DPDK patches and discussions
 help / color / mirror / Atom feed
From: Anoob Joseph <anoobj@marvell.com>
To: Akhil Goyal <akhil.goyal@nxp.com>,
	Pablo de Lara <pablo.de.lara.guarch@intel.com>
Cc: Jerin Jacob Kollanukkaran <jerinj@marvell.com>,
	"Narayana Prasad Raju Athreya" <pathreya@marvell.com>,
	"dev@dpdk.org" <dev@dpdk.org>
Subject: Re: [dpdk-dev] [PATCH 1/2] common/cpt: remove redundant bit swaps
Date: Sun, 7 Jul 2019 14:10:40 +0000	[thread overview]
Message-ID: <MN2PR18MB287735BEF74BE1662D4EBDBEDFF70@MN2PR18MB2877.namprd18.prod.outlook.com> (raw)
In-Reply-To: <1562419420-30278-1-git-send-email-anoobj@marvell.com>

Hi Akhil, Pablo

This patch is good to go if you don't have any comments.

Thanks,
Anoob

> -----Original Message-----
> From: Anoob Joseph <anoobj@marvell.com>
> Sent: Saturday, July 6, 2019 6:54 PM
> To: Akhil Goyal <akhil.goyal@nxp.com>; Pablo de Lara
> <pablo.de.lara.guarch@intel.com>
> Cc: Anoob Joseph <anoobj@marvell.com>; Jerin Jacob Kollanukkaran
> <jerinj@marvell.com>; Narayana Prasad Raju Athreya
> <pathreya@marvell.com>; dev@dpdk.org
> Subject: [PATCH 1/2] common/cpt: remove redundant bit swaps
> 
> The bit swaps can be removed by re-arranging the structure.
> 
> Signed-off-by: Anoob Joseph <anoobj@marvell.com>
> ---
>  drivers/common/cpt/cpt_hw_types.h |   7 +++
>  drivers/common/cpt/cpt_ucode.h    | 116 ++++++++++++--------------------------
>  2 files changed, 44 insertions(+), 79 deletions(-)
> 
> diff --git a/drivers/common/cpt/cpt_hw_types.h
> b/drivers/common/cpt/cpt_hw_types.h
> index 7be1d12..e2b127d 100644
> --- a/drivers/common/cpt/cpt_hw_types.h
> +++ b/drivers/common/cpt/cpt_hw_types.h
> @@ -30,10 +30,17 @@
>  typedef union {
>  	uint64_t u64;
>  	struct {
> +#if RTE_BYTE_ORDER == RTE_BIG_ENDIAN
>  		uint16_t opcode;
>  		uint16_t param1;
>  		uint16_t param2;
>  		uint16_t dlen;
> +#else
> +		uint16_t dlen;
> +		uint16_t param2;
> +		uint16_t param1;
> +		uint16_t opcode;
> +#endif
>  	} s;
>  } vq_cmd_word0_t;
> 
> diff --git a/drivers/common/cpt/cpt_ucode.h
> b/drivers/common/cpt/cpt_ucode.h index e02b34a..c589b58 100644
> --- a/drivers/common/cpt/cpt_ucode.h
> +++ b/drivers/common/cpt/cpt_ucode.h
> @@ -520,16 +520,15 @@ cpt_digest_gen_prep(uint32_t flags,
> 
>  	/*GP op header */
>  	vq_cmd_w0.u64 = 0;
> -	vq_cmd_w0.s.param2 = rte_cpu_to_be_16(((uint16_t)hash_type << 8));
> +	vq_cmd_w0.s.param2 = ((uint16_t)hash_type << 8);
>  	if (ctx->hmac) {
>  		opcode.s.major = CPT_MAJOR_OP_HMAC | CPT_DMA_MODE;
> -		vq_cmd_w0.s.param1 = rte_cpu_to_be_16(key_len);
> -		vq_cmd_w0.s.dlen =
> -			rte_cpu_to_be_16((data_len + ROUNDUP8(key_len)));
