[PPC] Replace PPCMergeStringPool with GlobalMerge for Linux

Author: syzaara

llvm/lib/CodeGen/GlobalMerge.cpp

@@ -79,6 +79,7 @@
 #include "llvm/IR/GlobalValue.h"
 #include "llvm/IR/GlobalVariable.h"
 #include "llvm/IR/Instruction.h"
+#include "llvm/IR/IntrinsicInst.h"
 #include "llvm/IR/Module.h"
 #include "llvm/IR/Type.h"
 #include "llvm/IR/Use.h"
@@ -713,6 +714,24 @@ bool GlobalMergeImpl::run(Module &M) {
     // Ignore all "required" globals:
     if (isMustKeepGlobalVariable(&GV))
       continue;
+    auto checkUsers = [] (const GlobalVariable *GV) {
+      for (const User *CurrentUser : GV->users()) {
+        if (auto *I = dyn_cast<Instruction>(CurrentUser)) {
+          // Do not merge globals in exception pads.
+          if (I->isEHPad())
+            return false;
+          if (auto *II = dyn_cast<IntrinsicInst>(I)) {
+            // Some intrinsics require a plain global.
+            if (II->getIntrinsicID() == Intrinsic::eh_typeid_for)
+              return false;
+          }
+        }
+      }
+      return true;
+    };
+
+    if (!checkUsers(&GV))
+      continue;
 
     // Don't merge tagged globals, as each global should have its own unique
     // memory tag at runtime. TODO(hctim): This can be relaxed: constant globals

llvm/lib/Target/PowerPC/PPCTargetMachine.cpp

@@ -503,17 +503,10 @@ bool PPCPassConfig::addPreISel() {
   // Specifying the command line option overrides the AIX default.
   if ((EnableGlobalMerge.getNumOccurrences() > 0)
           ? EnableGlobalMerge
-          : (TM->getTargetTriple().isOSAIX() &&
-             getOptLevel() != CodeGenOptLevel::None))
+          : getOptLevel() != CodeGenOptLevel::None)
     addPass(createGlobalMergePass(TM, GlobalMergeMaxOffset, false, false, true,
                                   true));
 
-  if ((MergeStringPool.getNumOccurrences() > 0)
-          ? MergeStringPool
-          : (TM->getTargetTriple().isOSLinux() &&
-             getOptLevel() != CodeGenOptLevel::None))
-    addPass(createPPCMergeStringPoolPass());
-
   if (!DisableInstrFormPrep && getOptLevel() != CodeGenOptLevel::None)
     addPass(createPPCLoopInstrFormPrepPass(getPPCTargetMachine()));
 

llvm/test/CodeGen/PowerPC/O3-pipeline.ll

@@ -70,6 +70,7 @@
 ; CHECK-NEXT:       CodeGen Prepare
 ; CHECK-NEXT:       Dominator Tree Construction
 ; CHECK-NEXT:       Exception handling preparation
+; CHECK-NEXT:       Merge internal globals
 ; CHECK-NEXT:       Natural Loop Information
 ; CHECK-NEXT:       Scalar Evolution Analysis
 ; CHECK-NEXT:       Prepare loop for ppc preferred instruction forms

