On Wed, 4 Dec 2024 at 12:39, Richard Earnshaw (lists)
<richard.earns...@arm.com> wrote:
>
> On 25/11/2024 20:08, Christophe Lyon wrote:
> > In this PR, we have to handle a case where MVE predicates are supplied
> > as a const_int, where individual predicates have illegal boolean
> > values (such as 0xc for a 4-bit boolean predicate). To avoid the ICE,
> > fix the constant (any non-zero value is converted to all 1s) and emit
> > a warning.
> >
> > On MVE, V8BI and V4BI multi-bit masks are interpreted byte-by-byte at
> > instruction level, but end-users should describe lanes rather than
> > bytes (so all bytes of a true-predicated lane should be '1'), see
> > https://developer.arm.com/documentation/101028/0012/14--M-profile-Vector-Extension--MVE--intrinsics.
> >
> > Since gen_lowpart can ICE on a subreg, we force predicates in a subreg
> > into a reg, after removing subreg of the same size as the target
> > (HImode) which would be made redundant by gen_lowpart and confuse the
> > DLSTP optimization.
> >
> > 2024-11-20 Christophe Lyon <christophe.l...@linaro.org>
> > Jakub Jelinek <ja...@redhat.com>
> >
> > PR target/114801
> > gcc/
> > * config/arm/arm-mve-builtins.cc
> > (function_expander::add_input_operand): Handle CONST_INT
> > predicates.
> >
> > gcc/testsuite/
> > * gcc.target/arm/mve/pr108443.c: Update predicate constant.
> > * gcc.target/arm/mve/pr114801.c: New test.
> > ---
> > gcc/config/arm/arm-mve-builtins.cc | 37 ++++++++++++++++++-
> > gcc/testsuite/gcc.target/arm/mve/pr108443.c | 4 +--
> > gcc/testsuite/gcc.target/arm/mve/pr114801.c | 39 +++++++++++++++++++++
> > 3 files changed, 77 insertions(+), 3 deletions(-)
> > create mode 100644 gcc/testsuite/gcc.target/arm/mve/pr114801.c
> >
> > diff --git a/gcc/config/arm/arm-mve-builtins.cc
> > b/gcc/config/arm/arm-mve-builtins.cc
> > index 255aed25600..5ff32ce06b7 100644
> > --- a/gcc/config/arm/arm-mve-builtins.cc
> > +++ b/gcc/config/arm/arm-mve-builtins.cc
> > @@ -2352,7 +2352,42 @@ function_expander::add_input_operand (insn_code
> > icode, rtx x)
> > mode = GET_MODE (x);
> > }
> > else if (VALID_MVE_PRED_MODE (mode))
> > - x = gen_lowpart (mode, x);
> > + {
> > + if (CONST_INT_P (x) && (mode == V8BImode || mode == V4BImode))
> > + {
> > + /* In V8BI or V4BI each element has 2 or 4 bits, if those bits
> > aren't
> > + all the same, gen_lowpart might ICE. Canonicalize all the 2 or 4
> > + bits to all ones if any of them is non-zero. V8BI and V4BI
> > + multi-bit masks are interpreted byte-by-byte at instruction
> > level,
> > + but such constants should describe lanes, rather than bytes. See
> > +
> > https://developer.arm.com/documentation/101028/0012/14--M-profile-Vector-Extension--MVE--intrinsics.
> > */
>
> Apart from being an overly long line, deep links like this are generally not
> very stable. I suggest we just say something like "See the section on MVE
> intrinsics in the Arm ACLE specification".
Right, I was wondering what was the best practice, I think I've seen
such links recently, not sure where.
I'll update the comment, and the commit message.
>
> > + unsigned HOST_WIDE_INT xi = UINTVAL (x);
> > + xi |= ((xi & 0x5555) << 1) | ((xi & 0xaaaa) >> 1);
> > + if (mode == V4BImode)
> > + xi |= ((xi & 0x3333) << 2) | ((xi & 0xcccc) >> 2);
> > + if (xi != UINTVAL (x))
> > + inform (location, "constant predicate argument %d (%wx) does"
> > + " not map to %d lane numbers, converted to %wx",
> > + opno, UINTVAL (x) & 0xffff, mode == V8BImode ? 8 : 4,
> > + xi & 0xffff);
>
> I think this should be a warning (so that werror can work with it).
> Otherwise such messages can't be faulted.
OK, I will change this.
>
> > +
> > + x = gen_int_mode (xi, HImode);
> > + }
> > + else if (SUBREG_P (x))
> > + {
> > + /* Already of the right size, drop the subreg which will be made
> > + redundant by gen_lowpart below. */
> > + if (GET_MODE_SIZE (GET_MODE (x)) == GET_MODE_SIZE (HImode)
> > + || SUBREG_BYTE (x) == 0)
> > + x = SUBREG_REG (x);
> > +
> > + /* gen_lowpart on a SUBREG can ICE. */
> > + if (gen_lowpart_common (mode, x) == 0)
> > + x = force_reg (GET_MODE (x), x);
> > + }
> > +
> > + x = gen_lowpart (mode, x);
>
> I wonder if this is overly complex. Wouldn't it be better to just write here:
>
> else if (!REG_P (x))
> x = force_reg (GET_MODE (x), x);
>
> and then let the optimizers clean things up?
