From patchwork Tue Nov 1 15:28:34 2016 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Claudiu Zissulescu X-Patchwork-Id: 689962 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@bilbo.ozlabs.org Received: from sourceware.org (server1.sourceware.org [209.132.180.131]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ozlabs.org (Postfix) with ESMTPS id 3t7Zw64zyGz9t2N for ; Wed, 2 Nov 2016 02:31:50 +1100 (AEDT) Authentication-Results: ozlabs.org; dkim=pass (1024-bit key; unprotected) header.d=gcc.gnu.org header.i=@gcc.gnu.org header.b=s/Czhf5K; dkim-atps=neutral DomainKey-Signature: a=rsa-sha1; c=nofws; d=gcc.gnu.org; h=list-id :list-unsubscribe:list-archive:list-post:list-help:sender:from :to:cc:subject:date:message-id:in-reply-to:references :mime-version:content-type; q=dns; s=default; b=VCkirkcdwByHTrL/ 9OOdQNcWQyxnEubnrbdDCM89BL30Z3qdHUyjK5g/7ULFeU/L1/Emeihann6s5ocg NipI5f2JCpYlEVUi4BJhYXeoGYDYQ3fn0oTNl8WGF6N2d+JKDRVQc8M4nQ7jeKgv ejMw55l04XH9sSTVX+HEfJI2ojs= DKIM-Signature: v=1; a=rsa-sha1; c=relaxed; d=gcc.gnu.org; h=list-id :list-unsubscribe:list-archive:list-post:list-help:sender:from :to:cc:subject:date:message-id:in-reply-to:references :mime-version:content-type; s=default; bh=WqODxEB5ChETgz3L65W87w MiD+0=; b=s/Czhf5KmxZYW6v4l6w7+Mu9QcvL7ZdY2zq8lRofhUY84lcVnJiPpv kmlzN06nJa412RbvksqBLtz/aaWjFu5ux8TyjHVmajdSXA84wGAEDAw+EnLx5HSN oYym06hOcx4IXRlxiUdYeAzr1sCPwIgJNqRVIUu6V107TCPYS4HDQ= Received: (qmail 55779 invoked by alias); 1 Nov 2016 15:31:39 -0000 Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Id: List-Unsubscribe: List-Archive: List-Post: List-Help: Sender: gcc-patches-owner@gcc.gnu.org Delivered-To: mailing list gcc-patches@gcc.gnu.org Received: (qmail 55764 invoked by uid 89); 1 Nov 2016 15:31:38 -0000 Authentication-Results: sourceware.org; auth=none X-Virus-Found: No X-Spam-SWARE-Status: No, score=-1.3 required=5.0 tests=AWL, BAYES_00, KAM_LAZY_DOMAIN_SECURITY, LIKELY_SPAM_BODY, RCVD_IN_DNSWL_NONE, RP_MATCHES_RCVD autolearn=no version=3.3.2 spammy=ww, Division, cal, incorporated X-HELO: smtprelay.synopsys.com Received: from smtprelay2.synopsys.com (HELO smtprelay.synopsys.com) (198.182.60.111) by sourceware.org (qpsmtpd/0.93/v0.84-503-g423c35a) with ESMTP; Tue, 01 Nov 2016 15:31:28 +0000 Received: from mailhost.synopsys.com (mailhost1.synopsys.com [10.12.238.239]) by smtprelay.synopsys.com (Postfix) with ESMTP id 80F8710C1366; Tue, 1 Nov 2016 08:31:26 -0700 (PDT) Received: from mailhost.synopsys.com (localhost [127.0.0.1]) by mailhost.synopsys.com (Postfix) with ESMTP id 68384A43; Tue, 1 Nov 2016 08:31:26 -0700 (PDT) Received: from US01WEHTC2.internal.synopsys.com (us01wehtc2-vip.internal.synopsys.com [10.12.239.238]) by mailhost.synopsys.com (Postfix) with ESMTP id 3AE9FA3E; Tue, 1 Nov 2016 