llvm/test/CodeGen/PowerPC/PR35812-neg-cmpxchg.ll

@@ -39,21 +39,21 @@ define signext i32 @main() nounwind {
 ; CHECK-NEXT:    cmplwi 3, 234
 ; CHECK-NEXT:    bne 0, .LBB0_7
 ; CHECK-NEXT:  # %bb.5: # %L.B0001
-; CHECK-NEXT:    addis 3, 2, .L__ModuleStringPool@toc@ha
-; CHECK-NEXT:    addi 3, 3, .L__ModuleStringPool@toc@l
+; CHECK-NEXT:    addis 3, 2, .L_MergedGlobals@toc@ha
+; CHECK-NEXT:    addi 3, 3, .L_MergedGlobals@toc@l
 ; CHECK-NEXT:    bl puts
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    li 3, 0
 ; CHECK-NEXT:    b .LBB0_9
 ; CHECK-NEXT:  .LBB0_6: # %L.B0003
-; CHECK-NEXT:    addis 3, 2, .L__ModuleStringPool@toc@ha
-; CHECK-NEXT:    addi 3, 3, .L__ModuleStringPool@toc@l
-; CHECK-NEXT:    addi 3, 3, 7
+; CHECK-NEXT:    addis 3, 2, .L_MergedGlobals@toc@ha
+; CHECK-NEXT:    addi 3, 3, .L_MergedGlobals@toc@l
+; CHECK-NEXT:    addi 3, 3, 16
 ; CHECK-NEXT:    b .LBB0_8
 ; CHECK-NEXT:  .LBB0_7: # %L.B0005
-; CHECK-NEXT:    addis 3, 2, .L__ModuleStringPool@toc@ha
-; CHECK-NEXT:    addi 3, 3, .L__ModuleStringPool@toc@l
-; CHECK-NEXT:    addi 3, 3, 53
+; CHECK-NEXT:    addis 3, 2, .L_MergedGlobals@toc@ha
+; CHECK-NEXT:    addi 3, 3, .L_MergedGlobals@toc@l
+; CHECK-NEXT:    addi 3, 3, 64
 ; CHECK-NEXT:  .LBB0_8: # %L.B0003
 ; CHECK-NEXT:    bl puts
 ; CHECK-NEXT:    nop
@@ -107,21 +107,21 @@ define signext i32 @main() nounwind {
 ; CHECK-P7-NEXT:    cmplwi 3, 234
 ; CHECK-P7-NEXT:    bne 0, .LBB0_7
 ; CHECK-P7-NEXT:  # %bb.5: # %L.B0001
-; CHECK-P7-NEXT:    addis 3, 2, .L__ModuleStringPool@toc@ha
-; CHECK-P7-NEXT:    addi 3, 3, .L__ModuleStringPool@toc@l
+; CHECK-P7-NEXT:    addis 3, 2, .L_MergedGlobals@toc@ha
+; CHECK-P7-NEXT:    addi 3, 3, .L_MergedGlobals@toc@l
 ; CHECK-P7-NEXT:    bl puts
 ; CHECK-P7-NEXT:    nop
 ; CHECK-P7-NEXT:    li 3, 0
 ; CHECK-P7-NEXT:    b .LBB0_9
 ; CHECK-P7-NEXT:  .LBB0_6: # %L.B0003
-; CHECK-P7-NEXT:    addis 3, 2, .L__ModuleStringPool@toc@ha
-; CHECK-P7-NEXT:    addi 3, 3, .L__ModuleStringPool@toc@l
-; CHECK-P7-NEXT:    addi 3, 3, 7
+; CHECK-P7-NEXT:    addis 3, 2, .L_MergedGlobals@toc@ha
+; CHECK-P7-NEXT:    addi 3, 3, .L_MergedGlobals@toc@l
+; CHECK-P7-NEXT:    addi 3, 3, 16
 ; CHECK-P7-NEXT:    b .LBB0_8
 ; CHECK-P7-NEXT:  .LBB0_7: # %L.B0005
-; CHECK-P7-NEXT:    addis 3, 2, .L__ModuleStringPool@toc@ha
-; CHECK-P7-NEXT:    addi 3, 3, .L__ModuleStringPool@toc@l
-; CHECK-P7-NEXT:    addi 3, 3, 53
+; CHECK-P7-NEXT:    addis 3, 2, .L_MergedGlobals@toc@ha
+; CHECK-P7-NEXT:    addi 3, 3, .L_MergedGlobals@toc@l
+; CHECK-P7-NEXT:    addi 3, 3, 64
 ; CHECK-P7-NEXT:  .LBB0_8: # %L.B0003
 ; CHECK-P7-NEXT:    bl puts
 ; CHECK-P7-NEXT:    nop

llvm/test/CodeGen/PowerPC/aix-xcoff-used-with-stringpool.ll

@@ -2,11 +2,11 @@
 ;; in llvm.used or in llvm.compiler.used.
 