>
The !REG_P(x) condition is not right, because force_reg crashes if x
== (const_int 1) for instance (we have mode=VOID in
int_mode_for_mode).
The first 'if' which looks at the mode sizes it to avoid regressions
in DLSTP transform, despite recent improvements there.
Without this, dlstp-compile-asm-2.c has regressions in test6 and test9.
For some reason we have:
(set (reg:HI 137 [ p1 ]
(subreg/s/v:HI (reg/v:SI 129 [ p1 ]) 0)) (nil))
(set (reg:V4BI 138 [ p1 ])
(subreg:V4BI (reg:HI 137 [ p1 ]) 0)) (nil))
which not hoisted, while with the check above we hoist this:
(set (reg:V4BI 136 [ p1 ])
(subreg:V4BI (reg/v:SI 129 [ p1 ]) 0)) (nil))
I was thinking this could be OK for gcc-15, and I would look at
removing this hack for gcc-16.
The gen_lowpart_common (mode, x) == 0 part is to call force_reg only
if gen_lowpart might have trouble with the current x, but avoid
unnecessary force_reg that could confuse DLSTP transform too.
In practice this is not exercised by the testsuite, but Jakub raised
concerns after gen_lowpart possibly ICEing.
Thanks,
Christophe
> R.
>
> > + }
> >
> > m_ops.safe_grow (m_ops.length () + 1, true);
> > create_input_operand (&m_ops.last (), x, mode);
> > diff --git a/gcc/testsuite/gcc.target/arm/mve/pr108443.c
> > b/gcc/testsuite/gcc.target/arm/mve/pr108443.c
> > index c5fbfa4a1bb..0c0e2dd6eb8 100644
> > --- a/gcc/testsuite/gcc.target/arm/mve/pr108443.c
> > +++ b/gcc/testsuite/gcc.target/arm/mve/pr108443.c
> > @@ -7,8 +7,8 @@
> > void
> > __attribute__ ((noipa)) partial_write_cst (uint32_t *a, uint32x4_t v)
> > {
> > - vstrwq_p_u32 (a, v, 0x00CC);
> > + vstrwq_p_u32 (a, v, 0x00FF);
> > }
> >
> > -/* { dg-final { scan-assembler {mov\tr[0-9]+, #204} } } */
> > +/* { dg-final { scan-assembler {mov\tr[0-9]+, #255} } } */
> >
> > diff --git a/gcc/testsuite/gcc.target/arm/mve/pr114801.c
> > b/gcc/testsuite/gcc.target/arm/mve/pr114801.c
> > new file mode 100644
> > index 00000000000..d051e309d0b
> > --- /dev/null
> > +++ b/gcc/testsuite/gcc.target/arm/mve/pr114801.c
> > @@ -0,0 +1,39 @@
> > +/* { dg-do compile } */
> > +/* { dg-require-effective-target arm_v8_1m_mve_ok } */
> > +/* { dg-options "-O2" } */
> > +/* { dg-add-options arm_v8_1m_mve } */
> > +/* { dg-final { check-function-bodies "**" "" "" } } */
> > +
> > +#include <arm_mve.h>
> > +
> > +/*
> > +** test_32:
> > +**...
> > +** mov r[0-9]+, #65295 @ movhi
> > +**...
> > +*/
> > +uint32x4_t test_32() {
> > + /* V4BI predicate converted to 0xff0f. */
> > + return vdupq_m_n_u32(vdupq_n_u32(0xffffffff), 0, 0x4f02); /* { dg-note
> > {constant predicate argument 3 \(0x4f02\) does not map to 4 lane numbers,
> > converted to 0xff0f} } */
> > +}
> > +
> > +/*
> > +** test_16:
> > +**...
> > +** mov r[0-9]+, #12339 @ movhi
> > +**...
> > +*/
> > +uint16x8_t test_16() {
> > + /* V8BI predicate converted to 0x3033. */
> > + return vdupq_m_n_u16(vdupq_n_u16(0xffff), 0, 0x3021); /* { dg-note
> > {constant predicate argument 3 \(0x3021\) does not map to 8 lane numbers,
> > converted to 0x3033} } */
> > +}
> > +
> > +/*
> > +** test_8:
> > +**...
> > +** mov r[0-9]+, #23055 @ movhi
> > +**...
> > +*/
> > +uint8x16_t test_8() {
> > + return vdupq_m_n_u8(vdupq_n_u8(0xff), 0, 0x5a0f);
> > +}
>
