Get rid of ykconfig -ao.

Not so long ago we added `yk-config -ao <level>` to allow us to run our
pre-canned franken-O1 pipeline for the entire test suite. This was only
ever designed as an interim hack, and we've wanted Yk to simply work
with standard optimisation flags like -O1, -O2 etc.

This change is a step in that direction.

We remove `-ao` and our non-standard defaults and add the ability for
individual tests to specify extra compiler flags. This gives us the
ability to specify optimisation parameters on a per-test basis.

I then ported as many tests as were practical to use -O1. The remaining
tests (that were either too involved for me to convert right now, or
actually rely on an IR shape generated by our old default pipeline) I've
left using our franken-O1 by passing:

-O0 -Xclang -disable-O0-optnone -Xlinker \
    --lto-newpm-passes=instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>

We do have to disable yklua in CI for now, as yklua still uses
`yk-config -ao` and we have to break a circular dependency.

.buildbot.sh

@@ -10,6 +10,7 @@ TRACERS="hwt swt"
 #  - `yk-config` must be in PATH.
 #  - YK_BUILD_TYPE must be set.
 test_yklua() {
+    return # FIXME: temporarily disable yklua to break a circular dep in CI.
     if [ ! -e "yklua" ]; then
         git clone https://github.com/ykjit/yklua
     fi

