Ricardo Jesus b563e76306
[AArch64] Improve scalar and Neon popcount with SVE CNT. (#143870)
When available, we can use SVE's CNT instruction to improve the lowering
of scalar and fixed-length popcount (CTPOP) since the SVE instruction
supports types that the Neon variant doesn't.

For the scalar types, I see the following speedups on NVIDIA Grace CPU:
| size (bits) | before (Gibit/s) | after (Gibit/s) | speedup |
|------------:|-----------------:|----------------:|--------:|
|         32  |           75.20  |          86.79  |   1.15  |
|         64  |          149.87  |         173.70  |   1.16  |
|        128  |          158.56  |         164.88  |   1.04  |
2025-06-30 09:04:45 +01:00

907 lines
28 KiB
LLVM

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -O0 -mtriple=aarch64-unknown-unknown | FileCheck %s --check-prefix=CHECKO0
; RUN: llc < %s -mtriple=aarch64-unknown-unknown -mattr=+neon | FileCheck %s --check-prefixes=CHECK,NEON
; RUN: llc < %s -mtriple=aarch64-unknown-unknown -mattr=+neon,+dotprod | FileCheck %s --check-prefixes=CHECK,DOT
; RUN: llc < %s -mtriple=aarch64-unknown-unknown -mattr=+sve | FileCheck %s --check-prefixes=CHECK,SVE
; RUN: llc < %s -mtriple=aarch64_be-unknown-unknown | FileCheck %s --check-prefixes=BE
; RUN: llc < %s -global-isel -mtriple=aarch64-unknown-unknown | FileCheck %s --check-prefix=GISEL
; RUN: llc < %s -O0 -global-isel -mtriple=aarch64-unknown-unknown | FileCheck %s --check-prefix=GISELO0
; RUN: llc < %s -global-isel -mtriple=aarch64-unknown-unknown -mattr=+neon | FileCheck %s --check-prefixes=GISEL,NEON-GISEL
; RUN: llc < %s -global-isel -mtriple=aarch64-unknown-unknown -mattr=+neon,+dotprod | FileCheck %s --check-prefixes=GISEL,DOT-GISEL
; RUN: llc < %s -global-isel -mtriple=aarch64-unknown-unknown -mattr=+sve | FileCheck %s --check-prefixes=GISEL,SVE-GISEL
; Function Attrs: nobuiltin nounwind readonly
define i8 @popcount128(ptr nocapture nonnull readonly %0) {
; CHECKO0-LABEL: popcount128:
; CHECKO0: // %bb.0: // %Entry
; CHECKO0-NEXT: ldr q0, [x0]
; CHECKO0-NEXT: cnt v0.16b, v0.16b
; CHECKO0-NEXT: uaddlv h0, v0.16b
; CHECKO0-NEXT: // kill: def $q0 killed $h0
; CHECKO0-NEXT: // kill: def $s0 killed $s0 killed $q0
; CHECKO0-NEXT: fmov w0, s0
; CHECKO0-NEXT: ret
;
; NEON-LABEL: popcount128:
; NEON: // %bb.0: // %Entry
; NEON-NEXT: ldr d0, [x0]
; NEON-NEXT: add x8, x0, #8
; NEON-NEXT: ld1 { v0.d }[1], [x8]
; NEON-NEXT: cnt v0.16b, v0.16b
; NEON-NEXT: addv b0, v0.16b
; NEON-NEXT: fmov w0, s0
; NEON-NEXT: ret
;
; DOT-LABEL: popcount128:
; DOT: // %bb.0: // %Entry
; DOT-NEXT: ldr d0, [x0]
; DOT-NEXT: add x8, x0, #8
; DOT-NEXT: ld1 { v0.d }[1], [x8]
; DOT-NEXT: cnt v0.16b, v0.16b
; DOT-NEXT: addv b0, v0.16b
; DOT-NEXT: fmov w0, s0
; DOT-NEXT: ret
;
; SVE-LABEL: popcount128:
; SVE: // %bb.0: // %Entry
; SVE-NEXT: ldr d0, [x0]
; SVE-NEXT: add x8, x0, #8
; SVE-NEXT: ptrue p0.d
; SVE-NEXT: ld1 { v0.d }[1], [x8]
; SVE-NEXT: cnt z0.d, p0/m, z0.d
; SVE-NEXT: addp d0, v0.2d
; SVE-NEXT: fmov w0, s0
; SVE-NEXT: ret
;
; BE-LABEL: popcount128:
; BE: // %bb.0: // %Entry
; BE-NEXT: ldr d0, [x0]
; BE-NEXT: add x8, x0, #8
; BE-NEXT: ld1 { v0.d }[1], [x8]
; BE-NEXT: rev64 v0.16b, v0.16b
; BE-NEXT: cnt v0.16b, v0.16b
; BE-NEXT: addv b0, v0.16b
; BE-NEXT: rev64 v0.4s, v0.4s
; BE-NEXT: mov w0, v0.s[1]
; BE-NEXT: ret
;
; GISEL-LABEL: popcount128:
; GISEL: // %bb.0: // %Entry
