Re: [PATCH v4 2/2] eal: provide leading and trailing zero bit count abstraction

[Ferruh Yigit <ferruh.yigit@amd.com>]

On 1/9/2023 5:36 PM, Tyler Retzlaff wrote:
> From: Tyler Retzlaff <roretzla@microsoft.com>
> 
> Provide an abstraction for leading and trailing zero bit counting
> functions to hide compiler specific intrinsics and builtins.
> 
> Include basic unit test of following functions added.
> 
>     rte_clz32
>     rte_clz64
>     rte_ctz32
>     rte_ctz64
> 
> Signed-off-by: Tyler Retzlaff <roretzla@linux.microsoft.com>
> ---
>  app/test/meson.build         |   2 +
>  app/test/test_bitcount.c     |  71 ++++++++++++++++++
>  lib/eal/include/rte_bitops.h | 168 +++++++++++++++++++++++++++++++++++++++++++
>  3 files changed, 241 insertions(+)
>  create mode 100644 app/test/test_bitcount.c
> 
> diff --git a/app/test/meson.build b/app/test/meson.build
> index f34d19e..d1277bc 100644
> --- a/app/test/meson.build
> +++ b/app/test/meson.build
> @@ -13,6 +13,7 @@ test_sources = files(
>          'test_alarm.c',
>          'test_atomic.c',
>          'test_barrier.c',
> +        'test_bitcount.c',
>          'test_bitops.c',
>          'test_bitmap.c',
>          'test_bpf.c',
> @@ -160,6 +161,7 @@ test_deps += ['bus_pci', 'bus_vdev']
>  fast_tests = [
>          ['acl_autotest', true, true],
>          ['atomic_autotest', false, true],
> +        ['bitcount_autotest', true, true],
>          ['bitmap_autotest', true, true],
>          ['bpf_autotest', true, true],
>          ['bpf_convert_autotest', true, true],
> diff --git a/app/test/test_bitcount.c b/app/test/test_bitcount.c
> new file mode 100644
> index 0000000..36eb05c
> --- /dev/null
> +++ b/app/test/test_bitcount.c
> @@ -0,0 +1,71 @@
> +/* SPDX-License-Identifier: BSD-3-Clause
> + * Copyright (C) 2022 Microsoft Corporation
> + */
> +
> +#include <string.h>
> +
> +#include <rte_bitops.h>
> +#include <rte_debug.h>
> +
> +#include "test.h"
> +
> +RTE_LOG_REGISTER(bitcount_logtype_test, test.bitcount, INFO);
> +
> +static int
> +test_clz32(void)
> +{
> +	uint32_t v = 1;
> +	RTE_TEST_ASSERT(rte_clz32(v) == sizeof(v) * CHAR_BIT - 1,
> +	    "Unexpected count.");
> +
> +	return 0;
> +}
> +
> +static int
> +test_clz64(void)
> +{
> +	uint64_t v = 1;
> +	RTE_TEST_ASSERT(rte_clz64(v) == sizeof(v) * CHAR_BIT - 1,
> +	    "Unexpected count.");
> +
> +	return 0;
> +}
> +
> +static int
> +test_ctz32(void)
> +{
> +	uint32_t v = 2;
> +	RTE_TEST_ASSERT(rte_ctz32(v) == 1, "Unexpected count.");
> +
> +	return 0;
> +}
> +
> +static int
> +test_ctz64(void)
> +{
> +	uint64_t v = 2;
> +	RTE_TEST_ASSERT(rte_ctz64(v) == 1, "Unexpected count.");
> +
> +	return 0;
> +}
> +
> +static struct unit_test_suite bitcount_test_suite = {
> +	.suite_name = "bitcount autotest",
> +	.setup = NULL,
> +	.teardown = NULL,
> +	.unit_test_cases = {
> +		TEST_CASE(test_clz32),
> +		TEST_CASE(test_clz64),
> +		TEST_CASE(test_ctz32),
> +		TEST_CASE(test_ctz64),
> +		TEST_CASES_END()
> +	}
> +};
> +
> +static int
> +test_bitcount(void)
> +{
> +	return unit_test_suite_runner(&bitcount_test_suite);
> +}
> +
> +REGISTER_TEST_COMMAND(bitcount_autotest, test_bitcount);
> diff --git a/lib/eal/include/rte_bitops.h b/lib/eal/include/rte_bitops.h
> index 531479e..387d7aa 100644
> --- a/lib/eal/include/rte_bitops.h
> +++ b/lib/eal/include/rte_bitops.h
> @@ -1,5 +1,7 @@
>  /* SPDX-License-Identifier: BSD-3-Clause
>   * Copyright(c) 2020 Arm Limited
> + * Copyright(c) 2010-2019 Intel Corporation
> + * Copyright(c) 2023 Microsoft Corporation
>   */
>  
>  #ifndef _RTE_BITOPS_H_
> @@ -275,6 +277,172 @@
>  	return val & mask;
>  }
>  
> +#ifdef RTE_TOOLCHAIN_MSVC
> +
> +/**
> + * @warning
> + * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
> + *
> + * Get the count of leading 0-bits in v.
> + *
> + * @param v
> + *   The value.
> + * @return
> + *   The count of leading zero bits.
> + */
> +__rte_experimental
> +static inline unsigned int
> +rte_clz32(uint32_t v)
> +{
> +	unsigned long rv;
> +
> +	(void)_BitScanReverse(&rv, v);
> +
> +	return (unsigned int)rv;
> +}


