diff options
author | Ludovic Courtès <ludo@gnu.org> | 2010-11-17 23:06:26 +0100 |
---|---|---|
committer | Ludovic Courtès <ludo@gnu.org> | 2010-11-19 13:34:43 +0100 |
commit | 0c57673a1176dec03d6ddb88fa5cfc82a2628cc3 (patch) | |
tree | e57d89e00876317f1bac4fb98295318073e9ad37 /libguile/vm-i-scheme.c | |
parent | f13f1e9f6ff111fa676e54090eb62f2a0f066d13 (diff) | |
download | guile-0c57673a1176dec03d6ddb88fa5cfc82a2628cc3.tar.gz |
Add optimized tagged integer addition/subtractions for x86_64.
This results in a 17% improvement in the execution time of the "+" and
"-" benchmarks for fixnums.
* libguile/vm-i-scheme.c (ASM_ADD, ASM_SUB)[defined __x86_64__ &&
SCM_GNUC_PREREQ (4, 5)]: New macros.
(add)[defined ASM_ADD]: Use `ASM_ADD' for the fast path.
(sub)[defined ASM_SUB]: Use `ASM_SUB' for the fast path.
* test-suite/tests/numbers.test ("+")["fixnum + fixnum = bignum
(32-bit)", "fixnum + fixnum = bignum (64-bit)", "bignum + fixnum =
fixnum", "wrong type"]: New tests.
("-")["fixnum - fixnum = bignum (32-bit)", "fixnum - fixnum = bignum
(64-bit)", "bignum - fixnum = fixnum", "wrong type"]: New tests.
* test-suite/tests/00-initial-env.test ("goopsless")["+ wrong type
argument"]: Use `with-test-prefix/c&e' instead of `with-test-prefix'.
["- wrong type argument"]: New test prefix.
Diffstat (limited to 'libguile/vm-i-scheme.c')
-rw-r--r-- | libguile/vm-i-scheme.c | 71 |
1 files changed, 71 insertions, 0 deletions
diff --git a/libguile/vm-i-scheme.c b/libguile/vm-i-scheme.c index 3e80a0ef5..5557e19b8 100644 --- a/libguile/vm-i-scheme.c +++ b/libguile/vm-i-scheme.c @@ -210,6 +210,8 @@ VM_DEFINE_FUNCTION (149, ge, "ge?", 2) */ /* The maximum/minimum tagged integers. */ +#undef INUM_MAX +#undef INUM_MIN #define INUM_MAX (INTPTR_MAX - 1) #define INUM_MIN (INTPTR_MIN + scm_tc2_int) @@ -227,9 +229,68 @@ VM_DEFINE_FUNCTION (149, ge, "ge?", 2) RETURN (SFUNC (x, y)); \ } +/* Assembly tagged integer arithmetic routines. This code uses the + `asm goto' feature introduced in GCC 4.5. */ + +#if defined __x86_64__ && SCM_GNUC_PREREQ (4, 5) + +/* The macros below check the CPU's overflow flag to improve fixnum + arithmetic. The %rcx register is explicitly clobbered because `asm + goto' can't have outputs, in which case the `r' constraint could be + used to let the register allocator choose a register. + + TODO: Use `cold' label attribute in GCC 4.6. + http://gcc.gnu.org/ml/gcc-patches/2010-10/msg01777.html */ + +# define ASM_ADD(x, y) \ + { \ + asm volatile goto ("mov %1, %%rcx; " \ + "test %[tag], %%cl; je %l[slow_add]; " \ + "test %[tag], %0; je %l[slow_add]; " \ + "add %0, %%rcx; jo %l[slow_add]; " \ + "sub %[tag], %%rcx; " \ + "mov %%rcx, (%[vsp])\n" \ + : /* no outputs */ \ + : "r" (x), "r" (y), \ + [vsp] "r" (sp), [tag] "i" (scm_tc2_int) \ + : "rcx", "memory" \ + : slow_add); \ + NEXT; \ + } \ + slow_add: \ + do { } while (0) + +# define ASM_SUB(x, y) \ + { \ + asm volatile goto ("mov %0, %%rcx; " \ + "test %[tag], %%cl; je %l[slow_sub]; " \ + "test %[tag], %1; je %l[slow_sub]; " \ + "sub %1, %%rcx; jo %l[slow_sub]; " \ + "add %[tag], %%rcx; " \ + "mov %%rcx, (%[vsp])\n" \ + : /* no outputs */ \ + : "r" (x), "r" (y), \ + [vsp] "r" (sp), [tag] "i" (scm_tc2_int) \ + : "rcx", "memory" \ + : slow_sub); \ + NEXT; \ + } \ + slow_sub: \ + do { } while (0) + +#endif + + VM_DEFINE_FUNCTION (150, add, "add", 2) { +#ifndef ASM_ADD FUNC2 (+, scm_sum); +#else + ARGS2 (x, y); + ASM_ADD (x, y); + SYNC_REGISTER (); + RETURN (scm_sum (x, y)); +#endif } VM_DEFINE_FUNCTION (151, add1, "add1", 1) @@ -256,7 +317,14 @@ VM_DEFINE_FUNCTION (151, add1, "add1", 1) VM_DEFINE_FUNCTION (152, sub, "sub", 2) { +#ifndef ASM_SUB FUNC2 (-, scm_difference); +#else + ARGS2 (x, y); + ASM_SUB (x, y); + SYNC_REGISTER (); + RETURN (scm_difference (x, y)); +#endif } VM_DEFINE_FUNCTION (153, sub1, "sub1", 1) @@ -281,6 +349,9 @@ VM_DEFINE_FUNCTION (153, sub1, "sub1", 1) RETURN (scm_difference (x, SCM_I_MAKINUM (1))); } +# undef ASM_ADD +# undef ASM_SUB + VM_DEFINE_FUNCTION (154, mul, "mul", 2) { ARGS2 (x, y); |