This is an archive of the discontinued LLVM Phabricator instance.

Differential D143857

[X86] Add tests for shuffle as shift/rotate; NFC
ClosedPublic

Authored by goldstein.w.n on Feb 12 2023, 2:14 PM.

Details

Reviewers
RKSimon
pengfei
Commits
rG015b3817ca1e: Add tests for shuffle as shift/rotate; NFC

Diff Detail

Event Timeline

goldstein.w.n created this revision.Feb 12 2023, 2:14 PM
Herald added a project: Restricted Project. · View Herald TranscriptFeb 12 2023, 2:14 PM
Herald added a subscriber: pengfei. · View Herald Transcript
goldstein.w.n requested review of this revision.Feb 12 2023, 2:14 PM
Herald added a project: Restricted Project. · View Herald TranscriptFeb 12 2023, 2:14 PM
Herald added a subscriber: llvm-commits. · View Herald Transcript
goldstein.w.n added reviewers: RKSimon, pengfei.Feb 12 2023, 2:18 PM
goldstein.w.n added a child revision: D143786: [X86] Add `TuningPreferShiftShuffle` for when Shifts are preferable to shuffles..
goldstein.w.n added a parent revision: D143785: [X86] Add Extend shuffle pattern to vNf32 shuffles..Feb 12 2023, 2:27 PM
goldstein.w.n updated this revision to Diff 497557.Feb 14 2023, 11:14 PM

Rebase

Harbormaster completed remote builds in B213812: Diff 497557.Feb 14 2023, 11:59 PM
goldstein.w.n updated this revision to Diff 498069.Feb 16 2023, 10:26 AM

Rebase

Harbormaster completed remote builds in B214195: Diff 498069.Feb 16 2023, 12:08 PM
RKSimon accepted this revision.Feb 17 2023, 3:54 AM

LGTM with a few minors

llvm/test/CodeGen/X86/shuffle-as-shifts.ll
4

Include a test run with x86-86-v4 (and maybe znver4)

7

Maybe rename these to include the shuffle mask - its often very useful when searching: shuffle_rot_v4i32_1032 etc.

This revision is now accepted and ready to land.Feb 17 2023, 3:54 AM
goldstein.w.n marked 2 inline comments as done.Feb 20 2023, 6:02 PM
goldstein.w.n updated this revision to Diff 498987.Feb 20 2023, 6:03 PM

Add more targets + common check + rename

goldstein.w.n updated this revision to Diff 498990.Feb 20 2023, 6:05 PM

Rename functions + add more targets + common checks

goldstein.w.n updated this revision to Diff 498996.Feb 20 2023, 6:08 PM

Rename functions + common check + more targets

Harbormaster completed remote builds in B214888: Diff 498996.Feb 20 2023, 7:32 PM
goldstein.w.n updated this revision to Diff 499021.Feb 20 2023, 7:52 PM

Rebase

Harbormaster completed remote builds in B214910: Diff 499021.Feb 20 2023, 10:18 PM
goldstein.w.n updated this revision to Diff 499379.Feb 21 2023, 11:35 PM

Rebase

goldstein.w.n updated this revision to Diff 499540.Feb 22 2023, 9:12 AM

Rebase

Harbormaster completed remote builds in B215280: Diff 499540.Feb 22 2023, 10:21 AM
goldstein.w.n updated this revision to Diff 499691.Feb 22 2023, 5:40 PM

Rebase

Harbormaster completed remote builds in B215399: Diff 499691.Feb 22 2023, 7:21 PM
goldstein.w.n updated this revision to Diff 500058.Feb 23 2023, 8:09 PM

Rebase

Harbormaster completed remote builds in B215659: Diff 500058.Feb 23 2023, 9:59 PM
Closed by commit rG015b3817ca1e: Add tests for shuffle as shift/rotate; NFC (authored by goldstein.w.n). · Explain WhyFeb 24 2023, 1:22 PM
This revision was automatically updated to reflect the committed changes.
goldstein.w.n added a commit: rG015b3817ca1e: Add tests for shuffle as shift/rotate; NFC.
goldstein.w.n removed a child revision: D143786: [X86] Add `TuningPreferShiftShuffle` for when Shifts are preferable to shuffles..Feb 27 2023, 4:21 PM

