[X86] [CodeView] Add codeview mappings for registers ST0-ST7

[mstorsjo]

Summary

These can end up needed after https://reviews.llvm.org/D116821.

Suggested by @aganea.

This doesn't come with any tests right now - do we need them here, and
are there any suggestion on what kind that should be? (The issue
that originally triggered this probably doesn't reproduce hitting
these register types, as that feature has been disabled/reverted for
now.)

[aganea]

The feature can still be enabled with -mllvm -experimental-debug-variable-locations=true, so D116821 wasn't completely reverted.

Why not adding your initial bug repro as a test, along with the flags above on the cmd line? It also looks like that prior mappings were added without test coverage.

[rnk]

lgtm

I think the plan is to reland the feature, so we'll have test coverage for this soon enough.

[mstorsjo]

Add a testcase.

[mstorsjo]

In D118072#3267462, @aganea wrote:

The feature can still be enabled with -mllvm -experimental-debug-variable-locations=true, so D116821 wasn't completely reverted.

Why not adding your initial bug repro as a test, along with the flags above on the cmd line? It also looks like that prior mappings were added without test coverage.

Ok, I made such a testcase. I'm a bit afraid that it might be a little brittle in the long run though...

[aganea]

In D118072#3267627, @mstorsjo wrote:

In D118072#3267462, @aganea wrote:

The feature can still be enabled with -mllvm -experimental-debug-variable-locations=true, so D116821 wasn't completely reverted.
(l
Why not adding your initial bug repro as a test, along with the flags above on the cmd line? It also looks like that prior mappings were added without test coverage.

Ok, I made such a testcase. I'm a bit afraid that it might be a little brittle in the long run though...

I was simply surprised that a such simple testcase wasn't catched elsewhere. Could you please explain 'brittle'? Do you think the outcome of this test isn't stable in the long run?
In any case LGTM.
I'll let the choice (to add a test or not) to your discretion ;-) I felt that it could make you loose time again later down the line...

In D118072#3267599, @rnk wrote:

I think the plan is to reland the feature, so we'll have test coverage for this soon enough.

I was under the impression that the discussions in D11681 were revolving around resources not correctness?

[mstorsjo]

In D118072#3267788, @aganea wrote:

In D118072#3267627, @mstorsjo wrote:

In D118072#3267462, @aganea wrote:

The feature can still be enabled with -mllvm -experimental-debug-variable-locations=true, so D116821 wasn't completely reverted.
(l
Why not adding your initial bug repro as a test, along with the flags above on the cmd line? It also looks like that prior mappings were added without test coverage.

Ok, I made such a testcase. I'm a bit afraid that it might be a little brittle in the long run though...

I was simply surprised that a such simple testcase wasn't catched elsewhere. Could you please explain 'brittle'? Do you think the outcome of this test isn't stable in the long run?
In any case LGTM.
I'll let the choice (to add a test or not) to your discretion ;-) I felt that it could make you loose time again later down the line...

Brittle, in the sense that it depends on all the internal machinery, whether this input piece of code actually produces the debug info we want to test. (Up until recently, it didn't, and now it does with a flag only.) I mean, it would be more robust if I'd take the intermediate output from a stage when everything has been settled about what debug info to produce, and then only produce codeview debug info out of it. I.e. at a point where the input specifically says "debug info for X86::ST0". But then on the other hand, that test would pretty much just test whether X86::ST0 is include in the table too.

So I guess this is fine - and in case the test causes problems later, we can reconsider if it has to be kept or not.

llvm/lib/Target/X86/MCTargetDesc/X86MCTargetDesc.cpp

MCPhysReg Reg;
       {codeview::RegisterId::EBX, X86::EBX},
       {codeview::RegisterId::ESP, X86::ESP},
       {codeview::RegisterId::EBP, X86::EBP},
       {codeview::RegisterId::ESI, X86::ESI},
       {codeview::RegisterId::EDI, X86::EDI},
 
       {codeview::RegisterId::EFLAGS, X86::EFLAGS},
 
+      {codeview::RegisterId::ST0, X86::ST0},
+      {codeview::RegisterId::ST1, X86::ST1},
+      {codeview::RegisterId::ST2, X86::ST2},
+      {codeview::RegisterId::ST3, X86::ST3},
+      {codeview::RegisterId::ST4, X86::ST4},
+      {codeview::RegisterId::ST5, X86::ST5},
+      {codeview::RegisterId::ST6, X86::ST6},
+      {codeview::RegisterId::ST7, X86::ST7},
+
       {codeview::RegisterId::ST0, X86::FP0},
       {codeview::RegisterId::ST1, X86::FP1},
       {codeview::RegisterId::ST2, X86::FP2},
       {codeview::RegisterId::ST3, X86::FP3},
       {codeview::RegisterId::ST4, X86::FP4},
       {codeview::RegisterId::ST5, X86::FP5},
       {codeview::RegisterId::ST6, X86::FP6},
       {codeview::RegisterId::ST7, X86::FP7},

llvm/test/DebugInfo/COFF/x87-registers.ll

+; RUN: llc -experimental-debug-variable-locations=true < %s -filetype=obj | llvm-readobj - --codeview | FileCheck %s
+
+; CHECK:      DefRangeRegisterSym {
+; CHECK-NEXT:   Kind: S_DEFRANGE_REGISTER
+; CHECK-NEXT:   Register: ST0
+
+target datalayout = "e-m:w-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-w64-windows-gnu"
+
+define i32 @a() !dbg !8 {
+entry:
+  call void @llvm.dbg.declare(metadata [6 x i8]* undef, metadata !13, metadata !DIExpression(DW_OP_LLVM_fragment, 80, 48)), !dbg !15
+  %0 = tail call x86_fp80 asm sideeffect "", "={st},~{dirflag},~{fpsr},~{flags}"(), !dbg !16, !srcloc !17
+  call void @llvm.dbg.value(metadata x86_fp80 %0, metadata !13, metadata !DIExpression()), !dbg !18
+  ret i32 undef, !dbg !19
+}
+
+declare void @llvm.dbg.declare(metadata, metadata, metadata)
+declare void @llvm.dbg.value(metadata, metadata, metadata)
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!2, !3, !4, !5, !6}
+!llvm.ident = !{!7}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 14.0.0", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, splitDebugInlining: false, nameTableKind: None)
+!1 = !DIFile(filename: "codeview-fp0.c", directory: "llvm/build")
+!2 = !{i32 2, !"CodeView", i32 1}
+!3 = !{i32 2, !"Debug Info Version", i32 3}
+!4 = !{i32 1, !"wchar_size", i32 2}
+!5 = !{i32 7, !"PIC Level", i32 2}
+!6 = !{i32 7, !"uwtable", i32 1}
+!7 = !{!"clang version 14.0.0"}
+!8 = distinct !DISubprogram(name: "a", scope: !1, file: !1, line: 1, type: !9, scopeLine: 1, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !12)
+!9 = !DISubroutineType(types: !10)
+!10 = !{!11}
+!11 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!12 = !{!13}
+!13 = !DILocalVariable(name: "b", scope: !8, file: !1, line: 2, type: !14)
+!14 = !DIBasicType(name: "long double", size: 128, encoding: DW_ATE_float)
+!15 = !DILocation(line: 2, scope: !8)
+!16 = !DILocation(line: 3, scope: !8)
+!17 = !{i64 40}
+!18 = !DILocation(line: 0, scope: !8)
+!19 = !DILocation(line: 4, scope: !8)