> +		vq_cmd_w0.s.param1 = key_len;
> +		vq_cmd_w0.s.dlen = data_len + ROUNDUP8(key_len);
>  	} else {
>  		opcode.s.major = CPT_MAJOR_OP_HASH | CPT_DMA_MODE;
>  		vq_cmd_w0.s.param1 = 0;
> -		vq_cmd_w0.s.dlen = rte_cpu_to_be_16(data_len);
> +		vq_cmd_w0.s.dlen = data_len;
>  	}
> 
>  	opcode.s.minor = 0;
> @@ -540,10 +539,10 @@ cpt_digest_gen_prep(uint32_t flags,
>  		/* Minor op is passthrough */
>  		opcode.s.minor = 0x03;
>  		/* Send out completion code only */
> -		vq_cmd_w0.s.param2 = rte_cpu_to_be_16(0x1);
> +		vq_cmd_w0.s.param2 = 0x1;
>  	}
> 
> -	vq_cmd_w0.s.opcode = rte_cpu_to_be_16(opcode.flags);
> +	vq_cmd_w0.s.opcode = opcode.flags;
> 
>  	/* DPTR has SG list */
>  	in_buffer = m_vaddr;
> @@ -622,7 +621,7 @@ cpt_digest_gen_prep(uint32_t flags,
>  	size = g_size_bytes + s_size_bytes + SG_LIST_HDR_SIZE;
> 
>  	/* This is DPTR len incase of SG mode */
> -	vq_cmd_w0.s.dlen = rte_cpu_to_be_16(size);
> +	vq_cmd_w0.s.dlen = size;
> 
>  	m_vaddr = (uint8_t *)m_vaddr + size;
>  	m_dma += size;
> @@ -635,11 +634,6 @@ cpt_digest_gen_prep(uint32_t flags,
> 
>  	req->ist.ei1 = dptr_dma;
>  	req->ist.ei2 = rptr_dma;
> -	/* First 16-bit swap then 64-bit swap */
> -	/* TODO: HACK: Reverse the vq_cmd and cpt_req bit field definitions
> -	 * to eliminate all the swapping
> -	 */
> -	vq_cmd_w0.u64 = rte_cpu_to_be_64(vq_cmd_w0.u64);
> 
>  	/* vq command w3 */
>  	vq_cmd_w3.u64 = 0;
> @@ -798,8 +792,8 @@ cpt_enc_hmac_prep(uint32_t flags,
> 
>  	/* GP op header */
>  	vq_cmd_w0.u64 = 0;
> -	vq_cmd_w0.s.param1 = rte_cpu_to_be_16(encr_data_len);
> -	vq_cmd_w0.s.param2 = rte_cpu_to_be_16(auth_data_len);
> +	vq_cmd_w0.s.param1 = encr_data_len;
> +	vq_cmd_w0.s.param2 = auth_data_len;
>  	/*
>  	 * In 83XX since we have a limitation of
>  	 * IV & Offset control word not part of instruction @@ -826,9 +820,9
> @@ cpt_enc_hmac_prep(uint32_t flags,
>  		req->alternate_caddr = (uint64_t *)((uint8_t *)dm_vaddr
>  						    + outputlen - iv_len);
> 
> -		vq_cmd_w0.s.dlen = rte_cpu_to_be_16(inputlen +
> OFF_CTRL_LEN);
> +		vq_cmd_w0.s.dlen = inputlen + OFF_CTRL_LEN;
> 
> -		vq_cmd_w0.s.opcode = rte_cpu_to_be_16(opcode.flags);
> +		vq_cmd_w0.s.opcode = opcode.flags;
> 
>  		if (likely(iv_len)) {
>  			uint64_t *dest = (uint64_t *)((uint8_t *)offset_vaddr
> @@ -861,7 +855,7 @@ cpt_enc_hmac_prep(uint32_t flags,
> 
>  		opcode.s.major |= CPT_DMA_MODE;
> 
> -		vq_cmd_w0.s.opcode = rte_cpu_to_be_16(opcode.flags);
> +		vq_cmd_w0.s.opcode = opcode.flags;
> 
>  		if (likely(iv_len)) {
>  			uint64_t *dest = (uint64_t *)((uint8_t *)offset_vaddr
