[X86] Model MXCSR for all SSE instructions

[craig.topper]

Summary

This patch adds MXCSR as a reserved physical register and models its use
by X86 SSE instructions. It also adds flag "mayRaiseFPException" for the
instructions that possibly can raise FP exception according to the
architecture definition.

Following what SystemZ and other targets does, only the current rounding
modes and the IEEE exception masks are modeled. *Changes* of the MXCSR
due to exceptions are not modeled.

[craig.topper]

Does this really not affect any tests? I would have thought some test would print MIR output that would show the implicit use that wasn't there before.

llvm/lib/Target/X86/X86InstrFormats.td

230

Is EXP an abbreviation for Exception? If so, I think EXC is probably a better abbreviation

[RKSimon]

Instead of attributing all these instructions with MXCSR first wouldn't we be better off starting just attributing (V)LDMXCSR/(V)STMXCSR in this patch and then future patches sets up groups of instructions with suitable tests?

llvm/lib/Target/X86/X86InstrFormats.td

230

mayRaiseMXCSRException might be better as well - to avoid x87/fpu ambiguity

[pengfei]

In D68121#1685679, @RKSimon wrote:

Instead of attributing all these instructions with MXCSR first wouldn't we be better off starting just attributing (V)LDMXCSR/(V)STMXCSR in this patch and then future patches sets up groups of instructions with suitable tests?

The difficult is we may have X87/SSE/AVX/AVX512 instructions under the same node, so I'm planning to mode the attribute of instructions by X87/SSE/AVX/AVX512 first, then set up instructions by node.
And we don't need to model MXCSR for (V)LDMXCSR/(V)STMXCSR, because they are using hasSideEffects and cannot be replaced due to we only modeling rounding and exception bits in MXCSR.

llvm/lib/Target/X86/X86InstrFormats.td

230

mayRaiseFPException is a target independent flag and is widely used to check if an instruction may raise exception during optimization. I think there's no need to distinguish it from x87 or SIMD during the optimization, and we can distinguish them from the cpcode of MI if necessary.

[pengfei]

In D68121#1685355, @craig.topper wrote:

Does this really not affect any tests? I would have thought some test would print MIR output that would show the implicit use that wasn't there before.

You are correct. Besides some MIR tests, there're also some machine combiner failure. I'm trying to fix it. Thanks for reminding.

llvm/lib/Target/X86/X86InstrFormats.td

230

Yes, I'll change the name.

[pengfei]

Address review comments, add affected tests.

[craig.topper]

llvm/test/CodeGen/X86/evex-to-vex-compress.mir

2317 ↗

(On Diff #222314)

Why are only COMI instructions updated?

[pengfei]

llvm/test/CodeGen/X86/evex-to-vex-compress.mir

2317 ↗

(On Diff #222314)

We don't have any special test case for checking the operands of MIs, and I checked other target don't either. This file may be the one that checks most AVX/AVX512 MIs, but we only modeled SSE instructions in this patch.
VCOMI shouldn't be updated either. The reason is VCOMISS/VCOMISD are defined is X86InstrSSE.td and using sse12_ord_cmp which I modeled MXCSR in.
In fact, I just made sure all SSE instructions are correctly modeled, not all modeled instructions are SSE. I think it doesn't matter, because it is NFC and we will model AVX instructions next.

[RKSimon]

In D68121#1685783, @pengfei wrote:

And we don't need to model MXCSR for (V)LDMXCSR/(V)STMXCSR, because they are using hasSideEffects and cannot be replaced due to we only modeling rounding and exception bits in MXCSR.

I'm not talking about removing the hasSideEffects tag (at least not yet), I'm just suggesting that it'd make sense to indicate that MXCSR is written/read by (V)LDMXCSR/(V)STMXCSR

[craig.topper]

Commandeering so I can rebase part of this patch to remove the X86InstrInfo.cpp changes.

[craig.topper]

Remove X86InstrInfo.cpp changes

[craig.topper]

LGTM

llvm/lib/Target/X86/X86InstrFormats.td

 class EVEX_K { bit hasEVEX_K = 1; }
 class EVEX_KZ : EVEX_K { bit hasEVEX_Z = 1; }
 class EVEX_B { bit hasEVEX_B = 1; }
 class EVEX_RC { bit hasEVEX_RC = 1; }
 class EVEX_V512 { bit hasEVEX_L2 = 1; bit hasVEX_L = 0; }
 class EVEX_V256 { bit hasEVEX_L2 = 0; bit hasVEX_L = 1; }
 class EVEX_V128 { bit hasEVEX_L2 = 0; bit hasVEX_L = 0; }
 class NOTRACK { bit hasNoTrackPrefix = 1; }
+class SIMD_EXP { list<Register> Uses = [MXCSR]; bit mayRaiseFPException = 1; }
 
 // Specify AVX512 8-bit compressed displacement encoding based on the vector
 // element size in bits (8, 16, 32, 64) and the CDisp8 form.
 class EVEX_CD8<int esize, CD8VForm form> {
   int CD8_EltSize = !srl(esize, 3);
   bits<3> CD8_Form = form.Value;
 }
 

llvm/lib/Target/X86/X86InstrSSE.td

def rm : SI<opc, MRMSrcMem, (outs DstRC:$dst), (ins x86memop:$src),
               mem#"\t{$src, $dst|$dst, $src}",
               [(set DstRC:$dst, (OpNode (ld_frag addr:$src)))]>,
               Sched<[sched.Folded]>;
 }
 
 multiclass sse12_cvt_p<bits<8> opc, RegisterClass RC, X86MemOperand x86memop,
                        ValueType DstTy, ValueType SrcTy, PatFrag ld_frag,
                        string asm, Domain d, X86FoldableSchedWrite sched> {
-let hasSideEffects = 0 in {
+let hasSideEffects = 0, Uses = [MXCSR], mayRaiseFPException = 1 in {
   def rr : I<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src), asm,
              [(set RC:$dst, (DstTy (sint_to_fp (SrcTy RC:$src))))], d>,
              Sched<[sched]>;
   let mayLoad = 1 in
   def rm : I<opc, MRMSrcMem, (outs RC:$dst), (ins x86memop:$src), asm,
              [(set RC:$dst, (DstTy (sint_to_fp
                                     (SrcTy (ld_frag addr:$src)))))], d>,
              Sched<[sched.Folded]>;
let hasSideEffects = 0, Predicates = [UseAVX] in {
   let mayLoad = 1 in
   def rm : SI<opc, MRMSrcMem, (outs DstRC:$dst),
               (ins DstRC:$src1, x86memop:$src),
               asm#"{"#mem#"}\t{$src, $src1, $dst|$dst, $src1, $src}", []>,
            Sched<[sched.Folded, sched.ReadAfterFold]>;
 } // hasSideEffects = 0
 }
 
-let isCodeGenOnly = 1, Predicates = [UseAVX] in {
+let isCodeGenOnly = 1, Predicates = [UseAVX], Uses = [MXCSR], mayRaiseFPException = 1 in {
 defm VCVTTSS2SI   : sse12_cvt_s<0x2C, FR32, GR32, fp_to_sint, f32mem, loadf32,
                                 "cvttss2si", "cvttss2si",
                                 WriteCvtSS2I>,
                                 XS, VEX, VEX_LIG;
 defm VCVTTSS2SI64 : sse12_cvt_s<0x2C, FR32, GR64, fp_to_sint, f32mem, loadf32,
                                 "cvttss2si", "cvttss2si",
                                 WriteCvtSS2I>,
                                 XS, VEX, VEX_W, VEX_LIG;
 defm VCVTTSD2SI   : sse12_cvt_s<0x2C, FR64, GR32, fp_to_sint, f64mem, loadf64,
                                 "cvttsd2si", "cvttsd2si",
                                 WriteCvtSD2I>,
                                 XD, VEX, VEX_LIG;
 defm VCVTTSD2SI64 : sse12_cvt_s<0x2C, FR64, GR64, fp_to_sint, f64mem, loadf64,
                                 "cvttsd2si", "cvttsd2si",
                                 WriteCvtSD2I>,
                                 XD, VEX, VEX_W, VEX_LIG;
 }
 
