aboutsummaryrefslogtreecommitdiff
path: root/lib/Target/ARM/ARMInstrVFP.td
diff options
context:
space:
mode:
Diffstat (limited to 'lib/Target/ARM/ARMInstrVFP.td')
-rw-r--r--lib/Target/ARM/ARMInstrVFP.td55
1 files changed, 38 insertions, 17 deletions
diff --git a/lib/Target/ARM/ARMInstrVFP.td b/lib/Target/ARM/ARMInstrVFP.td
index e83f8c850632..050cd1a445ad 100644
--- a/lib/Target/ARM/ARMInstrVFP.td
+++ b/lib/Target/ARM/ARMInstrVFP.td
@@ -20,7 +20,6 @@ def arm_cmpfp : SDNode<"ARMISD::CMPFP", SDT_ARMCmp, [SDNPOutGlue]>;
def arm_cmpfp0 : SDNode<"ARMISD::CMPFPw0", SDT_CMPFP0, [SDNPOutGlue]>;
def arm_fmdrr : SDNode<"ARMISD::VMOVDRR", SDT_VMOVDRR>;
-
//===----------------------------------------------------------------------===//
// Operand Definitions.
//
@@ -93,7 +92,7 @@ def VLDRD : ADI5<0b1101, 0b01, (outs DPR:$Dd), (ins addrmode5:$addr),
def VLDRS : ASI5<0b1101, 0b01, (outs SPR:$Sd), (ins addrmode5:$addr),
IIC_fpLoad32, "vldr", "\t$Sd, $addr",
- [(set SPR:$Sd, (load addrmode5:$addr))]> {
+ [(set SPR:$Sd, (alignedload32 addrmode5:$addr))]> {
// Some single precision VFP instructions may be executed on both NEON and VFP
// pipelines.
let D = VFPNeonDomain;
@@ -107,7 +106,7 @@ def VSTRD : ADI5<0b1101, 0b00, (outs), (ins DPR:$Dd, addrmode5:$addr),
def VSTRS : ASI5<0b1101, 0b00, (outs), (ins SPR:$Sd, addrmode5:$addr),
IIC_fpStore32, "vstr", "\t$Sd, $addr",
- [(store SPR:$Sd, addrmode5:$addr)]> {
+ [(alignedstore32 SPR:$Sd, addrmode5:$addr)]> {
// Some single precision VFP instructions may be executed on both NEON and VFP
// pipelines.
let D = VFPNeonDomain;
@@ -393,8 +392,8 @@ multiclass vmaxmin_inst<string op, bit opc, SDNode SD> {
}
}
-defm VMAXNM : vmaxmin_inst<"vmaxnm", 0, ARMvmaxnm>;
-defm VMINNM : vmaxmin_inst<"vminnm", 1, ARMvminnm>;
+defm VMAXNM : vmaxmin_inst<"vmaxnm", 0, fmaxnum>;
+defm VMINNM : vmaxmin_inst<"vminnm", 1, fminnum>;
// Match reassociated forms only if not sign dependent rounding.
def : Pat<(fmul (fneg DPR:$a), (f64 DPR:$b)),
@@ -541,19 +540,23 @@ def VCVTSD : VFPAI<(outs SPR:$Sd), (ins DPR:$Dm), VFPUnaryFrm,
// FIXME: Verify encoding after integrated assembler is working.
