mirror of
https://github.com/RPCS3/llvm.git
synced 2024-12-27 22:55:15 +00:00
Remove some ugly now-redundant casts.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28864 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Chris Lattner
parent
a354849c27
commit
30da68acce
@ -2265,16 +2265,16 @@ def : Pat<(v4i32 (undef)), (IMPLICIT_DEF_VR128)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v2i64 (undef)), (IMPLICIT_DEF_VR128)>, Requires<[HasSSE2]>;
|
||||
|
||||
// 128-bit vector all zero's.
|
||||
def : Pat<(v16i8 immAllZerosV), (v16i8 (V_SET0_PI))>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v8i16 immAllZerosV), (v8i16 (V_SET0_PI))>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v4i32 immAllZerosV), (v4i32 (V_SET0_PI))>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v16i8 immAllZerosV), (V_SET0_PI)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v8i16 immAllZerosV), (V_SET0_PI)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v4i32 immAllZerosV), (V_SET0_PI)>, Requires<[HasSSE2]>;
|
||||
|
||||
// 128-bit vector all one's.
|
||||
def : Pat<(v16i8 immAllOnesV), (v16i8 (V_SETALLONES))>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v8i16 immAllOnesV), (v8i16 (V_SETALLONES))>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v4i32 immAllOnesV), (v4i32 (V_SETALLONES))>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v2i64 immAllOnesV), (v2i64 (V_SETALLONES))>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v4f32 immAllOnesV), (v4f32 (V_SETALLONES))>, Requires<[HasSSE1]>;
|
||||
def : Pat<(v16i8 immAllOnesV), (V_SETALLONES)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v8i16 immAllOnesV), (V_SETALLONES)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v4i32 immAllOnesV), (V_SETALLONES)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v2i64 immAllOnesV), (V_SETALLONES)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v4f32 immAllOnesV), (V_SETALLONES)>, Requires<[HasSSE1]>;
|
||||
|
||||
// Store 128-bit integer vector values.
|
||||
def : Pat<(store (v16i8 VR128:$src), addr:$dst),
|
||||
@ -2286,9 +2286,9 @@ def : Pat<(store (v4i32 VR128:$src), addr:$dst),
|
||||
|
||||
// Scalar to v8i16 / v16i8. The source may be a GR32, but only the lower 8 or
|
||||
// 16-bits matter.
|
||||
def : Pat<(v8i16 (X86s2vec GR32:$src)), (v8i16 (MOVDI2PDIrr GR32:$src))>,
|
||||
def : Pat<(v8i16 (X86s2vec GR32:$src)), (MOVDI2PDIrr GR32:$src)>,
|
||||
Requires<[HasSSE2]>;
|
||||
def : Pat<(v16i8 (X86s2vec GR32:$src)), (v16i8 (MOVDI2PDIrr GR32:$src))>,
|
||||
def : Pat<(v16i8 (X86s2vec GR32:$src)), (MOVDI2PDIrr GR32:$src)>,
|
||||
Requires<[HasSSE2]>;
|
||||
|
||||
// bit_convert
|
||||
@ -2358,155 +2358,155 @@ def : Pat<(v2f64 (bitconvert (v4f32 VR128:$src))), (v2f64 VR128:$src)>,
|
||||
let AddedComplexity = 20 in {
|
||||
def : Pat<(v8i16 (vector_shuffle immAllZerosV,
|
||||
(v8i16 (X86s2vec GR32:$src)), MOVL_shuffle_mask)),
|
||||
(v8i16 (MOVZDI2PDIrr GR32:$src))>, Requires<[HasSSE2]>;
|
||||
(MOVZDI2PDIrr GR32:$src)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v16i8 (vector_shuffle immAllZerosV,
|
||||
(v16i8 (X86s2vec GR32:$src)), MOVL_shuffle_mask)),
|
||||
(v16i8 (MOVZDI2PDIrr GR32:$src))>, Requires<[HasSSE2]>;
|
||||
(MOVZDI2PDIrr GR32:$src)>, Requires<[HasSSE2]>;
|
||||
// Zeroing a VR128 then do a MOVS{S|D} to the lower bits.