Revision Contents

PathSize
llvm/
test/
CodeGen/
X86/
242 lines

Diff 500286

llvm/test/CodeGen/X86/shuffle-as-shifts.ll

  • This file was added.
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=skylake-avx512 | FileCheck %s --check-prefixes=CHECK,CHECK-SKX
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=icelake-server | FileCheck %s --check-prefixes=CHECK,CHECK-ICX
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64-v4 | FileCheck %s --check-prefixes=CHECK,CHECK-V4
RKSimonUnsubmitted
Done
ReplyInline Actions

Include a test run with x86-86-v4 (and maybe znver4)

RKSimon: Include a test run with x86-86-v4 (and maybe znver4)
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=znver4 | FileCheck %s --check-prefixes=CHECK,CHECK-ZNVER4
RKSimonUnsubmitted
Done
ReplyInline Actions

Maybe rename these to include the shuffle mask - its often very useful when searching: shuffle_rot_v4i32_1032 etc.

RKSimon: Maybe rename these to include the shuffle mask - its often very useful when searching…
define <4 x i32> @shuf_rot_v4i32_1032(<4 x i32> %x) {
; CHECK-LABEL: shuf_rot_v4i32_1032:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddd %xmm0, %xmm0, %xmm0
; CHECK-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,0,3,2]
; CHECK-NEXT: retq
%x1 = add <4 x i32> %x, %x
%r = shufflevector <4 x i32> %x1, <4 x i32> zeroinitializer, <4 x i32> <i32 1, i32 0, i32 3, i32 2>
ret <4 x i32> %r
}
define <8 x i32> @shuf_rot_v8i32_10325476(<8 x i32> %x) {
; CHECK-LABEL: shuf_rot_v8i32_10325476:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddd %ymm0, %ymm0, %ymm0
; CHECK-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
; CHECK-NEXT: retq
%x1 = add <8 x i32> %x, %x
%r = shufflevector <8 x i32> %x1, <8 x i32> zeroinitializer, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
ret <8 x i32> %r
}
define <16 x i32> @shuf_rot_v16i32_1032547698111013121514(<16 x i32> %x) {
; CHECK-LABEL: shuf_rot_v16i32_1032547698111013121514:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddd %zmm0, %zmm0, %zmm0
; CHECK-NEXT: vpshufd {{.*#+}} zmm0 = zmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
; CHECK-NEXT: retq
%x1 = add <16 x i32> %x, %x
%r = shufflevector <16 x i32> %x1, <16 x i32> zeroinitializer, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
ret <16 x i32> %r
}
define <8 x i16> @shuf_rot_v8i16_10325476(<8 x i16> %x) {
; CHECK-LABEL: shuf_rot_v8i16_10325476:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddw %xmm0, %xmm0, %xmm0
; CHECK-NEXT: vprold $16, %xmm0, %xmm0
; CHECK-NEXT: retq
%x1 = add <8 x i16> %x, %x
%r = shufflevector <8 x i16> %x1, <8 x i16> zeroinitializer, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
ret <8 x i16> %r
}
define <16 x i16> @shuf_rot_v16i16_1032547698111013121514(<16 x i16> %x) {
; CHECK-LABEL: shuf_rot_v16i16_1032547698111013121514:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddw %ymm0, %ymm0, %ymm0
; CHECK-NEXT: vprold $16, %ymm0, %ymm0
; CHECK-NEXT: retq
%x1 = add <16 x i16> %x, %x
%r = shufflevector <16 x i16> %x1, <16 x i16> zeroinitializer, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
ret <16 x i16> %r
}
define <32 x i16> @shuf_rot_v32i16_1234056749101181314151217181916212223202527272429303128(<32 x i16> %x) {