def VCVTBHS: ASuI<0b11101, 0b11, 0b0010, 0b01, 0, (outs SPR:$Sd), (ins SPR:$Sm),
/* FIXME */ IIC_fpCVTSH, "vcvtb", ".f32.f16\t$Sd, $Sm",
- [/* For disassembly only; pattern left blank */]>;
+ [/* For disassembly only; pattern left blank */]>,
+ Requires<[HasFP16]>;
def VCVTBSH: ASuI<0b11101, 0b11, 0b0011, 0b01, 0, (outs SPR:$Sd), (ins SPR:$Sm),
/* FIXME */ IIC_fpCVTHS, "vcvtb", ".f16.f32\t$Sd, $Sm",
- [/* For disassembly only; pattern left blank */]>;
+ [/* For disassembly only; pattern left blank */]>,
+ Requires<[HasFP16]>;
def VCVTTHS: ASuI<0b11101, 0b11, 0b0010, 0b11, 0, (outs SPR:$Sd), (ins SPR:$Sm),
/* FIXME */ IIC_fpCVTSH, "vcvtt", ".f32.f16\t$Sd, $Sm",
- [/* For disassembly only; pattern left blank */]>;
+ [/* For disassembly only; pattern left blank */]>,
+ Requires<[HasFP16]>;
def VCVTTSH: ASuI<0b11101, 0b11, 0b0011, 0b11, 0, (outs SPR:$Sd), (ins SPR:$Sm),
/* FIXME */ IIC_fpCVTHS, "vcvtt", ".f16.f32\t$Sd, $Sm",
- [/* For disassembly only; pattern left blank */]>;
+ [/* For disassembly only; pattern left blank */]>,
+ Requires<[HasFP16]>;
def VCVTBHD : ADuI<0b11101, 0b11, 0b0010, 0b01, 0,
(outs DPR:$Dd), (ins SPR:$Sm),
@@ -922,6 +925,22 @@ def VMOVDRR : AVConv5I<0b11000100, 0b1011,
let isRegSequence = 1;
}
+// Hoist an fabs or a fneg of a value coming from integer registers
+// and do the fabs/fneg on the integer value. This is never a lose
+// and could enable the conversion to float to be removed completely.
+def : Pat<(fabs (arm_fmdrr GPR:$Rl, GPR:$Rh)),
+ (VMOVDRR GPR:$Rl, (BFC GPR:$Rh, (i32 0x7FFFFFFF)))>,
+ Requires<[IsARM]>;
+def : Pat<(fabs (arm_fmdrr GPR:$Rl, GPR:$Rh)),
+ (VMOVDRR GPR:$Rl, (t2BFC GPR:$Rh, (i32 0x7FFFFFFF)))>,
+ Requires<[IsThumb2]>;
+def : Pat<(fneg (arm_fmdrr GPR:$Rl, GPR:$Rh)),
+ (VMOVDRR GPR:$Rl, (EORri GPR:$Rh, (i32 0x80000000)))>,
+ Requires<[IsARM]>;
+def : Pat<(fneg (arm_fmdrr GPR:$Rl, GPR:$Rh)),
+ (VMOVDRR GPR:$Rl, (t2EORri GPR:$Rh, (i32 0x80000000)))>,
+ Requires<[IsThumb2]>;
+
let hasSideEffects = 0 in
def VMOVSRR : AVConv5I<0b11000100, 0b1010,
(outs SPR:$dst1, SPR:$dst2), (ins GPR:$src1, GPR:$src2),
@@ -1003,7 +1022,7 @@ let Predicates=[HasVFP2, HasDPVFP] in {
def : VFPPat<(f64 (sint_to_fp GPR:$a)),
(VSITOD (COPY_TO_REGCLASS GPR:$a, SPR))>;
- def : VFPPat<(f64 (sint_to_fp (i32 (load addrmode5:$a)))),
+ def : VFPPat<(f64 (sint_to_fp (i32 (alignedload32 addrmode5:$a)))),
(VSITOD (VLDRS addrmode5:$a))>;
}
@@ -1021,7 +1040,7 @@ def VSITOS : AVConv1InSs_Encode<0b11101, 0b11, 0b1000, 0b1010,
def : VFPNoNEONPat<(f32 (sint_to_fp GPR:$a)),
(VSITOS (COPY_TO_REGCLASS GPR:$a, SPR))>;
