mikros/rust
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Auto merge of #115580 - eduardosm:stdarch-intrinsics, r=davidtwco,bjorn3

Browse Source 
Update stdarch submodule and remove special handling in cranelift codegen for some AVX and SSE2 LLVM intrinsics

https://github.com/rust-lang/stdarch/pull/1463 reimplemented some x86 intrinsics to avoid using some x86-specific LLVM intrinsics:

* Store unaligned (`_mm*_storeu_*`) use `<*mut _>::write_unaligned` instead of `llvm.x86.*.storeu.*`.
* Shift by immediate (`_mm*_s{ll,rl,ra}i_epi*`) use `if` (srl, sll) or `min` (sra) to simulate the behaviour when the RHS is out of range. RHS is constant, so the `if`/`min` will be optimized away.

This PR updates the stdarch submodule to pull these changes and removes special handling for those LLVM intrinsics from cranelift codegen. I left gcc codegen untouched because there are some autogenerated lists.
This commit is contained in:
bors 2023-09-06 22:26:37 +00:00
commit d4f12a5ed3
1 changed files with 0 additions and 248 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

248
src/intrinsics/llvm_x86.rs
View File

@ -177,244 +177,6 @@ pub(crate) fn codegen_x86_llvm_intrinsic_call<'tcx>(
bool_to_zero_or_max_uint(fx, res_lane_ty, res_lane)
});
}
"llvm.x86.sse2.psrli.d" => {
let (a, imm8) = match args {
[a, imm8] => (a, imm8),
_ => bug!("wrong number of args for intrinsic {intrinsic}"),
};
let a = codegen_operand(fx, a);
let imm8 = crate::constant::mir_operand_get_const_val(fx, imm8)
.expect("llvm.x86.sse2.psrli.d imm8 not const");
simd_for_each_lane(fx, a, ret, &|fx, _lane_ty, _res_lane_ty, lane| match imm8
.try_to_bits(Size::from_bytes(4))
.unwrap_or_else(|| panic!("imm8 not scalar: {:?}", imm8))
{
imm8 if imm8 < 32 => fx.bcx.ins().ushr_imm(lane, i64::from(imm8 as u8)),
_ => fx.bcx.ins().iconst(types::I32, 0),
});
}
"llvm.x86.sse2.psrai.d" => {
let (a, imm8) = match args {
[a, imm8] => (a, imm8),
_ => bug!("wrong number of args for intrinsic {intrinsic}"),
};
let a = codegen_operand(fx, a);
let imm8 = crate::constant::mir_operand_get_const_val(fx, imm8)
.expect("llvm.x86.sse2.psrai.d imm8 not const");
simd_for_each_lane(fx, a, ret, &|fx, _lane_ty, _res_lane_ty, lane| match imm8
.try_to_bits(Size::from_bytes(4))
.unwrap_or_else(|| panic!("imm8 not scalar: {:?}", imm8))
{
imm8 if imm8 < 32 => fx.bcx.ins().sshr_imm(lane, i64::from(imm8 as u8)),
_ => fx.bcx.ins().iconst(types::I32, 0),
});
}
"llvm.x86.sse2.pslli.d" => {
let (a, imm8) = match args {
[a, imm8] => (a, imm8),
_ => bug!("wrong number of args for intrinsic {intrinsic}"),
};
let a = codegen_operand(fx, a);
let imm8 = crate::constant::mir_operand_get_const_val(fx, imm8)
.expect("llvm.x86.sse2.pslli.d imm8 not const");
simd_for_each_lane(fx, a, ret, &|fx, _lane_ty, _res_lane_ty, lane| match imm8
.try_to_bits(Size::from_bytes(4))
.unwrap_or_else(|| panic!("imm8 not scalar: {:?}", imm8))
{
imm8 if imm8 < 32 => fx.bcx.ins().ishl_imm(lane, i64::from(imm8 as u8)),
_ => fx.bcx.ins().iconst(types::I32, 0),
});
}
"llvm.x86.sse2.psrli.w" => {
let (a, imm8) = match args {
[a, imm8] => (a, imm8),
_ => bug!("wrong number of args for intrinsic {intrinsic}"),
};
let a = codegen_operand(fx, a);
let imm8 = crate::constant::mir_operand_get_const_val(fx, imm8)
.expect("llvm.x86.sse2.psrli.d imm8 not const");
simd_for_each_lane(fx, a, ret, &|fx, _lane_ty, _res_lane_ty, lane| match imm8
.try_to_bits(Size::from_bytes(4))
