Add generations to the compilation cache for eval and regexp

Refactor the handling of generations in the compilation cache.

Add generations to the compilation cache for eval and regexp. The number of generations for these are set to two.

BUG=none
TEST=none

Committed: http://code.google.com/p/v8/source/detail?r=2233

[Søren Thygesen Gjesse, 2009-06-22 09:35:50]

I totally reworked the compilation cache change, so no reference to the previous
change (http://codereview.chromium.org/131100), but all comments are addressed,
even though there is still duplicated control structures.

[Kasper Lund, 2009-06-22 09:49:39]

LGTM with a few comments:

http://codereview.chromium.org/140056/diff/1/4
File src/compilation-cache.cc (right):

http://codereview.chromium.org/140056/diff/1/4#newcode37
Line 37: #define NUMBER_OF_SUB_CACHES                   4
Why not use constants for this? I would definitely have something like enum {
kNumberOfSubCaches = 4, ... } or static const ints.

http://codereview.chromium.org/140056/diff/1/4#newcode39
Line 39: // The number of generations for each sub cache.
This should probably also be constants.

http://codereview.chromium.org/140056/diff/1/4#newcode52
Line 52: static CompilationSubCache* entry[NUMBER_OF_SUB_CACHES] =
entry seems like a weird name. How about subcaches?

http://codereview.chromium.org/140056/diff/1/4#newcode73
Line 73: static const int kInitialCacheSize = 64;
Maybe move this constant out to the other constant definitions?

http://codereview.chromium.org/140056/diff/1/4#newcode87
Line 87: for (int generation = generations_ - 1;
I would probably rewrite this to use |i| as the counter (like it's done in
CompilationSubCache::Clear).

http://codereview.chromium.org/140056/diff/1/4#newcode210
Line 210: Handle<CompilationCacheTable> table = GetTable(generation);
Weird indentation.

http://codereview.chromium.org/140056/diff/1/4#newcode249
Line 249: for (generation = 0; generation < generations(); generation++) {
Weird indentation.

http://codereview.chromium.org/140056/diff/1/2
File src/compilation-cache.h (right):

http://codereview.chromium.org/140056/diff/1/2#newcode96
Line 96: // The compilation cache consists of several generational sub-caches
which uses
Could all this CompilationSubCache stuff be moved to compilation-cache.cc? It
seems like it's an implementation detail for the CompilationCache.

http://codereview.chromium.org/140056/diff/1/2#newcode103
Line 103: explicit CompilationSubCache(int generations):
generations_(generations) {
Space before :?

http://codereview.chromium.org/140056/diff/1/2#newcode104
Line 104: // tables_ = new Object*[generations];
Remove.

http://codereview.chromium.org/140056/diff/1/2#newcode107
Line 107: virtual ~CompilationSubCache() {
Why is this virtual? Is it ever called and needed?

[Søren Thygesen Gjesse, 2009-06-22 11:10:54]

http://codereview.chromium.org/140056/diff/1/4
File src/compilation-cache.cc (right):

http://codereview.chromium.org/140056/diff/1/4#newcode37
Line 37: #define NUMBER_OF_SUB_CACHES                   4
On 2009/06/22 09:49:39, Kasper Lund wrote:
> Why not use constants for this? I would definitely have something like enum {
> kNumberOfSubCaches = 4, ... } or static const ints.

Done.

http://codereview.chromium.org/140056/diff/1/4#newcode39
Line 39: // The number of generations for each sub cache.
On 2009/06/22 09:49:39, Kasper Lund wrote:
> This should probably also be constants.

Done.

http://codereview.chromium.org/140056/diff/1/4#newcode52
Line 52: static CompilationSubCache* entry[NUMBER_OF_SUB_CACHES] =
On 2009/06/22 09:49:39, Kasper Lund wrote:
> entry seems like a weird name. How about subcaches?

Done.

http://codereview.chromium.org/140056/diff/1/4#newcode73
Line 73: static const int kInitialCacheSize = 64;
On 2009/06/22 09:49:39, Kasper Lund wrote:
> Maybe move this constant out to the other constant definitions?

Done.

http://codereview.chromium.org/140056/diff/1/4#newcode87
Line 87: for (int generation = generations_ - 1;
On 2009/06/22 09:49:39, Kasper Lund wrote:
> I would probably rewrite this to use |i| as the counter (like it's done in
> CompilationSubCache::Clear).

Done.

http://codereview.chromium.org/140056/diff/1/4#newcode210
Line 210: Handle<CompilationCacheTable> table = GetTable(generation);
On 2009/06/22 09:49:39, Kasper Lund wrote:
> Weird indentation.

Done.

http://codereview.chromium.org/140056/diff/1/4#newcode249
Line 249: for (generation = 0; generation < generations(); generation++) {
On 2009/06/22 09:49:39, Kasper Lund wrote:
> Weird indentation.

Done.

http://codereview.chromium.org/140056/diff/1/2
File src/compilation-cache.h (right):

http://codereview.chromium.org/140056/diff/1/2#newcode96
Line 96: // The compilation cache consists of several generational sub-caches
which uses
On 2009/06/22 09:49:39, Kasper Lund wrote:
> Could all this CompilationSubCache stuff be moved to compilation-cache.cc? It
> seems like it's an implementation detail for the CompilationCache.

Done.

http://codereview.chromium.org/140056/diff/1/2#newcode103
Line 103: explicit CompilationSubCache(int generations):
generations_(generations) {
On 2009/06/22 09:49:39, Kasper Lund wrote:
> Space before :?

Done.

http://codereview.chromium.org/140056/diff/1/2#newcode104
Line 104: // tables_ = new Object*[generations];
On 2009/06/22 09:49:39, Kasper Lund wrote:
> Remove.

Done.

http://codereview.chromium.org/140056/diff/1/2#newcode107
Line 107: virtual ~CompilationSubCache() {
On 2009/06/22 09:49:39, Kasper Lund wrote:
> Why is this virtual? Is it ever called and needed?

It is never called with the current static allocation if the sub-caches -
removed.