From patchwork Fri Nov 10 03:39:00 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: liuhongt X-Patchwork-Id: 163772 Return-Path: Delivered-To: ouuuleilei@gmail.com Received: by 2002:a59:b129:0:b0:403:3b70:6f57 with SMTP id q9csp867146vqs; Thu, 9 Nov 2023 19:41:36 -0800 (PST) X-Google-Smtp-Source: AGHT+IGbv/lryZaqP0NsnLc5ifto1XGVjGLac2yO7rwJ6YWk2Sv6CTAVt8dYHLMDghBFRJKx5QH1 X-Received: by 2002:a05:6214:1d23:b0:670:d2dc:c8e6 with SMTP id f3-20020a0562141d2300b00670d2dcc8e6mr7537327qvd.27.1699587696047; Thu, 09 Nov 2023 19:41:36 -0800 (PST) ARC-Seal: i=2; a=rsa-sha256; t=1699587696; cv=pass; d=google.com; s=arc-20160816; b=b6I3Sgo3ShZGrW29Sww8BcJKS+6pmP3KLRgMC+IqGQqRbsHrIvvBrn0JIDHLp2E5cT qdKH0lf6+7pylxvZeb2qmte8WoHV9V75JI5Sw3fX/Ys7x2GIBEfC447j0bL3RBUpJCbT 2Ltenh3L/xbnj/zFts8qqe+utqWyStEgUHCXXNkCJwkwvAlJ8raP2Q8+Jo+chdZrx0bR Jwmx0DG0wQNwDXUM6ZdhARC2llKs0uDc+OnoJsNJ3nO9Yk42NeCte3RGQL0L84bmUVxH wJKIZnVhLrcoNvBxROVCwvbsHibG7OXcODBC/Avhr2LC4oFC7ojtoSecp4wzaEpsTPYq a2xA== ARC-Message-Signature: i=2; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=errors-to:list-subscribe:list-help:list-post:list-archive :list-unsubscribe:list-id:precedence:content-transfer-encoding :mime-version:message-id:date:subject:cc:to:from:dkim-signature :arc-filter:dmarc-filter:delivered-to; bh=cwcc100UwRzkVMHnm53wSht2UvO7w35ZHXwod8bPh40=; fh=ChXOctppJn0KECDRINafwUY5xHRufGHaa0Ju9pddrcQ=; b=ssh+42mmY0Uk53D4JPPqoSq7gFPBahKPK/FVc0Zc5KGIzMBtdvBuRfJKglR5A4R+QA z2jP2Unhu2Df8sMnQGDFXZTNHX+iD4MEtIDyUqLqViONiUF6A0m7RDrvJeDXK1DYc5SG hoSqp3wrppSkAOWOgzDzKLIiNbfykDvdhrDYSQOjGpGwQBxjxhtEkp7ubzGsEOmB+ULc kFNZ7Pcg1UjVFvwGUNIVtlxRlF0X7GUyztum/lf9t95ZIH9HO3h0gbJCa9ND1OaBul9p 1tKbhR5xERga4B5+pFb1mqKTnTWjlVB+QlS7/AJ7v0FsAZvfRhDC7YLor0inm+6W5lEw t1CA== ARC-Authentication-Results: i=2; mx.google.com; dkim=pass header.i=@intel.com header.s=Intel header.b=IFy94wce; arc=pass (i=1); spf=pass (google.com: domain of gcc-patches-bounces+ouuuleilei=gmail.com@gcc.gnu.org designates 2620:52:3:1:0:246e:9693:128c as permitted sender) smtp.mailfrom="gcc-patches-bounces+ouuuleilei=gmail.com@gcc.gnu.org"; dmarc=pass (p=NONE sp=NONE dis=NONE) header.from=intel.com Received: from server2.sourceware.org (server2.sourceware.org. [2620:52:3:1:0:246e:9693:128c]) by mx.google.com with ESMTPS id l18-20020a056214105200b0066d7ec3340bsi3961124qvr.616.2023.11.09.19.41.35 for (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Thu, 09 Nov 2023 19:41:36 -0800 (PST) Received-SPF: pass (google.com: domain of gcc-patches-bounces+ouuuleilei=gmail.com@gcc.gnu.org designates 2620:52:3:1:0:246e:9693:128c as permitted sender) client-ip=2620:52:3:1:0:246e:9693:128c; Authentication-Results: mx.google.com; dkim=pass header.i=@intel.com header.s=Intel header.b=IFy94wce; arc=pass (i=1); spf=pass (google.com: domain of gcc-patches-bounces+ouuuleilei=gmail.com@gcc.gnu.org designates 2620:52:3:1:0:246e:9693:128c as permitted sender) smtp.mailfrom="gcc-patches-bounces+ouuuleilei=gmail.com@gcc.gnu.org"; dmarc=pass (p=NONE sp=NONE dis=NONE) header.from=intel.com Received: from server2.sourceware.org (localhost [IPv6:::1]) by sourceware.org (Postfix) with ESMTP id B67A03857BB3 for ; Fri, 10 Nov 2023 03:41:35 +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 [192.55.52.115]) by sourceware.org (Postfix) with ESMTPS id 285803858D33 for ; Fri, 10 Nov 2023 03:41:08 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org 285803858D33 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 285803858D33 Authentication-Results: server2.sourceware.org; arc=none smtp.remote-ip=192.55.52.115 ARC-Seal: i=1; a=rsa-sha256; d=sourceware.org; s=key; t=1699587673; cv=none; b=LhemfPPVXkhj/DnP9nGY27jQZqW0/jy1hMHXQjHu5pD7fBkV2DFLTw9lGUMEgqKJT+dABfQzn9p0dmQje7THNPauq+G+GnFrKWt8SJxf1sdWqi2bZ5T3+3S4MNJeKLPnTwKXR/Vf9hLwPi8p9Joys9IqIY1nmpHp2lu61gL4Nh0= ARC-Message-Signature: i=1; a=rsa-sha256; d=sourceware.org; s=key; t=1699587673; c=relaxed/simple; bh=z4Vj9FfV9tZBFBk5DeG/o63DB9EFFs7vUC1uc2Kn5Z4=; h=DKIM-Signature:From:To:Subject:Date:Message-Id:MIME-Version; b=bXKY02Vw0C2E8O9rKZ/LIJcgnmr36+nBBJtIZ4T/wvyrwu1y2uf6wOggbu2+h9iTRAiNo8PYVIka8AVWIYibSlgV2UwsmcCpJxhbU0R4UCp72rOm8PHby7+seeBZF+zT2DK/cKfdnJofC2qjS4i5+gte0JoJKTiLkLsYjq3use8= 