From: Mathieu Desnoyers Date: Fri, 15 May 2009 20:31:18 +0000 (-0400) Subject: Update out of order memory models to include instruction scheduling X-Git-Tag: v0.1~223 X-Git-Url: https://git.lttng.org./?a=commitdiff_plain;h=dfa8abef9f88fd28e232b0eab9a105fc50e71959;p=userspace-rcu.git Update out of order memory models to include instruction scheduling Signed-off-by: Mathieu Desnoyers --- diff --git a/formal-model/ooomem-double-update-minimal/DEFINES b/formal-model/ooomem-double-update-minimal/DEFINES deleted file mode 100644 index 7cf7b24..0000000 --- a/formal-model/ooomem-double-update-minimal/DEFINES +++ /dev/null @@ -1,3 +0,0 @@ -#define read_one_is_one (read_one == 1) -#define read_two_is_one (read_two == 1) -#define read_two_done (read_two != 2) diff --git a/formal-model/ooomem-double-update-minimal/Makefile b/formal-model/ooomem-double-update-minimal/Makefile deleted file mode 100644 index 37422b5..0000000 --- a/formal-model/ooomem-double-update-minimal/Makefile +++ /dev/null @@ -1,93 +0,0 @@ -# This program is free software; you can redistribute it and/or modify -# it under the terms of the GNU General Public License as published by -# the Free Software Foundation; either version 2 of the License, or -# (at your option) any later version. -# -# This program is distributed in the hope that it will be useful, -# but WITHOUT ANY WARRANTY; without even the implied warranty of -# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the -# GNU General Public License for more details. -# -# You should have received a copy of the GNU General Public License -# along with this program; if not, write to the Free Software -# Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. -# -# Copyright (C) Mathieu Desnoyers, 2009 -# -# Authors: Mathieu Desnoyers - -#CFLAGS=-DSAFETY -#CFLAGS=-DHASH64 -DREACH -CFLAGS=-DHASH64 - -#try pan -i to get the smallest trace. - -SPINFILE=mem.spin - -default: - make read_order | tee read_order.log - make read_order_no_wmb | tee read_order_no_wmb.log - make read_order_no_rmb | tee read_order_no_rmb.log - make asserts | tee asserts.log - make summary - -#show trail : spin -v -t -N pan.ltl input.spin -# after each individual make. - -summary: - @echo - @echo "Verification summary" - @grep error *.log - -asserts: clean - cat DEFINES > .input.spin - cat ${SPINFILE} >> .input.spin - rm -f .input.spin.trail - spin -a -X .input.spin - gcc -w ${CFLAGS} -DSAFETY -o pan pan.c - ./pan -v -c1 -X -m10000 -w19 - cp .input.spin $@.spin.input - -cp .input.spin.trail $@.spin.input.trail - -read_order: clean read_order_ltl run - cp .input.spin $@.spin.input - -cp .input.spin.trail $@.spin.input.trail - -read_order_no_rmb: clean read_order_ltl read_order_no_rmb_define run - cp .input.spin $@.spin.input - -cp .input.spin.trail $@.spin.input.trail - -read_order_no_rmb_define: - cp read_order_no_rmb.define .input.define - -read_order_no_wmb: clean read_order_ltl read_order_no_wmb_define run - cp .input.spin $@.spin.input - -cp .input.spin.trail $@.spin.input.trail - -read_order_no_wmb_define: - cp read_order_no_wmb.define .input.define - -read_order_ltl: - touch .input.define - cat DEFINES > pan.ltl - cat .input.define >> pan.ltl - spin -f "!