[v12,2/3] drivers/common: add diagnostics macros to make code portable

Message ID 1736915238-779-3-git-send-email-andremue@linux.microsoft.com (mailing list archive)
State Superseded
Delegated to: Thomas Monjalon
Headers
Series add diagnostics macros to make code portable |

Checks

Context Check Description
ci/checkpatch success coding style OK

Commit Message

Andre Muezerie Jan. 15, 2025, 4:27 a.m. UTC
It was a common pattern to have "GCC diagnostic ignored" pragmas
sprinkled over the code and only activate these pragmas for certain
compilers (gcc and clang). Clang supports GCC's pragma for
compatibility with existing source code, so #pragma GCC diagnostic
and #pragma clang diagnostic are synonyms for Clang
(https://clang.llvm.org/docs/UsersManual.html).

Now that effort is being made to make the code compatible with MSVC
these expressions would become more complex. It makes sense to hide
this complexity behind macros. This makes maintenance easier as these
macros are defined in a single place. As a plus the code becomes
more readable as well.

Signed-off-by: Andre Muezerie <andremue@linux.microsoft.com>
---
 drivers/common/idpf/idpf_common_rxtx_avx512.c | 77 ++++++++++---------
 1 file changed, 39 insertions(+), 38 deletions(-)
  

Comments

Bruce Richardson Jan. 15, 2025, 11:13 a.m. UTC | #1
On Tue, Jan 14, 2025 at 08:27:17PM -0800, Andre Muezerie wrote:
> It was a common pattern to have "GCC diagnostic ignored" pragmas
> sprinkled over the code and only activate these pragmas for certain
> compilers (gcc and clang). Clang supports GCC's pragma for
> compatibility with existing source code, so #pragma GCC diagnostic
> and #pragma clang diagnostic are synonyms for Clang
> (https://clang.llvm.org/docs/UsersManual.html).
> 
> Now that effort is being made to make the code compatible with MSVC
> these expressions would become more complex. It makes sense to hide
> this complexity behind macros. This makes maintenance easier as these
> macros are defined in a single place. As a plus the code becomes
> more readable as well.
> 
> Signed-off-by: Andre Muezerie <andremue@linux.microsoft.com>
> ---
>  drivers/common/idpf/idpf_common_rxtx_avx512.c | 77 ++++++++++---------
>  1 file changed, 39 insertions(+), 38 deletions(-)
> 
> diff --git a/drivers/common/idpf/idpf_common_rxtx_avx512.c b/drivers/common/idpf/idpf_common_rxtx_avx512.c
> index b8450b03ae..fefc0a05ca 100644
> --- a/drivers/common/idpf/idpf_common_rxtx_avx512.c
> +++ b/drivers/common/idpf/idpf_common_rxtx_avx512.c
> @@ -6,10 +6,6 @@
>  #include "idpf_common_device.h"
>  #include "idpf_common_rxtx.h"
>  
> -#ifndef __INTEL_COMPILER
> -#pragma GCC diagnostic ignored "-Wcast-qual"
> -#endif
> -
>  #define IDPF_DESCS_PER_LOOP_AVX 8
>  #define PKTLEN_SHIFT 10
>  
> @@ -34,7 +30,7 @@ idpf_singleq_rearm_common(struct idpf_rx_queue *rxq)
>  			dma_addr0 = _mm_setzero_si128();
>  			for (i = 0; i < IDPF_VPMD_DESCS_PER_LOOP; i++) {
>  				rxp[i] = &rxq->fake_mbuf;
> -				_mm_store_si128((__m128i *)&rxdp[i].read,
> +				_mm_store_si128((__m128i *)RTE_IGNORE_CAST_QUAL(&rxdp[i].read),
>  						dma_addr0);
>  			}
>  		}

<SNIP for brevity>

Adding in the (void *) to the IGNORE macro will hugely reduce the amount of
casting being done in this file!