> @@ -1005,7 +999,7 @@ cpt_enc_hmac_prep(uint32_t flags,
>  		size = g_size_bytes + s_size_bytes + SG_LIST_HDR_SIZE;
> 
>  		/* This is DPTR len incase of SG mode */
> -		vq_cmd_w0.s.dlen = rte_cpu_to_be_16(size);
> +		vq_cmd_w0.s.dlen = size;
> 
>  		m_vaddr = (uint8_t *)m_vaddr + size;
>  		m_dma += size;
> @@ -1020,12 +1014,6 @@ cpt_enc_hmac_prep(uint32_t flags,
>  		req->ist.ei2 = rptr_dma;
>  	}
> 
> -	/* First 16-bit swap then 64-bit swap */
> -	/* TODO: HACK: Reverse the vq_cmd and cpt_req bit field definitions
> -	 * to eliminate all the swapping
> -	 */
> -	vq_cmd_w0.u64 = rte_cpu_to_be_64(vq_cmd_w0.u64);
> -
>  	ctx_dma = fc_params->ctx_buf.dma_addr +
>  		offsetof(struct cpt_ctx, fctx);
>  	/* vq command w3 */
> @@ -1175,8 +1163,8 @@ cpt_dec_hmac_prep(uint32_t flags,
>  		encr_offset = inputlen;
> 
>  	vq_cmd_w0.u64 = 0;
> -	vq_cmd_w0.s.param1 = rte_cpu_to_be_16(encr_data_len);
> -	vq_cmd_w0.s.param2 = rte_cpu_to_be_16(auth_data_len);
> +	vq_cmd_w0.s.param1 = encr_data_len;
> +	vq_cmd_w0.s.param2 = auth_data_len;
> 
>  	/*
>  	 * In 83XX since we have a limitation of @@ -1209,9 +1197,9 @@
> cpt_dec_hmac_prep(uint32_t flags,
>  		 * hmac.
>  		 */
> 
> -		vq_cmd_w0.s.dlen = rte_cpu_to_be_16(inputlen +
> OFF_CTRL_LEN);
> +		vq_cmd_w0.s.dlen = inputlen + OFF_CTRL_LEN;
> 
> -		vq_cmd_w0.s.opcode = rte_cpu_to_be_16(opcode.flags);
> +		vq_cmd_w0.s.opcode = opcode.flags;
> 
>  		if (likely(iv_len)) {
>  			uint64_t *dest = (uint64_t *)((uint8_t *)offset_vaddr +
> @@ -1245,7 +1233,7 @@ cpt_dec_hmac_prep(uint32_t flags,
> 
>  		opcode.s.major |= CPT_DMA_MODE;
> 
> -		vq_cmd_w0.s.opcode = rte_cpu_to_be_16(opcode.flags);
> +		vq_cmd_w0.s.opcode = opcode.flags;
> 
>  		if (likely(iv_len)) {
>  			uint64_t *dest = (uint64_t *)((uint8_t *)offset_vaddr +
> @@ -1401,7 +1389,7 @@ cpt_dec_hmac_prep(uint32_t flags,
>  		size = g_size_bytes + s_size_bytes + SG_LIST_HDR_SIZE;
> 
>  		/* This is DPTR len incase of SG mode */
> -		vq_cmd_w0.s.dlen = rte_cpu_to_be_16(size);
> +		vq_cmd_w0.s.dlen = size;
> 
>  		m_vaddr = (uint8_t *)m_vaddr + size;
>  		m_dma += size;
> @@ -1417,12 +1405,6 @@ cpt_dec_hmac_prep(uint32_t flags,
>  		req->ist.ei2 = rptr_dma;
>  	}
> 
> -	/* First 16-bit swap then 64-bit swap */
> -	/* TODO: HACK: Reverse the vq_cmd and cpt_req bit field definitions
> -	 * to eliminate all the swapping
> -	 */
> -	vq_cmd_w0.u64 = rte_cpu_to_be_64(vq_cmd_w0.u64);
> -
>  	ctx_dma = fc_params->ctx_buf.dma_addr +
>  		offsetof(struct cpt_ctx, fctx);
>  	/* vq command w3 */
> @@ -1579,8 +1561,8 @@ cpt_zuc_snow3g_enc_prep(uint32_t req_flags,
>  	 * GP op header, lengths are expected in bits.