; GISEL-NEXT: ldr q0, [x0]
; GISEL-NEXT: cnt v0.16b, v0.16b
; GISEL-NEXT: uaddlv h0, v0.16b
; GISEL-NEXT: fmov w0, s0
; GISEL-NEXT: ret
;
; GISELO0-LABEL: popcount128:
; GISELO0: // %bb.0: // %Entry
; GISELO0-NEXT: ldr q0, [x0]
; GISELO0-NEXT: cnt v0.16b, v0.16b
; GISELO0-NEXT: uaddlv h0, v0.16b
; GISELO0-NEXT: // kill: def $q0 killed $h0
; GISELO0-NEXT: // kill: def $s0 killed $s0 killed $q0
; GISELO0-NEXT: fmov w0, s0
; GISELO0-NEXT: ret
Entry:
%1 = load i128, ptr %0, align 16
%2 = tail call i128 @llvm.ctpop.i128(i128 %1)
%3 = trunc i128 %2 to i8
ret i8 %3
}
; Function Attrs: nounwind readnone speculatable willreturn
declare i128 @llvm.ctpop.i128(i128)
; Function Attrs: nobuiltin nounwind readonly
define i16 @popcount256(ptr nocapture nonnull readonly %0) {
; CHECKO0-LABEL: popcount256:
; CHECKO0: // %bb.0: // %Entry
; CHECKO0-NEXT: ldr x11, [x0]
; CHECKO0-NEXT: ldr x10, [x0, #8]
; CHECKO0-NEXT: ldr x9, [x0, #16]
; CHECKO0-NEXT: ldr x8, [x0, #24]
; CHECKO0-NEXT: // implicit-def: $q1
; CHECKO0-NEXT: mov v1.d[0], x11
; CHECKO0-NEXT: mov v1.d[1], x10
; CHECKO0-NEXT: // implicit-def: $q0
; CHECKO0-NEXT: mov v0.d[0], x9
; CHECKO0-NEXT: mov v0.d[1], x8
; CHECKO0-NEXT: cnt v1.16b, v1.16b
; CHECKO0-NEXT: uaddlv h1, v1.16b
; CHECKO0-NEXT: // kill: def $q1 killed $h1
; CHECKO0-NEXT: // kill: def $s1 killed $s1 killed $q1
; CHECKO0-NEXT: fmov w0, s1
; CHECKO0-NEXT: mov w10, wzr
; CHECKO0-NEXT: mov w9, w0
; CHECKO0-NEXT: mov w8, w10
; CHECKO0-NEXT: bfi x9, x8, #32, #32
; CHECKO0-NEXT: cnt v0.16b, v0.16b
; CHECKO0-NEXT: uaddlv h0, v0.16b
; CHECKO0-NEXT: // kill: def $q0 killed $h0
; CHECKO0-NEXT: // kill: def $s0 killed $s0 killed $q0
; CHECKO0-NEXT: fmov w0, s0
; CHECKO0-NEXT: mov w8, w0
; CHECKO0-NEXT: // kill: def $x10 killed $w10
; CHECKO0-NEXT: bfi x8, x10, #32, #32
; CHECKO0-NEXT: adds x8, x8, x9
; CHECKO0-NEXT: mov w0, w8
; CHECKO0-NEXT: ret
;
; NEON-LABEL: popcount256:
; NEON: // %bb.0: // %Entry
; NEON-NEXT: ldr d0, [x0, #16]
; NEON-NEXT: ldr d1, [x0]
; NEON-NEXT: add x8, x0, #8
; NEON-NEXT: add x9, x0, #24
; NEON-NEXT: ld1 { v0.d }[1], [x9]
; NEON-NEXT: ld1 { v1.d }[1], [x8]
; NEON-NEXT: cnt v0.16b, v0.16b
; NEON-NEXT: cnt v1.16b, v1.16b
; NEON-NEXT: addv b0, v0.16b
; NEON-NEXT: addv b1, v1.16b
; NEON-NEXT: fmov w8, s0
; NEON-NEXT: fmov w9, s1
; NEON-NEXT: add w0, w9, w8
; NEON-NEXT: ret
;
; DOT-LABEL: popcount256:
; DOT: // %bb.0: // %Entry
; DOT-NEXT: ldr d0, [x0, #16]
; DOT-NEXT: ldr d1, [x0]
; DOT-NEXT: add x8, x0, #8
; DOT-NEXT: add x9, x0, #24
; DOT-NEXT: ld1 { v0.d }[1], [x9]
; DOT-NEXT: ld1 { v1.d }[1], [x8]
; DOT-NEXT: cnt v0.16b, v0.16b
; DOT-NEXT: cnt v1.16b, v1.16b
; DOT-NEXT: addv b0, v0.16b
; DOT-NEXT: addv b1, v1.16b
; DOT-NEXT: fmov w8, s0
; DOT-NEXT: fmov w9, s1
; DOT-NEXT: add w0, w9, w8
; DOT-NEXT: ret
;
; SVE-LABEL: popcount256:
; SVE: // %bb.0: // %Entry
; SVE-NEXT: ldr d0, [x0, #16]
; SVE-NEXT: ldr d1, [x0]
; SVE-NEXT: add x8, x0, #8
; SVE-NEXT: add x9, x0, #24
; SVE-NEXT: ptrue p0.d
; SVE-NEXT: ld1 { v0.d }[1], [x9]
; SVE-NEXT: ld1 { v1.d }[1], [x8]
; SVE-NEXT: cnt z0.d, p0/m, z0.d
; SVE-NEXT: cnt z1.d, p0/m, z1.d
; SVE-NEXT: add v0.2d, v1.2d, v0.2d
; SVE-NEXT: addp d0, v0.2d
; SVE-NEXT: fmov w0, s0
; SVE-NEXT: ret
;
; BE-LABEL: popcount256:
; BE: // %bb.0: // %Entry
; BE-NEXT: ldr d0, [x0]
; BE-NEXT: ldr d1, [x0, #16]
; BE-NEXT: add x8, x0, #24
