From: Philip Reames Date: Thu, 12 Sep 2019 23:22:37 +0000 (+0000) Subject: [Test] Restructure check lines to show differences between modes more clearly X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=8fb3b56963373249e94583e01f2756db23322bef;p=llvm [Test] Restructure check lines to show differences between modes more clearly With the landing of the previous patch (in particular D66318) there are a lot fewer diffs now. I added an experimental O0 line, and updated all the tests to group experimental and non-experimental O0/O3 together. Skimming the remaining diffs, there's only a few which are obviously incorrect. There's a large number which are questionable, so more todo. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@371790 91177308-0d34-0410-b5e6-96231b3b80d8 --- diff --git a/test/CodeGen/X86/atomic-unordered.ll b/test/CodeGen/X86/atomic-unordered.ll index 722791e1279..db4fc2d0575 100644 --- a/test/CodeGen/X86/atomic-unordered.ll +++ b/test/CodeGen/X86/atomic-unordered.ll @@ -1,7 +1,8 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc -O0 < %s -mtriple=x86_64-linux-generic -verify-machineinstrs -mcpu=skylake | FileCheck --check-prefixes=CHECK,CHECK-NOX,CHECK-O0 %s -; RUN: llc -O3 < %s -mtriple=x86_64-linux-generic -verify-machineinstrs -mcpu=skylake | FileCheck --check-prefixes=CHECK,CHECK-NOX,CHECK-O3 %s -; RUN: llc -O3 < %s -mtriple=x86_64-linux-generic -verify-machineinstrs -mcpu=skylake -x86-experimental-unordered-atomic-isel | FileCheck --check-prefixes=CHECK,CHECK-EX %s +; RUN: llc -O0 < %s -mtriple=x86_64-linux-generic -verify-machineinstrs -mcpu=skylake | FileCheck --check-prefixes=CHECK,CHECK-O0,CHECK-O0-CUR %s +; RUN: llc -O3 < %s -mtriple=x86_64-linux-generic -verify-machineinstrs -mcpu=skylake | FileCheck --check-prefixes=CHECK,CHECK-O3,CHECK-O3-CUR %s +; RUN: llc -O0 < %s -mtriple=x86_64-linux-generic -verify-machineinstrs -mcpu=skylake -x86-experimental-unordered-atomic-isel | FileCheck --check-prefixes=CHECK,CHECK-O0,CHECK-O0-EX %s +; RUN: llc -O3 < %s -mtriple=x86_64-linux-generic -verify-machineinstrs -mcpu=skylake -x86-experimental-unordered-atomic-isel | FileCheck --check-prefixes=CHECK,CHECK-O3,CHECK-O3-EX %s define i8 @load_i8(i8* %ptr) { ; CHECK-LABEL: load_i8: @@ -23,11 +24,6 @@ define void @store_i8(i8* %ptr, i8 %v) { ; CHECK-O3: # %bb.0: ; CHECK-O3-NEXT: movb %sil, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: store_i8: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movb %sil, (%rdi) -; CHECK-EX-NEXT: retq store atomic i8 %v, i8* %ptr unordered, align 1 ret void } @@ -42,11 +38,6 @@ define i16 @load_i16(i16* %ptr) { ; CHECK-O3: # %bb.0: ; CHECK-O3-NEXT: movzwl (%rdi), %eax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_i16: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movzwl (%rdi), %eax -; CHECK-EX-NEXT: retq %v = load atomic i16, i16* %ptr unordered, align 2 ret i16 %v } @@ -63,11 +54,6 @@ define void @store_i16(i16* %ptr, i16 %v) { ; CHECK-O3: # %bb.0: ; CHECK-O3-NEXT: movw %si, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: store_i16: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movw %si, (%rdi) -; CHECK-EX-NEXT: retq store atomic i16 %v, i16* %ptr unordered, align 2 ret void } @@ -127,11 +113,6 @@ define void @narrow_writeback_or(i64* %ptr) { ; CHECK-O3: # %bb.0: ; CHECK-O3-NEXT: orq $7, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: narrow_writeback_or: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: orq $7, (%rdi) -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %ptr unordered, align 8 %v.new = or i64 %v, 7 store atomic i64 %v.new, i64* %ptr unordered, align 8 @@ -154,12 +135,6 @@ define void @narrow_writeback_and(i64* %ptr) { ; CHECK-O3-NEXT: movl $4294967040, %eax # imm = 0xFFFFFF00 ; CHECK-O3-NEXT: andq %rax, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: narrow_writeback_and: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movl $4294967040, %eax # imm = 0xFFFFFF00 -; CHECK-EX-NEXT: andq %rax, (%rdi) -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %ptr unordered, align 8 %v.new = and i64 %v, 4294967040 ;; 0xFFFF_FF00 store atomic i64 %v.new, i64* %ptr unordered, align 8 @@ -179,11 +154,6 @@ define void @narrow_writeback_xor(i64* %ptr) { ; CHECK-O3: # %bb.0: ; CHECK-O3-NEXT: xorq $7, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: narrow_writeback_xor: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: xorq $7, (%rdi) -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %ptr unordered, align 8 %v.new = xor i64 %v, 7 store atomic i64 %v.new, i64* %ptr unordered, align 8 @@ -194,7 +164,6 @@ define void @narrow_writeback_xor(i64* %ptr) { ;; improve codegeneration. Note that widening is only legal if the ;; resulting type would be atomic. Each tests has a well aligned, and ;; unaligned variant to ensure we get correct codegen here. -;; ;; Note: It's not a legality issue, but there's a gotcha here to be aware ;; of. Once we widen a pair of atomic stores, we loose the information ;; that the original atomicity requirement was half the width. Given that, @@ -285,20 +254,6 @@ define i128 @load_i128(i128* %ptr) { ; CHECK-O3-NEXT: popq %rbx ; CHECK-O3-NEXT: .cfi_def_cfa_offset 8 ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_i128: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: pushq %rbx -; CHECK-EX-NEXT: .cfi_def_cfa_offset 16 -; CHECK-EX-NEXT: .cfi_offset %rbx, -16 -; CHECK-EX-NEXT: xorl %eax, %eax -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: xorl %ecx, %ecx -; CHECK-EX-NEXT: xorl %ebx, %ebx -; CHECK-EX-NEXT: lock cmpxchg16b (%rdi) -; CHECK-EX-NEXT: popq %rbx -; CHECK-EX-NEXT: .cfi_def_cfa_offset 8 -; CHECK-EX-NEXT: retq %v = load atomic i128, i128* %ptr unordered, align 16 ret i128 %v } @@ -352,25 +307,6 @@ define void @store_i128(i128* %ptr, i128 %v) { ; CHECK-O3-NEXT: popq %rbx ; CHECK-O3-NEXT: .cfi_def_cfa_offset 8 ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: store_i128: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: pushq %rbx -; CHECK-EX-NEXT: .cfi_def_cfa_offset 16 -; CHECK-EX-NEXT: .cfi_offset %rbx, -16 -; CHECK-EX-NEXT: movq %rdx, %rcx -; CHECK-EX-NEXT: movq %rsi, %rbx -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: movq 8(%rdi), %rdx -; CHECK-EX-NEXT: .p2align 4, 0x90 -; CHECK-EX-NEXT: .LBB16_1: # %atomicrmw.start -; CHECK-EX-NEXT: # =>This Inner Loop Header: Depth=1 -; CHECK-EX-NEXT: lock cmpxchg16b (%rdi) -; CHECK-EX-NEXT: jne .LBB16_1 -; CHECK-EX-NEXT: # %bb.2: # %atomicrmw.end -; CHECK-EX-NEXT: popq %rbx -; CHECK-EX-NEXT: .cfi_def_cfa_offset 8 -; CHECK-EX-NEXT: retq store atomic i128 %v, i128* %ptr unordered, align 16 ret void } @@ -424,28 +360,6 @@ define i256 @load_i256(i256* %ptr) { ; CHECK-O3-NEXT: .cfi_def_cfa_offset 8 ; CHECK-O3-NEXT: vzeroupper ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_i256: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: pushq %rbx -; CHECK-EX-NEXT: .cfi_def_cfa_offset 16 -; CHECK-EX-NEXT: subq $32, %rsp -; CHECK-EX-NEXT: .cfi_def_cfa_offset 48 -; CHECK-EX-NEXT: .cfi_offset %rbx, -16 -; CHECK-EX-NEXT: movq %rdi, %rbx -; CHECK-EX-NEXT: movq %rsp, %rdx -; CHECK-EX-NEXT: movl $32, %edi -; CHECK-EX-NEXT: xorl %ecx, %ecx -; CHECK-EX-NEXT: callq __atomic_load -; CHECK-EX-NEXT: vmovups (%rsp), %ymm0 -; CHECK-EX-NEXT: vmovups %ymm0, (%rbx) -; CHECK-EX-NEXT: movq %rbx, %rax -; CHECK-EX-NEXT: addq $32, %rsp -; CHECK-EX-NEXT: .cfi_def_cfa_offset 16 -; CHECK-EX-NEXT: popq %rbx -; CHECK-EX-NEXT: .cfi_def_cfa_offset 8 -; CHECK-EX-NEXT: vzeroupper -; CHECK-EX-NEXT: retq %v = load atomic i256, i256* %ptr unordered, align 16 ret i256 %v } @@ -489,43 +403,39 @@ define void @store_i256(i256* %ptr, i256 %v) { ; CHECK-O3-NEXT: addq $40, %rsp ; CHECK-O3-NEXT: .cfi_def_cfa_offset 8 ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: store_i256: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: subq $40, %rsp -; CHECK-EX-NEXT: .cfi_def_cfa_offset 48 -; CHECK-EX-NEXT: movq %rdi, %rax -; CHECK-EX-NEXT: movq %r8, {{[0-9]+}}(%rsp) -; CHECK-EX-NEXT: movq %rcx, {{[0-9]+}}(%rsp) -; CHECK-EX-NEXT: movq %rdx, {{[0-9]+}}(%rsp) -; CHECK-EX-NEXT: movq %rsi, {{[0-9]+}}(%rsp) -; CHECK-EX-NEXT: leaq {{[0-9]+}}(%rsp), %rdx -; CHECK-EX-NEXT: movl $32, %edi -; CHECK-EX-NEXT: movq %rax, %rsi -; CHECK-EX-NEXT: xorl %ecx, %ecx -; CHECK-EX-NEXT: callq __atomic_store -; CHECK-EX-NEXT: addq $40, %rsp -; CHECK-EX-NEXT: .cfi_def_cfa_offset 8 -; CHECK-EX-NEXT: retq store atomic i256 %v, i256* %ptr unordered, align 16 ret void } ; Legal if wider type is also atomic (TODO) define void @vec_store(i32* %p0, <2 x i32> %vec) { -; CHECK-NOX-LABEL: vec_store: -; CHECK-NOX: # %bb.0: -; CHECK-NOX-NEXT: vmovd %xmm0, %eax -; CHECK-NOX-NEXT: vpextrd $1, %xmm0, %ecx -; CHECK-NOX-NEXT: movl %eax, (%rdi) -; CHECK-NOX-NEXT: movl %ecx, 4(%rdi) -; CHECK-NOX-NEXT: retq -; -; CHECK-EX-LABEL: vec_store: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: vmovss %xmm0, (%rdi) -; CHECK-EX-NEXT: vextractps $1, %xmm0, 4(%rdi) -; CHECK-EX-NEXT: retq +; CHECK-O0-CUR-LABEL: vec_store: +; CHECK-O0-CUR: # %bb.0: +; CHECK-O0-CUR-NEXT: vmovd %xmm0, %eax +; CHECK-O0-CUR-NEXT: vpextrd $1, %xmm0, %ecx +; CHECK-O0-CUR-NEXT: movl %eax, (%rdi) +; CHECK-O0-CUR-NEXT: movl %ecx, 4(%rdi) +; CHECK-O0-CUR-NEXT: retq +; +; CHECK-O3-CUR-LABEL: vec_store: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: vmovd %xmm0, %eax +; CHECK-O3-CUR-NEXT: vpextrd $1, %xmm0, %ecx +; CHECK-O3-CUR-NEXT: movl %eax, (%rdi) +; CHECK-O3-CUR-NEXT: movl %ecx, 4(%rdi) +; CHECK-O3-CUR-NEXT: retq +; +; CHECK-O0-EX-LABEL: vec_store: +; CHECK-O0-EX: # %bb.0: +; CHECK-O0-EX-NEXT: vmovd %xmm0, (%rdi) +; CHECK-O0-EX-NEXT: vpextrd $1, %xmm0, 4(%rdi) +; CHECK-O0-EX-NEXT: retq +; +; CHECK-O3-EX-LABEL: vec_store: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: vmovss %xmm0, (%rdi) +; CHECK-O3-EX-NEXT: vextractps $1, %xmm0, 4(%rdi) +; CHECK-O3-EX-NEXT: retq %v1 = extractelement <2 x i32> %vec, i32 0 %v2 = extractelement <2 x i32> %vec, i32 1 %p1 = getelementptr i32, i32* %p0, i64 1 @@ -536,19 +446,33 @@ define void @vec_store(i32* %p0, <2 x i32> %vec) { ; Not legal to widen due to alignment restriction define void @vec_store_unaligned(i32* %p0, <2 x i32> %vec) { -; CHECK-NOX-LABEL: vec_store_unaligned: -; CHECK-NOX: # %bb.0: -; CHECK-NOX-NEXT: vmovd %xmm0, %eax -; CHECK-NOX-NEXT: vpextrd $1, %xmm0, %ecx -; CHECK-NOX-NEXT: movl %eax, (%rdi) -; CHECK-NOX-NEXT: movl %ecx, 4(%rdi) -; CHECK-NOX-NEXT: retq -; -; CHECK-EX-LABEL: vec_store_unaligned: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: vmovss %xmm0, (%rdi) -; CHECK-EX-NEXT: vextractps $1, %xmm0, 4(%rdi) -; CHECK-EX-NEXT: retq +; CHECK-O0-CUR-LABEL: vec_store_unaligned: +; CHECK-O0-CUR: # %bb.0: +; CHECK-O0-CUR-NEXT: vmovd %xmm0, %eax +; CHECK-O0-CUR-NEXT: vpextrd $1, %xmm0, %ecx +; CHECK-O0-CUR-NEXT: movl %eax, (%rdi) +; CHECK-O0-CUR-NEXT: movl %ecx, 4(%rdi) +; CHECK-O0-CUR-NEXT: retq +; +; CHECK-O3-CUR-LABEL: vec_store_unaligned: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: vmovd %xmm0, %eax +; CHECK-O3-CUR-NEXT: vpextrd $1, %xmm0, %ecx +; CHECK-O3-CUR-NEXT: movl %eax, (%rdi) +; CHECK-O3-CUR-NEXT: movl %ecx, 4(%rdi) +; CHECK-O3-CUR-NEXT: retq +; +; CHECK-O0-EX-LABEL: vec_store_unaligned: +; CHECK-O0-EX: # %bb.0: +; CHECK-O0-EX-NEXT: vmovd %xmm0, (%rdi) +; CHECK-O0-EX-NEXT: vpextrd $1, %xmm0, 4(%rdi) +; CHECK-O0-EX-NEXT: retq +; +; CHECK-O3-EX-LABEL: vec_store_unaligned: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: vmovss %xmm0, (%rdi) +; CHECK-O3-EX-NEXT: vextractps $1, %xmm0, 4(%rdi) +; CHECK-O3-EX-NEXT: retq %v1 = extractelement <2 x i32> %vec, i32 0 %v2 = extractelement <2 x i32> %vec, i32 1 %p1 = getelementptr i32, i32* %p0, i64 1 @@ -562,18 +486,31 @@ define void @vec_store_unaligned(i32* %p0, <2 x i32> %vec) { ; Legal if wider type is also atomic (TODO) ; Also, can avoid register move from xmm to eax (TODO) define void @widen_broadcast2(i32* %p0, <2 x i32> %vec) { -; CHECK-NOX-LABEL: widen_broadcast2: -; CHECK-NOX: # %bb.0: -; CHECK-NOX-NEXT: vmovd %xmm0, %eax -; CHECK-NOX-NEXT: movl %eax, (%rdi) -; CHECK-NOX-NEXT: movl %eax, 4(%rdi) -; CHECK-NOX-NEXT: retq -; -; CHECK-EX-LABEL: widen_broadcast2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: vmovss %xmm0, (%rdi) -; CHECK-EX-NEXT: vmovss %xmm0, 4(%rdi) -; CHECK-EX-NEXT: retq +; CHECK-O0-CUR-LABEL: widen_broadcast2: +; CHECK-O0-CUR: # %bb.0: +; CHECK-O0-CUR-NEXT: vmovd %xmm0, %eax +; CHECK-O0-CUR-NEXT: movl %eax, (%rdi) +; CHECK-O0-CUR-NEXT: movl %eax, 4(%rdi) +; CHECK-O0-CUR-NEXT: retq +; +; CHECK-O3-CUR-LABEL: widen_broadcast2: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: vmovd %xmm0, %eax +; CHECK-O3-CUR-NEXT: movl %eax, (%rdi) +; CHECK-O3-CUR-NEXT: movl %eax, 4(%rdi) +; CHECK-O3-CUR-NEXT: retq +; +; CHECK-O0-EX-LABEL: widen_broadcast2: +; CHECK-O0-EX: # %bb.0: +; CHECK-O0-EX-NEXT: vmovd %xmm0, (%rdi) +; CHECK-O0-EX-NEXT: vmovd %xmm0, 4(%rdi) +; CHECK-O0-EX-NEXT: retq +; +; CHECK-O3-EX-LABEL: widen_broadcast2: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: vmovss %xmm0, (%rdi) +; CHECK-O3-EX-NEXT: vmovss %xmm0, 4(%rdi) +; CHECK-O3-EX-NEXT: retq %v1 = extractelement <2 x i32> %vec, i32 0 %p1 = getelementptr i32, i32* %p0, i64 1 store atomic i32 %v1, i32* %p0 unordered, align 8 @@ -583,18 +520,31 @@ define void @widen_broadcast2(i32* %p0, <2 x i32> %vec) { ; Not legal to widen due to alignment restriction define void @widen_broadcast2_unaligned(i32* %p0, <2 x i32> %vec) { -; CHECK-NOX-LABEL: widen_broadcast2_unaligned: -; CHECK-NOX: # %bb.0: -; CHECK-NOX-NEXT: vmovd %xmm0, %eax -; CHECK-NOX-NEXT: movl %eax, (%rdi) -; CHECK-NOX-NEXT: movl %eax, 4(%rdi) -; CHECK-NOX-NEXT: retq -; -; CHECK-EX-LABEL: widen_broadcast2_unaligned: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: vmovss %xmm0, (%rdi) -; CHECK-EX-NEXT: vmovss %xmm0, 4(%rdi) -; CHECK-EX-NEXT: retq +; CHECK-O0-CUR-LABEL: widen_broadcast2_unaligned: +; CHECK-O0-CUR: # %bb.0: +; CHECK-O0-CUR-NEXT: vmovd %xmm0, %eax +; CHECK-O0-CUR-NEXT: movl %eax, (%rdi) +; CHECK-O0-CUR-NEXT: movl %eax, 4(%rdi) +; CHECK-O0-CUR-NEXT: retq +; +; CHECK-O3-CUR-LABEL: widen_broadcast2_unaligned: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: vmovd %xmm0, %eax +; CHECK-O3-CUR-NEXT: movl %eax, (%rdi) +; CHECK-O3-CUR-NEXT: movl %eax, 4(%rdi) +; CHECK-O3-CUR-NEXT: retq +; +; CHECK-O0-EX-LABEL: widen_broadcast2_unaligned: +; CHECK-O0-EX: # %bb.0: +; CHECK-O0-EX-NEXT: vmovd %xmm0, (%rdi) +; CHECK-O0-EX-NEXT: vmovd %xmm0, 4(%rdi) +; CHECK-O0-EX-NEXT: retq +; +; CHECK-O3-EX-LABEL: widen_broadcast2_unaligned: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: vmovss %xmm0, (%rdi) +; CHECK-O3-EX-NEXT: vmovss %xmm0, 4(%rdi) +; CHECK-O3-EX-NEXT: retq %v1 = extractelement <2 x i32> %vec, i32 0 %p1 = getelementptr i32, i32* %p0, i64 1 store atomic i32 %v1, i32* %p0 unordered, align 4 @@ -655,12 +605,6 @@ define i64 @load_fold_add2(i64* %p, i64 %v2) { ; CHECK-O3-NEXT: movq %rsi, %rax ; CHECK-O3-NEXT: addq (%rdi), %rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_add2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq %rsi, %rax -; CHECK-EX-NEXT: addq (%rdi), %rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = add i64 %v, %v2 ret i64 %ret @@ -673,17 +617,17 @@ define i64 @load_fold_add3(i64* %p1, i64* %p2) { ; CHECK-O0-NEXT: addq (%rsi), %rax ; CHECK-O0-NEXT: retq ; -; CHECK-O3-LABEL: load_fold_add3: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: movq (%rsi), %rax -; CHECK-O3-NEXT: addq (%rdi), %rax -; CHECK-O3-NEXT: retq +; CHECK-O3-CUR-LABEL: load_fold_add3: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movq (%rsi), %rax +; CHECK-O3-CUR-NEXT: addq (%rdi), %rax +; CHECK-O3-CUR-NEXT: retq ; -; CHECK-EX-LABEL: load_fold_add3: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: addq (%rsi), %rax -; CHECK-EX-NEXT: retq +; CHECK-O3-EX-LABEL: load_fold_add3: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: movq (%rdi), %rax +; CHECK-O3-EX-NEXT: addq (%rsi), %rax +; CHECK-O3-EX-NEXT: retq %v = load atomic i64, i64* %p1 unordered, align 8 %v2 = load atomic i64, i64* %p2 unordered, align 8 %ret = add i64 %v, %v2 @@ -703,12 +647,6 @@ define i64 @load_fold_sub1(i64* %p) { ; CHECK-O3-NEXT: movq (%rdi), %rax ; CHECK-O3-NEXT: addq $-15, %rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_sub1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: addq $-15, %rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = sub i64 %v, 15 ret i64 %ret @@ -750,13 +688,6 @@ define i64 @load_fold_mul1(i64* %p) { ; CHECK-O3-NEXT: leaq (%rax,%rax,4), %rax ; CHECK-O3-NEXT: leaq (%rax,%rax,2), %rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_mul1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: leaq (%rax,%rax,4), %rax -; CHECK-EX-NEXT: leaq (%rax,%rax,2), %rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = mul i64 %v, 15 ret i64 %ret @@ -774,12 +705,6 @@ define i64 @load_fold_mul2(i64* %p, i64 %v2) { ; CHECK-O3-NEXT: movq %rsi, %rax ; CHECK-O3-NEXT: imulq (%rdi), %rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_mul2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq %rsi, %rax -; CHECK-EX-NEXT: imulq (%rdi), %rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = mul i64 %v, %v2 ret i64 %ret @@ -792,17 +717,17 @@ define i64 @load_fold_mul3(i64* %p1, i64* %p2) { ; CHECK-O0-NEXT: imulq (%rsi), %rax ; CHECK-O0-NEXT: retq ; -; CHECK-O3-LABEL: load_fold_mul3: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: movq (%rsi), %rax -; CHECK-O3-NEXT: imulq (%rdi), %rax -; CHECK-O3-NEXT: retq +; CHECK-O3-CUR-LABEL: load_fold_mul3: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movq (%rsi), %rax +; CHECK-O3-CUR-NEXT: imulq (%rdi), %rax +; CHECK-O3-CUR-NEXT: retq ; -; CHECK-EX-LABEL: load_fold_mul3: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: imulq (%rsi), %rax -; CHECK-EX-NEXT: retq +; CHECK-O3-EX-LABEL: load_fold_mul3: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: movq (%rdi), %rax +; CHECK-O3-EX-NEXT: imulq (%rsi), %rax +; CHECK-O3-EX-NEXT: retq %v = load atomic i64, i64* %p1 unordered, align 8 %v2 = load atomic i64, i64* %p2 unordered, align 8 %ret = mul i64 %v, %v2 @@ -831,19 +756,6 @@ define i64 @load_fold_sdiv1(i64* %p) { ; CHECK-O3-NEXT: sarq $3, %rdx ; CHECK-O3-NEXT: addq %rdx, %rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_sdiv1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rcx -; CHECK-EX-NEXT: movabsq $-8608480567731124087, %rdx # imm = 0x8888888888888889 -; CHECK-EX-NEXT: movq %rcx, %rax -; CHECK-EX-NEXT: imulq %rdx -; CHECK-EX-NEXT: addq %rcx, %rdx -; CHECK-EX-NEXT: movq %rdx, %rax -; CHECK-EX-NEXT: shrq $63, %rax -; CHECK-EX-NEXT: sarq $3, %rdx -; CHECK-EX-NEXT: addq %rdx, %rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = sdiv i64 %v, 15 ret i64 %ret @@ -875,24 +787,6 @@ define i64 @load_fold_sdiv2(i64* %p, i64 %v2) { ; CHECK-O3-NEXT: divl %esi ; CHECK-O3-NEXT: # kill: def $eax killed $eax def $rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_sdiv2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: movq %rax, %rcx -; CHECK-EX-NEXT: orq %rsi, %rcx -; CHECK-EX-NEXT: shrq $32, %rcx -; CHECK-EX-NEXT: je .LBB35_1 -; CHECK-EX-NEXT: # %bb.2: -; CHECK-EX-NEXT: cqto -; CHECK-EX-NEXT: idivq %rsi -; CHECK-EX-NEXT: retq -; CHECK-EX-NEXT: .LBB35_1: -; CHECK-EX-NEXT: # kill: def $eax killed $eax killed $rax -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: divl %esi -; CHECK-EX-NEXT: # kill: def $eax killed $eax def $rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = sdiv i64 %v, %v2 ret i64 %ret @@ -924,25 +818,6 @@ define i64 @load_fold_sdiv3(i64* %p1, i64* %p2) { ; CHECK-O3-NEXT: divl %ecx ; CHECK-O3-NEXT: # kill: def $eax killed $eax def $rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_sdiv3: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: movq (%rsi), %rcx -; CHECK-EX-NEXT: movq %rax, %rdx -; CHECK-EX-NEXT: orq %rcx, %rdx -; CHECK-EX-NEXT: shrq $32, %rdx -; CHECK-EX-NEXT: je .LBB36_1 -; CHECK-EX-NEXT: # %bb.2: -; CHECK-EX-NEXT: cqto -; CHECK-EX-NEXT: idivq %rcx -; CHECK-EX-NEXT: retq -; CHECK-EX-NEXT: .LBB36_1: -; CHECK-EX-NEXT: # kill: def $eax killed $eax killed $rax -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: divl %ecx -; CHECK-EX-NEXT: # kill: def $eax killed $eax def $rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p1 unordered, align 8 %v2 = load atomic i64, i64* %p2 unordered, align 8 %ret = sdiv i64 %v, %v2 @@ -960,22 +835,22 @@ define i64 @load_fold_udiv1(i64* %p) { ; CHECK-O0-NEXT: divq %rsi ; CHECK-O0-NEXT: retq ; -; CHECK-O3-LABEL: load_fold_udiv1: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: movq (%rdi), %rax -; CHECK-O3-NEXT: movabsq $-8608480567731124087, %rcx # imm = 0x8888888888888889 -; CHECK-O3-NEXT: mulq %rcx -; CHECK-O3-NEXT: movq %rdx, %rax -; CHECK-O3-NEXT: shrq $3, %rax -; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_udiv1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movabsq $-8608480567731124087, %rax # imm = 0x8888888888888889 -; CHECK-EX-NEXT: mulq (%rdi) -; CHECK-EX-NEXT: movq %rdx, %rax -; CHECK-EX-NEXT: shrq $3, %rax -; CHECK-EX-NEXT: retq +; CHECK-O3-CUR-LABEL: load_fold_udiv1: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movq (%rdi), %rax +; CHECK-O3-CUR-NEXT: movabsq $-8608480567731124087, %rcx # imm = 0x8888888888888889 +; CHECK-O3-CUR-NEXT: mulq %rcx +; CHECK-O3-CUR-NEXT: movq %rdx, %rax +; CHECK-O3-CUR-NEXT: shrq $3, %rax +; CHECK-O3-CUR-NEXT: retq +; +; CHECK-O3-EX-LABEL: load_fold_udiv1: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: movabsq $-8608480567731124087, %rax # imm = 0x8888888888888889 +; CHECK-O3-EX-NEXT: mulq (%rdi) +; CHECK-O3-EX-NEXT: movq %rdx, %rax +; CHECK-O3-EX-NEXT: shrq $3, %rax +; CHECK-O3-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = udiv i64 %v, 15 ret i64 %ret @@ -1007,24 +882,6 @@ define i64 @load_fold_udiv2(i64* %p, i64 %v2) { ; CHECK-O3-NEXT: divl %esi ; CHECK-O3-NEXT: # kill: def $eax killed $eax def $rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_udiv2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: movq %rax, %rcx -; CHECK-EX-NEXT: orq %rsi, %rcx -; CHECK-EX-NEXT: shrq $32, %rcx -; CHECK-EX-NEXT: je .LBB38_1 -; CHECK-EX-NEXT: # %bb.2: -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: divq %rsi -; CHECK-EX-NEXT: retq -; CHECK-EX-NEXT: .LBB38_1: -; CHECK-EX-NEXT: # kill: def $eax killed $eax killed $rax -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: divl %esi -; CHECK-EX-NEXT: # kill: def $eax killed $eax def $rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = udiv i64 %v, %v2 ret i64 %ret @@ -1057,25 +914,6 @@ define i64 @load_fold_udiv3(i64* %p1, i64* %p2) { ; CHECK-O3-NEXT: divl %ecx ; CHECK-O3-NEXT: # kill: def $eax killed $eax def $rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_udiv3: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: movq (%rsi), %rcx -; CHECK-EX-NEXT: movq %rax, %rdx -; CHECK-EX-NEXT: orq %rcx, %rdx -; CHECK-EX-NEXT: shrq $32, %rdx -; CHECK-EX-NEXT: je .LBB39_1 -; CHECK-EX-NEXT: # %bb.2: -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: divq %rcx -; CHECK-EX-NEXT: retq -; CHECK-EX-NEXT: .LBB39_1: -; CHECK-EX-NEXT: # kill: def $eax killed $eax killed $rax -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: divl %ecx -; CHECK-EX-NEXT: # kill: def $eax killed $eax def $rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p1 unordered, align 8 %v2 = load atomic i64, i64* %p2 unordered, align 8 %ret = udiv i64 %v, %v2 @@ -1109,23 +947,6 @@ define i64 @load_fold_srem1(i64* %p) { ; CHECK-O3-NEXT: subq %rax, %rcx ; CHECK-O3-NEXT: movq %rcx, %rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_srem1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rcx -; CHECK-EX-NEXT: movabsq $-8608480567731124087, %rdx # imm = 0x8888888888888889 -; CHECK-EX-NEXT: movq %rcx, %rax -; CHECK-EX-NEXT: imulq %rdx -; CHECK-EX-NEXT: addq %rcx, %rdx -; CHECK-EX-NEXT: movq %rdx, %rax -; CHECK-EX-NEXT: shrq $63, %rax -; CHECK-EX-NEXT: sarq $3, %rdx -; CHECK-EX-NEXT: addq %rax, %rdx -; CHECK-EX-NEXT: leaq (%rdx,%rdx,4), %rax -; CHECK-EX-NEXT: leaq (%rax,%rax,2), %rax -; CHECK-EX-NEXT: subq %rax, %rcx -; CHECK-EX-NEXT: movq %rcx, %rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = srem i64 %v, 15 ret i64 %ret @@ -1159,25 +980,6 @@ define i64 @load_fold_srem2(i64* %p, i64 %v2) { ; CHECK-O3-NEXT: divl %esi ; CHECK-O3-NEXT: movl %edx, %eax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_srem2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: movq %rax, %rcx -; CHECK-EX-NEXT: orq %rsi, %rcx -; CHECK-EX-NEXT: shrq $32, %rcx -; CHECK-EX-NEXT: je .LBB41_1 -; CHECK-EX-NEXT: # %bb.2: -; CHECK-EX-NEXT: cqto -; CHECK-EX-NEXT: idivq %rsi -; CHECK-EX-NEXT: movq %rdx, %rax -; CHECK-EX-NEXT: retq -; CHECK-EX-NEXT: .LBB41_1: -; CHECK-EX-NEXT: # kill: def $eax killed $eax killed $rax -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: divl %esi -; CHECK-EX-NEXT: movl %edx, %eax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = srem i64 %v, %v2 ret i64 %ret @@ -1211,26 +1013,6 @@ define i64 @load_fold_srem3(i64* %p1, i64* %p2) { ; CHECK-O3-NEXT: divl %ecx ; CHECK-O3-NEXT: movl %edx, %eax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_srem3: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: movq (%rsi), %rcx -; CHECK-EX-NEXT: movq %rax, %rdx -; CHECK-EX-NEXT: orq %rcx, %rdx -; CHECK-EX-NEXT: shrq $32, %rdx -; CHECK-EX-NEXT: je .LBB42_1 -; CHECK-EX-NEXT: # %bb.2: -; CHECK-EX-NEXT: cqto -; CHECK-EX-NEXT: idivq %rcx -; CHECK-EX-NEXT: movq %rdx, %rax -; CHECK-EX-NEXT: retq -; CHECK-EX-NEXT: .LBB42_1: -; CHECK-EX-NEXT: # kill: def $eax killed $eax killed $rax -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: divl %ecx -; CHECK-EX-NEXT: movl %edx, %eax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p1 unordered, align 8 %v2 = load atomic i64, i64* %p2 unordered, align 8 %ret = srem i64 %v, %v2 @@ -1261,19 +1043,6 @@ define i64 @load_fold_urem1(i64* %p) { ; CHECK-O3-NEXT: subq %rax, %rcx ; CHECK-O3-NEXT: movq %rcx, %rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_urem1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rcx -; CHECK-EX-NEXT: movabsq $-8608480567731124087, %rdx # imm = 0x8888888888888889 -; CHECK-EX-NEXT: movq %rcx, %rax -; CHECK-EX-NEXT: mulq %rdx -; CHECK-EX-NEXT: shrq $3, %rdx -; CHECK-EX-NEXT: leaq (%rdx,%rdx,4), %rax -; CHECK-EX-NEXT: leaq (%rax,%rax,2), %rax -; CHECK-EX-NEXT: subq %rax, %rcx -; CHECK-EX-NEXT: movq %rcx, %rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = urem i64 %v, 15 ret i64 %ret @@ -1308,25 +1077,6 @@ define i64 @load_fold_urem2(i64* %p, i64 %v2) { ; CHECK-O3-NEXT: divl %esi ; CHECK-O3-NEXT: movl %edx, %eax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_urem2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: movq %rax, %rcx -; CHECK-EX-NEXT: orq %rsi, %rcx -; CHECK-EX-NEXT: shrq $32, %rcx -; CHECK-EX-NEXT: je .LBB44_1 -; CHECK-EX-NEXT: # %bb.2: -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: divq %rsi -; CHECK-EX-NEXT: movq %rdx, %rax -; CHECK-EX-NEXT: retq -; CHECK-EX-NEXT: .LBB44_1: -; CHECK-EX-NEXT: # kill: def $eax killed $eax killed $rax -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: divl %esi -; CHECK-EX-NEXT: movl %edx, %eax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = urem i64 %v, %v2 ret i64 %ret @@ -1361,26 +1111,6 @@ define i64 @load_fold_urem3(i64* %p1, i64* %p2) { ; CHECK-O3-NEXT: divl %ecx ; CHECK-O3-NEXT: movl %edx, %eax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_urem3: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: movq (%rsi), %rcx -; CHECK-EX-NEXT: movq %rax, %rdx -; CHECK-EX-NEXT: orq %rcx, %rdx -; CHECK-EX-NEXT: shrq $32, %rdx -; CHECK-EX-NEXT: je .LBB45_1 -; CHECK-EX-NEXT: # %bb.2: -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: divq %rcx -; CHECK-EX-NEXT: movq %rdx, %rax -; CHECK-EX-NEXT: retq -; CHECK-EX-NEXT: .LBB45_1: -; CHECK-EX-NEXT: # kill: def $eax killed $eax killed $rax -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: divl %ecx -; CHECK-EX-NEXT: movl %edx, %eax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p1 unordered, align 8 %v2 = load atomic i64, i64* %p2 unordered, align 8 %ret = urem i64 %v, %v2 @@ -1412,11 +1142,6 @@ define i64 @load_fold_shl2(i64* %p, i64 %v2) { ; CHECK-O3: # %bb.0: ; CHECK-O3-NEXT: shlxq %rsi, (%rdi), %rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_shl2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: shlxq %rsi, (%rdi), %rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = shl i64 %v, %v2 ret i64 %ret @@ -1436,12 +1161,6 @@ define i64 @load_fold_shl3(i64* %p1, i64* %p2) { ; CHECK-O3-NEXT: movq (%rsi), %rax ; CHECK-O3-NEXT: shlxq %rax, (%rdi), %rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_shl3: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rsi), %rax -; CHECK-EX-NEXT: shlxq %rax, (%rdi), %rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p1 unordered, align 8 %v2 = load atomic i64, i64* %p2 unordered, align 8 %ret = shl i64 %v, %v2 @@ -1473,11 +1192,6 @@ define i64 @load_fold_lshr2(i64* %p, i64 %v2) { ; CHECK-O3: # %bb.0: ; CHECK-O3-NEXT: shrxq %rsi, (%rdi), %rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_lshr2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: shrxq %rsi, (%rdi), %rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = lshr i64 %v, %v2 ret i64 %ret @@ -1497,12 +1211,6 @@ define i64 @load_fold_lshr3(i64* %p1, i64* %p2) { ; CHECK-O3-NEXT: movq (%rsi), %rax ; CHECK-O3-NEXT: shrxq %rax, (%rdi), %rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_lshr3: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rsi), %rax -; CHECK-EX-NEXT: shrxq %rax, (%rdi), %rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p1 unordered, align 8 %v2 = load atomic i64, i64* %p2 unordered, align 8 %ret = lshr i64 %v, %v2 @@ -1534,11 +1242,6 @@ define i64 @load_fold_ashr2(i64* %p, i64 %v2) { ; CHECK-O3: # %bb.0: ; CHECK-O3-NEXT: sarxq %rsi, (%rdi), %rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_ashr2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: sarxq %rsi, (%rdi), %rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = ashr i64 %v, %v2 ret i64 %ret @@ -1558,12 +1261,6 @@ define i64 @load_fold_ashr3(i64* %p1, i64* %p2) { ; CHECK-O3-NEXT: movq (%rsi), %rax ; CHECK-O3-NEXT: sarxq %rax, (%rdi), %rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_ashr3: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rsi), %rax -; CHECK-EX-NEXT: sarxq %rax, (%rdi), %rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p1 unordered, align 8 %v2 = load atomic i64, i64* %p2 unordered, align 8 %ret = ashr i64 %v, %v2 @@ -1583,12 +1280,6 @@ define i64 @load_fold_and1(i64* %p) { ; CHECK-O3-NEXT: movq (%rdi), %rax ; CHECK-O3-NEXT: andl $15, %eax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_and1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: andl $15, %eax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = and i64 %v, 15 ret i64 %ret @@ -1606,12 +1297,6 @@ define i64 @load_fold_and2(i64* %p, i64 %v2) { ; CHECK-O3-NEXT: movq %rsi, %rax ; CHECK-O3-NEXT: andq (%rdi), %rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_and2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq %rsi, %rax -; CHECK-EX-NEXT: andq (%rdi), %rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = and i64 %v, %v2 ret i64 %ret @@ -1624,17 +1309,17 @@ define i64 @load_fold_and3(i64* %p1, i64* %p2) { ; CHECK-O0-NEXT: andq (%rsi), %rax ; CHECK-O0-NEXT: retq ; -; CHECK-O3-LABEL: load_fold_and3: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: movq (%rsi), %rax -; CHECK-O3-NEXT: andq (%rdi), %rax -; CHECK-O3-NEXT: retq +; CHECK-O3-CUR-LABEL: load_fold_and3: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movq (%rsi), %rax +; CHECK-O3-CUR-NEXT: andq (%rdi), %rax +; CHECK-O3-CUR-NEXT: retq ; -; CHECK-EX-LABEL: load_fold_and3: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: andq (%rsi), %rax -; CHECK-EX-NEXT: retq +; CHECK-O3-EX-LABEL: load_fold_and3: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: movq (%rdi), %rax +; CHECK-O3-EX-NEXT: andq (%rsi), %rax +; CHECK-O3-EX-NEXT: retq %v = load atomic i64, i64* %p1 unordered, align 8 %v2 = load atomic i64, i64* %p2 unordered, align 8 %ret = and i64 %v, %v2 @@ -1665,12 +1350,6 @@ define i64 @load_fold_or2(i64* %p, i64 %v2) { ; CHECK-O3-NEXT: movq %rsi, %rax ; CHECK-O3-NEXT: orq (%rdi), %rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_or2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq %rsi, %rax -; CHECK-EX-NEXT: orq (%rdi), %rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = or i64 %v, %v2 ret i64 %ret @@ -1683,17 +1362,17 @@ define i64 @load_fold_or3(i64* %p1, i64* %p2) { ; CHECK-O0-NEXT: orq (%rsi), %rax ; CHECK-O0-NEXT: retq ; -; CHECK-O3-LABEL: load_fold_or3: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: movq (%rsi), %rax -; CHECK-O3-NEXT: orq (%rdi), %rax -; CHECK-O3-NEXT: retq +; CHECK-O3-CUR-LABEL: load_fold_or3: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movq (%rsi), %rax +; CHECK-O3-CUR-NEXT: orq (%rdi), %rax +; CHECK-O3-CUR-NEXT: retq ; -; CHECK-EX-LABEL: load_fold_or3: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: orq (%rsi), %rax -; CHECK-EX-NEXT: retq +; CHECK-O3-EX-LABEL: load_fold_or3: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: movq (%rdi), %rax +; CHECK-O3-EX-NEXT: orq (%rsi), %rax +; CHECK-O3-EX-NEXT: retq %v = load atomic i64, i64* %p1 unordered, align 8 %v2 = load atomic i64, i64* %p2 unordered, align 8 %ret = or i64 %v, %v2 @@ -1724,12 +1403,6 @@ define i64 @load_fold_xor2(i64* %p, i64 %v2) { ; CHECK-O3-NEXT: movq %rsi, %rax ; CHECK-O3-NEXT: xorq (%rdi), %rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_xor2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq %rsi, %rax -; CHECK-EX-NEXT: xorq (%rdi), %rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = xor i64 %v, %v2 ret i64 %ret @@ -1742,17 +1415,17 @@ define i64 @load_fold_xor3(i64* %p1, i64* %p2) { ; CHECK-O0-NEXT: xorq (%rsi), %rax ; CHECK-O0-NEXT: retq ; -; CHECK-O3-LABEL: load_fold_xor3: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: movq (%rsi), %rax -; CHECK-O3-NEXT: xorq (%rdi), %rax -; CHECK-O3-NEXT: retq +; CHECK-O3-CUR-LABEL: load_fold_xor3: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movq (%rsi), %rax +; CHECK-O3-CUR-NEXT: xorq (%rdi), %rax +; CHECK-O3-CUR-NEXT: retq ; -; CHECK-EX-LABEL: load_fold_xor3: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: xorq (%rsi), %rax -; CHECK-EX-NEXT: retq +; CHECK-O3-EX-LABEL: load_fold_xor3: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: movq (%rdi), %rax +; CHECK-O3-EX-NEXT: xorq (%rsi), %rax +; CHECK-O3-EX-NEXT: retq %v = load atomic i64, i64* %p1 unordered, align 8 %v2 = load atomic i64, i64* %p2 unordered, align 8 %ret = xor i64 %v, %v2 @@ -1774,12 +1447,6 @@ define i1 @load_fold_icmp1(i64* %p) { ; CHECK-O3-NEXT: cmpq $15, (%rdi) ; CHECK-O3-NEXT: sete %al ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_icmp1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: cmpq $15, (%rdi) -; CHECK-EX-NEXT: sete %al -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = icmp eq i64 %v, 15 ret i1 %ret @@ -1800,12 +1467,6 @@ define i1 @load_fold_icmp2(i64* %p, i64 %v2) { ; CHECK-O3-NEXT: cmpq %rsi, (%rdi) ; CHECK-O3-NEXT: sete %al ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_fold_icmp2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: cmpq %rsi, (%rdi) -; CHECK-EX-NEXT: sete %al -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %ret = icmp eq i64 %v, %v2 ret i1 %ret @@ -1822,19 +1483,19 @@ define i1 @load_fold_icmp3(i64* %p1, i64* %p2) { ; CHECK-O0-NEXT: movb %dl, %al ; CHECK-O0-NEXT: retq ; -; CHECK-O3-LABEL: load_fold_icmp3: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: movq (%rsi), %rax -; CHECK-O3-NEXT: cmpq %rax, (%rdi) -; CHECK-O3-NEXT: sete %al -; CHECK-O3-NEXT: retq +; CHECK-O3-CUR-LABEL: load_fold_icmp3: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movq (%rsi), %rax +; CHECK-O3-CUR-NEXT: cmpq %rax, (%rdi) +; CHECK-O3-CUR-NEXT: sete %al +; CHECK-O3-CUR-NEXT: retq ; -; CHECK-EX-LABEL: load_fold_icmp3: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: cmpq (%rsi), %rax -; CHECK-EX-NEXT: sete %al -; CHECK-EX-NEXT: retq +; CHECK-O3-EX-LABEL: load_fold_icmp3: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: movq (%rdi), %rax +; CHECK-O3-EX-NEXT: cmpq (%rsi), %rax +; CHECK-O3-EX-NEXT: sete %al +; CHECK-O3-EX-NEXT: retq %v = load atomic i64, i64* %p1 unordered, align 8 %v2 = load atomic i64, i64* %p2 unordered, align 8 %ret = icmp eq i64 %v, %v2 @@ -1860,11 +1521,6 @@ define void @rmw_fold_add1(i64* %p, i64 %v) { ; CHECK-O3: # %bb.0: ; CHECK-O3-NEXT: addq $15, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_add1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: addq $15, (%rdi) -; CHECK-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = add i64 %prev, 15 store atomic i64 %val, i64* %p unordered, align 8 @@ -1884,11 +1540,6 @@ define void @rmw_fold_add2(i64* %p, i64 %v) { ; CHECK-O3: # %bb.0: ; CHECK-O3-NEXT: addq %rsi, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_add2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: addq %rsi, (%rdi) -; CHECK-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = add i64 %prev, %v store atomic i64 %val, i64* %p unordered, align 8 @@ -1908,11 +1559,6 @@ define void @rmw_fold_sub1(i64* %p, i64 %v) { ; CHECK-O3: # %bb.0: ; CHECK-O3-NEXT: addq $-15, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_sub1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: addq $-15, (%rdi) -; CHECK-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = sub i64 %prev, 15 store atomic i64 %val, i64* %p unordered, align 8 @@ -1932,11 +1578,6 @@ define void @rmw_fold_sub2(i64* %p, i64 %v) { ; CHECK-O3: # %bb.0: ; CHECK-O3-NEXT: subq %rsi, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_sub2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: subq %rsi, (%rdi) -; CHECK-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = sub i64 %prev, %v store atomic i64 %val, i64* %p unordered, align 8 @@ -1972,12 +1613,6 @@ define void @rmw_fold_mul2(i64* %p, i64 %v) { ; CHECK-O3-NEXT: imulq (%rdi), %rsi ; CHECK-O3-NEXT: movq %rsi, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_mul2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: imulq (%rdi), %rsi -; CHECK-EX-NEXT: movq %rsi, (%rdi) -; CHECK-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = mul i64 %prev, %v store atomic i64 %val, i64* %p unordered, align 8 @@ -2014,20 +1649,6 @@ define void @rmw_fold_sdiv1(i64* %p, i64 %v) { ; CHECK-O3-NEXT: addq %rax, %rdx ; CHECK-O3-NEXT: movq %rdx, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_sdiv1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rcx -; CHECK-EX-NEXT: movabsq $-8608480567731124087, %rdx # imm = 0x8888888888888889 -; CHECK-EX-NEXT: movq %rcx, %rax -; CHECK-EX-NEXT: imulq %rdx -; CHECK-EX-NEXT: addq %rcx, %rdx -; CHECK-EX-NEXT: movq %rdx, %rax -; CHECK-EX-NEXT: shrq $63, %rax -; CHECK-EX-NEXT: sarq $3, %rdx -; CHECK-EX-NEXT: addq %rax, %rdx -; CHECK-EX-NEXT: movq %rdx, (%rdi) -; CHECK-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = sdiv i64 %prev, 15 store atomic i64 %val, i64* %p unordered, align 8 @@ -2063,26 +1684,6 @@ define void @rmw_fold_sdiv2(i64* %p, i64 %v) { ; CHECK-O3-NEXT: # kill: def $eax killed $eax def $rax ; CHECK-O3-NEXT: movq %rax, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_sdiv2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: movq %rax, %rcx -; CHECK-EX-NEXT: orq %rsi, %rcx -; CHECK-EX-NEXT: shrq $32, %rcx -; CHECK-EX-NEXT: je .LBB74_1 -; CHECK-EX-NEXT: # %bb.2: -; CHECK-EX-NEXT: cqto -; CHECK-EX-NEXT: idivq %rsi -; CHECK-EX-NEXT: movq %rax, (%rdi) -; CHECK-EX-NEXT: retq -; CHECK-EX-NEXT: .LBB74_1: -; CHECK-EX-NEXT: # kill: def $eax killed $eax killed $rax -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: divl %esi -; CHECK-EX-NEXT: # kill: def $eax killed $eax def $rax -; CHECK-EX-NEXT: movq %rax, (%rdi) -; CHECK-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = sdiv i64 %prev, %v store atomic i64 %val, i64* %p unordered, align 8 @@ -2091,22 +1692,31 @@ define void @rmw_fold_sdiv2(i64* %p, i64 %v) { ; Legal, as expected define void @rmw_fold_udiv1(i64* %p, i64 %v) { -; CHECK-NOX-LABEL: rmw_fold_udiv1: -; CHECK-NOX: # %bb.0: -; CHECK-NOX-NEXT: movq (%rdi), %rax -; CHECK-NOX-NEXT: movabsq $-8608480567731124087, %rcx # imm = 0x8888888888888889 -; CHECK-NOX-NEXT: mulq %rcx -; CHECK-NOX-NEXT: shrq $3, %rdx -; CHECK-NOX-NEXT: movq %rdx, (%rdi) -; CHECK-NOX-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_udiv1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movabsq $-8608480567731124087, %rax # imm = 0x8888888888888889 -; CHECK-EX-NEXT: mulq (%rdi) -; CHECK-EX-NEXT: shrq $3, %rdx -; CHECK-EX-NEXT: movq %rdx, (%rdi) -; CHECK-EX-NEXT: retq +; CHECK-O0-LABEL: rmw_fold_udiv1: +; CHECK-O0: # %bb.0: +; CHECK-O0-NEXT: movq (%rdi), %rax +; CHECK-O0-NEXT: movabsq $-8608480567731124087, %rcx # imm = 0x8888888888888889 +; CHECK-O0-NEXT: mulq %rcx +; CHECK-O0-NEXT: shrq $3, %rdx +; CHECK-O0-NEXT: movq %rdx, (%rdi) +; CHECK-O0-NEXT: retq +; +; CHECK-O3-CUR-LABEL: rmw_fold_udiv1: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movq (%rdi), %rax +; CHECK-O3-CUR-NEXT: movabsq $-8608480567731124087, %rcx # imm = 0x8888888888888889 +; CHECK-O3-CUR-NEXT: mulq %rcx +; CHECK-O3-CUR-NEXT: shrq $3, %rdx +; CHECK-O3-CUR-NEXT: movq %rdx, (%rdi) +; CHECK-O3-CUR-NEXT: retq +; +; CHECK-O3-EX-LABEL: rmw_fold_udiv1: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: movabsq $-8608480567731124087, %rax # imm = 0x8888888888888889 +; CHECK-O3-EX-NEXT: mulq (%rdi) +; CHECK-O3-EX-NEXT: shrq $3, %rdx +; CHECK-O3-EX-NEXT: movq %rdx, (%rdi) +; CHECK-O3-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = udiv i64 %prev, 15 store atomic i64 %val, i64* %p unordered, align 8 @@ -2143,26 +1753,6 @@ define void @rmw_fold_udiv2(i64* %p, i64 %v) { ; CHECK-O3-NEXT: # kill: def $eax killed $eax def $rax ; CHECK-O3-NEXT: movq %rax, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_udiv2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: movq %rax, %rcx -; CHECK-EX-NEXT: orq %rsi, %rcx -; CHECK-EX-NEXT: shrq $32, %rcx -; CHECK-EX-NEXT: je .LBB76_1 -; CHECK-EX-NEXT: # %bb.2: -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: divq %rsi -; CHECK-EX-NEXT: movq %rax, (%rdi) -; CHECK-EX-NEXT: retq -; CHECK-EX-NEXT: .LBB76_1: -; CHECK-EX-NEXT: # kill: def $eax killed $eax killed $rax -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: divl %esi -; CHECK-EX-NEXT: # kill: def $eax killed $eax def $rax -; CHECK-EX-NEXT: movq %rax, (%rdi) -; CHECK-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = udiv i64 %prev, %v store atomic i64 %val, i64* %p unordered, align 8 @@ -2205,23 +1795,6 @@ define void @rmw_fold_srem1(i64* %p, i64 %v) { ; CHECK-O3-NEXT: subq %rax, %rcx ; CHECK-O3-NEXT: movq %rcx, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_srem1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rcx -; CHECK-EX-NEXT: movabsq $-8608480567731124087, %rdx # imm = 0x8888888888888889 -; CHECK-EX-NEXT: movq %rcx, %rax -; CHECK-EX-NEXT: imulq %rdx -; CHECK-EX-NEXT: addq %rcx, %rdx -; CHECK-EX-NEXT: movq %rdx, %rax -; CHECK-EX-NEXT: shrq $63, %rax -; CHECK-EX-NEXT: sarq $3, %rdx -; CHECK-EX-NEXT: addq %rax, %rdx -; CHECK-EX-NEXT: leaq (%rdx,%rdx,4), %rax -; CHECK-EX-NEXT: leaq (%rax,%rax,2), %rax -; CHECK-EX-NEXT: subq %rax, %rcx -; CHECK-EX-NEXT: movq %rcx, (%rdi) -; CHECK-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = srem i64 %prev, 15 store atomic i64 %val, i64* %p unordered, align 8 @@ -2257,26 +1830,6 @@ define void @rmw_fold_srem2(i64* %p, i64 %v) { ; CHECK-O3-NEXT: # kill: def $edx killed $edx def $rdx ; CHECK-O3-NEXT: movq %rdx, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_srem2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: movq %rax, %rcx -; CHECK-EX-NEXT: orq %rsi, %rcx -; CHECK-EX-NEXT: shrq $32, %rcx -; CHECK-EX-NEXT: je .LBB78_1 -; CHECK-EX-NEXT: # %bb.2: -; CHECK-EX-NEXT: cqto -; CHECK-EX-NEXT: idivq %rsi -; CHECK-EX-NEXT: movq %rdx, (%rdi) -; CHECK-EX-NEXT: retq -; CHECK-EX-NEXT: .LBB78_1: -; CHECK-EX-NEXT: # kill: def $eax killed $eax killed $rax -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: divl %esi -; CHECK-EX-NEXT: # kill: def $edx killed $edx def $rdx -; CHECK-EX-NEXT: movq %rdx, (%rdi) -; CHECK-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = srem i64 %prev, %v store atomic i64 %val, i64* %p unordered, align 8 @@ -2311,19 +1864,6 @@ define void @rmw_fold_urem1(i64* %p, i64 %v) { ; CHECK-O3-NEXT: subq %rax, %rcx ; CHECK-O3-NEXT: movq %rcx, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_urem1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rcx -; CHECK-EX-NEXT: movabsq $-8608480567731124087, %rdx # imm = 0x8888888888888889 -; CHECK-EX-NEXT: movq %rcx, %rax -; CHECK-EX-NEXT: mulq %rdx -; CHECK-EX-NEXT: shrq $3, %rdx -; CHECK-EX-NEXT: leaq (%rdx,%rdx,4), %rax -; CHECK-EX-NEXT: leaq (%rax,%rax,2), %rax -; CHECK-EX-NEXT: subq %rax, %rcx -; CHECK-EX-NEXT: movq %rcx, (%rdi) -; CHECK-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = urem i64 %prev, 15 store atomic i64 %val, i64* %p unordered, align 8 @@ -2360,26 +1900,6 @@ define void @rmw_fold_urem2(i64* %p, i64 %v) { ; CHECK-O3-NEXT: # kill: def $edx killed $edx def $rdx ; CHECK-O3-NEXT: movq %rdx, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_urem2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: movq %rax, %rcx -; CHECK-EX-NEXT: orq %rsi, %rcx -; CHECK-EX-NEXT: shrq $32, %rcx -; CHECK-EX-NEXT: je .LBB80_1 -; CHECK-EX-NEXT: # %bb.2: -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: divq %rsi -; CHECK-EX-NEXT: movq %rdx, (%rdi) -; CHECK-EX-NEXT: retq -; CHECK-EX-NEXT: .LBB80_1: -; CHECK-EX-NEXT: # kill: def $eax killed $eax killed $rax -; CHECK-EX-NEXT: xorl %edx, %edx -; CHECK-EX-NEXT: divl %esi -; CHECK-EX-NEXT: # kill: def $edx killed $edx def $rdx -; CHECK-EX-NEXT: movq %rdx, (%rdi) -; CHECK-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = urem i64 %prev, %v store atomic i64 %val, i64* %p unordered, align 8 @@ -2388,17 +1908,24 @@ define void @rmw_fold_urem2(i64* %p, i64 %v) { ; Legal to fold (TODO) define void @rmw_fold_shl1(i64* %p, i64 %v) { -; CHECK-NOX-LABEL: rmw_fold_shl1: -; CHECK-NOX: # %bb.0: -; CHECK-NOX-NEXT: movq (%rdi), %rax -; CHECK-NOX-NEXT: shlq $15, %rax -; CHECK-NOX-NEXT: movq %rax, (%rdi) -; CHECK-NOX-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_shl1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: shlq $15, (%rdi) -; CHECK-EX-NEXT: retq +; CHECK-O0-LABEL: rmw_fold_shl1: +; CHECK-O0: # %bb.0: +; CHECK-O0-NEXT: movq (%rdi), %rax +; CHECK-O0-NEXT: shlq $15, %rax +; CHECK-O0-NEXT: movq %rax, (%rdi) +; CHECK-O0-NEXT: retq +; +; CHECK-O3-CUR-LABEL: rmw_fold_shl1: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movq (%rdi), %rax +; CHECK-O3-CUR-NEXT: shlq $15, %rax +; CHECK-O3-CUR-NEXT: movq %rax, (%rdi) +; CHECK-O3-CUR-NEXT: retq +; +; CHECK-O3-EX-LABEL: rmw_fold_shl1: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: shlq $15, (%rdi) +; CHECK-O3-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = shl i64 %prev, 15 store atomic i64 %val, i64* %p unordered, align 8 @@ -2417,18 +1944,18 @@ define void @rmw_fold_shl2(i64* %p, i64 %v) { ; CHECK-O0-NEXT: movq %rax, (%rdi) ; CHECK-O0-NEXT: retq ; -; CHECK-O3-LABEL: rmw_fold_shl2: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: shlxq %rsi, (%rdi), %rax -; CHECK-O3-NEXT: movq %rax, (%rdi) -; CHECK-O3-NEXT: retq +; CHECK-O3-CUR-LABEL: rmw_fold_shl2: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: shlxq %rsi, (%rdi), %rax +; CHECK-O3-CUR-NEXT: movq %rax, (%rdi) +; CHECK-O3-CUR-NEXT: retq ; -; CHECK-EX-LABEL: rmw_fold_shl2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq %rsi, %rcx -; CHECK-EX-NEXT: # kill: def $cl killed $cl killed $rcx -; CHECK-EX-NEXT: shlq %cl, (%rdi) -; CHECK-EX-NEXT: retq +; CHECK-O3-EX-LABEL: rmw_fold_shl2: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: movq %rsi, %rcx +; CHECK-O3-EX-NEXT: # kill: def $cl killed $cl killed $rcx +; CHECK-O3-EX-NEXT: shlq %cl, (%rdi) +; CHECK-O3-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = shl i64 %prev, %v store atomic i64 %val, i64* %p unordered, align 8 @@ -2437,17 +1964,24 @@ define void @rmw_fold_shl2(i64* %p, i64 %v) { ; Legal to fold (TODO) define void @rmw_fold_lshr1(i64* %p, i64 %v) { -; CHECK-NOX-LABEL: rmw_fold_lshr1: -; CHECK-NOX: # %bb.0: -; CHECK-NOX-NEXT: movq (%rdi), %rax -; CHECK-NOX-NEXT: shrq $15, %rax -; CHECK-NOX-NEXT: movq %rax, (%rdi) -; CHECK-NOX-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_lshr1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: shrq $15, (%rdi) -; CHECK-EX-NEXT: retq +; CHECK-O0-LABEL: rmw_fold_lshr1: +; CHECK-O0: # %bb.0: +; CHECK-O0-NEXT: movq (%rdi), %rax +; CHECK-O0-NEXT: shrq $15, %rax +; CHECK-O0-NEXT: movq %rax, (%rdi) +; CHECK-O0-NEXT: retq +; +; CHECK-O3-CUR-LABEL: rmw_fold_lshr1: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movq (%rdi), %rax +; CHECK-O3-CUR-NEXT: shrq $15, %rax +; CHECK-O3-CUR-NEXT: movq %rax, (%rdi) +; CHECK-O3-CUR-NEXT: retq +; +; CHECK-O3-EX-LABEL: rmw_fold_lshr1: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: shrq $15, (%rdi) +; CHECK-O3-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = lshr i64 %prev, 15 store atomic i64 %val, i64* %p unordered, align 8 @@ -2466,18 +2000,18 @@ define void @rmw_fold_lshr2(i64* %p, i64 %v) { ; CHECK-O0-NEXT: movq %rax, (%rdi) ; CHECK-O0-NEXT: retq ; -; CHECK-O3-LABEL: rmw_fold_lshr2: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: shrxq %rsi, (%rdi), %rax -; CHECK-O3-NEXT: movq %rax, (%rdi) -; CHECK-O3-NEXT: retq +; CHECK-O3-CUR-LABEL: rmw_fold_lshr2: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: shrxq %rsi, (%rdi), %rax +; CHECK-O3-CUR-NEXT: movq %rax, (%rdi) +; CHECK-O3-CUR-NEXT: retq ; -; CHECK-EX-LABEL: rmw_fold_lshr2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq %rsi, %rcx -; CHECK-EX-NEXT: # kill: def $cl killed $cl killed $rcx -; CHECK-EX-NEXT: shrq %cl, (%rdi) -; CHECK-EX-NEXT: retq +; CHECK-O3-EX-LABEL: rmw_fold_lshr2: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: movq %rsi, %rcx +; CHECK-O3-EX-NEXT: # kill: def $cl killed $cl killed $rcx +; CHECK-O3-EX-NEXT: shrq %cl, (%rdi) +; CHECK-O3-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = lshr i64 %prev, %v store atomic i64 %val, i64* %p unordered, align 8 @@ -2486,17 +2020,24 @@ define void @rmw_fold_lshr2(i64* %p, i64 %v) { ; Legal to fold (TODO) define void @rmw_fold_ashr1(i64* %p, i64 %v) { -; CHECK-NOX-LABEL: rmw_fold_ashr1: -; CHECK-NOX: # %bb.0: -; CHECK-NOX-NEXT: movq (%rdi), %rax -; CHECK-NOX-NEXT: sarq $15, %rax -; CHECK-NOX-NEXT: movq %rax, (%rdi) -; CHECK-NOX-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_ashr1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: sarq $15, (%rdi) -; CHECK-EX-NEXT: retq +; CHECK-O0-LABEL: rmw_fold_ashr1: +; CHECK-O0: # %bb.0: +; CHECK-O0-NEXT: movq (%rdi), %rax +; CHECK-O0-NEXT: sarq $15, %rax +; CHECK-O0-NEXT: movq %rax, (%rdi) +; CHECK-O0-NEXT: retq +; +; CHECK-O3-CUR-LABEL: rmw_fold_ashr1: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movq (%rdi), %rax +; CHECK-O3-CUR-NEXT: sarq $15, %rax +; CHECK-O3-CUR-NEXT: movq %rax, (%rdi) +; CHECK-O3-CUR-NEXT: retq +; +; CHECK-O3-EX-LABEL: rmw_fold_ashr1: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: sarq $15, (%rdi) +; CHECK-O3-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = ashr i64 %prev, 15 store atomic i64 %val, i64* %p unordered, align 8 @@ -2515,18 +2056,18 @@ define void @rmw_fold_ashr2(i64* %p, i64 %v) { ; CHECK-O0-NEXT: movq %rax, (%rdi) ; CHECK-O0-NEXT: retq ; -; CHECK-O3-LABEL: rmw_fold_ashr2: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: sarxq %rsi, (%rdi), %rax -; CHECK-O3-NEXT: movq %rax, (%rdi) -; CHECK-O3-NEXT: retq +; CHECK-O3-CUR-LABEL: rmw_fold_ashr2: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: sarxq %rsi, (%rdi), %rax +; CHECK-O3-CUR-NEXT: movq %rax, (%rdi) +; CHECK-O3-CUR-NEXT: retq ; -; CHECK-EX-LABEL: rmw_fold_ashr2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq %rsi, %rcx -; CHECK-EX-NEXT: # kill: def $cl killed $cl killed $rcx -; CHECK-EX-NEXT: sarq %cl, (%rdi) -; CHECK-EX-NEXT: retq +; CHECK-O3-EX-LABEL: rmw_fold_ashr2: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: movq %rsi, %rcx +; CHECK-O3-EX-NEXT: # kill: def $cl killed $cl killed $rcx +; CHECK-O3-EX-NEXT: sarq %cl, (%rdi) +; CHECK-O3-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = ashr i64 %prev, %v store atomic i64 %val, i64* %p unordered, align 8 @@ -2548,11 +2089,6 @@ define void @rmw_fold_and1(i64* %p, i64 %v) { ; CHECK-O3: # %bb.0: ; CHECK-O3-NEXT: andq $15, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_and1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: andq $15, (%rdi) -; CHECK-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = and i64 %prev, 15 store atomic i64 %val, i64* %p unordered, align 8 @@ -2572,11 +2108,6 @@ define void @rmw_fold_and2(i64* %p, i64 %v) { ; CHECK-O3: # %bb.0: ; CHECK-O3-NEXT: andq %rsi, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_and2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: andq %rsi, (%rdi) -; CHECK-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = and i64 %prev, %v store atomic i64 %val, i64* %p unordered, align 8 @@ -2596,11 +2127,6 @@ define void @rmw_fold_or1(i64* %p, i64 %v) { ; CHECK-O3: # %bb.0: ; CHECK-O3-NEXT: orq $15, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_or1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: orq $15, (%rdi) -; CHECK-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = or i64 %prev, 15 store atomic i64 %val, i64* %p unordered, align 8 @@ -2620,11 +2146,6 @@ define void @rmw_fold_or2(i64* %p, i64 %v) { ; CHECK-O3: # %bb.0: ; CHECK-O3-NEXT: orq %rsi, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_or2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: orq %rsi, (%rdi) -; CHECK-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = or i64 %prev, %v store atomic i64 %val, i64* %p unordered, align 8 @@ -2644,11 +2165,6 @@ define void @rmw_fold_xor1(i64* %p, i64 %v) { ; CHECK-O3: # %bb.0: ; CHECK-O3-NEXT: xorq $15, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_xor1: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: xorq $15, (%rdi) -; CHECK-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = xor i64 %prev, 15 store atomic i64 %val, i64* %p unordered, align 8 @@ -2668,11 +2184,6 @@ define void @rmw_fold_xor2(i64* %p, i64 %v) { ; CHECK-O3: # %bb.0: ; CHECK-O3-NEXT: xorq %rsi, (%rdi) ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: rmw_fold_xor2: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: xorq %rsi, (%rdi) -; CHECK-EX-NEXT: retq %prev = load atomic i64, i64* %p unordered, align 8 %val = xor i64 %prev, %v store atomic i64 %val, i64* %p unordered, align 8 @@ -2709,13 +2220,6 @@ define i32 @fold_trunc_add(i64* %p, i32 %v2) { ; CHECK-O3-NEXT: addl %esi, %eax ; CHECK-O3-NEXT: # kill: def $eax killed $eax killed $rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: fold_trunc_add: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: addl %esi, %eax -; CHECK-EX-NEXT: # kill: def $eax killed $eax killed $rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %trunc = trunc i64 %v to i32 %ret = add i32 %trunc, %v2 @@ -2737,13 +2241,6 @@ define i32 @fold_trunc_and(i64* %p, i32 %v2) { ; CHECK-O3-NEXT: andl %esi, %eax ; CHECK-O3-NEXT: # kill: def $eax killed $eax killed $rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: fold_trunc_and: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: andl %esi, %eax -; CHECK-EX-NEXT: # kill: def $eax killed $eax killed $rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %trunc = trunc i64 %v to i32 %ret = and i32 %trunc, %v2 @@ -2765,13 +2262,6 @@ define i32 @fold_trunc_or(i64* %p, i32 %v2) { ; CHECK-O3-NEXT: orl %esi, %eax ; CHECK-O3-NEXT: # kill: def $eax killed $eax killed $rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: fold_trunc_or: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: orl %esi, %eax -; CHECK-EX-NEXT: # kill: def $eax killed $eax killed $rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %trunc = trunc i64 %v to i32 %ret = or i32 %trunc, %v2 @@ -2799,15 +2289,6 @@ define i32 @split_load(i64* %p) { ; CHECK-O3-NEXT: orl %eax, %ecx ; CHECK-O3-NEXT: movzbl %cl, %eax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: split_load: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: movq %rax, %rcx -; CHECK-EX-NEXT: shrq $32, %rcx -; CHECK-EX-NEXT: orl %eax, %ecx -; CHECK-EX-NEXT: movzbl %cl, %eax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %b1 = trunc i64 %v to i8 %v.shift = lshr i64 %v, 32 @@ -2834,16 +2315,22 @@ define i64 @constant_folding(i64* %p) { ; Legal to forward and fold (TODO) define i64 @load_forwarding(i64* %p) { -; CHECK-NOX-LABEL: load_forwarding: -; CHECK-NOX: # %bb.0: -; CHECK-NOX-NEXT: movq (%rdi), %rax -; CHECK-NOX-NEXT: orq (%rdi), %rax -; CHECK-NOX-NEXT: retq -; -; CHECK-EX-LABEL: load_forwarding: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq (%rdi), %rax -; CHECK-EX-NEXT: retq +; CHECK-O0-LABEL: load_forwarding: +; CHECK-O0: # %bb.0: +; CHECK-O0-NEXT: movq (%rdi), %rax +; CHECK-O0-NEXT: orq (%rdi), %rax +; CHECK-O0-NEXT: retq +; +; CHECK-O3-CUR-LABEL: load_forwarding: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movq (%rdi), %rax +; CHECK-O3-CUR-NEXT: orq (%rdi), %rax +; CHECK-O3-CUR-NEXT: retq +; +; CHECK-O3-EX-LABEL: load_forwarding: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: movq (%rdi), %rax +; CHECK-O3-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8 %v2 = load atomic i64, i64* %p unordered, align 8 %ret = or i64 %v, %v2 @@ -2949,31 +2436,32 @@ define i64 @fold_constant(i64 %arg) { ; CHECK-O3-NEXT: movq %rdi, %rax ; CHECK-O3-NEXT: addq {{.*}}(%rip), %rax ; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: fold_constant: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq %rdi, %rax -; CHECK-EX-NEXT: addq {{.*}}(%rip), %rax -; CHECK-EX-NEXT: retq %v = load atomic i64, i64* @Constant unordered, align 8 %ret = add i64 %v, %arg ret i64 %ret } define i64 @fold_constant_clobber(i64* %p, i64 %arg) { -; CHECK-NOX-LABEL: fold_constant_clobber: -; CHECK-NOX: # %bb.0: -; CHECK-NOX-NEXT: movq {{.*}}(%rip), %rax -; CHECK-NOX-NEXT: movq $5, (%rdi) -; CHECK-NOX-NEXT: addq %rsi, %rax -; CHECK-NOX-NEXT: retq -; -; CHECK-EX-LABEL: fold_constant_clobber: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq %rsi, %rax -; CHECK-EX-NEXT: movq $5, (%rdi) -; CHECK-EX-NEXT: addq {{.*}}(%rip), %rax -; CHECK-EX-NEXT: retq +; CHECK-O0-LABEL: fold_constant_clobber: +; CHECK-O0: # %bb.0: +; CHECK-O0-NEXT: movq {{.*}}(%rip), %rax +; CHECK-O0-NEXT: movq $5, (%rdi) +; CHECK-O0-NEXT: addq %rsi, %rax +; CHECK-O0-NEXT: retq +; +; CHECK-O3-CUR-LABEL: fold_constant_clobber: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movq {{.*}}(%rip), %rax +; CHECK-O3-CUR-NEXT: movq $5, (%rdi) +; CHECK-O3-CUR-NEXT: addq %rsi, %rax +; CHECK-O3-CUR-NEXT: retq +; +; CHECK-O3-EX-LABEL: fold_constant_clobber: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: movq %rsi, %rax +; CHECK-O3-EX-NEXT: movq $5, (%rdi) +; CHECK-O3-EX-NEXT: addq {{.*}}(%rip), %rax +; CHECK-O3-EX-NEXT: retq %v = load atomic i64, i64* @Constant unordered, align 8 store i64 5, i64* %p %ret = add i64 %v, %arg @@ -2981,19 +2469,26 @@ define i64 @fold_constant_clobber(i64* %p, i64 %arg) { } define i64 @fold_constant_fence(i64 %arg) { -; CHECK-NOX-LABEL: fold_constant_fence: -; CHECK-NOX: # %bb.0: -; CHECK-NOX-NEXT: movq {{.*}}(%rip), %rax -; CHECK-NOX-NEXT: mfence -; CHECK-NOX-NEXT: addq %rdi, %rax -; CHECK-NOX-NEXT: retq -; -; CHECK-EX-LABEL: fold_constant_fence: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq %rdi, %rax -; CHECK-EX-NEXT: mfence -; CHECK-EX-NEXT: addq {{.*}}(%rip), %rax -; CHECK-EX-NEXT: retq +; CHECK-O0-LABEL: fold_constant_fence: +; CHECK-O0: # %bb.0: +; CHECK-O0-NEXT: movq {{.*}}(%rip), %rax +; CHECK-O0-NEXT: mfence +; CHECK-O0-NEXT: addq %rdi, %rax +; CHECK-O0-NEXT: retq +; +; CHECK-O3-CUR-LABEL: fold_constant_fence: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movq {{.*}}(%rip), %rax +; CHECK-O3-CUR-NEXT: mfence +; CHECK-O3-CUR-NEXT: addq %rdi, %rax +; CHECK-O3-CUR-NEXT: retq +; +; CHECK-O3-EX-LABEL: fold_constant_fence: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: movq %rdi, %rax +; CHECK-O3-EX-NEXT: mfence +; CHECK-O3-EX-NEXT: addq {{.*}}(%rip), %rax +; CHECK-O3-EX-NEXT: retq %v = load atomic i64, i64* @Constant unordered, align 8 fence seq_cst %ret = add i64 %v, %arg @@ -3001,19 +2496,26 @@ define i64 @fold_constant_fence(i64 %arg) { } define i64 @fold_invariant_clobber(i64* dereferenceable(8) %p, i64 %arg) { -; CHECK-NOX-LABEL: fold_invariant_clobber: -; CHECK-NOX: # %bb.0: -; CHECK-NOX-NEXT: movq (%rdi), %rax -; CHECK-NOX-NEXT: movq $5, (%rdi) -; CHECK-NOX-NEXT: addq %rsi, %rax -; CHECK-NOX-NEXT: retq -; -; CHECK-EX-LABEL: fold_invariant_clobber: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq %rsi, %rax -; CHECK-EX-NEXT: movq $5, (%rdi) -; CHECK-EX-NEXT: addq (%rdi), %rax -; CHECK-EX-NEXT: retq +; CHECK-O0-LABEL: fold_invariant_clobber: +; CHECK-O0: # %bb.0: +; CHECK-O0-NEXT: movq (%rdi), %rax +; CHECK-O0-NEXT: movq $5, (%rdi) +; CHECK-O0-NEXT: addq %rsi, %rax +; CHECK-O0-NEXT: retq +; +; CHECK-O3-CUR-LABEL: fold_invariant_clobber: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movq (%rdi), %rax +; CHECK-O3-CUR-NEXT: movq $5, (%rdi) +; CHECK-O3-CUR-NEXT: addq %rsi, %rax +; CHECK-O3-CUR-NEXT: retq +; +; CHECK-O3-EX-LABEL: fold_invariant_clobber: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: movq %rsi, %rax +; CHECK-O3-EX-NEXT: movq $5, (%rdi) +; CHECK-O3-EX-NEXT: addq (%rdi), %rax +; CHECK-O3-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8, !invariant.load !