From patchwork Fri Feb 23 08:05:58 2024 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Li, Pan2" X-Patchwork-Id: 1903118 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@legolas.ozlabs.org Authentication-Results: legolas.ozlabs.org; dkim=pass (2048-bit key; unprotected) header.d=intel.com header.i=@intel.com header.a=rsa-sha256 header.s=Intel header.b=Nfgb5sYv; dkim-atps=neutral Authentication-Results: legolas.ozlabs.org; spf=pass (sender SPF authorized) smtp.mailfrom=gcc.gnu.org (client-ip=2620:52:3:1:0:246e:9693:128c; helo=server2.sourceware.org; envelope-from=gcc-patches-bounces+incoming=patchwork.ozlabs.org@gcc.gnu.org; receiver=patchwork.ozlabs.org) Received: from server2.sourceware.org (server2.sourceware.org [IPv6:2620:52:3:1:0:246e:9693:128c]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange X25519 server-signature ECDSA (secp384r1) server-digest SHA384) (No client certificate requested) by legolas.ozlabs.org (Postfix) with ESMTPS id 4Th2f41yTcz23pm for ; Fri, 23 Feb 2024 19:06:36 +1100 (AEDT) Received: from server2.sourceware.org (localhost [IPv6:::1]) by sourceware.org (Postfix) with ESMTP id 3219B385828D for ; Fri, 23 Feb 2024 08:06:34 +0000 (GMT) X-Original-To: gcc-patches@gcc.gnu.org Delivered-To: gcc-patches@gcc.gnu.org Received: from mgamail.intel.com (mgamail.intel.com [192.198.163.13]) by sourceware.org (Postfix) with ESMTPS id A9E54385840C for ; Fri, 23 Feb 2024 08:06:03 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org A9E54385840C Authentication-Results: sourceware.org; dmarc=pass (p=none dis=none) header.from=intel.com Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=intel.com ARC-Filter: OpenARC Filter v1.0.0 sourceware.org A9E54385840C Authentication-Results: server2.sourceware.org; arc=none smtp.remote-ip=192.198.163.13 ARC-Seal: i=1; a=rsa-sha256; d=sourceware.org; s=key; t=1708675568; cv=none; b=a8+ARtgp3QBaIOqIXkxGwBUJ+6kkKO1lFUGilqTh3V7cX+u89ef5xIntBr1UjYH37LkPnBKpvxPdMb2TPcQMr3ixZF4knRPtxgl4FkGPk5okL1mLRoFuWyqQuvbtbSn5lPLkRW70Og7D6qw68ncEGbNzkYDL0ccszvuS20C5EzI= ARC-Message-Signature: i=1; a=rsa-sha256; d=sourceware.org; s=key; t=1708675568; c=relaxed/simple; bh=61KkC/bjrASKVPpFhs42WNc1443PLVFwujeHX/sh5To=; h=DKIM-Signature:From:To:Subject:Date:Message-Id:MIME-Version; b=KUED2Js7T0nFy0PAol0/kxh3OW3Dnu8DAJKnpXUl1/ghtSgQOpqIXLaWnou1TSp65y2s9CjElYpgG95EuprnUGHr1nX3C3rLRI2ZemZlUabnOG9N9vipwyAm0y3ie6kH34iVT0IW7LdJjE7glgKDiecwGpv4kOmEBrIqtUj5tDI= ARC-Authentication-Results: i=1; server2.sourceware.org DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1708675564; x=1740211564; h=from:to:cc:subject:date:message-id:mime-version: content-transfer-encoding; bh=61KkC/bjrASKVPpFhs42WNc1443PLVFwujeHX/sh5To=; b=Nfgb5sYva+8xpyYKGI7BOwke/ZoYapmahPtSKYuqBtCm94dNAdcJd9Ke C00SIY7urUjMabp7/25s4i+ESnE7/ZFC4SbmEXMAMLXJkcVYel6aIxiOF ujzeetTEVSMqyQey8xH852ln1Ksyaa5zGvLDfntAwzHV/5psS5O01e0v9 h0r2N/0YNrhRBtOG+axKzkCXoijel2FZIpDSVsUfiXhL5wxGApvQmIU2R Y58GjXGH9SoyUT1h20vc4A8EHi2a/4EnpnhoaF+aNT37cz6vCT4JXMi7D WalzeONTVlQoqwIL/fDUh39w90ibUDeXoBKW0fNvXHRxtjHww/oBSQ2sw Q==; X-IronPort-AV: E=McAfee;i="6600,9927,10992"; a="5920551" X-IronPort-AV: E=Sophos;i="6.06,179,1705392000"; d="scan'208";a="5920551" Received: from orviesa001.jf.intel.com ([10.64.159.141]) by fmvoesa107.