>  	 */
>  	vq_cmd_w0.u64 = 0;
> -	vq_cmd_w0.s.param1 = rte_cpu_to_be_16(encr_data_len);
> -	vq_cmd_w0.s.param2 = rte_cpu_to_be_16(auth_data_len);
> +	vq_cmd_w0.s.param1 = encr_data_len;
> +	vq_cmd_w0.s.param2 = auth_data_len;
> 
>  	/*
>  	 * In 83XX since we have a limitation of @@ -1609,9 +1591,9 @@
> cpt_zuc_snow3g_enc_prep(uint32_t req_flags,
>  		req->alternate_caddr = (uint64_t *)((uint8_t *)dm_vaddr
>  						    + outputlen - iv_len);
> 
> -		vq_cmd_w0.s.dlen = rte_cpu_to_be_16(inputlen +
> OFF_CTRL_LEN);
> +		vq_cmd_w0.s.dlen = inputlen + OFF_CTRL_LEN;
> 
> -		vq_cmd_w0.s.opcode = rte_cpu_to_be_16(opcode.flags);
> +		vq_cmd_w0.s.opcode = opcode.flags;
> 
>  		if (likely(iv_len)) {
>  			uint32_t *iv_d = (uint32_t *)((uint8_t *)offset_vaddr
> @@ -1638,7 +1620,7 @@ cpt_zuc_snow3g_enc_prep(uint32_t req_flags,
> 
>  		opcode.s.major |= CPT_DMA_MODE;
> 
> -		vq_cmd_w0.s.opcode = rte_cpu_to_be_16(opcode.flags);
> +		vq_cmd_w0.s.opcode = opcode.flags;
> 
>  		/* DPTR has SG list */
>  		in_buffer = m_vaddr;
> @@ -1740,7 +1722,7 @@ cpt_zuc_snow3g_enc_prep(uint32_t req_flags,
>  		size = g_size_bytes + s_size_bytes + SG_LIST_HDR_SIZE;
> 
>  		/* This is DPTR len incase of SG mode */
> -		vq_cmd_w0.s.dlen = rte_cpu_to_be_16(size);
> +		vq_cmd_w0.s.dlen = size;
> 
>  		m_vaddr = (uint8_t *)m_vaddr + size;
>  		m_dma += size;
> @@ -1755,12 +1737,6 @@ cpt_zuc_snow3g_enc_prep(uint32_t req_flags,
>  		req->ist.ei2 = rptr_dma;
>  	}
> 
> -	/* First 16-bit swap then 64-bit swap */
> -	/* TODO: HACK: Reverse the vq_cmd and cpt_req bit field definitions
> -	 * to eliminate all the swapping
> -	 */
> -	vq_cmd_w0.u64 = rte_cpu_to_be_64(vq_cmd_w0.u64);
> -
>  	/* vq command w3 */
>  	vq_cmd_w3.u64 = 0;
>  	vq_cmd_w3.s.grp = 0;
> @@ -1886,7 +1862,7 @@ cpt_zuc_snow3g_dec_prep(uint32_t req_flags,
>  	 * GP op header, lengths are expected in bits.