bin/yk-config

@@ -10,59 +10,16 @@ if [ ! -d "${DIR}/../ykrt" ]; then
     exit 1
 fi
 
-set_aot_pipeline() {
-    level=$1
-    # env var always takes precident.
-    if [ ! -z "${YKB_AOT_OPTLEVEL}" ]; then
-        level=${YKB_AOT_OPTLEVEL}
-    fi
-    case ${level} in
-        0)
-            POSTLINK_PASSES=${POSTLINK_PASSES_AO0}
-            PRELINK_PASSES=${PRELINK_PASSES_AO0}
-            ;;
-        1)
-            POSTLINK_PASSES=${POSTLINK_PASSES_AO1}
-            PRELINK_PASSES=${PRELINK_PASSES_AO1}
-            ;;
-        *)
-            echo "error: invalid yk AOT optimisation level '${level}'" >&2
-            exit 1
-            ;;
-    esac
-}
-
-# Canned pipelines.
-#
-# Ideally we'd detect the optimisation level from the CFLAGS env var, but it's
-# not easy to reliably parse that info out of CFLAGS. So for now we allow `--ao
-# <n>` to yk-config (or set YKB_AOT_OPTLEVEL in the env) to specify the AOT
-# optimisation level.
-#
-# Note that clang is still passed -O0 regardless. We just add our own pipeline
-# specification on top.
-PRELINK_PASSES_AO0=""
-POSTLINK_PASSES_AO0="instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>"
-PRELINK_PASSES_AO1=""
-# The whole -O1 pipeline, as reported by `opt -O1 -print-pipeline-passes
-# /dev/null`.
-POSTLINK_PASSES_AO1="annotation2metadata,forceattrs,inferattrs,coro-early,function<eager-inv>(lower-expect,simplifycfg<bonus-inst-threshold=1;no-forward-switch-cond;no-switch-range-to-icmp;no-switch-to-lookup;keep-loops;no-hoist-common-insts;no-sink-common-insts;speculate-blocks;simplify-cond-branch>,sroa<modify-cfg>,early-cse<>),openmp-opt,ipsccp,called-value-propagation,globalopt,function<eager-inv>(mem2reg,instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>,simplifycfg<bonus-inst-threshold=1;no-forward-switch-cond;switch-range-to-icmp;no-switch-to-lookup;keep-loops;no-hoist-common-insts;no-sink-common-insts;speculate-blocks;simplify-cond-branch>),require<globals-aa>,function(invalidate<aa>),require<profile-summary>,cgscc(devirt<4>(inline<only-mandatory>,inline,function-attrs<skip-non-recursive-function-attrs>,function<eager-inv;no-rerun>(sroa<modify-cfg>,early-cse<memssa>,simplifycfg<bonus-inst-threshold=1;no-forward-switch-cond;switch-range-to-icmp;no-switch-to-lookup;keep-loops;no-hoist-common-insts;no-sink-common-insts;speculate-blocks;simplify-cond-branch>,instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>,libcalls-shrinkwrap,simplifycfg<bonus-inst-threshold=1;no-forward-switch-cond;switch-range-to-icmp;no-switch-to-lookup;keep-loops;no-hoist-common-insts;no-sink-common-insts;speculate-blocks;simplify-cond-branch>,reassociate,loop-mssa(loop-instsimplify,loop-simplifycfg,licm<no-allowspeculation>,loop-rotate<header-duplication;no-prepare-for-lto>,licm<allowspeculation>,simple-loop-unswitch<no-nontrivial;trivial>),simplifycfg<bonus-inst-threshold=1;no-forward-switch-cond;switch-range-to-icmp;no-switch-to-lookup;keep-loops;no-hoist-common-insts;no-sink-common-insts;speculate-blocks;simplify-cond-branch>,instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>,loop(loop-idiom,indvars,loop-deletion,loop-unroll-full),sroa<modify-cfg>,memcpyopt,sccp,bdce,instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>,coro-elide,adce,simplifycfg<bonus-inst-threshold=1;no-forward-switch-cond;switch-range-to-icmp;no-switch-to-lookup;keep-loops;no-hoist-common-insts;no-sink-common-insts;speculate-blocks;simplify-cond-branch>,instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>),function-attrs,function(require<should-not-run-function-passes>),coro-split)),deadargelim,coro-cleanup,globalopt,globaldce,elim-avail-extern,rpo-function-attrs,recompute-globalsaa,function<eager-inv>(float2int,lower-constant-intrinsics,loop(loop-rotate<header-duplication;no-prepare-for-lto>,loop-deletion),loop-distribute,inject-tli-mappings,loop-vectorize<no-interleave-forced-only;vectorize-forced-only;>,infer-alignment,loop-load-elim,instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>,simplifycfg<bonus-inst-threshold=1;forward-switch-cond;switch-range-to-icmp;switch-to-lookup;no-keep-loops;hoist-common-insts;sink-common-insts;speculate-blocks;simplify-cond-branch>,vector-combine,instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>,loop-unroll<O1>,transform-warning,sroa<preserve-cfg>,infer-alignment,instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>,loop-mssa(licm<allowspeculation>),alignment-from-assumptions,loop-sink,instsimplify,div-rem-pairs,tailcallelim,simplifycfg<bonus-inst-threshold=1;no-forward-switch-cond;switch-range-to-icmp;no-switch-to-lookup;keep-loops;no-hoist-common-insts;no-sink-common-insts;speculate-blocks;simplify-cond-branch>),globaldce,constmerge,cg-profile,rel-lookup-table-converter,function(annotation-remarks),verify"
-# Initialise the AOT pipeline to level 0 unless YKB_AOT_OPTLEVEL is set.
-set_aot_pipeline 0
-
 OUTPUT=""
 
 usage() {
     echo "Generate C compiler flags for building against the yk JIT.\n"
     echo "Usage:"
-    echo "    yk-config <profile> [--ao <0|1>] \\"
+    echo "    yk-config <profile> \\"
     echo "        <--cc|--cxx|--ar|--ranlib|--cppflags|--cflags|--ldflags>\n"
     echo "    Where <profile> is a Rust cargo profile starting with either 'debug' or 'release'.\n"
     echo "    --cc, --cxx, --ar, --ranlib, --cppflags, --cflags and --ldflags specify "
-    echo "    what flags to output.\n"
-    echo "    --ao specifies the yk-specific AOT optimisation pipeline to use for flags "
-    echo "    that follow. Defaults to 0, but (if set) the YKB_AOT_OPTLEVEL environment "
-    echo "    variable always takes precedent."
+    echo "    what flags to output.\n\n"
 }
 