OK, this looks wrong to me,


'_BitScanReverse' is [1]:
"Search the mask data from most significant bit (MSB) to least
significant bit (LSB) for a set bit (1)."

As far as I can see index starts from zero and from lsb to msb.

So _BitScanReverse() returns following index values:
0x2 => 1
0xffffffff => 31

If above is correct, above function doesn't return number of leading
zeros, but it should return "31 - (unsigned int)rv".

Please check godbolt experiment for above:
https://godbolt.org/z/znYn54f57


As far as I can see unit test is correct, so above should fail with
windows compiler, and I assume you run unit test with windows compiler,
so probably I am missing something but please help me understand.



[1]
https://learn.microsoft.com/en-us/cpp/intrinsics/bitscanreverse-bitscanreverse64?view=msvc-170

> +
> +/**
> + * @warning
> + * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
> + *
> + * Get the count of leading 0-bits in v.
> + *
> + * @param v
> + *   The value.
> + * @return
> + *   The count of leading zero bits.
> + */
> +__rte_experimental
> +static inline unsigned int
> +rte_clz64(uint64_t v)
> +{
> +	unsigned long rv;
> +
> +	(void)_BitScanReverse64(&rv, v);
> +
> +	return (unsigned int)rv;
> +}
> +
> +/**
> + * @warning
> + * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
> + *
> + * Get the count of trailing 0-bits in v.
> + *
> + * @param v
> + *   The value.
> + * @return
> + *   The count of trailing zero bits.
> + */
> +__rte_experimental
> +static inline unsigned int
> +rte_ctz32(uint32_t v)
> +{
> +	unsigned long rv;
> +
> +	(void)_BitScanForward(&rv, v);
> +
> +	return (unsigned int)rv;
> +}
> +
> +/**
> + * @warning
> + * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
> + *
> + * Get the count of trailing 0-bits in v.
> + *
> + * @param v
> + *   The value.
> + * @return
> + *   The count of trailing zero bits.
> + */
> +__rte_experimental
> +static inline unsigned int
> +rte_ctz64(uint64_t v)
> +{
> +	unsigned long rv;
> +
> +	(void)_BitScanForward64(&rv, v);
> +
> +	return (unsigned int)rv;
> +}
> +
> +#else
> +
> +/**
> + * @warning
> + * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
> + *
> + * Get the count of leading 0-bits in v.
> + *
> + * @param v
> + *   The value.
> + * @return
> + *   The count of leading zero bits.
> + */
> +__rte_experimental
> +static inline unsigned int
> +rte_clz32(uint32_t v)
> +{
> +	return (unsigned int)__builtin_clz(v);
> +}
> +
> +/**
> + * @warning
> + * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
> + *
> + * Get the count of leading 0-bits in v.
> + *
> + * @param v
> + *   The value.
> + * @return
> + *   The count of leading zero bits.
> + */
> +__rte_experimental
> +static inline unsigned int
> +rte_clz64(uint64_t v)
> +{
> +	return (unsigned int)__builtin_clzll(v);
> +}
> +
> +/**
> + * @warning
> + * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
> + *
> + * Get the count of trailing 0-bits in v.
> + *
> + * @param v
> + *   The value.
> + * @return
> + *   The count of trailing zero bits.
> + */
> +__rte_experimental
> +static inline unsigned int
> +rte_ctz32(uint32_t v)
> +{
> +	return (unsigned int)__builtin_ctz(v);
> +}
> +
> +/**
> + * @warning
> + * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
> + *
> + * Get the count of trailing 0-bits in v.
> + *
> + * @param v
> + *   The value.
> + * @return
> + *   The count of trailing zero bits.
> + */
> +__rte_experimental
> +static inline unsigned int
> +rte_ctz64(uint64_t v)
> +{
> +	return (unsigned int)__builtin_ctzll(v);
> +}

'rte_ctz32()' & 'rte_ctz64()' are practically exact same with
'rte_bsf32()' & 'rte_bsf64()',

Although I can see description is different, do we need same
functionality with new function name,
why not add 'rte_bsf32()' & 'rte_bsf64()' versions for the Windows?



btw, do you guys know what 'bsf' & 'fls' (rte_fls_u32) stands for?

> +
> +#endif
> +
>  /**
>   * Combines 32b inputs most significant set bits into the least
>   * significant bits to construct a value with the same MSBs as x