 // The assembler can recognize rr 64-bit instructions by seeing a rxx
 // register, but the same isn't true when only using memory operands,
 // provide other assembly "l" and "q" forms to address this explicitly
 // where appropriate to do so.
 let isCodeGenOnly = 1 in {
 defm VCVTSI2SS   : sse12_vcvt_avx<0x2A, GR32, FR32, i32mem, "cvtsi2ss", "l",
-                                  WriteCvtI2SS>, XS, VEX_4V, VEX_LIG;
+                                  WriteCvtI2SS>, XS, VEX_4V, VEX_LIG, SIMD_EXP;
 defm VCVTSI642SS : sse12_vcvt_avx<0x2A, GR64, FR32, i64mem, "cvtsi2ss", "q",
-                                  WriteCvtI2SS>, XS, VEX_4V, VEX_W, VEX_LIG;
+                                  WriteCvtI2SS>, XS, VEX_4V, VEX_W, VEX_LIG, SIMD_EXP;
 defm VCVTSI2SD   : sse12_vcvt_avx<0x2A, GR32, FR64, i32mem, "cvtsi2sd", "l",
                                   WriteCvtI2SD>, XD, VEX_4V, VEX_LIG;
 defm VCVTSI642SD : sse12_vcvt_avx<0x2A, GR64, FR64, i64mem, "cvtsi2sd", "q",
-                                  WriteCvtI2SD>, XD, VEX_4V, VEX_W, VEX_LIG;
+                                  WriteCvtI2SD>, XD, VEX_4V, VEX_W, VEX_LIG, SIMD_EXP;
 } // isCodeGenOnly = 1
 
 let Predicates = [UseAVX] in {
   def : Pat<(f32 (sint_to_fp (loadi32 addr:$src))),
             (VCVTSI2SSrm (f32 (IMPLICIT_DEF)), addr:$src)>;
   def : Pat<(f32 (sint_to_fp (loadi64 addr:$src))),
             (VCVTSI642SSrm (f32 (IMPLICIT_DEF)), addr:$src)>;
   def : Pat<(f64 (sint_to_fp (loadi32 addr:$src))),
def : Pat<(f64 (sint_to_fp GR32:$src)),
             (VCVTSI2SDrr (f64 (IMPLICIT_DEF)), GR32:$src)>;
   def : Pat<(f64 (sint_to_fp GR64:$src)),
             (VCVTSI642SDrr (f64 (IMPLICIT_DEF)), GR64:$src)>;
 }
 
 let isCodeGenOnly = 1 in {
 defm CVTTSS2SI : sse12_cvt_s<0x2C, FR32, GR32, fp_to_sint, f32mem, loadf32,
                       "cvttss2si", "cvttss2si",
-                      WriteCvtSS2I>, XS;
+                      WriteCvtSS2I>, XS, SIMD_EXP;
 defm CVTTSS2SI64 : sse12_cvt_s<0x2C, FR32, GR64, fp_to_sint, f32mem, loadf32,
                       "cvttss2si", "cvttss2si",
-                      WriteCvtSS2I>, XS, REX_W;
+                      WriteCvtSS2I>, XS, REX_W, SIMD_EXP;
 defm CVTTSD2SI : sse12_cvt_s<0x2C, FR64, GR32, fp_to_sint, f64mem, loadf64,
                       "cvttsd2si", "cvttsd2si",
-                      WriteCvtSD2I>, XD;
+                      WriteCvtSD2I>, XD, SIMD_EXP;
 defm CVTTSD2SI64 : sse12_cvt_s<0x2C, FR64, GR64, fp_to_sint, f64mem, loadf64,
                       "cvttsd2si", "cvttsd2si",
-                      WriteCvtSD2I>, XD, REX_W;
+                      WriteCvtSD2I>, XD, REX_W, SIMD_EXP;
 defm CVTSI2SS  : sse12_cvt_s<0x2A, GR32, FR32, sint_to_fp, i32mem, loadi32,
                       "cvtsi2ss", "cvtsi2ss{l}",
-                      WriteCvtI2SS, ReadInt2Fpu>, XS;
+                      WriteCvtI2SS, ReadInt2Fpu>, XS, SIMD_EXP;
 defm CVTSI642SS : sse12_cvt_s<0x2A, GR64, FR32, sint_to_fp, i64mem, loadi64,
                       "cvtsi2ss", "cvtsi2ss{q}",
-                      WriteCvtI2SS, ReadInt2Fpu>, XS, REX_W;
+                      WriteCvtI2SS, ReadInt2Fpu>, XS, REX_W, SIMD_EXP;
 defm CVTSI2SD  : sse12_cvt_s<0x2A, GR32, FR64, sint_to_fp, i32mem, loadi32,
                       "cvtsi2sd", "cvtsi2sd{l}",
                       WriteCvtI2SD, ReadInt2Fpu>, XD;
 defm CVTSI642SD : sse12_cvt_s<0x2A, GR64, FR64, sint_to_fp, i64mem, loadi64,
                       "cvtsi2sd", "cvtsi2sd{q}",
-                      WriteCvtI2SD, ReadInt2Fpu>, XD, REX_W;
+                      WriteCvtI2SD, ReadInt2Fpu>, XD, REX_W, SIMD_EXP;
 } // isCodeGenOnly = 1
 
 // Conversion Instructions Intrinsics - Match intrinsics which expect MM
 // and/or XMM operand(s).
 
 multiclass sse12_cvt_sint<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC,
                           ValueType DstVT, ValueType SrcVT, SDNode OpNode,
                           Operand memop, ComplexPattern mem_cpat, string asm,
def rm_Int : SI<opc, MRMSrcMem, (outs DstRC:$dst),
                   (ins DstRC:$src1, x86memop:$src2),
                   !if(Is2Addr,
                       asm#"{"#mem#"}\t{$src2, $dst|$dst, $src2}",
                       asm#"{"#mem#"}\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
                   []>, Sched<[sched.Folded, sched.ReadAfterFold]>;
 }
 }
 
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
 let Predicates = [UseAVX] in {
 defm VCVTSD2SI : sse12_cvt_sint<0x2D, VR128, GR32, i32, v2f64,
                   X86cvts2si, sdmem, sse_load_f64, "cvtsd2si",
                   WriteCvtSD2I>, XD, VEX, VEX_LIG;
 defm VCVTSD2SI64 : sse12_cvt_sint<0x2D, VR128, GR64, i64, v2f64,
                     X86cvts2si, sdmem, sse_load_f64, "cvtsd2si",
                     WriteCvtSD2I>, XD, VEX, VEX_W, VEX_LIG;
 }
 defm CVTSD2SI : sse12_cvt_sint<0x2D, VR128, GR32, i32, v2f64, X86cvts2si,
                  sdmem, sse_load_f64, "cvtsd2si", WriteCvtSD2I>, XD;
 defm CVTSD2SI64 : sse12_cvt_sint<0x2D, VR128, GR64, i64, v2f64, X86cvts2si,
                    sdmem, sse_load_f64, "cvtsd2si", WriteCvtSD2I>, XD, REX_W;
-
+}
 
 let Predicates = [UseAVX] in {
 defm VCVTSI2SS : sse12_cvt_sint_3addr<0x2A, GR32, VR128,
-          i32mem, "cvtsi2ss", "l", WriteCvtI2SS, 0>, XS, VEX_4V, VEX_LIG;
+          i32mem, "cvtsi2ss", "l", WriteCvtI2SS, 0>, XS, VEX_4V, VEX_LIG, SIMD_EXP;
 defm VCVTSI642SS : sse12_cvt_sint_3addr<0x2A, GR64, VR128,
-          i64mem, "cvtsi2ss", "q", WriteCvtI2SS, 0>, XS, VEX_4V, VEX_LIG, VEX_W;
+          i64mem, "cvtsi2ss", "q", WriteCvtI2SS, 0>, XS, VEX_4V, VEX_LIG, VEX_W, SIMD_EXP;
 defm VCVTSI2SD : sse12_cvt_sint_3addr<0x2A, GR32, VR128,
           i32mem, "cvtsi2sd", "l", WriteCvtI2SD, 0>, XD, VEX_4V, VEX_LIG;
 defm VCVTSI642SD : sse12_cvt_sint_3addr<0x2A, GR64, VR128,
-          i64mem, "cvtsi2sd", "q", WriteCvtI2SD, 0>, XD, VEX_4V, VEX_LIG, VEX_W;
+          i64mem, "cvtsi2sd", "q", WriteCvtI2SD, 0>, XD, VEX_4V, VEX_LIG, VEX_W, SIMD_EXP;
 }
 let Constraints = "$src1 = $dst" in {
   defm CVTSI2SS : sse12_cvt_sint_3addr<0x2A, GR32, VR128,
-                        i32mem, "cvtsi2ss", "l", WriteCvtI2SS>, XS;
+                        i32mem, "cvtsi2ss", "l", WriteCvtI2SS>, XS, SIMD_EXP;
   defm CVTSI642SS : sse12_cvt_sint_3addr<0x2A, GR64, VR128,
-                        i64mem, "cvtsi2ss", "q", WriteCvtI2SS>, XS, REX_W;
+                        i64mem, "cvtsi2ss", "q", WriteCvtI2SS>, XS, REX_W, SIMD_EXP;
   defm CVTSI2SD : sse12_cvt_sint_3addr<0x2A, GR32, VR128,
                         i32mem, "cvtsi2sd", "l", WriteCvtI2SD>, XD;
   defm CVTSI642SD : sse12_cvt_sint_3addr<0x2A, GR64, VR128,
-                        i64mem, "cvtsi2sd", "q", WriteCvtI2SD>, XD, REX_W;
+                        i64mem, "cvtsi2sd", "q", WriteCvtI2SD>, XD, REX_W, SIMD_EXP;
 }
 
 def : InstAlias<"vcvtsi2ss{l}\t{$src2, $src1, $dst|$dst, $src1, $src2}",
                (VCVTSI2SSrr_Int VR128:$dst, VR128:$src1, GR32:$src2), 0, "att">;
 def : InstAlias<"vcvtsi2ss{q}\t{$src2, $src1, $dst|$dst, $src1, $src2}",
                (VCVTSI642SSrr_Int VR128:$dst, VR128:$src1, GR64:$src2), 0, "att">;
 def : InstAlias<"vcvtsi2sd{l}\t{$src2, $src1, $dst|$dst, $src1, $src2}",
                (VCVTSI2SDrr_Int VR128:$dst, VR128:$src1, GR32:$src2), 0, "att">;

 def : InstAlias<"cvtsi2ss\t{$src, $dst|$dst, $src}",
                 (CVTSI2SSrm_Int VR128:$dst, i32mem:$src), 0, "att">;
 def : InstAlias<"cvtsi2sd\t{$src, $dst|$dst, $src}",
                 (CVTSI2SDrm_Int VR128:$dst, i32mem:$src), 0, "att">;
 
