r200: invalidate texture paths in some more places
[mesa.git] / src / mesa / drivers / dri / r200 / r200_span.c
index e1ad976ce3b8e333cf7ab3136c234d5e979df012..17c651395b94f018a46d5c5d05b85f1235e67109 100644 (file)
@@ -1,4 +1,3 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_span.c,v 1.1 2002/10/30 12:51:52 alanh Exp $ */
 /*
 Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
 
@@ -33,12 +32,13 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  *   Keith Whitwell <keith@tungstengraphics.com>
  */
 
-#include "glheader.h"
-#include "imports.h"
+#include "main/glheader.h"
+#include "main/imports.h"
+#include "main/colormac.h"
 #include "swrast/swrast.h"
-#include "colormac.h"
 
 #include "r200_context.h"
+#include "radeon_buffer.h"
 #include "r200_ioctl.h"
 #include "r200_state.h"
 #include "r200_span.h"
@@ -46,71 +46,30 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #define DBG 0
 
-#define LOCAL_VARS                                                     \
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);                   \
-   r200ScreenPtr r200Screen = rmesa->r200Screen;                       \
-   __DRIscreenPrivate *sPriv = rmesa->dri.screen;                      \
-   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;                  \
-   GLuint pitch = r200Screen->frontPitch * r200Screen->cpp;            \
-   GLuint height = dPriv->h;                                           \
-   char *buf = (char *)(sPriv->pFB +                                   \
-                       rmesa->state.color.drawOffset +                 \
-                       (dPriv->x * r200Screen->cpp) +          \
-                       (dPriv->y * pitch));                            \
-   char *read_buf = (char *)(sPriv->pFB +                              \
-                            rmesa->state.pixel.readOffset +            \
-                            (dPriv->x * r200Screen->cpp) +             \
-                            (dPriv->y * pitch));                       \
-   GLuint p;                                                           \
-   (void) read_buf; (void) buf; (void) p
-
-#define LOCAL_DEPTH_VARS                                               \
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);                   \
-   r200ScreenPtr r200Screen = rmesa->r200Screen;                       \
-   __DRIscreenPrivate *sPriv = rmesa->dri.screen;                      \
-   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;                  \
-   GLuint height = dPriv->h;                                           \
-   GLuint xo = dPriv->x;                                               \
-   GLuint yo = dPriv->y;                                               \
-   char *buf = (char *)(sPriv->pFB + r200Screen->depthOffset); \
-   (void) buf
-
-#define LOCAL_STENCIL_VARS     LOCAL_DEPTH_VARS
-
-
-#define CLIPPIXEL( _x, _y )                                            \
-   ((_x >= minx) && (_x < maxx) && (_y >= miny) && (_y < maxy))
-
-
-#define CLIPSPAN( _x, _y, _n, _x1, _n1, _i )                           \
-   if ( _y < miny || _y >= maxy ) {                                    \
-      _n1 = 0, _x1 = x;                                                        \
-   } else {                                                            \
-      _n1 = _n;                                                                \
-      _x1 = _x;                                                                \
-      if ( _x1 < minx ) _i += (minx-_x1), n1 -= (minx-_x1), _x1 = minx; \
-      if ( _x1 + _n1 >= maxx ) n1 -= (_x1 + n1 - maxx);                        \
-   }
-
-#define Y_FLIP( _y )           (height - _y - 1)
+/*
+ * Note that all information needed to access pixels in a renderbuffer
+ * should be obtained through the gl_renderbuffer parameter, not per-context
+ * information.
+ */
+#define LOCAL_VARS                                             \
+   struct radeon_renderbuffer *rrb = (void *) rb;              \
+   const __DRIdrawablePrivate *dPriv = rrb->dPriv;             \
+   const GLuint bottom = dPriv->h - 1;                         \
+   GLuint p;                                                   \
+   (void) p;
 
+#define LOCAL_DEPTH_VARS                               \
+   struct radeon_renderbuffer *rrb = (void *) rb;              \
+   const __DRIdrawablePrivate *dPriv = rrb->dPriv;     \
+   const GLuint bottom = dPriv->h - 1;                 \
+   GLuint xo = dPriv->x;                               \
+   GLuint yo = dPriv->y;
 
-#define HW_LOCK() 
+#define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS
 
