* Keith Whitwell <keith@tungstengraphics.com>
*
*/
-#include <unistd.h>
-#include "glheader.h"
-#include "imports.h"
+
+#include "main/glheader.h"
#include "swrast/swrast.h"
#include "r300_state.h"
#include "drirenderbuffer.h"
-
#define DBG 0
-
/*
* Note that all information needed to access pixels in a renderbuffer
* should be obtained through the gl_renderbuffer parameter, not per-context
#define HW_UNLOCK()
-
-
/* ================================================================
* Color buffer
*/
#define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 2)
#include "spantmp2.h"
-
/* 32 bit, ARGB8888 color spanline and pixel functions
*/
#define SPANTMP_PIXEL_FMT GL_BGRA
#define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 4)
#include "spantmp2.h"
-
/* ================================================================
* Depth buffer
*/
* too...
*/
-static GLuint
-radeon_mba_z32( const driRenderbuffer *drb, GLint x, GLint y )
+static GLuint radeon_mba_z32(const driRenderbuffer * drb, GLint x, GLint y)
{
- GLuint pitch = drb->pitch;
- if (1 /*|| drb->depthHasSurface */) {
- return 4 * (x + y * pitch);
- }
- else {
- GLuint ba, address = 0; /* a[0..1] = 0 */
-
- ba = (y / 8) * (pitch / 8) + (x / 8);
-
- address |= (x & 0x7) << 2; /* a[2..4] = x[0..2] */
- address |= (y & 0x3) << 5; /* a[5..6] = y[0..1] */
- address |=
- (((x & 0x10) >> 2) ^ (y & 0x4)) << 5; /* a[7] = x[4] ^ y[2] */
- address |= (ba & 0x3) << 8; /* a[8..9] = ba[0..1] */
-
- address |= (y & 0x8) << 7; /* a[10] = y[3] */
- address |=
- (((x & 0x8) << 1) ^ (y & 0x10)) << 7; /* a[11] = x[3] ^ y[4] */
- address |= (ba & ~0x3) << 10; /* a[12..] = ba[2..] */
-
- return address;
- }
+ GLuint pitch = drb->pitch;
+ if (drb->depthHasSurface) {
+ return 4 * (x + y * pitch);
+ } else {
+ GLuint ba, address = 0; /* a[0..1] = 0 */
+
+#ifdef COMPILE_R300
+ ba = (y / 8) * (pitch / 8) + (x / 8);
+#else
+ ba = (y / 16) * (pitch / 16) + (x / 16);
+#endif
+
+ address |= (x & 0x7) << 2; /* a[2..4] = x[0..2] */
+ address |= (y & 0x3) << 5; /* a[5..6] = y[0..1] */
+ address |= (((x & 0x10) >> 2) ^ (y & 0x4)) << 5; /* a[7] = x[4] ^ y[2] */
+ address |= (ba & 0x3) << 8; /* a[8..9] = ba[0..1] */
+
+ address |= (y & 0x8) << 7; /* a[10] = y[3] */
+ address |= (((x & 0x8) << 1) ^ (y & 0x10)) << 7; /* a[11] = x[3] ^ y[4] */
+ address |= (ba & ~0x3) << 10; /* a[12..] = ba[2..] */
+
+ return address;
+ }
}
-
static INLINE GLuint
-radeon_mba_z16( const driRenderbuffer *drb, GLint x, GLint y )
+radeon_mba_z16(const driRenderbuffer * drb, GLint x, GLint y)
{
- GLuint pitch = drb->pitch;
- if (1 /*|| drb->depthHasSurface */) {
- return 2 * (x + y * pitch);
- }
- else {
- GLuint ba, address = 0; /* a[0] = 0 */
-
- ba = (y / 16) * (pitch / 32) + (x / 32);
-
- address |= (x & 0x7) << 1; /* a[1..3] = x[0..2] */
- address |= (y & 0x7) << 4; /* a[4..6] = y[0..2] */
- address |= (x & 0x8) << 4; /* a[7] = x[3] */
- address |= (ba & 0x3) << 8; /* a[8..9] = ba[0..1] */
- address |= (y & 0x8) << 7; /* a[10] = y[3] */
- address |= ((x & 0x10) ^ (y & 0x10)) << 7;/* a[11] = x[4] ^ y[4] */
- address |= (ba & ~0x3) << 10; /* a[12..] = ba[2..] */