(`cat read_order.ltl | grep -v ^//`)" >> pan.ltl - -run: pan - ./pan -a -v -c1 -X -m10000 -w19 - -pan: pan.c - gcc -w ${CFLAGS} -o pan pan.c - -pan.c: pan.ltl ${SPINFILE} - cat DEFINES > .input.spin - cat .input.define >> .input.spin - cat ${SPINFILE} >> .input.spin - rm -f .input.spin.trail - spin -a -X -N pan.ltl .input.spin - -.PHONY: clean default distclean summary -clean: - rm -f pan* trail.out .input.spin* *.spin.trail .input.define -distclean: - rm -f *.trail *.input *.log diff --git a/formal-model/ooomem-double-update-minimal/mem.sh b/formal-model/ooomem-double-update-minimal/mem.sh deleted file mode 100644 index 56e8123..0000000 --- a/formal-model/ooomem-double-update-minimal/mem.sh +++ /dev/null @@ -1,29 +0,0 @@ -#!/bin/sh -# -# Compiles and runs the urcu.spin Promela model. -# -# This program is free software; you can redistribute it and/or modify -# it under the terms of the GNU General Public License as published by -# the Free Software Foundation; either version 2 of the License, or -# (at your option) any later version. -# -# This program is distributed in the hope that it will be useful, -# but WITHOUT ANY WARRANTY; without even the implied warranty of -# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the -# GNU General Public License for more details. -# -# You should have received a copy of the GNU General Public License -# along with this program; if not, write to the Free Software -# Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. -# -# Copyright (C) IBM Corporation, 2009 -# Mathieu Desnoyers, 2009 -# -# Authors: Paul E. McKenney -# Mathieu Desnoyers - -# Basic execution, without LTL clauses. See Makefile. - -spin -a mem.spin -cc -DSAFETY -o pan pan.c -./pan -v -c1 -X -m10000000 -w21 diff --git a/formal-model/ooomem-double-update-minimal/mem.spin b/formal-model/ooomem-double-update-minimal/mem.spin deleted file mode 100644 index bae92d6..0000000 --- a/formal-model/ooomem-double-update-minimal/mem.spin +++ /dev/null @@ -1,241 +0,0 @@ -/* - * mem.spin: Promela code to validate memory barriers with OOO memory. - * - * This program is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License as published by - * the Free Software Foundation; either version 2 of the License, or - * (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. - * - * You should have received a copy of the GNU General Public License - * along with this program; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. - * - * Copyright (c) 2009 Mathieu Desnoyers - */ - -/* Promela validation variables. */ - -/* - * Produced process data flow. Updated after each instruction to show which - * variables are ready. Assigned using SSA (static single assignment) (defuse - * analysis must be done on the program to map "real" variables to single define - * followed by use). Using one-hot bit encoding per variable to save state - * space. Used as triggers to execute the instructions having those variables - * as input. - */ - -#define PRODUCE_TOKENS(state, bits) \ - state = (state) | (bits) - -/* All bits must be active to consume. All notbits must be inactive. */ -/* Consuming a token does not clear it, it just waits for it. */ -#define CONSUME_TOKENS(state, bits, notbits) \ - ((!