[Bug 89597] Implement SSBOs in GLSL front-end and i965

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=89597

--- Comment #43 from Iago Toral <itoral at igalia.com> ---
Francisco, since you have already landed the untyped write messages in master I
have started to rewrite my code to use them but it seems as if these messages
don't write anything for me. Looking at the generated assembly it seems that
everything is fine though. For example:

mov(8)    g2<1>UD    0x00000000UD     { align1 WE_all 1Q compacted };
mov(16)   g3<1>UD    0x00000000UD     { align1 WE_all 1H compacted };
mov(16)   g5<1>UD    0x0000000aUD     { align1 WE_all 1H compacted };
mov(1)    g2.7<1>UD  g1.7<0,1,0>UD    { align1 WE_all };
send(8)   null       g2<8,8,1>UD
     data ( DC untyped surface write, 1, 46) mlen 5 rlen 0 { align1 1Q };

that is supposed to write the value 10 (0xa) to offset 0 of surface index 1,
which comes from a single assignment of that value to a integer variable at the
beginning of the buffer.

The assembly looks correct to me:
- g2 is the header and carries the pixel mask as delivered to the thread
- g3/g4 have the offset payload for SIMD16 (0 for all channels)
- g5/g6 have the data payload for SIMD16  (0xa for all channels)
- The message length is 5 for SIMD16 (header, offsetx2, datax2)
- surface index is 1, which is correct
- msg_control is 46, meaning only red channel enabled and SIMD16 mode.

but it does not seem to take any effect: the buffer contents are not modified
at any offset after this.

The surface I am writing to is a BRW_SURFACE_BUFFER with format
BRW_SURFACEFORMAT_RAW and a pitch of 1.

Am I missing something obvious here?

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/intel-3d-bugs/attachments/20150505/b88d2b40/attachment.html>