 ; RUN: llc -verify-machineinstrs -mcpu=pwr8 -mtriple powerpc-ibm-aix-xcoff \
-; RUN:   -ppc-merge-string-pool=true -global-merge-all-const=false -data-sections=false < %s | \
+; RUN:   -data-sections=false < %s | \
 ; RUN:   FileCheck %s
 
 ; RUN: llc -verify-machineinstrs -mcpu=pwr8 -mtriple powerpc64-ibm-aix-xcoff \
-; RUN:   -ppc-merge-string-pool=true -global-merge-all-const=false -data-sections=false < %s | \
+; RUN:   -data-sections=false < %s | \
 ; RUN:   FileCheck %s
 
 @keep_this = internal constant [5 x i8] c"keep1", align 1
@@ -53,7 +53,7 @@ entry:
 ; CHECK:  keep_this:
 ; CHECK:    .lglobl keep_this2
 ; CHECK:  keep_this2:
-; CHECK:  L..__ModuleStringPool:
+; CHECK:  L.._MergedGlobals:
 ; CHECK:    .string "str1_STRING"
 ; CHECK:    .string "str2_STRING"
 ; CHECK:    .string "str3_STRING"

llvm/test/CodeGen/PowerPC/ctrloop-fp128.ll

@@ -9,24 +9,17 @@
 define void @fmul_ctrloop_fp128() nounwind {
 ; PWR9-LABEL: fmul_ctrloop_fp128:
 ; PWR9:       # %bb.0: # %entry
-; PWR9-NEXT:    addis 3, 2, a@toc@ha
+; PWR9-NEXT:    addis 3, 2, .L_MergedGlobals@toc@ha
 ; PWR9-NEXT:    li 4, 4
-; PWR9-NEXT:    addi 3, 3, a@toc@l
+; PWR9-NEXT:    addi 3, 3, .L_MergedGlobals@toc@l
 ; PWR9-NEXT:    lxv 34, 0(3)
-; PWR9-NEXT:    addis 3, 2, y@toc@ha
 ; PWR9-NEXT:    mtctr 4
-; PWR9-NEXT:    addis 4, 2, x@toc@ha
-; PWR9-NEXT:    addi 3, 3, y@toc@l
-; PWR9-NEXT:    addi 4, 4, x@toc@l
-; PWR9-NEXT:    addi 3, 3, -16
-; PWR9-NEXT:    addi 4, 4, -16
 ; PWR9-NEXT:    .p2align 5
 ; PWR9-NEXT:  .LBB0_1: # %for.body
 ; PWR9-NEXT:    #
-; PWR9-NEXT:    lxv 35, 16(4)
-; PWR9-NEXT:    addi 4, 4, 16
+; PWR9-NEXT:    lxv 35, 16(3)
 ; PWR9-NEXT:    xsmulqp 3, 2, 3
-; PWR9-NEXT:    stxv 35, 16(3)
+; PWR9-NEXT:    stxv 35, 80(3)
 ; PWR9-NEXT:    addi 3, 3, 16
 ; PWR9-NEXT:    bdnz .LBB0_1
 ; PWR9-NEXT:  # %bb.2: # %for.end
@@ -38,40 +31,33 @@ define void @fmul_ctrloop_fp128() nounwind {
 ; PWR8-NEXT:    stdu 1, -112(1)
 ; PWR8-NEXT:    li 3, 48