; BE-NEXT: add x9, x0, #8
; BE-NEXT: ld1 { v0.d }[1], [x9]
; BE-NEXT: ld1 { v1.d }[1], [x8]
; BE-NEXT: rev64 v0.16b, v0.16b
; BE-NEXT: rev64 v1.16b, v1.16b
; BE-NEXT: cnt v0.16b, v0.16b
; BE-NEXT: cnt v1.16b, v1.16b
; BE-NEXT: addv b0, v0.16b
; BE-NEXT: addv b1, v1.16b
; BE-NEXT: rev64 v0.4s, v0.4s
; BE-NEXT: rev64 v1.4s, v1.4s
; BE-NEXT: mov w8, v0.s[1]
; BE-NEXT: mov w9, v1.s[1]
; BE-NEXT: add w0, w9, w8
; BE-NEXT: ret
;
; GISEL-LABEL: popcount256:
; GISEL: // %bb.0: // %Entry
; GISEL-NEXT: ldp x8, x9, [x0]
; GISEL-NEXT: mov v0.d[0], x8
; GISEL-NEXT: ldp x8, x10, [x0, #16]
; GISEL-NEXT: mov v1.d[0], x8
; GISEL-NEXT: mov v0.d[1], x9
; GISEL-NEXT: mov v1.d[1], x10
; GISEL-NEXT: cnt v0.16b, v0.16b
; GISEL-NEXT: cnt v1.16b, v1.16b
; GISEL-NEXT: uaddlv h0, v0.16b
; GISEL-NEXT: uaddlv h1, v1.16b
; GISEL-NEXT: mov w8, v0.s[0]
; GISEL-NEXT: fmov w9, s1
; GISEL-NEXT: add x0, x8, w9, uxtw
; GISEL-NEXT: // kill: def $w0 killed $w0 killed $x0
; GISEL-NEXT: ret
;
; GISELO0-LABEL: popcount256:
; GISELO0: // %bb.0: // %Entry
; GISELO0-NEXT: ldr x11, [x0]
; GISELO0-NEXT: ldr x10, [x0, #8]
; GISELO0-NEXT: ldr x9, [x0, #16]
; GISELO0-NEXT: ldr x8, [x0, #24]
; GISELO0-NEXT: // implicit-def: $q1
; GISELO0-NEXT: mov v1.d[0], x11
; GISELO0-NEXT: mov v1.d[1], x10
; GISELO0-NEXT: // implicit-def: $q0
; GISELO0-NEXT: mov v0.d[0], x9
; GISELO0-NEXT: mov v0.d[1], x8
; GISELO0-NEXT: cnt v1.16b, v1.16b
; GISELO0-NEXT: uaddlv h1, v1.16b
; GISELO0-NEXT: // kill: def $q1 killed $h1
; GISELO0-NEXT: // kill: def $s1 killed $s1 killed $q1
; GISELO0-NEXT: fmov w0, s1
; GISELO0-NEXT: mov w10, wzr
; GISELO0-NEXT: mov w9, w0
; GISELO0-NEXT: mov w8, w10
; GISELO0-NEXT: bfi x9, x8, #32, #32
; GISELO0-NEXT: cnt v0.16b, v0.16b
; GISELO0-NEXT: uaddlv h0, v0.16b
; GISELO0-NEXT: // kill: def $q0 killed $h0
; GISELO0-NEXT: // kill: def $s0 killed $s0 killed $q0
; GISELO0-NEXT: fmov w0, s0
; GISELO0-NEXT: mov w8, w0
; GISELO0-NEXT: // kill: def $x10 killed $w10
; GISELO0-NEXT: bfi x8, x10, #32, #32
; GISELO0-NEXT: adds x8, x8, x9
; GISELO0-NEXT: mov w0, w8
; GISELO0-NEXT: ret
Entry:
%1 = load i256, ptr %0, align 16
%2 = tail call i256 @llvm.ctpop.i256(i256 %1)
%3 = trunc i256 %2 to i16
ret i16 %3
}
; Function Attrs: nounwind readnone speculatable willreturn
declare i256 @llvm.ctpop.i256(i256)
define <1 x i128> @popcount1x128(<1 x i128> %0) {
; CHECKO0-LABEL: popcount1x128:
; CHECKO0: // %bb.0: // %Entry
; CHECKO0-NEXT: // implicit-def: $q0
; CHECKO0-NEXT: mov v0.d[0], x0
; CHECKO0-NEXT: mov v0.d[1], x1
; CHECKO0-NEXT: cnt v0.16b, v0.16b
; CHECKO0-NEXT: uaddlv h0, v0.16b
; CHECKO0-NEXT: // kill: def $q0 killed $h0
; CHECKO0-NEXT: mov x1, xzr
; CHECKO0-NEXT: // kill: def $s0 killed $s0 killed $q0
; CHECKO0-NEXT: fmov w0, s0
; CHECKO0-NEXT: mov w8, wzr
; CHECKO0-NEXT: // kill: def $x0 killed $w0
; CHECKO0-NEXT: // kill: def $x8 killed $w8
; CHECKO0-NEXT: bfi x0, x8, #32, #32
; CHECKO0-NEXT: ret
;
; NEON-LABEL: popcount1x128:
; NEON: // %bb.0: // %Entry
; NEON-NEXT: fmov d0, x0
; NEON-NEXT: mov v0.d[1], x1
; NEON-NEXT: mov x1, xzr
; NEON-NEXT: cnt v0.16b, v0.16b
; NEON-NEXT: addv b0, v0.16b
; NEON-NEXT: fmov x0, d0
; NEON-NEXT: ret
;
; DOT-LABEL: popcount1x128:
; DOT: // %bb.0: // %Entry
; DOT-NEXT: fmov d0, x0
; DOT-NEXT: mov v0.d[1], x1
; DOT-NEXT: mov x1, xzr
; DOT-NEXT: cnt v0.16b, v0.16b
; DOT-NEXT: addv b0, v0.16b
; DOT-NEXT: fmov x0, d0