 handle_arg() {
@@ -77,32 +34,18 @@ handle_arg() {
     fi
 
     case $1 in
-        --ao) set_aot_pipeline $2;;
         --cc) OUTPUT="${ykllvm_bin_dir}/clang" ;;
         --cxx) OUTPUT="${ykllvm_bin_dir}/clang++" ;;
         --ar) OUTPUT="${ykllvm_bin_dir}/llvm-ar" ;;
         --ranlib) OUTPUT="${ykllvm_bin_dir}/llvm-ranlib" ;;
         --cflags)
-            # FIXME: we can only do -O0 for now.
-            OUTPUT="${OUTPUT} -O0"
             # Enable LTO.
             OUTPUT="${OUTPUT} -flto"
             # Outline functions containing loops during AOT compilation. Needed
             # for `yk_unroll_safe`.
             OUTPUT="${OUTPUT} -fyk-noinline-funcs-with-loops"
-            # Required so that our curated set of optimisation passes can run.
-            # Without this, clang will slap `optnone` attributes on every
-            # function, causing optimisations to skip them.
-            OUTPUT="${OUTPUT} -Xclang -disable-O0-optnone"
-            # Run these "prelink" optimisation passes.
-            #
-            # These are run on the individual pre-LTO-merged LLVM modules.
-            if [ ! -z "${PRELINK_PASSES}" ]; then
-                OUTPUT="${OUTPUT} -mllvm --newpm-passes=${PRELINK_PASSES}"
-            fi
-            case $profile in
-                debug*) OUTPUT="$OUTPUT -g" ;;
-            esac
+            # Don't optimise functions by changing their calling convention.
+            OUTPUT="${OUTPUT} -mllvm -yk-dont-opt-func-abi"
             ;;
         --cppflags)
             # Path to yk.h
@@ -167,18 +110,11 @@ handle_arg() {
             if [ "${YKB_TRACER}" = "swt" ]; then
                 OUTPUT="${OUTPUT} -Wl,--mllvm=--yk-basicblock-tracer"
             fi
-            # Add known yk-friendly optimisations.
-            #
-            # This pairs with `-Xclang -disable-O0-optnone`. See above.
+            # Disable backend optimisations that Yk can't handle by adding the
+            # `optnone` attribute onto every function after high-level IR
+            # optimisations have finished.
             OUTPUT="${OUTPUT} -Wl,--mllvm=--yk-optnone-after-ir-passes"
 
-            # Run these post-link passes.
-            #
-            # These are run on the merged LTO LLVM module.
-            if [ ! -z "${POSTLINK_PASSES}" ]; then
-                OUTPUT="${OUTPUT} -Xlinker --lto-newpm-passes=${POSTLINK_PASSES}"
-            fi
-
             # Emit a basic block map section. Used for block mapping.
             OUTPUT="${OUTPUT} -Wl,--lto-basic-block-sections=labels"
 

tests/c/aot_debuginfo.c

@@ -1,4 +1,5 @@
-// ignore-if: test $YK_CARGO_PROFILE != "debug"
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O1 -g
 // Run-time:
 //   env-var: YKD_LOG_IR=-:aot
 //   env-var: YKD_SERIALISE_COMPILATION=1
@@ -9,13 +10,15 @@
 //     %{{_}}: ptr = call yk_mt_new(0x0)
 //     ...
 //     # aot_debuginfo.c:{{_}}: yk_mt_hot_threshold_set(mt, 0);
+//     ...
 //     call yk_mt_hot_threshold_set(%{{_}}, 0i32)
 //     ...
 //     # aot_debuginfo.c:{{_}}: YkLocation loc = yk_location_new();
 //     ...
 //     %{{_}}: i64 = call yk_location_new()
 //     ...
 //     # aot_debuginfo.c:{{_}}: int i = 4;
+//     ...
 //     *%{{_}} = 4i32
 //     ...
 //     # aot_debuginfo.c:{{_}}: while (i > 0) {

tests/c/arithmetic.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O1
 // Run-time:
 //   env-var: YKD_LOG_IR=-:jit-pre-opt,jit-post-opt
 //   env-var: YKD_SERIALISE_COMPILATION=1