-
- return address;
- }
+ GLuint pitch = drb->pitch;
+ if (drb->depthHasSurface) {
+ return 2 * (x + y * pitch);
+ } else {
+ GLuint ba, address = 0; /* a[0] = 0 */
+
+ ba = (y / 16) * (pitch / 32) + (x / 32);
+
+ address |= (x & 0x7) << 1; /* a[1..3] = x[0..2] */
+ address |= (y & 0x7) << 4; /* a[4..6] = y[0..2] */
+ address |= (x & 0x8) << 4; /* a[7] = x[3] */
+ address |= (ba & 0x3) << 8; /* a[8..9] = ba[0..1] */
+ address |= (y & 0x8) << 7; /* a[10] = y[3] */
+ address |= ((x & 0x10) ^ (y & 0x10)) << 7; /* a[11] = x[4] ^ y[4] */
+ address |= (ba & ~0x3) << 10; /* a[12..] = ba[2..] */
+
+ return address;
+ }
}
-
/* 16-bit depth buffer functions
*/
+#define VALUE_TYPE GLushort
+
#define WRITE_DEPTH( _x, _y, d ) \
*(GLushort *)(buf + radeon_mba_z16( drb, _x + xo, _y + yo )) = d;
#define TAG(x) radeon##x##_z16
#include "depthtmp.h"
-
/* 24 bit depth, 8 bit stencil depthbuffer functions
*
* Careful: It looks like the R300 uses ZZZS byte order while the R200
* uses SZZZ for 24 bit depth, 8 bit stencil mode.
*/
+#define VALUE_TYPE GLuint
+
+#ifdef COMPILE_R300
#define WRITE_DEPTH( _x, _y, d ) \
do { \
GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo ); \
tmp |= ((d << 8) & 0xffffff00); \
*(GLuint *)(buf + offset) = tmp; \
} while (0)
+#else
+#define WRITE_DEPTH( _x, _y, d ) \
+do { \
+ GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo ); \
+ GLuint tmp = *(GLuint *)(buf + offset); \
+ tmp &= 0xff000000; \
+ tmp |= ((d) & 0x00ffffff); \
+ *(GLuint *)(buf + offset) = tmp; \
+} while (0)
+#endif
+#ifdef COMPILE_R300
#define READ_DEPTH( d, _x, _y ) \
do { \
d = (*(GLuint *)(buf + radeon_mba_z32( drb, _x + xo, \
_y + yo )) & 0xffffff00) >> 8; \
}while(0)
+#else
+#define READ_DEPTH( d, _x, _y ) \
+ d = *(GLuint *)(buf + radeon_mba_z32( drb, _x + xo, \
+ _y + yo )) & 0x00ffffff;
+#endif
#define TAG(x) radeon##x##_z24_s8
#include "depthtmp.h"
-
/* ================================================================
* Stencil buffer
*/
/* 24 bit depth, 8 bit stencil depthbuffer functions
*/
+#ifdef COMPILE_R300
#define WRITE_STENCIL( _x, _y, d ) \
do { \
GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo ); \
tmp |= (d) & 0xff; \
*(GLuint *)(buf + offset) = tmp; \
} while (0)
+#else
+#define WRITE_STENCIL( _x, _y, d ) \
+do { \
+ GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo ); \
+ GLuint tmp = *(GLuint *)(buf + offset); \
+ tmp &= 0x00ffffff; \
+ tmp |= (((d) & 0xff) << 24); \
+ *(GLuint *)(buf + offset) = tmp; \
+} while (0)
+#endif
+#ifdef COMPILE_R300
#define READ_STENCIL( d, _x, _y ) \
do { \
GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo ); \
GLuint tmp = *(GLuint *)(buf + offset); \
d = tmp & 0x000000ff; \
} while (0)
+#else
+#define READ_STENCIL( d, _x, _y ) \
+do { \
+ GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo ); \
+ GLuint tmp = *(GLuint *)(buf + offset); \
+ d = (tmp & 0xff000000) >> 24; \
+} while (0)
+#endif
#define TAG(x) radeon##x##_z24_s8
#include "stenciltmp.h"
-
-
/* Move locking out to get reasonable span performance (10x better
* than doing this in HW_LOCK above). WaitForIdle() is the main
* culprit.