 /// SSE 1 Only
 
 // Aliases for intrinsics
-let Predicates = [UseAVX] in {
+let Predicates = [UseAVX], Uses = [MXCSR], mayRaiseFPException = 1 in {
 defm VCVTTSS2SI : sse12_cvt_sint<0x2C, VR128, GR32, i32, v4f32, X86cvtts2Int,
                                 ssmem, sse_load_f32, "cvttss2si",
                                 WriteCvtSS2I>, XS, VEX, VEX_LIG;
 defm VCVTTSS2SI64 : sse12_cvt_sint<0x2C, VR128, GR64, i64, v4f32,
                                X86cvtts2Int, ssmem, sse_load_f32,
                                "cvttss2si", WriteCvtSS2I>,
                                XS, VEX, VEX_LIG, VEX_W;
 defm VCVTTSD2SI : sse12_cvt_sint<0x2C, VR128, GR32, i32, v2f64, X86cvtts2Int,
                                 sdmem, sse_load_f64, "cvttsd2si",
                                 WriteCvtSS2I>, XD, VEX, VEX_LIG;
 defm VCVTTSD2SI64 : sse12_cvt_sint<0x2C, VR128, GR64, i64, v2f64,
                               X86cvtts2Int, sdmem, sse_load_f64,
                               "cvttsd2si", WriteCvtSS2I>,
                               XD, VEX, VEX_LIG, VEX_W;
 }
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
 defm CVTTSS2SI : sse12_cvt_sint<0x2C, VR128, GR32, i32, v4f32, X86cvtts2Int,
                                     ssmem, sse_load_f32, "cvttss2si",
                                     WriteCvtSS2I>, XS;
 defm CVTTSS2SI64 : sse12_cvt_sint<0x2C, VR128, GR64, i64, v4f32,
                                    X86cvtts2Int, ssmem, sse_load_f32,
                                    "cvttss2si", WriteCvtSS2I>, XS, REX_W;
 defm CVTTSD2SI : sse12_cvt_sint<0x2C, VR128, GR32, i32, v2f64, X86cvtts2Int,
                                     sdmem, sse_load_f64, "cvttsd2si",
                                     WriteCvtSD2I>, XD;
 defm CVTTSD2SI64 : sse12_cvt_sint<0x2C, VR128, GR64, i64, v2f64,
                                   X86cvtts2Int, sdmem, sse_load_f64,
                                   "cvttsd2si", WriteCvtSD2I>, XD, REX_W;
+}
 
 def : InstAlias<"vcvttss2si{l}\t{$src, $dst|$dst, $src}",
                 (VCVTTSS2SIrr_Int GR32:$dst, VR128:$src), 0, "att">;
 def : InstAlias<"vcvttss2si{l}\t{$src, $dst|$dst, $src}",
                 (VCVTTSS2SIrm_Int GR32:$dst, f32mem:$src), 0, "att">;
 def : InstAlias<"vcvttsd2si{l}\t{$src, $dst|$dst, $src}",
                 (VCVTTSD2SIrr_Int GR32:$dst, VR128:$src), 0, "att">;
 def : InstAlias<"vcvttsd2si{l}\t{$src, $dst|$dst, $src}",
def : InstAlias<"cvttss2si{q}\t{$src, $dst|$dst, $src}",
                 (CVTTSS2SI64rr_Int GR64:$dst, VR128:$src), 0, "att">;
 def : InstAlias<"cvttss2si{q}\t{$src, $dst|$dst, $src}",
                 (CVTTSS2SI64rm_Int GR64:$dst, f32mem:$src), 0, "att">;
 def : InstAlias<"cvttsd2si{q}\t{$src, $dst|$dst, $src}",
                 (CVTTSD2SI64rr_Int GR64:$dst, VR128:$src), 0, "att">;
 def : InstAlias<"cvttsd2si{q}\t{$src, $dst|$dst, $src}",
                 (CVTTSD2SI64rm_Int GR64:$dst, f64mem:$src), 0, "att">;
 
-let Predicates = [UseAVX] in {
+let Predicates = [UseAVX], Uses = [MXCSR], mayRaiseFPException = 1 in {
 defm VCVTSS2SI   : sse12_cvt_sint<0x2D, VR128, GR32, i32, v4f32, X86cvts2si,
                                   ssmem, sse_load_f32, "cvtss2si",
                                   WriteCvtSS2I>, XS, VEX, VEX_LIG;
 defm VCVTSS2SI64 : sse12_cvt_sint<0x2D, VR128, GR64, i64, v4f32, X86cvts2si,
                                   ssmem, sse_load_f32, "cvtss2si",
                                   WriteCvtSS2I>, XS, VEX, VEX_W, VEX_LIG;
 }
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
 defm CVTSS2SI : sse12_cvt_sint<0x2D, VR128, GR32, i32, v4f32, X86cvts2si,
                                ssmem, sse_load_f32, "cvtss2si",
                                WriteCvtSS2I>, XS;
 defm CVTSS2SI64 : sse12_cvt_sint<0x2D, VR128, GR64, i64, v4f32, X86cvts2si,
                                  ssmem, sse_load_f32, "cvtss2si",
                                  WriteCvtSS2I>, XS, REX_W;
 
 defm VCVTDQ2PS   : sse12_cvt_p<0x5B, VR128, i128mem, v4f32, v4i32, load,
                                "vcvtdq2ps\t{$src, $dst|$dst, $src}",
                                SSEPackedSingle, WriteCvtI2PS>,
                                PS, VEX, Requires<[HasAVX, NoVLX]>, VEX_WIG;
 defm VCVTDQ2PSY  : sse12_cvt_p<0x5B, VR256, i256mem, v8f32, v8i32, load,
                                "vcvtdq2ps\t{$src, $dst|$dst, $src}",
                                SSEPackedSingle, WriteCvtI2PSY>,
                                PS, VEX, VEX_L, Requires<[HasAVX, NoVLX]>, VEX_WIG;
 
 defm CVTDQ2PS : sse12_cvt_p<0x5B, VR128, i128mem, v4f32, v4i32, memop,
                             "cvtdq2ps\t{$src, $dst|$dst, $src}",
                             SSEPackedSingle, WriteCvtI2PS>,
                             PS, Requires<[UseSSE2]>;
+}
 
 // AVX aliases
 def : InstAlias<"vcvtss2si{l}\t{$src, $dst|$dst, $src}",
                 (VCVTSS2SIrr_Int GR32:$dst, VR128:$src), 0, "att">;
 def : InstAlias<"vcvtss2si{l}\t{$src, $dst|$dst, $src}",
                 (VCVTSS2SIrm_Int GR32:$dst, ssmem:$src), 0, "att">;
 def : InstAlias<"vcvtsd2si{l}\t{$src, $dst|$dst, $src}",
                 (VCVTSD2SIrr_Int GR32:$dst, VR128:$src), 0, "att">;

 /// SSE 2 Only
 
 // Convert scalar double to scalar single
 let isCodeGenOnly = 1, hasSideEffects = 0, Predicates = [UseAVX] in {
 def VCVTSD2SSrr  : VSDI<0x5A, MRMSrcReg, (outs FR32:$dst),
                         (ins FR32:$src1, FR64:$src2),
                         "cvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}", []>,
                         VEX_4V, VEX_LIG, VEX_WIG,
-                        Sched<[WriteCvtSD2SS]>;
+                        Sched<[WriteCvtSD2SS]>, SIMD_EXP;
 let mayLoad = 1 in
 def VCVTSD2SSrm  : I<0x5A, MRMSrcMem, (outs FR32:$dst),
                      (ins FR32:$src1, f64mem:$src2),
                      "vcvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}", []>,
                      XD, VEX_4V, VEX_LIG, VEX_WIG,
-                     Sched<[WriteCvtSD2SS.Folded, WriteCvtSD2SS.ReadAfterFold]>;
+                     Sched<[WriteCvtSD2SS.Folded, WriteCvtSD2SS.ReadAfterFold]>, SIMD_EXP;
 }
 
 def : Pat<(f32 (fpround FR64:$src)),
             (VCVTSD2SSrr (f32 (IMPLICIT_DEF)), FR64:$src)>,
           Requires<[UseAVX]>;
 
 let isCodeGenOnly = 1 in {
 def CVTSD2SSrr  : SDI<0x5A, MRMSrcReg, (outs FR32:$dst), (ins FR64:$src),
                       "cvtsd2ss\t{$src, $dst|$dst, $src}",
                       [(set FR32:$dst, (fpround FR64:$src))]>,
-                      Sched<[WriteCvtSD2SS]>;
+                      Sched<[WriteCvtSD2SS]>, SIMD_EXP;
 def CVTSD2SSrm  : I<0x5A, MRMSrcMem, (outs FR32:$dst), (ins f64mem:$src),
                     "cvtsd2ss\t{$src, $dst|$dst, $src}",
                     [(set FR32:$dst, (fpround (loadf64 addr:$src)))]>,
                     XD, Requires<[UseSSE2, OptForSize]>,
-                    Sched<[WriteCvtSD2SS.Folded]>;
+                    Sched<[WriteCvtSD2SS.Folded]>, SIMD_EXP;
 }
 
