Gathers are (unsurprisingly) a notable exception to the rule that R-V V
gets faster with larger group multipliers. So roll the function to speed
it up.
Before:
vector_fmul_reverse_fixed_c: 2840.7
vector_fmul_reverse_fixed_rvv_i32: 2430.2
After:
vector_fmul_reverse_fixed_c: 2841.0
vector_fmul_reverse_fixed_rvv_i32: 962.2
It might be possible to further optimise the function by moving the
reverse-subtract out of the loop and adding ad-hoc tail handling.
The gather index vector is only used as double-length (due to register
pressure), so no need to initialise it for quad-length. Basically this
matches the multiplier in the prologue to the the multipler in the loop.
This revectors the inner loop to reverse vectors element in vectors,
thus eliminating the negative register stride. Note that RVV does not
have a vector reverse instruction, so this uses a gather.
It does not make much sense to me, but GCC somehow optimises the
inline assembler even though the output is very obviously used and
having observable side effects.
This reverts commit 09731fbfc3.
So far, AV_READ_TIME would return the cycle counter. This posed two
problems:
1) On recent systems, it would just raise an illegal instruction
exception. Indeed RDCYCLE is blocked in user space to ward off some
side channel attacks. In particular, this would cause the random
number generator to crash.
2) It does not match the x86 behaviour and the apparent original intent
of AV_READ_TIME in the functional code base (outside test cases).
So this replaces the cycle counter with the time counter. The unit is
a platform-dependent constant fraction of time, and the value should be
stable across harts (RISC-V lingo for physical CPU thread).
1) Take the reductive sum out of the loop,
leaving a regular vector addition in the loop.
2) Merge the addition and the multiplication.
3) Unroll.
Before:
scalarproduct_float_rvv_f32: 832.5
After:
scalarproduct_float_rvv_f32: 275.2
The code was blindly assuming that Zbb or V implied Zba. While the
earlier is practically always true, the later broke some QEMU setups,
as V was introduced earlier than Zba.
As with the earlier bswap change, all versions of GCC and Clang that
support RISC-V support the popcount built-ins, so we can just use them
instead of inline assembler.
av_bswapXX() are used in context that expect exact size types, notably
variable arguments to av_log(). On Linux RV64, uint_fast32_t is an
unsigned long, so the current inline assembler does not work properly.
Since GCC and Clang gained their byte-swap built-ins before they
supported RISC-V, we can simply defer to them. As an added bonus, the
compiler can do instruction scheduling, which it couldn't with the Zbb
inline assembler.
VSETVLI xd, x0, ...' has rather nonobvious semantics:
- If xd is x0, then it preserves the current vector length.
- If xd is not x0, it sets the vector length to the supported maximum.
Also somewhat confusingly, while VMV.X.S always does its thing
regardless of the selected vector length, VMV.S.X does _nothing_ if the
selected vector length is zero.
So the current code breaks fails to initialise the accumulator if we
are unlucky to have a selected vector length of zero on entry. Fix it
by forcing the vector length to one.
On most cases, the vector type (VTYPE) for the RISC-V Vector extension
is supplied as an immediate value, with either of the VSETVLI or
VSETIVLI instructions. There is however a third instruction VSETVL
which takes the vector type from a general purpose register. That is so
the type can be selected at run-time.
This introduces a macro to load a (valid) vector type into a register.
The syntax follows that of VSETVLI and VSETIVLI, with element size,
group multiplier, then tail and mask policies.
Unfortunately, it is common, and will remain so, that the Bit
manipulations are not enabled at compilation time. This is an official
policy for Debian ports in general (though they do not support RISC-V
officially as of yet) to stick to the minimal target baseline, which
does not include the B extension or even its Zbb subset.
For inline helpers (CPOP, REV8), compiler builtins (CTZ, CLZ) or
even plain C code (MIN, MAX, MINU, MAXU), run-time detection seems
impractical. But at least it can work for the byte-swap DSP functions.
RVV defines a total of 12 different extensions, including:
- 5 different instruction subsets:
- Zve32x: 8-, 16- and 32-bit integers,
- Zve32f: Zve32x plus single precision floats,
- Zve64x: Zve32x plus 64-bit integers,
- Zve64f: Zve32f plus Zve64x,
- Zve64d: Zve64f plus double precision floats.
- 6 different vector lengths:
- Zvl32b (embedded only),
- Zvl64b (embedded only),
- Zvl128b,
- Zvl256b,
- Zvl512b,
- Zvl1024b,
- and the V extension proper: equivalent to Zve64f and Zvl128b.
In total, there are 6 different possible sets of supported instructions
(including the empty set), but for convenience we allocate one bit for
each type sets: up-to-32-bit ints (RVV_I32), floats (RVV_F32),
64-bit ints (RVV_I64) and doubles (RVV_F64).
Whence the vector size is needed, it can be retrieved by reading the
unprivileged read-only vlenb CSR. This should probably be a separate
helper macro if needed at a later point.
This introduces compile-time and run-time CPU detection on RISC-V. In
practice, I doubt that FFmpeg will ever see a RISC-V CPU without all of
I, F and D extensions, and if it does, it probably won't have run-time
detection. So the flags are essentially always set.
But as things stand, checkasm wants them that way. Compare the ARMV8
flag on AArch64. We are nowhere near running short on CPU flag bits.
If the target supports the Basic bit-manipulation (Zbb) extension, then
the REV8 instruction is available to reverse byte order.
Note that this instruction only exists at the "XLEN" register size,
so we need to right shift the result down to the data width.
If Zbb is not supported, then this patchset does nothing. Support for
run-time detection is left for the future. Currently, there are no
bits in auxv/ELF HWCAP for Z-extensions, so there are no clean ways to
do this.
This uses the architected RISC-V 64-bit cycle counter from the
RISC-V unprivileged instruction set.
In 64-bit and 128-bit, this is a straightforward CSR read.
In 32-bit mode, the 64-bit value is exposed as two CSRs, which
cannot be read atomically, so a loop is necessary to detect and fix up
the race condition where the bottom half wraps exactly between the two
reads.
Rémi Denis-Courmont