; DOT-NEXT: ret
;
; SVE-LABEL: popcount1x128:
; SVE: // %bb.0: // %Entry
; SVE-NEXT: fmov d0, x0
; SVE-NEXT: ptrue p0.d
; SVE-NEXT: mov v0.d[1], x1
; SVE-NEXT: mov x1, xzr
; SVE-NEXT: cnt z0.d, p0/m, z0.d
; SVE-NEXT: addp d0, v0.2d
; SVE-NEXT: fmov x0, d0
; SVE-NEXT: ret
;
; BE-LABEL: popcount1x128:
; BE: // %bb.0: // %Entry
; BE-NEXT: fmov d0, x0
; BE-NEXT: mov x0, xzr
; BE-NEXT: mov v0.d[1], x1
; BE-NEXT: rev64 v0.16b, v0.16b
; BE-NEXT: cnt v0.16b, v0.16b
; BE-NEXT: addv b0, v0.16b
; BE-NEXT: fmov x1, d0
; BE-NEXT: ret
;
; GISEL-LABEL: popcount1x128:
; GISEL: // %bb.0: // %Entry
; GISEL-NEXT: mov v0.d[0], x0
; GISEL-NEXT: mov v0.d[1], x1
; GISEL-NEXT: mov x1, xzr
; GISEL-NEXT: cnt v0.16b, v0.16b
; GISEL-NEXT: uaddlv h0, v0.16b
; GISEL-NEXT: mov w0, v0.s[0]
; GISEL-NEXT: ret
;
; GISELO0-LABEL: popcount1x128:
; GISELO0: // %bb.0: // %Entry
; GISELO0-NEXT: // implicit-def: $q0
; GISELO0-NEXT: mov v0.d[0], x0
; GISELO0-NEXT: mov v0.d[1], x1
; GISELO0-NEXT: cnt v0.16b, v0.16b
; GISELO0-NEXT: uaddlv h0, v0.16b
; GISELO0-NEXT: // kill: def $q0 killed $h0
; GISELO0-NEXT: mov x1, xzr
; GISELO0-NEXT: // kill: def $s0 killed $s0 killed $q0
; GISELO0-NEXT: fmov w0, s0
; GISELO0-NEXT: mov w8, wzr
; GISELO0-NEXT: // kill: def $x0 killed $w0
; GISELO0-NEXT: // kill: def $x8 killed $w8
; GISELO0-NEXT: bfi x0, x8, #32, #32
; GISELO0-NEXT: ret
Entry:
%1 = tail call <1 x i128> @llvm.ctpop.v1i128(<1 x i128> %0)
ret <1 x i128> %1
}
declare <1 x i128> @llvm.ctpop.v1i128(<1 x i128>)
define <2 x i64> @popcount2x64(<2 x i64> %0) {
; CHECKO0-LABEL: popcount2x64:
; CHECKO0: // %bb.0: // %Entry
; CHECKO0-NEXT: cnt v0.16b, v0.16b
; CHECKO0-NEXT: uaddlp v0.8h, v0.16b
; CHECKO0-NEXT: uaddlp v0.4s, v0.8h
; CHECKO0-NEXT: uaddlp v0.2d, v0.4s
; CHECKO0-NEXT: ret
;
; NEON-LABEL: popcount2x64:
; NEON: // %bb.0: // %Entry
; NEON-NEXT: cnt v0.16b, v0.16b
; NEON-NEXT: uaddlp v0.8h, v0.16b
; NEON-NEXT: uaddlp v0.4s, v0.8h
; NEON-NEXT: uaddlp v0.2d, v0.4s
; NEON-NEXT: ret
;
; DOT-LABEL: popcount2x64:
; DOT: // %bb.0: // %Entry
; DOT-NEXT: movi v1.16b, #1
; DOT-NEXT: cnt v0.16b, v0.16b
; DOT-NEXT: movi v2.2d, #0000000000000000
; DOT-NEXT: udot v2.4s, v1.16b, v0.16b
; DOT-NEXT: uaddlp v0.2d, v2.4s
; DOT-NEXT: ret
;
; SVE-LABEL: popcount2x64:
; SVE: // %bb.0: // %Entry
; SVE-NEXT: ptrue p0.d, vl2
; SVE-NEXT: // kill: def $q0 killed $q0 def $z0
; SVE-NEXT: cnt z0.d, p0/m, z0.d
; SVE-NEXT: // kill: def $q0 killed $q0 killed $z0
; SVE-NEXT: ret
;
; BE-LABEL: popcount2x64:
; BE: // %bb.0: // %Entry
; BE-NEXT: rev64 v0.16b, v0.16b
; BE-NEXT: ext v0.16b, v0.16b, v0.16b, #8
; BE-NEXT: cnt v0.16b, v0.16b
; BE-NEXT: uaddlp v0.8h, v0.16b
; BE-NEXT: uaddlp v0.4s, v0.8h
; BE-NEXT: uaddlp v0.2d, v0.4s
; BE-NEXT: ext v0.16b, v0.16b, v0.16b, #8
; BE-NEXT: ret
;
; GISELO0-LABEL: popcount2x64:
; GISELO0: // %bb.0: // %Entry
; GISELO0-NEXT: cnt v0.16b, v0.16b
; GISELO0-NEXT: uaddlp v0.8h, v0.16b
; GISELO0-NEXT: uaddlp v0.4s, v0.8h
; GISELO0-NEXT: uaddlp v0.2d, v0.4s
; GISELO0-NEXT: ret
;
; NEON-GISEL-LABEL: popcount2x64:
; NEON-GISEL: // %bb.0: // %Entry
; NEON-GISEL-NEXT: cnt v0.16b, v0.16b
; NEON-GISEL-NEXT: uaddlp v0.8h, v0.16b
; NEON-GISEL-NEXT: uaddlp v0.4s, v0.8h
; NEON-GISEL-NEXT: uaddlp v0.2d, v0.4s
; NEON-GISEL-NEXT: ret
;
; DOT-GISEL-LABEL: popcount2x64:
; DOT-GISEL: // %bb.0: // %Entry
; DOT-GISEL-NEXT: movi v1.2d, #0000000000000000