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
 def VCVTSD2SSrr_Int: I<0x5A, MRMSrcReg,
                        (outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
                        "vcvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}",
                        [(set VR128:$dst,
                          (v4f32 (X86frounds VR128:$src1, (v2f64 VR128:$src2))))]>,
                        XD, VEX_4V, VEX_LIG, VEX_WIG, Requires<[UseAVX]>,
                        Sched<[WriteCvtSD2SS]>;
 def VCVTSD2SSrm_Int: I<0x5A, MRMSrcMem,

 def CVTSD2SSrm_Int: I<0x5A, MRMSrcMem,
                        (outs VR128:$dst), (ins VR128:$src1, sdmem:$src2),
                        "cvtsd2ss\t{$src2, $dst|$dst, $src2}",
                        [(set VR128:$dst,
                          (v4f32 (X86frounds VR128:$src1,sse_load_f64:$src2)))]>,
                        XD, Requires<[UseSSE2]>,
                        Sched<[WriteCvtSD2SS.Folded, WriteCvtSD2SS.ReadAfterFold]>;
 }
+}
 
 // Convert scalar single to scalar double
 // SSE2 instructions with XS prefix
 let isCodeGenOnly = 1, hasSideEffects = 0 in {
 def VCVTSS2SDrr : I<0x5A, MRMSrcReg, (outs FR64:$dst),
                     (ins FR64:$src1, FR32:$src2),
                     "vcvtss2sd\t{$src2, $src1, $dst|$dst, $src1, $src2}", []>,
                     XS, VEX_4V, VEX_LIG, VEX_WIG,
-                    Sched<[WriteCvtSS2SD]>, Requires<[UseAVX]>;
+                    Sched<[WriteCvtSS2SD]>, Requires<[UseAVX]>, SIMD_EXP;
 let mayLoad = 1 in
 def VCVTSS2SDrm : I<0x5A, MRMSrcMem, (outs FR64:$dst),
                     (ins FR64:$src1, f32mem:$src2),
                     "vcvtss2sd\t{$src2, $src1, $dst|$dst, $src1, $src2}", []>,
                     XS, VEX_4V, VEX_LIG, VEX_WIG,
                     Sched<[WriteCvtSS2SD.Folded, WriteCvtSS2SD.ReadAfterFold]>,
-                    Requires<[UseAVX, OptForSize]>;
+                    Requires<[UseAVX, OptForSize]>, SIMD_EXP;
 } // isCodeGenOnly = 1, hasSideEffects = 0
 
 def : Pat<(f64 (fpextend FR32:$src)),
     (VCVTSS2SDrr (f64 (IMPLICIT_DEF)), FR32:$src)>, Requires<[UseAVX]>;
 def : Pat<(fpextend (loadf32 addr:$src)),
     (VCVTSS2SDrm (f64 (IMPLICIT_DEF)), addr:$src)>, Requires<[UseAVX, OptForSize]>;
 
 let isCodeGenOnly = 1 in {
 def CVTSS2SDrr : I<0x5A, MRMSrcReg, (outs FR64:$dst), (ins FR32:$src),
                    "cvtss2sd\t{$src, $dst|$dst, $src}",
                    [(set FR64:$dst, (fpextend FR32:$src))]>,
-                   XS, Requires<[UseSSE2]>, Sched<[WriteCvtSS2SD]>;
+                   XS, Requires<[UseSSE2]>, Sched<[WriteCvtSS2SD]>, SIMD_EXP;
 def CVTSS2SDrm : I<0x5A, MRMSrcMem, (outs FR64:$dst), (ins f32mem:$src),
                    "cvtss2sd\t{$src, $dst|$dst, $src}",
                    [(set FR64:$dst, (fpextend (loadf32 addr:$src)))]>,
                    XS, Requires<[UseSSE2, OptForSize]>,
-                   Sched<[WriteCvtSS2SD.Folded]>;
+                   Sched<[WriteCvtSS2SD.Folded]>, SIMD_EXP;
 } // isCodeGenOnly = 1
 
-let hasSideEffects = 0 in {
+let hasSideEffects = 0, Uses = [MXCSR], mayRaiseFPException = 1 in {
 def VCVTSS2SDrr_Int: I<0x5A, MRMSrcReg,
                       (outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
                     "vcvtss2sd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
                     []>, XS, VEX_4V, VEX_LIG, VEX_WIG,
                     Requires<[HasAVX]>, Sched<[WriteCvtSS2SD]>;
 let mayLoad = 1 in
 def VCVTSS2SDrm_Int: I<0x5A, MRMSrcMem,
                       (outs VR128:$dst), (ins VR128:$src1, ssmem:$src2),
def : Pat<(v4f32 (X86Movss
           (CVTSI2SSrm_Int VR128:$dst, addr:$src)>;
 } // Predicates = [UseSSE1]
 
 let Predicates = [HasAVX, NoVLX] in {
 // Convert packed single/double fp to doubleword
 def VCVTPS2DQrr : VPDI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                        "cvtps2dq\t{$src, $dst|$dst, $src}",
                        [(set VR128:$dst, (v4i32 (X86cvtp2Int (v4f32 VR128:$src))))]>,
-                       VEX, Sched<[WriteCvtPS2I]>, VEX_WIG;
+                       VEX, Sched<[WriteCvtPS2I]>, VEX_WIG, SIMD_EXP;
 def VCVTPS2DQrm : VPDI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
                        "cvtps2dq\t{$src, $dst|$dst, $src}",
                        [(set VR128:$dst,
                          (v4i32 (X86cvtp2Int (loadv4f32 addr:$src))))]>,
-                       VEX, Sched<[WriteCvtPS2ILd]>, VEX_WIG;
+                       VEX, Sched<[WriteCvtPS2ILd]>, VEX_WIG, SIMD_EXP;
 def VCVTPS2DQYrr : VPDI<0x5B, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
                         "cvtps2dq\t{$src, $dst|$dst, $src}",
                         [(set VR256:$dst,
                           (v8i32 (X86cvtp2Int (v8f32 VR256:$src))))]>,
-                        VEX, VEX_L, Sched<[WriteCvtPS2IY]>, VEX_WIG;
+                        VEX, VEX_L, Sched<[WriteCvtPS2IY]>, VEX_WIG, SIMD_EXP;
 def VCVTPS2DQYrm : VPDI<0x5B, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
                         "cvtps2dq\t{$src, $dst|$dst, $src}",
                         [(set VR256:$dst,
                           (v8i32 (X86cvtp2Int (loadv8f32 addr:$src))))]>,
-                        VEX, VEX_L, Sched<[WriteCvtPS2IYLd]>, VEX_WIG;
+                        VEX, VEX_L, Sched<[WriteCvtPS2IYLd]>, VEX_WIG, SIMD_EXP;
 }
 def CVTPS2DQrr : PDI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                      "cvtps2dq\t{$src, $dst|$dst, $src}",
                      [(set VR128:$dst, (v4i32 (X86cvtp2Int (v4f32 VR128:$src))))]>,
-                     Sched<[WriteCvtPS2I]>;
+                     Sched<[WriteCvtPS2I]>, SIMD_EXP;
 def CVTPS2DQrm : PDI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
                      "cvtps2dq\t{$src, $dst|$dst, $src}",
                      [(set VR128:$dst,
                        (v4i32 (X86cvtp2Int (memopv4f32 addr:$src))))]>,
-                     Sched<[WriteCvtPS2ILd]>;
+                     Sched<[WriteCvtPS2ILd]>, SIMD_EXP;
 
 
 // Convert Packed Double FP to Packed DW Integers
-let Predicates = [HasAVX, NoVLX] in {
+let Predicates = [HasAVX, NoVLX], Uses = [MXCSR], mayRaiseFPException = 1 in {
 // The assembler can recognize rr 256-bit instructions by seeing a ymm
 // register, but the same isn't true when using memory operands instead.
 // Provide other assembly rr and rm forms to address this explicitly.
 def VCVTPD2DQrr  : SDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                        "vcvtpd2dq\t{$src, $dst|$dst, $src}",
                        [(set VR128:$dst,
                          (v4i32 (X86cvtp2Int (v2f64 VR128:$src))))]>,
                        VEX, Sched<[WriteCvtPD2I]>, VEX_WIG;
def : InstAlias<"vcvtpd2dqx\t{$src, $dst|$dst, $src}",
                 (VCVTPD2DQrr VR128:$dst, VR128:$src), 0, "att">;
 def : InstAlias<"vcvtpd2dqy\t{$src, $dst|$dst, $src}",
                 (VCVTPD2DQYrr VR128:$dst, VR256:$src), 0, "att">;
 
 def CVTPD2DQrm  : SDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
                       "cvtpd2dq\t{$src, $dst|$dst, $src}",
                       [(set VR128:$dst,
                         (v4i32 (X86cvtp2Int (memopv2f64 addr:$src))))]>,
-                      Sched<[WriteCvtPD2ILd]>;
+                      Sched<[WriteCvtPD2ILd]>, SIMD_EXP;
 def CVTPD2DQrr  : SDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                       "cvtpd2dq\t{$src, $dst|$dst, $src}",
                       [(set VR128:$dst,
                         (v4i32 (X86cvtp2Int (v2f64 VR128:$src))))]>,
-                      Sched<[WriteCvtPD2I]>;
+                      Sched<[WriteCvtPD2I]>, SIMD_EXP;
 
