21/06/2022 17:50, Rahul Bhansali пишет:
-----Original Message----- From: Konstantin Ananyev <konstantin.v.anan...@yandex.ru> Sent: Tuesday, June 21, 2022 4:43 AM To: Rahul Bhansali <rbhans...@marvell.com>; dev@dpdk.org; Ruifeng Wang <ruifeng.w...@arm.com> Cc: Jerin Jacob Kollanukkaran <jer...@marvell.com> Subject: [EXT] Re: [PATCH v2 1/2] examples/l3fwd: common packet group functionality External Email ---------------------------------------------------------------------- 17/06/2022 08:50, Rahul Bhansali пишет:CC: Konstantin Ananyev-----Original Message----- From: Rahul Bhansali <rbhans...@marvell.com> Sent: Friday, June 17, 2022 1:13 PM To: dev@dpdk.org; Ruifeng Wang <ruifeng.w...@arm.com> Cc: Jerin Jacob Kollanukkaran <jer...@marvell.com>; Rahul Bhansali <rbhans...@marvell.com> Subject: [PATCH v2 1/2] examples/l3fwd: common packet group functionality This will make the packet grouping function common, so that other examples can utilize as per need. Signed-off-by: Rahul Bhansali <rbhans...@marvell.com> --- Changes in v2: New patch to address review comment. examples/common/neon_common.h | 50 ++++++++++++ examples/common/pkt_group.h | 139 ++++++++++++++++++++++++++++++++++ examples/l3fwd/Makefile | 5 +- examples/l3fwd/l3fwd.h | 2 - examples/l3fwd/l3fwd_common.h | 129 +------------------------------ examples/l3fwd/l3fwd_neon.h | 43 +---------- examples/meson.build | 2 +- 7 files changed, 198 insertions(+), 172 deletions(-) create mode 100644 examples/common/neon_common.h create mode 100644 examples/common/pkt_group.h diff --git a/examples/common/neon_common.h b/examples/common/neon_common.h new file mode 100644 index 0000000000..f01b5ab6bc --- /dev/null +++ b/examples/common/neon_common.h @@ -0,0 +1,50 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2016-2018 Intel Corporation. + * Copyright(c) 2017-2018 Linaro Limited. + * Copyright(C) 2022 Marvell. + */ + +#ifndef _NEON_COMMON_H_ +#define _NEON_COMMON_H_ + +#include "pkt_group.h" + +/* + * Group consecutive packets with the same destination port in bursts of 4. + * Suppose we have array of destination ports: + * dst_port[] = {a, b, c, d,, e, ... } + * dp1 should contain: <a, b, c, d>, dp2: <b, c, d, e>. + * We doing 4 comparisons at once and the result is 4 bit mask. + * This mask is used as an index into prebuild array of pnum values. + */ +static inline uint16_t * +neon_port_groupx4(uint16_t pn[FWDSTEP + 1], uint16_t *lp, uint16x8_tdp1,+ uint16x8_t dp2) +{ + union { + uint16_t u16[FWDSTEP + 1]; + uint64_t u64; + } *pnum = (void *)pn; + + uint16x8_t mask = {1, 2, 4, 8, 0, 0, 0, 0}; + int32_t v; + + dp1 = vceqq_u16(dp1, dp2); + dp1 = vandq_u16(dp1, mask); + v = vaddvq_u16(dp1); + + /* update last port counter. */hh + lp[0] += gptbl[v].lpv; + rte_compiler_barrier(); + + /* if dest port value has changed. */ + if (v != GRPMSK) { + pnum->u64 = gptbl[v].pnum; + pnum->u16[FWDSTEP] = 1; + lp = pnum->u16 + gptbl[v].idx; + } + + return lp; +}Thanks for the effort. As I can see this function: port_groupx4() is nearly identical for all 3 platforms: sse/nenon/altivec (except of course built-in arch-specific instincts). In fact, even comemnts are identical. I wonder can we have something like: examples/common/<arch>/port_group.h and for each arch will have defined port_groupx4(...) ?Yes, It’s a good point. I was thinking to have arch in file name itself. But we can have arch specific directory and have different header files. Do you want me to make changes for all 3 sse/neon/altivec or just neon ?
My thought was to move headers for all archs.
I can check compilation for all but functionality/perf validate for Neon only.
