73cddef788
Fixes https://github.com/llvm/llvm-project/issues/169270 Changes the implementation of `is_finite` to emit fewer instructions, e.g. X86_64 ```asm old: # 18 bytes movd %xmm0, %eax andl $2147483647, %eax cmpl $2139095040, %eax setl %al retq new: # 15 bytes movd %xmm0, %eax addl %eax, %eax cmpl $-16777216, %eax setb %al retq ``` Aarch64 ```asm old: fmov w9, s0 mov w8, #2139095040 and w9, w9, #0x7fffffff cmp w9, w8 cset w0, lt ret new: fmov w8, s0 ubfx w8, w8, #23, #8 cmp w8, #255 cset w0, lo ret ``` See the issue for more information.
105 lines
3.5 KiB
LLVM
105 lines
3.5 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
|
|
; RUN: llc -mtriple=aarch64 -verify-machineinstrs -global-isel=0 %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD
|
|
; RUN: llc -mtriple=aarch64 -verify-machineinstrs -mattr=-fp-armv8 %s -o - | FileCheck %s --check-prefix=CHECK-NOFP
|
|
|
|
define i1 @isfinite_bf16(bfloat %x) {
|
|
; CHECK-LABEL: isfinite_bf16:
|
|
; CHECK: // %bb.0: // %entry
|
|
; CHECK-NEXT: // kill: def $h0 killed $h0 def $s0
|
|
; CHECK-NEXT: fmov w8, s0
|
|
; CHECK-NEXT: ubfx w8, w8, #7, #8
|
|
; CHECK-NEXT: cmp w8, #255
|
|
; CHECK-NEXT: cset w0, lo
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; CHECK-NOFP-LABEL: isfinite_bf16:
|
|
; CHECK-NOFP: // %bb.0: // %entry
|
|
; CHECK-NOFP-NEXT: ubfx w8, w0, #7, #8
|
|
; CHECK-NOFP-NEXT: cmp w8, #255
|
|
; CHECK-NOFP-NEXT: cset w0, lo
|
|
; CHECK-NOFP-NEXT: ret
|
|
entry:
|
|
%0 = tail call i1 @llvm.is.fpclass.bf16(bfloat %x, i32 504) ; 0x1f8 = "finite"
|
|
ret i1 %0
|
|
}
|
|
|
|
define i1 @not_isfinite_bf16(bfloat %x) {
|
|
; CHECK-LABEL: not_isfinite_bf16:
|
|
; CHECK: // %bb.0: // %entry
|
|
; CHECK-NEXT: // kill: def $h0 killed $h0 def $s0
|
|
; CHECK-NEXT: fmov w8, s0
|
|
; CHECK-NEXT: ubfx w8, w8, #7, #8
|
|
; CHECK-NEXT: cmp w8, #254
|
|
; CHECK-NEXT: cset w0, hi
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; CHECK-NOFP-LABEL: not_isfinite_bf16:
|
|
; CHECK-NOFP: // %bb.0: // %entry
|
|
; CHECK-NOFP-NEXT: ubfx w8, w0, #7, #8
|
|
; CHECK-NOFP-NEXT: cmp w8, #254
|
|
; CHECK-NOFP-NEXT: cset w0, hi
|
|
; CHECK-NOFP-NEXT: ret
|
|
entry:
|
|
%0 = tail call i1 @llvm.is.fpclass.bf16(bfloat %x, i32 519) ; ~0x1f8 = "~finite"
|
|
ret i1 %0
|
|
}
|
|
|
|
|
|
define <4 x i1> @isfinite_v4bf16(<4 x bfloat> %x) {
|
|
; CHECK-LABEL: isfinite_v4bf16:
|
|
; CHECK: // %bb.0: // %entry
|
|
; CHECK-NEXT: movi d1, #0xff00ff00ff00ff00
|
|
; CHECK-NEXT: add v0.4h, v0.4h, v0.4h
|
|
; CHECK-NEXT: cmhi v0.4h, v1.4h, v0.4h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; CHECK-NOFP-LABEL: isfinite_v4bf16:
|
|
; CHECK-NOFP: // %bb.0: // %entry
|
|
; CHECK-NOFP-NEXT: ubfx w8, w0, #7, #8
|
|
; CHECK-NOFP-NEXT: ubfx w9, w1, #7, #8
|
|
; CHECK-NOFP-NEXT: cmp w8, #255
|
|
; CHECK-NOFP-NEXT: ubfx w8, w2, #7, #8
|
|
; CHECK-NOFP-NEXT: cset w0, lo
|
|
; CHECK-NOFP-NEXT: cmp w9, #255
|
|
; CHECK-NOFP-NEXT: ubfx w9, w3, #7, #8
|
|
; CHECK-NOFP-NEXT: cset w1, lo
|
|
; CHECK-NOFP-NEXT: cmp w8, #255
|
|
; CHECK-NOFP-NEXT: cset w2, lo
|
|
; CHECK-NOFP-NEXT: cmp w9, #255
|
|
; CHECK-NOFP-NEXT: cset w3, lo
|
|
; CHECK-NOFP-NEXT: ret
|
|
entry:
|
|
%0 = tail call <4 x i1> @llvm.is.fpclass.v4bf16(<4 x bfloat> %x, i32 504) ; 0x1f8 = "finite"
|
|
ret <4 x i1> %0
|
|
}
|
|
|
|
define <4 x i1> @not_isfinite_v4bf16(<4 x bfloat> %x) {
|
|
; CHECK-LABEL: not_isfinite_v4bf16:
|
|
; CHECK: // %bb.0: // %entry
|
|
; CHECK-NEXT: movi d1, #0xff00ff00ff00ff00
|
|
; CHECK-NEXT: add v0.4h, v0.4h, v0.4h
|
|
; CHECK-NEXT: cmhs v0.4h, v0.4h, v1.4h
|
|
; CHECK-NEXT: ret
|
|
;
|
|
; CHECK-NOFP-LABEL: not_isfinite_v4bf16:
|
|
; CHECK-NOFP: // %bb.0: // %entry
|
|
; CHECK-NOFP-NEXT: ubfx w8, w0, #7, #8
|
|
; CHECK-NOFP-NEXT: ubfx w9, w1, #7, #8
|
|
; CHECK-NOFP-NEXT: cmp w8, #254
|
|
; CHECK-NOFP-NEXT: ubfx w8, w2, #7, #8
|
|
; CHECK-NOFP-NEXT: cset w0, hi
|
|
; CHECK-NOFP-NEXT: cmp w9, #254
|
|
; CHECK-NOFP-NEXT: ubfx w9, w3, #7, #8
|
|
; CHECK-NOFP-NEXT: cset w1, hi
|
|
; CHECK-NOFP-NEXT: cmp w8, #254
|
|
; CHECK-NOFP-NEXT: cset w2, hi
|
|
; CHECK-NOFP-NEXT: cmp w9, #254
|
|
; CHECK-NOFP-NEXT: cset w3, hi
|
|
; CHECK-NOFP-NEXT: ret
|
|
entry:
|
|
%0 = tail call <4 x i1> @llvm.is.fpclass.v4bf16(<4 x bfloat> %x, i32 519) ; ~0x1f8 = "~finite"
|
|
ret <4 x i1> %0
|
|
}
|
|
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
|
|
; CHECK-SD: {{.*}}
|