tests/c/ashr_exact.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O0 -Xclang -disable-O0-optnone -Xlinker --lto-newpm-passes=instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>
 // Run-time:
 //   env-var: YKD_LOG_IR=-:jit-pre-opt
 //   env-var: YKD_SERIALISE_COMPILATION=1

tests/c/bf.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O0 -Xclang -disable-O0-optnone -Xlinker --lto-newpm-passes=instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>
 // Run-time:
 //   env-var: YKD_SERIALISE_COMPILATION=1
 //   env-var: YK_LOG=4

tests/c/call_ext_simple.c

@@ -1,10 +1,11 @@
 // Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O1
 // Run-time:
 //   env-var: YKD_LOG_IR=-:jit-pre-opt
 //   env-var: YKD_SERIALISE_COMPILATION=1
 //   stderr:
 //     ...
-//     func_decl putchar (i32) -> i32
+//     func_decl putc (i32, ptr) -> i32
 //     ...
 //     %{{6}}: i32 = call @putc...
 //     ...

tests/c/double.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O1
 // Run-time:
 //   env-var: YKD_LOG_IR=-:aot,jit-pre-opt
 //   env-var: YKD_SERIALISE_COMPILATION=1

tests/c/doubleinline.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O1
 // Run-time:
 //   env-var: YKD_LOG_IR=-:aot,jit-pre-opt
 //   env-var: YKD_SERIALISE_COMPILATION=1

tests/c/dyn_ptradd_mixed.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O1
 // Run-time:
 //   env-var: YKD_LOG_IR=-:aot,jit-pre-opt
 //   env-var: YKD_SERIALISE_COMPILATION=1

tests/c/dyn_ptradd_multidim.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O0 -Xclang -disable-O0-optnone -Xlinker --lto-newpm-passes=instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>
 // Run-time:
 //   env-var: YKD_LOG_IR=-:aot,jit-pre-opt
 //   env-var: YKD_SERIALISE_COMPILATION=1
@@ -51,8 +53,8 @@ int main(int argc, char **argv) {
 
   int i = 0;
   NOOPT_VAL(loc);
-  NOOPT_VAL(i);
   NOOPT_VAL(array);
+  NOOPT_VAL(i);
   while (i < 4) {
     yk_mt_control_point(mt, &loc);
     fprintf(stderr, "i=%d, elem=%03d\n", i, array[i][i][i]);

tests/c/float.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O1
 // Run-time:
 //   env-var: YKD_LOG_IR=-:aot,jit-pre-opt
 //   env-var: YKD_SERIALISE_COMPILATION=1

tests/c/float_binop.c

@@ -1,4 +1,5 @@
-// ## yk-config-env: YKB_AOT_OPTLEVEL=1
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O1
 // Run-time:
 //   env-var: YKD_LOG_IR=-:aot,jit-pre-opt
 //   env-var: YKD_SERIALISE_COMPILATION=1

tests/c/float_div.c

@@ -1,4 +1,5 @@
-// ## yk-config-env: YKB_AOT_OPTLEVEL=1
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O1
 // Run-time:
 //   env-var: YKD_LOG_IR=-:aot,jit-pre-opt
 //   env-var: YKD_SERIALISE_COMPILATION=1

tests/c/float_mul.c

@@ -1,4 +1,5 @@
-// ## yk-config-env: YKB_AOT_OPTLEVEL=1
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O1
 // Run-time:
 //   env-var: YKD_LOG_IR=-:aot,jit-pre-opt
 //   env-var: YKD_SERIALISE_COMPILATION=1

tests/c/floats.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O1
 // Run-time:
 //   env-var: YKD_SERIALISE_COMPILATION=1
 //   status: success

tests/c/floats_return.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O0 -Xclang -disable-O0-optnone -Xlinker --lto-newpm-passes=instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>
 // Run-time:
 //   env-var: YKD_LOG_IR=-:jit-pre-opt
 //   env-var: YKD_SERIALISE_COMPILATION=1