-#define HW_CLIPLOOP()                                                  \
-   do {                                                                        \
-      __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;               \
-      int _nc = dPriv->numClipRects;                                   \
-                                                                       \
-      while ( _nc-- ) {                                                        \
-        int minx = dPriv->pClipRects[_nc].x1 - dPriv->x;               \
-        int miny = dPriv->pClipRects[_nc].y1 - dPriv->y;               \
-        int maxx = dPriv->pClipRects[_nc].x2 - dPriv->x;               \
-        int maxy = dPriv->pClipRects[_nc].y2 - dPriv->y;
-
-#define HW_ENDCLIPLOOP()                                               \
-      }                                                                        \
-   } while (0)
+#define Y_FLIP(Y) (bottom - (Y))
+
+#define HW_LOCK() 
 
 #define HW_UNLOCK()                                                    
 
@@ -122,61 +81,23 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 /* 16 bit, RGB565 color spanline and pixel functions
  */
-#define INIT_MONO_PIXEL(p, color) \
-  p = PACK_COLOR_565( color[0], color[1], color[2] )
+#define SPANTMP_PIXEL_FMT GL_RGB
+#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5
 
-#define WRITE_RGBA( _x, _y, r, g, b, a )                               \
-   *(GLushort *)(buf + _x*2 + _y*pitch) = ((((int)r & 0xf8) << 8) |    \
-                                          (((int)g & 0xfc) << 3) |     \
-                                          (((int)b & 0xf8) >> 3))
-
-#define WRITE_PIXEL( _x, _y, p )                                       \
-   *(GLushort *)(buf + _x*2 + _y*pitch) = p
-
-#define READ_RGBA( rgba, _x, _y )                                      \
-   do {                                                                        \
-      GLushort p = *(GLushort *)(read_buf + _x*2 + _y*pitch);          \
-      rgba[0] = ((p >> 8) & 0xf8) * 255 / 0xf8;                                \
-      rgba[1] = ((p >> 3) & 0xfc) * 255 / 0xfc;                                \
-      rgba[2] = ((p << 3) & 0xf8) * 255 / 0xf8;                                \
-      rgba[3] = 0xff;                                                  \
-   } while (0)
-
-#define TAG(x) r200##x##_RGB565
-#include "spantmp.h"
+#define TAG(x)    radeon##x##_RGB565
+#define TAG2(x,y) radeon##x##_RGB565##y
+#define GET_PTR(X,Y) radeon_ptr16(rrb, (X), (Y))
+#include "spantmp2.h"
 
 /* 32 bit, ARGB8888 color spanline and pixel functions
  */
-#undef INIT_MONO_PIXEL
-#define INIT_MONO_PIXEL(p, color) \
-  p = PACK_COLOR_8888( color[3], color[0], color[1], color[2] )
-
-#define WRITE_RGBA( _x, _y, r, g, b, a )                       \
-do {                                                           \
-   *(GLuint *)(buf + _x*4 + _y*pitch) = ((b <<  0) |           \
-                                        (g <<  8) |            \
-                                        (r << 16) |            \
-                                        (a << 24) );           \
-} while (0)
-
-#define WRITE_PIXEL( _x, _y, p )                       \
-do {                                                   \
-   *(GLuint *)(buf + _x*4 + _y*pitch) = p;             \
-} while (0)
-
-#define READ_RGBA( rgba, _x, _y )                              \
-do {                                                           \
-   volatile GLuint *ptr = (volatile GLuint *)(read_buf + _x*4 + _y*pitch); \
-   GLuint p = *ptr;                                    \
-   rgba[0] = (p >> 16) & 0xff;                                 \
-   rgba[1] = (p >>  8) & 0xff;                                 \
-   rgba[2] = (p >>  0) & 0xff;                                 \
-   rgba[3] = (p >> 24) & 0xff;                                 \
-} while (0)
-
-#define TAG(x) r200##x##_ARGB8888
-#include "spantmp.h"
+#define SPANTMP_PIXEL_FMT GL_BGRA
+#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
 
+#define TAG(x)    radeon##x##_ARGB8888
+#define TAG2(x,y) radeon##x##_ARGB8888##y
+#define GET_PTR(X,Y) radeon_ptr32(rrb, (X), (Y))
+#include "spantmp2.h"
 
 
 /* ================================================================
@@ -188,77 +109,43 @@ do {                                                              \
  * manner as the engine.  In each case, the linear block address (ba)
  * is calculated, and then wired with x and y to produce the final
  * memory address.
+ * The chip will do address translation on its own if the surface registers
+ * are set up correctly. It is not quite enough to get it working with hyperz too...
  */
 
