ia64
arch/kalimba
arch/kvx
arch/lanai
arch/lm32
arch/loongarch32
arch/loongarch64
arch/m32r
arch/m68k
arch/m88k
arch/mcore
arch/microblaze
arch/mips
arch/mips64
arch/mmix
arch/moxie
arch/mrisc32
arch/msp430
arch/nds32
arch/ns32k
arch/nvptx
arch/or1k
arch/powerpc
arch/powerpc64
arch/propeller
arch/riscv32
arch/riscv64
arch/rl78
arch/rx
arch/s390x
arch/sh
arch/sparc
arch/sparc64
arch/spirv
arch/spu
arch/tricore
arch/v850
arch/vax
arch/vc4
arch/ve
arch/wasm
arch/x86
arch/x86_64
arch/xcore
arch/xtensa
autodoc
backend/c
backend/llvm
backend/self-hosted
binutils
breaking
build system
debug info
docs
error message
frontend
fuzzing
incremental
lib/c
lib/compiler-rt
lib/cxx
lib/std
lib/tsan
lib/ubsan-rt
lib/unwind
linking
miscompilation
os/android
os/contiki
os/dragonfly
os/driverkit
os/emscripten
os/freebsd
os/fuchsia
os/haiku
os/hermit
os/hurd
os/illumos
os/ios
os/linux
os/maccatalyst
os/macos
os/managarm
os/netbsd
os/ohos
os/openbsd
os/plan9
os/redox
os/rtems
os/serenity
os/tvos
os/uefi
os/visionos
os/wasi
os/watchos
os/windows
proposal
release notes
testing
zig cc
zig fmt
zig reduce
bounty
bug
contributor-friendly
downstream
enhancement
infra
optimization
question
regression
upstream
use case
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
3 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
ziglang/zig!35285
Reference in a new issue
ziglang/zig
No description provided.
Delete branch "MasonRemaley/zig:muladd-langref"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?