08:31:26 -0700 (PDT) Received: from IN01WEHTCA.internal.synopsys.com (10.144.199.104) by US01WEHTC2.internal.synopsys.com (10.12.239.237) with Microsoft SMTP Server (TLS) id 14.3.266.1; Tue, 1 Nov 2016 08:30:17 -0700 Received: from IN01WEHTCB.internal.synopsys.com (10.144.199.105) by IN01WEHTCA.internal.synopsys.com (10.144.199.103) with Microsoft SMTP Server (TLS) id 14.3.266.1; Tue, 1 Nov 2016 21:00:14 +0530 Received: from nl20droid1.internal.synopsys.com (10.100.24.228) by IN01WEHTCB.internal.synopsys.com (10.144.199.243) with Microsoft SMTP Server (TLS) id 14.3.266.1; Tue, 1 Nov 2016 21:00:13 +0530 From: Claudiu Zissulescu To: , CC: , Subject: [PATCH] [ARC] Various small miscellaneous fixes. Date: Tue, 1 Nov 2016 16:28:34 +0100 Message-ID: <1478014114-25720-1-git-send-email-claziss@synopsys.com> In-Reply-To: <20160708104811.GK10229@embecosm.com> References: <20160708104811.GK10229@embecosm.com> MIME-Version: 1.0 This is an updated version of the patch that can be applied as is. Ok to apply? Claudiu gcc/ 2016-05-09 Claudiu Zissulescu * config/arc/arc.c (arc_process_double_reg_moves): Change. * config/arc/arc.md (movsi_insn): Disable unsupported move instructions for ARCv2 cores. (movdi): Use prepare_move_operands. (movsf, movdf): Use move_dest_operand predicate. (arc_process_double_reg_moves): Change. * config/arc/constraints.md (Chs): Enable when barrel shifter is present. * config/arc/fpu.md (divsf3): Change to divsf3_fpu. * config/arc/fpx.md (dexcl_3op_peep2_insn): Dx data register is also a destination. (dexcl_3op_peep2_insn_nores): Likewise. * config/arc/arc.h (SHIFT_COUNT_TRUNCATED): Define to one. (LINK_COMMAND_SPEC): Remove. --- gcc/config/arc/arc.c | 5 +---- gcc/config/arc/arc.h | 27 +++------------------------ gcc/config/arc/arc.md | 35 +++++++++++++++++++---------------- gcc/config/arc/constraints.md | 3 ++- gcc/config/arc/fpu.md | 6 ++++-- gcc/config/arc/fpx.md | 26 ++++++++++++-------------- 6 files changed, 41 insertions(+), 61 deletions(-) diff --git a/gcc/config/arc/arc.c b/gcc/config/arc/arc.c index 0e7b63d..c927d5b 100644 --- a/gcc/config/arc/arc.c +++ b/gcc/config/arc/arc.c @@ -9021,10 +9021,7 @@ arc_process_double_reg_moves (rtx *operands) rtx srcLow = simplify_gen_subreg (SImode, src, DFmode, TARGET_BIG_ENDIAN ? 4 : 0); - emit_insn (gen_rtx_UNSPEC_VOLATILE (Pmode, - gen_rtvec (3, dest, srcHigh, srcLow), - VUNSPEC_ARC_DEXCL_NORES)); - + emit_insn (gen_dexcl_2op (dest, srcHigh, srcLow)); } else gcc_unreachable (); diff --git a/gcc/config/arc/arc.h b/gcc/config/arc/arc.h index b146f3a..17285a7 100644 --- a/gcc/config/arc/arc.h +++ b/gcc/config/arc/arc.h @@ -128,24 +128,6 @@ along with GCC; see the file COPYING3. If not see %{!marclinux*: %{pg|p|profile:-marclinux_prof;: -marclinux}} \ %{!z:-z max-page-size=0x2000 -z common-page-size=0x2000} \ %{shared:-shared}" -/* Like the standard LINK_COMMAND_SPEC, but add %G when building - a shared library with -nostdlib, so that the hidden functions of libgcc - will be incorporated. - N.B., we don't want a plain -lgcc, as this would lead to re-exporting - non-hidden functions, so we have to consider libgcc_s.so.