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=1699587668; x=1731123668; h=from:to:cc:subject:date:message-id:mime-version: content-transfer-encoding; bh=z4Vj9FfV9tZBFBk5DeG/o63DB9EFFs7vUC1uc2Kn5Z4=; b=IFy94wceHRjZYL0eIYPwMQCVABsko9m+Paeot2LM7VS4cj6k+sMcirnw x11NOeeBpx7vCv46PFhCSD/2YAnu5PZX9Beovqz7MbByBb/eh5sCko/zG oIV37Jh8mk4+Eu7GtFKzUni8RSxR3zl3yoD6pVfYp3bFpoiHMI9R4oQsL YByL29MxuzlPP1m2tB+osuysv/blqP6qXg/POb/tL5ErtPMZdIdVdd/oI R+OMvlp7fcccPLiKIvnD5Cr3LNBpXld7UYzj43OfkjReF0Zi4COhHbahE L5nY/8pY/D8Liau3bQith6+nlPY+jPnetFvziVc8Q+J9MPRepxNwa+caP w==; X-IronPort-AV: E=McAfee;i="6600,9927,10889"; a="389933404" X-IronPort-AV: E=Sophos;i="6.03,291,1694761200"; d="scan'208";a="389933404" Received: from orsmga001.jf.intel.com ([10.7.209.18]) by fmsmga103.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 09 Nov 2023 19:41:07 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=McAfee;i="6600,9927,10889"; a="798507118" X-IronPort-AV: E=Sophos;i="6.03,291,1694761200"; d="scan'208";a="798507118" Received: from shvmail03.sh.intel.com ([10.239.245.20]) by orsmga001.jf.intel.com with ESMTP; 09 Nov 2023 19:41:01 -0800 Received: from shliclel4217.sh.intel.com (shliclel4217.sh.intel.com [10.239.240.127]) by shvmail03.sh.intel.com (Postfix) with ESMTP id 4880610056F3; Fri, 10 Nov 2023 11:41:00 +0800 (CST) From: liuhongt To: gcc-patches@gcc.gnu.org Cc: crazylht@gmail.com, hjl.tools@gmail.com Subject: [PATCH] Support vec_set/vec_extract/vec_init for V4HF/V2HF. Date: Fri, 10 Nov 2023 11:39:00 +0800 Message-Id: <20231110033900.246872-1-hongtao.liu@intel.com> X-Mailer: git-send-email 2.31.1 MIME-Version: 1.0 X-Spam-Status: No, score=-12.0 required=5.0 tests=BAYES_00, DKIMWL_WL_HIGH, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, DKIM_VALID_EF, GIT_PATCH_0, KAM_SHORT, 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+ouuuleilei=gmail.com@gcc.gnu.org X-getmail-retrieved-from-mailbox: INBOX X-GMAIL-THRID: 1782146867556484356 X-GMAIL-MSGID: 1782146867556484356 Bootstrapped and regtested on x86_64-pc-linux-gnu{-m32,}. Ready push to trunk. gcc/ChangeLog: * config/i386/i386-expand.cc (ix86_expand_vector_init_duplicate): Handle V4HF/V4BF and V2HF/V2BF. (ix86_expand_vector_init_one_nonzero): Ditto. (ix86_expand_vector_init_one_var): Ditto. (ix86_expand_vector_init_general): Ditto. (ix86_expand_vector_set_var): Ditto. (ix86_expand_vector_set): Ditto. (ix86_expand_vector_extract): Ditto. * config/i386/mmx.md (mmxdoublevecmode): Extend to V4HF/V4BF/V2HF/V2BF. (*mmx_pinsrw): Extend to V4FI_64, add a new alternative (&x, x, x), add a new define_split after the pattern. (*mmx_pextrw): New define_insn. (mmx_pshufw_1): Rename to .. (mmx_pshufw_1): .. this, extend to V4FI_64. (*mmx_pblendw64): Extend to V4FI_64. (*vec_dup): New define_insn. (vec_setv4hi): Rename to .. (vec_set): .. this, and extend to V4FI_64 (vec_extractv4hihi): Rename to .. (vec_extract): .. this, and extend to V4FI_64. (vec_init): New define_insn. (*pinsrw): Extend to V2FI_32, add a new alternative (&x, x, x), and add a new define_split after it. (*pextrw): New define_insn. (vec_setv2hi): Rename to .. (vec_set): .. this, extend to V2FI_32. (vec_extractv2hihi): Rename to .. (vec_extract): .. this, extend to V2FI_32. (*punpckwd): Extend to V2FI_32. (*pshufw_1): Rename to .. (*pshufw_1): .. this, extend to V2FI_32. (vec_initv2hihi): Rename to .. (vec_init): .. this, and extend to V2FI_32. (*vec_dup): New define_insn. * config/i386/sse.md (*vec_extract): Refine constraint from v to Yw. gcc/testsuite/ChangeLog: * gcc.target/i386/part-vect-vec_elem-1.c: New test. * gcc.target/i386/part-vect-vec_elem-2.c: New test. --- gcc/config/i386/i386-expand.cc | 60 ++++ gcc/config/i386/mmx.md | 271 ++++++++++++++---- gcc/config/i386/sse.md | 4 +- .../gcc.target/i386/part-vect-vec_elem-1.c | 135 +++++++++ .../gcc.target/i386/part-vect-vec_elem-2.c | 135 +++++++++ 5 files changed, 541 insertions(+), 64 deletions(-) create mode 100644 