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 23 Feb 2024 00:06:03 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="6.06,179,1705392000"; d="scan'208";a="43283994" Received: from shvmail02.sh.intel.com ([10.239.244.9]) by orviesa001.jf.intel.com with ESMTP; 23 Feb 2024 00:06:00 -0800 Received: from pli-ubuntu.sh.intel.com (pli-ubuntu.sh.intel.com [10.239.159.47]) by shvmail02.sh.intel.com (Postfix) with ESMTP id 7AF3F1005667; Fri, 23 Feb 2024 16:05:59 +0800 (CST) From: pan2.li@intel.com To: gcc-patches@gcc.gnu.org Cc: juzhe.zhong@rivai.ai, pan2.li@intel.com, yanzhang.wang@intel.com, kito.cheng@gmail.com Subject: [PATCH v1] RISC-V: Introduce gcc option mrvv-vector-bits for RVV Date: Fri, 23 Feb 2024 16:05:58 +0800 Message-Id: <20240223080558.2644800-1-pan2.li@intel.com> X-Mailer: git-send-email 2.34.1 MIME-Version: 1.0 X-Spam-Status: No, score=-11.1 required=5.0 tests=BAYES_00, DKIMWL_WL_HIGH, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, DKIM_VALID_EF, GIT_PATCH_0, KAM_SHORT, RCVD_IN_MSPIKE_H2, SPF_HELO_NONE, SPF_NONE, TXREP, T_SCC_BODY_TEXT_LINE autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on server2.sourceware.org X-BeenThere: gcc-patches@gcc.gnu.org X-Mailman-Version: 2.1.30 Precedence: list List-Id: Gcc-patches mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: gcc-patches-bounces+incoming=patchwork.ozlabs.org@gcc.gnu.org From: Pan Li This patch would like to introduce one new gcc option for RVV. To appoint the bits size of one RVV vector register. Valid arguments to '-mrvv-vector-bits=' are: * 64 * 128 * 256 * 512 * 1024 * 2048 * 4096 * 8192 * 16384 * 32768 * 65536 * scalable * zvl 1. The scalable will be the default values which take min_vlen for the riscv_vector_chunks. 2. The zvl will pick up the zvl*b from the march option. For example, the mrvv-vector-bits will be 1024 when march=rv64gcv_zvl1024b. 3. Otherwise, it will take the value provide and complain error if none of above valid value is given. This option may influence the code gen when auto-vector. For example, void test_rvv_vector_bits (int *a, int *b, int *out) { for (int i = 0; i < 8; i++) out[i] = a[i] + b[i]; } It will generate code similar to below when build with -march=rv64gcv_zvl128b -mabi=lp64 -mrvv-vector-bits=zvl test_rvv_vector_bits: ... vsetivli zero,4,e32,m1,ta,ma vle32.v v1,0(a0) vle32.v v2,0(a1) vadd.vv v1,v1,v2 vse32.v v1,0(a2) ... vle32.v v1,0(a0) vle32.v v2,0(a1) vadd.vv v1,v1,v2 vse32.v v1,0(a2) And it will become more simply similar to below when build with -march=rv64gcv_zvl128b -mabi=lp64 -mrvv-vector-bits=256 test_rvv_vector_bits: ... vsetivli zero,8,e32,m2,ta,ma vle32.v v2,0(a0) vle32.v v4,0(a1) vadd.vv v2,v2,v4 vse32.v v2,0(a2) Passed the regression test of rvv. gcc/ChangeLog: * config/riscv/riscv-opts.h (enum rvv_vector_bits_enum): New enum for different RVV vector bits. * config/riscv/riscv.cc (riscv_convert_vector_bits): New func to get the RVV vector bits, with given min_vlen. (riscv_convert_vector_chunks): Combine the mrvv-vector-bits option with min_vlen to RVV vector chunks. (riscv_override_options_internal): Update comments and rename the vector chunks. * config/riscv/riscv.opt: Add option mrvv-vector-bits. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/rvv-vector-bits-1.c: New test. * gcc.target/riscv/rvv/base/rvv-vector-bits-2.c: New test. * gcc.target/riscv/rvv/base/rvv-vector-bits-3.c: New test. * gcc.target/riscv/rvv/base/rvv-vector-bits-4.c: New test. Signed-off-by: Pan Li --- gcc/config/riscv/riscv-opts.h | 16 ++++++ gcc/config/riscv/riscv.cc | 49 ++++++++++++++++--- gcc/config/riscv/riscv.opt | 47 ++++++++++++++++++ .../riscv/rvv/base/rvv-vector-bits-1.c | 6 +++ .../riscv/rvv/base/rvv-vector-bits-2.c | 20 ++++++++ .../riscv/rvv/base/rvv-vector-bits-3.c | 25 ++++++++++ .../riscv/rvv/base/rvv-vector-bits-4.c | 6 +++ 7 files changed, 163 insertions(+), 6 deletions(-) create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/base/rvv-vector-bits-1.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/base/rvv-vector-bits-2.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/base/rvv-vector-bits-3.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/base/rvv-vector-bits-4.c diff --git a/gcc/config/riscv/riscv-opts.h b/gcc/config/riscv/riscv-opts.h index 4edddbadc37..b2141190731 100644 --- a/gcc/config/riscv/riscv-opts.h +++ b/gcc/config/riscv/riscv-opts.h @@ -129,6 +129,22 @@ enum vsetvl_strategy_enum { VSETVL_OPT_NO_FUSION, }; +enum rvv_vector_bits_enum { + RVV_VECTOR_BITS_SCALABLE, + RVV_VECTOR_BITS_ZVL, + RVV_VECTOR_BITS_64 = 64, + RVV_VECTOR_BITS_128 = 128, + RVV_VECTOR_BITS_256 = 256, + RVV_VECTOR_BITS_512 = 512, + RVV_VECTOR_BITS_1024 = 1024, + RVV_VECTOR_BITS_2048 = 2048, + RVV_VECTOR_BITS_4096 = 4096, + RVV_VECTOR_BITS_8192 = 8192, + RVV_VECTOR_BITS_16384 = 16384, + RVV_VECTOR_BITS_32768 = 32768, + RVV_VECTOR_BITS_65536 = 65536, +}; + #define TARGET_ZICOND_LIKE (TARGET_ZICOND || (TARGET_XVENTANACONDOPS && TARGET_64BIT)) /* Bit of riscv_zvl_flags will set contintuly, N-1 bit will set if N-bit is diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 5e984ee2a55..366d7ece383 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -8801,13 +8801,50 @@ riscv_init_machine_status (void) return ggc_cleared_alloc (); } -/* Return the VLEN value associated with -march. +static int +riscv_convert_vector_bits (int min_vlen) +{ + int rvv_bits = 0; + + switch (rvv_vector_bits) + { + case RVV_VECTOR_BITS_SCALABLE: + case RVV_VECTOR_BITS_ZVL: + rvv_bits = min_vlen; + break; + case RVV_VECTOR_BITS_64: + case RVV_VECTOR_BITS_128: + case RVV_VECTOR_BITS_256: + case RVV_VECTOR_BITS_512: + case RVV_VECTOR_BITS_1024: + case RVV_VECTOR_BITS_2048: + case RVV_VECTOR_BITS_4096: + case RVV_VECTOR_BITS_8192: + case RVV_VECTOR_BITS_16384: + case RVV_VECTOR_BITS_32768: + case RVV_VECTOR_BITS_65536: + rvv_bits = rvv_vector_bits; + break; + default: + gcc_unreachable (); + } + + if (rvv_bits < min_vlen) + error ("RVV vector bits %d cannot be less than minimal vector length %d", + rvv_bits, min_vlen); + + return rvv_bits; +} + +/* Return the VLEN value