|
||||
def : Pat<(v2f64 (vector_shuffle immAllZerosV,
|
||||
(v2f64 (scalar_to_vector FR64:$src)), MOVL_shuffle_mask)),
|
||||
(v2f64 (MOVLSD2PDrr (V_SET0_PD), FR64:$src))>, Requires<[HasSSE2]>;
|
||||
(MOVLSD2PDrr (V_SET0_PD), FR64:$src)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v4f32 (vector_shuffle immAllZerosV,
|
||||
(v4f32 (scalar_to_vector FR32:$src)), MOVL_shuffle_mask)),
|
||||
(v4f32 (MOVLSS2PSrr (V_SET0_PS), FR32:$src))>, Requires<[HasSSE2]>;
|
||||
(MOVLSS2PSrr (V_SET0_PS), FR32:$src)>, Requires<[HasSSE2]>;
|
||||
}
|
||||
|
||||
// Splat v2f64 / v2i64
|
||||
let AddedComplexity = 10 in {
|
||||
def : Pat<(vector_shuffle (v2f64 VR128:$src), (undef), SSE_splat_v2_mask:$sm),
|
||||
(v2f64 (UNPCKLPDrr VR128:$src, VR128:$src))>, Requires<[HasSSE2]>;
|
||||
(UNPCKLPDrr VR128:$src, VR128:$src)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(vector_shuffle (v2i64 VR128:$src), (undef), SSE_splat_v2_mask:$sm),
|
||||
(v2i64 (PUNPCKLQDQrr VR128:$src, VR128:$src))>, Requires<[HasSSE2]>;
|
||||
(PUNPCKLQDQrr VR128:$src, VR128:$src)>, Requires<[HasSSE2]>;
|
||||
}
|
||||
|
||||
// Splat v4f32
|
||||
def : Pat<(vector_shuffle (v4f32 VR128:$src), (undef), SSE_splat_mask:$sm),
|
||||
(v4f32 (SHUFPSrri VR128:$src, VR128:$src, SSE_splat_mask:$sm))>,
|
||||
(SHUFPSrri VR128:$src, VR128:$src, SSE_splat_mask:$sm)>,
|
||||
Requires<[HasSSE1]>;
|
||||
|
||||
// Special unary SHUFPSrri case.
|
||||
// FIXME: when we want non two-address code, then we should use PSHUFD?
|
||||
def : Pat<(vector_shuffle (v4f32 VR128:$src1), (undef),
|
||||
SHUFP_unary_shuffle_mask:$sm),
|
||||
(v4f32 (SHUFPSrri VR128:$src1, VR128:$src1, SHUFP_unary_shuffle_mask:$sm))>,
|
||||
(SHUFPSrri VR128:$src1, VR128:$src1, SHUFP_unary_shuffle_mask:$sm)>,
|
||||
Requires<[HasSSE1]>;
|
||||
// Unary v4f32 shuffle with PSHUF* in order to fold a load.
|
||||
def : Pat<(vector_shuffle (loadv4f32 addr:$src1), (undef),
|
||||
SHUFP_unary_shuffle_mask:$sm),
|
||||
(v4f32 (PSHUFDmi addr:$src1, SHUFP_unary_shuffle_mask:$sm))>,
|
||||
(PSHUFDmi addr:$src1, SHUFP_unary_shuffle_mask:$sm)>,
|
||||
Requires<[HasSSE2]>;
|
||||
// Special binary v4i32 shuffle cases with SHUFPS.
|
||||
def : Pat<(vector_shuffle (v4i32 VR128:$src1), (v4i32 VR128:$src2),
|
||||
PSHUFD_binary_shuffle_mask:$sm),
|
||||
(v4i32 (SHUFPSrri VR128:$src1, VR128:$src2,
|
||||
PSHUFD_binary_shuffle_mask:$sm))>, Requires<[HasSSE2]>;
|
||||
(SHUFPSrri VR128:$src1, VR128:$src2, PSHUFD_binary_shuffle_mask:$sm)>,
|
||||
Requires<[HasSSE2]>;
|
||||
def : Pat<(vector_shuffle (v4i32 VR128:$src1),
|
||||
(bc_v4i32 (loadv2i64 addr:$src2)), PSHUFD_binary_shuffle_mask:$sm),
|
||||
(v4i32 (SHUFPSrmi VR128:$src1, addr:$src2,
|
||||
PSHUFD_binary_shuffle_mask:$sm))>, Requires<[HasSSE2]>;
