[X86] WaitPKG instructions

Three new instructions:

umonitor - Sets up a linear address range to be
monitored by hardware and activates the monitor.
The address range should be a writeback memory
caching type.

umwait - A hint that allows the processor to
stop instruction execution and enter an
implementation-dependent optimized state
until occurrence of a class of events.

tpause - Directs the processor to enter an
implementation-dependent optimized state
until the TSC reaches the value in EDX:EAX.

Also modifying the description of the mfence
instruction, as the rep prefix (0xF3) was allowed
before, which would conflict with umonitor during
disassembly.

Before:
$ echo 0xf3,0x0f,0xae,0xf0 | llvm-mc -disassemble
.text
mfence

After:
$ echo 0xf3,0x0f,0xae,0xf0 | llvm-mc -disassemble
.text
umonitor        %rax

Reviewers: craig.topper, zvi

Reviewed By: craig.topper

Differential Revision: https://reviews.llvm.org/D45253

llvm-svn: 330462

7 files changed, 152 insertions, 0 deletions

diff --git a/llvm/test/CodeGen/X86/waitpkg-intrinsics.ll b/llvm/test/CodeGen/X86/waitpkg-intrinsics.ll
new file mode 100644
index 00000000000..c0be395aa66
--- /dev/null
+++ b/llvm/test/CodeGen/X86/waitpkg-intrinsics.ll
@@ -0,0 +1,67 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-linux -mattr=+waitpkg | FileCheck %s --check-prefix=X64
+; RUN: llc < %s -mtriple=i386-pc-linux -mattr=+waitpkg | FileCheck %s --check-prefix=X32
+
+define void @test_umonitor(i8* %address) {
+; X64-LABEL: test_umonitor:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    umonitor %rdi
+; X64-NEXT:    retq
+;
+; X32-LABEL: test_umonitor:
+; X32:       # %bb.0: # %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    umonitor %eax
+; X32-NEXT:    retl
+entry:
+  call void @llvm.x86.umonitor(i8* %address)
+  ret void
+}
+
+define i8 @test_umwait(i32 %control, i32 %counter_high, i32 %counter_low) {
+; X64-LABEL: test_umwait:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movl %edx, %eax
+; X64-NEXT:    movl %esi, %edx
+; X64-NEXT:    umwait %edi
+; X64-NEXT:    setb %al
+; X64-NEXT:    retq
+;
+; X32-LABEL: test_umwait:
+; X32:       # %bb.0: # %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT:    umwait %ecx
+; X32-NEXT:    setb %al
+; X32-NEXT:    retl
+entry:
+  call i8 @llvm.x86.umwait(i32 %control, i32 %counter_high, i32 %counter_low)
+  ret i8 %0
+}
+
+define i8 @test_tpause(i32 %control, i32 %counter_high, i32 %counter_low) {
+; X64-LABEL: test_tpause:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movl %edx, %eax
+; X64-NEXT:    movl %esi, %edx
+; X64-NEXT:    tpause %edi
+; X64-NEXT:    setb %al
+; X64-NEXT:    retq
+;
+; X32-LABEL: test_tpause:
+; X32:       # %bb.0: # %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT:    tpause %ecx
+; X32-NEXT:    setb %al
+; X32-NEXT:    retl
+entry:
+  call i8 @llvm.x86.tpause(i32 %control, i32 %counter_high, i32 %counter_low)
+  ret i8 %0
+}
+
+declare void @llvm.x86.umonitor(i8*)
+declare i8 @llvm.x86.umwait(i32, i32, i32)
+declare i8 @llvm.x86.tpause(i32, i32, i32)
diff --git a/llvm/test/MC/Disassembler/X86/x86-16.txt b/llvm/test/MC/Disassembler/X86/x86-16.txt
index 60341ad61f0..cf534351157 100644
--- a/llvm/test/MC/Disassembler/X86/x86-16.txt
+++ b/llvm/test/MC/Disassembler/X86/x86-16.txt
@@ -794,3 +794,9 @@
 
 # CHECK: wbnoinvd
 0xf3 0x0f 0x09
+
+# CHECK: umonitor %ax
+0xf3 0x0f 0xae 0xf0
+
+# CHECK: umonitor %eax
+0x67 0xf3 0x0f 0xae 0xf0
diff --git a/llvm/test/MC/Disassembler/X86/x86-32.txt b/llvm/test/MC/Disassembler/X86/x86-32.txt
index f4adf9f0d5e..20a82b5a582 100644
--- a/llvm/test/MC/Disassembler/X86/x86-32.txt
+++ b/llvm/test/MC/Disassembler/X86/x86-32.txt
@@ -847,3 +847,15 @@
 0x0f 0xb7 0x00
 # CHECK: movzww (%eax), %ax
 0x66 0x0f 0xb7 0x00
+
+# CHECK: umonitor %eax
+0xf3 0x0f 0xae 0xf0
+
+# CHECK: umonitor %ax
+0x67 0xf3 0x0f 0xae 0xf0
+
+# CHECK: umwait %eax
+0xf2 0x0f 0xae 0xf0
+
+# CHECK: tpause %eax
+0x66 0x0f 0xae 0xf0
diff --git a/llvm/test/MC/Disassembler/X86/x86-64.txt b/llvm/test/MC/Disassembler/X86/x86-64.txt
index f066feead4d..eeba5e3f9de 100644
--- a/llvm/test/MC/Disassembler/X86/x86-64.txt
+++ b/llvm/test/MC/Disassembler/X86/x86-64.txt
@@ -525,3 +525,18 @@
 