 // Convert with truncation packed single/double fp to doubleword
 // SSE2 packed instructions with XS prefix
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
 let Predicates = [HasAVX, NoVLX] in {
 def VCVTTPS2DQrr : VS2SI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                          "cvttps2dq\t{$src, $dst|$dst, $src}",
                          [(set VR128:$dst,
                            (v4i32 (X86cvttp2si (v4f32 VR128:$src))))]>,
                          VEX, Sched<[WriteCvtPS2I]>, VEX_WIG;
 def VCVTTPS2DQrm : VS2SI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
                          "cvttps2dq\t{$src, $dst|$dst, $src}",
def CVTTPS2DQrr : S2SI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                        [(set VR128:$dst,
                          (v4i32 (X86cvttp2si (v4f32 VR128:$src))))]>,
                        Sched<[WriteCvtPS2I]>;
 def CVTTPS2DQrm : S2SI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
                        "cvttps2dq\t{$src, $dst|$dst, $src}",
                        [(set VR128:$dst,
                          (v4i32 (X86cvttp2si (memopv4f32 addr:$src))))]>,
                        Sched<[WriteCvtPS2ILd]>;
+}
 
 // The assembler can recognize rr 256-bit instructions by seeing a ymm
 // register, but the same isn't true when using memory operands instead.
 // Provide other assembly rr and rm forms to address this explicitly.
-let Predicates = [HasAVX, NoVLX] in {
+let Predicates = [HasAVX, NoVLX], Uses = [MXCSR], mayRaiseFPException = 1 in {
 // XMM only
 def VCVTTPD2DQrr : VPDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                         "cvttpd2dq\t{$src, $dst|$dst, $src}",
                         [(set VR128:$dst,
                           (v4i32 (X86cvttp2si (v2f64 VR128:$src))))]>,
                         VEX, Sched<[WriteCvtPD2I]>, VEX_WIG;
 def VCVTTPD2DQrm : VPDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
                         "cvttpd2dq{x}\t{$src, $dst|$dst, $src}",
let Predicates = [HasAVX, NoVLX] in {
   def : Pat<(v4i32 (fp_to_sint (loadv4f64 addr:$src))),
             (VCVTTPD2DQYrm addr:$src)>;
 }
 
 def CVTTPD2DQrr : PDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                       "cvttpd2dq\t{$src, $dst|$dst, $src}",
                       [(set VR128:$dst,
                         (v4i32 (X86cvttp2si (v2f64 VR128:$src))))]>,
-                      Sched<[WriteCvtPD2I]>;
+                      Sched<[WriteCvtPD2I]>, SIMD_EXP;
 def CVTTPD2DQrm : PDI<0xE6, MRMSrcMem, (outs VR128:$dst),(ins f128mem:$src),
                       "cvttpd2dq\t{$src, $dst|$dst, $src}",
                       [(set VR128:$dst,
                         (v4i32 (X86cvttp2si (memopv2f64 addr:$src))))]>,
-                      Sched<[WriteCvtPD2ILd]>;
+                      Sched<[WriteCvtPD2ILd]>, SIMD_EXP;
 
 // Convert packed single to packed double
-let Predicates = [HasAVX, NoVLX] in {
+let Predicates = [HasAVX, NoVLX], Uses = [MXCSR], mayRaiseFPException = 1 in {
                   // SSE2 instructions without OpSize prefix
 def VCVTPS2PDrr : I<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                     "vcvtps2pd\t{$src, $dst|$dst, $src}",
                     [(set VR128:$dst, (v2f64 (X86vfpext (v4f32 VR128:$src))))]>,
                     PS, VEX, Sched<[WriteCvtPS2PD]>, VEX_WIG;
 def VCVTPS2PDrm : I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src),
                     "vcvtps2pd\t{$src, $dst|$dst, $src}",
                     [(set VR128:$dst, (v2f64 (extloadv2f32 addr:$src)))]>,
                     PS, VEX, Sched<[WriteCvtPS2PD.Folded]>, VEX_WIG;
 def VCVTPS2PDYrr : I<0x5A, MRMSrcReg, (outs VR256:$dst), (ins VR128:$src),
                      "vcvtps2pd\t{$src, $dst|$dst, $src}",
                      [(set VR256:$dst, (v4f64 (fpextend (v4f32 VR128:$src))))]>,
                      PS, VEX, VEX_L, Sched<[WriteCvtPS2PDY]>, VEX_WIG;
 def VCVTPS2PDYrm : I<0x5A, MRMSrcMem, (outs VR256:$dst), (ins f128mem:$src),
                      "vcvtps2pd\t{$src, $dst|$dst, $src}",
                      [(set VR256:$dst, (v4f64 (extloadv4f32 addr:$src)))]>,
                      PS, VEX, VEX_L, Sched<[WriteCvtPS2PDY.Folded]>, VEX_WIG;
 }
 
-let Predicates = [UseSSE2] in {
+let Predicates = [UseSSE2], Uses = [MXCSR], mayRaiseFPException = 1 in {
 def CVTPS2PDrr : I<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                    "cvtps2pd\t{$src, $dst|$dst, $src}",
                    [(set VR128:$dst, (v2f64 (X86vfpext (v4f32 VR128:$src))))]>,
                    PS, Sched<[WriteCvtPS2PD]>;
 def CVTPS2PDrm : I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src),
                    "cvtps2pd\t{$src, $dst|$dst, $src}",
                    [(set VR128:$dst, (v2f64 (extloadv2f32 addr:$src)))]>,
                    PS, Sched<[WriteCvtPS2PD.Folded]>;
let Predicates = [UseSSE2] in {
   def : Pat<(v2f64 (X86VSintToFP (bc_v4i32 (v2i64 (X86vzload64 addr:$src))))),
             (CVTDQ2PDrm addr:$src)>;
 } // Predicates = [UseSSE2]
 
 // Convert packed double to packed single
 // The assembler can recognize rr 256-bit instructions by seeing a ymm
 // register, but the same isn't true when using memory operands instead.
 // Provide other assembly rr and rm forms to address this explicitly.
-let Predicates = [HasAVX, NoVLX] in {
+let Predicates = [HasAVX, NoVLX], Uses = [MXCSR], mayRaiseFPException = 1 in {
 // XMM only
 def VCVTPD2PSrr : VPDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                        "cvtpd2ps\t{$src, $dst|$dst, $src}",
                        [(set VR128:$dst, (X86vfpround (v2f64 VR128:$src)))]>,
                        VEX, Sched<[WriteCvtPD2PS]>, VEX_WIG;
 def VCVTPD2PSrm : VPDI<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
                        "cvtpd2ps{x}\t{$src, $dst|$dst, $src}",
                        [(set VR128:$dst, (X86vfpround (loadv2f64 addr:$src)))]>,

 def : InstAlias<"vcvtpd2psx\t{$src, $dst|$dst, $src}",
                 (VCVTPD2PSrr VR128:$dst, VR128:$src), 0, "att">;
 def : InstAlias<"vcvtpd2psy\t{$src, $dst|$dst, $src}",
                 (VCVTPD2PSYrr VR128:$dst, VR256:$src), 0, "att">;
 
 def CVTPD2PSrr : PDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
                      "cvtpd2ps\t{$src, $dst|$dst, $src}",
                      [(set VR128:$dst, (X86vfpround (v2f64 VR128:$src)))]>,
-                     Sched<[WriteCvtPD2PS]>;
+                     Sched<[WriteCvtPD2PS]>, SIMD_EXP;
 def CVTPD2PSrm : PDI<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
                      "cvtpd2ps\t{$src, $dst|$dst, $src}",
                      [(set VR128:$dst, (X86vfpround (memopv2f64 addr:$src)))]>,
-                     Sched<[WriteCvtPD2PS.Folded]>;
+                     Sched<[WriteCvtPD2PS.Folded]>, SIMD_EXP;
 
 let Predicates = [HasAVX, NoVLX] in {
   def : Pat<(v4f32 (fpround (v4f64 VR256:$src))),
             (VCVTPD2PSYrr VR256:$src)>;
   def : Pat<(v4f32 (fpround (loadv4f64 addr:$src))),
             (VCVTPD2PSYrm addr:$src)>;
 }
 
 //===----------------------------------------------------------------------===//
 // SSE 1 & 2 - Compare Instructions
 //===----------------------------------------------------------------------===//
 
