4 files changed, 46 insertions, 18 deletions

diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c
index 46acf79721..33b8843e33 100644
--- a/src/mesa/drivers/dri/intel/intel_context.c
+++ b/src/mesa/drivers/dri/intel/intel_context.c
@@ -697,6 +697,9 @@ intelInitContext(struct intel_context *intel,
       intel->no_rast = 1;
    }
 
+   intel->tiling_swizzle_mode = driQueryOptioni(&intel->optionCache,
+						"swizzle_mode");
+
    /* Disable all hardware rendering (skip emitting batches and fences/waits
     * to the kernel)
     */
diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h
index f1116d2747..6ed9a377e4 100644
--- a/src/mesa/drivers/dri/intel/intel_context.h
+++ b/src/mesa/drivers/dri/intel/intel_context.h
@@ -266,6 +266,7 @@ struct intel_context
    GLuint lastStamp;
 
    GLboolean no_hw;
+   int tiling_swizzle_mode;
 
    /**
     * Configuration cache
diff --git a/src/mesa/drivers/dri/intel/intel_screen.c b/src/mesa/drivers/dri/intel/intel_screen.c
index 8fd503ee8b..6597dbffed 100644
--- a/src/mesa/drivers/dri/intel/intel_screen.c
+++ b/src/mesa/drivers/dri/intel/intel_screen.c
@@ -69,13 +69,20 @@ PUBLIC const char __driConfigOptions[] =
    DRI_CONF_SECTION_QUALITY
       DRI_CONF_FORCE_S3TC_ENABLE(false)
       DRI_CONF_ALLOW_LARGE_TEXTURES(2)
+      DRI_CONF_OPT_BEGIN_V(swizzle_mode, enum, 0, "0:2")
+	 DRI_CONF_DESC_BEGIN(en, "Tiling swizzle mode for software fallbacks")
+	    DRI_CONF_ENUM(0, "No swizzling")
+	    DRI_CONF_ENUM(1, "addr[6] = addr[6] ^ addr[9]")
+	    DRI_CONF_ENUM(2, "addr[6] = addr[6] ^ addr[9] ^ addr[10]")
+	 DRI_CONF_DESC_END
+      DRI_CONF_OPT_END
    DRI_CONF_SECTION_END
    DRI_CONF_SECTION_DEBUG
      DRI_CONF_NO_RAST(false)
    DRI_CONF_SECTION_END
 DRI_CONF_END;
 
-const GLuint __driNConfigOptions = 6;
+const GLuint __driNConfigOptions = 7;
 
 #ifdef USE_NEW_INTERFACE
 static PFNGLXCREATECONTEXTMODES create_context_modes = NULL;
diff --git a/src/mesa/drivers/dri/intel/intel_span.c b/src/mesa/drivers/dri/intel/intel_span.c
index c6778b16ff..8d7d913ca9 100644
--- a/src/mesa/drivers/dri/intel/intel_span.c
+++ b/src/mesa/drivers/dri/intel/intel_span.c
@@ -106,29 +106,46 @@ static GLubyte *x_tile_swizzle(struct intel_renderbuffer *irb, struct intel_cont
 	x_tile_off = xbyte & 0x1ff;
 	y_tile_off = y & 7;
 
-#ifndef I915
-	/* The documentation says that X tile layout is arranged in 8 512-byte
-	 * lines of pixel data.  However, that doesn't appear to be the case
-	 * on GM965, tested by drawing a 128x8 quad in no_rast mode.  For lines
-	 * 1,2,4, and 7 of each tile, each consecutive pair of 64-byte spans
-	 * has the locations of those spans swapped.
+	x_tile_number = xbyte >> 9;
+	y_tile_number = y >> 3;
+
+	tile_off = (y_tile_off << 9) + x_tile_off;
+
+	/* bit swizzling tricks your parents never told you about:
+	 *
+	 * The specs say that the X tiling layout is just 8 512-byte rows
+	 * packed into a page.  It turns out that there's some additional
+	 * swizzling of bit 6 to reduce cache aliasing issues.  Experimental
+	 * results below:
+	 *
+	 * line    bit	 GM965	945G/Q965
+	 *	9 10 11
+	 * 0	0  0  0  0	0
+	 * 1	0  1  0  1	1
+	 * 2	1  0  0  1	1
+	 * 3	1  1  0  0	0
+	 * 4	0  0  1  1	0
+	 * 5	0  1  1  0	1
+	 * 6	1  0  1  0	1
+	 * 7	1  1  1  1	0
+	 *
+	 * So we see that the GM965 is bit 6 ^ 9 ^ 10 ^ 11, while other
+	 * parts were just 6 ^ 9 ^ 10.  However, some systems, including a
+	 * GM965 we've seen, don't perform the swizzling at all.  Information
+	 * on how to detect it through register reads is expected soon.
 	 */
-	switch (y_tile_off) {
+	switch (intel->tiling_swizzle_mode) {
+	case 0:
+	   break;
 	case 1:
+	   tile_off ^= ((tile_off >> 3) & 64) ^ ((tile_off >> 4) & 64);
+	   break;
 	case 2:
-	case 4:
-	case 7:
-		x_tile_off ^= 64;
-		break;
-	default:
+	   tile_off ^= ((tile_off >> 3) & 64) ^ ((tile_off >> 4) & 64) ^
+	      ((tile_off >> 5) & 64);
 	   break;
 	}
-#endif
-
-	x_tile_number = xbyte >> 9;
-	y_tile_number = y >> 3;
 
-	tile_off = (y_tile_off << 9) + x_tile_off;
 	tile_base = (x_tile_number << 12) + y_tile_number * tile_stride;
 
 #if 0