summaryrefslogtreecommitdiffstats
path: root/llvm/lib/Target/X86/X86InstrAVX512.td
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/lib/Target/X86/X86InstrAVX512.td')
-rw-r--r--llvm/lib/Target/X86/X86InstrAVX512.td68
1 files changed, 24 insertions, 44 deletions
diff --git a/llvm/lib/Target/X86/X86InstrAVX512.td b/llvm/lib/Target/X86/X86InstrAVX512.td
index 29e18dd931f..84d8a2bbc80 100644
--- a/llvm/lib/Target/X86/X86InstrAVX512.td
+++ b/llvm/lib/Target/X86/X86InstrAVX512.td
@@ -5360,50 +5360,31 @@ let Defs = [EFLAGS], Predicates = [HasAVX512] in {
}
/// avx512_fp14_s rcp14ss, rcp14sd, rsqrt14ss, rsqrt14sd
-multiclass avx512_fp14_s<bits<8> opc, string OpcodeStr, RegisterClass RC,
- X86MemOperand x86memop> {
- let hasSideEffects = 0 in {
- def rr : AVX5128I<opc, MRMSrcReg, (outs RC:$dst),
- (ins RC:$src1, RC:$src2),
- !strconcat(OpcodeStr,
- "\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>, EVEX_4V;
+multiclass avx512_fp14_s<bits<8> opc, string OpcodeStr, SDNode OpNode,
+ X86VectorVTInfo _> {
+ let hasSideEffects = 0, AddedComplexity = 20 , Predicates = [HasAVX512] in {
+ defm rr : AVX512_maskable_scalar<opc, MRMSrcReg, _, (outs _.RC:$dst),
+ (ins _.RC:$src1, _.RC:$src2), OpcodeStr,
+ "$src2, $src1", "$src1, $src2",
+ (OpNode (_.VT _.RC:$src1), (_.VT _.RC:$src2))>, EVEX_4V;
let mayLoad = 1 in {
- def rm : AVX5128I<opc, MRMSrcMem, (outs RC:$dst),
- (ins RC:$src1, x86memop:$src2),
- !strconcat(OpcodeStr,
- "\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>, EVEX_4V;
+ defm rm : AVX512_maskable_scalar<opc, MRMSrcMem, _, (outs _.RC:$dst),
+ (ins _.RC:$src1, _.MemOp:$src2), OpcodeStr,
+ "$src2, $src1", "$src1, $src2",
+ (OpNode (_.VT _.RC:$src1),
+ (_.VT (scalar_to_vector (_.ScalarLdFrag addr:$src2))))>, EVEX_4V;
}
}
}
-defm VRCP14SS : avx512_fp14_s<0x4D, "vrcp14ss", FR32X, f32mem>,
- EVEX_CD8<32, CD8VT1>;
-defm VRCP14SD : avx512_fp14_s<0x4D, "vrcp14sd", FR64X, f64mem>,
- VEX_W, EVEX_CD8<64, CD8VT1>;
-defm VRSQRT14SS : avx512_fp14_s<0x4F, "vrsqrt14ss", FR32X, f32mem>,
- EVEX_CD8<32, CD8VT1>;
-defm VRSQRT14SD : avx512_fp14_s<0x4F, "vrsqrt14sd", FR64X, f64mem>,
- VEX_W, EVEX_CD8<64, CD8VT1>;
-
-def : Pat <(v4f32 (int_x86_avx512_rcp14_ss (v4f32 VR128X:$src1),
- (v4f32 VR128X:$src2), (bc_v4f32 (v4i32 immAllZerosV)), (i8 -1))),
- (COPY_TO_REGCLASS (VRCP14SSrr (COPY_TO_REGCLASS VR128X:$src1, FR32X),
