bouffa / gpuocelot

Automatically exported from code.google.com/p/gpuocelot
0 stars 0 forks source link

Ocelot won't interpretate ld.volatile.v4.f32 instruction #69

Open GoogleCodeExporter opened 8 years ago

GoogleCodeExporter commented 8 years ago
What steps will reproduce the problem?
1. Compile particles app in NVIDIA SDK 4.1 using nvcc 4.2 (nvcc -arch sm_20), 
exporting ptx file.
2. Use ptxOptimizer to parse output ptx file.

What is the expected output? What do you see instead?
Should create new ptx code.
Prints line and error:
1199 ld.volatile.v4.f32 {%f123, %f124, %f125, %f126}, [%rl19];
(1199, 12):  syntax error, unexpected TOKEN_V4

What version of the product are you using? On what operating system?
ocelot svn 1940

ptxgramar.ll won't accept a ".v4" token after a ".volatile" token, it requires 
to have a "addressSpace" between them.

Possible solution:
Add construction

ldModifier : TOKEN_VOLATILE instructionVectorType
{
    state.volatileFlag( true );
};

Original issue reported on code.google.com by unde...@gmail.com on 4 Jun 2012 at 10:00

Attachments:

GoogleCodeExporter commented 8 years ago
Actually, should add this construction:
ldModifier : TOKEN_VOLATILE instructionVectorType
{
    state.noAddressSpace();
    state.volatileFlag( true );
};

Original comment by unde...@gmail.com on 5 Jun 2012 at 1:16