/Bruce
  

Patch

diff --git a/drivers/common/idpf/idpf_common_rxtx_avx512.c b/drivers/common/idpf/idpf_common_rxtx_avx512.c
index b8450b03ae..fefc0a05ca 100644
--- a/drivers/common/idpf/idpf_common_rxtx_avx512.c
+++ b/drivers/common/idpf/idpf_common_rxtx_avx512.c
@@ -6,10 +6,6 @@ 
 #include "idpf_common_device.h"
 #include "idpf_common_rxtx.h"
 
-#ifndef __INTEL_COMPILER
-#pragma GCC diagnostic ignored "-Wcast-qual"
-#endif
-
 #define IDPF_DESCS_PER_LOOP_AVX 8
 #define PKTLEN_SHIFT 10
 
@@ -34,7 +30,7 @@  idpf_singleq_rearm_common(struct idpf_rx_queue *rxq)
 			dma_addr0 = _mm_setzero_si128();
 			for (i = 0; i < IDPF_VPMD_DESCS_PER_LOOP; i++) {
 				rxp[i] = &rxq->fake_mbuf;
-				_mm_store_si128((__m128i *)&rxdp[i].read,
+				_mm_store_si128((__m128i *)RTE_IGNORE_CAST_QUAL(&rxdp[i].read),
 						dma_addr0);
 			}
 		}
@@ -108,8 +104,10 @@  idpf_singleq_rearm_common(struct idpf_rx_queue *rxq)
 		dma_addr4_7 = _mm512_add_epi64(dma_addr4_7, hdr_room);
 
 		/* flush desc with pa dma_addr */
-		_mm512_store_si512((__m512i *)&rxdp->read, dma_addr0_3);
-		_mm512_store_si512((__m512i *)&(rxdp + 4)->read, dma_addr4_7);
+		_mm512_store_si512((__m512i *)RTE_IGNORE_CAST_QUAL(&rxdp->read),
+				dma_addr0_3);
+		_mm512_store_si512((__m512i *)RTE_IGNORE_CAST_QUAL(&(rxdp + 4)->read),
+				dma_addr4_7);
 	}
 
 	rxq->rxrearm_start += IDPF_RXQ_REARM_THRESH;
@@ -164,8 +162,8 @@  idpf_singleq_rearm(struct idpf_rx_queue *rxq)
 				dma_addr0 = _mm_setzero_si128();
 				for (i = 0; i < IDPF_VPMD_DESCS_PER_LOOP; i++) {
 					rxp[i] = &rxq->fake_mbuf;
-					_mm_storeu_si128((__m128i *)&rxdp[i].read,
-							 dma_addr0);
+					_mm_storeu_si128((__m128i *)RTE_IGNORE_CAST_QUAL
+							(&rxdp[i].read), dma_addr0);
 				}
 			}
 			rte_atomic_fetch_add_explicit(&rxq->rx_stats.mbuf_alloc_failed,
@@ -216,10 +214,10 @@  idpf_singleq_rearm(struct idpf_rx_queue *rxq)
 				 iovas1);
 		const __m512i desc6_7 = _mm512_bsrli_epi128(desc4_5, 8);
 
-		_mm512_storeu_si512((void *)rxdp, desc0_1);
-		_mm512_storeu_si512((void *)(rxdp + 2), desc2_3);
-		_mm512_storeu_si512((void *)(rxdp + 4), desc4_5);
-		_mm512_storeu_si512((void *)(rxdp + 6), desc6_7);
+		_mm512_storeu_si512((void *)RTE_IGNORE_CAST_QUAL(rxdp), desc0_1);
+		_mm512_storeu_si512((void *)RTE_IGNORE_CAST_QUAL((rxdp + 2)), desc2_3);
+		_mm512_storeu_si512((void *)RTE_IGNORE_CAST_QUAL((rxdp + 4)), desc4_5);
+		_mm512_storeu_si512((void *)RTE_IGNORE_CAST_QUAL((rxdp + 6)), desc6_7);
 
 		rxp += IDPF_DESCS_PER_LOOP_AVX;
 		rxdp += IDPF_DESCS_PER_LOOP_AVX;
@@ -337,28 +335,28 @@  _idpf_singleq_recv_raw_pkts_avx512(struct idpf_rx_queue *rxq,
 