- (COPY_TO_REGCLASS VR128X:$src2, FR32X)), VR128X)>;
-
-def : Pat <(v2f64 (int_x86_avx512_rcp14_sd (v2f64 VR128X:$src1),
- (v2f64 VR128X:$src2), (bc_v2f64 (v4i32 immAllZerosV)), (i8 -1))),
- (COPY_TO_REGCLASS (VRCP14SDrr (COPY_TO_REGCLASS VR128X:$src1, FR64X),
- (COPY_TO_REGCLASS VR128X:$src2, FR64X)), VR128X)>;
-
-def : Pat <(v4f32 (int_x86_avx512_rsqrt14_ss (v4f32 VR128X:$src1),
- (v4f32 VR128X:$src2), (bc_v4f32 (v4i32 immAllZerosV)), (i8 -1))),
- (COPY_TO_REGCLASS (VRSQRT14SSrr (COPY_TO_REGCLASS VR128X:$src1, FR32X),
- (COPY_TO_REGCLASS VR128X:$src2, FR32X)), VR128X)>;
-
-def : Pat <(v2f64 (int_x86_avx512_rsqrt14_sd (v2f64 VR128X:$src1),
- (v2f64 VR128X:$src2), (bc_v2f64 (v4i32 immAllZerosV)), (i8 -1))),
- (COPY_TO_REGCLASS (VRSQRT14SDrr (COPY_TO_REGCLASS VR128X:$src1, FR64X),
- (COPY_TO_REGCLASS VR128X:$src2, FR64X)), VR128X)>;
+defm VRCP14SS : avx512_fp14_s<0x4D, "vrcp14ss", X86frcp14s, f32x_info>,
+ EVEX_CD8<32, CD8VT1>, T8PD;
+defm VRCP14SD : avx512_fp14_s<0x4D, "vrcp14sd", X86frcp14s, f64x_info>,
+ VEX_W, EVEX_CD8<64, CD8VT1>, T8PD;
+defm VRSQRT14SS : avx512_fp14_s<0x4F, "vrsqrt14ss", X86frsqrt14s, f32x_info>,
+ EVEX_CD8<32, CD8VT1>, T8PD;
+defm VRSQRT14SD : avx512_fp14_s<0x4F, "vrsqrt14sd", X86frsqrt14s, f64x_info>,
+ VEX_W, EVEX_CD8<64, CD8VT1>, T8PD;
/// avx512_fp14_p rcp14ps, rcp14pd, rsqrt14ps, rsqrt14pd
multiclass avx512_fp14_p<bits<8> opc, string OpcodeStr, SDNode OpNode,
@@ -5685,15 +5666,14 @@ defm VSQRT : avx512_sqrt_scalar_all<0x51, "vsqrt">, VEX_LIG;
let Predicates = [HasAVX512] in {
def : Pat<(f32 (X86frsqrt FR32X:$src)),
- (VRSQRT14SSrr (f32 (IMPLICIT_DEF)), FR32X:$src)>;
+ (COPY_TO_REGCLASS (VRSQRT14SSrr (v4f32 (IMPLICIT_DEF)), (COPY_TO_REGCLASS FR32X:$src, VR128X)), VR128X)>;
def : Pat<(f32 (X86frsqrt (load addr:$src))),
- (VRSQRT14SSrm (f32 (IMPLICIT_DEF)), addr:$src)>,
+ (COPY_TO_REGCLASS (VRSQRT14SSrm (v4f32 (IMPLICIT_DEF)), addr:$src), VR128X)>,
Requires<[OptForSize]>;
-
def : Pat<(f32 (X86frcp FR32X:$src)),
- (VRCP14SSrr (f32 (IMPLICIT_DEF)), FR32X:$src)>;
+ (COPY_TO_REGCLASS (VRCP14SSrr (v4f32 (IMPLICIT_DEF)), (COPY_TO_REGCLASS FR32X:$src, VR128X)), VR128X )>;
def : Pat<(f32 (X86frcp (load addr:$src))),
- (VRCP14SSrm (f32 (IMPLICIT_DEF)), addr:$src)>,
+ (COPY_TO_REGCLASS (VRCP14SSrm (v4f32 (IMPLICIT_DEF)), addr:$src), VR128X)>,
Requires<[OptForSize]>;
}
OpenPOWER on IntegriCloud