associated with -march and -mwrvv-vector-bits. TODO: So far we only support length-agnostic value. */ static poly_uint16 -riscv_convert_vector_bits (struct gcc_options *opts) +riscv_convert_vector_chunks (struct gcc_options *opts) { int chunk_num; int min_vlen = TARGET_MIN_VLEN_OPTS (opts); + int rvv_bits = riscv_convert_vector_bits (min_vlen); + if (min_vlen > 32) { /* When targetting minimum VLEN > 32, we should use 64-bit chunk size. @@ -8826,7 +8863,7 @@ riscv_convert_vector_bits (struct gcc_options *opts) - TARGET_MIN_VLEN = 2048bit: [256,256] - TARGET_MIN_VLEN = 4096bit: [512,512] FIXME: We currently DON'T support TARGET_MIN_VLEN > 4096bit. */ - chunk_num = min_vlen / 64; + chunk_num = rvv_bits / 64; } else { @@ -8848,7 +8885,7 @@ riscv_convert_vector_bits (struct gcc_options *opts) if (TARGET_VECTOR_OPTS_P (opts)) { if (opts->x_riscv_autovec_preference == RVV_FIXED_VLMAX) - return (int) min_vlen / (riscv_bytes_per_vector_chunk * 8); + return (int) rvv_bits / (riscv_bytes_per_vector_chunk * 8); else return poly_uint16 (chunk_num, chunk_num); } @@ -8920,8 +8957,8 @@ riscv_override_options_internal (struct gcc_options *opts) if (TARGET_VECTOR && TARGET_BIG_ENDIAN) sorry ("Current RISC-V GCC does not support RVV in big-endian mode"); - /* Convert -march to a chunks count. */ - riscv_vector_chunks = riscv_convert_vector_bits (opts); + /* Convert -march and -mrvv-vector-bits to a chunks count. */ + riscv_vector_chunks = riscv_convert_vector_chunks (opts); } /* Implement TARGET_OPTION_OVERRIDE. */ diff --git a/gcc/config/riscv/riscv.opt b/gcc/config/riscv/riscv.opt index 20685c42aed..73ae6abe871 100644 --- a/gcc/config/riscv/riscv.opt +++ b/gcc/config/riscv/riscv.opt @@ -607,3 +607,50 @@ Enum(stringop_strategy) String(vector) Value(STRATEGY_VECTOR) mstringop-strategy= Target RejectNegative Joined Enum(stringop_strategy) Var(stringop_strategy) Init(STRATEGY_AUTO) Specify stringop expansion strategy. + +Enum +Name(rvv_vector_bits) Type(enum rvv_vector_bits_enum) +The possible RVV vector register lengths: + +EnumValue +Enum(rvv_vector_bits) String(scalable) Value(RVV_VECTOR_BITS_SCALABLE) + +EnumValue +Enum(rvv_vector_bits) String(64) Value(RVV_VECTOR_BITS_64) + +EnumValue +Enum(rvv_vector_bits) String(128) Value(RVV_VECTOR_BITS_128) + +EnumValue +Enum(rvv_vector_bits) String(256) Value(RVV_VECTOR_BITS_256) + +EnumValue +Enum(rvv_vector_bits) String(512) Value(RVV_VECTOR_BITS_512) + +EnumValue +Enum(rvv_vector_bits) String(1024) Value(RVV_VECTOR_BITS_1024) + +EnumValue +Enum(rvv_vector_bits) String(2048) Value(RVV_VECTOR_BITS_2048) + +EnumValue +Enum(rvv_vector_bits) String(4096) Value(RVV_VECTOR_BITS_4096) + +EnumValue +Enum(rvv_vector_bits) String(8192) Value(RVV_VECTOR_BITS_8192) + +EnumValue +Enum(rvv_vector_bits) String(16384) Value(RVV_VECTOR_BITS_16384) + +EnumValue +Enum(rvv_vector_bits) String(32768) Value(RVV_VECTOR_BITS_32768) + +EnumValue +Enum(rvv_vector_bits) String(65536) Value(RVV_VECTOR_BITS_65536) + +EnumValue +Enum(rvv_vector_bits) String(zvl) Value(RVV_VECTOR_BITS_ZVL) + +mrvv-vector-bits= +Target RejectNegative Joined