 // sse12_cmp_scalar - sse 1 & 2 compare scalar instructions
 multiclass sse12_cmp_scalar<RegisterClass RC, X86MemOperand x86memop,
                             SDNode OpNode, ValueType VT,
                             PatFrag ld_frag, string asm,
                             X86FoldableSchedWrite sched> {
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
   let isCommutable = 1 in
   def rr : SIi8<0xC2, MRMSrcReg,
                 (outs RC:$dst), (ins RC:$src1, RC:$src2, u8imm:$cc), asm,
                 [(set RC:$dst, (OpNode (VT RC:$src1), RC:$src2, timm:$cc))]>,
                 Sched<[sched]>;
   def rm : SIi8<0xC2, MRMSrcMem,
                 (outs RC:$dst), (ins RC:$src1, x86memop:$src2, u8imm:$cc), asm,
                 [(set RC:$dst, (OpNode (VT RC:$src1),
                                          (ld_frag addr:$src2), timm:$cc))]>,
                 Sched<[sched.Folded, sched.ReadAfterFold]>;
 }
+}
 
 let isCodeGenOnly = 1 in {
   let ExeDomain = SSEPackedSingle in
   defm VCMPSS : sse12_cmp_scalar<FR32, f32mem, X86cmps, f32, loadf32,
                    "cmpss\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}",
                    SchedWriteFCmpSizes.PS.Scl>, XS, VEX_4V, VEX_LIG, VEX_WIG;
   let ExeDomain = SSEPackedDouble in
   defm VCMPSD : sse12_cmp_scalar<FR64, f64mem, X86cmps, f64, loadf64,
defm CMPSD : sse12_cmp_scalar<FR64, f64mem, X86cmps, f64, loadf64,
                     "cmpsd\t{$cc, $src2, $dst|$dst, $src2, $cc}",
                     SchedWriteFCmpSizes.PD.Scl>, XD;
   }
 }
 
 multiclass sse12_cmp_scalar_int<Operand memop,
                          Intrinsic Int, string asm, X86FoldableSchedWrite sched,
                          ComplexPattern mem_cpat> {
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
   def rr_Int : SIi8<0xC2, MRMSrcReg, (outs VR128:$dst),
                       (ins VR128:$src1, VR128:$src, u8imm:$cc), asm,
                         [(set VR128:$dst, (Int VR128:$src1,
                                                VR128:$src, timm:$cc))]>,
            Sched<[sched]>;
 let mayLoad = 1 in
   def rm_Int : SIi8<0xC2, MRMSrcMem, (outs VR128:$dst),
                       (ins VR128:$src1, memop:$src, u8imm:$cc), asm,
                         [(set VR128:$dst, (Int VR128:$src1,
                                                mem_cpat:$src, timm:$cc))]>,
            Sched<[sched.Folded, sched.ReadAfterFold]>;
 }
+}
 
 // Aliases to match intrinsics which expect XMM operand(s).
 let ExeDomain = SSEPackedSingle in
 defm VCMPSS  : sse12_cmp_scalar_int<ssmem, int_x86_sse_cmp_ss,
                      "cmpss\t{$cc, $src, $src1, $dst|$dst, $src1, $src, $cc}",
                      SchedWriteFCmpSizes.PS.Scl, sse_load_f32>,
                      XS, VEX_4V, VEX_LIG, VEX_WIG;
 let ExeDomain = SSEPackedDouble in

 }
 
 
 // sse12_ord_cmp - Unordered/Ordered scalar fp compare and set EFLAGS
 multiclass sse12_ord_cmp<bits<8> opc, RegisterClass RC, SDNode OpNode,
                          ValueType vt, X86MemOperand x86memop,
                          PatFrag ld_frag, string OpcodeStr,
                          X86FoldableSchedWrite sched> {
-let hasSideEffects = 0 in {
+let hasSideEffects = 0, Uses = [MXCSR], mayRaiseFPException = 1 in {
   def rr: SI<opc, MRMSrcReg, (outs), (ins RC:$src1, RC:$src2),
                      !strconcat(OpcodeStr, "\t{$src2, $src1|$src1, $src2}"),
                      [(set EFLAGS, (OpNode (vt RC:$src1), RC:$src2))]>,
           Sched<[sched]>;
 let mayLoad = 1 in
   def rm: SI<opc, MRMSrcMem, (outs), (ins RC:$src1, x86memop:$src2),
                      !strconcat(OpcodeStr, "\t{$src2, $src1|$src1, $src2}"),
                      [(set EFLAGS, (OpNode (vt RC:$src1),
                                            (ld_frag addr:$src2)))]>,
           Sched<[sched.Folded, sched.ReadAfterFold]>;
 }
 }
 
 // sse12_ord_cmp_int - Intrinsic version of sse12_ord_cmp
 multiclass sse12_ord_cmp_int<bits<8> opc, RegisterClass RC, SDNode OpNode,
                              ValueType vt, Operand memop,
                              ComplexPattern mem_cpat, string OpcodeStr,
                              X86FoldableSchedWrite sched> {
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
   def rr_Int: SI<opc, MRMSrcReg, (outs), (ins RC:$src1, RC:$src2),
                      !strconcat(OpcodeStr, "\t{$src2, $src1|$src1, $src2}"),
                      [(set EFLAGS, (OpNode (vt RC:$src1), RC:$src2))]>,
           Sched<[sched]>;
 let mayLoad = 1 in
   def rm_Int: SI<opc, MRMSrcMem, (outs), (ins RC:$src1, memop:$src2),
                      !strconcat(OpcodeStr, "\t{$src2, $src1|$src1, $src2}"),
                      [(set EFLAGS, (OpNode (vt RC:$src1),
                                            mem_cpat:$src2))]>,
           Sched<[sched.Folded, sched.ReadAfterFold]>;
 }
+}
 
 let Defs = [EFLAGS] in {
   defm VUCOMISS : sse12_ord_cmp<0x2E, FR32, X86cmp, f32, f32mem, loadf32,
                                "ucomiss", WriteFCom>, PS, VEX, VEX_LIG, VEX_WIG;
   defm VUCOMISD : sse12_ord_cmp<0x2E, FR64, X86cmp, f64, f64mem, loadf64,
                                "ucomisd", WriteFCom>, PD, VEX, VEX_LIG, VEX_WIG;
   let Pattern = []<dag> in {
     defm VCOMISS  : sse12_ord_cmp<0x2F, FR32, undef, f32, f32mem, loadf32,
let Defs = [EFLAGS] in {
   }
 } // Defs = [EFLAGS]
 
 // sse12_cmp_packed - sse 1 & 2 compare packed instructions
 multiclass sse12_cmp_packed<RegisterClass RC, X86MemOperand x86memop,
                             ValueType VT, string asm,
                             X86FoldableSchedWrite sched,
                             Domain d, PatFrag ld_frag> {
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
   let isCommutable = 1 in
   def rri : PIi8<0xC2, MRMSrcReg,
              (outs RC:$dst), (ins RC:$src1, RC:$src2, u8imm:$cc), asm,
              [(set RC:$dst, (VT (X86cmpp RC:$src1, RC:$src2, timm:$cc)))], d>,
             Sched<[sched]>;
   def rmi : PIi8<0xC2, MRMSrcMem,
              (outs RC:$dst), (ins RC:$src1, x86memop:$src2, u8imm:$cc), asm,
              [(set RC:$dst,
                (VT (X86cmpp RC:$src1, (ld_frag addr:$src2), timm:$cc)))], d>,
             Sched<[sched.Folded, sched.ReadAfterFold]>;
 }
+}
 
 defm VCMPPS : sse12_cmp_packed<VR128, f128mem, v4f32,
                "cmpps\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}",
                SchedWriteFCmpSizes.PS.XMM, SSEPackedSingle, loadv4f32>, PS, VEX_4V, VEX_WIG;
 defm VCMPPD : sse12_cmp_packed<VR128, f128mem, v2f64,
                "cmppd\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}",
                SchedWriteFCmpSizes.PD.XMM, SSEPackedDouble, loadv2f64>, PD, VEX_4V, VEX_WIG;
 defm VCMPPSY : sse12_cmp_packed<VR256, f256mem, v8f32,

 ///
 /// These three forms can each be reg+reg or reg+mem.
 ///
 
 /// FIXME: once all 256-bit intrinsics are matched, cleanup and refactor those
 /// classes below
 multiclass basic_sse12_fp_binop_p<bits<8> opc, string OpcodeStr,
                                   SDNode OpNode, X86SchedWriteSizes sched> {
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
   let Predicates = [HasAVX, NoVLX] in {
   defm V#NAME#PS : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode,
                                VR128, v4f32, f128mem, loadv4f32,
                                SSEPackedSingle, sched.PS.XMM, 0>, PS, VEX_4V, VEX_WIG;
   defm V#NAME#PD : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode,
                                VR128, v2f64, f128mem, loadv2f64,
                                SSEPackedDouble, sched.PD.XMM, 0>, PD, VEX_4V, VEX_WIG;
 
let Constraints = "$src1 = $dst" in {
     defm PS : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode, VR128,
                               v4f32, f128mem, memopv4f32, SSEPackedSingle,
                               sched.PS.XMM>, PS;
     defm PD : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode, VR128,
                               v2f64, f128mem, memopv2f64, SSEPackedDouble,
                               sched.PD.XMM>, PD;
   }
 }
+}
 