; CHECK-LABEL: shuf_rot_v32i16_1234056749101181314151217181916212223202527272429303128:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddw %zmm0, %zmm0, %zmm0
; CHECK-NEXT: vprolq $48, %zmm0, %zmm0
; CHECK-NEXT: retq
%x1 = add <32 x i16> %x, %x
%r = shufflevector <32 x i16> %x1, <32 x i16> zeroinitializer, <32 x i32> <i32 1,i32 2,i32 3,i32 0,i32 5,i32 6,i32 7,i32 4,i32 9,i32 10,i32 11,i32 8,i32 13,i32 14,i32 15,i32 12,i32 17,i32 18,i32 19,i32 16,i32 21,i32 22,i32 23,i32 20,i32 25,i32 26,i32 27,i32 24,i32 29,i32 30,i32 31,i32 28>
ret <32 x i16> %r
}
define <16 x i8> @shuf_rot_v16i8_2301674510118914151213(<16 x i8> %x) {
; CHECK-LABEL: shuf_rot_v16i8_2301674510118914151213:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddb %xmm0, %xmm0, %xmm0
; CHECK-NEXT: vprold $16, %xmm0, %xmm0
; CHECK-NEXT: retq
%x1 = add <16 x i8> %x, %x
%r = shufflevector <16 x i8> %x1, <16 x i8> zeroinitializer, <16 x i32> <i32 2, i32 3, i32 0, i32 1, i32 6, i32 7, i32 4, i32 5, i32 10, i32 11, i32 8, i32 9, i32 14, i32 15, i32 12, i32 13>
ret <16 x i8> %r
}
define <32 x i8> @shuf_rot_v32i8_230167451011891415121318191617222320212627242530312829(<32 x i8> %x) {
; CHECK-LABEL: shuf_rot_v32i8_230167451011891415121318191617222320212627242530312829:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddb %ymm0, %ymm0, %ymm0
; CHECK-NEXT: vprold $16, %ymm0, %ymm0
; CHECK-NEXT: retq
%x1 = add <32 x i8> %x, %x
%r = shufflevector <32 x i8> %x1, <32 x i8> zeroinitializer, <32 x i32> <i32 2, i32 3, i32 0, i32 1, i32 6, i32 7, i32 4, i32 5, i32 10, i32 11, i32 8, i32 9, i32 14, i32 15, i32 12, i32 13, i32 18, i32 19, i32 16, i32 17, i32 22, i32 23, i32 20, i32 21, i32 26, i32 27, i32 24, i32 25, i32 30, i32 31, i32 28, i32 29>
ret <32 x i8> %r
}
define <64 x i8> @shuf_rot_v64i8_3012745611891015121314191617182320212227242526312829303532333439363738434041424744454651484950555253545956575863606162(<64 x i8> %x) {
; CHECK-LABEL: shuf_rot_v64i8_3012745611891015121314191617182320212227242526312829303532333439363738434041424744454651484950555253545956575863606162:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddb %zmm0, %zmm0, %zmm0
; CHECK-NEXT: vprold $8, %zmm0, %zmm0
; CHECK-NEXT: retq
%x1 = add <64 x i8> %x, %x
%r = shufflevector <64 x i8> %x1, <64 x i8> zeroinitializer, <64 x i32> <i32 3,i32 0,i32 1,i32 2,i32 7,i32 4,i32 5,i32 6,i32 11,i32 8,i32 9,i32 10,i32 15,i32 12,i32 13,i32 14,i32 19,i32 16,i32 17,i32 18,i32 23,i32 20,i32 21,i32 22,i32 27,i32 24,i32 25,i32 26,i32 31,i32 28,i32 29,i32 30,i32 35,i32 32,i32 33,i32 34,i32 39,i32 36,i32 37,i32 38,i32 43,i32 40,i32 41,i32 42,i32 47,i32 44,i32 45,i32 46,i32 51,i32 48,i32 49,i32 50,i32 55,i32 52,i32 53,i32 54,i32 59,i32 56,i32 57,i32 58,i32 63,i32 60,i32 61,i32 62>
ret <64 x i8> %r
}
define <4 x i32> @shuf_shr_v4i32_1U3U(<4 x i32> %x) {
; CHECK-LABEL: shuf_shr_v4i32_1U3U:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddd %xmm0, %xmm0, %xmm0
; CHECK-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
; CHECK-NEXT: retq
%x1 = add <4 x i32> %x, %x
%r = shufflevector <4 x i32> %x1, <4 x i32> zeroinitializer, <4 x i32> <i32 1, i32 undef, i32 3, i32 undef>
ret <4 x i32> %r
}
define <8 x i32> @shuf_shr_v8i32_1U3U5U7U(<8 x i32> %x) {