-def : VFPNoNEONPat<(f32 (sint_to_fp (i32 (load addrmode5:$a)))),
+def : VFPNoNEONPat<(f32 (sint_to_fp (i32 (alignedload32 addrmode5:$a)))),
(VSITOS (VLDRS addrmode5:$a))>;
def VUITOD : AVConv1IDs_Encode<0b11101, 0b11, 0b1000, 0b1011,
@@ -1035,7 +1054,7 @@ let Predicates=[HasVFP2, HasDPVFP] in {
def : VFPPat<(f64 (uint_to_fp GPR:$a)),
(VUITOD (COPY_TO_REGCLASS GPR:$a, SPR))>;
- def : VFPPat<(f64 (uint_to_fp (i32 (load addrmode5:$a)))),
+ def : VFPPat<(f64 (uint_to_fp (i32 (alignedload32 addrmode5:$a)))),
(VUITOD (VLDRS addrmode5:$a))>;
}
@@ -1053,7 +1072,7 @@ def VUITOS : AVConv1InSs_Encode<0b11101, 0b11, 0b1000, 0b1010,
def : VFPNoNEONPat<(f32 (uint_to_fp GPR:$a)),
(VUITOS (COPY_TO_REGCLASS GPR:$a, SPR))>;
-def : VFPNoNEONPat<(f32 (uint_to_fp (i32 (load addrmode5:$a)))),
+def : VFPNoNEONPat<(f32 (uint_to_fp (i32 (alignedload32 addrmode5:$a)))),
(VUITOS (VLDRS addrmode5:$a))>;
// FP -> Int:
@@ -1106,7 +1125,7 @@ let Predicates=[HasVFP2, HasDPVFP] in {
def : VFPPat<(i32 (fp_to_sint (f64 DPR:$a))),
(COPY_TO_REGCLASS (VTOSIZD DPR:$a), GPR)>;
- def : VFPPat<(store (i32 (fp_to_sint (f64 DPR:$a))), addrmode5:$ptr),
+ def : VFPPat<(alignedstore32 (i32 (fp_to_sint (f64 DPR:$a))), addrmode5:$ptr),
(VSTRS (VTOSIZD DPR:$a), addrmode5:$ptr)>;
}
@@ -1124,7 +1143,8 @@ def VTOSIZS : AVConv1InsS_Encode<0b11101, 0b11, 0b1101, 0b1010,
def : VFPNoNEONPat<(i32 (fp_to_sint SPR:$a)),
(COPY_TO_REGCLASS (VTOSIZS SPR:$a), GPR)>;
-def : VFPNoNEONPat<(store (i32 (fp_to_sint (f32 SPR:$a))), addrmode5:$ptr),
+def : VFPNoNEONPat<(alignedstore32 (i32 (fp_to_sint (f32 SPR:$a))),
+ addrmode5:$ptr),
(VSTRS (VTOSIZS SPR:$a), addrmode5:$ptr)>;
def VTOUIZD : AVConv1IsD_Encode<0b11101, 0b11, 0b1100, 0b1011,
@@ -1138,7 +1158,7 @@ let Predicates=[HasVFP2, HasDPVFP] in {
def : VFPPat<(i32 (fp_to_uint (f64 DPR:$a))),
(COPY_TO_REGCLASS (VTOUIZD DPR:$a), GPR)>;
- def : VFPPat<(store (i32 (fp_to_uint (f64 DPR:$a))), addrmode5:$ptr),
+ def : VFPPat<(alignedstore32 (i32 (fp_to_uint (f64 DPR:$a))), addrmode5:$ptr),
(VSTRS (VTOUIZD DPR:$a), addrmode5:$ptr)>;
}
@@ -1156,7 +1176,8 @@ def VTOUIZS : AVConv1InsS_Encode<0b11101, 0b11, 0b1100, 0b1010,
def : VFPNoNEONPat<(i32 (fp_to_uint SPR:$a)),
(COPY_TO_REGCLASS (VTOUIZS SPR:$a), GPR)>;
-def : VFPNoNEONPat<(store (i32 (fp_to_uint (f32 SPR:$a))), addrmode5:$ptr),
+def : VFPNoNEONPat<(alignedstore32 (i32 (fp_to_uint (f32 SPR:$a))),
+ addrmode5:$ptr),
(VSTRS (VTOUIZS SPR:$a), addrmode5:$ptr)>;
// And the Z bit '0' variants, i.e. use the rounding mode specified by FPSCR.
generated by cgit v1.2.3 (git 2.25.1) at 2024-11-08 02:34:05 +0000