azwolskiandClaude Sonnet 4.6
6e021c57e5
Replace all occurrences of the deprecated undef keyword with poison in ISPC's built-in IR library files (builtins/*.ll) and M4 macro files (builtins/util.m4, builtins/util-xe.m4, builtins/svml.m4). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
36 lines
1.7 KiB
LLVM
36 lines
1.7 KiB
LLVM
;; Copyright (c) 2020-2026, Intel Corporation
|
|
;;
|
|
;; SPDX-License-Identifier: BSD-3-Clause
|
|
|
|
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
|
|
;; half precision rcp and rsqrt using native AVX512FP16 instructions
|
|
;; they will be invoked only when hardware support is available.
|
|
|
|
declare <8 x half> @llvm.x86.avx512fp16.mask.rcp.sh(<8 x half>, <8 x half>, <8 x half>, i8)
|
|
define half @__rcp_uniform_half(half) nounwind readonly alwaysinline {
|
|
%vec = insertelement <8 x half> poison, half %0, i32 0
|
|
%rcp = tail call <8 x half> @llvm.x86.avx512fp16.mask.rcp.sh(<8 x half> %vec, <8 x half> %vec, <8 x half> poison, i8 -1)
|
|
%ret = extractelement <8 x half> %rcp, i32 0
|
|
ret half %ret
|
|
}
|
|
|
|
declare <32 x half> @llvm.x86.avx512fp16.mask.rcp.ph.512(<32 x half>, <32 x half>, i32)
|
|
define <32 x half> @__rcp_varying_half(<32 x half>) nounwind readonly alwaysinline {
|
|
%ret = tail call <32 x half> @llvm.x86.avx512fp16.mask.rcp.ph.512(<32 x half> %0, <32 x half> poison, i32 -1)
|
|
ret <32 x half> %ret
|
|
}
|
|
|
|
declare <8 x half> @llvm.x86.avx512fp16.mask.rsqrt.sh(<8 x half>, <8 x half>, <8 x half>, i8)
|
|
define half @__rsqrt_uniform_half(half) nounwind readnone alwaysinline {
|
|
%vec = insertelement <8 x half> poison, half %0, i32 0
|
|
%rsqrt = tail call <8 x half> @llvm.x86.avx512fp16.mask.rsqrt.sh(<8 x half> %vec, <8 x half> %vec, <8 x half> poison, i8 -1)
|
|
%ret = extractelement <8 x half> %rsqrt, i32 0
|
|
ret half %ret
|
|
}
|
|
|
|
declare <32 x half> @llvm.x86.avx512fp16.mask.rsqrt.ph.512(<32 x half>, <32 x half>, i32)
|
|
define <32 x half> @__rsqrt_varying_half(<32 x half>) nounwind readnone alwaysinline {
|
|
%ret = tail call <32 x half> @llvm.x86.avx512fp16.mask.rsqrt.ph.512(<32 x half> %0, <32 x half> poison, i32 -1)
|
|
ret <32 x half> %ret
|
|
}
|