Matt Pharr
d2fa735ef1
Provide ISPC_POINTER_SIZE predefined maacro (32 or 64)
2011-12-09 16:37:42 -08:00
Matt Pharr
20f34b67da
Fix typo in documentation
2011-12-09 16:31:03 -08:00
Matt Pharr
03f3db1e89
Fix bugs in ForeachStmt::TypeCheck() and Optimize() methods.
...
Specifically, we weren't storing the results passed back from when we called
those methods of the start and end exprs. This manifested itself as overloaded
functions there not resolving properly.
2011-12-08 15:29:20 -08:00
Matt Pharr
9805b0742d
Switch to avx-x2 for the stencil workload
2011-12-08 14:36:09 -08:00
Matt Pharr
6000c696b2
Small fixes to optimization disabling code.
2011-12-08 14:35:57 -08:00
Matt Pharr
5a2edf723b
Update with latest performance numbers.
2011-12-08 14:35:22 -08:00
Matt Pharr
aec7da740a
Fix malformed program crashes.
2011-12-08 14:35:12 -08:00
Matt Pharr
a79bc75b72
Add a number of symbol names to list to make internal after loading builtins.
...
Fixes issue #131 ; because they weren't being marked as internal before, when
compiling to multiple targets these would lead to multiply-defined symbols.
2011-12-07 08:30:38 -08:00
Matt Pharr
eaaebf7928
Small documentation cleanups
2011-12-06 16:52:02 -08:00
Matt Pharr
198aa9620e
Fix bug with mask used for gather/scatter code generation.
...
We should always use the full mask for this, never the internal mask.
Added tests for this.
2011-12-06 15:51:56 -08:00
Matt Pharr
27c53a3c25
Try 3 on warning about no output file specified
2011-12-06 14:44:41 -08:00
Matt Pharr
bd70182369
Add some additional tests
2011-12-06 14:26:52 -08:00
Matt Pharr
04df63d955
Update run_tests.py to work on Windows. Removed JIT-based testing path entirely.
2011-12-06 13:46:20 -08:00
Matt Pharr
d59131d670
Fix warning to not print "Warning" twice
2011-12-06 09:03:44 -08:00
Matt Pharr
9475e13d81
Issue a warning if no output file is specified.
2011-12-06 08:21:34 -08:00
Matt Pharr
765d86076f
Basic support for AVX2 when building with LLVM3.1svn
...
For now this target just uses the same builtins-*.ll files as the
regular AVX1 target. Once the gather intrinsic is available from
LLVM, we'll want to have custom target files that call out to that
for gathers. (The integer min/max intrinsics should be wired up to
the __{min,max}_varying_{int,uint}*() builtins at that point as
well.)
2011-12-06 08:20:53 -08:00
Matt Pharr
e2b6ed3db8
Fix built for LLVM2.9 and 3.1svn
2011-12-06 08:08:41 -08:00
Matt Pharr
b22943b4a4
Update release notes and doxygen for 1.1.0 release
2011-12-05 14:44:53 -08:00
Matt Pharr
13df5f1cb9
Performance results page
2011-12-05 14:24:42 -08:00
Matt Pharr
f19c2aba40
Windows build fixes for examples, update options task granularity
2011-12-05 14:23:50 -08:00
Matt Pharr
ffc1d97df7
Fix aobench_instrumented build on Windows
2011-12-05 13:33:29 -08:00
Matt Pharr
9dd498718b
Updated options pricing example to have a tasking-based path as well.
2011-12-05 13:24:34 -08:00
Matt Pharr
6181ce59ae
FunctionCallExpr bug: launch count wasn't being type checked, optimized.
...
This manifested itself by a call to an overloaded function in a launch count
expression that wasn't being resolved.
2011-12-05 13:23:20 -08:00
Matt Pharr
48a6c2a35b
Fix test for 16-wide case
2011-12-05 11:45:06 -08:00
Matt Pharr
0388f46a3b
Remove test that was failing (now recorded as issue #130 ).