.unwrap_or_else(|| panic!("imm8 not scalar: {:?}", imm8))
{
imm8 if imm8 < 16 => fx.bcx.ins().ushr_imm(lane, i64::from(imm8 as u8)),
_ => fx.bcx.ins().iconst(types::I32, 0),
});
}
"llvm.x86.sse2.psrai.w" => {
let (a, imm8) = match args {
[a, imm8] => (a, imm8),
_ => bug!("wrong number of args for intrinsic {intrinsic}"),
};
let a = codegen_operand(fx, a);
let imm8 = crate::constant::mir_operand_get_const_val(fx, imm8)
.expect("llvm.x86.sse2.psrai.d imm8 not const");
simd_for_each_lane(fx, a, ret, &|fx, _lane_ty, _res_lane_ty, lane| match imm8
.try_to_bits(Size::from_bytes(4))
.unwrap_or_else(|| panic!("imm8 not scalar: {:?}", imm8))
{
imm8 if imm8 < 16 => fx.bcx.ins().sshr_imm(lane, i64::from(imm8 as u8)),
_ => fx.bcx.ins().iconst(types::I32, 0),
});
}
"llvm.x86.sse2.pslli.w" => {
let (a, imm8) = match args {
[a, imm8] => (a, imm8),
_ => bug!("wrong number of args for intrinsic {intrinsic}"),
};
let a = codegen_operand(fx, a);
let imm8 = crate::constant::mir_operand_get_const_val(fx, imm8)
.expect("llvm.x86.sse2.pslli.d imm8 not const");
simd_for_each_lane(fx, a, ret, &|fx, _lane_ty, _res_lane_ty, lane| match imm8
.try_to_bits(Size::from_bytes(4))
.unwrap_or_else(|| panic!("imm8 not scalar: {:?}", imm8))
{
imm8 if imm8 < 16 => fx.bcx.ins().ishl_imm(lane, i64::from(imm8 as u8)),
_ => fx.bcx.ins().iconst(types::I32, 0),
});
}
"llvm.x86.avx.psrli.d" => {
let (a, imm8) = match args {
[a, imm8] => (a, imm8),
_ => bug!("wrong number of args for intrinsic {intrinsic}"),
};
let a = codegen_operand(fx, a);
let imm8 = crate::constant::mir_operand_get_const_val(fx, imm8)
.expect("llvm.x86.avx.psrli.d imm8 not const");
simd_for_each_lane(fx, a, ret, &|fx, _lane_ty, _res_lane_ty, lane| match imm8
.try_to_bits(Size::from_bytes(4))
.unwrap_or_else(|| panic!("imm8 not scalar: {:?}", imm8))
{
imm8 if imm8 < 32 => fx.bcx.ins().ushr_imm(lane, i64::from(imm8 as u8)),
_ => fx.bcx.ins().iconst(types::I32, 0),
});
}
"llvm.x86.avx.psrai.d" => {
let (a, imm8) = match args {
[a, imm8] => (a, imm8),
_ => bug!("wrong number of args for intrinsic {intrinsic}"),
};
let a = codegen_operand(fx, a);
let imm8 = crate::constant::mir_operand_get_const_val(fx, imm8)
.expect("llvm.x86.avx.psrai.d imm8 not const");
simd_for_each_lane(fx, a, ret, &|fx, _lane_ty, _res_lane_ty, lane| match imm8
.try_to_bits(Size::from_bytes(4))
.unwrap_or_else(|| panic!("imm8 not scalar: {:?}", imm8))
{
imm8 if imm8 < 32 => fx.bcx.ins().sshr_imm(lane, i64::from(imm8 as u8)),
_ => fx.bcx.ins().iconst(types::I32, 0),
});
}
"llvm.x86.sse2.psrli.q" => {
let (a, imm8) = match args {
[a, imm8] => (a, imm8),
_ => bug!("wrong number of args for intrinsic {intrinsic}"),
};
let a = codegen_operand(fx, a);
let imm8 = crate::constant::mir_operand_get_const_val(fx, imm8)
.expect("llvm.x86.avx.psrli.q imm8 not const");
simd_for_each_lane(fx, a, ret, &|fx, _lane_ty, _res_lane_ty, lane| match imm8
.try_to_bits(Size::from_bytes(4))
.unwrap_or_else(|| panic!("imm8 not scalar: {:?}", imm8))
{
imm8 if imm8 < 64 => fx.bcx.ins().ushr_imm(lane, i64::from(imm8 as u8)),
_ => fx.bcx.ins().iconst(types::I32, 0),
});
}
"llvm.x86.sse2.pslli.q" => {
let (a, imm8) = match args {
[a, imm8] => (a, imm8),
_ => bug!("wrong number of args for intrinsic {intrinsic}"),
};
let a = codegen_operand(fx, a);
let imm8 = crate::constant::mir_operand_get_const_val(fx, imm8)
.expect("llvm.x86.avx.pslli.q imm8 not const");
simd_for_each_lane(fx, a, ret, &|fx, _lane_ty, _res_lane_ty, lane| match imm8