|
||||
(SHUFPSrmi VR128:$src1, addr:$src2, PSHUFD_binary_shuffle_mask:$sm)>,
|
||||
Requires<[HasSSE2]>;
|
||||
|
||||
// vector_shuffle v1, <undef>, <0, 0, 1, 1, ...>
|
||||
let AddedComplexity = 10 in {
|
||||
def : Pat<(v4f32 (vector_shuffle VR128:$src, (undef),
|
||||
UNPCKL_v_undef_shuffle_mask)),
|
||||
(v4f32 (UNPCKLPSrr VR128:$src, VR128:$src))>, Requires<[HasSSE2]>;
|
||||
(UNPCKLPSrr VR128:$src, VR128:$src)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v16i8 (vector_shuffle VR128:$src, (undef),
|
||||
UNPCKL_v_undef_shuffle_mask)),
|
||||
(v16i8 (PUNPCKLBWrr VR128:$src, VR128:$src))>, Requires<[HasSSE2]>;
|
||||
(PUNPCKLBWrr VR128:$src, VR128:$src)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v8i16 (vector_shuffle VR128:$src, (undef),
|
||||
UNPCKL_v_undef_shuffle_mask)),
|
||||
(v8i16 (PUNPCKLWDrr VR128:$src, VR128:$src))>, Requires<[HasSSE2]>;
|
||||
(PUNPCKLWDrr VR128:$src, VR128:$src)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v4i32 (vector_shuffle VR128:$src, (undef),
|
||||
UNPCKL_v_undef_shuffle_mask)),
|
||||
(v4i32 (PUNPCKLDQrr VR128:$src, VR128:$src))>, Requires<[HasSSE1]>;
|
||||
(PUNPCKLDQrr VR128:$src, VR128:$src)>, Requires<[HasSSE1]>;
|
||||
}
|
||||
|
||||
let AddedComplexity = 20 in {
|
||||
// vector_shuffle v1, <undef> <1, 1, 3, 3>
|
||||
def : Pat<(v4i32 (vector_shuffle VR128:$src, (undef),
|
||||
MOVSHDUP_shuffle_mask)),
|
||||
(v4i32 (MOVSHDUPrr VR128:$src))>, Requires<[HasSSE3]>;
|
||||
(MOVSHDUPrr VR128:$src)>, Requires<[HasSSE3]>;
|
||||
def : Pat<(v4i32 (vector_shuffle (bc_v4i32 (loadv2i64 addr:$src)), (undef),
|
||||
MOVSHDUP_shuffle_mask)),
|
||||
(v4i32 (MOVSHDUPrm addr:$src))>, Requires<[HasSSE3]>;
|
||||
(MOVSHDUPrm addr:$src)>, Requires<[HasSSE3]>;
|
||||
|
||||
// vector_shuffle v1, <undef> <0, 0, 2, 2>
|
||||
def : Pat<(v4i32 (vector_shuffle VR128:$src, (undef),
|
||||
MOVSLDUP_shuffle_mask)),
|
||||
(v4i32 (MOVSLDUPrr VR128:$src))>, Requires<[HasSSE3]>;
|
||||
(MOVSLDUPrr VR128:$src)>, Requires<[HasSSE3]>;
|
||||
def : Pat<(v4i32 (vector_shuffle (bc_v4i32 (loadv2i64 addr:$src)), (undef),
|
||||
MOVSLDUP_shuffle_mask)),
|
||||
(v4i32 (MOVSLDUPrm addr:$src))>, Requires<[HasSSE3]>;
|
||||
(MOVSLDUPrm addr:$src)>, Requires<[HasSSE3]>;
|
||||
}
|
||||
|
||||
let AddedComplexity = 20 in {
|
||||
// vector_shuffle v1, v2 <0, 1, 4, 5> using MOVLHPS
|
||||
def : Pat<(v4i32 (vector_shuffle VR128:$src1, VR128:$src2,
|
||||
MOVHP_shuffle_mask)),
|
||||
(v4i32 (MOVLHPSrr VR128:$src1, VR128:$src2))>;
|
||||
(MOVLHPSrr VR128:$src1, VR128:$src2)>;
|
||||
|
||||
// vector_shuffle v1, v2 <6, 7, 2, 3> using MOVHLPS
|
||||
def : Pat<(v4i32 (vector_shuffle VR128:$src1, VR128:$src2,
|
||||
MOVHLPS_shuffle_mask)),
|
||||
