-
Notifications
You must be signed in to change notification settings - Fork 15
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
[AIE2P][AA] added AIE2P intrinsics to AA
- Loading branch information
1 parent
ce3dc2c
commit a149a57
Showing
3 changed files
with
970 additions
and
1 deletion.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
214 changes: 214 additions & 0 deletions
214
llvm/test/CodeGen/AIE/alias-analysis-fifo-vst-vld-scheduling.mir
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,214 @@ | ||
| # NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py | ||
| # | ||
| # This file is licensed under the Apache License v2.0 with LLVM Exceptions. | ||
| # See https://llvm.org/LICENSE.txt for license information. | ||
| # SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception | ||
| # | ||
| # (c) Copyright 2025 Advanced Micro Devices, Inc. or its affiliates | ||
| # RUN: llc -mtriple=aie2p --start-after=machine-scheduler -o - %s | FileCheck %s | ||
|
|
||
| # no alias conflicts between load and stores | ||
|
|
||
| --- | | ||
| ; ModuleID = 'llvm/test/CodeGen/AIE/alias-analysis-fifo-vst-vld-scheduling.ll' | ||
| source_filename = "0_0/src/0_0.cc" | ||
| target datalayout = "e-m:e-p:20:32-i1:8:32-i8:8:32-i16:16:32-i32:32:32-f32:32:32-i64:32-f64:32-a:0:32-n32" | ||
| target triple = "aie2p" | ||
|
|
||
| $concurrentLoadStore = comdat any | ||
|
|
||
| ; Function Attrs: mustprogress noinline | ||
| define weak_odr dso_local void @concurrentLoadStore(ptr noalias %in, ptr noalias %out, ptr noalias nonnull align 64 dereferenceable(64) %params) local_unnamed_addr #0 comdat { | ||
| ; CHECK-LABEL: concurrentLoadStore: | ||
| ; CHECK: .p2align 4 | ||
| ; CHECK-NEXT: // %bb.0: // %entry | ||
| ; CHECK-NEXT: mova r26, #0; nopb ; nops ; nopx ; mov p2, p1; nopv | ||
| ; CHECK-NEXT: nopx ; vbcst.32 x0, r26 | ||
| ; CHECK-NEXT: vbcst.8 x2, r26 | ||
| ; CHECK-NEXT: mov r24, r26 | ||
| ; CHECK-NEXT: mov r27, r26 | ||
| ; CHECK-NEXT: mov el2, r26 | ||
| ; CHECK-NEXT: mov eh2, r27 | ||
| ; CHECK-NEXT: vlda.fill.512 [p0, lf0, r24]; vmov x1, x0 | ||
| ; CHECK-NEXT: vldb.pop.512 x4, [p0, lf0, r24]; vmov lfl0, x0 | ||
| ; CHECK-NEXT: vmov lfh0, x1 | ||
| ; CHECK-NEXT: ret lr; vmov sfl, x0 | ||
| ; CHECK-NEXT: vmov sfh, x1 // Delay Slot 5 | ||
| ; CHECK-NEXT: vst.push.576 ex2, [p2, sf, r26] // Delay Slot 4 | ||
| ; CHECK-NEXT: vst.flush.512 [p2, sf, r26] // Delay Slot 3 | ||
| ; CHECK-NEXT: nop // Delay Slot 2 | ||
| ; CHECK-NEXT: nop // Delay Slot 1 | ||
| entry: | ||
| br label %bb.exit | ||
|
|
||
| bb.exit: ; preds = %entry | ||
| %I.7 = call { ptr, <32 x i32>, i32 } @llvm.aie2p.fifo.ld.fill(ptr %in, <32 x i32> zeroinitializer, i32 0) | ||