>  	 */
>  	vq_cmd_w0.u64 = 0;
> -	vq_cmd_w0.s.param1 = rte_cpu_to_be_16(encr_data_len);
> +	vq_cmd_w0.s.param1 = encr_data_len;
> 
>  	/*
>  	 * In 83XX since we have a limitation of @@ -1915,9 +1891,9 @@
> cpt_zuc_snow3g_dec_prep(uint32_t req_flags,
>  		req->alternate_caddr = (uint64_t *)((uint8_t *)dm_vaddr
>  						    + outputlen - iv_len);
> 
> -		vq_cmd_w0.s.dlen = rte_cpu_to_be_16(inputlen +
> OFF_CTRL_LEN);
> +		vq_cmd_w0.s.dlen = inputlen + OFF_CTRL_LEN;
> 
> -		vq_cmd_w0.s.opcode = rte_cpu_to_be_16(opcode.flags);
> +		vq_cmd_w0.s.opcode = opcode.flags;
> 
>  		if (likely(iv_len)) {
>  			uint32_t *iv_d = (uint32_t *)((uint8_t *)offset_vaddr
> @@ -1945,7 +1921,7 @@ cpt_zuc_snow3g_dec_prep(uint32_t req_flags,
> 
>  		opcode.s.major |= CPT_DMA_MODE;
> 
> -		vq_cmd_w0.s.opcode = rte_cpu_to_be_16(opcode.flags);
> +		vq_cmd_w0.s.opcode = opcode.flags;
> 
>  		/* DPTR has SG list */
>  		in_buffer = m_vaddr;
> @@ -2020,7 +1996,7 @@ cpt_zuc_snow3g_dec_prep(uint32_t req_flags,
>  		size = g_size_bytes + s_size_bytes + SG_LIST_HDR_SIZE;
> 
>  		/* This is DPTR len incase of SG mode */
> -		vq_cmd_w0.s.dlen = rte_cpu_to_be_16(size);
> +		vq_cmd_w0.s.dlen = size;
> 
>  		m_vaddr = (uint8_t *)m_vaddr + size;
>  		m_dma += size;
> @@ -2035,12 +2011,6 @@ cpt_zuc_snow3g_dec_prep(uint32_t req_flags,
>  		req->ist.ei2 = rptr_dma;
>  	}
> 
> -	/* First 16-bit swap then 64-bit swap */
> -	/* TODO: HACK: Reverse the vq_cmd and cpt_req bit field definitions
> -	 * to eliminate all the swapping
> -	 */
> -	vq_cmd_w0.u64 = rte_cpu_to_be_64(vq_cmd_w0.u64);
> -
>  	/* vq command w3 */
>  	vq_cmd_w3.u64 = 0;
>  	vq_cmd_w3.s.grp = 0;
> @@ -2150,9 +2120,9 @@ cpt_kasumi_enc_prep(uint32_t req_flags,
>  	 * GP op header, lengths are expected in bits.