; DOT-GISEL-NEXT: cnt v0.16b, v0.16b
; DOT-GISEL-NEXT: movi v2.16b, #1
; DOT-GISEL-NEXT: udot v1.4s, v2.16b, v0.16b
; DOT-GISEL-NEXT: uaddlp v0.2d, v1.4s
; DOT-GISEL-NEXT: ret
;
; SVE-GISEL-LABEL: popcount2x64:
; SVE-GISEL: // %bb.0: // %Entry
; SVE-GISEL-NEXT: cnt v0.16b, v0.16b
; SVE-GISEL-NEXT: uaddlp v0.8h, v0.16b
; SVE-GISEL-NEXT: uaddlp v0.4s, v0.8h
; SVE-GISEL-NEXT: uaddlp v0.2d, v0.4s
; SVE-GISEL-NEXT: ret
Entry:
%1 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
ret <2 x i64> %1
}
declare <2 x i64> @llvm.ctpop.v2i64(<2 x i64>)
define <1 x i64> @popcount1x64(<1 x i64> %0) {
; CHECKO0-LABEL: popcount1x64:
; CHECKO0: // %bb.0: // %Entry
; CHECKO0-NEXT: fmov x0, d0
; CHECKO0-NEXT: fmov d0, x0
; CHECKO0-NEXT: cnt v0.8b, v0.8b
; CHECKO0-NEXT: uaddlv h0, v0.8b
; CHECKO0-NEXT: // kill: def $q0 killed $h0
; CHECKO0-NEXT: mov w8, v0.s[0]
; CHECKO0-NEXT: // kill: def $x8 killed $w8
; CHECKO0-NEXT: fmov d0, x8
; CHECKO0-NEXT: ret
;
; NEON-LABEL: popcount1x64:
; NEON: // %bb.0: // %Entry
; NEON-NEXT: cnt v0.8b, v0.8b
; NEON-NEXT: uaddlp v0.4h, v0.8b
; NEON-NEXT: uaddlp v0.2s, v0.4h
; NEON-NEXT: uaddlp v0.1d, v0.2s
; NEON-NEXT: ret
;
; DOT-LABEL: popcount1x64:
; DOT: // %bb.0: // %Entry
; DOT-NEXT: cnt v0.8b, v0.8b
; DOT-NEXT: uaddlp v0.4h, v0.8b
; DOT-NEXT: uaddlp v0.2s, v0.4h
; DOT-NEXT: uaddlp v0.1d, v0.2s
; DOT-NEXT: ret
;
; SVE-LABEL: popcount1x64:
; SVE: // %bb.0: // %Entry
; SVE-NEXT: ptrue p0.d, vl1
; SVE-NEXT: // kill: def $d0 killed $d0 def $z0
; SVE-NEXT: cnt z0.d, p0/m, z0.d
; SVE-NEXT: // kill: def $d0 killed $d0 killed $z0
; SVE-NEXT: ret
;
; BE-LABEL: popcount1x64:
; BE: // %bb.0: // %Entry
; BE-NEXT: rev64 v0.8b, v0.8b
; BE-NEXT: cnt v0.8b, v0.8b
; BE-NEXT: uaddlp v0.4h, v0.8b
; BE-NEXT: uaddlp v0.2s, v0.4h
; BE-NEXT: uaddlp v0.1d, v0.2s
; BE-NEXT: ret
;
; GISEL-LABEL: popcount1x64:
; GISEL: // %bb.0: // %Entry
; GISEL-NEXT: cnt v0.8b, v0.8b
; GISEL-NEXT: uaddlv h0, v0.8b
; GISEL-NEXT: mov w8, v0.s[0]
; GISEL-NEXT: fmov d0, x8
; GISEL-NEXT: ret
;
; GISELO0-LABEL: popcount1x64:
; GISELO0: // %bb.0: // %Entry
; GISELO0-NEXT: fmov x0, d0
; GISELO0-NEXT: fmov d0, x0
; GISELO0-NEXT: cnt v0.8b, v0.8b
; GISELO0-NEXT: uaddlv h0, v0.8b
; GISELO0-NEXT: // kill: def $q0 killed $h0
; GISELO0-NEXT: mov w8, v0.s[0]
; GISELO0-NEXT: // kill: def $x8 killed $w8
; GISELO0-NEXT: fmov d0, x8
; GISELO0-NEXT: ret
Entry:
%1 = tail call <1 x i64> @llvm.ctpop.v1i64(<1 x i64> %0)
ret <1 x i64> %1
}
declare <1 x i64> @llvm.ctpop.v1i64(<1 x i64>)
define <4 x i32> @popcount4x32(<4 x i32> %0) {
; CHECKO0-LABEL: popcount4x32:
; CHECKO0: // %bb.0: // %Entry
; CHECKO0-NEXT: cnt v0.16b, v0.16b
; CHECKO0-NEXT: uaddlp v0.8h, v0.16b
; CHECKO0-NEXT: uaddlp v0.4s, v0.8h
; CHECKO0-NEXT: ret
;
; NEON-LABEL: popcount4x32:
; NEON: // %bb.0: // %Entry
; NEON-NEXT: cnt v0.16b, v0.16b
; NEON-NEXT: uaddlp v0.8h, v0.16b
; NEON-NEXT: uaddlp v0.4s, v0.8h
; NEON-NEXT: ret
;
; DOT-LABEL: popcount4x32:
; DOT: // %bb.0: // %Entry
; DOT-NEXT: movi v1.16b, #1
; DOT-NEXT: cnt v2.16b, v0.16b
; DOT-NEXT: movi v0.2d, #0000000000000000
; DOT-NEXT: udot v0.4s, v1.16b, v2.16b
; DOT-NEXT: ret
;
; SVE-LABEL: popcount4x32:
; SVE: // %bb.0: // %Entry
; SVE-NEXT: ptrue p0.s, vl4
; SVE-NEXT: // kill: def $q0 killed $q0 def $z0
; SVE-NEXT: cnt z0.s, p0/m, z0.s
; SVE-NEXT: // kill: def $q0 killed $q0 killed $z0
; SVE-NEXT: ret
;