| %I.8 = extractvalue { ptr, <32 x i32>, i32 } %I.7, 0 | ||
| %I.9 = extractvalue { ptr, <32 x i32>, i32 } %I.7, 1 | ||
| %I.10 = extractvalue { ptr, <32 x i32>, i32 } %I.7, 2 | ||
| %I.11 = tail call { <64 x i8>, ptr, <32 x i32>, i32 } @llvm.aie2p.fifo.ld.pop.unaligned(ptr %I.8, <32 x i32> %I.9, i32 %I.10) | ||
| %I.12 = extractvalue { <64 x i8>, ptr, <32 x i32>, i32 } %I.11, 1 | ||
| %I.13 = extractvalue { <64 x i8>, ptr, <32 x i32>, i32 } %I.11, 2 | ||
| %I.14 = extractvalue { <64 x i8>, ptr, <32 x i32>, i32 } %I.11, 3 | ||
| %I.15 = extractvalue { <64 x i8>, ptr, <32 x i32>, i32 } %I.11, 0 | ||
| %vec = or i64 0, 0 | ||
| %i8vec = bitcast i64 %vec to <8 x i8> | ||
| %0 = call { ptr, <32 x i32>, i32 } @llvm.aie2p.fifo.st.push.576.bfp16(ptr %out, <64 x i8> zeroinitializer, <8 x i8> %i8vec, <32 x i32> zeroinitializer, i32 0) | ||
| %1 = extractvalue { ptr, <32 x i32>, i32 } %0, 0 | ||
| %2 = extractvalue { ptr, <32 x i32>, i32 } %0, 1 | ||
| %3 = extractvalue { ptr, <32 x i32>, i32 } %0, 2 | ||
| %4 = call { ptr, <32 x i32>, i32 } @llvm.aie2p.fifo.st.flush(ptr %1, <32 x i32> %2, i32 %3) | ||
| %5 = extractvalue { ptr, <32 x i32>, i32 } %4, 0 | ||
| %6 = extractvalue { ptr, <32 x i32>, i32 } %4, 1 | ||
| %7 = extractvalue { ptr, <32 x i32>, i32 } %4, 2 | ||
| ret void | ||
| } | ||
|
|
||
| ; Function Attrs: nounwind memory(argmem: read) | ||
| declare { ptr, <32 x i32>, i32 } @llvm.aie2p.fifo.ld.fill(ptr, <32 x i32>, i32) #1 | ||
|
|
||
| ; Function Attrs: nounwind memory(none) | ||
| declare <32 x float> @llvm.aie2p.v32bf16.to.v32accfloat(<32 x bfloat>) #2 | ||
|
|
||
| ; Function Attrs: nounwind memory(argmem: read) | ||
| declare { <64 x i8>, ptr, <32 x i32>, i32 } @llvm.aie2p.fifo.ld.pop.unaligned(ptr, <32 x i32>, i32) #1 | ||
|
|
||
| ; Function Attrs: nounwind memory(argmem: read) | ||
| declare { <64 x i8>, ptr, <32 x i32>, i32, i20 } @llvm.aie2p.fifo.ld.pop.2d.unaligned(ptr, <32 x i32>, i32, i20, i20, i20, i20) #1 | ||
|
|
||
| ; Function Attrs: nounwind memory(argmem: write) | ||
| declare { ptr, <32 x i32>, i32 } @llvm.aie2p.fifo.st.push.576.bfp16(ptr, <64 x i8>, <8 x i8>, <32 x i32>, i32) #3 | ||
|
|
||
| ; Function Attrs: nounwind memory(inaccessiblemem: read) | ||
| declare { <64 x i8>, <8 x i8> } @llvm.aie2p.v64accfloat.to.v64bfp16ebs8(<64 x i8>, <8 x i8>, <64 x float>) #4 | ||
|
|
||
| ; Function Attrs: nounwind memory(argmem: write) | ||
| declare { ptr, <32 x i32>, i32 } @llvm.aie2p.fifo.st.flush(ptr, <32 x i32>, i32) #3 | ||
|
|
||
| ; Function Attrs: nounwind memory(argmem: write) | ||
| declare { ptr, <32 x i32>, i32, i20 } @llvm.aie2p.fifo.st.flush.2d(ptr, <32 x i32>, i32, i20, i20, i20, i20) #3 | ||
|
|
||