2011-12-05 09:39:50 -08:00
Matt Pharr
e3cae098fe
Update test
2011-12-05 09:27:53 -08:00
Matt Pharr
455d963962
Don't ignore return value from getcwd()
2011-12-05 09:26:33 -08:00
Matt Pharr
d748c501c9
Fully automate building final HTML files from docs
2011-12-05 09:08:51 -08:00
Matt Pharr
5b8596102a
Add entry about using valgrind with ispc to FAQ
2011-12-05 05:57:18 -08:00
Matt Pharr
dc525f281d
Fix bug where declarations of arrays of func. ptrs would be lost.
...
Issue #126 .
2011-12-05 05:35:47 -08:00
Matt Pharr
f95504fb5e
Symbol table now properly handles scopes for function declarations.
...
Previously, they all went into one big pile that was never cleaned up;
this was the wrong thing to do in a world where one might have a
function declaration inside another functions, say.
2011-12-04 17:37:13 -08:00
Matt Pharr
32904dfa11
Fix uninitialized memory error introduced in d65c02f3
2011-12-04 16:39:56 -08:00
Matt Pharr
186d0223d2
Fix AoS/SoA stdlib functions to match documentation
...
(i.e. actually remove the old offset parameter stuff now that
we can actually pass pointers.)
2011-12-03 22:44:16 -08:00
Matt Pharr
3efbfc30b7
Issue an error if a varying lvalue is passed to a reference function parameter.
...
(Previously, we crashed.)
2011-12-03 15:35:50 -08:00
Matt Pharr
0fd7811344
Small documentation edits and updates.
2011-12-03 15:35:50 -08:00
Matt Pharr
d492ba08e6
Fix bugs that broke typedefs in function definitions.
...
Issue #118 .
2011-12-03 15:35:44 -08:00
Matt Pharr
a1c0b4f95a
Allow 'continue' statements in 'foreach' loops.
2011-12-03 09:31:02 -08:00
Matt Pharr
c3b55de1ad
Fix volume rendering example for command-line args change
2011-12-03 09:30:10 -08:00
Matt Pharr
e07ef6d46a
1.1 Users guide final (for now)
2011-12-02 17:04:39 -08:00
Matt Pharr
3e4d69cbd3
Checkpoint work on specifying execution model
2011-12-02 16:01:05 -08:00
Matt Pharr
511a3ab15a
Checkpoint documentation work
2011-12-01 17:00:41 -08:00
Matt Pharr
24ef9dac8f
Use foreach in the deferred shading example
2011-12-01 17:00:30 -08:00
Matt Pharr
3bb6bff15d
Add tests of things the docs claim will cause an error to be issued
2011-12-01 17:00:13 -08:00
Matt Pharr
1390aed99c
Make 32-bit addressing the default.
...
Also renamed the command-line flag to --addressing={32,64}.
2011-12-01 13:38:40 -08:00
Matt Pharr
82aa6efd12
Checkpoint user's guide edits
2011-12-01 13:38:17 -08:00
Matt Pharr
f90aa172a6
Documentation work; first pass perf guide complete
2011-12-01 09:42:56 -08:00
Matt Pharr
a2f118a14e
FAQ and perf guide updates
2011-11-30 19:38:37 -08:00
Matt Pharr
c5aecd51e9
Fix indentation in usage message.
2011-11-30 17:11:12 -08:00
Matt Pharr
4d6bcdf41c
Documentation refactoring, initial pass at FAQ
2011-11-30 17:11:03 -08:00
Matt Pharr
8bc7367109
Add foreach and foreach_tiled looping constructs
...
These make it easier to iterate over arbitrary amounts of data
elements; specifically, they automatically handle the "ragged
extra bits" that come up when the number of elements to be
processed isn't evenly divided by programCount.
TODO: documentation
2011-11-30 13:17:31 -08:00