((state) & (notbits))) && ((state) & (bits)) == (bits)) - -#define CLEAR_TOKENS(state, bits) \ - state = (state) & ~(bits) - -#define NR_PROCS 2 - -#define get_pid() (_pid) - -/* - * Each process have its own data in cache. Caches are randomly updated. - * smp_wmb and smp_rmb forces cache updates (write and read), wmb_mb forces - * both. - */ - -#define DECLARE_CACHED_VAR(type, x, v) \ - type mem_##x = v; \ - type cached_##x[NR_PROCS] = v; \ - bit cache_dirty_##x[NR_PROCS] = 0; - -#define IS_CACHE_DIRTY(x, id) (cache_dirty_##x[id]) - -#define READ_CACHED_VAR(x) \ - (cached_##x[get_pid()]) - -#define WRITE_CACHED_VAR(x, v) \ - atomic { \ - cached_##x[get_pid()] = v; \ - cache_dirty_##x[get_pid()] = 1; \ - } - -#define CACHE_WRITE_TO_MEM(x, id) \ - if \ - :: IS_CACHE_DIRTY(x, id) -> \ - mem_##x = cached_##x[id]; \ - cache_dirty_##x[id] = 0; \ - :: else -> \ - skip \ - fi; - -#define CACHE_READ_FROM_MEM(x, id) \ - if \ - :: !IS_CACHE_DIRTY(x, id) -> \ - cached_##x[id] = mem_##x; \ - :: else -> \ - skip \ - fi; - -/* - * May update other caches if cache is dirty, or not. - */ -#define RANDOM_CACHE_WRITE_TO_MEM(x, id) \ - if \ - :: 1 -> CACHE_WRITE_TO_MEM(x, id); \ - :: 1 -> skip \ - fi; - -#define RANDOM_CACHE_READ_FROM_MEM(x, id)\ - if \ - :: 1 -> CACHE_READ_FROM_MEM(x, id); \ - :: 1 -> skip \ - fi; - -inline ooo_mem() -{ - atomic { - RANDOM_CACHE_WRITE_TO_MEM(alpha, get_pid()); - RANDOM_CACHE_WRITE_TO_MEM(beta, get_pid()); - RANDOM_CACHE_READ_FROM_MEM(alpha, get_pid()); - RANDOM_CACHE_READ_FROM_MEM(beta, get_pid()); - } -} - -/* must consume all prior read tokens */ -inline smp_rmb() -{ - atomic { - /* todo : consume all read tokens .. ? */ - CACHE_READ_FROM_MEM(alpha, get_pid()); - CACHE_READ_FROM_MEM(beta, get_pid()); - } -} - -/* must consume all prior write tokens */ -inline smp_wmb() -{ - atomic { - CACHE_WRITE_TO_MEM(alpha, get_pid()); - CACHE_WRITE_TO_MEM(beta, get_pid()); - } -} - -/* sync_core() must consume all prior read and write tokens, including rmb/wmb - * tokens */ - -/* must consume all prior read and write tokens */ -inline smp_mb() -{ - atomic { - smp_wmb(); - /* sync_core() */ - smp_rmb(); - } -} - -/* Keep in sync manually with smp_rmb, wmp_wmb and ooo_mem */ -DECLARE_CACHED_VAR(byte, alpha, 0); -DECLARE_CACHED_VAR(byte, beta, 0); - -/* value 2 is uninitialized */ -byte read_one = 2; -byte read_two = 2; - -/* - * Bit encoding, proc_one_produced : - */ - -#define P1_PROD_NONE (1 << 0) - -#define P1_READ_ONE (1 << 1) -#define P1_RMB (1 << 2) -#define P1_READ_TWO (1 << 3) - -int proc_one_produced; - -active proctype test_proc_one() -{ - assert(get_pid() < NR_PROCS); - - PRODUCE_TOKENS(proc_one_produced, P1_PROD_NONE); -#ifdef NO_RMB - PRODUCE_TOKENS(proc_one_produced, P1_RMB); -#endif - - do - :: CONSUME_TOKENS(proc_one_produced, P1_PROD_NONE, P1_READ_ONE) -> - ooo_mem(); - read_one = READ_CACHED_VAR(beta); - ooo_mem(); - PRODUCE_TOKENS(proc_one_produced, P1_READ_ONE); - :: CONSUME_TOKENS(proc_one_produced, P1_READ_ONE, P1_RMB) -> - smp_rmb(); - PRODUCE_TOKENS(proc_one_produced, P1_RMB); - :: CONSUME_TOKENS(proc_one_produced, P1_RMB, P1_READ_TWO) -> - ooo_mem(); - read_two = READ_CACHED_VAR(alpha); - ooo_mem(); - PRODUCE_TOKENS(proc_one_produced, P1_READ_TWO); - :: CONSUME_TOKENS(proc_one_produced, P1_PROD_NONE | P1_READ_ONE - | P1_RMB | P1_READ_TWO, 0) -> - break; - od; - - //CLEAR_TOKENS(proc_one_produced, - // P1_PROD_NONE | P1_READ_ONE | P1_RMB | P2_READ_TWO); - - // test : [] (read_one == 1 -> read_two == 1) - assert(read_one != 1 || read_two == 1); -} - - -/* - * Bit encoding, proc_two_produced : - */ - -#define P2_PROD_NONE (1 << 0) - -#define P2_WRITE_ONE (1 << 1) -#define P2_WMB (1 << 2) -#define P2_WRITE_TWO (1 << 3) - -int proc_two_produced; - -active proctype test_proc_two() -{ - assert(get_pid() < NR_PROCS); - - PRODUCE_TOKENS(proc_two_produced, P2_PROD_NONE); -#ifdef NO_WMB - PRODUCE_TOKENS(proc_two_produced, P2_WMB); -#endif - - do - :: CONSUME_TOKENS(proc_two_produced, P2_PROD_NONE, P2_WRITE_ONE) -> - ooo_mem(); - WRITE_CACHED_VAR(alpha, 1); - ooo_mem(); - PRODUCE_TOKENS(proc_two_produced, P2_WRITE_ONE); - :: CONSUME_TOKENS(proc_two_produced, P2_WRITE_ONE, P2_WMB) -> - smp_wmb(); - PRODUCE_TOKENS(proc_two_produced, P2_WMB); - :: CONSUME_TOKENS(proc_two_produced, P2_WMB, P2_WRITE_TWO) -> - ooo_mem(); - WRITE_CACHED_VAR(beta, 1); - ooo_mem(); - PRODUCE_TOKENS(proc_two_produced, P2_WRITE_TWO); - :: CONSUME_TOKENS(proc_two_produced, P2_PROD_NONE | P2_WRITE_ONE - | P2_WMB | P2_WRITE_TWO, 0) -> - break; - od; - - //CLEAR_TOKENS(proc_two_produced, - // P2_PROD_NONE | P2_WRITE_ONE | P2_WMB | P2_WRITE_TWO); -} diff --git a/formal-model/ooomem-double-update-minimal/read_order.ltl b/formal-model/ooomem-double-update-minimal/read_order.ltl deleted file mode 100644 index dec26f2..0000000 --- a/formal-model/ooomem-double-update-minimal/read_order.ltl +++ /dev/null @@ -1 +0,0 @@ -[] (read_one_is_one -> (!read_two_done || read_two_is_one)) diff --git a/formal-model/ooomem-double-update-minimal/read_order_no_rmb.define b/formal-model/ooomem-double-update-minimal/read_order_no_rmb.define deleted file mode 100644 index 73e61a4..0000000 --- a/formal-model/ooomem-double-update-minimal/read_order_no_rmb.define +++ /dev/null @@ -1 +0,0 @@ -#define NO_RMB diff --git a/formal-model/ooomem-double-update-minimal/read_order_no_wmb.define b/formal-model/ooomem-double-update-minimal/read_order_no_wmb.define deleted file mode 100644 index 710f29d..0000000 --- a/formal-model/ooomem-double-update-minimal/read_order_no_wmb.define +++ /dev/null @@ -1 +0,0 @@ -#define NO_WMB diff --git a/formal-model/ooomem-double-update-minimal/references.txt b/formal-model/ooomem-double-update-minimal/references.txt deleted file mode 100644 index ca6798f..0000000 --- a/formal-model/ooomem-double-update-minimal/references.txt +++ /dev/null @@ -1,7 +0,0 @@ -http://spinroot.com/spin/Man/ltl.html -http://en.wikipedia.org/wiki/Linear_temporal_logic -http://www.dcs.gla.ac.uk/~muffy/MRS4-2002/lect11.ppt - -http://www.lsv.ens-cachan.fr/~gastin/ltl2ba/index.php -http://spinroot.com/spin/Man/index.html -http://spinroot.com/spin/Man/promela.html diff --git a/formal-model/ooomem-double-update/DEFINES b/formal-model/ooomem-double-update/DEFINES index e74150d..7cf7b24 100644 --- a/formal-model/ooomem-double-update/DEFINES +++ b/formal-model/ooomem-double-update/DEFINES @@ -1,2 +1,3 @@ -#define read_one_is_zero (read_one == 0) -#define read_two_is_zero (read_two == 0) +#define read_one_is_one (read_one == 1) +#define read_two_is_one (read_two == 1) +#define read_two_done (read_two != 2) diff --git a/formal-model/ooomem-double-update/mem.spin b/formal-model/ooomem-double-update/mem.spin index 4c5d6fd..48c40f1 100644 --- a/formal-model/ooomem-double-update/mem.spin +++ b/formal-model/ooomem-double-update/mem.spin @@ -20,8 +20,23 @@ /* Promela validation variables. */ -#define NR_READERS 1 -#define NR_WRITERS 1 +/* + * Produced process control and data flow. Updated after each instruction to + * show which variables are ready. Using one-hot bit encoding per variable to + * save state space. Used as triggers to execute the instructions having those + * variables as input. Leaving bits active to inhibit instruction execution. + * Scheme used to make instruction disabling and automatic dependency fall-back + * automatic. + */ + +#define CONSUME_TOKENS(state, bits, notbits) \ + ((!(state & (notbits))) && (state & (bits)) == (bits)) + +#define PRODUCE_TOKENS(state, bits) \ + state = state | (bits); + +#define CLEAR_TOKENS(state, bits) \ + state = state & ~(bits) #define NR_PROCS 2 @@ -33,17 +48,18 @@ * both. */ -#define DECLARE_CACHED_VAR(type, x, v) \ - type mem_##x = v; \ - type cached_##x[NR_PROCS] = v; \ - bit cache_dirty_##x[NR_PROCS] = 0 +#define DECLARE_CACHED_VAR(type, x, v) \ + type mem_##x = v; \ + type cached_##x[NR_PROCS] = v; \ + bit cache_dirty_##x[NR_PROCS] = 0; #define IS_CACHE_DIRTY(x, id) (cache_dirty_##x[id]) -#define READ_CACHED_VAR(x) (cached_##x[get_pid()]) +#define READ_CACHED_VAR(x) \ + (cached_##x[get_pid()]) -#define WRITE_CACHED_VAR(x, v) \ - atomic { \ +#define WRITE_CACHED_VAR(x, v) \ + atomic { \ cached_##x[get_pid()] = v; \ cache_dirty_##x[get_pid()] = 1; \ } @@ -57,37 +73,50 @@ skip \ fi; -#define CACHE_READ_FROM_MEM(x, id) \ - if \ - :: !IS_CACHE_DIRTY(x, id) -> \ - cached_##x[id] = mem_##x;\ - :: else -> \ - skip \ +#define CACHE_READ_FROM_MEM(x, id) \ + if \ + :: !IS_CACHE_DIRTY(x, id) -> \ + cached_##x[id] = mem_##x; \ + :: else -> \ + skip \ fi; /* * May update other caches if cache is dirty, or not. */ -#define RANDOM_CACHE_WRITE_TO_MEM(x, id)\ - if \ +#define RANDOM_CACHE_WRITE_TO_MEM(x, id) \ + if \ :: 1 -> CACHE_WRITE_TO_MEM(x, id); \ - :: 1 -> skip \ + :: 1 -> skip \ fi; #define RANDOM_CACHE_READ_FROM_MEM(x, id)\ - if \ + if \ :: 1 -> CACHE_READ_FROM_MEM(x, id); \ - :: 1 -> skip \ + :: 1 -> skip \ fi; +inline ooo_mem() +{ + atomic { + RANDOM_CACHE_WRITE_TO_MEM(alpha, get_pid()); + RANDOM_CACHE_WRITE_TO_MEM(beta, get_pid()); + RANDOM_CACHE_READ_FROM_MEM(alpha, get_pid()); + RANDOM_CACHE_READ_FROM_MEM(beta, get_pid()); + } +} + +/* must consume all prior read tokens */ inline smp_rmb() { atomic { + /* todo : consume all read tokens .. ? */ CACHE_READ_FROM_MEM(alpha, get_pid()); CACHE_READ_FROM_MEM(beta, get_pid()); } } +/* must consume all prior write tokens */ inline smp_wmb() { atomic { @@ -96,10 +125,15 @@ inline smp_wmb() } } +/* sync_core() must consume all prior read and write tokens, including rmb/wmb + * tokens */ + +/* must consume all prior read and write tokens */ inline smp_mb() { atomic { smp_wmb(); + /* sync_core() */ smp_rmb(); } } @@ -108,60 +142,108 @@ inline