.try_to_bits(Size::from_bytes(4))
.unwrap_or_else(|| panic!("imm8 not scalar: {:?}", imm8))
{
imm8 if imm8 < 64 => fx.bcx.ins().ishl_imm(lane, i64::from(imm8 as u8)),
_ => fx.bcx.ins().iconst(types::I32, 0),
});
}
"llvm.x86.avx.pslli.d" => {
let (a, imm8) = match args {
[a, imm8] => (a, imm8),
_ => bug!("wrong number of args for intrinsic {intrinsic}"),
};
let a = codegen_operand(fx, a);
let imm8 = crate::constant::mir_operand_get_const_val(fx, imm8)
.expect("llvm.x86.avx.pslli.d imm8 not const");
simd_for_each_lane(fx, a, ret, &|fx, _lane_ty, _res_lane_ty, lane| match imm8
.try_to_bits(Size::from_bytes(4))
.unwrap_or_else(|| panic!("imm8 not scalar: {:?}", imm8))
{
imm8 if imm8 < 32 => fx.bcx.ins().ishl_imm(lane, i64::from(imm8 as u8)),
_ => fx.bcx.ins().iconst(types::I32, 0),
});
}
"llvm.x86.avx2.psrli.w" => {
let (a, imm8) = match args {
[a, imm8] => (a, imm8),
_ => bug!("wrong number of args for intrinsic {intrinsic}"),
};
let a = codegen_operand(fx, a);
let imm8 = crate::constant::mir_operand_get_const_val(fx, imm8)
.expect("llvm.x86.avx.psrli.w imm8 not const");
simd_for_each_lane(fx, a, ret, &|fx, _lane_ty, _res_lane_ty, lane| match imm8
.try_to_bits(Size::from_bytes(4))
.unwrap_or_else(|| panic!("imm8 not scalar: {:?}", imm8))
{
imm8 if imm8 < 16 => fx.bcx.ins().ushr_imm(lane, i64::from(imm8 as u8)),
_ => fx.bcx.ins().iconst(types::I32, 0),
});
}
"llvm.x86.avx2.psrai.w" => {
let (a, imm8) = match args {
[a, imm8] => (a, imm8),
_ => bug!("wrong number of args for intrinsic {intrinsic}"),
};
let a = codegen_operand(fx, a);
let imm8 = crate::constant::mir_operand_get_const_val(fx, imm8)
.expect("llvm.x86.avx.psrai.w imm8 not const");
simd_for_each_lane(fx, a, ret, &|fx, _lane_ty, _res_lane_ty, lane| match imm8
.try_to_bits(Size::from_bytes(4))
.unwrap_or_else(|| panic!("imm8 not scalar: {:?}", imm8))
{
imm8 if imm8 < 16 => fx.bcx.ins().sshr_imm(lane, i64::from(imm8 as u8)),
_ => fx.bcx.ins().iconst(types::I32, 0),
});
}
"llvm.x86.avx2.pslli.w" => {
let (a, imm8) = match args {
[a, imm8] => (a, imm8),
_ => bug!("wrong number of args for intrinsic {intrinsic}"),
};
let a = codegen_operand(fx, a);
let imm8 = crate::constant::mir_operand_get_const_val(fx, imm8)
.expect("llvm.x86.avx.pslli.w imm8 not const");
simd_for_each_lane(fx, a, ret, &|fx, _lane_ty, _res_lane_ty, lane| match imm8
.try_to_bits(Size::from_bytes(4))
.unwrap_or_else(|| panic!("imm8 not scalar: {:?}", imm8))
{
imm8 if imm8 < 16 => fx.bcx.ins().ishl_imm(lane, i64::from(imm8 as u8)),
_ => fx.bcx.ins().iconst(types::I32, 0),
});
}
"llvm.x86.ssse3.pshuf.b.128" | "llvm.x86.avx2.pshuf.b" => {
let (a, b) = match args {
[a, b] => (a, b),
@ -506,14 +268,6 @@ fn select4(
ret.place_lane(fx, 2).to_ptr().store(fx, res_2, MemFlags::trusted());
ret.place_lane(fx, 3).to_ptr().store(fx, res_3, MemFlags::trusted());
}
"llvm.x86.sse2.storeu.dq" | "llvm.x86.sse2.storeu.pd" => {
intrinsic_args!(fx, args => (mem_addr, a); intrinsic);
let mem_addr = mem_addr.load_scalar(fx);
// FIXME correctly handle the unalignment
let dest = CPlace::for_ptr(Pointer::new(mem_addr), a.layout());
dest.write_cvalue(fx, a);
}
"llvm.x86.ssse3.pabs.b.128" | "llvm.x86.ssse3.pabs.w.128" | "llvm.x86.ssse3.pabs.d.128" => {
let a = match args {
[a] => a,
@ -571,8 +325,6 @@ fn select4(
// llvm.x86.avx2.vperm2i128
// llvm.x86.ssse3.pshuf.b.128
// llvm.x86.avx2.pshuf.b
// llvm.x86.avx2.psrli.w
// llvm.x86.sse2.psrli.w
fn llvm_add_sub<'tcx>(
fx: &mut FunctionCx<'_, '_, 'tcx>,