(v4i32 (MOVHLPSrr VR128:$src1, VR128:$src2))>;
|
||||
(MOVHLPSrr VR128:$src1, VR128:$src2)>;
|
||||
|
||||
// vector_shuffle v1, undef <2, 3, ?, ?> using MOVHLPS
|
||||
def : Pat<(v4f32 (vector_shuffle VR128:$src1, (undef),
|
||||
UNPCKH_shuffle_mask)),
|
||||
(v4f32 (MOVHLPSrr VR128:$src1, VR128:$src1))>;
|
||||
(MOVHLPSrr VR128:$src1, VR128:$src1)>;
|
||||
def : Pat<(v4i32 (vector_shuffle VR128:$src1, (undef),
|
||||
UNPCKH_shuffle_mask)),
|
||||
(v4i32 (MOVHLPSrr VR128:$src1, VR128:$src1))>;
|
||||
(MOVHLPSrr VR128:$src1, VR128:$src1)>;
|
||||
|
||||
// vector_shuffle v1, (load v2) <4, 5, 2, 3> using MOVLPS
|
||||
// vector_shuffle v1, (load v2) <0, 1, 4, 5> using MOVHPS
|
||||
def : Pat<(v4f32 (vector_shuffle VR128:$src1, (loadv4f32 addr:$src2),
|
||||
MOVLP_shuffle_mask)),
|
||||
(v4f32 (MOVLPSrm VR128:$src1, addr:$src2))>, Requires<[HasSSE1]>;
|
||||
(MOVLPSrm VR128:$src1, addr:$src2)>, Requires<[HasSSE1]>;
|
||||
def : Pat<(v2f64 (vector_shuffle VR128:$src1, (loadv2f64 addr:$src2),
|
||||
MOVLP_shuffle_mask)),
|
||||
(v2f64 (MOVLPDrm VR128:$src1, addr:$src2))>, Requires<[HasSSE2]>;
|
||||
(MOVLPDrm VR128:$src1, addr:$src2)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v4f32 (vector_shuffle VR128:$src1, (loadv4f32 addr:$src2),
|
||||
MOVHP_shuffle_mask)),
|
||||
(v4f32 (MOVHPSrm VR128:$src1, addr:$src2))>, Requires<[HasSSE1]>;
|
||||
(MOVHPSrm VR128:$src1, addr:$src2)>, Requires<[HasSSE1]>;
|
||||
def : Pat<(v2f64 (vector_shuffle VR128:$src1, (loadv2f64 addr:$src2),
|
||||
MOVHP_shuffle_mask)),
|
||||
(v2f64 (MOVHPDrm VR128:$src1, addr:$src2))>, Requires<[HasSSE2]>;
|
||||
(MOVHPDrm VR128:$src1, addr:$src2)>, Requires<[HasSSE2]>;
|
||||
|
||||
def : Pat<(v4i32 (vector_shuffle VR128:$src1, (bc_v4i32 (loadv2i64 addr:$src2)),
|
||||
MOVLP_shuffle_mask)),
|
||||
(v4i32 (MOVLPSrm VR128:$src1, addr:$src2))>, Requires<[HasSSE2]>;
|
||||
(MOVLPSrm VR128:$src1, addr:$src2)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v2i64 (vector_shuffle VR128:$src1, (loadv2i64 addr:$src2),
|
||||
MOVLP_shuffle_mask)),
|
||||
(v2i64 (MOVLPDrm VR128:$src1, addr:$src2))>, Requires<[HasSSE2]>;
|
||||
(MOVLPDrm VR128:$src1, addr:$src2)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v4i32 (vector_shuffle VR128:$src1, (bc_v4i32 (loadv2i64 addr:$src2)),
|
||||
MOVHP_shuffle_mask)),
|
||||
(v4i32 (MOVHPSrm VR128:$src1, addr:$src2))>, Requires<[HasSSE1]>;
|
||||
(MOVHPSrm VR128:$src1, addr:$src2)>, Requires<[HasSSE1]>;
|
||||
def : Pat<(v2i64 (vector_shuffle VR128:$src1, (loadv2i64 addr:$src2),
|
||||
MOVLP_shuffle_mask)),
|
||||
(v2i64 (MOVLPDrm VR128:$src1, addr:$src2))>, Requires<[HasSSE2]>;
|
||||
(MOVLPDrm VR128:$src1, addr:$src2)>, Requires<[HasSSE2]>;
|
||||
|
||||
// Setting the lowest element in the vector.