smp_mb() DECLARE_CACHED_VAR(byte, alpha, 0); DECLARE_CACHED_VAR(byte, beta, 0); -inline ooo_mem() -{ - atomic { - RANDOM_CACHE_WRITE_TO_MEM(alpha, get_pid()); - RANDOM_CACHE_WRITE_TO_MEM(beta, get_pid()); - RANDOM_CACHE_READ_FROM_MEM(alpha, get_pid()); - RANDOM_CACHE_READ_FROM_MEM(beta, get_pid()); - } -} - /* value 2 is uninitialized */ byte read_one = 2; byte read_two = 2; +/* + * Bit encoding, proc_one_produced : + */ + +#define P1_PROD_NONE (1 << 0) + +#define P1_READ_ONE (1 << 1) +#define P1_RMB (1 << 2) +#define P1_READ_TWO (1 << 3) + +/* Only need a single color. */ +byte proc_one_produced; + active proctype test_proc_one() { assert(get_pid() < NR_PROCS); - ooo_mem(); - WRITE_CACHED_VAR(alpha, 1); - ooo_mem(); -#ifndef NO_WMB - smp_wmb(); - ooo_mem(); -#endif -#ifndef NO_RMB - smp_rmb(); - ooo_mem(); + PRODUCE_TOKENS(proc_one_produced, P1_PROD_NONE); +#ifdef NO_RMB + PRODUCE_TOKENS(proc_one_produced, P1_RMB); #endif - read_one = READ_CACHED_VAR(beta); - ooo_mem(); - // test : [] (read_one == 0 -> read_two != 0) - // test : [] (read_two == 0 -> read_one != 0) - assert(!(read_one == 0 && read_two == 0)); + + do + :: CONSUME_TOKENS(proc_one_produced, + P1_PROD_NONE, P1_READ_ONE) -> + ooo_mem(); + read_one = READ_CACHED_VAR(beta); + ooo_mem(); + PRODUCE_TOKENS(proc_one_produced, P1_READ_ONE); + :: CONSUME_TOKENS(proc_one_produced, + P1_READ_ONE, P1_RMB) -> + smp_rmb(); + PRODUCE_TOKENS(proc_one_produced, P1_RMB); + :: CONSUME_TOKENS(proc_one_produced, + P1_RMB, P1_READ_TWO) -> + ooo_mem(); + read_two = READ_CACHED_VAR(alpha); + ooo_mem(); + PRODUCE_TOKENS(proc_one_produced, P1_READ_TWO); + :: CONSUME_TOKENS(proc_one_produced, + P1_PROD_NONE | P1_READ_ONE | P1_RMB + | P1_READ_TWO, 0) -> + break; + od; + + //CLEAR_TOKENS(proc_one_produced, + // P1_PROD_NONE | P1_READ_ONE | P1_RMB | P1_READ_TWO); + + // test : [] (read_one == 1 -> read_two == 1) + assert(read_one != 1 || read_two == 1); } + +/* + * Bit encoding, proc_two_produced : + */ + +#define P2_PROD_NONE (1 << 0) + +#define P2_WRITE_ONE (1 << 1) +#define P2_WMB (1 << 2) +#define P2_WRITE_TWO (1 << 3) + +/* Only need a single color. */ +byte proc_two_produced; + active proctype test_proc_two() { assert(get_pid() < NR_PROCS); - ooo_mem(); - WRITE_CACHED_VAR(beta, 1); - ooo_mem(); -#ifndef NO_WMB - smp_wmb(); - ooo_mem(); + PRODUCE_TOKENS(proc_two_produced, P2_PROD_NONE); +#ifdef NO_WMB + PRODUCE_TOKENS(proc_two_produced, P2_WMB); #endif -#ifndef NO_RMB - smp_rmb(); - ooo_mem(); -#endif - read_two = READ_CACHED_VAR(alpha); - ooo_mem(); - // test : [] (read_one == 0 -> read_two != 0) - // test : [] (read_two == 0 -> read_one != 0) - assert(!(read_one == 0 && read_two == 0)); + + do + :: CONSUME_TOKENS(proc_two_produced, + P2_PROD_NONE, P2_WRITE_ONE) -> + ooo_mem(); + WRITE_CACHED_VAR(alpha, 1); + ooo_mem(); + PRODUCE_TOKENS(proc_two_produced, P2_WRITE_ONE); + :: CONSUME_TOKENS(proc_two_produced, + P2_WRITE_ONE, P2_WMB) -> + smp_wmb(); + PRODUCE_TOKENS(proc_two_produced, P2_WMB); + :: CONSUME_TOKENS(proc_two_produced, + P2_WMB, P2_WRITE_TWO) -> + ooo_mem(); + WRITE_CACHED_VAR(beta, 1); + ooo_mem(); + PRODUCE_TOKENS(proc_two_produced, P2_WRITE_TWO); + :: CONSUME_TOKENS(proc_two_produced, + P2_PROD_NONE | P2_WRITE_ONE + | P2_WMB | P2_WRITE_TWO, 0) -> + break; + od; + + //CLEAR_TOKENS(proc_two_produced, + // P2_PROD_NONE | P2_WRITE_ONE | P2_WMB | P2_WRITE_TWO); } diff --git a/formal-model/ooomem-double-update/read_order.ltl b/formal-model/ooomem-double-update/read_order.ltl index f4dbf03..dec26f2 100644 --- a/formal-model/ooomem-double-update/read_order.ltl +++ b/formal-model/ooomem-double-update/read_order.ltl @@ -1 +1 @@ -[] (read_one_is_zero -> !read_two_is_zero) +[] (read_one_is_one -> (!read_two_done || read_two_is_one)) diff --git a/formal-model/ooomem-two-writes/Makefile b/formal-model/ooomem-two-writes/Makefile index 37422b5..6e91788 100644 --- a/formal-model/ooomem-two-writes/Makefile +++ b/formal-model/ooomem-two-writes/Makefile @@ -28,6 +28,7 @@ default: make read_order | tee read_order.log make read_order_no_wmb | tee read_order_no_wmb.log make read_order_no_rmb | tee read_order_no_rmb.log + make read_order_no_sync | tee read_order_no_sync.log make asserts | tee asserts.log make summary @@ -53,6 +54,13 @@ read_order: clean read_order_ltl run cp .input.spin $@.spin.input -cp .input.spin.trail $@.spin.input.trail +read_order_no_sync: clean read_order_ltl read_order_no_sync_define run + cp .input.spin $@.spin.input + -cp .input.spin.trail $@.spin.input.trail + +read_order_no_sync_define: + cp read_order_no_sync.define .input.define + read_order_no_rmb: clean read_order_ltl read_order_no_rmb_define run cp .input.spin $@.spin.input -cp .input.spin.trail $@.spin.input.trail diff --git a/formal-model/ooomem-two-writes/mem.spin b/formal-model/ooomem-two-writes/mem.spin index cb206d2..9202043 100644 --- a/formal-model/ooomem-two-writes/mem.spin +++ b/formal-model/ooomem-two-writes/mem.spin @@ -21,24 +21,22 @@ /* Promela validation variables. */ /* - * Produced process data flow. Updated after each instruction to show which - * variables are ready. Assigned using SSA (static single assignment) (defuse - * analysis must be done on the program to map "real" variables to single define - * followed by use). Using one-hot bit encoding per variable to save state - * space. Used as triggers to execute the instructions having those variables - * as input. + * Produced process control and data flow. Updated after each instruction to + * show which variables are ready. Using one-hot bit encoding per variable to + * save state space. Used as triggers to execute the instructions having those + * variables as input. Leaving bits active to inhibit instruction execution. + * Scheme used to make instruction disabling and automatic dependency fall-back + * automatic. */ -#define PRODUCE_TOKENS(state, bits) \ - state = (state) | (bits) +#define CONSUME_TOKENS(state, bits, notbits) \ + ((!(state & (notbits))) && (state & (bits)) == (bits)) -/* All bits must be active to consume. All notbits must be inactive. */ -/* Consuming a token does not clear it, it just waits for it. */ -#define CONSUME_TOKENS(state, bits, notbits) \ - ((!((state) & (notbits))) && ((state) & (bits)) == (bits)) +#define PRODUCE_TOKENS(state, bits) \ + state = state | (bits); -#define CLEAR_TOKENS(state, bits) \ - state = (state) & ~(bits) +#define CLEAR_TOKENS(state, bits) \ + state = state & ~(bits) #define NR_PROCS 2