[X86] Remove GCCBuiltins from 512-bit cvt(u)qqtops, cvt(u)qqtopd, and cvt(u)dqtops intrinsics. Add new variadic uitofp/sitofp with rounding mode intrinsics.

[craig.topper]

Summary

See clang patch D56998 for a full description.

[RKSimon]

include/llvm/IR/IntrinsicsX86.td

2691

AFAICT these are only ever going to be used by avx512 targets - add avx512 prefix?

[craig.topper]

Rename the intrinsics and add autoupgrade support

[craig.topper]

Ping

[RKSimon]

LGTM

include/llvm/IR/IntrinsicsX86.td

def int_x86_avx512_packuswb_512 : GCCBuiltin<"__builtin_ia32_packuswb512">,
                          [IntrNoMem]>;
   def int_x86_avx512_packusdw_512 : GCCBuiltin<"__builtin_ia32_packusdw512">,
               Intrinsic<[llvm_v32i16_ty], [llvm_v16i32_ty, llvm_v16i32_ty],
                          [IntrNoMem]>;
 }
 
 // Vector convert
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
-  def int_x86_avx512_mask_cvtdq2ps_512 :
-        GCCBuiltin<"__builtin_ia32_cvtdq2ps512_mask">,
+  def int_x86_sitofp_round :
+          Intrinsic<[llvm_anyfloat_ty], [llvm_anyint_ty, llvm_i32_ty],
+          [IntrNoMem]>;
+
+  def int_x86_uitofp_round :
+          Intrinsic<[llvm_anyfloat_ty], [llvm_anyint_ty, llvm_i32_ty],
+          [IntrNoMem]>;
+
+  def int_x86_avx512_mask_cvtdq2ps_512 : // FIXME remove
           Intrinsic<[llvm_v16f32_ty],
           [llvm_v16i32_ty, llvm_v16f32_ty,  llvm_i16_ty,  llvm_i32_ty],
           [IntrNoMem]>;
 
   def int_x86_avx512_mask_cvtpd2dq_128 :
         GCCBuiltin<"__builtin_ia32_cvtpd2dq128_mask">,
           Intrinsic<[llvm_v4i32_ty],
           [llvm_v2f64_ty, llvm_v4i32_ty,  llvm_i8_ty],
def int_x86_avx512_mask_cvtps2uqq_256 :
           [IntrNoMem]>;
 
   def int_x86_avx512_mask_cvtps2uqq_512 :
         GCCBuiltin<"__builtin_ia32_cvtps2uqq512_mask">,
           Intrinsic<[llvm_v8i64_ty],
           [llvm_v8f32_ty, llvm_v8i64_ty,  llvm_i8_ty,  llvm_i32_ty],
           [IntrNoMem]>;
 
-  def int_x86_avx512_mask_cvtqq2pd_512 :
-        GCCBuiltin<"__builtin_ia32_cvtqq2pd512_mask">,
+  def int_x86_avx512_mask_cvtqq2pd_512 : // FIXME remove
           Intrinsic<[llvm_v8f64_ty],
           [llvm_v8i64_ty, llvm_v8f64_ty,  llvm_i8_ty,  llvm_i32_ty],
           [IntrNoMem]>;
 
   def int_x86_avx512_mask_cvtqq2ps_128 :
         GCCBuiltin<"__builtin_ia32_cvtqq2ps128_mask">,
           Intrinsic<[llvm_v4f32_ty],
           [llvm_v2i64_ty, llvm_v4f32_ty,  llvm_i8_ty],
           [IntrNoMem]>;
 
-  def int_x86_avx512_mask_cvtqq2ps_256 :
-        GCCBuiltin<"__builtin_ia32_cvtqq2ps256_mask">,
+  def int_x86_avx512_mask_cvtqq2ps_256 : // FIXME: remove
           Intrinsic<[llvm_v4f32_ty],
           [llvm_v4i64_ty, llvm_v4f32_ty,  llvm_i8_ty],
           [IntrNoMem]>;
 