 ; PWR8-NEXT:    std 0, 128(1)
+; PWR8-NEXT:    std 30, 96(1) # 8-byte Folded Spill
+; PWR8-NEXT:    std 27, 72(1) # 8-byte Folded Spill
 ; PWR8-NEXT:    std 28, 80(1) # 8-byte Folded Spill
+; PWR8-NEXT:    li 28, 16
+; PWR8-NEXT:    li 27, 80
 ; PWR8-NEXT:    std 29, 88(1) # 8-byte Folded Spill
-; PWR8-NEXT:    std 30, 96(1) # 8-byte Folded Spill
-; PWR8-NEXT:    li 30, 4
-; PWR8-NEXT:    li 29, 16
-; PWR8-NEXT:    std 26, 64(1) # 8-byte Folded Spill
 ; PWR8-NEXT:    stxvd2x 63, 1, 3 # 16-byte Folded Spill
-; PWR8-NEXT:    addis 3, 2, a@toc@ha
-; PWR8-NEXT:    addi 3, 3, a@toc@l
-; PWR8-NEXT:    std 27, 72(1) # 8-byte Folded Spill
-; PWR8-NEXT:    lxvd2x 0, 0, 3
-; PWR8-NEXT:    addis 3, 2, y@toc@ha
-; PWR8-NEXT:    addi 3, 3, y@toc@l
-; PWR8-NEXT:    addi 28, 3, -16
-; PWR8-NEXT:    addis 3, 2, x@toc@ha
-; PWR8-NEXT:    addi 3, 3, x@toc@l
-; PWR8-NEXT:    addi 3, 3, -16
+; PWR8-NEXT:    addis 3, 2, .L_MergedGlobals@toc@ha
+; PWR8-NEXT:    li 29, 4
+; PWR8-NEXT:    addi 30, 3, .L_MergedGlobals@toc@l
+; PWR8-NEXT:    std 26, 64(1) # 8-byte Folded Spill
+; PWR8-NEXT:    lxvd2x 0, 0, 30
 ; PWR8-NEXT:    xxswapd 63, 0
 ; PWR8-NEXT:    .p2align 4
 ; PWR8-NEXT:  .LBB0_1: # %for.body
 ; PWR8-NEXT:    #
-; PWR8-NEXT:    lxvd2x 0, 3, 29
+; PWR8-NEXT:    lxvd2x 0, 30, 28
 ; PWR8-NEXT:    vmr 2, 31
-; PWR8-NEXT:    addi 27, 28, 16
-; PWR8-NEXT:    addi 26, 3, 16
+; PWR8-NEXT:    addi 26, 30, 16
 ; PWR8-NEXT:    xxswapd 35, 0
 ; PWR8-NEXT:    bl __mulkf3
 ; PWR8-NEXT:    nop
-; PWR8-NEXT:    addi 30, 30, -1
+; PWR8-NEXT:    addi 29, 29, -1
 ; PWR8-NEXT:    xxswapd 0, 34
-; PWR8-NEXT:    mr 3, 26
-; PWR8-NEXT:    cmpldi 30, 0
-; PWR8-NEXT:    stxvd2x 0, 28, 29
-; PWR8-NEXT:    mr 28, 27
+; PWR8-NEXT:    cmpldi 29, 0
+; PWR8-NEXT:    stxvd2x 0, 30, 27
+; PWR8-NEXT:    mr 30, 26
 ; PWR8-NEXT:    bc 12, 1, .LBB0_1
 ; PWR8-NEXT:  # %bb.2: # %for.end
 ; PWR8-NEXT:    li 3, 48
@@ -110,9 +96,9 @@ define void @fpext_ctrloop_fp128(ptr %a) nounwind {