| attributes #0 = { mustprogress noinline "no-jump-tables"="true" "no-trapping-math"="true" "stack-protector-buffer-size"="8" } | ||
| attributes #1 = { nounwind memory(argmem: read) } | ||
| attributes #2 = { nounwind memory(none) } | ||
| attributes #3 = { nounwind memory(argmem: write) } | ||
| attributes #4 = { nounwind memory(inaccessiblemem: read) } | ||
|
|
||
| !llvm.linker.options = !{} | ||
| !llvm.module.flags = !{!0} | ||
| !llvm.ident = !{!1} | ||
|
|
||
| !0 = !{i32 1, !"wchar_size", i32 4} | ||
| !1 = !{!"clang version 19.0.0git (/scratch/llvm-aie/clang 5c68f0e173ec03a2b4f908018663ca52923e52fb)"} | ||
|
|
||
| ... | ||
| --- | ||
| name: concurrentLoadStore | ||
| alignment: 16 | ||
| exposesReturnsTwice: false | ||
| legalized: true | ||
| regBankSelected: true | ||
| selected: true | ||
| failedISel: false | ||
| tracksRegLiveness: true | ||
| hasWinCFI: false | ||
| callsEHReturn: false | ||
| callsUnwindInit: false | ||
| hasEHCatchret: false | ||
| hasEHScopes: false | ||
| hasEHFunclets: false | ||
| isOutlined: false | ||
| debugInstrRef: false | ||
| failsVerification: false | ||
| tracksDebugUserValues: false | ||
| registers: | ||
| - { id: 0, class: eps, preferred-register: '' } | ||
| - { id: 1, class: mpfs, preferred-register: '' } | ||
| - { id: 2, class: _, preferred-register: '' } | ||
| - { id: 3, class: eps, preferred-register: '' } | ||
| - { id: 4, class: eldfiforeg, preferred-register: '' } | ||
| - { id: 5, class: erf2, preferred-register: '' } | ||
| - { id: 6, class: vec1024, preferred-register: '' } | ||
| - { id: 7, class: mr26_fifo_st, preferred-register: '' } | ||
| - { id: 8, class: vec512, preferred-register: '' } | ||
| - { id: 9, class: eps, preferred-register: '' } | ||
| - { id: 10, class: eldfiforeg, preferred-register: '' } | ||
| - { id: 11, class: erf2, preferred-register: '' } | ||
| - { id: 12, class: _, preferred-register: '' } | ||
| - { id: 13, class: el_with_sub_l_even_in_er26, preferred-register: '' } | ||
| - { id: 14, class: el, preferred-register: '' } | ||
| - { id: 15, class: mpfs, preferred-register: '' } | ||
| - { id: 16, class: mstfifo, preferred-register: '' } | ||
| - { id: 17, class: mr26_fifo_st, preferred-register: '' } | ||
| - { id: 18, class: vec512, preferred-register: '' } | ||
| - { id: 19, class: _, preferred-register: '' } | ||
| - { id: 20, class: ep_as_32bit_and_mpfs, preferred-register: '' } | ||
| - { id: 21, class: mstfifo, preferred-register: '' } | ||
| - { id: 22, class: mr26_fifo_st, preferred-register: '' } | ||
| - { id: 23, class: exe, preferred-register: '' } | ||
| - { id: 24, class: _, preferred-register: '' } | ||
| - { id: 25, class: er, preferred-register: '' } | ||
| - { id: 26, class: eldfiforeg, preferred-register: '' } | ||