-  def int_x86_avx512_mask_cvtqq2ps_512 :
-        GCCBuiltin<"__builtin_ia32_cvtqq2ps512_mask">,
+  def int_x86_avx512_mask_cvtqq2ps_512 : // FIXME: remove
           Intrinsic<[llvm_v8f32_ty],
           [llvm_v8i64_ty, llvm_v8f32_ty,  llvm_i8_ty,  llvm_i32_ty],
           [IntrNoMem]>;
 
   def int_x86_avx512_mask_cvttpd2dq_128 :
         GCCBuiltin<"__builtin_ia32_cvttpd2dq128_mask">,
           Intrinsic<[llvm_v4i32_ty],
           [llvm_v2f64_ty, llvm_v4i32_ty,  llvm_i8_ty],
def int_x86_avx512_mask_cvttps2uqq_256 :
           [IntrNoMem]>;
 
   def int_x86_avx512_mask_cvttps2uqq_512 :
         GCCBuiltin<"__builtin_ia32_cvttps2uqq512_mask">,
           Intrinsic<[llvm_v8i64_ty],
           [llvm_v8f32_ty, llvm_v8i64_ty,  llvm_i8_ty,  llvm_i32_ty],
           [IntrNoMem]>;
 
-  def int_x86_avx512_mask_cvtudq2ps_512 :
-        GCCBuiltin<"__builtin_ia32_cvtudq2ps512_mask">,
+  def int_x86_avx512_mask_cvtudq2ps_512 : // FIXME remove
           Intrinsic<[llvm_v16f32_ty],
           [llvm_v16i32_ty, llvm_v16f32_ty,  llvm_i16_ty,  llvm_i32_ty],
           [IntrNoMem]>;
 
-  def int_x86_avx512_mask_cvtuqq2pd_512 :
-        GCCBuiltin<"__builtin_ia32_cvtuqq2pd512_mask">,
+  def int_x86_avx512_mask_cvtuqq2pd_512 : // FIXME remove
           Intrinsic<[llvm_v8f64_ty],
           [llvm_v8i64_ty, llvm_v8f64_ty,  llvm_i8_ty,  llvm_i32_ty],
           [IntrNoMem]>;
 
   def int_x86_avx512_mask_cvtuqq2ps_128 :
         GCCBuiltin<"__builtin_ia32_cvtuqq2ps128_mask">,
           Intrinsic<[llvm_v4f32_ty],
           [llvm_v2i64_ty, llvm_v4f32_ty,  llvm_i8_ty],
           [IntrNoMem]>;
 
-  def int_x86_avx512_mask_cvtuqq2ps_256 :
-        GCCBuiltin<"__builtin_ia32_cvtuqq2ps256_mask">,
+  def int_x86_avx512_mask_cvtuqq2ps_256 : // FIXME remove
           Intrinsic<[llvm_v4f32_ty],
           [llvm_v4i64_ty, llvm_v4f32_ty,  llvm_i8_ty],
           [IntrNoMem]>;
 
-  def int_x86_avx512_mask_cvtuqq2ps_512 :
-        GCCBuiltin<"__builtin_ia32_cvtuqq2ps512_mask">,
+  def int_x86_avx512_mask_cvtuqq2ps_512 : // FIXME remove
           Intrinsic<[llvm_v8f32_ty],
           [llvm_v8i64_ty, llvm_v8f32_ty,  llvm_i8_ty,  llvm_i32_ty],
           [IntrNoMem]>;
 
   def int_x86_avx512_mask_rndscale_pd_128 : GCCBuiltin<"__builtin_ia32_rndscalepd_128_mask">,
         Intrinsic<[llvm_v2f64_ty], [llvm_v2f64_ty, llvm_i32_ty,
                                      llvm_v2f64_ty,  llvm_i8_ty], [IntrNoMem]>;
   def int_x86_avx512_mask_rndscale_pd_256 : GCCBuiltin<"__builtin_ia32_rndscalepd_256_mask">,