 ; PWR9-NEXT:    li 4, 4
 ; PWR9-NEXT:    addi 3, 3, -8
 ; PWR9-NEXT:    mtctr 4
-; PWR9-NEXT:    addis 4, 2, y@toc@ha
-; PWR9-NEXT:    addi 4, 4, y@toc@l
-; PWR9-NEXT:    addi 4, 4, -16
+; PWR9-NEXT:    addis 4, 2, .L_MergedGlobals@toc@ha
+; PWR9-NEXT:    addi 4, 4, .L_MergedGlobals@toc@l
+; PWR9-NEXT:    addi 4, 4, 64
 ; PWR9-NEXT:    .p2align 5
 ; PWR9-NEXT:  .LBB1_1: # %for.body
 ; PWR9-NEXT:    #
@@ -133,11 +119,11 @@ define void @fpext_ctrloop_fp128(ptr %a) nounwind {
 ; PWR8-NEXT:    std 30, -16(1) # 8-byte Folded Spill
 ; PWR8-NEXT:    stdu 1, -64(1)
 ; PWR8-NEXT:    addi 30, 3, -8
-; PWR8-NEXT:    addis 3, 2, y@toc@ha
+; PWR8-NEXT:    addis 3, 2, .L_MergedGlobals@toc@ha
 ; PWR8-NEXT:    li 29, 4
 ; PWR8-NEXT:    std 0, 80(1)
-; PWR8-NEXT:    addi 3, 3, y@toc@l
-; PWR8-NEXT:    addi 28, 3, -16
+; PWR8-NEXT:    addi 3, 3, .L_MergedGlobals@toc@l
+; PWR8-NEXT:    addi 28, 3, 64
 ; PWR8-NEXT:    .p2align 4
 ; PWR8-NEXT:  .LBB1_1: # %for.body
 ; PWR8-NEXT:    #
@@ -182,9 +168,8 @@ define void @fptrunc_ctrloop_fp128(ptr %a) nounwind {
 ; PWR9-NEXT:    li 4, 4
 ; PWR9-NEXT:    addi 3, 3, -8
 ; PWR9-NEXT:    mtctr 4
-; PWR9-NEXT:    addis 4, 2, x@toc@ha
-; PWR9-NEXT:    addi 4, 4, x@toc@l
-; PWR9-NEXT:    addi 4, 4, -16
+; PWR9-NEXT:    addis 4, 2, .L_MergedGlobals@toc@ha
+; PWR9-NEXT:    addi 4, 4, .L_MergedGlobals@toc@l
 ; PWR9-NEXT:    .p2align 5
 ; PWR9-NEXT:  .LBB2_1: # %for.body
 ; PWR9-NEXT:    #
@@ -205,11 +190,10 @@ define void @fptrunc_ctrloop_fp128(ptr %a) nounwind {
 ; PWR8-NEXT:    std 30, -16(1) # 8-byte Folded Spill
 ; PWR8-NEXT:    stdu 1, -64(1)
 ; PWR8-NEXT:    addi 30, 3, -8
-; PWR8-NEXT:    addis 3, 2, x@toc@ha
+; PWR8-NEXT:    addis 3, 2, .L_MergedGlobals@toc@ha
 ; PWR8-NEXT:    li 29, 4
 ; PWR8-NEXT:    std 0, 80(1)
-; PWR8-NEXT:    addi 3, 3, x@toc@l
-; PWR8-NEXT:    addi 28, 3, -16
+; PWR8-NEXT:    addi 28, 3, .L_MergedGlobals@toc@l
 ; PWR8-NEXT:    .p2align 4
 ; PWR8-NEXT:  .LBB2_1: # %for.body
 ; PWR8-NEXT:    #