; BE-LABEL: popcount4x32:
; BE: // %bb.0: // %Entry
; BE-NEXT: rev64 v0.16b, v0.16b
; BE-NEXT: ext v0.16b, v0.16b, v0.16b, #8
; BE-NEXT: cnt v0.16b, v0.16b
; BE-NEXT: uaddlp v0.8h, v0.16b
; BE-NEXT: uaddlp v0.4s, v0.8h
; BE-NEXT: rev64 v0.4s, v0.4s
; BE-NEXT: ext v0.16b, v0.16b, v0.16b, #8
; BE-NEXT: ret
;
; GISELO0-LABEL: popcount4x32:
; GISELO0: // %bb.0: // %Entry
; GISELO0-NEXT: cnt v0.16b, v0.16b
; GISELO0-NEXT: uaddlp v0.8h, v0.16b
; GISELO0-NEXT: uaddlp v0.4s, v0.8h
; GISELO0-NEXT: ret
;
; NEON-GISEL-LABEL: popcount4x32:
; NEON-GISEL: // %bb.0: // %Entry
; NEON-GISEL-NEXT: cnt v0.16b, v0.16b
; NEON-GISEL-NEXT: uaddlp v0.8h, v0.16b
; NEON-GISEL-NEXT: uaddlp v0.4s, v0.8h
; NEON-GISEL-NEXT: ret
;
; DOT-GISEL-LABEL: popcount4x32:
; DOT-GISEL: // %bb.0: // %Entry
; DOT-GISEL-NEXT: movi v1.2d, #0000000000000000
; DOT-GISEL-NEXT: cnt v0.16b, v0.16b
; DOT-GISEL-NEXT: movi v2.16b, #1
; DOT-GISEL-NEXT: udot v1.4s, v2.16b, v0.16b
; DOT-GISEL-NEXT: mov v0.16b, v1.16b
; DOT-GISEL-NEXT: ret
;
; SVE-GISEL-LABEL: popcount4x32:
; SVE-GISEL: // %bb.0: // %Entry
; SVE-GISEL-NEXT: cnt v0.16b, v0.16b
; SVE-GISEL-NEXT: uaddlp v0.8h, v0.16b
; SVE-GISEL-NEXT: uaddlp v0.4s, v0.8h
; SVE-GISEL-NEXT: ret
Entry:
%1 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
ret <4 x i32> %1
}
declare <4 x i32> @llvm.ctpop.v4i32(<4 x i32>)
define <2 x i32> @popcount2x32(<2 x i32> %0) {
; CHECKO0-LABEL: popcount2x32:
; CHECKO0: // %bb.0: // %Entry
; CHECKO0-NEXT: cnt v0.8b, v0.8b
; CHECKO0-NEXT: uaddlp v0.4h, v0.8b
; CHECKO0-NEXT: uaddlp v0.2s, v0.4h
; CHECKO0-NEXT: ret
;
; NEON-LABEL: popcount2x32:
; NEON: // %bb.0: // %Entry
; NEON-NEXT: cnt v0.8b, v0.8b
; NEON-NEXT: uaddlp v0.4h, v0.8b
; NEON-NEXT: uaddlp v0.2s, v0.4h
; NEON-NEXT: ret
;
; DOT-LABEL: popcount2x32:
; DOT: // %bb.0: // %Entry
; DOT-NEXT: movi v1.2d, #0000000000000000
; DOT-NEXT: cnt v0.8b, v0.8b
; DOT-NEXT: movi v2.8b, #1
; DOT-NEXT: udot v1.2s, v2.8b, v0.8b
; DOT-NEXT: fmov d0, d1
; DOT-NEXT: ret
;
; SVE-LABEL: popcount2x32:
; SVE: // %bb.0: // %Entry
; SVE-NEXT: ptrue p0.s, vl2
; SVE-NEXT: // kill: def $d0 killed $d0 def $z0
; SVE-NEXT: cnt z0.s, p0/m, z0.s
; SVE-NEXT: // kill: def $d0 killed $d0 killed $z0
; SVE-NEXT: ret
;
; BE-LABEL: popcount2x32:
; BE: // %bb.0: // %Entry
; BE-NEXT: rev64 v0.8b, v0.8b
; BE-NEXT: cnt v0.8b, v0.8b
; BE-NEXT: uaddlp v0.4h, v0.8b
; BE-NEXT: uaddlp v0.2s, v0.4h
; BE-NEXT: rev64 v0.2s, v0.2s
; BE-NEXT: ret
;
; GISELO0-LABEL: popcount2x32:
; GISELO0: // %bb.0: // %Entry
; GISELO0-NEXT: cnt v0.8b, v0.8b
; GISELO0-NEXT: uaddlp v0.4h, v0.8b
; GISELO0-NEXT: uaddlp v0.2s, v0.4h
; GISELO0-NEXT: ret
;
; NEON-GISEL-LABEL: popcount2x32:
; NEON-GISEL: // %bb.0: // %Entry
; NEON-GISEL-NEXT: cnt v0.8b, v0.8b
; NEON-GISEL-NEXT: uaddlp v0.4h, v0.8b
; NEON-GISEL-NEXT: uaddlp v0.2s, v0.4h
; NEON-GISEL-NEXT: ret
;
; DOT-GISEL-LABEL: popcount2x32:
; DOT-GISEL: // %bb.0: // %Entry
; DOT-GISEL-NEXT: movi v1.2d, #0000000000000000
; DOT-GISEL-NEXT: cnt v0.8b, v0.8b
; DOT-GISEL-NEXT: movi v2.8b, #1
; DOT-GISEL-NEXT: udot v1.2s, v2.8b, v0.8b
; DOT-GISEL-NEXT: fmov d0, d1
; DOT-GISEL-NEXT: ret
;
; SVE-GISEL-LABEL: popcount2x32:
; SVE-GISEL: // %bb.0: // %Entry
; SVE-GISEL-NEXT: cnt v0.8b, v0.8b
; SVE-GISEL-NEXT: uaddlp v0.4h, v0.8b
; SVE-GISEL-NEXT: uaddlp v0.2s, v0.4h
; SVE-GISEL-NEXT: ret
Entry:
%1 = tail call <2 x i32> @llvm.ctpop.v2i32(<2 x i32> %0)