tests/c/inline_const.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O1
 // Run-time:
 //   env-var: YKD_LOG_IR=-:jit-pre-opt
 //   env-var: YKD_SERIALISE_COMPILATION=1

tests/c/inst_type_depends_global.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O0 -Xclang -disable-O0-optnone -Xlinker --lto-newpm-passes=instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>
 // Run-time:
 //   env-var: YKD_LOG_IR=-:aot
 //   env-var: YKD_SERIALISE_COMPILATION=1

tests/c/intrinsics.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O0 -Xclang -disable-O0-optnone -Xlinker --lto-newpm-passes=instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>
 // Run-time:
 //   env-var: YKD_LOG_IR=-:jit-pre-opt
 //   env-var: YKD_SERIALISE_COMPILATION=1

tests/c/phi1.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O0 -Xclang -disable-O0-optnone -Xlinker --lto-newpm-passes=instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>
 // Run-time:
 //   env-var: YKD_LOG_IR=-:aot,jit-pre-opt
 //   env-var: YK_LOG=4

tests/c/phi2.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O0 -Xclang -disable-O0-optnone -Xlinker --lto-newpm-passes=instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>
 // Run-time:
 //   env-var: YKD_LOG_IR=-:aot,jit-pre-opt
 //   env-var: YK_LOG=4

tests/c/phi3.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O0 -Xclang -disable-O0-optnone -Xlinker --lto-newpm-passes=instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>
 // Run-time:
 //   env-var: YKD_LOG_IR=-:aot,jit-pre-opt
 //   env-var: YK_LOG=4

tests/c/ptradd.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O0 -Xclang -disable-O0-optnone -Xlinker --lto-newpm-passes=instcombine<max-iterations=1;no-use-loop-info;no-verify-fixpoint>
 // Run-time:
 //   env-var: YKD_LOG_IR=-:aot,jit-pre-opt
 //   env-var: YK_LOG=4

tests/c/sdiv.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O1
 // Run-time:
 //   env-var: YKD_LOG_IR=-:jit-pre-opt
 //   env-var: YKD_SERIALISE_COMPILATION=1

tests/c/simple_inline.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O1
 // Run-time:
 //   env-var: YKD_LOG_IR=-:jit-pre-opt
 //   env-var: YKD_SERIALISE_COMPILATION=1

tests/c/simplecall.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O1
 // Run-time:
 //   env-var: YKD_LOG_IR=-:aot,jit-pre-opt
 //   env-var: YKD_SERIALISE_COMPILATION=1
@@ -25,8 +27,8 @@
 //     2
 //     foo
 //     1
-//     yk-jit-event: deoptimise
 //     bar
+//     yk-jit-event: deoptimise
 //     0
 //     exit
 

tests/c/udiv.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O1
 // Run-time:
 //   env-var: YKD_LOG_IR=-:jit-pre-opt
 //   env-var: YKD_SERIALISE_COMPILATION=1

tests/c/unintptr_t_to_ptr.c

@@ -1,3 +1,5 @@
+// Compiler:
+//   env-var: YKB_EXTRA_CC_FLAGS=-O1
 // Run-time:
 //   env-var: YKD_LOG_IR=-:aot
 //   env-var: YKD_SERIALISE_COMPILATION=1

ykbuild/completion-wrapper

@@ -2,8 +2,11 @@
 
 # This wrapper script is used to record compiler invocations in sub-builds.
 # That information can then be used for e.g. code completion purposes.
+#
+# It is also used to allow tests to insert extra flags into the compiler
+# invocation.
 
 set -e
 
-echo "${YK_COMPILER_PATH} $@" > $(mktemp -p ${YK_COMPILER_TEMPDIR})
-${YK_COMPILER_PATH} $@
+echo "${YK_COMPILER_PATH} ${YKB_EXTRA_CC_FLAGS} $@" > $(mktemp -p ${YK_COMPILER_TEMPDIR})
+${YK_COMPILER_PATH} ${YKB_EXTRA_CC_FLAGS} $@