[Bug 27901] New: GLSL cos/sin functions broken on Mesa R600 driver

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



https://bugs.freedesktop.org/show_bug.cgi?id=27901

           Summary: GLSL cos/sin functions broken on Mesa R600 driver
           Product: Mesa
           Version: git
          Platform: All
        OS/Version: All
            Status: NEW
          Severity: normal
          Priority: medium
         Component: Drivers/DRI/R600
        AssignedTo: dri-devel@xxxxxxxxxxxxxxxxxxxxx
        ReportedBy: alain.perrot@xxxxxxxxx


As I was playing with OpenGL examples from Joe Groff's blog at
http://duriansoftware.com/joe/An-intro-to-modern-OpenGL.-Table-of-Contents.html,
I noticed that the cos and sin GLSL functions are broken on the Mesa R600
driver. Other trigonometric functions may be broken as well.

The blog entries provide two versions of a sample program that blend two images
together using GLSL. The blend factor is computed from a timestamp using the
sin function.

There's no problem with the example from chapter 2 where the blend factor is
computed on the CPU (source code at http://github.com/jckarter/hello-gl).

Mesa R600 driver fails to properly run the example from chapter 3 where the
blend factor is computed on the GPU using the GLSL sin function. Mesa software
renderer has no problem running this example (source code at
http://github.com/jckarter/hello-gl-ch3).


Looking at GLSL 1.10 spec, R600 ISA doc and R600 driver code (function
assemble_TRIG in r700_assembler.c), I guess there is two problems :

1. both the GLSL cos/sin functions and R600 hardware expects the angle to be
specified in radians, but the assemble_TRIG function divides the specified
angle by 2 * PI (the result is no more radians).

2. R600 hardware expects the angle to be in the range [-PI, PI] while GLSL does
not specify a range, and the assemble_TRIG function does not clip the angle in
the required range.


A quick and dirty hack to make the example from chapter 3 work with R600 driver
is to clip the timestamp to the [-PI, PI] range and then to multiply it by 2 *
PI :

--- hello-gl.c.old      2010-04-29 20:55:55.000000000 +0200
+++ hello-gl.c  2010-04-29 20:56:12.000000000 +0200
@@ -202,7 +202,10 @@
 static void update_timer(void)
 {
     int milliseconds = glutGet(GLUT_ELAPSED_TIME);
-    g_resources.timer = (float)milliseconds * 0.001f;
+    g_resources.timer = fmodf((float)milliseconds * 0.001f, 2.0f *
3.1415926535897f);
+    if(g_resources.timer > 3.1415926535897f)
+        g_resources.timer -= 2.0f * 3.1415926535897f;
+    g_resources.timer *= 2.0f * 3.1415926535897f;
     glutPostRedisplay();
 }


My config:
RV670 (Radeon HD 3870)
Kubuntu 10.04 "Lucid Lynx" 64-bit
Linux kernel 2.6.34-rc5 from Ubuntu Mainline Kernel PPA
libdrm, radeon, Mesa, Xorg from Xorg-edgers PPA

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.
_______________________________________________
dri-devel mailing list
dri-devel@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Index of Archives]     [Linux DRI Users]     [Linux Intel Graphics]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]
  Powered by Linux