diff options
Diffstat (limited to 'src/passes/pass.cpp')
| -rw-r--r-- | src/passes/pass.cpp | 10 |
1 files changed, 10 insertions, 0 deletions
diff --git a/src/passes/pass.cpp b/src/passes/pass.cpp index eaf20eb51..c548b66c5 100644 --- a/src/passes/pass.cpp +++ b/src/passes/pass.cpp @@ -128,6 +128,9 @@ void PassRegistry::registerPasses() { createDeNaNPass); registerPass( "directize", "turns indirect calls into direct ones", createDirectizePass); + registerPass("discard-global-effects", + "discards global effect info", + createDiscardGlobalEffectsPass); registerPass( "dfo", "optimizes using the DataFlow SSA IR", createDataFlowOptsPass); registerPass("dwarfdump", @@ -165,6 +168,9 @@ void PassRegistry::registerPasses() { "functions with i64 in their signature (which cannot be invoked " "via the wasm table without JavaScript BigInt support).", createGenerateI64DynCallsPass); + registerPass("generate-global-effects", + "generate global effect info (helps later passes)", + createGenerateGlobalEffectsPass); registerPass( "generate-stack-ir", "generate Stack IR", createGenerateStackIRPass); registerPass( @@ -581,6 +587,10 @@ void PassRunner::addDefaultGlobalOptimizationPrePasses() { addIfNoDWARFIssues("cfp"); addIfNoDWARFIssues("gsi"); } + // TODO: generate-global-effects here, right before function passes, then + // discard in addDefaultGlobalOptimizationPostPasses? the benefit seems + // quite minor so far, except perhaps when using call.without.effects + // which can lead to more opportunities for global effects to matter. } void PassRunner::addDefaultGlobalOptimizationPostPasses() { |