llvm/test/CodeGen/PowerPC/expand-contiguous-isel.ll

@@ -110,11 +110,12 @@ define i64 @_Z3fn1N4llvm9StringRefE([2 x i64] %Str.coerce) {
 ; CHECK-GEN-ISEL-TRUE-NEXT:    mtlr r0
 ; CHECK-GEN-ISEL-TRUE-NEXT:    blr
 ; CHECK-GEN-ISEL-TRUE-NEXT:  .LBB0_15: # %cond.false.i
-; CHECK-GEN-ISEL-TRUE-NEXT:    addis r3, r2, .L__ModuleStringPool@toc@ha
+; CHECK-GEN-ISEL-TRUE-NEXT:    addis r3, r2, .L_MergedGlobals@toc@ha
+; CHECK-GEN-ISEL-TRUE-NEXT:    addi r5, r3, .L_MergedGlobals@toc@l
+; CHECK-GEN-ISEL-TRUE-NEXT:    addi r3, r5, 3
+; CHECK-GEN-ISEL-TRUE-NEXT:    addi r4, r5, 134
+; CHECK-GEN-ISEL-TRUE-NEXT:    addi r6, r5, 38
 ; CHECK-GEN-ISEL-TRUE-NEXT:    li r5, 225
-; CHECK-GEN-ISEL-TRUE-NEXT:    addi r4, r3, .L__ModuleStringPool@toc@l
-; CHECK-GEN-ISEL-TRUE-NEXT:    addi r3, r4, 53
-; CHECK-GEN-ISEL-TRUE-NEXT:    addi r6, r4, 88
 ; CHECK-GEN-ISEL-TRUE-NEXT:    bl __assert_fail
 ; CHECK-GEN-ISEL-TRUE-NEXT:    nop
 ; CHECK-GEN-ISEL-TRUE-NEXT:  .LBB0_16: # %if.then9
@@ -226,11 +227,12 @@ define i64 @_Z3fn1N4llvm9StringRefE([2 x i64] %Str.coerce) {
 ; CHECK-NEXT:    mtlr r0
 ; CHECK-NEXT:    blr
 ; CHECK-NEXT:  .LBB0_20: # %cond.false.i
-; CHECK-NEXT:    addis r3, r2, .L__ModuleStringPool@toc@ha
+; CHECK-NEXT:    addis r3, r2, .L_MergedGlobals@toc@ha
+; CHECK-NEXT:    addi r5, r3, .L_MergedGlobals@toc@l
+; CHECK-NEXT:    addi r3, r5, 3
+; CHECK-NEXT:    addi r4, r5, 134
+; CHECK-NEXT:    addi r6, r5, 38
 ; CHECK-NEXT:    li r5, 225
-; CHECK-NEXT:    addi r4, r3, .L__ModuleStringPool@toc@l
-; CHECK-NEXT:    addi r3, r4, 53
-; CHECK-NEXT:    addi r6, r4, 88
 ; CHECK-NEXT:    bl __assert_fail
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:  .LBB0_21: # %if.then9

llvm/test/CodeGen/PowerPC/licm-remat.ll

@@ -20,10 +20,10 @@ declare void @llvm.memcpy.p0.p0.i64(ptr nocapture writeonly, ptr nocapture reado
 define linkonce_odr void @ZN6snappyDecompressor_(ptr %this, ptr %writer) {
 ; CHECK-LABEL: ZN6snappyDecompressor_:
 ; CHECK:       # %bb.0: # %entry
-; CHECK:       addis 4, 2, .L__ModuleStringPool@toc@ha
-; CHECK:       addi 26, 4, .L__ModuleStringPool@toc@l
+; CHECK:       addis 4, 2, .L_MergedGlobals@toc@ha
+; CHECK:       addi 26, 4, .L_MergedGlobals@toc@l
 ; CHECK:       .LBB0_2: # %for.cond
-; CHECK-NOT:   addis {{[0-9]+}}, 2, .L__ModuleStringPool@toc@ha
+; CHECK-NOT:   addis {{[0-9]+}}, 2, .L_MergedGlobals@toc@ha
 ; CHECK:       bctrl
 entry:
   %ip_limit_ = getelementptr inbounds %"class.snappy::SnappyDecompressor", ptr %this, i64 0, i32 2

llvm/test/CodeGen/PowerPC/merge-private.ll

@@ -6,10 +6,10 @@
 ; RUN:     -ppc-asm-full-reg-names < %s | FileCheck %s \
 ; RUN:     --check-prefix=AIX32
 ; RUN: llc -verify-machineinstrs -mtriple powerpc64le-unknown-linux -mcpu=pwr8 \
-; RUN:     -ppc-asm-full-reg-names -ppc-global-merge=true < %s | FileCheck %s \
+; RUN:     -ppc-asm-full-reg-names < %s | FileCheck %s \
 ; RUN:     --check-prefix=LINUX64LE
 ; RUN: llc -verify-machineinstrs -mtriple powerpc64-unknown-linux -mcpu=pwr8 \
-; RUN:     -ppc-asm-full-reg-names -ppc-global-merge=true < %s | FileCheck %s \
+; RUN:     -ppc-asm-full-reg-names < %s | FileCheck %s \
 ; RUN:     --check-prefix=LINUX64BE
 ; The below run line is added to ensure that the assembly corresponding to
 ; the following check-prefix is generated by default on AIX (without any