297714 - Memory cache should use PR_GetPhysicalMemorySize

Status: RESOLVED FIXED

(Core :: Networking: Cache, defect)

Description

[Brian Ryner (not reading)]

Comment 1

[Brian Ryner (not reading)]

Attachment: patch

Comment 2

[Darin Fisher]

Comment on attachment 186258 [details] [diff] [review]
patch

>Index: nsCacheService.cpp

>+    PRUint64 bytes = PR_GetPhysicalMemorySize();
> 
>+    if (bytes == 0)  return 0;

nit: return on new line so a breakpoint can be set on return.

also, shouldn't this be comparing against LL_ZERO instead?


>+    // Conversion from unsigned int64 to double doesn't work on all platforms.
>+    // We need to truncate the value at LL_MAXINT to make sure we don't
>+    // overflow.
>+    if (LL_CMP(bytes, >, LL_MAXINT))
>+        bytes = LL_MAXINT;
> 
>+    double bytesD;
>+    LL_L2D(bytesD, (PRInt64) bytes);
> 
>+    double x = log(bytesD)/log((double)2) - 14;

nit: "2.0" instead of "(double)2"


otherwise, it looks good.

Comment 3

[Brian Ryner (not reading)]

Comment on attachment 186258 [details] [diff] [review]
patch

Requesting approval for 1.8b3.	This is pretty safe, as we're just moving
functionality into NSPR.

Comment 4

[Brian Ryner (not reading)]

checked in

Comment 5

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

This caused a large Tp drop on btek.  Was it supposed to?  It doesn't look like
it was.

Comment 6

[Darin Fisher]

Yikes!  That seems to suggest that the memory cache size might have increased in
size unexpectedly.