-#define BIT(x,b) ((x & (1<<b))>>b)
-static GLuint r200_mba_z32( r200ContextPtr rmesa,
-                                      GLint x, GLint y )
-{
-   GLuint pitch = rmesa->r200Screen->frontPitch;
-   GLuint b = ((y & 0x3FF) >> 4) * ((pitch & 0xFFF) >> 5) + ((x & 0x3FF) >> 5);
-   GLuint a = 
-      (BIT(x,0) << 2) |
-      (BIT(y,0) << 3) |
-      (BIT(x,1) << 4) |
-      (BIT(y,1) << 5) |
-      (BIT(x,3) << 6) |
-      (BIT(x,4) << 7) |
-      (BIT(x,2) << 8) |
-      (BIT(y,2) << 9) |
-      (BIT(y,3) << 10) |
-      (((pitch & 0x20) ? (b & 0x01) : ((b & 0x01) ^ (BIT(y,4)))) << 11) |
-      ((b >> 1) << 12);
-   return a;
-}
-
-static GLuint r200_mba_z16( r200ContextPtr rmesa, GLint x, GLint y )
-{
-   GLuint pitch = rmesa->r200Screen->frontPitch;
-   GLuint b = ((y & 0x3FF) >> 4) * ((pitch & 0xFFF) >> 6) + ((x & 0x3FF) >> 6);
-   GLuint a = 
-      (BIT(x,0) << 1) |
-      (BIT(y,0) << 2) |
-      (BIT(x,1) << 3) |
-      (BIT(y,1) << 4) |
-      (BIT(x,2) << 5) |
-      (BIT(x,4) << 6) |
-      (BIT(x,5) << 7) |
-      (BIT(x,3) << 8) |
-      (BIT(y,2) << 9) |
-      (BIT(y,3) << 10) |
-      (((pitch & 0x40) ? (b & 0x01) : ((b & 0x01) ^ (BIT(y,4)))) << 11) |
-      ((b >> 1) << 12);
-   return a;
-}
-
-
 /* 16-bit depth buffer functions
  */
+#define VALUE_TYPE GLushort
+
 #define WRITE_DEPTH( _x, _y, d )                                       \
-   *(GLushort *)(buf + r200_mba_z16( rmesa, _x + xo, _y + yo )) = d;
+   *(GLushort *)radeon_ptr(rrb, _x + xo, _y + yo) = d
 
 #define READ_DEPTH( d, _x, _y )                                                \
-   d = *(GLushort *)(buf + r200_mba_z16( rmesa, _x + xo, _y + yo ));
+   d = *(GLushort *)radeon_ptr(rrb, _x + xo, _y + yo)
 
-#define TAG(x) r200##x##_16
+#define TAG(x) radeon##x##_z16
 #include "depthtmp.h"
 
+
 /* 24 bit depth, 8 bit stencil depthbuffer functions
  */
+#define VALUE_TYPE GLuint
+
 #define WRITE_DEPTH( _x, _y, d )                                       \
 do {                                                                   \
-   GLuint offset = r200_mba_z32( rmesa, _x + xo, _y + yo );            \
-   GLuint tmp = *(GLuint *)(buf + offset);                             \
+   GLuint *_ptr = (GLuint*)radeon_ptr32(rrb, _x + xo, _y + yo);                \
+   GLuint tmp = *_ptr;                                                 \
    tmp &= 0xff000000;                                                  \
    tmp |= ((d) & 0x00ffffff);                                          \
-   *(GLuint *)(buf + offset) = tmp;                                    \
+   *_ptr = tmp;                                        \
 } while (0)
 
 #define READ_DEPTH( d, _x, _y )                                                \
-   d = *(GLuint *)(buf + r200_mba_z32( rmesa, _x + xo,         \
-                                        _y + yo )) & 0x00ffffff;
+   do {                                                                        \
+      d = (*(GLuint*)(radeon_ptr32(rrb, _x + xo, _y + yo)) & 0x00ffffff); \
+   }while(0)
 
-#define TAG(x) r200##x##_24_8
+#define TAG(x) radeon##x##_z24_s8
 #include "depthtmp.h"
 