gcc/testsuite/gcc.target/i386/part-vect-vec_elem-1.c create mode 100644 gcc/testsuite/gcc.target/i386/part-vect-vec_elem-2.c diff --git a/gcc/config/i386/i386-expand.cc b/gcc/config/i386/i386-expand.cc index 8fad73c1549..b52ec51fbe4 100644 --- a/gcc/config/i386/i386-expand.cc +++ b/gcc/config/i386/i386-expand.cc @@ -15592,6 +15592,17 @@ ix86_expand_vector_init_duplicate (bool mmx_ok, machine_mode mode, } goto widen; + case E_V4HFmode: + case E_V4BFmode: + if (TARGET_MMX_WITH_SSE) + { + val = force_reg (GET_MODE_INNER (mode), val); + rtx x = gen_rtx_VEC_DUPLICATE (mode, val); + emit_insn (gen_rtx_SET (target, x)); + return true; + } + return false; + case E_V2HImode: if (TARGET_SSE2) { @@ -15605,6 +15616,17 @@ ix86_expand_vector_init_duplicate (bool mmx_ok, machine_mode mode, } return false; + case E_V2HFmode: + case E_V2BFmode: + if (TARGET_SSE2) + { + val = force_reg (GET_MODE_INNER (mode), val); + rtx x = gen_rtx_VEC_DUPLICATE (mode, val); + emit_insn (gen_rtx_SET (target, x)); + return true; + } + return false; + case E_V8QImode: case E_V4QImode: if (!mmx_ok) @@ -15815,6 +15837,8 @@ ix86_expand_vector_init_one_nonzero (bool mmx_ok, machine_mode mode, use_vector_set = TARGET_MMX_WITH_SSE && TARGET_SSE4_1; break; case E_V4HImode: + case E_V4HFmode: + case E_V4BFmode: use_vector_set = TARGET_SSE || TARGET_3DNOW_A; break; case E_V4QImode: @@ -16051,6 +16075,8 @@ ix86_expand_vector_init_one_var (bool mmx_ok, machine_mode mode, case E_V4SImode: case E_V8HImode: case E_V4HImode: + case E_V4HFmode: + case E_V4BFmode: break; case E_V16QImode: @@ -16438,6 +16464,7 @@ ix86_expand_vector_init_general (bool mmx_ok, machine_mode mode, rtx ops[64], op0, op1, op2, op3, op4, op5; machine_mode half_mode = VOIDmode; machine_mode quarter_mode = VOIDmode; + machine_mode int_inner_mode = VOIDmode; int n, i; switch (mode) @@ -16582,6 +16609,13 @@ quarter: ix86_expand_vector_init_interleave (mode, target, ops, n >> 1); return; + case E_V4HFmode: + case E_V4BFmode: + case E_V2HFmode: + case E_V2BFmode: + int_inner_mode = HImode; + break; + case E_V4HImode: case E_V8QImode: @@ -16613,6 +16647,16 @@ quarter: for (j = 0; j < n_elt_per_word; ++j) { rtx elt = XVECEXP (vals, 0, (i+1)*n_elt_per_word - j - 1); + if (int_inner_mode != E_VOIDmode) + { + gcc_assert (TARGET_SSE2 && int_inner_mode == HImode); + rtx tmp = gen_reg_rtx (int_inner_mode); + elt = lowpart_subreg (int_inner_mode, + force_reg (inner_mode, elt), + inner_mode); + emit_move_insn (tmp, elt); + elt = tmp; + } elt = convert_modes (tmp_mode, inner_mode, elt, true); if (j == 0) @@ -16839,6 +16883,14 @@ ix86_expand_vector_set_var (rtx target, rtx val, rtx idx) case E_V16SFmode: cmp_mode = V16SImode; break; + case E_V2HFmode: + case E_V2BFmode: + cmp_mode = V2HImode; + break; + case E_V4HFmode: + case E_V4BFmode: + cmp_mode = V4HImode; + break; case E_V8HFmode: cmp_mode = V8HImode; break; @@ -17085,9 +17137,13 @@ ix86_expand_vector_set (bool mmx_ok, rtx target, rtx val, int elt) case E_V8HFmode: case E_V8BFmode: case E_V2HImode: + case E_V2HFmode: + case E_V2BFmode: use_vec_merge = TARGET_SSE2; break; case E_V4HImode: + case E_V4HFmode: + case E_V4BFmode: use_vec_merge = mmx_ok && (TARGET_SSE || TARGET_3DNOW_A); break; @@ -17428,9 +17484,13 @@ ix86_expand_vector_extract (bool mmx_ok, rtx target, rtx vec, int elt) case E_V8HFmode: case E_V8BFmode: case E_V2HImode: + case E_V2HFmode: + case E_V2BFmode: use_vec_extr = TARGET_SSE2; break; case E_V4HImode: + case E_V4HFmode: + case E_V4BFmode: use_vec_extr = mmx_ok && (TARGET_SSE || TARGET_3DNOW_A); break; diff --git a/gcc/config/i386/mmx.md b/gcc/config/i386/mmx.md index 50402b5b544..a3d08bb9d3b 100644 --- a/gcc/config/i386/mmx.md +++ b/gcc/config/i386/mmx.md @@ -112,11 +112,21 @@ (define_mode_attr mmxintvecmodelower ;; Mapping of vector modes to a vector mode of double size (define_mode_attr mmxdoublevecmode - [(V2SF "V4SF") (V2SI "V4SI") (V4HF "V8HF") (V4HI "V8HI")]) + [(V2SF "V4SF") (V2SI "V4SI") (V4HF "V8HF") (V4HI "V8HI") + (V2HI "V4HI") (V2HF "V4HF") (V2BF "V4BF")]) ;; Mapping of vector modes back to the scalar modes (define_mode_attr mmxscalarmode - [(V2SI "SI") (V2SF "SF")]) + [(V2SI "SI") (V2SF "SF") + (V4HF "HF") (V4BF "BF") + (V2HF "HF") (V2BF "BF") + (V4HI "HI") (V2HI "HI")]) + +(define_mode_attr mmxscalarmodelower + [(V2SI "si") (V2SF "sf") + (V4HF "hf") (V4BF "bf") + (V2HF "hf") (V2BF "bf") + (V4HI "hi") (V2HI "hi")]) (define_mode_attr Yv_Yw [(V8QI "Yw") (V4HI "Yw") (V2SI "Yv") (V1DI "Yv") (V2SF "Yv")]) @@ -4882,11 +4892,11 @@ (define_insn "*mmx_pinsrd" (set_attr "mode" "TI")]) (define_insn "*mmx_pinsrw" - [(set (match_operand:V4HI 0 "register_operand" "=y,x,YW") - (vec_merge:V4HI - (vec_duplicate:V4HI - (match_operand:HI 2 "nonimmediate_operand" "rm,rm,rm")) - (match_operand:V4HI 1 "register_operand" "0,0,YW") + [(set (match_operand:V4FI_64 0 "register_operand" "=y,x,YW,&x") + (vec_merge:V4FI_64 + (vec_duplicate:V4FI_64 + (match_operand: 2 "nonimmediate_operand" "rm,rm,rm,x")) + (match_operand:V4FI_64 1 "register_operand" "0,0,YW,x") (match_operand:SI 3 "const_int_operand")))] "(TARGET_MMX || TARGET_MMX_WITH_SSE) && (TARGET_SSE || TARGET_3DNOW_A) @@ -4896,6 +4906,8 @@ (define_insn "*mmx_pinsrw" operands[3] = GEN_INT (exact_log2 (INTVAL (operands[3]))); switch (which_alternative) { + case 3: + return "#"; case 2: if (MEM_P (operands[2])) return "vpinsrw\t{%3, %2, %1, %0|%0, %1, %2, %3}"; @@ -4911,11 +4923,28 @@ (define_insn "*mmx_pinsrw" gcc_unreachable (); } } - [(set_attr "isa" "*,sse2_noavx,avx") - (set_attr "mmx_isa" "native,*,*") - (set_attr "type" "mmxcvt,sselog,sselog") + [(set_attr "isa" "*,sse2_noavx,avx,sse4") + (set_attr "mmx_isa" "native,*,*,*") + (set_attr "type" "mmxcvt,sselog,sselog,sselog") (set_attr "length_immediate" "1") - (set_attr "mode" "DI,TI,TI")]) + (set_attr "mode" "DI,TI,TI,TI")]) + +;; For TARGET_SSE2, implement insert from XMM reg with PSHULFW + PBLENDW. +(define_split + [(set (match_operand:V4FI_64 0 "sse_reg_operand") + (vec_merge:V4FI_64 + (vec_duplicate:V4FI_64 + (match_operand: 2 "sse_reg_operand")) + (match_operand:V4FI_64 1 "sse_reg_operand") + (match_operand:SI 3 "const_int_operand")))] + "TARGET_MMX_WITH_SSE && TARGET_SSE4_1 && reload_completed + && ((unsigned) exact_log2 (INTVAL (operands[3])) + < GET_MODE_NUNITS (mode))" + [(set (match_dup 0) + (vec_duplicate:V4FI_64 (match_dup 2))) + (set (match_dup 0) + (vec_merge:V4FI_64 (match_dup 1) (match_dup 0) (match_dup 3)))] + "operands[3] = GEN_INT (~INTVAL (operands[3]) & 0xf);") (define_insn "*mmx_pinsrb" [(set (match_operand:V8QI 0 "register_operand" "=x,YW") @@ -4973,6 +5002,41 @@ (define_insn "*mmx_pextrw" (set_attr "prefix" "orig,maybe_vex,maybe_vex,maybe_evex") (set_attr "mode" "DI,TI,TI,TI")]) +(define_insn "*mmx_pextrw" + [(set (match_operand: 0 "register_sse4nonimm_operand" "=?r,?r,jm,m,x,Yw") + (vec_select: + (match_operand:V4F_64 1 "register_operand" "y,YW,YW,YW,0,YW") + (parallel [(match_operand:SI 2 "const_0_to_3_operand")])))] + "(TARGET_MMX || TARGET_MMX_WITH_SSE) + && (TARGET_SSE || TARGET_3DNOW_A)" +{ + switch (which_alternative) + { + case 0: + case 1: + return "%vpextrw\t{%2, %1, %k0|%k0, %1, %2}"; + case 2: + case 3: + return "%vpextrw\t{%2, %1, %0|%0, %1, %2}"; + case 4: + operands[2] = GEN_INT (INTVAL (operands[2]) * 2); + return "psrldq\t{%2, %0|%0, %2}"; + case 5: + operands[2] = GEN_INT (INTVAL (operands[2]) * 2); + return "vpsrldq\t{%2, %1, %0|%0, %1, %2}"; + + default: + gcc_unreachable (); + } +} + [(set_attr "isa" "*,sse2,sse4_noavx,avx,noavx,avx") + (set_attr "addr" "*,*,gpr16,*,*,*") + (set_attr "mmx_isa" "native,*,*,*,*,*") + (set_attr "type" "mmxcvt,sselog1,sselog1,sselog1,sseishft1,sseishft1") + (set_attr "length_immediate" "1") + (set_attr "prefix" "orig,maybe_vex,maybe_vex,maybe_evex,orig,maybe_evex") + (set_attr "mode" "DI,TI,TI,TI,TI,TI")]) + (define_insn "*mmx_pextrw_zext" [(set (match_operand:SWI48 0 "register_operand" "=r,r") (zero_extend:SWI48 @@ -5069,18 +5133,18 @@ (define_expand "mmx_pshufw" && (TARGET_SSE || TARGET_3DNOW_A)" { int mask = INTVAL (operands[2]); - emit_insn (gen_mmx_pshufw_1 (operands[0], operands[1], - GEN_INT ((mask >> 0) & 3), - GEN_INT ((mask >> 2) & 