lib/Target/X86/X86IntrinsicsInfo.h

static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx512_vpmadd52l_uq_512 , IFMA_OP, X86ISD::VPMADD52L, 0),
   X86_INTRINSIC_DATA(avx512_vpshufbitqmb_128, INTR_TYPE_2OP, X86ISD::VPSHUFBITQMB, 0),
   X86_INTRINSIC_DATA(avx512_vpshufbitqmb_256, INTR_TYPE_2OP, X86ISD::VPSHUFBITQMB, 0),
   X86_INTRINSIC_DATA(avx512_vpshufbitqmb_512, INTR_TYPE_2OP, X86ISD::VPSHUFBITQMB, 0),
   X86_INTRINSIC_DATA(bmi_bextr_32,         INTR_TYPE_2OP, X86ISD::BEXTR, 0),
   X86_INTRINSIC_DATA(bmi_bextr_64,         INTR_TYPE_2OP, X86ISD::BEXTR, 0),
   X86_INTRINSIC_DATA(bmi_bzhi_32,          INTR_TYPE_2OP, X86ISD::BZHI, 0),
   X86_INTRINSIC_DATA(bmi_bzhi_64,          INTR_TYPE_2OP, X86ISD::BZHI, 0),
+  X86_INTRINSIC_DATA(sitofp_round,      INTR_TYPE_1OP, ISD::SINT_TO_FP, X86ISD::SINT_TO_FP_RND),
   X86_INTRINSIC_DATA(sse_cmp_ps,        INTR_TYPE_3OP, X86ISD::CMPP, 0),
   X86_INTRINSIC_DATA(sse_comieq_ss,     COMI, X86ISD::COMI, ISD::SETEQ),
   X86_INTRINSIC_DATA(sse_comige_ss,     COMI, X86ISD::COMI, ISD::SETGE),
   X86_INTRINSIC_DATA(sse_comigt_ss,     COMI, X86ISD::COMI, ISD::SETGT),
   X86_INTRINSIC_DATA(sse_comile_ss,     COMI, X86ISD::COMI, ISD::SETLE),
   X86_INTRINSIC_DATA(sse_comilt_ss,     COMI, X86ISD::COMI, ISD::SETLT),
   X86_INTRINSIC_DATA(sse_comineq_ss,    COMI, X86ISD::COMI, ISD::SETNE),
   X86_INTRINSIC_DATA(sse_cvtss2si,      INTR_TYPE_1OP, X86ISD::CVTS2SI, 0),
static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(ssse3_phsub_w_128, INTR_TYPE_2OP, X86ISD::HSUB, 0),
   X86_INTRINSIC_DATA(ssse3_pmadd_ub_sw_128, INTR_TYPE_2OP, X86ISD::VPMADDUBSW, 0),
   X86_INTRINSIC_DATA(ssse3_pmul_hr_sw_128, INTR_TYPE_2OP, X86ISD::MULHRS, 0),
   X86_INTRINSIC_DATA(ssse3_pshuf_b_128, INTR_TYPE_2OP, X86ISD::PSHUFB, 0),
   X86_INTRINSIC_DATA(subborrow_32,      ADX, X86ISD::SBB, X86ISD::SUB),
   X86_INTRINSIC_DATA(subborrow_64,      ADX, X86ISD::SBB, X86ISD::SUB),
   X86_INTRINSIC_DATA(tbm_bextri_u32,    INTR_TYPE_2OP, X86ISD::BEXTR, 0),
   X86_INTRINSIC_DATA(tbm_bextri_u64,    INTR_TYPE_2OP, X86ISD::BEXTR, 0),
+  X86_INTRINSIC_DATA(uitofp_round,      INTR_TYPE_1OP, ISD::UINT_TO_FP, X86ISD::UINT_TO_FP_RND),
   X86_INTRINSIC_DATA(vcvtph2ps_128,     INTR_TYPE_1OP, X86ISD::CVTPH2PS, 0),
   X86_INTRINSIC_DATA(vcvtph2ps_256,     INTR_TYPE_1OP, X86ISD::CVTPH2PS, 0),
   X86_INTRINSIC_DATA(vcvtps2ph_128,     INTR_TYPE_2OP, X86ISD::CVTPS2PH, 0),
   X86_INTRINSIC_DATA(vcvtps2ph_256,     INTR_TYPE_2OP, X86ISD::CVTPS2PH, 0),
 
   X86_INTRINSIC_DATA(vgf2p8affineinvqb_128, INTR_TYPE_3OP,
                      X86ISD::GF2P8AFFINEINVQB, 0),
   X86_INTRINSIC_DATA(vgf2p8affineinvqb_256, INTR_TYPE_3OP,