 
@@ -270,163 +157,48 @@ do {                                                                     \
  */
 #define WRITE_STENCIL( _x, _y, d )                                     \
 do {                                                                   \
-   GLuint offset = r200_mba_z32( rmesa, _x + xo, _y + yo );            \
-   GLuint tmp = *(GLuint *)(buf + offset);                             \
+   GLuint *_ptr = (GLuint*)radeon_ptr32(rrb, _x + xo, _y + yo);                \
+   GLuint tmp = *_ptr;                         \
    tmp &= 0x00ffffff;                                                  \
    tmp |= (((d) & 0xff) << 24);                                                \
-   *(GLuint *)(buf + offset) = tmp;                                    \
+   *_ptr = tmp;                                        \
 } while (0)
 
 #define READ_STENCIL( d, _x, _y )                                      \
 do {                                                                   \
-   GLuint offset = r200_mba_z32( rmesa, _x + xo, _y + yo );            \
-   GLuint tmp = *(GLuint *)(buf + offset);                             \
+   GLuint *_ptr = (GLuint*)radeon_ptr32(rrb, _x + xo, _y + yo);                \
+   GLuint tmp = *_ptr;                                                 \
    tmp &= 0xff000000;                                                  \
    d = tmp >> 24;                                                      \
 } while (0)
 
-#define TAG(x) r200##x##_24_8
+#define TAG(x) radeon##x##_z24_s8
 #include "stenciltmp.h"
 
 
-/*
- * This function is called to specify which buffer to read and write
- * for software rasterization (swrast) fallbacks.  This doesn't necessarily
- * correspond to glDrawBuffer() or glReadBuffer() calls.
- */
-static void r200SetBuffer( GLcontext *ctx,
-                           GLframebuffer *colorBuffer,
-                           GLuint bufferBit )
+void r200InitSpanFuncs( GLcontext *ctx )
 {
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   switch ( bufferBit ) {
-   case DD_FRONT_LEFT_BIT:
-      if ( rmesa->doPageFlip && rmesa->sarea->pfCurrentPage == 1 ) {
-        rmesa->state.pixel.readOffset = rmesa->r200Screen->backOffset;
-        rmesa->state.pixel.readPitch  = rmesa->r200Screen->backPitch;
-        rmesa->state.color.drawOffset = rmesa->r200Screen->backOffset;
-        rmesa->state.color.drawPitch  = rmesa->r200Screen->backPitch;
-      } else {
-       rmesa->state.pixel.readOffset = rmesa->r200Screen->frontOffset;
-       rmesa->state.pixel.readPitch  = rmesa->r200Screen->frontPitch;
-       rmesa->state.color.drawOffset = rmesa->r200Screen->frontOffset;
-       rmesa->state.color.drawPitch  = rmesa->r200Screen->frontPitch;
-      }
-      break;
-   case DD_BACK_LEFT_BIT:
-      if ( rmesa->doPageFlip && rmesa->sarea->pfCurrentPage == 1 ) {
-       rmesa->state.pixel.readOffset = rmesa->r200Screen->frontOffset;
-       rmesa->state.pixel.readPitch  = rmesa->r200Screen->frontPitch;
-       rmesa->state.color.drawOffset = rmesa->r200Screen->frontOffset;
-       rmesa->state.color.drawPitch  = rmesa->r200Screen->frontPitch;
-      } else {
-        rmesa->state.pixel.readOffset = rmesa->r200Screen->backOffset;
-        rmesa->state.pixel.readPitch  = rmesa->r200Screen->backPitch;
-        rmesa->state.color.drawOffset = rmesa->r200Screen->backOffset;
-        rmesa->state.color.drawPitch  = rmesa->r200Screen->backPitch;
-      }
-      break;
-   default:
-      _mesa_problem(ctx, "Bad bufferBit in %s", __FUNCTION__);
-      break;
-   }
+   struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx);
+   swdd->SpanRenderStart          = radeonSpanRenderStart;
+   swdd->SpanRenderFinish         = radeonSpanRenderFinish; 
 }
 
-/* Move locking out to get reasonable span performance (10x better
- * than doing this in HW_LOCK above).  WaitForIdle() is the main
- * culprit.
- */
 
-static void r200SpanRenderStart( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-
-   R200_FIREVERTICES( rmesa );
-   LOCK_HARDWARE( rmesa );
-   r200WaitForIdleLocked( rmesa );
-
-   /* Read & rewrite the first pixel in the frame buffer.  This should
-    * be a noop, right?  In fact without this conform fails as reading
-    * from the framebuffer sometimes produces old results -- the
-    * on-card read cache gets mixed up and doesn't notice that the
-    * framebuffer has been updated.
-    *
-    * In the worst case this is buggy too as p might get the wrong
-    * value first time, so really need a hidden pixel somewhere for this.
-    */
-   {
-      int p;
-      volatile int *read_buf = (volatile int *)(rmesa->dri.screen->pFB + 
-                                               rmesa->state.pixel.readOffset);
-      p = *read_buf;
-      *read_buf = p;
-   }
-}
-
-static void r200SpanRenderFinish( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-   _swrast_flush( ctx );
-   UNLOCK_HARDWARE( rmesa );
-}
 