*/
-static void radeonSpanRenderStart( GLcontext *ctx )
+static void radeonSpanRenderStart(GLcontext * ctx)
{
- radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
- {
- static int first = 1;
- r300ContextPtr r300 = (r300ContextPtr)rmesa;
-
- if (first) {
- r300->span_dlocking = getenv("R300_SPAN_DISABLE_LOCKING") ? 1 : 0;
- if (r300->span_dlocking == 0) {
- fprintf(stderr, "Try R300_SPAN_DISABLE_LOCKING env var if this hangs.\n");
- fflush(stderr);
- sleep(1);
- }
- first = 0;
- }
-
- if (r300->span_dlocking) {
- r300Flush(ctx);
- LOCK_HARDWARE( rmesa );
- radeonWaitForIdleLocked( rmesa );
- UNLOCK_HARDWARE( rmesa );
-
- return;
+ radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+#ifdef COMPILE_R300
+ r300ContextPtr r300 = (r300ContextPtr) rmesa;
+ R300_FIREVERTICES(r300);
+#else
+ RADEON_FIREVERTICES(rmesa);
+#endif
+ LOCK_HARDWARE(rmesa);
+ radeonWaitForIdleLocked(rmesa);
+
+ /* Read 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.
+ *
+ * Note that we should probably be reading some otherwise unused
+ * region of VRAM, otherwise we might get incorrect results when
+ * reading pixels from the top left of the screen.
+ *
+ * I found this problem on an R420 with glean's texCube test.
+ * Note that the R200 span code also *writes* the first pixel in the
+ * framebuffer, but I've found this to be unnecessary.
+ * -- Nicolai Hähnle, June 2008
+ */
+ {
+ int p;
+ driRenderbuffer *drb =
+ (driRenderbuffer *) ctx->WinSysDrawBuffer->_ColorDrawBuffers[0];
+ volatile int *buf =
+ (volatile int *)(rmesa->dri.screen->pFB + drb->offset);
+ p = *buf;
}
- }
- // R300_FIREVERTICES( rmesa );
- // old code has flush
- r300Flush(ctx);
- LOCK_HARDWARE( rmesa );
- radeonWaitForIdleLocked( rmesa );
}
-static void radeonSpanRenderFinish( GLcontext *ctx )
+static void radeonSpanRenderFinish(GLcontext * ctx)
{
- radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
- r300ContextPtr r300 = (r300ContextPtr)rmesa;
- _swrast_flush( ctx );
- if (r300->span_dlocking == 0)
- UNLOCK_HARDWARE( rmesa );
+ radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+ _swrast_flush(ctx);
+ UNLOCK_HARDWARE(rmesa);
}
-void radeonInitSpanFuncs( GLcontext *ctx )
+void radeonInitSpanFuncs(GLcontext * ctx)
{
- struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx);
- swdd->SpanRenderStart = radeonSpanRenderStart;
- swdd->SpanRenderFinish = radeonSpanRenderFinish;
+ struct swrast_device_driver *swdd =
+ _swrast_GetDeviceDriverReference(ctx);
+ swdd->SpanRenderStart = radeonSpanRenderStart;
+ swdd->SpanRenderFinish = radeonSpanRenderFinish;
}
-
/**
* Plug in the Get/Put routines for the given driRenderbuffer.
*/
-void
-radeonSetSpanFunctions(driRenderbuffer *drb, const GLvisual *vis)
+void radeonSetSpanFunctions(driRenderbuffer * drb, const GLvisual * vis)
{
- if (drb->Base.InternalFormat == GL_RGBA) {
- if (vis->redBits == 5 && vis->greenBits == 6 && vis->blueBits == 5) {
- radeonInitPointers_RGB565(&drb->Base);
- }
- else {
- radeonInitPointers_ARGB8888(&drb->Base);
- }
- }
- else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT16) {
- radeonInitDepthPointers_z16(&drb->Base);
- }
- else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT24) {
- radeonInitDepthPointers_z24_s8(&drb->Base);
- }
- else if (drb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) {
- radeonInitStencilPointers_z24_s8(&drb->Base);
- }
+ if (drb->Base.InternalFormat == GL_RGBA) {
+ if (vis->redBits == 5 && vis->greenBits == 6
+ && vis->blueBits == 5) {
+ radeonInitPointers_RGB565(&drb->Base);
+ } else {
+ radeonInitPointers_ARGB8888(&drb->Base);
+ }
+ } else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT16) {
+ radeonInitDepthPointers_z16(&drb->Base);
+ } else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT24) {
+ radeonInitDepthPointers_z24_s8(&drb->Base);
+ } else if (drb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) {
+ radeonInitStencilPointers_z24_s8(&drb->Base);
+ }
}