3), - GEN_INT ((mask >> 4) & 3), - GEN_INT ((mask >> 6) & 3))); + emit_insn (gen_mmx_pshufwv4hi_1 (operands[0], operands[1], + GEN_INT ((mask >> 0) & 3), + GEN_INT ((mask >> 2) & 3), + GEN_INT ((mask >> 4) & 3), + GEN_INT ((mask >> 6) & 3))); DONE; }) -(define_insn "mmx_pshufw_1" - [(set (match_operand:V4HI 0 "register_operand" "=y,Yw") - (vec_select:V4HI - (match_operand:V4HI 1 "register_mmxmem_operand" "ym,Yw") +(define_insn "mmx_pshufw_1" + [(set (match_operand:V4FI_64 0 "register_operand" "=y,Yw") + (vec_select:V4FI_64 + (match_operand:V4FI_64 1 "register_mmxmem_operand" "ym,Yw") (parallel [(match_operand 2 "const_0_to_3_operand") (match_operand 3 "const_0_to_3_operand") (match_operand 4 "const_0_to_3_operand") @@ -5134,10 +5198,10 @@ (define_insn "*mmx_pshufd_1" (set_attr "mode" "TI")]) (define_insn "*mmx_pblendw64" - [(set (match_operand:V4HI 0 "register_operand" "=Yr,*x,x") - (vec_merge:V4HI - (match_operand:V4HI 2 "register_operand" "Yr,*x,x") - (match_operand:V4HI 1 "register_operand" "0,0,x") + [(set (match_operand:V4FI_64 0 "register_operand" "=Yr,*x,x") + (vec_merge:V4FI_64 + (match_operand:V4FI_64 2 "register_operand" "Yr,*x,x") + (match_operand:V4FI_64 1 "register_operand" "0,0,x") (match_operand:SI 3 "const_0_to_15_operand")))] "TARGET_SSE4_1 && TARGET_MMX_WITH_SSE" "@ @@ -5152,10 +5216,10 @@ (define_insn "*mmx_pblendw64" (set_attr "mode" "TI")]) (define_insn "*mmx_pblendw32" - [(set (match_operand:V2HI 0 "register_operand" "=Yr,*x,x") - (vec_merge:V2HI - (match_operand:V2HI 2 "register_operand" "Yr,*x,x") - (match_operand:V2HI 1 "register_operand" "0,0,x") + [(set (match_operand:V2FI_32 0 "register_operand" "=Yr,*x,x") + (vec_merge:V2FI_32 + (match_operand:V2FI_32 2 "register_operand" "Yr,*x,x") + (match_operand:V2FI_32 1 "register_operand" "0,0,x") (match_operand:SI 3 "const_0_to_7_operand")))] "TARGET_SSE4_1" "@ @@ -5212,6 +5276,16 @@ (define_insn "*vec_dupv4hi" (set_attr "length_immediate" "1") (set_attr "mode" "DI,TI")]) +(define_insn "*vec_dup" + [(set (match_operand:V4F_64 0 "register_operand" "=Yw") + (vec_duplicate:V4F_64 + (match_operand: 1 "register_operand" "Yw")))] + "TARGET_MMX_WITH_SSE" + "%vpshuflw\t{$0, %1, %0|%0, %1, 0}" + [(set_attr "isa" "sse2") + (set_attr "type" "sselog1") + (set_attr "length_immediate" "1") + (set_attr "mode" "TI")]) (define_insn "*vec_dupv2si" [(set (match_operand:V2SI 0 "register_operand" "=y,Yv") @@ -5405,9 +5479,9 @@ (define_expand "vec_initv2sisi" DONE; }) -(define_expand "vec_setv4hi" - [(match_operand:V4HI 0 "register_operand") - (match_operand:HI 1 "register_operand") +(define_expand "vec_set" + [(match_operand:V4FI_64 0 "register_operand") + (match_operand: 1 "register_operand") (match_operand 2 "vec_setm_mmx_operand")] "TARGET_MMX || TARGET_MMX_WITH_SSE" { @@ -5419,9 +5493,9 @@ (define_expand "vec_setv4hi" DONE; }) -(define_expand "vec_extractv4hihi" - [(match_operand:HI 0 "register_operand") - (match_operand:V4HI 1 "register_operand") +(define_expand "vec_extract" + [(match_operand: 0 "register_operand") + (match_operand:V4FI_64 1 "register_operand") (match_operand 2 "const_int_operand")] "TARGET_MMX || TARGET_MMX_WITH_SSE" { @@ -5440,6 +5514,16 @@ (define_expand "vec_initv4hihi" DONE; }) +(define_expand "vec_init" + [(match_operand:V4F_64 0 "register_operand") + (match_operand 1)] + "TARGET_MMX_WITH_SSE" +{ + ix86_expand_vector_init (TARGET_MMX_WITH_SSE, operands[0], + operands[1]); + DONE; +}) + (define_expand "vec_setv8qi" [(match_operand:V8QI 0 "register_operand") (match_operand:QI 1 "register_operand") @@ -5476,11 +5560,11 @@ (define_expand "vec_initv8qiqi" }) (define_insn "*pinsrw" - [(set (match_operand:V2HI 0 "register_operand" "=x,YW") - (vec_merge:V2HI - (vec_duplicate:V2HI - (match_operand:HI 2 "nonimmediate_operand" "rm,rm")) - (match_operand:V2HI 1 "register_operand" "0,YW") + [(set (match_operand:V2FI_32 0 "register_operand" "=x,YW,&x") + (vec_merge:V2FI_32 + (vec_duplicate:V2FI_32 + (match_operand: 2 "nonimmediate_operand" "rm,rm,x")) + (match_operand:V2FI_32 1 "register_operand" "0,YW,x") (match_operand:SI 3 "const_int_operand")))] "TARGET_SSE2 && ((unsigned) exact_log2 (INTVAL (operands[3])) @@ -5489,6 +5573,8 @@ (define_insn "*pinsrw" operands[3] = GEN_INT (exact_log2 (INTVAL (operands[3]))); switch (which_alternative) { + case 2: + return "#"; case 1: if (MEM_P (operands[2])) return "vpinsrw\t{%3, %2, %1, %0|%0, %1, %2, %3}"; @@ -5503,11 +5589,29 @@ (define_insn "*pinsrw" gcc_unreachable (); } } - [(set_attr "isa" "noavx,avx") + [(set_attr "isa" "noavx,avx,sse4") (set_attr "type" "sselog") (set_attr "length_immediate" "1") (set_attr "mode" "TI")]) +;; For TARGET_SSE2, implement insert from XMM reg with PSHULFW + PBLENDW. +(define_split + [(set (match_operand:V2FI_32 0 "sse_reg_operand") + (vec_merge:V2FI_32 + (vec_duplicate:V2FI_32 + (match_operand: 2 "sse_reg_operand")) + (match_operand:V2FI_32 1 "sse_reg_operand") + (match_operand:SI 3 "const_int_operand")))] + "TARGET_SSE4_1 && reload_completed + && ((unsigned) exact_log2 (INTVAL (operands[3])) + < GET_MODE_NUNITS (mode))" + [(set (match_dup 0) + (vec_duplicate:V2FI_32 (match_dup 2))) + (set (match_dup 0) + (vec_merge:V2FI_32 (match_dup 1) (match_dup 0) (match_dup 3)))] + "operands[3] = GEN_INT (~INTVAL (operands[3]) & 0x3);") + + (define_insn "*pinsrb" [(set (match_operand:V4QI 0 "register_operand" "=x,YW") (vec_merge:V4QI @@ -5561,6 +5665,39 @@ (define_insn "*pextrw" (set_attr "prefix" "maybe_vex") (set_attr "mode" "TI")]) +(define_insn "*pextrw" + [(set (match_operand: 0 "register_sse4nonimm_operand" "=?r,jm,m,x,Yw") + (vec_select: + (match_operand:V2F_32 1 "register_operand" "YW,YW,YW,0,YW") + (parallel [(match_operand:SI 2 "const_0_to_1_operand")])))] + "TARGET_SSE2" +{ + switch (which_alternative) + { + case 0: + return "%vpextrw\t{%2, %1, %k0|%k0, %1, %2}"; + case 1: + return "pextrw\t{%2, %1, %0|%0, %1, %2}"; + case 2: + return "vpextrw\t{%2, %1, %0|%0, %1, %2}"; + case 3: + operands[2] = GEN_INT (INTVAL (operands[2]) * 2); + return "psrldq\t{%2, %0|%0, %2}"; + case 4: + operands[2] = GEN_INT (INTVAL (operands[2]) * 2); + return "vpsrldq\t{%2, %1, %0|%0, %1, %2}"; + + default: + gcc_unreachable (); + } +} + [(set_attr "isa" "*,sse4_noavx,avx,noavx,avx") + (set_attr "addr" "*,gpr16,*,*,*") + (set_attr "type" "sselog1,sselog1,sselog1,sseishft1,sseishft1") + (set_attr "length_immediate" "1") + (set_attr "prefix" "maybe_vex,orig,maybe_evex,orig,maybe_evex") + (set_attr "mode" "TI")]) + (define_insn "*pextrw_zext" [(set (match_operand:SWI48 0 "register_operand" "=r") (zero_extend:SWI48 @@ -5608,9 +5745,9 @@ (define_insn "*pextrb_zext" (set_attr "prefix" "maybe_vex") (set_attr "mode" "TI")]) -(define_expand "vec_setv2hi" - [(match_operand:V2HI 0 "register_operand") - (match_operand:HI 1 "register_operand") +(define_expand "vec_set" + [(match_operand:V2FI_32 0 "register_operand") + (match_operand: 1 "register_operand") (match_operand 2 "vec_setm_sse41_operand")] "TARGET_SSE2" { @@ -5622,9 +5759,9 @@ (define_expand "vec_setv2hi" DONE; }) -(define_expand "vec_extractv2hihi" - [(match_operand:HI 0 "register_operand") - (match_operand:V2HI 1 "register_operand") +(define_expand "vec_extract" + [(match_operand: 0 "register_operand") + (match_operand:V2FI_32 1 "register_operand") (match_operand 2 "const_int_operand")] "TARGET_SSE2" { @@ -5659,29 +5796,29 @@ (define_expand "vec_extractv4qiqi" }) (define_insn_and_split "*punpckwd" - [(set (match_operand:V2HI 0 "register_operand" "=x,Yw") - (vec_select:V2HI - (vec_concat:V4HI - (match_operand:V2HI 1 "register_operand" "0,Yw") - (match_operand:V2HI 2 "register_operand" "x,Yw")) + [(set (match_operand:V2FI_32 0 "register_operand" "=x,Yw") + (vec_select:V2FI_32 + (vec_concat: + (match_operand:V2FI_32 1 "register_operand" "0,Yw") + (match_operand:V2FI_32 2 "register_operand" "x,Yw")) (parallel [(match_operand 3 "const_0_to_3_operand") (match_operand 4 "const_0_to_3_operand")])))] "TARGET_SSE2" "#" "&& reload_completed" [(set (match_dup 5) - (vec_select:V8HI + (vec_select: (match_dup 5) (parallel [(match_dup 3) (match_dup 4) (const_int 2) (const_int 3) (const_int 4) (const_int 5) (const_int 6) (const_int 7)])))] { - rtx