-void r200InitSpanFuncs( GLcontext *ctx )
+/**
+ * Plug in the Get/Put routines for the given driRenderbuffer.
+ */
+void radeonSetSpanFunctions(struct radeon_renderbuffer *rrb)
 {
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx);
-
-   swdd->SetBuffer = r200SetBuffer;
-
-   switch ( rmesa->r200Screen->cpp ) {
-   case 2:
-      swdd->WriteRGBASpan      = r200WriteRGBASpan_RGB565;
-      swdd->WriteRGBSpan       = r200WriteRGBSpan_RGB565;
-      swdd->WriteMonoRGBASpan  = r200WriteMonoRGBASpan_RGB565;
-      swdd->WriteRGBAPixels    = r200WriteRGBAPixels_RGB565;
-      swdd->WriteMonoRGBAPixels        = r200WriteMonoRGBAPixels_RGB565;
-      swdd->ReadRGBASpan       = r200ReadRGBASpan_RGB565;
-      swdd->ReadRGBAPixels      = r200ReadRGBAPixels_RGB565;
-      break;
-
-   case 4:
-      swdd->WriteRGBASpan      = r200WriteRGBASpan_ARGB8888;
-      swdd->WriteRGBSpan       = r200WriteRGBSpan_ARGB8888;
-      swdd->WriteMonoRGBASpan   = r200WriteMonoRGBASpan_ARGB8888;
-      swdd->WriteRGBAPixels     = r200WriteRGBAPixels_ARGB8888;
-      swdd->WriteMonoRGBAPixels = r200WriteMonoRGBAPixels_ARGB8888;
-      swdd->ReadRGBASpan       = r200ReadRGBASpan_ARGB8888;
-      swdd->ReadRGBAPixels      = r200ReadRGBAPixels_ARGB8888;
-      break;
-
-   default:
-      break;
-   }
-
-   switch ( rmesa->glCtx->Visual.depthBits ) {
-   case 16:
-      swdd->ReadDepthSpan      = r200ReadDepthSpan_16;
-      swdd->WriteDepthSpan     = r200WriteDepthSpan_16;
-      swdd->ReadDepthPixels    = r200ReadDepthPixels_16;
-      swdd->WriteDepthPixels   = r200WriteDepthPixels_16;
-      break;
-
-   case 24:
-      swdd->ReadDepthSpan      = r200ReadDepthSpan_24_8;
-      swdd->WriteDepthSpan     = r200WriteDepthSpan_24_8;
-      swdd->ReadDepthPixels    = r200ReadDepthPixels_24_8;
-      swdd->WriteDepthPixels   = r200WriteDepthPixels_24_8;
-
-      swdd->ReadStencilSpan    = r200ReadStencilSpan_24_8;
-      swdd->WriteStencilSpan   = r200WriteStencilSpan_24_8;
-      swdd->ReadStencilPixels  = r200ReadStencilPixels_24_8;
-      swdd->WriteStencilPixels = r200WriteStencilPixels_24_8;
-      break;
-
-   default:
-      break;
-   }
-
-   swdd->SpanRenderStart          = r200SpanRenderStart;
-   swdd->SpanRenderFinish         = r200SpanRenderFinish; 
+       if (rrb->base.InternalFormat == GL_RGB5) {
+               radeonInitPointers_RGB565(&rrb->base);
+       } else if (rrb->base.InternalFormat == GL_RGBA8) {
+               radeonInitPointers_ARGB8888(&rrb->base);
+       } else if (rrb->base.InternalFormat == GL_DEPTH_COMPONENT16) {
+               radeonInitDepthPointers_z16(&rrb->base);
+       } else if (rrb->base.InternalFormat == GL_DEPTH_COMPONENT24) {
+               radeonInitDepthPointers_z24_s8(&rrb->base);
+       } else if (rrb->base.InternalFormat == GL_STENCIL_INDEX8_EXT) {
+               radeonInitStencilPointers_z24_s8(&rrb->base);
+       }
 }