; CHECK-LABEL: shuf_shr_v8i32_1U3U5U7U:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddd %ymm0, %ymm0, %ymm0
; CHECK-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
; CHECK-NEXT: retq
%x1 = add <8 x i32> %x, %x
%r = shufflevector <8 x i32> %x1, <8 x i32> zeroinitializer, <8 x i32> <i32 1, i32 undef, i32 3, i32 undef, i32 5, i32 undef, i32 7, i32 undef>
ret <8 x i32> %r
}
define <16 x i32> @shuf_shr_v16i32_U3U5U7U9U11U13U15(<16 x i32> %x) {
; CHECK-LABEL: shuf_shr_v16i32_U3U5U7U9U11U13U15:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddd %zmm0, %zmm0, %zmm0
; CHECK-NEXT: vpshufd {{.*#+}} zmm0 = zmm0[1,1,3,3,5,5,7,7,9,9,11,11,13,13,15,15]
; CHECK-NEXT: retq
%x1 = add <16 x i32> %x, %x
%r = shufflevector <16 x i32> %x1, <16 x i32> zeroinitializer, <16 x i32> <i32 1, i32 undef, i32 3, i32 undef, i32 5, i32 undef, i32 7, i32 undef, i32 9, i32 undef, i32 11, i32 undef, i32 13, i32 undef, i32 15, i32 undef>
ret <16 x i32> %r
}
define <8 x i16> @shuf_shr_v8i16_123U567U(<8 x i16> %x) {
; CHECK-LABEL: shuf_shr_v8i16_123U567U:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddw %xmm0, %xmm0, %xmm0
; CHECK-NEXT: vpsrlq $16, %xmm0, %xmm0
; CHECK-NEXT: retq
%x1 = add <8 x i16> %x, %x
%r = shufflevector <8 x i16> %x1, <8 x i16> zeroinitializer, <8 x i32> <i32 1, i32 2, i32 3, i32 undef, i32 5, i32 6, i32 7, i32 undef>
ret <8 x i16> %r
}
define <32 x i16> @shuf_shr_v32i16_1U3U5U7U9U11U13U15U17U19U21U23U25U27U29U31U(<32 x i16> %x) {
; CHECK-LABEL: shuf_shr_v32i16_1U3U5U7U9U11U13U15U17U19U21U23U25U27U29U31U:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddw %zmm0, %zmm0, %zmm0
; CHECK-NEXT: vpsrld $16, %zmm0, %zmm0
; CHECK-NEXT: retq
%x1 = add <32 x i16> %x, %x
%r = shufflevector <32 x i16> %x1, <32 x i16> zeroinitializer, <32 x i32> <i32 1, i32 undef, i32 3, i32 undef, i32 5, i32 undef, i32 7, i32 undef, i32 9, i32 undef, i32 11, i32 undef, i32 13, i32 undef, i32 15, i32 undef, i32 17, i32 undef, i32 19, i32 undef, i32 21, i32 undef, i32 23, i32 undef, i32 25, i32 undef, i32 27, i32 undef, i32 29, i32 undef, i32 31, i32 undef>
ret <32 x i16> %r
}
define <32 x i8> @shuf_shr_v32i8_1U3U5U7U9U11U13U15U17U19U21U23U25U27U29U31U(<32 x i8> %x) {
; CHECK-LABEL: shuf_shr_v32i8_1U3U5U7U9U11U13U15U17U19U21U23U25U27U29U31U:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddb %ymm0, %ymm0, %ymm0
; CHECK-NEXT: vpsrlw $8, %ymm0, %ymm0
; CHECK-NEXT: retq
%x1 = add <32 x i8> %x, %x
%r = shufflevector <32 x i8> %x1, <32 x i8> zeroinitializer, <32 x i32> <i32 1, i32 undef, i32 3, i32 undef, i32 5, i32 undef, i32 7, i32 undef, i32 9, i32 undef, i32 11, i32 undef, i32 13, i32 undef, i32 15, i32 undef, i32 17, i32 undef, i32 19, i32 undef, i32 21, i32 undef, i32 23, i32 undef, i32 25, i32 undef, i32 27, i32 undef, i32 29, i32 undef, i32 31, i32 undef>
ret <32 x i8> %r
}
define <4 x i32> @shuf_shl_v4i32_U0U2(<4 x i32> %x) {
; CHECK-LABEL: shuf_shl_v4i32_U0U2:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddd %xmm0, %xmm0, %xmm0
; CHECK-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,2,2]
; CHECK-NEXT: retq
%x1 = add <4 x i32> %x, %x
%r = shufflevector <4 x i32> %x1, <4 x i32> zeroinitializer, <4 x i32> <i32 undef, i32 0, i32 undef, i32 2>
ret <4 x i32> %r
}
define <8 x i32> @shuf_shl_v8i32_U0U2U4U6(<8 x i32> %x) {
