[MC][RISCV][LoongArch] Add AlignFragment size if layout is available and not need insert nops (llvm#76552)

Due to delayed decision for ADD/SUB relocations, RISCV and LoongArch may
go slow fragment walk path with available layout. When RISCV (or
LoongArch in the future) don't need insert nops, that means relax is
disabled. With available layout and not needing insert nops, the size of
AlignFragment should be a constant. So we can add it to Displacement for
folding A-B.

Author: MQ-mengqing

llvm/lib/MC/MCExpr.cpp

@@ -704,8 +704,14 @@ static void AttemptToFoldSymbolOffsetDifference(
       }
 
       int64_t Num;
+      unsigned Count;
       if (DF) {
         Displacement += DF->getContents().size();
+      } else if (auto *AF = dyn_cast<MCAlignFragment>(FI);
+                 AF && Layout &&
+                 !Asm->getBackend().shouldInsertExtraNopBytesForCodeAlign(
+                     *AF, Count)) {
+        Displacement += Asm->computeFragmentSize(*Layout, *AF);
       } else if (auto *FF = dyn_cast<MCFillFragment>(FI);
                  FF && FF->getNumValues().evaluateAsAbsolute(Num)) {
         Displacement += Num * FF->getValueSize();

llvm/test/MC/LoongArch/Misc/cfi-advance.s

@@ -0,0 +1,27 @@
+# RUN: llvm-mc --filetype=obj --triple=loongarch64 -mattr=-relax %s -o %t.o
+# RUN: llvm-readobj -r %t.o | FileCheck --check-prefix=RELOC %s
+# RUN: llvm-dwarfdump --debug-frame %t.o | FileCheck --check-prefix=DWARFDUMP %s
+
+# RELOC:       Relocations [
+# RELOC-NEXT:    .rela.eh_frame {
+# RELOC-NEXT:       0x1C R_LARCH_32_PCREL .text 0x0
+# RELOC-NEXT:    }
+# RELOC-NEXT:  ]
+# DWARFDUMP:       DW_CFA_advance_loc: 4
+# DWARFDUMP-NEXT:  DW_CFA_def_cfa_offset: +8
+# DWARFDUMP-NEXT:  DW_CFA_advance_loc: 8
+# DWARFDUMP-NEXT:  DW_CFA_def_cfa_offset: +8
+
+        .text
+        .globl test
+        .p2align 2
+        .type   test,@function
+test:
+        .cfi_startproc
+        nop
+        .cfi_def_cfa_offset 8
+        .p2align 3
+        nop
+        .cfi_def_cfa_offset 8
+        nop
+        .cfi_endproc

llvm/test/MC/LoongArch/Relocations/relax-addsub.s

@@ -23,14 +23,6 @@
 # RELAX-NEXT:      0x14 R_LARCH_RELAX - 0x0
 # RELAX-NEXT:    }
 # RELAX-NEXT:    Section ({{.*}}) .rela.data {
-# RELAX-NEXT:      0xF R_LARCH_ADD8 .L3 0x0
-# RELAX-NEXT:      0xF R_LARCH_SUB8 .L2 0x0
-# RELAX-NEXT:      0x10 R_LARCH_ADD16 .L3 0x0
-# RELAX-NEXT:      0x10 R_LARCH_SUB16 .L2 0x0
-# RELAX-NEXT:      0x12 R_LARCH_ADD32 .L3 0x0
-# RELAX-NEXT:      0x12 R_LARCH_SUB32 .L2 0x0
-# RELAX-NEXT:      0x16 R_LARCH_ADD64 .L3 0x0
-# RELAX-NEXT:      0x16 R_LARCH_SUB64 .L2 0x0
 # RELAX-NEXT:      0x1E R_LARCH_ADD8 .L4 0x0
 # RELAX-NEXT:      0x1E R_LARCH_SUB8 .L3 0x0
 # RELAX-NEXT:      0x1F R_LARCH_ADD16 .L4 0x0
@@ -43,8 +35,8 @@
 # RELAX-NEXT:  ]
 
 # RELAX:      Hex dump of section '.data':
-# RELAX-NEXT: 0x00000000 04040004 00000004 00000000 00000000
-# RELAX-NEXT: 0x00000010 00000000 00000000 00000000 00000000
+# RELAX-NEXT: 0x00000000 04040004 00000004 00000000 0000000c
+# RELAX-NEXT: 0x00000010 0c000c00 00000c00 00000000 00000000
 # RELAX-NEXT: 0x00000020 00000000 00000000 00000000 00
 