>  	 */
>  	vq_cmd_w0.u64 = 0;
> -	vq_cmd_w0.s.param1 = rte_cpu_to_be_16(encr_data_len);
> -	vq_cmd_w0.s.param2 = rte_cpu_to_be_16(auth_data_len);
> -	vq_cmd_w0.s.opcode = rte_cpu_to_be_16(opcode.flags);
> +	vq_cmd_w0.s.param1 = encr_data_len;
> +	vq_cmd_w0.s.param2 = auth_data_len;
> +	vq_cmd_w0.s.opcode = opcode.flags;
> 
>  	/* consider iv len */
>  	if (flags == 0x0) {
> @@ -2279,7 +2249,7 @@ cpt_kasumi_enc_prep(uint32_t req_flags,
>  	size = g_size_bytes + s_size_bytes + SG_LIST_HDR_SIZE;
> 
>  	/* This is DPTR len incase of SG mode */
> -	vq_cmd_w0.s.dlen = rte_cpu_to_be_16(size);
> +	vq_cmd_w0.s.dlen = size;
> 
>  	m_vaddr = (uint8_t *)m_vaddr + size;
>  	m_dma += size;
> @@ -2293,12 +2263,6 @@ cpt_kasumi_enc_prep(uint32_t req_flags,
>  	req->ist.ei1 = dptr_dma;
>  	req->ist.ei2 = rptr_dma;
> 
> -	/* First 16-bit swap then 64-bit swap */
> -	/* TODO: HACK: Reverse the vq_cmd and cpt_req bit field definitions
> -	 * to eliminate all the swapping
> -	 */
> -	vq_cmd_w0.u64 = rte_cpu_to_be_64(vq_cmd_w0.u64);
> -
>  	/* vq command w3 */
>  	vq_cmd_w3.u64 = 0;
>  	vq_cmd_w3.s.grp = 0;
> @@ -2394,8 +2358,8 @@ cpt_kasumi_dec_prep(uint64_t d_offs,
>  	 * GP op header, lengths are expected in bits.
>  	 */
>  	vq_cmd_w0.u64 = 0;
> -	vq_cmd_w0.s.param1 = rte_cpu_to_be_16(encr_data_len);
> -	vq_cmd_w0.s.opcode = rte_cpu_to_be_16(opcode.flags);
> +	vq_cmd_w0.s.param1 = encr_data_len;
> +	vq_cmd_w0.s.opcode = opcode.flags;
> 
>  	/* consider iv len */
>  	encr_offset += iv_len;
> @@ -2480,7 +2444,7 @@ cpt_kasumi_dec_prep(uint64_t d_offs,
>  	size = g_size_bytes + s_size_bytes + SG_LIST_HDR_SIZE;
> 
>  	/* This is DPTR len incase of SG mode */
> -	vq_cmd_w0.s.dlen = rte_cpu_to_be_16(size);
> +	vq_cmd_w0.s.dlen = size;
> 
>  	m_vaddr = (uint8_t *)m_vaddr + size;
>  	m_dma += size;
> @@ -2494,12 +2458,6 @@ cpt_kasumi_dec_prep(uint64_t d_offs,
>  	req->ist.ei1 = dptr_dma;
>  	req->ist.ei2 = rptr_dma;
> 
> -	/* First 16-bit swap then 64-bit swap */
> -	/* TODO: HACK: Reverse the vq_cmd and cpt_req bit field definitions
> -	 * to eliminate all the swapping
> -	 */
> -	vq_cmd_w0.u64 = rte_cpu_to_be_64(vq_cmd_w0.u64);
> -
>  	/* vq command w3 */
>  	vq_cmd_w3.u64 = 0;
>  	vq_cmd_w3.s.grp = 0;
> --
> 2.7.4


  parent reply	other threads:[~2019-07-07 14:10 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-07-06 13:23 Anoob Joseph
2019-07-06 13:23 ` [dpdk-dev] [PATCH 2/2] common/cpt: remove redundant code in datapath Anoob Joseph
2019-07-07 14:11   ` Anoob Joseph
2019-07-07 14:10 ` Anoob Joseph [this message]
2019-07-14 13:55 ` [dpdk-dev] [PATCH 1/2] common/cpt: remove redundant bit swaps Thomas Monjalon

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=MN2PR18MB287735BEF74BE1662D4EBDBEDFF70@MN2PR18MB2877.namprd18.prod.outlook.com \
    --to=anoobj@marvell.com \
    --cc=akhil.goyal@nxp.com \
    --cc=dev@dpdk.org \
    --cc=jerinj@marvell.com \
    --cc=pablo.de.lara.guarch@intel.com \
    --cc=pathreya@marvell.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).