 multiclass basic_sse12_fp_binop_s<bits<8> opc, string OpcodeStr, SDNode OpNode,
                                   X86SchedWriteSizes sched> {
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
   defm V#NAME#SS : sse12_fp_scalar<opc, !strconcat(OpcodeStr, "ss"),
                          OpNode, FR32, f32mem, SSEPackedSingle, sched.PS.Scl, 0>,
                          XS, VEX_4V, VEX_LIG, VEX_WIG;
   defm V#NAME#SD : sse12_fp_scalar<opc, !strconcat(OpcodeStr, "sd"),
                          OpNode, FR64, f64mem, SSEPackedDouble, sched.PD.Scl, 0>,
                          XD, VEX_4V, VEX_LIG, VEX_WIG;
 
   let Constraints = "$src1 = $dst" in {
     defm SS : sse12_fp_scalar<opc, !strconcat(OpcodeStr, "ss"),
                               OpNode, FR32, f32mem, SSEPackedSingle,
                               sched.PS.Scl>, XS;
     defm SD : sse12_fp_scalar<opc, !strconcat(OpcodeStr, "sd"),
                               OpNode, FR64, f64mem, SSEPackedDouble,
                               sched.PD.Scl>, XD;
   }
 }
+}
 
 multiclass basic_sse12_fp_binop_s_int<bits<8> opc, string OpcodeStr,
                                       SDPatternOperator OpNode,
                                       X86SchedWriteSizes sched> {
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
   defm V#NAME#SS : sse12_fp_scalar_int<opc, OpcodeStr, OpNode, VR128, v4f32,
                    !strconcat(OpcodeStr, "ss"), ssmem, sse_load_f32,
                    SSEPackedSingle, sched.PS.Scl, 0>, XS, VEX_4V, VEX_LIG, VEX_WIG;
   defm V#NAME#SD : sse12_fp_scalar_int<opc, OpcodeStr, OpNode, VR128, v2f64,
                    !strconcat(OpcodeStr, "sd"), sdmem, sse_load_f64,
                    SSEPackedDouble, sched.PD.Scl, 0>, XD, VEX_4V, VEX_LIG, VEX_WIG;
 
   let Constraints = "$src1 = $dst" in {
     defm SS : sse12_fp_scalar_int<opc, OpcodeStr, OpNode, VR128, v4f32,
                    !strconcat(OpcodeStr, "ss"), ssmem, sse_load_f32,
                    SSEPackedSingle, sched.PS.Scl>, XS;
     defm SD : sse12_fp_scalar_int<opc, OpcodeStr, OpNode, VR128, v2f64,
                    !strconcat(OpcodeStr, "sd"), sdmem, sse_load_f64,
                    SSEPackedDouble, sched.PD.Scl>, XD;
   }
 }
+}
 
 // Binary Arithmetic instructions
 defm ADD : basic_sse12_fp_binop_p<0x58, "add", fadd, SchedWriteFAddSizes>,
            basic_sse12_fp_binop_s<0x58, "add", fadd, SchedWriteFAddSizes>,
            basic_sse12_fp_binop_s_int<0x58, "add", null_frag, SchedWriteFAddSizes>;
 defm MUL : basic_sse12_fp_binop_p<0x59, "mul", fmul, SchedWriteFMulSizes>,
            basic_sse12_fp_binop_s<0x59, "mul", fmul, SchedWriteFMulSizes>,
            basic_sse12_fp_binop_s_int<0x59, "mul", null_frag, SchedWriteFMulSizes>;
defm V#NAME#SD  : avx_fp_unop_s<opc, "v"#OpcodeStr##sd, FR64, f64,
                          f64mem, sdmem, OpNode, SSEPackedDouble, sched.Scl, AVXTarget>,
                          XD, VEX_4V, VEX_LIG, VEX_WIG;
 }
 
 // Square root.
 defm SQRT  : sse1_fp_unop_s<0x51, "sqrt", fsqrt, SchedWriteFSqrt, UseAVX>,
              sse1_fp_unop_p<0x51, "sqrt", fsqrt, SchedWriteFSqrt, [HasAVX, NoVLX]>,
              sse2_fp_unop_s<0x51, "sqrt", fsqrt, SchedWriteFSqrt64, UseAVX>,
-             sse2_fp_unop_p<0x51, "sqrt", fsqrt, SchedWriteFSqrt64>;
+             sse2_fp_unop_p<0x51, "sqrt", fsqrt, SchedWriteFSqrt64>, SIMD_EXP;
 
 // Reciprocal approximations. Note that these typically require refinement
 // in order to obtain suitable precision.
 defm RSQRT : sse1_fp_unop_s<0x52, "rsqrt", X86frsqrt, SchedWriteFRsqrt, HasAVX>,
              sse1_fp_unop_s_intr<0x52, "rsqrt", X86frsqrt, SchedWriteFRsqrt, HasAVX>,
              sse1_fp_unop_p<0x52, "rsqrt", X86frsqrt, SchedWriteFRsqrt, [HasAVX]>;
 defm RCP   : sse1_fp_unop_s<0x53, "rcp", X86frcp, SchedWriteFRcp, HasAVX>,
              sse1_fp_unop_s_intr<0x53, "rcp", X86frcp, SchedWriteFRcp, HasAVX>,

 
 //===---------------------------------------------------------------------===//
 // SSE3 - Arithmetic
 //===---------------------------------------------------------------------===//
 
 multiclass sse3_addsub<string OpcodeStr, ValueType vt, RegisterClass RC,
                        X86MemOperand x86memop, X86FoldableSchedWrite sched,
                        PatFrag ld_frag, bit Is2Addr = 1> {
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
   def rr : I<0xD0, MRMSrcReg,
        (outs RC:$dst), (ins RC:$src1, RC:$src2),
        !if(Is2Addr,
            !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
            !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
        [(set RC:$dst, (vt (X86Addsub RC:$src1, RC:$src2)))]>,
        Sched<[sched]>;
   def rm : I<0xD0, MRMSrcMem,
        (outs RC:$dst), (ins RC:$src1, x86memop:$src2),
        !if(Is2Addr,
            !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
            !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
        [(set RC:$dst, (vt (X86Addsub RC:$src1, (ld_frag addr:$src2))))]>,
        Sched<[sched.Folded, sched.ReadAfterFold]>;
 }
+}
 
 let Predicates = [HasAVX] in {
   let ExeDomain = SSEPackedSingle in {
     defm VADDSUBPS : sse3_addsub<"vaddsubps", v4f32, VR128, f128mem,
                                  SchedWriteFAddSizes.PS.XMM, loadv4f32, 0>,
                                  XD, VEX_4V, VEX_WIG;
     defm VADDSUBPSY : sse3_addsub<"vaddsubps", v8f32, VR256, f256mem,
                                   SchedWriteFAddSizes.PS.YMM, loadv8f32, 0>,

 // SSE3 Instructions
 //===---------------------------------------------------------------------===//
 
 // Horizontal ops
 multiclass S3D_Int<bits<8> o, string OpcodeStr, ValueType vt, RegisterClass RC,
                    X86MemOperand x86memop, SDNode OpNode,
                    X86FoldableSchedWrite sched, PatFrag ld_frag,
                    bit Is2Addr = 1> {
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
   def rr : S3DI<o, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2),
        !if(Is2Addr,
          !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
          !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
       [(set RC:$dst, (vt (OpNode RC:$src1, RC:$src2)))]>,
       Sched<[sched]>;
 
   def rm : S3DI<o, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2),
        !if(Is2Addr,
          !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
          !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
       [(set RC:$dst, (vt (OpNode RC:$src1, (ld_frag addr:$src2))))]>,
       Sched<[sched.Folded, sched.ReadAfterFold]>;
 }
+}
 multiclass S3_Int<bits<8> o, string OpcodeStr, ValueType vt, RegisterClass RC,
                   X86MemOperand x86memop, SDNode OpNode,
                   X86FoldableSchedWrite sched, PatFrag ld_frag,
                   bit Is2Addr = 1> {
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
   def rr : S3I<o, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2),
        !if(Is2Addr,
          !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
          !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
       [(set RC:$dst, (vt (OpNode RC:$src1, RC:$src2)))]>,
         Sched<[sched]>;
 
   def rm : S3I<o, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2),
        !if(Is2Addr,
          !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
          !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
       [(set RC:$dst, (vt (OpNode RC:$src1, (ld_frag addr:$src2))))]>,
         Sched<[sched.Folded, sched.ReadAfterFold]>;
 }
+}
 
 let Predicates = [HasAVX] in {
   let ExeDomain = SSEPackedSingle in {
     defm VHADDPS  : S3D_Int<0x7C, "vhaddps", v4f32, VR128, f128mem,
                             X86fhadd, WriteFHAdd, loadv4f32, 0>, VEX_4V, VEX_WIG;
     defm VHSUBPS  : S3D_Int<0x7D, "vhsubps", v4f32, VR128, f128mem,
                             X86fhsub, WriteFHAdd, loadv4f32, 0>, VEX_4V, VEX_WIG;
     defm VHADDPSY : S3D_Int<0x7C, "vhaddps", v8f32, VR256, f256mem,

 //===----------------------------------------------------------------------===//
 
 multiclass sse41_fp_unop_p<bits<8> opc, string OpcodeStr,
                            X86MemOperand x86memop, RegisterClass RC,
                            ValueType VT, PatFrag mem_frag, SDNode OpNode,
                            X86FoldableSchedWrite sched> {
   // Intrinsic operation, reg.
   // Vector intrinsic operation, reg
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
   def r : SS4AIi8<opc, MRMSrcReg,
                   (outs RC:$dst), (ins RC:$src1, i32u8imm:$src2),
                   !strconcat(OpcodeStr,
                   "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
                   [(set RC:$dst, (VT (OpNode RC:$src1, timm:$src2)))]>,
                   Sched<[sched]>;
 
