//===----------------------------------------------------------------------===//
// Prefetch intrinsic.
-def PREFETCHT0 : VoPSI<0x18, MRM1m, (outs), (ins i8mem:$src),
- "prefetcht0\t$src", [(prefetch addr:$src, imm, (i32 3), (i32 1))]>;
-def PREFETCHT1 : VoPSI<0x18, MRM2m, (outs), (ins i8mem:$src),
- "prefetcht1\t$src", [(prefetch addr:$src, imm, (i32 2), (i32 1))]>;
-def PREFETCHT2 : VoPSI<0x18, MRM3m, (outs), (ins i8mem:$src),
- "prefetcht2\t$src", [(prefetch addr:$src, imm, (i32 1), (i32 1))]>;
-def PREFETCHNTA : VoPSI<0x18, MRM0m, (outs), (ins i8mem:$src),
- "prefetchnta\t$src", [(prefetch addr:$src, imm, (i32 0), (i32 1))]>;
+let Predicates = [HasXMM] in {
+def PREFETCHT0 : I<0x18, MRM1m, (outs), (ins i8mem:$src),
+ "prefetcht0\t$src", [(prefetch addr:$src, imm, (i32 3), (i32 1))]>, TB;
+def PREFETCHT1 : I<0x18, MRM2m, (outs), (ins i8mem:$src),
+ "prefetcht1\t$src", [(prefetch addr:$src, imm, (i32 2), (i32 1))]>, TB;
+def PREFETCHT2 : I<0x18, MRM3m, (outs), (ins i8mem:$src),
+ "prefetcht2\t$src", [(prefetch addr:$src, imm, (i32 1), (i32 1))]>, TB;
+def PREFETCHNTA : I<0x18, MRM0m, (outs), (ins i8mem:$src),
+ "prefetchnta\t$src", [(prefetch addr:$src, imm, (i32 0), (i32 1))]>, TB;
+}
// Flush cache
def CLFLUSH : I<0xAE, MRM7m, (outs), (ins i8mem:$src),
"clflush\t$src", [(int_x86_sse2_clflush addr:$src)]>,
- TB, Requires<[HasSSE2]>;
+ TB, Requires<[HasXMMInt]>;
// Pause. This "instruction" is encoded as "rep; nop", so even though it
// was introduced with SSE2, it's backward compatible.
// Load, store, and memory fence
def SFENCE : I<0xAE, MRM_F8, (outs), (ins),
- "sfence", [(int_x86_sse_sfence)]>, TB, Requires<[HasSSE1]>;
+ "sfence", [(int_x86_sse_sfence)]>, TB, Requires<[HasXMM]>;
def LFENCE : I<0xAE, MRM_E8, (outs), (ins),
- "lfence", [(int_x86_sse2_lfence)]>, TB, Requires<[HasSSE2]>;
+ "lfence", [(int_x86_sse2_lfence)]>, TB, Requires<[HasXMMInt]>;
def MFENCE : I<0xAE, MRM_F0, (outs), (ins),
- "mfence", [(int_x86_sse2_mfence)]>, TB, Requires<[HasSSE2]>;
+ "mfence", [(int_x86_sse2_mfence)]>, TB, Requires<[HasXMMInt]>;
def : Pat<(X86SFence), (SFENCE)>;
def : Pat<(X86LFence), (LFENCE)>;
let usesCustomInserter = 1 in {
def MONITOR : PseudoI<(outs), (ins i32mem:$src1, GR32:$src2, GR32:$src3),
- [(int_x86_sse3_monitor addr:$src1, GR32:$src2, GR32:$src3)]>;
+ [(int_x86_sse3_monitor addr:$src1, GR32:$src2, GR32:$src3)]>,
+ Requires<[HasSSE3orAVX]>;
def MWAIT : PseudoI<(outs), (ins GR32:$src1, GR32:$src2),
- [(int_x86_sse3_mwait GR32:$src1, GR32:$src2)]>;
+ [(int_x86_sse3_mwait GR32:$src1, GR32:$src2)]>,
+ Requires<[HasSSE3orAVX]>;
}
let Uses = [EAX, ECX, EDX] in
def MONITORrrr : I<0x01, MRM_C8, (outs), (ins), "monitor", []>, TB,
- Requires<[HasSSE3]>;
+ Requires<[HasSSE3orAVX]>;
let Uses = [ECX, EAX] in
def MWAITrr : I<0x01, MRM_C9, (outs), (ins), "mwait", []>, TB,
- Requires<[HasSSE3]>;
+ Requires<[HasSSE3orAVX]>;
def : InstAlias<"mwait %eax, %ecx", (MWAITrr)>, Requires<[In32BitMode]>;
def : InstAlias<"mwait %rax, %rcx", (MWAITrr)>, Requires<[In64BitMode]>;
}
declare void @llvm.x86.avx.vzeroupper() nounwind
+; Make sure instructions with no AVX equivalents, but are associated with SSEX feature flags still work
+; CHECK: monitor
+define void @monitor(i8* %P, i32 %E, i32 %H) nounwind {
+entry:
+ tail call void @llvm.x86.sse3.monitor(i8* %P, i32 %E, i32 %H)
+ ret void
+}
+declare void @llvm.x86.sse3.monitor(i8*, i32, i32) nounwind
+
+; CHECK: mwait
+define void @mwait(i32 %E, i32 %H) nounwind {
+entry:
+ tail call void @llvm.x86.sse3.mwait(i32 %E, i32 %H)
+ ret void
+}
+declare void @llvm.x86.sse3.mwait(i32, i32) nounwind
+
+; CHECK: sfence
+define void @sfence() nounwind {
+entry:
+ tail call void @llvm.x86.sse.sfence()
+ ret void
+}
+declare void @llvm.x86.sse.sfence() nounwind
+
+; CHECK: lfence
+define void @lfence() nounwind {
+entry:
+ tail call void @llvm.x86.sse2.lfence()
+ ret void
+}
+declare void @llvm.x86.sse2.lfence() nounwind
+
+; CHECK: mfence
+define void @mfence() nounwind {
+entry:
+ tail call void @llvm.x86.sse2.mfence()
+ ret void
+}
+declare void @llvm.x86.sse2.mfence() nounwind
+
+; CHECK: clflush
+define void @clflush(i8* %p) nounwind {
+entry:
+ tail call void @llvm.x86.sse2.clflush(i8* %p)
+ ret void
+}
+declare void @llvm.x86.sse2.clflush(i8*) nounwind