ret <2 x i32> %1
}
declare <2 x i32> @llvm.ctpop.v2i32(<2 x i32>)
define <8 x i16> @popcount8x16(<8 x i16> %0) {
; CHECKO0-LABEL: popcount8x16:
; CHECKO0: // %bb.0: // %Entry
; CHECKO0-NEXT: cnt v0.16b, v0.16b
; CHECKO0-NEXT: uaddlp v0.8h, v0.16b
; CHECKO0-NEXT: ret
;
; NEON-LABEL: popcount8x16:
; NEON: // %bb.0: // %Entry
; NEON-NEXT: cnt v0.16b, v0.16b
; NEON-NEXT: uaddlp v0.8h, v0.16b
; NEON-NEXT: ret
;
; DOT-LABEL: popcount8x16:
; DOT: // %bb.0: // %Entry
; DOT-NEXT: cnt v0.16b, v0.16b
; DOT-NEXT: uaddlp v0.8h, v0.16b
; DOT-NEXT: ret
;
; SVE-LABEL: popcount8x16:
; SVE: // %bb.0: // %Entry
; SVE-NEXT: ptrue p0.h, vl8
; SVE-NEXT: // kill: def $q0 killed $q0 def $z0
; SVE-NEXT: cnt z0.h, p0/m, z0.h
; SVE-NEXT: // kill: def $q0 killed $q0 killed $z0
; SVE-NEXT: ret
;
; BE-LABEL: popcount8x16:
; BE: // %bb.0: // %Entry
; BE-NEXT: rev64 v0.16b, v0.16b
; BE-NEXT: ext v0.16b, v0.16b, v0.16b, #8
; BE-NEXT: cnt v0.16b, v0.16b
; BE-NEXT: uaddlp v0.8h, v0.16b
; BE-NEXT: rev64 v0.8h, v0.8h
; BE-NEXT: ext v0.16b, v0.16b, v0.16b, #8
; BE-NEXT: ret
;
; GISEL-LABEL: popcount8x16:
; GISEL: // %bb.0: // %Entry
; GISEL-NEXT: cnt v0.16b, v0.16b
; GISEL-NEXT: uaddlp v0.8h, v0.16b
; GISEL-NEXT: ret
;
; GISELO0-LABEL: popcount8x16:
; GISELO0: // %bb.0: // %Entry
; GISELO0-NEXT: cnt v0.16b, v0.16b
; GISELO0-NEXT: uaddlp v0.8h, v0.16b
; GISELO0-NEXT: ret
Entry:
%1 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
ret <8 x i16> %1
}
declare <8 x i16> @llvm.ctpop.v8i16(<8 x i16>)
define <4 x i16> @popcount4x16(<4 x i16> %0) {
; CHECKO0-LABEL: popcount4x16:
; CHECKO0: // %bb.0: // %Entry
; CHECKO0-NEXT: cnt v0.8b, v0.8b
; CHECKO0-NEXT: uaddlp v0.4h, v0.8b
; CHECKO0-NEXT: ret
;
; NEON-LABEL: popcount4x16:
; NEON: // %bb.0: // %Entry
; NEON-NEXT: cnt v0.8b, v0.8b
; NEON-NEXT: uaddlp v0.4h, v0.8b
; NEON-NEXT: ret
;
; DOT-LABEL: popcount4x16:
; DOT: // %bb.0: // %Entry
; DOT-NEXT: cnt v0.8b, v0.8b
; DOT-NEXT: uaddlp v0.4h, v0.8b
; DOT-NEXT: ret
;
; SVE-LABEL: popcount4x16:
; SVE: // %bb.0: // %Entry
; SVE-NEXT: ptrue p0.h, vl4
; SVE-NEXT: // kill: def $d0 killed $d0 def $z0
; SVE-NEXT: cnt z0.h, p0/m, z0.h
; SVE-NEXT: // kill: def $d0 killed $d0 killed $z0
; SVE-NEXT: ret
;
; BE-LABEL: popcount4x16:
; BE: // %bb.0: // %Entry
; BE-NEXT: rev64 v0.8b, v0.8b
; BE-NEXT: cnt v0.8b, v0.8b
; BE-NEXT: uaddlp v0.4h, v0.8b
; BE-NEXT: rev64 v0.4h, v0.4h
; BE-NEXT: ret
;
; GISEL-LABEL: popcount4x16:
; GISEL: // %bb.0: // %Entry
; GISEL-NEXT: cnt v0.8b, v0.8b
; GISEL-NEXT: uaddlp v0.4h, v0.8b
; GISEL-NEXT: ret
;
; GISELO0-LABEL: popcount4x16:
; GISELO0: // %bb.0: // %Entry
; GISELO0-NEXT: cnt v0.8b, v0.8b
; GISELO0-NEXT: uaddlp v0.4h, v0.8b
; GISELO0-NEXT: ret
Entry:
%1 = tail call <4 x i16> @llvm.ctpop.v4i16(<4 x i16> %0)
ret <4 x i16> %1
}
define i32 @ctpop_into_extract(ptr %p) {
; CHECKO0-LABEL: ctpop_into_extract:
; CHECKO0: // %bb.0:
; CHECKO0-NEXT: mov w8, #-1 // =0xffffffff
; CHECKO0-NEXT: // implicit-def: $d2
; CHECKO0-NEXT: fmov s2, w8
; CHECKO0-NEXT: ldr d0, [x0]
; CHECKO0-NEXT: fmov s1, s0
; CHECKO0-NEXT: fmov w8, s1
; CHECKO0-NEXT: fmov s1, w8
; CHECKO0-NEXT: // kill: def $d1 killed $s1
; CHECKO0-NEXT: cnt v1.8b, v1.8b
; CHECKO0-NEXT: uaddlv h1, v1.8b
; CHECKO0-NEXT: // kill: def $q1 killed $h1
; CHECKO0-NEXT: // kill: def $s1 killed $s1 killed $q1
; CHECKO0-NEXT: fmov w8, s1
; CHECKO0-NEXT: // implicit-def: $q1