| - { id: 27, class: mstfifo, preferred-register: '' } | ||
| - { id: 28, class: vec576, preferred-register: '' } | ||
| - { id: 29, class: erf2, preferred-register: '' } | ||
| - { id: 30, class: exo, preferred-register: '' } | ||
| - { id: 31, class: epsrfldf, preferred-register: '' } | ||
| - { id: 32, class: epsrfldf, preferred-register: '' } | ||
| liveins: [] | ||
| calleeSavedRegisters: [ '$lr', '$r8', '$r9', '$r10', '$r11', '$r12', '$r13', | ||
| '$r14', '$r15', '$p6', '$p7' ] | ||
| frameInfo: | ||
| isFrameAddressTaken: false | ||
| isReturnAddressTaken: false | ||
| hasStackMap: false | ||
| hasPatchPoint: false | ||
| stackSize: 0 | ||
| offsetAdjustment: 0 | ||
| maxAlignment: 1 | ||
| adjustsStack: false | ||
| hasCalls: false | ||
| stackProtector: '' | ||
| functionContext: '' | ||
| maxCallFrameSize: 4294967295 | ||
| cvBytesOfCalleeSavedRegisters: 0 | ||
| hasOpaqueSPAdjustment: false | ||
| hasVAStart: false | ||
| hasMustTailInVarArgFunc: false | ||
| hasTailCall: false | ||
| localFrameSize: 0 | ||
| savePoint: '' | ||
| restorePoint: '' | ||
| fixedStack: [] | ||
| stack: [] | ||
| entry_values: [] | ||
| callSites: [] | ||
| debugValueSubstitutions: [] | ||
| constants: [] | ||
| machineFunctionInfo: | ||
| varArgsFrameIndex: 0 | ||
| body: | | ||
| bb.0.entry: | ||
| successors: %bb.1(0x80000000) | ||
| liveins: $p0, $p1, $p2 | ||
| undef %32.sub_ptr:epsrfldf = COPY $p0 | ||
| %20:ep_as_32bit_and_mpfs = COPY $p1 | ||
| undef %13.sub_l_even:el_with_sub_l_even_in_er26 = MOV_RLC_imm11_pseudo 0 | ||
| undef %6.sub_512_lo:vec1024 = VBCST_32 %13.sub_l_even | ||
| %6.sub_512_hi:vec1024 = COPY %6.sub_512_lo | ||
| undef %28.sub_bfp16_x:vec576 = VBCST_8 %13.sub_l_even | ||
| bb.1.bb.exit: | ||
| %32.sub_fifo:epsrfldf = COPY %6 | ||
| %32.sub_avail:epsrfldf = COPY %13.sub_l_even | ||
| %32.sub_ptr:epsrfldf, %32.sub_fifo:epsrfldf, %32.sub_avail:epsrfldf = VLDA_FILL_512 %32.sub_ptr, %32.sub_fifo, %32.sub_avail | ||
| dead %8:vec512, dead %32.sub_ptr:epsrfldf, dead %32.sub_fifo:epsrfldf, dead %32.sub_avail:epsrfldf = VLDB_POP_512_normal_pop %32.sub_ptr, %32.sub_fifo, %32.sub_avail, implicit-def $srfifo_uf | ||
| %13.sub_l_odd:el_with_sub_l_even_in_er26 = COPY %13.sub_l_even | ||
| %21:mstfifo = COPY %6 | ||
| %28.sub_bfp16_e:vec576 = COPY %13 | ||
| %21:mstfifo, %20:ep_as_32bit_and_mpfs, undef %13.sub_l_even:el_with_sub_l_even_in_er26 = VST_PUSH_576 %21, %28, %20, %13.sub_l_even, implicit-def $srfifo_of | ||
| dead %21:mstfifo, dead %20:ep_as_32bit_and_mpfs, dead %13.sub_l_even:el_with_sub_l_even_in_er26 = VST_FLUSH_512_normal_flush %21, %20, %13.sub_l_even, implicit-def $srfifo_of | ||
| PseudoRET implicit $lr | ||
| ... |
Oops, something went wrong.