 # CHECK: cldemote -559038737(%rbx,%rcx,8)
 0x0f,0x1c,0x84,0xcb,0xef,0xbe,0xad,0xde
+
+# CHECK: umonitor %rax
+0xf3 0x0f 0xae 0xf0
+
+# CHECK: umonitor %eax
+0x67 0xf3 0x0f 0xae 0xf0
+
+# CHECK: umonitor %r13
+0xf3 0x41 0x0f 0xae 0xf5
+
+# CHECK: umwait %r15
+0xf2 0x41 0x0f 0xae 0xf7
+
+# CHECK: tpause %r15
+0x66 0x41 0x0f 0xae 0xf7
diff --git a/llvm/test/MC/X86/x86-16.s b/llvm/test/MC/X86/x86-16.s
index b890be30500..7326c2a9bfa 100644
--- a/llvm/test/MC/X86/x86-16.s
+++ b/llvm/test/MC/X86/x86-16.s
@@ -973,3 +973,11 @@ data32 lgdt 4(%eax)
 // CHECK: wbnoinvd
 // CHECK:  encoding: [0xf3,0x0f,0x09]
 wbnoinvd
+
+// CHECK: umonitor %ax
+// CHECK:  encoding: [0xf3,0x0f,0xae,0xf0]
+umonitor %ax
+
+// CHECK: umonitor %eax
+// CHECK:  encoding: [0x67,0xf3,0x0f,0xae,0xf0]
+umonitor %eax
diff --git a/llvm/test/MC/X86/x86-32-coverage.s b/llvm/test/MC/X86/x86-32-coverage.s
index 65ea392fa72..79dc2e506ef 100644
--- a/llvm/test/MC/X86/x86-32-coverage.s
+++ b/llvm/test/MC/X86/x86-32-coverage.s
@@ -10752,3 +10752,19 @@ btcl $4, (%eax)
 // CHECK: cldemote 3735928559(%ebx,%ecx,8)
 // CHECK:  encoding: [0x0f,0x1c,0x84,0xcb,0xef,0xbe,0xad,0xde]
         	cldemote 0xdeadbeef(%ebx,%ecx,8)
+
+// CHECK: umonitor %eax
+// CHECK:  encoding: [0xf3,0x0f,0xae,0xf0]
+	umonitor %eax
+
+// CHECK: umonitor %ax
+// CHECK:  encoding: [0x67,0xf3,0x0f,0xae,0xf0]
+	umonitor %ax
+
+// CHECK: umwait %eax
+// CHECK:  encoding: [0xf2,0x0f,0xae,0xf0]
+	umwait %eax
+
+// CHECK: tpause %eax
+// CHECK:  encoding: [0x66,0x0f,0xae,0xf0]
+	tpause %eax
diff --git a/llvm/test/MC/X86/x86-64.s b/llvm/test/MC/X86/x86-64.s
index ea5be29cbfc..31e8888f398 100644
--- a/llvm/test/MC/X86/x86-64.s
+++ b/llvm/test/MC/X86/x86-64.s
@@ -1571,6 +1571,34 @@ cldemote 4(%rax)
 // CHECK:  encoding: [0x0f,0x1c,0x84,0xcb,0xef,0xbe,0xad,0xde]
 cldemote 0xdeadbeef(%rbx,%rcx,8)
 
+// CHECK: umonitor %r13
+// CHECK:  encoding: [0xf3,0x41,0x0f,0xae,0xf5]
+umonitor %r13
+
+// CHECK: umonitor %rax
+// CHECK:  encoding: [0xf3,0x0f,0xae,0xf0]
+umonitor %rax
+
+// CHECK: umonitor %eax
+// CHECK:  encoding: [0x67,0xf3,0x0f,0xae,0xf0]
+umonitor %eax
+
+// CHECK: umwait %r15
+// CHECK:  encoding: [0xf2,0x41,0x0f,0xae,0xf7]
+umwait %r15
+
+// CHECK: umwait %ebx
+// CHECK:  encoding: [0xf2,0x0f,0xae,0xf3]
+umwait %ebx
+
+// CHECK: tpause %r15
+// CHECK:  encoding: [0x66,0x41,0x0f,0xae,0xf7]
+tpause %r15
+
+// CHECK: tpause %ebx
+// CHECK:  encoding: [0x66,0x0f,0xae,0xf3]
+tpause %ebx
+
 //  __asm __volatile(
 //    "pushf        \n\t"
 //    "popf       \n\t"