This ensures we clip the index to be in bounds of the vector we are
inserting into. If the index is out of bounds the results of the insert
element is poison. If we don't clip the index we can write memory that
was not part of the original store.
Fixes#74248#75557.
Now we have default abi lp64 for rv64if and ilp32 for rv32if, which is
different with riscv-gnu-toolchain. In
8e9fb09a0c/configure (L3385)
when have f and not d, it prefers lp64f/ilp32f but no soft float. This
patch tries to make their behaviors consistent.
For non-GlobalValue references, the small and medium code models can use
32 bit constants.
For GlobalValue references, use TargetMachine::isLargeGlobalObject().
Look through aliases for determining if a GlobalValue is small or large.
Even the large code model can reference small objects with 32 bit
constants as long as we're in no-pic mode, or if the reference is offset
from the GOT.
Original commit broke the build...
First reland broke large PIC builds referencing small data since it was using GOTOFF as a 32-bit constant.
... by lowering them as lazy resolve-on-first-use symbol resolvers. Note that this is subtly different timing than on ELF platforms, where ifunc resolution happens at load time.
Since ld64 and ld-prime don't support all the cases we need for these, we lower them manually in the AsmPrinter.
For non-GlobalValue references, the small and medium code models can use
32 bit constants.
For GlobalValue references, use TargetMachine::isLargeGlobalObject().
Look through aliases for determining if a GlobalValue is small or large.
Even the large code model can reference small objects with 32 bit
constants as long as we're in no-pic mode, or if the reference is offset
from the GOT.
Original commit broke the build...
For non-GlobalValue references, the small and medium code models can use
32 bit constants.
For GlobalValue references, use TargetMachine::isLargeGlobalObject().
Look through aliases for determining if a GlobalValue is small or large.
Even the large code model can reference small objects with 32 bit
constants as long as we're in no-pic mode, or if the reference is offset
from the GOT.
When doing our backwards walk, we were not handling the case where the
AVL was defined by a register whose definition was an ADDI xN, x0,
<imm>. Doing so (as we already do in the forward pass) allows us to
prune a few more transitions.
The SystemZ implementation of shouldCoalesce() is merely a workaround
for the fact that regalloc can run out of registers when extending 128-bit
intervals with subreg (GPR64/GPR32) COPYs.
This patch adds more freedom to the coalescer as it now only checks that
the subreg interval is local to MBB and does not have too many physreg
clobbers.
In some cases, the machine outliner needs to preserve LR across an
outlined call by pushing it onto the stack. Previously, this also
generated unwind table instructions, which is incorrect because EHABI
unwind tables cannot represent different stack frames a different points
in the function, so the extra unwind info applied to the entire
function.
The outliner code already avoided generating CFI instructions, but EHABI
unwind data is generated later from the actual instructions, so we need
to avoid using the FrameSetup and FrameDestroy flags to prevent unwind
data being generated.
If we know the exact VLEN, then we can tell if the AVL for particular
operation is equivalent to the vsetvli xN, zero, <vtype> encoding. Using
this encoding is better than having to materialize an immediate in a
register, but worse than being able to use the vsetivli zero, imm,
<type> encoding.
Middle end up optimizations can speculate away the short circuit
behavior of C/C++ && and ||. Using i1 and/or or logical select
instructions and a single branch.
SelectionDAGBuilder can turn i1 and/or/select back into multiple
branches, but this is disabled when jump is expensive.
RISC-V can use slt(u)(i) to evaluate a condition into any GPR which
makes us better than other targets that use a flag register. RISC-V also
has single instruction compare and branch. So its not clear from a code
size perspective that using compare+and/or is better.
If the full condition is dependent on multiple loads, using a logic
delays the branch resolution until all the loads are resolved even if
there is a cheap condition that makes the loads unnecessary.
PowerPC and Lanai are the only CPU targets that use setJumpIsExpensive.
NVPTX and AMDGPU also use it but they are GPU targets. PowerPC appears
to have a MachineIR pass that turns AND/OR of CR bits into multiple
branches. I don't know anything about Lanai and their reason for using
setJumpIsExpensive.
I think the decision to use logic vs branches is much more nuanced than
this big hammer. So I propose to make RISC-V match other CPU targets.
Anyone who wants the old behavior can still pass -mllvm
-jump-is-expensive=true.
This patch implements the builtins in Clang
and the LLVM-IR intrinsic for the following:
// Variants are also available for:
// _s8, _s16, _u16, _s32, _u32, _s64, _u64,
// _f16, _f32, _f64uint8x16_t svaddqv[_u8](svbool_t pg, svuint8_t zn);
// Variants are also available for:
// _s8, _u16, _s16, _u32, _s32, _u64, _s64
uint8x16_t svandqv[_u8](svbool_t pg, svuint8_t zn); uint8x16_t
sveorqv[_u8](svbool_t pg, svuint8_t zn); uint8x16_t svorqv[_u8](svbool_t
pg, svuint8_t zn);
// Variants are also available for:
// _s8, _u16, _s16, _u32, _s32, _u64, _s64;
uint8x16_t svmaxqv[_u8](svbool_t pg, svuint8_t zn); uint8x16_t
svminqv[_u8](svbool_t pg, svuint8_t zn);
// Variants are also available for _f32, _f64
float16x8_t svmaxnmqv[_f16](svbool_t pg, svfloat16_t zn); float16x8_t
svminnmqv[_f16](svbool_t pg, svfloat16_t zn);
According to the PR#257[1]
The reduction instruction uses scalable vectors as input and fixed
vectors as output, therefore we changed SVEEmitter to emit fixed vector
types in case the neon header(arm_neon.h) is not present.
[1]https://github.com/ARM-software/acle/pull/257
Co-author: Dinar Temirbulatov <dinar.temirbulatov@arm.com>
Add empty AMDGPUGlobalISelDivergenceLowering pass. This pass will
implement
- selection of divergent i1 phis as lane mask phis, requires lane mask
merging in some cases
- lower uses of divergent i1 values outside of the cycle using lane mask
merging
- lowering of all cases of temporal divergence:
- lower uses of uniform i1 values outside of the cycle using lane mask
merging
- lower uses of uniform non-i1 values outside of the cycle using a copy
to vgpr inside of the cycle
Add very detailed set of regression tests for cases mentioned above.
patch 1 from: https://github.com/llvm/llvm-project/pull/73337
When doing sink-and-fold, the MachineSink clears the "killed" flags of
the operands of the sunk (and deleted) instruction. However, this is not
always sufficient. In some cases we can create the new load/store
instruction with operands other than the ones present in the deleted
instruction. One such example is folding a zero word extend into a
memory load on AArch64. The zero-extend is represented by a pair of
instructions - `MOV` (i.e. `ORRwrs`) followed by a `SUBREG_TO_REG`. The
`SUBREG_TO_REG` is deleted (it is the sunk instruction), but the new
load instruction mentions operands "killed" in the `MOV`, which is no
longer correct.
To fix this, clear the "killed" flags of the registers participating in
the addressing mode.
If the rematerialize was placing a subregister into a super register,
and implicit operands referenced the original register, we need to add
undef flags to the now-subregister indexed implicit operands.
Depends #75152