|
||||
def : Pat<(v4i32 (vector_shuffle VR128:$src1, VR128:$src2,
|
||||
MOVL_shuffle_mask)),
|
||||
(v4i32 (MOVLPSrr VR128:$src1, VR128:$src2))>, Requires<[HasSSE2]>;
|
||||
(MOVLPSrr VR128:$src1, VR128:$src2)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v2i64 (vector_shuffle VR128:$src1, VR128:$src2,
|
||||
MOVL_shuffle_mask)),
|
||||
(v2i64 (MOVLPDrr VR128:$src1, VR128:$src2))>, Requires<[HasSSE2]>;
|
||||
(MOVLPDrr VR128:$src1, VR128:$src2)>, Requires<[HasSSE2]>;
|
||||
|
||||
// vector_shuffle v1, v2 <4, 5, 2, 3> using MOVLPDrr (movsd)
|
||||
def : Pat<(v4f32 (vector_shuffle VR128:$src1, VR128:$src2,
|
||||
MOVLP_shuffle_mask)),
|
||||
(v4f32 (MOVLPDrr VR128:$src1, VR128:$src2))>, Requires<[HasSSE2]>;
|
||||
(MOVLPDrr VR128:$src1, VR128:$src2)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v4i32 (vector_shuffle VR128:$src1, VR128:$src2,
|
||||
MOVLP_shuffle_mask)),
|
||||
(v4i32 (MOVLPDrr VR128:$src1, VR128:$src2))>, Requires<[HasSSE2]>;
|
||||
(MOVLPDrr VR128:$src1, VR128:$src2)>, Requires<[HasSSE2]>;
|
||||
|
||||
// Set lowest element and zero upper elements.
|
||||
def : Pat<(bc_v2i64 (vector_shuffle immAllZerosV,
|
||||
(v2f64 (scalar_to_vector (loadf64 addr:$src))),
|
||||
MOVL_shuffle_mask)),
|
||||
(v2i64 (MOVZQI2PQIrm addr:$src))>, Requires<[HasSSE2]>;
|
||||
(MOVZQI2PQIrm addr:$src)>, Requires<[HasSSE2]>;
|
||||
}
|
||||
|
||||
// FIXME: Temporary workaround since 2-wide shuffle is broken.
|
||||
@ -2550,20 +2550,20 @@ def : Pat<(int_x86_sse2_psrl_dq VR128:$src1, imm:$src2),
|
||||
// Some special case pandn patterns.
|
||||
def : Pat<(v2i64 (and (xor VR128:$src1, (bc_v2i64 (v4i32 immAllOnesV))),
|
||||
VR128:$src2)),
|
||||
(v2i64 (PANDNrr VR128:$src1, VR128:$src2))>, Requires<[HasSSE2]>;
|
||||
(PANDNrr VR128:$src1, VR128:$src2)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v2i64 (and (xor VR128:$src1, (bc_v2i64 (v8i16 immAllOnesV))),
|
||||
VR128:$src2)),
|
||||
(v2i64 (PANDNrr VR128:$src1, VR128:$src2))>, Requires<[HasSSE2]>;
|
||||
(PANDNrr VR128:$src1, VR128:$src2)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v2i64 (and (xor VR128:$src1, (bc_v2i64 (v16i8 immAllOnesV))),
|
||||
VR128:$src2)),
|
||||
(v2i64 (PANDNrr VR128:$src1, VR128:$src2))>, Requires<[HasSSE2]>;
|
||||
(PANDNrr VR128:$src1, VR128:$src2)>, Requires<[HasSSE2]>;
|
||||
|
||||
def : Pat<(v2i64 (and (xor VR128:$src1, (bc_v2i64 (v4i32 immAllOnesV))),
|
||||
(load addr:$src2))),
|
||||
(v2i64 (PANDNrm VR128:$src1, addr:$src2))>, Requires<[HasSSE2]>;
|
||||
(PANDNrm VR128:$src1, addr:$src2)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v2i64 (and (xor VR128:$src1, (bc_v2i64 (v8i16 immAllOnesV))),
|
||||
(load addr:$src2))),
|
||||
(v2i64 (PANDNrm VR128:$src1, addr:$src2))>, Requires<[HasSSE2]>;
|
||||
(PANDNrm VR128:$src1, addr:$src2)>, Requires<[HasSSE2]>;
|
||||
def : Pat<(v2i64 (and (xor VR128:$src1, (bc_v2i64 (v16i8 immAllOnesV))),
|
||||
(load addr:$src2))),
|
||||
(v2i64 (PANDNrm VR128:$src1, addr:$src2))>, Requires<[HasSSE2]>;
|
||||
(PANDNrm VR128:$src1, addr:$src2)>, Requires<[HasSSE2]>;
|
||||
|
||||
Loading…
Reference in New Issue
Block a user