* first, which in - turn should be wrapped with --as-needed. */ -#define LINK_COMMAND_SPEC "\ -%{!fsyntax-only:%{!c:%{!M:%{!MM:%{!E:%{!S:\ - %(linker) %l " LINK_PIE_SPEC "%X %{o*} %{A} %{d} %{e*} %{m} %{N} %{n} %{r}\ - %{s} %{t} %{u*} %{x} %{z} %{Z} %{!A:%{!nostdlib:%{!nostartfiles:%S}}}\ - %{static:} %{L*} %(mfwrap) %(link_libgcc) %o\ - %{fopenacc|fopenmp|%:gt(%{ftree-parallelize-loops=*:%*} 1):\ - %:include(libgomp.spec)%(link_gomp)}\ - %(mflib)\ - %{fprofile-arcs|fprofile-generate|coverage:-lgcov}\ - %{!nostdlib:%{!nodefaultlibs:%(link_ssp) %(link_gcc_c_sequence)}}\ - %{!A:%{!nostdlib:%{!nostartfiles:%E}}} %{T*} }}}}}}" - #else #define LINK_SPEC "%{mbig-endian:-EB} %{EB} %{EL}\ %{pg|p:-marcelf_prof;mA7|mARC700|mcpu=arc700|mcpu=ARC700: -marcelf}" @@ -1570,13 +1552,10 @@ extern int arc_return_address_regs[4]; /* Undo the effects of the movmem pattern presence on STORE_BY_PIECES_P . */ #define MOVE_RATIO(SPEED) ((SPEED) ? 15 : 3) -/* Define this to be nonzero if shift instructions ignore all but the low-order - few bits. Changed from 1 to 0 for rotate pattern testcases - (e.g. 20020226-1.c). This change truncates the upper 27 bits of a word - while rotating a word. Came to notice through a combine phase - optimization viz. a << (32-b) is equivalent to a << (-b). +/* Define this to be nonzero if shift instructions ignore all but the + low-order few bits. */ -#define SHIFT_COUNT_TRUNCATED 0 +#define SHIFT_COUNT_TRUNCATED 1 /* Value is 1 if truncating an integer of INPREC bits to OUTPREC bits is done just by pretending it is already truncated. */ diff --git a/gcc/config/arc/arc.md b/gcc/config/arc/arc.md index e127d5b..7147fbd 100644 --- a/gcc/config/arc/arc.md +++ b/gcc/config/arc/arc.md @@ -704,9 +704,9 @@ ; the iscompact attribute allows the epilogue expander to know for which ; insns it should lengthen the return insn. ; N.B. operand 1 of alternative 7 expands into pcl,symbol@gotpc . -(define_insn "*movsi_insn" ; 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 - [(set (match_operand:SI 0 "move_dest_operand" "=Rcq,Rcq#q, w, h, w,w, w, w, w, w,???w, ?w, w,Rcq#q, w,Rcq, S, Us<,RcqRck,!*x, r,!*Rsd,!*Rcd,r,Ucm, Usd,m,???m,VUsc,VUsc") - (match_operand:SI 1 "move_src_operand" " cL, cP,Rcq#q,hCm1,cL,I,Crr,Clo,Chi,Cbi,?Rac,Cpc,Clb, ?Cal,?Cal, T,Rcq,RcqRck, Us>,Usd,Ucm, Usd, Ucd,m, w,!