; CHECKO0-NEXT: fmov d1, d2
; CHECKO0-NEXT: mov v1.s[1], w8
; CHECKO0-NEXT: // kill: def $d1 killed $d1 killed $q1
; CHECKO0-NEXT: sub v0.2s, v0.2s, v1.2s
; CHECKO0-NEXT: str d0, [x0]
; CHECKO0-NEXT: mov w0, wzr
; CHECKO0-NEXT: ret
;
; NEON-LABEL: ctpop_into_extract:
; NEON: // %bb.0:
; NEON-NEXT: ldr d0, [x0]
; NEON-NEXT: movi v2.2d, #0xffffffffffffffff
; NEON-NEXT: mov x8, x0
; NEON-NEXT: mov w0, wzr
; NEON-NEXT: fmov w9, s0
; NEON-NEXT: fmov s1, w9
; NEON-NEXT: cnt v1.8b, v1.8b
; NEON-NEXT: addv b1, v1.8b
; NEON-NEXT: mov v2.s[1], v1.s[0]
; NEON-NEXT: sub v0.2s, v0.2s, v2.2s
; NEON-NEXT: str d0, [x8]
; NEON-NEXT: ret
;
; DOT-LABEL: ctpop_into_extract:
; DOT: // %bb.0:
; DOT-NEXT: ldr d0, [x0]
; DOT-NEXT: movi v2.2d, #0xffffffffffffffff
; DOT-NEXT: mov x8, x0
; DOT-NEXT: mov w0, wzr
; DOT-NEXT: fmov w9, s0
; DOT-NEXT: fmov s1, w9
; DOT-NEXT: cnt v1.8b, v1.8b
; DOT-NEXT: addv b1, v1.8b
; DOT-NEXT: mov v2.s[1], v1.s[0]
; DOT-NEXT: sub v0.2s, v0.2s, v2.2s
; DOT-NEXT: str d0, [x8]
; DOT-NEXT: ret
;
; SVE-LABEL: ctpop_into_extract:
; SVE: // %bb.0:
; SVE-NEXT: ldr d0, [x0]
; SVE-NEXT: ptrue p0.s
; SVE-NEXT: movi v2.2d, #0xffffffffffffffff
; SVE-NEXT: mov x8, x0
; SVE-NEXT: mov w0, wzr
; SVE-NEXT: movprfx z1, z0
; SVE-NEXT: cnt z1.s, p0/m, z0.s
; SVE-NEXT: mov v2.s[1], v1.s[0]
; SVE-NEXT: sub v0.2s, v0.2s, v2.2s
; SVE-NEXT: str d0, [x8]
; SVE-NEXT: ret
;
; BE-LABEL: ctpop_into_extract:
; BE: // %bb.0:
; BE-NEXT: ld1 { v0.2s }, [x0]
; BE-NEXT: movi v2.2d, #0xffffffffffffffff
; BE-NEXT: mov x8, x0
; BE-NEXT: mov w0, wzr
; BE-NEXT: fmov w9, s0
; BE-NEXT: fmov s1, w9
; BE-NEXT: cnt v1.8b, v1.8b
; BE-NEXT: addv b1, v1.8b
; BE-NEXT: mov v2.s[1], v1.s[0]
; BE-NEXT: sub v0.2s, v0.2s, v2.2s
; BE-NEXT: st1 { v0.2s }, [x8]
; BE-NEXT: ret
;
; GISEL-LABEL: ctpop_into_extract:
; GISEL: // %bb.0:
; GISEL-NEXT: ldr d0, [x0]
; GISEL-NEXT: mov x8, x0
; GISEL-NEXT: mov w0, wzr
; GISEL-NEXT: fmov w9, s0
; GISEL-NEXT: fmov s1, w9
; GISEL-NEXT: mov w9, #-1 // =0xffffffff
; GISEL-NEXT: fmov s2, w9
; GISEL-NEXT: cnt v1.8b, v1.8b
; GISEL-NEXT: uaddlv h1, v1.8b
; GISEL-NEXT: mov v2.s[1], v1.s[0]
; GISEL-NEXT: sub v0.2s, v0.2s, v2.2s
; GISEL-NEXT: str d0, [x8]
; GISEL-NEXT: ret
;
; GISELO0-LABEL: ctpop_into_extract:
; GISELO0: // %bb.0:
; GISELO0-NEXT: mov w8, #-1 // =0xffffffff
; GISELO0-NEXT: // implicit-def: $d2
; GISELO0-NEXT: fmov s2, w8
; GISELO0-NEXT: ldr d0, [x0]
; GISELO0-NEXT: fmov s1, s0
; GISELO0-NEXT: fmov w8, s1
; GISELO0-NEXT: fmov s1, w8
; GISELO0-NEXT: // kill: def $d1 killed $s1
; GISELO0-NEXT: cnt v1.8b, v1.8b
; GISELO0-NEXT: uaddlv h1, v1.8b
; GISELO0-NEXT: // kill: def $q1 killed $h1
; GISELO0-NEXT: // kill: def $s1 killed $s1 killed $q1
; GISELO0-NEXT: fmov w8, s1
; GISELO0-NEXT: // implicit-def: $q1
; GISELO0-NEXT: fmov d1, d2
; GISELO0-NEXT: mov v1.s[1], w8
; GISELO0-NEXT: // kill: def $d1 killed $d1 killed $q1
; GISELO0-NEXT: sub v0.2s, v0.2s, v1.2s
; GISELO0-NEXT: str d0, [x0]
; GISELO0-NEXT: mov w0, wzr
; GISELO0-NEXT: ret
%1 = load <2 x i32>, ptr %p, align 4
%2 = extractelement <2 x i32> %1, i64 0
%3 = call i32 @llvm.ctpop.i32(i32 %2)
%4 = insertelement <2 x i32> <i32 -1, i32 poison>, i32 %3, i64 1
%5 = sub <2 x i32> %1, %4
store <2 x i32> %5, ptr %p, align 4
ret i32 0
}
declare <4 x i16> @llvm.ctpop.v4i16(<4 x i16>)
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; CHECK: {{.*}}