dest = lowpart_subreg (V8HImode, operands[0], V2HImode); - rtx op1 = lowpart_subreg (V8HImode, operands[1], V2HImode); - rtx op2 = lowpart_subreg (V8HImode, operands[2], V2HImode); + rtx dest = lowpart_subreg (mode, operands[0], mode); + rtx op1 = lowpart_subreg (mode, operands[1], mode); + rtx op2 = lowpart_subreg (mode, operands[2], mode); - emit_insn (gen_vec_interleave_lowv8hi (dest, op1, op2)); + emit_insn (gen_vec_interleave_low (dest, op1, op2)); static const int map[4] = { 0, 2, 1, 3 }; @@ -5699,10 +5836,10 @@ (define_insn_and_split "*punpckwd" (set_attr "type" "sselog") (set_attr "mode" "TI")]) -(define_insn "*pshufw_1" - [(set (match_operand:V2HI 0 "register_operand" "=Yw") - (vec_select:V2HI - (match_operand:V2HI 1 "register_operand" "Yw") +(define_insn "*pshufw_1" + [(set (match_operand:V2FI_32 0 "register_operand" "=Yw") + (vec_select:V2FI_32 + (match_operand:V2FI_32 1 "register_operand" "Yw") (parallel [(match_operand 2 "const_0_to_1_operand") (match_operand 3 "const_0_to_1_operand")])))] "TARGET_SSE2" @@ -5731,8 +5868,18 @@ (define_insn "*vec_dupv2hi" (set_attr "length_immediate" "1") (set_attr "mode" "TI")]) -(define_expand "vec_initv2hihi" - [(match_operand:V2HI 0 "register_operand") +(define_insn "*vec_dup" + [(set (match_operand:V2F_32 0 "register_operand" "=Yw") + (vec_duplicate:V2F_32 + (match_operand: 1 "register_operand" "Yw")))] + "TARGET_SSE2" + "%vpshuflw\t{$0, %1, %0|%0, %1, 0}" + [(set_attr "type" "sselog1") + (set_attr "length_immediate" "1") + (set_attr "mode" "TI")]) + +(define_expand "vec_init" + [(match_operand:V2FI_32 0 "register_operand") (match_operand 1)] "TARGET_SSE2" { diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md index 33198756bb0..48a9bd99576 100644 --- a/gcc/config/i386/sse.md +++ b/gcc/config/i386/sse.md @@ -12372,9 +12372,9 @@ (define_insn_and_split "*vec_extract_0" "operands[1] = gen_lowpart (mode, operands[1]);") (define_insn "*vec_extract" - [(set (match_operand:HFBF 0 "register_sse4nonimm_operand" "=?r,jm,m,x,v") + [(set (match_operand:HFBF 0 "register_sse4nonimm_operand" "=?r,jm,m,x,Yw") (vec_select:HFBF - (match_operand: 1 "register_operand" "v,x,v,0,v") + (match_operand: 1 "register_operand" "v,x,v,0,YW") (parallel [(match_operand:SI 2 "const_0_to_7_operand")])))] "TARGET_SSE2" diff --git a/gcc/testsuite/gcc.target/i386/part-vect-vec_elem-1.c b/gcc/testsuite/gcc.target/i386/part-vect-vec_elem-1.c new file mode 100644 index 00000000000..dba98aa4810 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/part-vect-vec_elem-1.c @@ -0,0 +1,135 @@ +/* { dg-do run { target { ! ia32 } } } */ +/* { dg-options "-O1 -msse4.1" } */ +/* { dg-require-effective-target sse4 } */ + +#include "sse4_1-check.h" + +typedef _Float16 v4hf __attribute__((vector_size(8))); + +v4hf +__attribute__((noipa)) +vector_init_dupv4hf (_Float16 a) +{ + return __extension__(v4hf){a, a, a, a}; +} + +v4hf +__attribute__((noipa)) +vector_init_allzero (_Float16 a) +{ + return __extension__(v4hf){0, 0, 0, 0}; +} + +v4hf +__attribute__((noipa)) +vector_init_one_nonzero (_Float16 a) +{ + return __extension__(v4hf){0, 0, a, 0}; +} + +v4hf +__attribute__((noipa)) +vector_init_one_var (_Float16 a) +{ + return __extension__(v4hf){1, 2, a, 4}; +} + +v4hf +__attribute__((noipa)) +vector_init_general (_Float16 a, _Float16 a1, _Float16 a2, _Float16 a3) +{ + return __extension__(v4hf){a3, a2, a1, a}; +} + +v4hf +__attribute__((noipa)) +vec_set (_Float16 a, v4hf b) +{ + b[1] = a; + return b; +} + +v4hf +__attribute__((noipa)) +vec_set_var (_Float16 a, v4hf b, int c) +{ + b[c] = a; + return b; +} + +_Float16 +__attribute__((noipa)) +vec_extract (v4hf b) +{ + return b[2]; +} + +static void +sse4_1_test () +{ + typedef union { + _Float16 a[4]; + v4hf x;}union64hf; + union64hf res, exp, src; + + res.x = vector_init_dupv4hf (1.0f16); + for (int i = 0; i != 4; i++) + exp.a[i] = 1.0f16; + if (__builtin_memcmp (&res.a[0], &exp.a[0], 8) != 0) + __builtin_abort (); + + res.x = vector_init_allzero (1.0f16); + for (int i = 0; i != 4; i++) + exp.a[i] = 0.0f16; + if (__builtin_memcmp (&res.a[0], &exp.a[0], 8) != 0) + __builtin_abort (); + + res.x = vector_init_one_nonzero (1.0f16); + for (int i = 0; i != 4; i++) + exp.a[i] = 0.0f16; + exp.a[2] = 1.0f16; + if (__builtin_memcmp (&res.a[0], &exp.a[0], 8) != 0) + __builtin_abort (); + + res.x = vector_init_one_var (3.0f16); + for (int i = 0; i != 4; i++) + exp.a[i] = i + 1; + if (__builtin_memcmp (&res.a[0], &exp.a[0], 8) != 0) + __builtin_abort (); + + res.x = vector_init_general (4.0, 3.0f, 2.0f, 1.0); + for (int i = 0; i != 4; i++) + exp.a[i] = 1 + i; + if (__builtin_memcmp (&res.a[0], &exp.a[0], 8) != 0) + __builtin_abort (); + + for (int i = 0; i != 4; i++) + { + src.a[i] = i; + exp.a[i] = i; + } + res.x = vec_set (3.0f, src.x); + exp.a[1] = 3.0f; + if (__builtin_memcmp (&res.a[0], &exp.a[0], 8) != 0) + __builtin_abort (); + + for (int i = 0; i != 4; i++) + { + src.a[i] = i; + exp.a[i] = i; + } + res.x = vec_set_var (3.0f, src.x, 1); + exp.a[1] = 3.0f; + if (__builtin_memcmp (&res.a[0], &exp.a[0], 8) != 0) + __builtin_abort (); + + for (int i = 0; i != 4; i++) + { + src.a[i] = i; + exp.a[i] = i; + } + _Float16 res_scalar = vec_extract (src.x); + if (res_scalar != 2.0f) + __builtin_abort (); + return ; +} diff --git a/gcc/testsuite/gcc.target/i386/part-vect-vec_elem-2.c b/gcc/testsuite/gcc.target/i386/part-vect-vec_elem-2.c new file mode 100644 index 00000000000..cc195638bff --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/part-vect-vec_elem-2.c @@ -0,0 +1,135 @@ +/* { dg-do run { target { ! ia32 } } } */ +/* { dg-options "-O1 -msse4.1" } */ +/* { dg-require-effective-target sse4 } */ + +#include "sse4_1-check.h" + +typedef _Float16 v2hf __attribute__((vector_size(4))); + +v2hf +__attribute__((noipa)) +vector_init_dupv2hf (_Float16 a) +{ + return __extension__(v2hf){a, a}; +} + +v2hf +__attribute__((noipa)) +vector_init_allzero (_Float16 a) +{ + return __extension__(v2hf){0, 0}; +} + +v2hf +__attribute__((noipa)) +vector_init_one_nonzero (_Float16 a) +{ + return __extension__(v2hf){0, a}; +} + +v2hf +__attribute__((noipa)) +vector_init_one_var (_Float16 a) +{ + return __extension__(v2hf){1, a}; +} + +v2hf +__attribute__((noipa)) +vector_init_general (_Float16 a1, _Float16 a2) +{ + return __extension__(v2hf){a2, a1}; +} + +v2hf +__attribute__((noipa)) +vec_set (_Float16 a, v2hf b) +{ + b[1] = a; + return b; +} + +v2hf +__attribute__((noipa)) +vec_set_var (_Float16 a, v2hf b, int c) +{ + b[c] = a; + return b; +} + +_Float16 +__attribute__((noipa)) +vec_extract (v2hf b) +{ + return b[1]; +} + +static void +sse4_1_test () +{ + typedef union { + _Float16 a[2]; + v2hf x;}union64hf; + union64hf res, exp, src; + + res.x = vector_init_dupv2hf (1.0f16); + for (int i = 0; i != 2; i++) + exp.a[i] = 1.0f16; + if (__builtin_memcmp (&res.a[0], &exp.a[0], 4) != 0) + __builtin_abort (); + + res.x = vector_init_allzero (1.0f16); + for (int i = 0; i != 2; i++) + exp.a[i] = 0.0f16; + if (__builtin_memcmp (&res.a[0], &exp.a[0], 4) != 0) + __builtin_abort (); + + res.x = vector_init_one_nonzero (1.0f16); + for (int i = 0; i != 2; i++) + exp.a[i] = 0.0f16; + exp.a[1] = 1.0f16; + if (__builtin_memcmp (&res.a[0], &exp.a[0], 4) != 0) + __builtin_abort (); + + res.x = vector_init_one_var (3.0f16); + exp.a[0] = 1; + exp.a[1] = 3; + if (__builtin_memcmp (&res.a[0], &exp.a[0], 4) != 0) + __builtin_abort (); + + res.x = vector_init_general (2.0f, 1.0); + for (int i = 0; i != 2; i++) + exp.a[i] = 1 + i; + if (__builtin_memcmp (&res.a[0], &exp.a[0], 4) != 0) + __builtin_abort (); + + for (int i = 0; i != 2; i++) + { + src.a[i] = i; + exp.a[i] = i; + } + res.x = vec_set (3.0f, src.x); + exp.a[1] = 3.0f; + if (__builtin_memcmp (&res.a[0], &exp.a[0], 4) != 0) + __builtin_abort (); + + for (int i = 0; i != 2; i++) + { + src.a[i] = i; + exp.a[i] = i; + } + res.x = vec_set_var (3.0f, src.x, 1); + exp.a[1] = 3.0f; + if (__builtin_memcmp (&res.a[0], &exp.a[0], 4) != 0) + __builtin_abort (); + + for (int i = 0; i != 2; i++) + { + src.a[i] = i; + exp.a[i] = i; + } + _Float16 res_scalar = vec_extract (src.x); + if (res_scalar != 1.0f) + __builtin_abort (); + return ; +}