 		__m512i raw_desc0_3, raw_desc4_7;
 		const __m128i raw_desc7 =
-			_mm_load_si128((void *)(rxdp + 7));
+			_mm_load_si128((void *)RTE_IGNORE_CAST_QUAL(rxdp + 7));
 		rte_compiler_barrier();
 		const __m128i raw_desc6 =
-			_mm_load_si128((void *)(rxdp + 6));
+			_mm_load_si128((void *)RTE_IGNORE_CAST_QUAL(rxdp + 6));
 		rte_compiler_barrier();
 		const __m128i raw_desc5 =
-			_mm_load_si128((void *)(rxdp + 5));
+			_mm_load_si128((void *)RTE_IGNORE_CAST_QUAL(rxdp + 5));
 		rte_compiler_barrier();
 		const __m128i raw_desc4 =
-			_mm_load_si128((void *)(rxdp + 4));
+			_mm_load_si128((void *)RTE_IGNORE_CAST_QUAL(rxdp + 4));
 		rte_compiler_barrier();
 		const __m128i raw_desc3 =
-			_mm_load_si128((void *)(rxdp + 3));
+			_mm_load_si128((void *)RTE_IGNORE_CAST_QUAL(rxdp + 3));
 		rte_compiler_barrier();
 		const __m128i raw_desc2 =
-			_mm_load_si128((void *)(rxdp + 2));
+			_mm_load_si128((void *)RTE_IGNORE_CAST_QUAL(rxdp + 2));
 		rte_compiler_barrier();
 		const __m128i raw_desc1 =
-			_mm_load_si128((void *)(rxdp + 1));
+			_mm_load_si128((void *)RTE_IGNORE_CAST_QUAL(rxdp + 1));
 		rte_compiler_barrier();
 		const __m128i raw_desc0 =
-			_mm_load_si128((void *)(rxdp + 0));
+			_mm_load_si128((void *)RTE_IGNORE_CAST_QUAL(rxdp + 0));
 
 		raw_desc4_7 = _mm512_broadcast_i32x4(raw_desc4);
 		raw_desc4_7 = _mm512_inserti32x4(raw_desc4_7, raw_desc5, 1);
@@ -560,7 +558,7 @@  idpf_splitq_rearm_common(struct idpf_rx_queue *rx_bufq)
 			dma_addr0 = _mm_setzero_si128();
 			for (i = 0; i < IDPF_VPMD_DESCS_PER_LOOP; i++) {
 				rxp[i] = &rx_bufq->fake_mbuf;
-				_mm_store_si128((__m128i *)&rxdp[i],
+				_mm_store_si128((__m128i *)RTE_IGNORE_CAST_QUAL(&rxdp[i]),
 						dma_addr0);
 			}
 		}
@@ -634,7 +632,7 @@  idpf_splitq_rearm(struct idpf_rx_queue *rx_bufq)
 				dma_addr0 = _mm_setzero_si128();
 				for (i = 0; i < IDPF_VPMD_DESCS_PER_LOOP; i++) {
 					rxp[i] = &rx_bufq->fake_mbuf;
-					_mm_storeu_si128((__m128i *)&rxdp[i],
+					_mm_storeu_si128((__m128i *)RTE_IGNORE_CAST_QUAL(&rxdp[i]),
 							 dma_addr0);
 				}
 			}
@@ -798,28 +796,28 @@  _idpf_splitq_recv_raw_pkts_avx512(struct idpf_rx_queue *rxq,
 
 		__m512i raw_desc0_3, raw_desc4_7;
 		const __m128i raw_desc7 =
-			_mm_load_si128((void *)(rxdp + 7));
+			_mm_load_si128((void *)RTE_IGNORE_CAST_QUAL(rxdp + 7));
 		rte_compiler_barrier();
 		const __m128i raw_desc6 =
-			_mm_load_si128((void *)(rxdp + 6));
+			_mm_load_si128((void *)RTE_IGNORE_CAST_QUAL(rxdp + 6));
 		rte_compiler_barrier();
 		const __m128i raw_desc5 =
-			_mm_load_si128((void *)(rxdp + 5));
+			_mm_load_si128((void *)RTE_IGNORE_CAST_QUAL(rxdp + 5));
 		rte_compiler_barrier();
 		const __m128i raw_desc4 =
-			_mm_load_si128((void *)(rxdp + 4));
+			_mm_load_si128((void *)RTE_IGNORE_CAST_QUAL(rxdp + 4));
 		rte_compiler_barrier();
 		const __m128i raw_desc3 =
-			_mm_load_si128((void *)(rxdp + 3));
+			_mm_load_si128((void *)RTE_IGNORE_CAST_QUAL(rxdp + 3));
 		rte_compiler_barrier();
 		const __m128i raw_desc2 =
-			_mm_load_si128((void *)(rxdp + 2));
+			_mm_load_si128((void *)RTE_IGNORE_CAST_QUAL(rxdp + 2));
 		rte_compiler_barrier();
 		const __m128i raw_desc1 =
-			_mm_load_si128((void *)(rxdp + 1));
+			_mm_load_si128((void *)RTE_IGNORE_CAST_QUAL(rxdp + 1));
 		rte_compiler_barrier();
 		const __m128i raw_desc0 =
-			_mm_load_si128((void *)(rxdp + 0));
+			_mm_load_si128((void *)RTE_IGNORE_CAST_QUAL(rxdp + 0));
 
