From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 57633 invoked by alias); 14 Oct 2015 10:16:22 -0000 Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Id: List-Archive: List-Post: List-Help: Sender: gcc-patches-owner@gcc.gnu.org Received: (qmail 57620 invoked by uid 89); 14 Oct 2015 10:16:21 -0000 Authentication-Results: sourceware.org; auth=none X-Virus-Found: No X-Spam-SWARE-Status: No, score=-2.0 required=5.0 tests=AWL,BAYES_00,FREEMAIL_FROM,RCVD_IN_DNSWL_LOW,SPF_PASS autolearn=ham version=3.3.2 X-HELO: mail-yk0-f171.google.com Received: from mail-yk0-f171.google.com (HELO mail-yk0-f171.google.com) (209.85.160.171) by sourceware.org (qpsmtpd/0.93/v0.84-503-g423c35a) with (AES128-GCM-SHA256 encrypted) ESMTPS; Wed, 14 Oct 2015 10:16:19 +0000 Received: by ykaz22 with SMTP id z22so16852623yka.2 for ; Wed, 14 Oct 2015 03:16:17 -0700 (PDT) MIME-Version: 1.0 X-Received: by 10.129.107.197 with SMTP id g188mr1451819ywc.267.1444817777709; Wed, 14 Oct 2015 03:16:17 -0700 (PDT) Received: by 10.37.117.136 with HTTP; Wed, 14 Oct 2015 03:16:17 -0700 (PDT) In-Reply-To: References: <561CF734.7090802@redhat.com> Date: Wed, 14 Oct 2015 10:16:00 -0000 Message-ID: Subject: Re: [PATCH] Allow FSM to thread single block cases too From: Richard Biener To: Jeff Law , Jan Hubicka Cc: GCC Patches Content-Type: text/plain; charset=UTF-8 X-IsSubscribed: yes X-SW-Source: 2015-10/txt/msg01335.txt.bz2 On Tue, Oct 13, 2015 at 2:52 PM, Richard Biener wrote: > On Tue, Oct 13, 2015 at 2:21 PM, Jeff Law wrote: >> >> One of the cases that was missing in the FSM support is threading when the >> path is a single block. ie, a control statement's output can be statically >> determined just by looking at PHIs in the control statement's block for one >> or incoming edges. >> >> This is necessary to fix a regression if I turn off the old jump threader's >> backedge support. Just as important, Jan has in the past asked about a >> trivial jump threader to be run during early optimizations. Limiting the >> FSM bits to this case would likely satisfy that need in the future. > > I think he asked for trivial forward threads though due to repeated tests. > I hacked FRE to do this (I think), but maybe some trivial cleanup opportunities > are still left here. Honza? This or other related patches in the range r228731:228774 has caused a quite big jump in SPEC CPU 2000 binary sizes (notably 176.gcc - so maybe affecting bootstrap as well, at -O3). Are you sure this doesn't re-introduce DOM effectively peeling all loops once? Richard. > Richard. > >> Bootstrapped and regression tested on x86_64-linux-gnu. Installed on the >> trunk. >> >> Jeff >> >> commit a53bb29a1dffd329aa6235b88b0c2a830aa5a59e >> Author: Jeff Law >> Date: Tue Oct 13 06:19:20 2015 -0600 >> >> [PATCH] Allow FSM to thread single block cases too >> >> * tree-ssa-threadbackward.c >> (fsm_find_control_statement_thread_paths): >> Allow single block jump threading paths. >> >> * gcc.dg/tree-ssa/ssa-thread-13.c: New test. >> >> diff --git a/gcc/ChangeLog b/gcc/ChangeLog >> index d71bcd2..caab533 100644 >> --- a/gcc/ChangeLog >> +++ b/gcc/ChangeLog >> @@ -1,3 +1,8 @@ >> +2015-10-13 Jeff Law >> + >> + * tree-ssa-threadbackward.c >> (fsm_find_control_statement_thread_paths): >> + Allow single block jump threading paths. >> + >> 2015-10-13 Tom de Vries >> >> PR tree-optimization/67476 >> diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog >> index 4a08f0f..acf6df5 100644 >> --- a/gcc/testsuite/ChangeLog >> +++ b/gcc/testsuite/ChangeLog >> @@ -1,3 +1,7 @@ >> +2015-10-13 Jeff Law >> + >> + * gcc.dg/tree-ssa/ssa-thread-13.c: New test. >> + >> 2015-10-12 Jeff Law >> >> * gcc.dg/tree-ssa/ssa-thread-12.c: New test. >> diff --git a/gcc/testsuite/gcc.dg/tree-ssa/ssa-thread-13.c >> b/gcc/testsuite/gcc.dg/tree-ssa/ssa-thread-13.c >> new file mode 100644 >> index 0000000..5051d11 >> --- /dev/null >> +++ b/gcc/testsuite/gcc.dg/tree-ssa/ssa-thread-13.c >> @@ -0,0 +1,70 @@ >> +/* { dg-do compile } */ >> +/* { dg-options "-O2 -fdump-tree-vrp1-details" } */ >> +/* { dg-final { scan-tree-dump "FSM" "vrp1" } } */ >> + >> +typedef struct rtx_def *rtx; >> +typedef const struct rtx_def *const_rtx; >> +enum rtx_code >> +{ >> + UNKNOWN, VALUE, DEBUG_EXPR, EXPR_LIST, INSN_LIST, SEQUENCE, ADDRESS, >> + DEBUG_INSN, INSN, JUMP_INSN, CALL_INSN, BARRIER, CODE_LABEL, NOTE, >> + COND_EXEC, PARALLEL, ASM_INPUT, ASM_OPERANDS, UNSPEC, UNSPEC_VOLATILE, >> + ADDR_VEC, ADDR_DIFF_VEC, PREFETCH, SET, USE, CLOBBER, CALL, RETURN, >> + EH_RETURN, TRAP_IF, CONST_INT, CONST_FIXED, CONST_DOUBLE, CONST_VECTOR, >> + CONST_STRING, CONST, PC, REG, SCRATCH, SUBREG, STRICT_LOW_PART, CONCAT, >> + CONCATN, MEM, LABEL_REF, SYMBOL_REF, CC0, IF_THEN_ELSE, COMPARE, PLUS, >> + MINUS, NEG, MULT, SS_MULT, US_MULT, DIV, SS_DIV, US_DIV, MOD, UDIV, >> UMOD, >> + AND, IOR, XOR, NOT, ASHIFT, ROTATE, ASHIFTRT, LSHIFTRT, ROTATERT, SMIN, >> + SMAX, UMIN, UMAX, PRE_DEC, PRE_INC, POST_DEC, POST_INC, PRE_MODIFY, >> + POST_MODIFY, NE, EQ, GE, GT, LE, LT, GEU, GTU, LEU, LTU, UNORDERED, >> + ORDERED, UNEQ, UNGE, UNGT, UNLE, UNLT, LTGT, SIGN_EXTEND, ZERO_EXTEND, >> + TRUNCATE, FLOAT_EXTEND, FLOAT_TRUNCATE, FLOAT, FIX, UNSIGNED_FLOAT, >> + UNSIGNED_FIX, FRACT_CONVERT, UNSIGNED_FRACT_CONVERT, SAT_FRACT, >> + UNSIGNED_SAT_FRACT, ABS, SQRT, BSWAP, FFS, CLZ, CTZ, POPCOUNT, PARITY, >> + SIGN_EXTRACT, ZERO_EXTRACT, HIGH, LO_SUM, VEC_MERGE, VEC_SELECT, >> + VEC_CONCAT, VEC_DUPLICATE, SS_PLUS, US_PLUS, SS_MINUS, SS_NEG, US_NEG, >> + SS_ABS, SS_ASHIFT, US_ASHIFT, US_MINUS, SS_TRUNCATE, US_TRUNCATE, FMA, >> + VAR_LOCATION, DEBUG_IMPLICIT_PTR, ENTRY_VALUE, LAST_AND_UNUSED_RTX_CODE >> +}; >> +union rtunion_def >> +{ >> + rtx rt_rtx; >> +}; >> +typedef union rtunion_def rtunion; >> +struct rtx_def >> +{ >> + __extension__ enum rtx_code code:16; >> + union u >> + { >> + rtunion fld[1]; >> + } >> + u; >> +}; >> + >> +unsigned int rtx_cost (rtx, enum rtx_code, unsigned char); >> +rtx single_set_2 (const_rtx, rtx); >> + >> +unsigned >> +seq_cost (const_rtx seq, unsigned char speed) >> +{ >> + unsigned cost = 0; >> + rtx set; >> + for (; seq; seq = (((seq)->u.fld[2]).rt_rtx)) >> + { >> + set = >> + (((((enum rtx_code) (seq)->code) == INSN) >> + || (((enum rtx_code) (seq)->code) == DEBUG_INSN) >> + || (((enum rtx_code) (seq)->code) == JUMP_INSN) >> + || (((enum rtx_code) (seq)->code) == >> + CALL_INSN)) ? (((enum rtx_code) ((((seq)->u.fld[4]).rt_rtx))-> >> + code) == >> + SET ? (((seq)->u.fld[4]). >> + rt_rtx) : single_set_2 (seq, >> + (((seq)->u. >> + fld[4]). >> + rt_rtx))) : >> (rtx) >> + 0); >> + if (set) >> + cost += rtx_cost (set, SET, speed); >> + } >> +} >> diff --git a/gcc/tree-ssa-threadbackward.c b/gcc/tree-ssa-threadbackward.c >> index 5be6ee4..9128094 100644 >> --- a/gcc/tree-ssa-threadbackward.c >> +++ b/gcc/tree-ssa-threadbackward.c >> @@ -211,10 +211,6 @@ fsm_find_control_statement_thread_paths (tree name, >> continue; >> >> int path_length = path->length (); >> - /* A path with less than 2 basic blocks should not be jump-threaded. >> */ >> - if (path_length < 2) >> - continue; >> - >> if (path_length > PARAM_VALUE (PARAM_MAX_FSM_THREAD_LENGTH)) >> { >> if (dump_file && (dump_flags & TDF_DETAILS)) >>