*Rzd,c,?Rac, Cm3, C32"))] +(define_insn "*movsi_insn" ; 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 + [(set (match_operand:SI 0 "move_dest_operand" "=Rcq,Rcq#q, w, h, w,w, w, w, w, w,???w, ?w, w,Rcq#q, w,Rcq, S, Us<,RcqRck,!*x, r,!*Rsd,!*Rcd,r,Ucm, Usd,m,???m,VUsc,VUsc") + (match_operand:SI 1 "move_src_operand" " cL, cP,Rcq#q,hPCm1,cL,I,Crr,Clo,Chi,Cbi,?Rac,Cpc,Clb, ?Cal,?Cal, T,Rcq,RcqRck, Us>,Usd,Ucm, Usd, Ucd,m, w,!*Rzd,c,?Rac, Cm3, C32"))] "register_operand (operands[0], SImode) || register_operand (operands[1], SImode) || (CONSTANT_P (operands[1]) @@ -751,7 +751,7 @@ ; of Crr to 4. (set_attr "length" "*,*,*,*,4,4,4,4,4,4,4,8,8,*,8,*,*,*,*,*,4,*,4,*,*,*,*,*,4,8") (set_attr "predicable" "yes,no,yes,no,yes,no,no,no,no,no,yes,no,no,yes,yes,no,no,no,no,no,no,no,no,no,no,no,no,no,no,no") - (set_attr "cpu_facility" "*,*,av1,av2,*,*,*,*,*,*,*,*,*,*,*,*,*,*,*,*,*,av2,av2,*,*,av2,*,*,av2,*")]) + (set_attr "cpu_facility" "av1,av1,av1,av2,*,*,*,*,*,*,*,*,*,*,*,*,*,*,*,*,*,av2,av2,*,*,av2,*,*,av2,*")]) ;; Sometimes generated by the epilogue code. We don't want to ;; recognize these addresses in general, because the limm is costly, @@ -1088,12 +1088,9 @@ (match_operand:DI 1 "general_operand" ""))] "" " -{ - /* Everything except mem = const or mem = mem can be done easily. */ - - if (GET_CODE (operands[0]) == MEM) - operands[1] = force_reg (DImode, operands[1]); -}") + if (prepare_move_operands (operands, DImode)) + DONE; + ") (define_insn_and_split "*movdi_insn" [(set (match_operand:DI 0 "move_dest_operand" "=w, w,r,m") @@ -1140,7 +1137,7 @@ ;; Floating point move insns. (define_expand "movsf" - [(set (match_operand:SF 0 "general_operand" "") + [(set (match_operand:SF 0 "move_dest_operand" "") (match_operand:SF 1 "general_operand" ""))] "" "if (prepare_move_operands (operands, SFmode)) DONE;") @@ -1161,7 +1158,7 @@ (set_attr "iscompact" "true,false,false,false,false")]) (define_expand "movdf" - [(set (match_operand:DF 0 "nonimmediate_operand" "") + [(set (match_operand:DF 0 "move_dest_operand" "") (match_operand:DF 1 "general_operand" ""))] "" "if (prepare_move_operands (operands, DFmode)) DONE;") @@ -1231,12 +1228,18 @@ ; second time to put back the contents which the first DEXCLx ; will have overwritten ; dexcl2 r0, r1, r0 - (set (match_dup 4) ; aka r0result - ; aka DF, r1, r0 - (unspec_volatile:SI [(match_dup 1) (match_dup 5) (match_dup 4)] VUNSPEC_ARC_DEXCL )) + (parallel [ + (set (match_dup 4) ; aka r0result + ; aka DF, r1, r0 + (unspec_volatile:SI [(match_dup 5) (match_dup 4)] + VUNSPEC_ARC_DEXCL)) + (clobber (match_dup 1)) + ]) ; Generate the second, which makes sure operand5 and operand4 values ; are put back in the Dx register properly. - (unspec_volatile:SI [(match_dup 1) (match_dup 5) (match_dup 4)] VUNSPEC_ARC_DEXCL_NORES ) + (set (match_dup 1) (unspec_volatile:DF + [(match_dup 5) (match_dup 4)] + VUNSPEC_ARC_DEXCL_NORES)) ; Note: we cannot use a (clobber (match_scratch)) here because ; the combine pass will end up replacing uses of it with 0 diff --git a/gcc/config/arc/constraints.md b/gcc/config/arc/constraints.md index b7bf2d3..8eccb2f 100644 --- a/gcc/config/arc/constraints.md +++ b/gcc/config/arc/constraints.md @@ -256,7 +256,8 @@ "@internal constant for a highpart that can be checked with a shift (asr.f 0,rn,m)" (and (match_code "const_int") - (match_test "IS_POWEROF2_P (-ival)"))) + (match_test "IS_POWEROF2_P (-ival)") + (match_test "TARGET_BARREL_SHIFTER"))) (define_constraint "Clo" "@internal diff --git a/gcc/config/arc/fpu.md b/gcc/config/arc/fpu.md index 9b0a65d..e43047c 100644 --- a/gcc/config/arc/fpu.md +++ b/gcc/config/arc/fpu.md @@ -193,11 +193,13 @@ (set_attr "type" "fpu")]) ;; Division -(define_insn "divsf3" +(define_insn "*divsf3_fpu" [(set (match_operand:SF 0 "register_operand" "=r,r,r,r,r") (div:SF (match_operand:SF 1 "nonmemory_operand" "0,r,0,r,F") (match_operand:SF 2 "nonmemory_operand" "r,r,F,F,r")))] - "TARGET_FP_SP_SQRT" + "TARGET_FP_SP_SQRT + && (register_operand (operands[1], SFmode) + || register_operand (operands[2], SFmode))" "fsdiv%? %0,%1,%2" [(set_attr "length" "4,4,8,8,8") (set_attr "iscompact" "false") diff --git a/gcc/config/arc/fpx.md b/gcc/config/arc/fpx.md index 2e11157..9ed2b7c 100644 --- a/gcc/config/arc/fpx.md +++ b/gcc/config/arc/fpx.md @@ -168,28 +168,26 @@ (set_attr "type" "lr")] ) - (define_insn "*dexcl_3op_peep2_insn" [(set (match_operand:SI 0 "dest_reg_operand" "=r") ; not register_operand, to accept SUBREG - (unspec_volatile:SI [ - (match_operand:DF 1 "arc_double_register_operand" "D") - (match_operand:SI 2 "shouldbe_register_operand" "r") ; r1 - (match_operand:SI 3 "shouldbe_register_operand" "r") ; r0 - ] VUNSPEC_ARC_DEXCL )) - ] + (unspec_volatile:SI + [(match_operand:SI 1 "shouldbe_register_operand" "r") ; r1 + (match_operand:SI 2 "shouldbe_register_operand" "r") ; r0 + ] VUNSPEC_ARC_DEXCL )) + (clobber (match_operand:DF 3 "arc_double_register_operand" "=&D"))] "TARGET_DPFP" - "dexcl%F1 %0, %2, %3" + "dexcl%F3 %0, %1, %2" [(set_attr "type" "move") (set_attr "length" "4")] ) ;; version which will not overwrite operand0 -(define_insn "*dexcl_3op_peep2_insn_nores" - [ (unspec_volatile:SI [ - (match_operand:DF 0 "arc_double_register_operand" "D") - (match_operand:SI 1 "shouldbe_register_operand" "r") ; r1 - (match_operand:SI 2 "shouldbe_register_operand" "r") ; r0 - ] VUNSPEC_ARC_DEXCL_NORES ) +(define_insn "dexcl_2op" + [(set (match_operand:DF 0 "arc_double_register_operand" "=D") + (unspec_volatile:DF + [(match_operand:SI 1 "shouldbe_register_operand" "r") ; r1 + (match_operand:SI 2 "shouldbe_register_operand" "r") ; r0 + ] VUNSPEC_ARC_DEXCL_NORES)) ] "TARGET_DPFP" "dexcl%F0 0, %1, %2"