I can do quick functional test for x86. Plus I think l3fwd is part of release cycle testing anyway. Thanks Konstantin
+ +#endif /* _NEON_COMMON_H_ */ diff --git a/examples/common/pkt_group.h b/examples/common/pkt_group.h new file mode 100644 index 0000000000..8b26d9380f --- /dev/null +++ b/examples/common/pkt_group.h @@ -0,0 +1,139 @@ +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2016-2018 Intel Corporation. + * Copyright(c) 2017-2018 Linaro Limited. + * Copyright(C) 2022 Marvell. + */ + +#ifndef _PKT_GROUP_H_ +#define _PKT_GROUP_H_ + +#define FWDSTEP 4 + +/* + * Group consecutive packets with the same destination port into one burst. + * To avoid extra latency this is done together with some other +packet + * processing, but after we made a final decision about packet's destination. + * To do this we maintain: + * pnum - array of number of consecutive packets with the same dest +port for + * each packet in the input burst. + * lp - pointer to the last updated element in the pnum. + * dlp - dest port value lp corresponds to. + */ + +#define GRPSZ (1 << FWDSTEP) +#define GRPMSK (GRPSZ - 1) + +#define GROUP_PORT_STEP(dlp, dcp, lp, pn, idx) do { \ + if (likely((dlp) == (dcp)[(idx)])) { \ + (lp)[0]++; \ + } else { \ + (dlp) = (dcp)[idx]; \ + (lp) = (pn) + (idx); \ + (lp)[0] = 1; \ + } \ +} while (0) + +static const struct { + uint64_t pnum; /* prebuild 4 values for pnum[]. */ + int32_t idx; /* index for new last updated elemnet. */ + uint16_t lpv; /* add value to the last updated element. */ } +gptbl[GRPSZ] = { + { + /* 0: a != b, b != c, c != d, d != e */ + .pnum = UINT64_C(0x0001000100010001), + .idx = 4, + .lpv = 0, + }, + { + /* 1: a == b, b != c, c != d, d != e */ + .pnum = UINT64_C(0x0001000100010002), + .idx = 4, + .lpv = 1, + }, + { + /* 2: a != b, b == c, c != d, d != e */ + .pnum = UINT64_C(0x0001000100020001), + .idx = 4, + .lpv = 0, + }, + { + /* 3: a == b, b == c, c != d, d != e */ + .pnum = UINT64_C(0x0001000100020003), + .idx = 4, + .lpv = 2, + }, + { + /* 4: a != b, b != c, c == d, d != e */ + .pnum = UINT64_C(0x0001000200010001), + .idx = 4, + .lpv = 0, + }, + { + /* 5: a == b, b != c, c == d, d != e */ + .pnum = UINT64_C(0x0001000200010002), + .idx = 4, + .lpv = 1, + }, + { + /* 6: a != b, b == c, c == d, d != e */ + .pnum = UINT64_C(0x0001000200030001), + .idx = 4, + .lpv = 0, + }, + { + /* 7: a == b, b == c, c == d, d != e */ + .pnum = UINT64_C(0x0001000200030004), + .idx = 4, + .lpv = 3, + }, + { + /* 8: a != b, b != c, c != d, d == e */ + .pnum = UINT64_C(0x0002000100010001), + .idx = 3, + .lpv = 0, + }, + { + /* 9: a == b, b != c, c != d, d == e */ + .pnum = UINT64_C(0x0002000100010002), + .idx = 3, + .lpv = 1, + }, + { + /* 0xa: a != b, b == c, c != d, d == e */ + .pnum = UINT64_C(0x0002000100020001), + .idx = 3, + .lpv = 0, + }, + { + /* 0xb: a == b, b == c, c != d, d == e */ + .pnum = UINT64_C(0x0002000100020003), + .idx = 3, + .lpv = 2, + }, + { + /* 0xc: a != b, b != c, c == d, d == e */ + .pnum = UINT64_C(0x0002000300010001), + .idx = 2, + .lpv = 0, + }, + { + /* 0xd: a == b, b != c, c == d, d == e */ + .pnum = UINT64_C(0x0002000300010002), + .idx = 2, + .lpv = 1, + }, + { + /* 0xe: a != b, b == c, c == d, d == e */ + .pnum = UINT64_C(0x0002000300040001), + .idx = 1, + .lpv = 0, + }, + { + /* 0xf: a == b, b == c, c == d, d == e */ + .pnum = UINT64_C(0x0002000300040005), + .idx = 0, + .lpv = 4, + }, +}; + +#endif /* _PKT_GROUP_H_ */ diff --git a/examples/l3fwd/Makefile b/examples/l3fwd/Makefile index 8efe6378e2..8dbe85c2e6 100644 --- a/examples/l3fwd/Makefile +++ b/examples/l3fwd/Makefile @@ -22,6 +22,7 @@ shared: build/$(APP)-shared static: build/$(APP)-static ln -sf $(APP)-static build/$(APP) +INCLUDES =-I../common PC_FILE := $(shell $(PKGCONF) --path libdpdk 2>/dev/null) CFLAGS += -O3 $(shell $(PKGCONF) --cflags libdpdk) # Added for'rte_eth_link_to_str()'@@ -38,10 +39,10 @@ endif endif build/$(APP)-shared: $(SRCS-y) Makefile $(PC_FILE) | build - $(CC) $(CFLAGS) $(SRCS-y) -o $@ $(LDFLAGS) $(LDFLAGS_SHARED) + $(CC) $(CFLAGS) $(SRCS-y) $(INCLUDES) -o $@ $(LDFLAGS) +$(LDFLAGS_SHARED) build/$(APP)-static: $(SRCS-y) Makefile $(PC_FILE) | build - $(CC) $(CFLAGS) $(SRCS-y) -o $@ $(LDFLAGS) $(LDFLAGS_STATIC) + $(CC) $(CFLAGS) $(SRCS-y) $(INCLUDES) -o $@ $(LDFLAGS) +$(LDFLAGS_STATIC) build: @mkdir -p $@ diff --git a/examples/l3fwd/l3fwd.h b/examples/l3fwd/l3fwd.h index 8a52c90755..40b5f32a9e 100644 --- a/examples/l3fwd/l3fwd.h +++ b/examples/l3fwd/l3fwd.h @@ -44,8 +44,6 @@ /* Used to mark destination port as 'invalid'. */ #define BAD_PORT ((uint16_t)-1) -#define FWDSTEP 4 - /* replace first 12B of the ethernet header. */ #define MASK_ETH 0x3f diff --git a/examples/l3fwd/l3fwd_common.h b/examples/l3fwd/l3fwd_common.h index 8e4c27218f..224b1c08e8 100644 --- a/examples/l3fwd/l3fwd_common.h +++ b/examples/l3fwd/l3fwd_common.h @@ -7,6 +7,8 @@ #ifndef _L3FWD_COMMON_H_ #define _L3FWD_COMMON_H_ +#include "pkt_group.h" + #ifdef DO_RFC_1812_CHECKS #define IPV4_MIN_VER_IHL 0x45 @@ -50,133 +52,6 @@ rfc1812_process(struct rte_ipv4_hdr *ipv4_hdr, uint16_t *dp, uint32_t ptype) #define rfc1812_process(mb, dp, ptype) do { } while (0) #endif /* DO_RFC_1812_CHECKS */ -/* - * We group consecutive packets with the same destination port into oneburst.- * To avoid extra latency this is done together with some other packet - * processing, but after we made a final decision about packet's destination. - * To do this we maintain: - * pnum - array of number of consecutive packets with the same dest port for - * each packet in the input burst. - * lp - pointer to the last updated element in the pnum. - * dlp - dest port value lp corresponds to. - */ - -#define GRPSZ (1 << FWDSTEP) -#define GRPMSK (GRPSZ - 1) - -#define GROUP_PORT_STEP(dlp, dcp, lp, pn, idx) do { \ - if (likely((dlp) == (dcp)[(idx)])) { \ - (lp)[0]++; \ - } else { \ - (dlp) = (dcp)[idx]; \ - (lp) = (pn) + (idx); \ - (lp)[0] = 1; \ - } \ -} while (0) - -static const struct { - uint64_t pnum; /* prebuild 4 values for pnum[]. */ - int32_t idx; /* index for new last updated element. */ - uint16_t lpv; /* add value to the last updated element. */ -} gptbl[GRPSZ] = { - { - /* 0: a != b, b != c, c != d, d != e */ - .pnum = UINT64_C(0x0001000100010001), - .idx = 4, - .lpv = 0, - }, - { - /* 1: a == b, b != c, c != d, d != e */ - .pnum = UINT64_C(0x0001000100010002), - .idx = 4, - .lpv = 1, - }, - { - /* 2: a != b, b == c, c != d, d != e */ - .pnum = UINT64_C(0x0001000100020001), - .idx = 4, - .lpv = 0, - }, - { - /* 3: a == b, b == c, c != d, d != e */ - .pnum = UINT64_C(0x0001000100020003), - .idx = 4, - .lpv = 2, - }, - { - /* 4: a != b, b != c, c == d, d != e */ - .pnum = UINT64_C(0x0001000200010001), - .idx = 4, - .lpv = 0, - }, - { - /* 5: a == b, b != c, c == d, d != e */ - .pnum = UINT64_C(0x0001000200010002), - .idx = 4, - .lpv = 1, - }, - { - /* 6: a != b, b == c, c == d, d != e */ - .pnum = UINT64_C(0x0001000200030001), - .idx = 4, - .lpv = 0, - }, - { - /* 7: a == b, b == c, c == d, d != e */ - .pnum = UINT64_C(0x0001000200030004), - .idx = 4, - .lpv = 3, - }, - { - /* 8: a != b, b != c, c != d, d == e */ - .pnum = UINT64_C(0x0002000100010001), - .idx = 3, - .lpv = 0, - }, - { - /* 9: a == b, b != c, c != d, d == e */ - .pnum = UINT64_C(0x0002000100010002), - .idx = 3, - .lpv = 1, - }, - { - /* 0xa: a != b, b == c, c != d, d == e */ - .pnum = UINT64_C(0x0002000100020001), - .idx = 3, - .lpv = 0, - }, - { - /* 0xb: a == b, b == c, c != d, d == e */ - .pnum = UINT64_C(0x0002000100020003), - .idx = 3, - .lpv = 2, - }, - { - /* 0xc: a != b, b != c, c == d, d == e */ - .pnum = UINT64_C(0x0002000300010001), - .idx = 2, - .lpv = 0, - }, - { - /* 0xd: a == b, b != c, c == d, d == e */ - .pnum = UINT64_C(0x0002000300010002), - .idx = 2, - .lpv = 1, - }, - { - /* 0xe: a != b, b == c, c == d, d == e */ - .pnum = UINT64_C(0x0002000300040001), - .idx = 1, - .lpv = 0, - }, - { - /* 0xf: a == b, b == c, c == d, d == e */ - .pnum = UINT64_C(0x0002000300040005), - .idx = 0, - .lpv = 4, - }, -}; - static __rte_always_inline void send_packetsx4(struct lcore_conf *qconf, uint16_t port, struct rte_mbuf*m[],uint32_t num) diff --git a/examples/l3fwd/l3fwd_neon.h b/examples/l3fwd/l3fwd_neon.h index e3d33a5229..5fa765b640 100644 --- a/examples/l3fwd/l3fwd_neon.h +++ b/examples/l3fwd/l3fwd_neon.h @@ -7,6 +7,7 @@ #define _L3FWD_NEON_H_ #include "l3fwd.h" +#include "neon_common.h" #include "l3fwd_common.h" /* @@ -62,44 +63,6 @@ processx4_step3(struct rte_mbuf *pkt[FWDSTEP], uint16_t dst_port[FWDSTEP]) &dst_port[3], pkt[3]->packet_type); } -/* - * Group consecutive packets with the same destination port in bursts of 4. - * Suppose we have array of destination ports: - * dst_port[] = {a, b, c, d,, e, ... } - * dp1 should contain: <a, b, c, d>, dp2: <b, c, d, e>. - * We doing 4 comparisons at once and the result is 4 bit mask. - * This mask is used as an index into prebuild array of pnum values. - */ -static inline uint16_t * -port_groupx4(uint16_t pn[FWDSTEP + 1], uint16_t *lp, uint16x8_t dp1, - uint16x8_t dp2) -{ - union { - uint16_t u16[FWDSTEP + 1]; - uint64_t u64; - } *pnum = (void *)pn; - - int32_t v; - uint16x8_t mask = {1, 2, 4, 8, 0, 0, 0, 0}; - - dp1 = vceqq_u16(dp1, dp2); - dp1 = vandq_u16(dp1, mask); - v = vaddvq_u16(dp1); - - /* update last port counter. */ - lp[0] += gptbl[v].lpv; - rte_compiler_barrier(); - - /* if dest port value has changed. */ - if (v != GRPMSK) { - pnum->u64 = gptbl[v].pnum; - pnum->u16[FWDSTEP] = 1; - lp = pnum->u16 + gptbl[v].idx; - } - - return lp; -} - /** * Process one packet: * Update source and destination MAC addresses in the ethernet header. @@ -161,7 +124,7 @@ send_packets_multi(struct lcore_conf *qconf, struct rte_mbuf **pkts_burst, * <d[j-3], d[j-2], d[j-1], d[j], ... > */ dp2 = vld1q_u16(&dst_port[j - FWDSTEP + 1]); - lp = port_groupx4(&pnum[j - FWDSTEP], lp, dp1, dp2); + lp = neon_port_groupx4(&pnum[j - FWDSTEP], lp, dp1, dp2); /* * dp1: @@ -175,7 +138,7 @@ send_packets_multi(struct lcore_conf *qconf, struct rte_mbuf **pkts_burst, */ dp2 = vextq_u16(dp1, dp1, 1); dp2 = vsetq_lane_u16(vgetq_lane_u16(dp2, 2), dp2, 3); - lp = port_groupx4(&pnum[j - FWDSTEP], lp, dp1, dp2); + lp = neon_port_groupx4(&pnum[j - FWDSTEP], lp, dp1, dp2); /* * remove values added by the last repeated diff --git a/examples/meson.build b/examples/meson.build index 78de0e1f37..81e93799f2 100644 --- a/examples/meson.build +++ b/examples/meson.build @@ -97,7 +97,7 @@ foreach example: examples ldflags = default_ldflags ext_deps = [] - includes = [include_directories(example)] + includes = [include_directories(example, 'common')] deps = ['eal', 'mempool', 'net', 'mbuf', 'ethdev', 'cmdline'] subdir(example) -- 2.25.1