 .text
@@ -63,13 +55,12 @@
 .short .L2 - .L1
 .word  .L2 - .L1
 .dword .L2 - .L1
-## With relaxation, emit relocs because of the .align making the diff variable.
-## TODO Handle alignment directive. Why they emit relocs now? They returns
-## without folding symbols offset in AttemptToFoldSymbolOffsetDifference().
+## TODO Handle alignment directive.
 .byte  .L3 - .L2
 .short .L3 - .L2
 .word  .L3 - .L2
 .dword .L3 - .L2
+## With relaxation, emit relocs because the la.pcrel makes the diff variable.
 .byte  .L4 - .L3
 .short .L4 - .L3
 .word  .L4 - .L3

llvm/test/MC/RISCV/cfi-advance.s

@@ -5,12 +5,16 @@
 
 # CHECK:      .rela.eh_frame {
 # CHECK-NEXT:   0x1C R_RISCV_32_PCREL <null> 0x0
+# CHECK-NEXT:   0x35 R_RISCV_SET6 <null> 0x0
+# CHECK-NEXT:   0x35 R_RISCV_SUB6 <null> 0x0
 # CHECK-NEXT: }
 # CHECK-DWARFDUMP: DW_CFA_advance_loc1: 104
 # CHECK-DWARFDUMP-NEXT: DW_CFA_def_cfa_offset: +8
 # CHECK-DWARFDUMP-NEXT: DW_CFA_advance_loc2: 259
 # CHECK-DWARFDUMP-NEXT: DW_CFA_def_cfa_offset: +8
 # CHECK-DWARFDUMP-NEXT: DW_CFA_advance_loc4: 65539
+# CHECK-DWARFDUMP-NEXT: DW_CFA_def_cfa_offset: +8
+# CHECK-DWARFDUMP-NEXT: DW_CFA_advance_loc: 10
 # CHECK-DWARFDUMP-NEXT: DW_CFA_def_cfa_offset: +8
         .text
         .globl  test                            # -- Begin function test
@@ -28,4 +32,7 @@ test:
         .zero 65535, 0x90
         .cfi_def_cfa_offset 8
         nop
+        .p2align 3
+        .cfi_def_cfa_offset 8
+        nop
         .cfi_endproc

llvm/test/MC/RISCV/fixups-expr.s

@@ -16,11 +16,15 @@
 
 .globl G1
 .globl G2
+.globl G3
 .L1:
 G1:
   call extern
 .L2:
 G2:
+  .p2align 3
+.L3:
+G3:
 
 .data
 .dword .L2-.L1
@@ -31,6 +35,14 @@ G2:
 .half G2-G1
 .byte .L2-.L1
 .byte G2-G1
+.dword .L3-.L2
+.dword G3-G2
+.word .L3-.L2
+.word G3-G2
+.half .L3-.L2
+.half G3-G2
+.byte .L3-.L2
+.byte G3-G2
 # RELAX:      .rela.data {
 # RELAX-NEXT:   0x0 R_RISCV_ADD64 .L2 0x0
 # RELAX-NEXT:   0x0 R_RISCV_SUB64 .L1 0x0
@@ -48,4 +60,20 @@ G2:
 # RELAX-NEXT:   0x1C R_RISCV_SUB8 .L1 0x0
 # RELAX-NEXT:   0x1D R_RISCV_ADD8 G2 0x0
 # RELAX-NEXT:   0x1D R_RISCV_SUB8 G1 0x0
+# RELAX-NEXT:   0x1E R_RISCV_ADD64 .L3 0x0
+# RELAX-NEXT:   0x1E R_RISCV_SUB64 .L2 0x0
+# RELAX-NEXT:   0x26 R_RISCV_ADD64 G3 0x0
+# RELAX-NEXT:   0x26 R_RISCV_SUB64 G2 0x0
+# RELAX-NEXT:   0x2E R_RISCV_ADD32 .L3 0x0
+# RELAX-NEXT:   0x2E R_RISCV_SUB32 .L2 0x0
+# RELAX-NEXT:   0x32 R_RISCV_ADD32 G3 0x0
+# RELAX-NEXT:   0x32 R_RISCV_SUB32 G2 0x0
+# RELAX-NEXT:   0x36 R_RISCV_ADD16 .L3 0x0
+# RELAX-NEXT:   0x36 R_RISCV_SUB16 .L2 0x0
+# RELAX-NEXT:   0x38 R_RISCV_ADD16 G3 0x0
+# RELAX-NEXT:   0x38 R_RISCV_SUB16 G2 0x0
+# RELAX-NEXT:   0x3A R_RISCV_ADD8 .L3 0x0
+# RELAX-NEXT:   0x3A R_RISCV_SUB8 .L2 0x0
+# RELAX-NEXT:   0x3B R_RISCV_ADD8 G3 0x0
+# RELAX-NEXT:   0x3B R_RISCV_SUB8 G2 0x0
 # RELAX-NEXT: }