{} store i64 5, i64* %p %ret = add i64 %v, %arg @@ -3022,19 +2524,26 @@ define i64 @fold_invariant_clobber(i64* dereferenceable(8) %p, i64 %arg) { define i64 @fold_invariant_fence(i64* dereferenceable(8) %p, i64 %arg) { -; CHECK-NOX-LABEL: fold_invariant_fence: -; CHECK-NOX: # %bb.0: -; CHECK-NOX-NEXT: movq (%rdi), %rax -; CHECK-NOX-NEXT: mfence -; CHECK-NOX-NEXT: addq %rsi, %rax -; CHECK-NOX-NEXT: retq -; -; CHECK-EX-LABEL: fold_invariant_fence: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: movq %rsi, %rax -; CHECK-EX-NEXT: mfence -; CHECK-EX-NEXT: addq (%rdi), %rax -; CHECK-EX-NEXT: retq +; CHECK-O0-LABEL: fold_invariant_fence: +; CHECK-O0: # %bb.0: +; CHECK-O0-NEXT: movq (%rdi), %rax +; CHECK-O0-NEXT: mfence +; CHECK-O0-NEXT: addq %rsi, %rax +; CHECK-O0-NEXT: retq +; +; CHECK-O3-CUR-LABEL: fold_invariant_fence: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movq (%rdi), %rax +; CHECK-O3-CUR-NEXT: mfence +; CHECK-O3-CUR-NEXT: addq %rsi, %rax +; CHECK-O3-CUR-NEXT: retq +; +; CHECK-O3-EX-LABEL: fold_invariant_fence: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: movq %rsi, %rax +; CHECK-O3-EX-NEXT: mfence +; CHECK-O3-EX-NEXT: addq (%rdi), %rax +; CHECK-O3-EX-NEXT: retq %v = load atomic i64, i64* %p unordered, align 8, !invariant.load !{} fence seq_cst %ret = add i64 %v, %arg @@ -3045,26 +2554,33 @@ define i64 @fold_invariant_fence(i64* dereferenceable(8) %p, i64 %arg) { ; Exercise a few cases involving any extend idioms define i16 @load_i8_anyext_i16(i8* %ptr) { -; CHECK-O0-LABEL: load_i8_anyext_i16: -; CHECK-O0: # %bb.0: -; CHECK-O0-NEXT: movb (%rdi), %al -; CHECK-O0-NEXT: movzbl %al, %ecx -; CHECK-O0-NEXT: # kill: def $cx killed $cx killed $ecx -; CHECK-O0-NEXT: movw %cx, %ax -; CHECK-O0-NEXT: retq -; -; CHECK-O3-LABEL: load_i8_anyext_i16: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: movzbl (%rdi), %eax -; CHECK-O3-NEXT: # kill: def $ax killed $ax killed $eax -; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_i8_anyext_i16: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: vpbroadcastb (%rdi), %xmm0 -; CHECK-EX-NEXT: vmovd %xmm0, %eax -; CHECK-EX-NEXT: # kill: def $ax killed $ax killed $eax -; CHECK-EX-NEXT: retq +; CHECK-O0-CUR-LABEL: load_i8_anyext_i16: +; CHECK-O0-CUR: # %bb.0: +; CHECK-O0-CUR-NEXT: movb (%rdi), %al +; CHECK-O0-CUR-NEXT: movzbl %al, %ecx +; CHECK-O0-CUR-NEXT: # kill: def $cx killed $cx killed $ecx +; CHECK-O0-CUR-NEXT: movw %cx, %ax +; CHECK-O0-CUR-NEXT: retq +; +; CHECK-O3-CUR-LABEL: load_i8_anyext_i16: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movzbl (%rdi), %eax +; CHECK-O3-CUR-NEXT: # kill: def $ax killed $ax killed $eax +; CHECK-O3-CUR-NEXT: retq +; +; CHECK-O0-EX-LABEL: load_i8_anyext_i16: +; CHECK-O0-EX: # %bb.0: +; CHECK-O0-EX-NEXT: vpbroadcastb (%rdi), %xmm0 +; CHECK-O0-EX-NEXT: vmovd %xmm0, %eax +; CHECK-O0-EX-NEXT: # kill: def $ax killed $ax killed $eax +; CHECK-O0-EX-NEXT: retq +; +; CHECK-O3-EX-LABEL: load_i8_anyext_i16: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: vpbroadcastb (%rdi), %xmm0 +; CHECK-O3-EX-NEXT: vmovd %xmm0, %eax +; CHECK-O3-EX-NEXT: # kill: def $ax killed $ax killed $eax +; CHECK-O3-EX-NEXT: retq %v = load atomic i8, i8* %ptr unordered, align 2 %vec = insertelement <2 x i8> undef, i8 %v, i32 0 %res = bitcast <2 x i8> %vec to i16 @@ -3072,22 +2588,28 @@ define i16 @load_i8_anyext_i16(i8* %ptr) { } define i32 @load_i8_anyext_i32(i8* %ptr) { -; CHECK-O0-LABEL: load_i8_anyext_i32: -; CHECK-O0: # %bb.0: -; CHECK-O0-NEXT: movb (%rdi), %al -; CHECK-O0-NEXT: movzbl %al, %eax -; CHECK-O0-NEXT: retq -; -; CHECK-O3-LABEL: load_i8_anyext_i32: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: movzbl (%rdi), %eax -; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_i8_anyext_i32: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: vpbroadcastb (%rdi), %xmm0 -; CHECK-EX-NEXT: vmovd %xmm0, %eax -; CHECK-EX-NEXT: retq +; CHECK-O0-CUR-LABEL: load_i8_anyext_i32: +; CHECK-O0-CUR: # %bb.0: +; CHECK-O0-CUR-NEXT: movb (%rdi), %al +; CHECK-O0-CUR-NEXT: movzbl %al, %eax +; CHECK-O0-CUR-NEXT: retq +; +; CHECK-O3-CUR-LABEL: load_i8_anyext_i32: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movzbl (%rdi), %eax +; CHECK-O3-CUR-NEXT: retq +; +; CHECK-O0-EX-LABEL: load_i8_anyext_i32: +; CHECK-O0-EX: # %bb.0: +; CHECK-O0-EX-NEXT: vpbroadcastb (%rdi), %xmm0 +; CHECK-O0-EX-NEXT: vmovd %xmm0, %eax +; CHECK-O0-EX-NEXT: retq +; +; CHECK-O3-EX-LABEL: load_i8_anyext_i32: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: vpbroadcastb (%rdi), %xmm0 +; CHECK-O3-EX-NEXT: vmovd %xmm0, %eax +; CHECK-O3-EX-NEXT: retq %v = load atomic i8, i8* %ptr unordered, align 4 %vec = insertelement <4 x i8> undef, i8 %v, i32 0 %res = bitcast <4 x i8> %vec to i32 @@ -3095,24 +2617,30 @@ define i32 @load_i8_anyext_i32(i8* %ptr) { } define i32 @load_i16_anyext_i32(i16* %ptr) { -; CHECK-O0-LABEL: load_i16_anyext_i32: -; CHECK-O0: # %bb.0: -; CHECK-O0-NEXT: movw (%rdi), %ax -; CHECK-O0-NEXT: # implicit-def: $ecx -; CHECK-O0-NEXT: movw %ax, %cx -; CHECK-O0-NEXT: movl %ecx, %eax -; CHECK-O0-NEXT: retq -; -; CHECK-O3-LABEL: load_i16_anyext_i32: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: movzwl (%rdi), %eax -; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_i16_anyext_i32: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: vpbroadcastw (%rdi), %xmm0 -; CHECK-EX-NEXT: vmovd %xmm0, %eax -; CHECK-EX-NEXT: retq +; CHECK-O0-CUR-LABEL: load_i16_anyext_i32: +; CHECK-O0-CUR: # %bb.0: +; CHECK-O0-CUR-NEXT: movw (%rdi), %ax +; CHECK-O0-CUR-NEXT: # implicit-def: $ecx +; CHECK-O0-CUR-NEXT: movw %ax, %cx +; CHECK-O0-CUR-NEXT: movl %ecx, %eax +; CHECK-O0-CUR-NEXT: retq +; +; CHECK-O3-CUR-LABEL: load_i16_anyext_i32: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movzwl (%rdi), %eax +; CHECK-O3-CUR-NEXT: retq +; +; CHECK-O0-EX-LABEL: load_i16_anyext_i32: +; CHECK-O0-EX: # %bb.0: +; CHECK-O0-EX-NEXT: vpbroadcastw (%rdi), %xmm0 +; CHECK-O0-EX-NEXT: vmovd %xmm0, %eax +; CHECK-O0-EX-NEXT: retq +; +; CHECK-O3-EX-LABEL: load_i16_anyext_i32: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: vpbroadcastw (%rdi), %xmm0 +; CHECK-O3-EX-NEXT: vmovd %xmm0, %eax +; CHECK-O3-EX-NEXT: retq %v = load atomic i16, i16* %ptr unordered, align 4 %vec = insertelement <2 x i16> undef, i16 %v, i64 0 %res = bitcast <2 x i16> %vec to i32 @@ -3120,27 +2648,33 @@ define i32 @load_i16_anyext_i32(i16* %ptr) { } define i64 @load_i16_anyext_i64(i16* %ptr) { -; CHECK-O0-LABEL: load_i16_anyext_i64: -; CHECK-O0: # %bb.0: -; CHECK-O0-NEXT: movw (%rdi), %ax -; CHECK-O0-NEXT: # implicit-def: $ecx -; CHECK-O0-NEXT: movw %ax, %cx -; CHECK-O0-NEXT: vmovd %ecx, %xmm0 -; CHECK-O0-NEXT: vmovq %xmm0, %rax -; CHECK-O0-NEXT: retq -; -; CHECK-O3-LABEL: load_i16_anyext_i64: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: movzwl (%rdi), %eax -; CHECK-O3-NEXT: vmovd %eax, %xmm0 -; CHECK-O3-NEXT: vmovq %xmm0, %rax -; CHECK-O3-NEXT: retq -; -; CHECK-EX-LABEL: load_i16_anyext_i64: -; CHECK-EX: # %bb.0: -; CHECK-EX-NEXT: vpbroadcastw (%rdi), %xmm0 -; CHECK-EX-NEXT: vmovq %xmm0, %rax -; CHECK-EX-NEXT: retq +; CHECK-O0-CUR-LABEL: load_i16_anyext_i64: +; CHECK-O0-CUR: # %bb.0: +; CHECK-O0-CUR-NEXT: movw (%rdi), %ax +; CHECK-O0-CUR-NEXT: # implicit-def: $ecx +; CHECK-O0-CUR-NEXT: movw %ax, %cx +; CHECK-O0-CUR-NEXT: vmovd %ecx, %xmm0 +; CHECK-O0-CUR-NEXT: vmovq %xmm0, %rax +; CHECK-O0-CUR-NEXT: retq +; +; CHECK-O3-CUR-LABEL: load_i16_anyext_i64: +; CHECK-O3-CUR: # %bb.0: +; CHECK-O3-CUR-NEXT: movzwl (%rdi), %eax +; CHECK-O3-CUR-NEXT: vmovd %eax, %xmm0 +; CHECK-O3-CUR-NEXT: vmovq %xmm0, %rax +; CHECK-O3-CUR-NEXT: retq +; +; CHECK-O0-EX-LABEL: load_i16_anyext_i64: +; CHECK-O0-EX: # %bb.0: +; CHECK-O0-EX-NEXT: vpbroadcastw (%rdi), %xmm0 +; CHECK-O0-EX-NEXT: vmovq %xmm0, %rax +; CHECK-O0-EX-NEXT: retq +; +; CHECK-O3-EX-LABEL: load_i16_anyext_i64: +; CHECK-O3-EX: # %bb.0: +; CHECK-O3-EX-NEXT: vpbroadcastw (%rdi), %xmm0 +; CHECK-O3-EX-NEXT: vmovq %xmm0, %rax +; CHECK-O3-EX-NEXT: retq %v = load atomic i16, i16* %ptr unordered, align 8 %vec = insertelement <4 x i16> undef, i16 %v, i64 0 %res = bitcast <4 x i16> %vec to i64