 		raw_desc4_7 = _mm512_broadcast_i32x4(raw_desc4);
 		raw_desc4_7 = _mm512_inserti32x4(raw_desc4_7, raw_desc5, 1);
@@ -1131,7 +1129,7 @@  idpf_singleq_vtx1(volatile struct idpf_base_tx_desc *txdp,
 
 	__m128i descriptor = _mm_set_epi64x(high_qw,
 					    pkt->buf_iova + pkt->data_off);
-	_mm_storeu_si128((__m128i *)txdp, descriptor);
+	_mm_storeu_si128((__m128i *)RTE_IGNORE_CAST_QUAL(txdp), descriptor);
 }
 
 #define IDPF_TX_LEN_MASK 0xAA
@@ -1178,7 +1176,7 @@  idpf_singleq_vtx(volatile struct idpf_base_tx_desc *txdp,
 				 pkt[1]->buf_iova + pkt[1]->data_off,
 				 hi_qw0,
 				 pkt[0]->buf_iova + pkt[0]->data_off);
-		_mm512_storeu_si512((void *)txdp, desc0_3);
+		_mm512_storeu_si512((void *)RTE_IGNORE_CAST_QUAL(txdp), desc0_3);
 	}
 
 	/* do any last ones */
@@ -1435,7 +1433,7 @@  idpf_splitq_vtx1(volatile struct idpf_flex_tx_sched_desc *txdp,
 
 	__m128i descriptor = _mm_set_epi64x(high_qw,
 					    pkt->buf_iova + pkt->data_off);
-	_mm_storeu_si128((__m128i *)txdp, descriptor);
+	_mm_storeu_si128((__m128i *)RTE_IGNORE_CAST_QUAL(txdp), descriptor);
 }
 
 static __rte_always_inline void
@@ -1480,7 +1478,7 @@  idpf_splitq_vtx(volatile struct idpf_flex_tx_sched_desc *txdp,
 				 pkt[1]->buf_iova + pkt[1]->data_off,
 				 hi_qw0,
 				 pkt[0]->buf_iova + pkt[0]->data_off);
-		_mm512_storeu_si512((void *)txdp, desc0_3);
+		_mm512_storeu_si512((void *)RTE_IGNORE_CAST_QUAL(txdp), desc0_3);
 	}
 
 	/* do any last ones */
@@ -1521,11 +1519,13 @@  idpf_splitq_xmit_fixed_burst_vec_avx512(void *tx_queue, struct rte_mbuf **tx_pkt
 	if (nb_commit >= n) {
 		tx_backlog_entry_avx512(txep, tx_pkts, n);
 
-		idpf_splitq_vtx((void *)txdp, tx_pkts, n - 1, cmd_dtype);
+		idpf_splitq_vtx((void *)RTE_IGNORE_CAST_QUAL(txdp), tx_pkts, n - 1,
+				cmd_dtype);
 		tx_pkts += (n - 1);
 		txdp += (n - 1);
 
-		idpf_splitq_vtx1((void *)txdp, *tx_pkts++, cmd_dtype);
+		idpf_splitq_vtx1((void *)RTE_IGNORE_CAST_QUAL(txdp), *tx_pkts++,
+				cmd_dtype);
 
 		nb_commit = (uint16_t)(nb_commit - n);
 
@@ -1540,7 +1540,8 @@  idpf_splitq_xmit_fixed_burst_vec_avx512(void *tx_queue, struct rte_mbuf **tx_pkt
 
 	tx_backlog_entry_avx512(txep, tx_pkts, nb_commit);
 
-	idpf_splitq_vtx((void *)txdp, tx_pkts, nb_commit, cmd_dtype);
+	idpf_splitq_vtx((void *)RTE_IGNORE_CAST_QUAL(txdp), tx_pkts, nb_commit,
+			cmd_dtype);
 
 	tx_id = (uint16_t)(tx_id + nb_commit);
 	if (tx_id > txq->next_rs)