From f9d5c95fb4b4989ce4da8005793d3730452a33a9 Mon Sep 17 00:00:00 2001 From: Simon Marlow Date: Fri, 3 Apr 2009 12:18:16 +0000 Subject: [PATCH] On x86, use thread-local storage instead of stealing a reg for gct Benchmarks show that using TLS instead of stealing a register is better by a few percent on x86, due to the lack of registers. This only affects -threaded; without -threaded we're (now) using static storage for the GC data. --- rts/sm/GCThread.h | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/rts/sm/GCThread.h b/rts/sm/GCThread.h index c563d95..5646edd 100644 --- a/rts/sm/GCThread.h +++ b/rts/sm/GCThread.h @@ -209,10 +209,15 @@ extern gc_thread **gc_threads; #define SET_GCT(to) gct = (to) -#if defined(sparc_HOST_ARCH) +#if defined(sparc_HOST_ARCH) || defined(i386_HOST_ARCH) // Don't use REG_base or R1 for gct on SPARC because they're getting clobbered // by something else. Not sure what yet. -- BL 2009/01/03 +// Using __thread is better than stealing a register on x86, because +// we have too few registers available. In my tests it was worth +// about 5% in GC performance, but of course that might change as gcc +// improves. -- SDM 2009/04/03 + extern __thread gc_thread* gct; #define DECLARE_GCT __thread gc_thread* gct; -- 1.7.10.4