Enum(rvv_vector_bits) Var(rvv_vector_bits) Init(RVV_VECTOR_BITS_SCALABLE) +-mrvv-vector-bits= Set the number of bits in an RVV vector register. diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/rvv-vector-bits-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/rvv-vector-bits-1.c new file mode 100644 index 00000000000..b06d791f383 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/rvv-vector-bits-1.c @@ -0,0 +1,6 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvl256b -mabi=lp64 -mrvv-vector-bits=128 -O3" } */ + +#include "riscv_vector.h" + +/* { dg-error "RVV vector bits 128 cannot be less than minimal vector length 256" "" { target { "riscv*-*-*" } } 0 } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/rvv-vector-bits-2.c b/gcc/testsuite/gcc.target/riscv/rvv/base/rvv-vector-bits-2.c new file mode 100644 index 00000000000..37744339080 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/rvv-vector-bits-2.c @@ -0,0 +1,20 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvl128b -mabi=lp64 -mrvv-vector-bits=256 -O3 -fno-schedule-insns -fno-schedule-insns2" } */ +/* { dg-final { check-function-bodies "**" "" } } */ + +/* +** test_rvv_vector_bits_256: +** ... +** vsetivli\s+zero,\s*8,\s*e32,\s*m2,\s*ta,\s*ma +** vle32\.v\s+v[0-9]+,\s*0\(a0\) +** vle32\.v\s+v[0-9]+,\s*0\(a1\) +** vadd\.vv\s+v[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** vse32\.v\s+v[0-9]+,\s*0\(a2\) +** ret +** ... +*/ +void test_rvv_vector_bits_256 (int *a, int *b, int *out) +{ + for (int i = 0; i < 8; i++) + out[i] = a[i] + b[i]; +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/rvv-vector-bits-3.c b/gcc/testsuite/gcc.target/riscv/rvv/base/rvv-vector-bits-3.c new file mode 100644 index 00000000000..962cc8ffa6d --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/rvv-vector-bits-3.c @@ -0,0 +1,25 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvl128b -mabi=lp64 -mrvv-vector-bits=zvl -O3 -fno-schedule-insns -fno-schedule-insns2" } */ +/* { dg-final { check-function-bodies "**" "" } } */ + +/* +** test_rvv_vector_bits_zvl: +** ... +** vsetivli\s+zero,\s*4,\s*e32,\s*m1,\s*ta,\s*ma +** vle32\.v\s+v[0-9]+,\s*0\(a0\) +** vle32\.v\s+v[0-9]+,\s*0\(a1\) +** vadd\.vv\s+v[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** vse32\.v\s+v[0-9]+,\s*0\(a2\) +** ... +** vle32\.v\s+v[0-9]+,\s*0\(a0\) +** vle32\.v\s+v[0-9]+,\s*0\(a1\) +** vadd\.vv\s+v[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** vse32\.v\s+v[0-9]+,\s*0\(a2\) +** ret +** ... +*/ +void test_rvv_vector_bits_zvl (int *a, int *b, int *out) +{ + for (int i = 0; i < 8; i++) + out[i] = a[i] + b[i]; +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/rvv-vector-bits-4.c b/gcc/testsuite/gcc.target/riscv/rvv/base/rvv-vector-bits-4.c new file mode 100644 index 00000000000..863f96187e1 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/rvv-vector-bits-4.c @@ -0,0 +1,6 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvl256b -mabi=lp64 -mrvv-vector-bits=invalid-bits -O3" } */ + +#include "riscv_vector.h" + +/* { dg-error "unrecognized argument in option '-mrvv-vector-bits=invalid-bits" "" { target { "riscv*-*-*" } } 0 } */