   // Vector intrinsic operation, mem
   def m : SS4AIi8<opc, MRMSrcMem,
                   (outs RC:$dst), (ins x86memop:$src1, i32u8imm:$src2),
                   !strconcat(OpcodeStr,
                   "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
                   [(set RC:$dst,
                         (VT (OpNode (mem_frag addr:$src1), timm:$src2)))]>,
                   Sched<[sched.Folded]>;
 }
+}
 
 multiclass avx_fp_unop_rm<bits<8> opcss, bits<8> opcsd,
                           string OpcodeStr, X86FoldableSchedWrite sched> {
 let ExeDomain = SSEPackedSingle, hasSideEffects = 0, isCodeGenOnly = 1 in {
   def SSr : SS4AIi8<opcss, MRMSrcReg,
         (outs FR32:$dst), (ins FR32:$src1, FR32:$src2, i32u8imm:$src3),
         !strconcat(OpcodeStr,
             "ss\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
def SDm : SS4AIi8<opcsd, MRMSrcMem,
         !strconcat(OpcodeStr,
              "sd\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
         []>, Sched<[sched.Folded, sched.ReadAfterFold]>;
 } // ExeDomain = SSEPackedDouble, hasSideEffects = 0
 }
 
 multiclass sse41_fp_unop_s<bits<8> opcss, bits<8> opcsd,
                            string OpcodeStr, X86FoldableSchedWrite sched> {
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
 let ExeDomain = SSEPackedSingle, hasSideEffects = 0, isCodeGenOnly = 1 in {
   def SSr : SS4AIi8<opcss, MRMSrcReg,
                     (outs FR32:$dst), (ins FR32:$src1, i32u8imm:$src2),
                     !strconcat(OpcodeStr,
                                "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
                     []>, Sched<[sched]>;
 
   let mayLoad = 1 in
let ExeDomain = SSEPackedDouble, hasSideEffects = 0, isCodeGenOnly = 1 in {
   let mayLoad = 1 in
   def SDm : SS4AIi8<opcsd, MRMSrcMem,
                     (outs FR64:$dst), (ins f64mem:$src1, i32u8imm:$src2),
                     !strconcat(OpcodeStr,
                                "sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
                     []>, Sched<[sched.Folded, sched.ReadAfterFold]>;
 } // ExeDomain = SSEPackedDouble, hasSideEffects = 0
 }
+}
 
 multiclass sse41_fp_binop_s<bits<8> opcss, bits<8> opcsd,
                             string OpcodeStr, X86FoldableSchedWrite sched,
                             ValueType VT32, ValueType VT64,
                             SDNode OpNode, bit Is2Addr = 1> {
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
 let ExeDomain = SSEPackedSingle in {
   def SSr_Int : SS4AIi8<opcss, MRMSrcReg,
         (outs VR128:$dst), (ins VR128:$src1, VR128:$src2, i32u8imm:$src3),
         !if(Is2Addr,
             !strconcat(OpcodeStr,
                 "ss\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
             !strconcat(OpcodeStr,
                 "ss\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}")),
def SDm_Int : SS4AIi8<opcsd, MRMSrcMem,
                 "sd\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
             !strconcat(OpcodeStr,
                 "sd\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}")),
         [(set VR128:$dst,
               (OpNode VR128:$src1, sse_load_f64:$src2, timm:$src3))]>,
         Sched<[sched.Folded, sched.ReadAfterFold]>;
 } // ExeDomain = SSEPackedDouble, isCodeGenOnly = 1
 }
+}
 
 // FP round - roundss, roundps, roundsd, roundpd
 let Predicates = [HasAVX, NoVLX] in {
   let ExeDomain = SSEPackedSingle in {
     // Intrinsic form
     defm VROUNDPS  : sse41_fp_unop_p<0x08, "vroundps", f128mem, VR128, v4f32,
                                      loadv4f32, X86VRndScale, SchedWriteFRnd.XMM>,
                                    VEX, VEX_WIG;

 
 let Predicates = [HasAVX] in {
   let isCommutable = 0 in {
     defm VMPSADBW : SS41I_binop_rmi_int<0x42, "vmpsadbw", int_x86_sse41_mpsadbw,
                                         VR128, load, i128mem, 0,
                                         SchedWriteMPSAD.XMM>, VEX_4V, VEX_WIG;
   }
 
+let Uses = [MXCSR], mayRaiseFPException = 1 in {
   let ExeDomain = SSEPackedSingle in
   defm VDPPS : SS41I_binop_rmi_int<0x40, "vdpps", int_x86_sse41_dpps,
                                    VR128, load, f128mem, 0,
                                    SchedWriteDPPS.XMM>, VEX_4V, VEX_WIG;
   let ExeDomain = SSEPackedDouble in
   defm VDPPD : SS41I_binop_rmi_int<0x41, "vdppd", int_x86_sse41_dppd,
                                    VR128, load, f128mem, 0,
                                    SchedWriteDPPD.XMM>, VEX_4V, VEX_WIG;
   let ExeDomain = SSEPackedSingle in
   defm VDPPSY : SS41I_binop_rmi_int<0x40, "vdpps", int_x86_avx_dp_ps_256,
                                     VR256, load, i256mem, 0,
                                     SchedWriteDPPS.YMM>, VEX_4V, VEX_L, VEX_WIG;
 }
+}
 
 let Predicates = [HasAVX2] in {
   let isCommutable = 0 in {
   defm VMPSADBWY : SS41I_binop_rmi_int<0x42, "vmpsadbw", int_x86_avx2_mpsadbw,
                                   VR256, load, i256mem, 0,
                                   SchedWriteMPSAD.YMM>, VEX_4V, VEX_L, VEX_WIG;
   }
 }
 
 let Constraints = "$src1 = $dst" in {
   let isCommutable = 0 in {
   defm MPSADBW : SS41I_binop_rmi_int<0x42, "mpsadbw", int_x86_sse41_mpsadbw,
                                      VR128, memop, i128mem, 1,
                                      SchedWriteMPSAD.XMM>;
   }
 
   let ExeDomain = SSEPackedSingle in
   defm DPPS : SS41I_binop_rmi_int<0x40, "dpps", int_x86_sse41_dpps,
                                   VR128, memop, f128mem, 1,
-                                  SchedWriteDPPS.XMM>;
+                                  SchedWriteDPPS.XMM>, SIMD_EXP;
   let ExeDomain = SSEPackedDouble in
   defm DPPD : SS41I_binop_rmi_int<0x41, "dppd", int_x86_sse41_dppd,
                                   VR128, memop, f128mem, 1,
-                                  SchedWriteDPPD.XMM>;
+                                  SchedWriteDPPD.XMM>, SIMD_EXP;
 }
 
 /// SS41I_blend_rmi - SSE 4.1 blend with 8-bit immediate
 multiclass SS41I_blend_rmi<bits<8> opc, string OpcodeStr, SDNode OpNode,
                            ValueType OpVT, RegisterClass RC, PatFrag memop_frag,
                            X86MemOperand x86memop, bit Is2Addr, Domain d,
                            X86FoldableSchedWrite sched, SDNodeXForm commuteXForm> {
 let ExeDomain = d, Constraints = !if(Is2Addr, "$src1 = $dst", "") in {

llvm/lib/Target/X86/X86RegisterInfo.cpp

 
 BitVector X86RegisterInfo::getReservedRegs(const MachineFunction &MF) const {
   BitVector Reserved(getNumRegs());
   const X86FrameLowering *TFI = getFrameLowering(MF);
 
   // Set the floating point control register as reserved.
   Reserved.set(X86::FPCW);
 
+  // Set the SIMD floating point control register as reserved.
+  Reserved.set(X86::MXCSR);
+
   // Set the stack-pointer register and its aliases as reserved.
   for (MCSubRegIterator I(X86::RSP, this, /*IncludeSelf=*/true); I.isValid();
        ++I)
     Reserved.set(*I);
 
   // Set the Shadow Stack Pointer as reserved.
   Reserved.set(X86::SSP);
 

llvm/lib/Target/X86/X86RegisterInfo.td

 def ST7 : X86Reg<"st(7)", 7>, DwarfRegNum<[40, 19, 18]>;
 
 // Floating-point status word
 def FPSW : X86Reg<"fpsr", 0>;
 
 // Floating-point control word
 def FPCW : X86Reg<"fpcr", 0>;
 
+// SIMD Floating-point control register.
+// Note: We only model the current rounding modes and the IEEE masks.
+// IEEE flags, FTZ and DAZ are not modeled here.
+def MXCSR : X86Reg<"mxcsr", 0>;
+
 // Status flags register.
 //
 // Note that some flags that are commonly thought of as part of the status
 // flags register are modeled separately. Typically this is due to instructions
 // reading and updating those flags independently of all the others. We don't
 // want to create false dependencies between these instructions and so we use
 // a separate register to model them.
 def EFLAGS : X86Reg<"flags", 0>;