; CHECK-LABEL: shuf_shl_v8i32_U0U2U4U6:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddd %ymm0, %ymm0, %ymm0
; CHECK-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6]
; CHECK-NEXT: retq
%x1 = add <8 x i32> %x, %x
%r = shufflevector <8 x i32> %x1, <8 x i32> zeroinitializer, <8 x i32> <i32 undef, i32 0, i32 undef, i32 2, i32 undef, i32 4, i32 undef, i32 6>
ret <8 x i32> %r
}
define <16 x i32> @shuf_shl_v16i32_U0U2U4U6U8U10U12U14(<16 x i32> %x) {
; CHECK-LABEL: shuf_shl_v16i32_U0U2U4U6U8U10U12U14:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddd %zmm0, %zmm0, %zmm0
; CHECK-NEXT: vpshufd {{.*#+}} zmm0 = zmm0[0,0,2,2,4,4,6,6,8,8,10,10,12,12,14,14]
; CHECK-NEXT: retq
%x1 = add <16 x i32> %x, %x
%r = shufflevector <16 x i32> %x1, <16 x i32> zeroinitializer, <16 x i32> <i32 undef, i32 0, i32 undef, i32 2, i32 undef, i32 4, i32 undef, i32 6, i32 undef, i32 8, i32 undef, i32 10, i32 undef, i32 12, i32 undef, i32 14>
ret <16 x i32> %r
}
define <16 x i16> @shuf_shl_v16i16_U0U2U4U6U8U10U12U14(<16 x i16> %x) {
; CHECK-LABEL: shuf_shl_v16i16_U0U2U4U6U8U10U12U14:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddw %ymm0, %ymm0, %ymm0
; CHECK-NEXT: vpslld $16, %ymm0, %ymm0
; CHECK-NEXT: retq
%x1 = add <16 x i16> %x, %x
%r = shufflevector <16 x i16> %x1, <16 x i16> zeroinitializer, <16 x i32> <i32 undef, i32 0, i32 undef, i32 2, i32 undef, i32 4, i32 undef, i32 6, i32 undef, i32 8, i32 undef, i32 10, i32 undef, i32 12, i32 undef, i32 14>
ret <16 x i16> %r
}
define <16 x i8> @shuf_shl_v16i8_U0U2U4U6U8U10U12U14(<16 x i8> %x) {
; CHECK-LABEL: shuf_shl_v16i8_U0U2U4U6U8U10U12U14:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddb %xmm0, %xmm0, %xmm0
; CHECK-NEXT: vpsllw $8, %xmm0, %xmm0
; CHECK-NEXT: retq
%x1 = add <16 x i8> %x, %x
%r = shufflevector <16 x i8> %x1, <16 x i8> zeroinitializer, <16 x i32> <i32 undef, i32 0, i32 undef, i32 2, i32 undef, i32 4, i32 undef, i32 6, i32 undef, i32 8, i32 undef, i32 10, i32 undef, i32 12, i32 undef, i32 14>
ret <16 x i8> %r
}
define <64 x i8> @shuf_shl_v64i8_U0U2U4U6U8U10U12U14U16U18U20U22U24U26U28U30U32U34U36U38U40U42U44U46U48U50U52U54U56U58U60U62(<64 x i8> %x) {
; CHECK-LABEL: shuf_shl_v64i8_U0U2U4U6U8U10U12U14U16U18U20U22U24U26U28U30U32U34U36U38U40U42U44U46U48U50U52U54U56U58U60U62:
; CHECK: # %bb.0:
; CHECK-NEXT: vpaddb %zmm0, %zmm0, %zmm0
; CHECK-NEXT: vpsllw $8, %zmm0, %zmm0
; CHECK-NEXT: retq
%x1 = add <64 x i8> %x, %x
%r = shufflevector <64 x i8> %x1, <64 x i8> zeroinitializer, <64 x i32> <i32 undef, i32 0, i32 undef, i32 2, i32 undef, i32 4, i32 undef, i32 6, i32 undef, i32 8, i32 undef, i32 10, i32 undef, i32 12, i32 undef, i32 14, i32 undef, i32 16, i32 undef, i32 18, i32 undef, i32 20, i32 undef, i32 22, i32 undef, i32 24, i32 undef, i32 26, i32 undef, i32 28, i32 undef, i32 30, i32 undef, i32 32, i32 undef, i32 34, i32 undef, i32 36, i32 undef, i32 38, i32 undef, i32 40, i32 undef, i32 42, i32 undef, i32 44, i32 undef, i32 46, i32 undef, i32 48, i32 undef, i32 50, i32 undef, i32 52, i32 undef, i32 54, i32 undef, i32 56, i32 undef, i32 58, i32 undef, i32 60, i32 undef, i32 62>
ret <64 x i8> %r
}
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; CHECK-ICX: {{.*}}
; CHECK-SKX: {{.*}}
; CHECK-V4: {{.*}}
; CHECK-ZNVER4: {{.*}}