diff options
Diffstat (limited to 'shared')
39 files changed, 971 insertions, 1137 deletions
diff --git a/shared/intel_batchbuffer.c b/shared/intel_batchbuffer.c index 6aa36d1..ca6e2fa 100644 --- a/shared/intel_batchbuffer.c +++ b/shared/intel_batchbuffer.c @@ -201,11 +201,6 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file, drm_intel_bo_reference(intel->first_post_swapbuffers_batch); } - if (intel->first_post_swapbuffers_batch == NULL) { - intel->first_post_swapbuffers_batch = intel->batch->buf; - drm_intel_bo_reference(intel->first_post_swapbuffers_batch); - } - if (used == 0) { batch->cliprect_mode = IGNORE_CLIPRECTS; return; @@ -215,10 +210,10 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file, fprintf(stderr, "%s:%d: Batchbuffer flush with %db used\n", file, line, used); + batch->reserved_space = 0; /* Emit a flush if the bufmgr doesn't do it for us. */ if (intel->always_flush_cache || !intel->ttm) { - *(GLuint *) (batch->ptr) = intel->vtbl.flush_cmd(); - batch->ptr += 4; + intel_batchbuffer_emit_mi_flush(batch); used = batch->ptr - batch->map; } @@ -249,6 +244,8 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file, if (intel->vtbl.finish_batch) intel->vtbl.finish_batch(intel); + batch->reserved_space = BATCH_RESERVED; + /* TODO: Just pass the relocation list and dma buffer up to the * kernel. */ @@ -304,3 +301,31 @@ intel_batchbuffer_data(struct intel_batchbuffer *batch, __memcpy(batch->ptr, data, bytes); batch->ptr += bytes; } + +/* Emit a pipelined flush to either flush render and texture cache for + * reading from a FBO-drawn texture, or flush so that frontbuffer + * render appears on the screen in DRI1. + * + * This is also used for the always_flush_cache driconf debug option. + */ +void +intel_batchbuffer_emit_mi_flush(struct intel_batchbuffer *batch) +{ + struct intel_context *intel = batch->intel; + + if (intel->gen >= 4) { + BEGIN_BATCH(4, IGNORE_CLIPRECTS); + OUT_BATCH(_3DSTATE_PIPE_CONTROL | + PIPE_CONTROL_INSTRUCTION_FLUSH | + PIPE_CONTROL_WRITE_FLUSH | + PIPE_CONTROL_NO_WRITE); + OUT_BATCH(0); /* write address */ + OUT_BATCH(0); /* write data */ + OUT_BATCH(0); /* write data */ + ADVANCE_BATCH(); + } else { + BEGIN_BATCH(1, IGNORE_CLIPRECTS); + OUT_BATCH(MI_FLUSH); + ADVANCE_BATCH(); + } +} diff --git a/shared/intel_batchbuffer.h b/shared/intel_batchbuffer.h index 51579df..d4a9445 100644 --- a/shared/intel_batchbuffer.h +++ b/shared/intel_batchbuffer.h @@ -62,6 +62,7 @@ struct intel_batchbuffer } emit; GLuint dirty_state; + GLuint reserved_space; }; struct intel_batchbuffer *intel_batchbuffer_alloc(struct intel_context @@ -95,6 +96,7 @@ GLboolean intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, uint32_t read_domains, uint32_t write_domain, uint32_t offset); +void intel_batchbuffer_emit_mi_flush(struct intel_batchbuffer *batch); /* Inline functions - might actually be better off with these * non-inlined. Certainly better off switching all command packets to @@ -104,7 +106,7 @@ GLboolean intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, static INLINE GLint intel_batchbuffer_space(struct intel_batchbuffer *batch) { - return (batch->size - BATCH_RESERVED) - (batch->ptr - batch->map); + return (batch->size - batch->reserved_space) - (batch->ptr - batch->map); } @@ -157,7 +159,7 @@ intel_batchbuffer_require_space(struct intel_batchbuffer *batch, #define OUT_BATCH(d) intel_batchbuffer_emit_dword(intel->batch, d) #define OUT_RELOC(buf, read_domains, write_domain, delta) do { \ - assert((delta) >= 0); \ + assert((unsigned) (delta) < buf->size); \ intel_batchbuffer_emit_reloc(intel->batch, buf, \ read_domains, write_domain, delta); \ } while (0) @@ -173,12 +175,4 @@ intel_batchbuffer_require_space(struct intel_batchbuffer *batch, intel->batch->emit.start_ptr = NULL; \ } while(0) - -static INLINE void -intel_batchbuffer_emit_mi_flush(struct intel_batchbuffer *batch) -{ - intel_batchbuffer_require_space(batch, 4, IGNORE_CLIPRECTS); - intel_batchbuffer_emit_dword(batch, MI_FLUSH); -} - #endif diff --git a/shared/intel_blit.c b/shared/intel_blit.c index 0c5be4c..9f638b0 100644 --- a/shared/intel_blit.c +++ b/shared/intel_blit.c @@ -26,13 +26,9 @@ **************************************************************************/ -#include <stdio.h> -#include <errno.h> - #include "main/mtypes.h" #include "main/context.h" #include "main/enums.h" -#include "main/texformat.h" #include "main/colormac.h" #include "intel_blit.h" @@ -374,8 +370,6 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) skipBuffers = BUFFER_BIT_STENCIL; } - /* XXX Move this flush/lock into the following conditional? */ - intelFlush(&intel->ctx); LOCK_HARDWARE(intel); intel_get_cliprects(intel, &cliprects, &num_cliprects, &x_off, &y_off); @@ -441,6 +435,10 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) intel_region_buffer(intel, irb->region, all ? INTEL_WRITE_FULL : INTEL_WRITE_PART); + int x1 = b.x1 + irb->region->draw_x; + int y1 = b.y1 + irb->region->draw_y; + int x2 = b.x2 + irb->region->draw_x; + int y2 = b.y2 + irb->region->draw_y; GLuint clearVal; GLint pitch, cpp; @@ -449,11 +447,10 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) pitch = irb->region->pitch; cpp = irb->region->cpp; - DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", + DBG("%s dst:buf(%p)/%d %d,%d sz:%dx%d\n", __FUNCTION__, irb->region->buffer, (pitch * cpp), - irb->region->draw_offset, - b.x1, b.y1, b.x2 - b.x1, b.y2 - b.y1); + x1, y1, x2 - x1, y2 - y1); BR13 = 0xf0 << 16; CMD = XY_COLOR_BLT_CMD; @@ -499,8 +496,9 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) CLAMPED_FLOAT_TO_UBYTE(clear[2], color[2]); CLAMPED_FLOAT_TO_UBYTE(clear[3], color[3]); - switch (irb->texformat->MesaFormat) { + switch (irb->Base.Format) { case MESA_FORMAT_ARGB8888: + case MESA_FORMAT_XRGB8888: clearVal = intel->ClearColor8888; break; case MESA_FORMAT_RGB565: @@ -516,7 +514,7 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) break; default: _mesa_problem(ctx, "Unexpected renderbuffer format: %d\n", - irb->texformat->MesaFormat); + irb->Base.Format); clearVal = 0; } } @@ -526,17 +524,17 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) buf, irb->Base.Name); */ - assert(b.x1 < b.x2); - assert(b.y1 < b.y2); + assert(x1 < x2); + assert(y1 < y2); BEGIN_BATCH(6, REFERENCES_CLIPRECTS); OUT_BATCH(CMD); OUT_BATCH(BR13); - OUT_BATCH((b.y1 << 16) | b.x1); - OUT_BATCH((b.y2 << 16) | b.x2); + OUT_BATCH((y1 << 16) | x1); + OUT_BATCH((y2 << 16) | x2); OUT_RELOC(write_buffer, I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER, - irb->region->draw_offset); + 0); OUT_BATCH(clearVal); ADVANCE_BATCH(); clearMask &= ~bufBit; /* turn off bit, for faster loop exit */ diff --git a/shared/intel_buffer_objects.c b/shared/intel_buffer_objects.c index a022593..3b7015b 100644 --- a/shared/intel_buffer_objects.c +++ b/shared/intel_buffer_objects.c @@ -209,9 +209,23 @@ intel_bufferobj_subdata(GLcontext * ctx, memcpy((char *)intel_obj->sys_buffer + offset, data, size); else { /* Flush any existing batchbuffer that might reference this data. */ - intelFlush(ctx); + if (drm_intel_bo_busy(intel_obj->buffer) || + drm_intel_bo_references(intel->batch->buf, intel_obj->buffer)) { + drm_intel_bo *temp_bo; + + temp_bo = drm_intel_bo_alloc(intel->bufmgr, "subdata temp", size, 64); + + drm_intel_bo_subdata(temp_bo, 0, size, data); - dri_bo_subdata(intel_obj->buffer, offset, size, data); + intel_emit_linear_blit(intel, + intel_obj->buffer, offset, + temp_bo, 0, + size); + + drm_intel_bo_unreference(temp_bo); + } else { + dri_bo_subdata(intel_obj->buffer, offset, size, data); + } } } @@ -254,13 +268,14 @@ intel_bufferobj_map(GLcontext * ctx, if (intel_obj->sys_buffer) { obj->Pointer = intel_obj->sys_buffer; + obj->Length = obj->Size; + obj->Offset = 0; return obj->Pointer; } - /* Flush any existing batchbuffer that might have written to this - * buffer. - */ - intelFlush(ctx); + /* Flush any existing batchbuffer that might reference this data. */ + if (drm_intel_bo_references(intel->batch->buf, intel_obj->buffer)) + intelFlush(ctx); if (intel_obj->region) intel_bufferobj_cow(intel, intel_obj); @@ -330,7 +345,8 @@ intel_bufferobj_map_range(GLcontext * ctx, * the batchbuffer so that GEM knows about the buffer access for later * syncing. */ - if (!(access & GL_MAP_UNSYNCHRONIZED_BIT)) + if (!(access & GL_MAP_UNSYNCHRONIZED_BIT) && + drm_intel_bo_references(intel->batch->buf, intel_obj->buffer)) intelFlush(ctx); if (intel_obj->buffer == NULL) { diff --git a/shared/intel_buffers.c b/shared/intel_buffers.c index e7357e7..0564318 100644 --- a/shared/intel_buffers.c +++ b/shared/intel_buffers.c @@ -133,6 +133,25 @@ intel_get_cliprects(struct intel_context *intel, /** + * Check if we're about to draw into the front color buffer. + * If so, set the intel->front_buffer_dirty field to true. + */ +void +intel_check_front_buffer_rendering(struct intel_context *intel) +{ + const struct gl_framebuffer *fb = intel->ctx.DrawBuffer; + if (fb->Name == 0) { + /* drawing to window system buffer */ + if (fb->_NumColorDrawBuffers > 0) { + if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) { + intel->front_buffer_dirty = GL_TRUE; + } + } + } +} + + +/** * Update the hardware state for drawing into a window or framebuffer object. * * Called by glDrawBuffer, glBindFramebufferEXT, MakeCurrent, and other @@ -172,10 +191,15 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) return; } - /* - * How many color buffers are we drawing into? + /* How many color buffers are we drawing into? + * + * If there are zero buffers or the buffer is too big, don't configure any + * regions for hardware drawing. We'll fallback to software below. Not + * having regions set makes some of the software fallback paths faster. */ - if (fb->_NumColorDrawBuffers == 0) { + if ((fb->Width > ctx->Const.MaxRenderbufferSize) + || (fb->Height > ctx->Const.MaxRenderbufferSize) + || (fb->_NumColorDrawBuffers == 0)) { /* writing to 0 */ colorRegions[0] = NULL; intel->constant_cliprect = GL_TRUE; @@ -192,7 +216,7 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) } else { /* Get the intel_renderbuffer for the single colorbuffer we're drawing - * into, and set up cliprects if it's . + * into, and set up cliprects if it's a DRI1 window front buffer. */ if (fb->Name == 0) { intel->constant_cliprect = intel->driScreen->dri2.enabled; @@ -202,14 +226,12 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) intel_batchbuffer_flush(intel->batch); intel->front_cliprects = GL_TRUE; colorRegions[0] = intel_get_rb_region(fb, BUFFER_FRONT_LEFT); - - intel->front_buffer_dirty = GL_TRUE; } else { if (!intel->constant_cliprect && intel->front_cliprects) intel_batchbuffer_flush(intel->batch); intel->front_cliprects = GL_FALSE; - colorRegions[0]= intel_get_rb_region(fb, BUFFER_BACK_LEFT); + colorRegions[0] = intel_get_rb_region(fb, BUFFER_BACK_LEFT); } } else { @@ -257,7 +279,7 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) if (fb->_StencilBuffer && fb->_StencilBuffer->Wrapped) { irbStencil = intel_renderbuffer(fb->_StencilBuffer->Wrapped); if (irbStencil && irbStencil->region) { - ASSERT(irbStencil->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + ASSERT(irbStencil->Base.Format == MESA_FORMAT_S8_Z24); FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE); } else { diff --git a/shared/intel_buffers.h b/shared/intel_buffers.h index 6069d38..d7800f2 100644 --- a/shared/intel_buffers.h +++ b/shared/intel_buffers.h @@ -45,6 +45,8 @@ extern struct intel_region *intel_readbuf_region(struct intel_context *intel); extern struct intel_region *intel_drawbuf_region(struct intel_context *intel); +extern void intel_check_front_buffer_rendering(struct intel_context *intel); + extern void intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb); extern void intelInitBufferFuncs(struct dd_function_table *functions); diff --git a/shared/intel_clear.c b/shared/intel_clear.c index 1b0e221..f682ee3 100644 --- a/shared/intel_clear.c +++ b/shared/intel_clear.c @@ -75,6 +75,10 @@ intelClear(GLcontext *ctx, GLbitfield mask) struct gl_framebuffer *fb = ctx->DrawBuffer; GLuint i; + if (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_FRONT_RIGHT)) { + intel->front_buffer_dirty = GL_TRUE; + } + if (0) fprintf(stderr, "%s\n", __FUNCTION__); @@ -172,7 +176,7 @@ intelClear(GLcontext *ctx, GLbitfield mask) DBG("\n"); } - _mesa_meta_clear(&intel->ctx, tri_mask); + _mesa_meta_Clear(&intel->ctx, tri_mask); } if (swrast_mask) { diff --git a/shared/intel_context.c b/shared/intel_context.c index fce42e9..dd24b71 100644 --- a/shared/intel_context.c +++ b/shared/intel_context.c @@ -28,7 +28,7 @@ #include "main/glheader.h" #include "main/context.h" -#include "main/arrayobj.h" +/* #include "main/arrayobj.h" */ #include "main/extensions.h" #include "main/framebuffer.h" #include "main/imports.h" @@ -68,7 +68,7 @@ int INTEL_DEBUG = (0); #endif -#define DRIVER_DATE "20090712 2009Q2 RC3" +#define DRIVER_DATE "20091221 2009Q4" #define DRIVER_DATE_GEM "GEM " DRIVER_DATE @@ -189,19 +189,7 @@ intelGetString(GLcontext * ctx, GLenum name) static unsigned intel_bits_per_pixel(const struct intel_renderbuffer *rb) { - switch (rb->Base._ActualFormat) { - case GL_RGB5: - case GL_DEPTH_COMPONENT16: - return 16; - case GL_RGB8: - case GL_RGBA8: - case GL_DEPTH_COMPONENT24: - case GL_DEPTH24_STENCIL8_EXT: - case GL_STENCIL_INDEX8_EXT: - return 32; - default: - return 0; - } + return _mesa_get_format_bytes(rb->Base.Format) * 8; } void @@ -489,14 +477,14 @@ intel_flush(GLcontext *ctx, GLboolean needs_mi_flush) if (intel->Fallback) _swrast_flush(ctx); - if (!IS_965(intel->intelScreen->deviceID)) + if (intel->gen < 4) INTEL_FIREVERTICES(intel); /* Emit a flush so that any frontbuffer rendering that might have occurred * lands onscreen in a timely manner, even if the X Server doesn't trigger * a flush for us. */ - if (needs_mi_flush) + if (!intel->driScreen->dri2.enabled && needs_mi_flush) intel_batchbuffer_emit_mi_flush(intel->batch); if (intel->batch->map != intel->batch->ptr) @@ -588,11 +576,6 @@ intelInitDriverFunctions(struct dd_function_table *functions) functions->GetString = intelGetString; functions->UpdateState = intelInvalidateState; - functions->CopyColorTable = _swrast_CopyColorTable; - functions->CopyColorSubTable = _swrast_CopyColorSubTable; - functions->CopyConvolutionFilter1D = _swrast_CopyConvolutionFilter1D; - functions->CopyConvolutionFilter2D = _swrast_CopyConvolutionFilter2D; - intelInitTextureFuncs(functions); intelInitTextureImageFuncs(functions); intelInitTextureSubImageFuncs(functions); @@ -631,6 +614,13 @@ intelInitContext(struct intel_context *intel, intel->sarea = intelScreen->sarea; intel->driContext = driContextPriv; + if (IS_965(intel->intelScreen->deviceID)) + intel->gen = 4; + else if (IS_9XX(intel->intelScreen->deviceID)) + intel->gen = 3; + else + intel->gen = 2; + /* Dri stuff */ intel->hHWContext = driContextPriv->hHWContext; intel->driFd = sPriv->fd; @@ -638,17 +628,13 @@ intelInitContext(struct intel_context *intel, driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache, intel->driScreen->myNum, - IS_965(intelScreen->deviceID) ? "i965" : "i915"); + (intel->gen >= 4) ? "i965" : "i915"); if (intelScreen->deviceID == PCI_CHIP_I865_G) intel->maxBatchSize = 4096; else intel->maxBatchSize = BATCH_SZ; intel->bufmgr = intelScreen->bufmgr; - - if (0) /* for debug */ - drm_intel_bufmgr_set_debug(intel->bufmgr, 1); - intel->ttm = intelScreen->ttm; if (intel->ttm) { int bo_reuse_mode; @@ -704,7 +690,7 @@ intelInitContext(struct intel_context *intel, meta_init_metaops(ctx, &intel->meta); ctx->Const.MaxColorAttachments = 4; /* XXX FBO: review this */ - if (IS_965(intelScreen->deviceID)) { + if (intel->gen >= 4) { if (MAX_WIDTH > 8192) ctx->Const.MaxRenderbufferSize = 8192; } else { @@ -741,7 +727,7 @@ intelInitContext(struct intel_context *intel, break; } - if (IS_965(intelScreen->deviceID)) + if (intel->gen >= 4) intel->polygon_offset_scale /= 0xffff; intel->RenderIndex = ~0; @@ -754,12 +740,12 @@ intelInitContext(struct intel_context *intel, intel->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS); - if (IS_965(intelScreen->deviceID) && !intel->intelScreen->irq_active) { + if (intel->gen >= 4 && !intel->intelScreen->irq_active) { _mesa_printf("IRQs not active. Exiting\n"); exit(1); } - intelInitExtensions(ctx, GL_FALSE); + intelInitExtensions(ctx); INTEL_DEBUG = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); if (INTEL_DEBUG & DEBUG_BUFMGR) @@ -839,7 +825,7 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) _vbo_DestroyContext(&intel->ctx); _swrast_DestroyContext(&intel->ctx); - intel->Fallback = 0; /* don't call _swrast_Flush later */ + intel->Fallback = 0x0; /* don't call _swrast_Flush later */ intel_batchbuffer_free(intel->batch); intel->batch = NULL; @@ -944,10 +930,23 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, __DRIdrawablePrivate * driReadPriv) { __DRIscreenPrivate *psp = driDrawPriv->driScreenPriv; + struct intel_context *intel; + GET_CURRENT_CONTEXT(curCtx); + + if (driContextPriv) + intel = (struct intel_context *) driContextPriv->driverPrivate; + else + intel = NULL; + + /* According to the glXMakeCurrent() man page: "Pending commands to + * the previous context, if any, are flushed before it is released." + * But only flush if we're actually changing contexts. + */ + if (intel_context(curCtx) && intel_context(curCtx) != intel) { + _mesa_flush(curCtx); + } if (driContextPriv) { - struct intel_context *intel = - (struct intel_context *) driContextPriv->driverPrivate; struct intel_framebuffer *intel_fb = (struct intel_framebuffer *) driDrawPriv->driverPrivate; GLframebuffer *readFb = (GLframebuffer *) driReadPriv->driverPrivate; @@ -993,41 +992,35 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, _mesa_make_current(&intel->ctx, &intel_fb->Base, readFb); - /* The drawbuffer won't always be updated by _mesa_make_current: - */ - if (intel->ctx.DrawBuffer == &intel_fb->Base) { - - if (intel->driReadDrawable != driReadPriv) - intel->driReadDrawable = driReadPriv; - - if (intel->driDrawable != driDrawPriv) { - if (driDrawPriv->swap_interval == (unsigned)-1) { - int i; - - driDrawPriv->vblFlags = (intel->intelScreen->irq_active != 0) - ? driGetDefaultVBlankFlags(&intel->optionCache) - : VBLANK_FLAG_NO_IRQ; - - /* Prevent error printf if one crtc is disabled, this will - * be properly calculated in intelWindowMoved() next. - */ - driDrawPriv->vblFlags = intelFixupVblank(intel, driDrawPriv); - - (*psp->systemTime->getUST) (&intel_fb->swap_ust); - driDrawableInitVBlank(driDrawPriv); - intel_fb->vbl_waited = driDrawPriv->vblSeq; - - for (i = 0; i < 2; i++) { - if (intel_fb->color_rb[i]) - intel_fb->color_rb[i]->vbl_pending = driDrawPriv->vblSeq; - } - } - intel->driDrawable = driDrawPriv; - intelWindowMoved(intel); - } + intel->driReadDrawable = driReadPriv; + + if (intel->driDrawable != driDrawPriv) { + if (driDrawPriv->swap_interval == (unsigned)-1) { + int i; - intel_draw_buffer(&intel->ctx, &intel_fb->Base); + driDrawPriv->vblFlags = (intel->intelScreen->irq_active != 0) + ? driGetDefaultVBlankFlags(&intel->optionCache) + : VBLANK_FLAG_NO_IRQ; + + /* Prevent error printf if one crtc is disabled, this will + * be properly calculated in intelWindowMoved() next. + */ + driDrawPriv->vblFlags = intelFixupVblank(intel, driDrawPriv); + + (*psp->systemTime->getUST) (&intel_fb->swap_ust); + driDrawableInitVBlank(driDrawPriv); + intel_fb->vbl_waited = driDrawPriv->vblSeq; + + for (i = 0; i < 2; i++) { + if (intel_fb->color_rb[i]) + intel_fb->color_rb[i]->vbl_pending = driDrawPriv->vblSeq; + } + } + intel->driDrawable = driDrawPriv; + intelWindowMoved(intel); } + + intel_draw_buffer(&intel->ctx, &intel_fb->Base); } else { _mesa_make_current(NULL, NULL, NULL); diff --git a/shared/intel_context.h b/shared/intel_context.h index 03e7cf3..eb7be7d 100644 --- a/shared/intel_context.h +++ b/shared/intel_context.h @@ -61,6 +61,10 @@ typedef void (*intel_line_func) (struct intel_context *, intelVertex *, intelVertex *); typedef void (*intel_point_func) (struct intel_context *, intelVertex *); +/** + * Bits for intel->Fallback field + */ +/*@{*/ #define INTEL_FALLBACK_DRAW_BUFFER 0x1 #define INTEL_FALLBACK_READ_BUFFER 0x2 #define INTEL_FALLBACK_DEPTH_BUFFER 0x4 @@ -68,8 +72,10 @@ typedef void (*intel_point_func) (struct intel_context *, intelVertex *); #define INTEL_FALLBACK_USER 0x10 #define INTEL_FALLBACK_RENDERMODE 0x20 #define INTEL_FALLBACK_TEXTURE 0x40 +#define INTEL_FALLBACK_DRIVER 0x1000 /**< first for drivers */ +/*@}*/ -extern void intelFallback(struct intel_context *intel, GLuint bit, +extern void intelFallback(struct intel_context *intel, GLbitfield bit, GLboolean mode); #define FALLBACK( intel, bit, mode ) intelFallback( intel, bit, mode ) @@ -111,8 +117,6 @@ struct intel_context struct intel_region * depth_region, GLuint num_regions); - GLuint (*flush_cmd) (void); - void (*reduced_primitive_state) (struct intel_context * intel, GLenum rprim); @@ -170,13 +174,17 @@ struct intel_context struct dri_metaops meta; - GLint refcount; - GLuint Fallback; + GLbitfield Fallback; /**< mask of INTEL_FALLBACK_x bits */ GLuint NewGLState; dri_bufmgr *bufmgr; unsigned int maxBatchSize; + /** + * Generation number of the hardware: 2 is 8xx, 3 is 9xx pre-965, 4 is 965. + */ + int gen; + struct intel_region *front_region; struct intel_region *back_region; struct intel_region *depth_region; @@ -190,7 +198,6 @@ struct intel_context struct intel_batchbuffer *batch; drm_intel_bo *first_post_swapbuffers_batch; GLboolean no_batch_wrap; - unsigned batch_id; struct { @@ -254,9 +261,6 @@ struct intel_context intel_line_func draw_line; intel_tri_func draw_tri; - /* These refer to the current drawing buffer: - */ - struct gl_texture_object *frame_buffer_texobj; /** * Set to true if a single constant cliprect should be used in the * batchbuffer. Otherwise, cliprects must be calculated at batchbuffer @@ -296,7 +300,6 @@ struct intel_context GLboolean use_texture_tiling; GLboolean use_early_z; - drm_clip_rect_t fboRect; /**< cliprect for FBO rendering */ int perf_boxes; @@ -351,6 +354,19 @@ extern char *__progname; #define ALIGN(value, alignment) ((value + alignment - 1) & ~(alignment - 1)) #define IS_POWER_OF_TWO(val) (((val) & (val - 1)) == 0) +static inline uint32_t +U_FIXED(float value, uint32_t frac_bits) +{ + value *= (1 << frac_bits); + return value < 0 ? 0 : value; +} + +static inline uint32_t +S_FIXED(float value, uint32_t frac_bits) +{ + return value * (1 << frac_bits); +} + #define INTEL_FIREVERTICES(intel) \ do { \ if ((intel)->prim.flush) \ @@ -572,4 +588,25 @@ is_power_of_two(uint32_t value) return (value & (value - 1)) == 0; } +static inline void +intel_bo_map_gtt_preferred(struct intel_context *intel, + drm_intel_bo *bo, + GLboolean write) +{ + if (intel->intelScreen->kernel_exec_fencing) + drm_intel_gem_bo_map_gtt(bo); + else + drm_intel_bo_map(bo, write); +} + +static inline void +intel_bo_unmap_gtt_preferred(struct intel_context *intel, + drm_intel_bo *bo) +{ + if (intel->intelScreen->kernel_exec_fencing) + drm_intel_gem_bo_unmap_gtt(bo); + else + drm_intel_bo_unmap(bo); +} + #endif diff --git a/shared/intel_depthtmp.h b/shared/intel_depthtmp.h index 16d7708..a9c75d4 100644 --- a/shared/intel_depthtmp.h +++ b/shared/intel_depthtmp.h @@ -31,6 +31,16 @@ */ #define VALUE_TYPE INTEL_VALUE_TYPE +#define WRITE_DEPTH(_x, _y, d) \ + (*(INTEL_VALUE_TYPE *)(irb->region->buffer->virtual + \ + NO_TILE(_x, _y)) = d) +#define READ_DEPTH(d, _x, _y) \ + d = *(INTEL_VALUE_TYPE *)(irb->region->buffer->virtual + \ + NO_TILE(_x, _y)) +#define TAG(x) INTEL_TAG(intel_gttmap_##x) +#include "depthtmp.h" + +#define VALUE_TYPE INTEL_VALUE_TYPE #define WRITE_DEPTH(_x, _y, d) INTEL_WRITE_DEPTH(NO_TILE(_x, _y), d) #define READ_DEPTH(d, _x, _y) d = INTEL_READ_DEPTH(NO_TILE(_x, _y)) #define TAG(x) INTEL_TAG(intel##x) diff --git a/shared/intel_extensions.c b/shared/intel_extensions.c index 2e61c55..48cdae5 100644 --- a/shared/intel_extensions.c +++ b/shared/intel_extensions.c @@ -28,9 +28,11 @@ #include "intel_chipset.h" #include "intel_context.h" #include "intel_extensions.h" +#include "utils.h" #define need_GL_ARB_copy_buffer +#define need_GL_ARB_draw_elements_base_vertex #define need_GL_ARB_framebuffer_object #define need_GL_ARB_map_buffer_range #define need_GL_ARB_occlusion_query @@ -62,7 +64,7 @@ #define need_GL_VERSION_2_0 #define need_GL_VERSION_2_1 -#include "extension_helper.h" +#include "main/remap_helper.h" /** @@ -73,11 +75,15 @@ */ static const struct dri_extension card_extensions[] = { { "GL_ARB_copy_buffer", GL_ARB_copy_buffer_functions }, + { "GL_ARB_draw_elements_base_vertex", GL_ARB_draw_elements_base_vertex_functions }, { "GL_ARB_half_float_pixel", NULL }, { "GL_ARB_map_buffer_range", GL_ARB_map_buffer_range_functions }, { "GL_ARB_multitexture", NULL }, { "GL_ARB_point_parameters", GL_ARB_point_parameters_functions }, { "GL_ARB_point_sprite", NULL }, + { "GL_ARB_shader_objects", GL_ARB_shader_objects_functions }, + { "GL_ARB_shading_language_100", GL_VERSION_2_0_functions }, + { "GL_ARB_shading_language_120", GL_VERSION_2_1_functions }, { "GL_ARB_sync", GL_ARB_sync_functions }, { "GL_ARB_texture_border_clamp", NULL }, { "GL_ARB_texture_cube_map", NULL }, @@ -89,6 +95,7 @@ static const struct dri_extension card_extensions[] = { { "GL_ARB_texture_rectangle", NULL }, { "GL_ARB_vertex_array_object", GL_ARB_vertex_array_object_functions}, { "GL_ARB_vertex_program", GL_ARB_vertex_program_functions }, + { "GL_ARB_vertex_shader", GL_ARB_vertex_shader_functions }, { "GL_ARB_window_pos", GL_ARB_window_pos_functions }, { "GL_EXT_blend_color", GL_EXT_blend_color_functions }, { "GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions }, @@ -114,7 +121,6 @@ static const struct dri_extension card_extensions[] = { { "GL_MESA_pack_invert", NULL }, { "GL_MESA_ycbcr_texture", NULL }, { "GL_NV_blend_square", NULL }, - { "GL_NV_point_sprite", GL_NV_point_sprite_functions }, { "GL_NV_vertex_program", GL_NV_vertex_program_functions }, { "GL_NV_vertex_program1_1", NULL }, { "GL_SGIS_generate_mipmap", NULL }, @@ -139,6 +145,7 @@ static const struct dri_extension i915_extensions[] = { /** i965-only extensions */ static const struct dri_extension brw_extensions[] = { + { "GL_ARB_depth_clamp", NULL }, { "GL_ARB_depth_texture", NULL }, { "GL_ARB_fragment_program", NULL }, { "GL_ARB_fragment_program_shadow", NULL }, @@ -147,13 +154,9 @@ static const struct dri_extension brw_extensions[] = { { "GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions }, { "GL_ARB_point_sprite", NULL }, { "GL_ARB_seamless_cube_map", NULL }, - { "GL_ARB_shader_objects", GL_ARB_shader_objects_functions }, - { "GL_ARB_shading_language_100", GL_VERSION_2_0_functions }, - { "GL_ARB_shading_language_120", GL_VERSION_2_1_functions }, { "GL_ARB_shadow", NULL }, { "GL_MESA_texture_signed_rgba", NULL }, { "GL_ARB_texture_non_power_of_two", NULL }, - { "GL_ARB_vertex_shader", GL_ARB_vertex_shader_functions }, { "GL_EXT_shadow_funcs", NULL }, { "GL_EXT_stencil_two_side", GL_EXT_stencil_two_side_functions }, { "GL_EXT_texture_sRGB", NULL }, @@ -168,6 +171,7 @@ static const struct dri_extension brw_extensions[] = { static const struct dri_extension arb_oq_extensions[] = { + { "GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions }, { NULL, NULL } }; @@ -179,29 +183,38 @@ static const struct dri_extension ttm_extensions[] = { { NULL, NULL } }; +static const struct dri_extension fragment_shader_extensions[] = { + { "GL_ARB_fragment_shader", NULL }, + { NULL, NULL } +}; /** * Initializes potential list of extensions if ctx == NULL, or actually enables * extensions for a context. */ void -intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging) +intelInitExtensions(GLcontext *ctx) { - struct intel_context *intel = ctx?intel_context(ctx):NULL; + struct intel_context *intel = intel_context(ctx); /* Disable imaging extension until convolution is working in teximage paths. */ - enable_imaging = GL_FALSE; + driInitExtensions(ctx, card_extensions, GL_FALSE); - driInitExtensions(ctx, card_extensions, enable_imaging); - - if (intel == NULL || intel->ttm) + if (intel->ttm) driInitExtensions(ctx, ttm_extensions, GL_FALSE); - if (intel == NULL || IS_965(intel->intelScreen->deviceID)) + if (IS_965(intel->intelScreen->deviceID)) driInitExtensions(ctx, brw_extensions, GL_FALSE); - if (intel == NULL || IS_915(intel->intelScreen->deviceID) - || IS_945(intel->intelScreen->deviceID)) + if (IS_915(intel->intelScreen->deviceID) + || IS_945(intel->intelScreen->deviceID)) { driInitExtensions(ctx, i915_extensions, GL_FALSE); + + if (driQueryOptionb(&intel->optionCache, "fragment_shader")) + driInitExtensions(ctx, fragment_shader_extensions, GL_FALSE); + + if (driQueryOptionb(&intel->optionCache, "stub_occlusion_query")) + driInitExtensions(ctx, arb_oq_extensions, GL_FALSE); + } } diff --git a/shared/intel_extensions.h b/shared/intel_extensions.h index 97147ec..1d1c97a 100644 --- a/shared/intel_extensions.h +++ b/shared/intel_extensions.h @@ -30,7 +30,7 @@ extern void -intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging); +intelInitExtensions(GLcontext *ctx); #endif diff --git a/shared/intel_fbo.c b/shared/intel_fbo.c index 804c034..608f75b 100644 --- a/shared/intel_fbo.c +++ b/shared/intel_fbo.c @@ -33,11 +33,11 @@ #include "main/framebuffer.h" #include "main/renderbuffer.h" #include "main/context.h" -#include "main/texformat.h" #include "main/texrender.h" #include "drivers/common/meta.h" #include "intel_context.h" +#include "intel_batchbuffer.h" #include "intel_buffers.h" #include "intel_fbo.h" #include "intel_mipmap_tree.h" @@ -106,8 +106,8 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, { struct intel_context *intel = intel_context(ctx); struct intel_renderbuffer *irb = intel_renderbuffer(rb); - GLboolean softwareBuffer = GL_FALSE; int cpp; + GLuint pitch; ASSERT(rb->Name != 0); @@ -115,27 +115,16 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, case GL_R3_G3_B2: case GL_RGB4: case GL_RGB5: - rb->_ActualFormat = GL_RGB5; + rb->Format = MESA_FORMAT_RGB565; rb->DataType = GL_UNSIGNED_BYTE; - rb->RedBits = 5; - rb->GreenBits = 6; - rb->BlueBits = 5; - irb->texformat = &_mesa_texformat_rgb565; - cpp = 2; break; case GL_RGB: case GL_RGB8: case GL_RGB10: case GL_RGB12: case GL_RGB16: - rb->_ActualFormat = GL_RGB8; + rb->Format = MESA_FORMAT_XRGB8888; rb->DataType = GL_UNSIGNED_BYTE; - rb->RedBits = 8; - rb->GreenBits = 8; - rb->BlueBits = 8; - rb->AlphaBits = 0; - irb->texformat = &_mesa_texformat_argb8888; /* XXX: Need xrgb8888 */ - cpp = 4; break; case GL_RGBA: case GL_RGBA2: @@ -145,14 +134,8 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, case GL_RGB10_A2: case GL_RGBA12: case GL_RGBA16: - rb->_ActualFormat = GL_RGBA8; + rb->Format = MESA_FORMAT_ARGB8888; rb->DataType = GL_UNSIGNED_BYTE; - rb->RedBits = 8; - rb->GreenBits = 8; - rb->BlueBits = 8; - rb->AlphaBits = 8; - irb->texformat = &_mesa_texformat_argb8888; - cpp = 4; break; case GL_STENCIL_INDEX: case GL_STENCIL_INDEX1_EXT: @@ -160,36 +143,23 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, case GL_STENCIL_INDEX8_EXT: case GL_STENCIL_INDEX16_EXT: /* alloc a depth+stencil buffer */ - rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; + rb->Format = MESA_FORMAT_S8_Z24; rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - rb->StencilBits = 8; - cpp = 4; - irb->texformat = &_mesa_texformat_s8_z24; break; case GL_DEPTH_COMPONENT16: - rb->_ActualFormat = GL_DEPTH_COMPONENT16; + rb->Format = MESA_FORMAT_Z16; rb->DataType = GL_UNSIGNED_SHORT; - rb->DepthBits = 16; - cpp = 2; - irb->texformat = &_mesa_texformat_z16; break; case GL_DEPTH_COMPONENT: case GL_DEPTH_COMPONENT24: case GL_DEPTH_COMPONENT32: - rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; + rb->Format = MESA_FORMAT_S8_Z24; rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - rb->DepthBits = 24; - cpp = 4; - irb->texformat = &_mesa_texformat_s8_z24; break; case GL_DEPTH_STENCIL_EXT: case GL_DEPTH24_STENCIL8_EXT: - rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; + rb->Format = MESA_FORMAT_S8_Z24; rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - rb->DepthBits = 24; - rb->StencilBits = 8; - cpp = 4; - irb->texformat = &_mesa_texformat_s8_z24; break; default: _mesa_problem(ctx, @@ -197,6 +167,9 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, return GL_FALSE; } + rb->_BaseFormat = _mesa_base_fbo_format(ctx, internalFormat); + cpp = _mesa_get_format_bytes(rb->Format); + intelFlush(ctx); /* free old region */ @@ -205,32 +178,25 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, } /* allocate new memory region/renderbuffer */ - if (softwareBuffer) { - return _mesa_soft_renderbuffer_storage(ctx, rb, internalFormat, - width, height); - } - else { - /* Choose a pitch to match hardware requirements: - */ - GLuint pitch = ((cpp * width + 63) & ~63) / cpp; - /* alloc hardware renderbuffer */ - DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width, - height, pitch); + /* Choose a pitch to match hardware requirements: + */ + pitch = ((cpp * width + 63) & ~63) / cpp; - irb->region = intel_region_alloc(intel, I915_TILING_NONE, - cpp, width, height, pitch, - GL_TRUE); - if (!irb->region) - return GL_FALSE; /* out of memory? */ + /* alloc hardware renderbuffer */ + DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width, height, pitch); - ASSERT(irb->region->buffer); + irb->region = intel_region_alloc(intel, I915_TILING_NONE, cpp, + width, height, pitch, GL_TRUE); + if (!irb->region) + return GL_FALSE; /* out of memory? */ - rb->Width = width; - rb->Height = height; + ASSERT(irb->region->buffer); - return GL_TRUE; - } + rb->Width = width; + rb->Height = height; + + return GL_TRUE; } @@ -246,7 +212,7 @@ intel_alloc_window_storage(GLcontext * ctx, struct gl_renderbuffer *rb, ASSERT(rb->Name == 0); rb->Width = width; rb->Height = height; - rb->_ActualFormat = internalFormat; + rb->InternalFormat = internalFormat; return GL_TRUE; } @@ -307,12 +273,11 @@ intel_renderbuffer_set_region(struct intel_renderbuffer *rb, * not a user-created renderbuffer. */ struct intel_renderbuffer * -intel_create_renderbuffer(GLenum intFormat) +intel_create_renderbuffer(gl_format format) { GET_CURRENT_CONTEXT(ctx); struct intel_renderbuffer *irb; - const GLuint name = 0; irb = CALLOC_STRUCT(intel_renderbuffer); if (!irb) { @@ -320,75 +285,43 @@ intel_create_renderbuffer(GLenum intFormat) return NULL; } - _mesa_init_renderbuffer(&irb->Base, name); + _mesa_init_renderbuffer(&irb->Base, 0); irb->Base.ClassID = INTEL_RB_CLASS; - switch (intFormat) { - case GL_RGB5: - irb->Base._ActualFormat = GL_RGB5; - irb->Base._BaseFormat = GL_RGBA; - irb->Base.RedBits = 5; - irb->Base.GreenBits = 6; - irb->Base.BlueBits = 5; + switch (format) { + case MESA_FORMAT_RGB565: + irb->Base._BaseFormat = GL_RGB; irb->Base.DataType = GL_UNSIGNED_BYTE; - irb->texformat = &_mesa_texformat_rgb565; break; - case GL_RGB8: - irb->Base._ActualFormat = GL_RGB8; + case MESA_FORMAT_XRGB8888: irb->Base._BaseFormat = GL_RGB; - irb->Base.RedBits = 8; - irb->Base.GreenBits = 8; - irb->Base.BlueBits = 8; - irb->Base.AlphaBits = 0; irb->Base.DataType = GL_UNSIGNED_BYTE; - irb->texformat = &_mesa_texformat_argb8888; /* XXX: Need xrgb8888 */ break; - case GL_RGBA8: - irb->Base._ActualFormat = GL_RGBA8; + case MESA_FORMAT_ARGB8888: irb->Base._BaseFormat = GL_RGBA; - irb->Base.RedBits = 8; - irb->Base.GreenBits = 8; - irb->Base.BlueBits = 8; - irb->Base.AlphaBits = 8; - irb->Base.DataType = GL_UNSIGNED_BYTE; - irb->texformat = &_mesa_texformat_argb8888; - break; - case GL_STENCIL_INDEX8_EXT: - irb->Base._ActualFormat = GL_STENCIL_INDEX8_EXT; - irb->Base._BaseFormat = GL_STENCIL_INDEX; - irb->Base.StencilBits = 8; irb->Base.DataType = GL_UNSIGNED_BYTE; - irb->texformat = &_mesa_texformat_s8_z24; break; - case GL_DEPTH_COMPONENT16: - irb->Base._ActualFormat = GL_DEPTH_COMPONENT16; + case MESA_FORMAT_Z16: irb->Base._BaseFormat = GL_DEPTH_COMPONENT; - irb->Base.DepthBits = 16; irb->Base.DataType = GL_UNSIGNED_SHORT; - irb->texformat = &_mesa_texformat_z16; break; - case GL_DEPTH_COMPONENT24: - irb->Base._ActualFormat = GL_DEPTH24_STENCIL8_EXT; + case MESA_FORMAT_X8_Z24: irb->Base._BaseFormat = GL_DEPTH_COMPONENT; - irb->Base.DepthBits = 24; irb->Base.DataType = GL_UNSIGNED_INT; - irb->texformat = &_mesa_texformat_s8_z24; break; - case GL_DEPTH24_STENCIL8_EXT: - irb->Base._ActualFormat = GL_DEPTH24_STENCIL8_EXT; - irb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; - irb->Base.DepthBits = 24; - irb->Base.StencilBits = 8; + case MESA_FORMAT_S8_Z24: + irb->Base._BaseFormat = GL_DEPTH_STENCIL; irb->Base.DataType = GL_UNSIGNED_INT_24_8_EXT; - irb->texformat = &_mesa_texformat_s8_z24; break; default: _mesa_problem(NULL, "Unexpected intFormat in intel_create_renderbuffer"); + _mesa_free(irb); return NULL; } - irb->Base.InternalFormat = intFormat; + irb->Base.Format = format; + irb->Base.InternalFormat = irb->Base._BaseFormat; /* intel-specific methods */ irb->Base.Delete = intel_delete_renderbuffer; @@ -465,59 +398,49 @@ static GLboolean intel_update_wrapper(GLcontext *ctx, struct intel_renderbuffer *irb, struct gl_texture_image *texImage) { - irb->texformat = texImage->TexFormat; + gl_format texFormat; - if (texImage->TexFormat == &_mesa_texformat_argb8888) { - irb->Base._ActualFormat = GL_RGBA8; - irb->Base._BaseFormat = GL_RGBA; + if (texImage->TexFormat == MESA_FORMAT_ARGB8888) { irb->Base.DataType = GL_UNSIGNED_BYTE; DBG("Render to RGBA8 texture OK\n"); } - else if (texImage->TexFormat == &_mesa_texformat_rgb565) { - irb->Base._ActualFormat = GL_RGB5; - irb->Base._BaseFormat = GL_RGB; + else if (texImage->TexFormat == MESA_FORMAT_XRGB8888) { + irb->Base.DataType = GL_UNSIGNED_BYTE; + DBG("Render to XGBA8 texture OK\n"); + } + else if (texImage->TexFormat == MESA_FORMAT_RGB565) { irb->Base.DataType = GL_UNSIGNED_BYTE; DBG("Render to RGB5 texture OK\n"); } - else if (texImage->TexFormat == &_mesa_texformat_argb1555) { - irb->Base._ActualFormat = GL_RGB5_A1; - irb->Base._BaseFormat = GL_RGBA; + else if (texImage->TexFormat == MESA_FORMAT_ARGB1555) { irb->Base.DataType = GL_UNSIGNED_BYTE; DBG("Render to ARGB1555 texture OK\n"); } - else if (texImage->TexFormat == &_mesa_texformat_argb4444) { - irb->Base._ActualFormat = GL_RGBA4; - irb->Base._BaseFormat = GL_RGBA; + else if (texImage->TexFormat == MESA_FORMAT_ARGB4444) { irb->Base.DataType = GL_UNSIGNED_BYTE; DBG("Render to ARGB4444 texture OK\n"); } - else if (texImage->TexFormat == &_mesa_texformat_z16) { - irb->Base._ActualFormat = GL_DEPTH_COMPONENT16; - irb->Base._BaseFormat = GL_DEPTH_COMPONENT; + else if (texImage->TexFormat == MESA_FORMAT_Z16) { irb->Base.DataType = GL_UNSIGNED_SHORT; DBG("Render to DEPTH16 texture OK\n"); } - else if (texImage->TexFormat == &_mesa_texformat_s8_z24) { - irb->Base._ActualFormat = GL_DEPTH24_STENCIL8_EXT; - irb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; + else if (texImage->TexFormat == MESA_FORMAT_S8_Z24) { irb->Base.DataType = GL_UNSIGNED_INT_24_8_EXT; DBG("Render to DEPTH_STENCIL texture OK\n"); } else { - DBG("Render to texture BAD FORMAT %d\n", - texImage->TexFormat->MesaFormat); + DBG("Render to texture BAD FORMAT %d\n", texImage->TexFormat); return GL_FALSE; } - irb->Base.InternalFormat = irb->Base._ActualFormat; + irb->Base.Format = texImage->TexFormat; + + texFormat = texImage->TexFormat; + + irb->Base.InternalFormat = texImage->InternalFormat; + irb->Base._BaseFormat = _mesa_base_fbo_format(ctx, irb->Base.InternalFormat); irb->Base.Width = texImage->Width; irb->Base.Height = texImage->Height; - irb->Base.RedBits = texImage->TexFormat->RedBits; - irb->Base.GreenBits = texImage->TexFormat->GreenBits; - irb->Base.BlueBits = texImage->TexFormat->BlueBits; - irb->Base.AlphaBits = texImage->TexFormat->AlphaBits; - irb->Base.DepthBits = texImage->TexFormat->DepthBits; - irb->Base.StencilBits = texImage->TexFormat->StencilBits; irb->Base.Delete = intel_delete_renderbuffer; irb->Base.AllocStorage = intel_nop_alloc_storage; @@ -571,7 +494,7 @@ intel_render_texture(GLcontext * ctx, = att->Texture->Image[att->CubeMapFace][att->TextureLevel]; struct intel_renderbuffer *irb = intel_renderbuffer(att->Renderbuffer); struct intel_texture_image *intel_image; - GLuint imageOffset; + GLuint dst_x, dst_y; (void) fb; @@ -618,18 +541,17 @@ intel_render_texture(GLcontext * ctx, } /* compute offset of the particular 2D image within the texture region */ - imageOffset = intel_miptree_image_offset(intel_image->mt, - att->CubeMapFace, - att->TextureLevel); - - if (att->Texture->Target == GL_TEXTURE_3D) { - const GLuint *offsets = intel_miptree_depth_offsets(intel_image->mt, - att->TextureLevel); - imageOffset += offsets[att->Zoffset]; - } - - /* store that offset in the region */ - intel_image->mt->region->draw_offset = imageOffset; + intel_miptree_get_image_offset(intel_image->mt, + att->TextureLevel, + att->CubeMapFace, + att->Zoffset, + &dst_x, &dst_y); + + intel_image->mt->region->draw_offset = (dst_y * intel_image->mt->pitch + + dst_x) * intel_image->mt->cpp; + intel_image->mt->region->draw_x = dst_x; + intel_image->mt->region->draw_y = dst_y; + intel_image->used_as_render_target = GL_TRUE; /* update drawing region, etc */ intel_draw_buffer(ctx, fb); @@ -643,19 +565,23 @@ static void intel_finish_render_texture(GLcontext * ctx, struct gl_renderbuffer_attachment *att) { - /* no-op - * Previously we released the renderbuffer's intel_region but - * that's not necessary and actually caused problems when trying - * to do a glRead/CopyPixels from the renderbuffer later. - * The region will be released later if the texture is replaced - * or the renderbuffer deleted. - * - * The intention of this driver hook is more of a "done rendering - * to texture, please re-twiddle/etc if necessary". + struct intel_context *intel = intel_context(ctx); + struct gl_texture_object *tex_obj = att->Texture; + struct gl_texture_image *image = + tex_obj->Image[att->CubeMapFace][att->TextureLevel]; + struct intel_texture_image *intel_image = intel_texture_image(image); + + /* Flag that this image may now be validated into the object's miptree. */ + intel_image->used_as_render_target = GL_FALSE; + + /* Since we've (probably) rendered to the texture and will (likely) use + * it in the texture domain later on in this batchbuffer, flush the + * batch. Once again, we wish for a domain tracker in libdrm to cover + * usage inside of a batchbuffer like GEM does in the kernel. */ + intel_batchbuffer_emit_mi_flush(intel->batch); } - /** * Do additional "completeness" testing of a framebuffer object. */ @@ -687,8 +613,9 @@ intel_validate_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb) continue; } - switch (irb->texformat->MesaFormat) { + switch (irb->Base.Format) { case MESA_FORMAT_ARGB8888: + case MESA_FORMAT_XRGB8888: case MESA_FORMAT_RGB565: case MESA_FORMAT_ARGB1555: case MESA_FORMAT_ARGB4444: @@ -715,5 +642,5 @@ intel_fbo_init(struct intel_context *intel) intel->ctx.Driver.FinishRenderTexture = intel_finish_render_texture; intel->ctx.Driver.ResizeBuffers = intel_resize_buffers; intel->ctx.Driver.ValidateFramebuffer = intel_validate_framebuffer; - intel->ctx.Driver.BlitFramebuffer = _mesa_meta_blit_framebuffer; + intel->ctx.Driver.BlitFramebuffer = _mesa_meta_BlitFramebuffer; } diff --git a/shared/intel_fbo.h b/shared/intel_fbo.h index f0665af..fa43077 100644 --- a/shared/intel_fbo.h +++ b/shared/intel_fbo.h @@ -28,6 +28,7 @@ #ifndef INTEL_FBO_H #define INTEL_FBO_H +#include "main/formats.h" #include "intel_screen.h" struct intel_context; @@ -61,8 +62,6 @@ struct intel_renderbuffer struct gl_renderbuffer Base; struct intel_region *region; - const struct gl_texture_format *texformat; - GLuint vbl_pending; /**< vblank sequence number of pending flip */ uint8_t *span_cache; @@ -114,7 +113,7 @@ intel_renderbuffer_set_region(struct intel_renderbuffer *irb, extern struct intel_renderbuffer * -intel_create_renderbuffer(GLenum intFormat); +intel_create_renderbuffer(gl_format format); extern void diff --git a/shared/intel_generatemipmap.c b/shared/intel_generatemipmap.c deleted file mode 100644 index 12059e1..0000000 --- a/shared/intel_generatemipmap.c +++ /dev/null @@ -1,304 +0,0 @@ -/* - * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. - * Copyright © 2009 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS - * IN THE SOFTWARE. - * - * Authors: - * Eric Anholt <eric@anholt.net> - * - */ - -#include "main/glheader.h" -#include "main/enums.h" -#include "main/image.h" -#include "main/mtypes.h" -#include "main/macros.h" -#include "main/bufferobj.h" -#include "main/teximage.h" -#include "main/texenv.h" -#include "main/texobj.h" -#include "main/texstate.h" -#include "main/texparam.h" -#include "main/varray.h" -#include "main/attrib.h" -#include "main/enable.h" -#include "main/buffers.h" -#include "main/fbobject.h" -#include "main/framebuffer.h" -#include "main/renderbuffer.h" -#include "main/depth.h" -#include "main/hash.h" -#include "main/mipmap.h" -#include "main/blend.h" -#include "glapi/dispatch.h" -#include "swrast/swrast.h" - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_batchbuffer.h" -#include "intel_pixel.h" -#include "intel_tex.h" -#include "intel_mipmap_tree.h" - -static const char *intel_fp_tex2d = - "!!ARBfp1.0\n" - "TEX result.color, fragment.texcoord[0], texture[0], 2D;\n" - "END\n"; - -static GLboolean -intel_generate_mipmap_level(GLcontext *ctx, GLuint tex_name, - int level, int width, int height) -{ - struct intel_context *intel = intel_context(ctx); - GLfloat vertices[4][2]; - GLint status; - - /* Set to source from the previous level */ - _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_BASE_LEVEL, level - 1); - _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, level - 1); - - /* Set to draw into the current level */ - _mesa_FramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT, - GL_COLOR_ATTACHMENT0_EXT, - GL_TEXTURE_2D, - tex_name, - level); - /* Choose to render to the color attachment. */ - _mesa_DrawBuffer(GL_COLOR_ATTACHMENT0_EXT); - - status = _mesa_CheckFramebufferStatusEXT (GL_FRAMEBUFFER_EXT); - if (status != GL_FRAMEBUFFER_COMPLETE_EXT) - return GL_FALSE; - - meta_set_passthrough_transform(&intel->meta); - - /* XXX: Doing it right would involve setting up the transformation to do - * 0-1 mapping or something, and not changing the vertex data. - */ - vertices[0][0] = 0; - vertices[0][1] = 0; - vertices[1][0] = width; - vertices[1][1] = 0; - vertices[2][0] = width; - vertices[2][1] = height; - vertices[3][0] = 0; - vertices[3][1] = height; - - _mesa_VertexPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &vertices); - _mesa_Enable(GL_VERTEX_ARRAY); - meta_set_default_texrect(&intel->meta); - - _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4); - - meta_restore_texcoords(&intel->meta); - meta_restore_transform(&intel->meta); - - return GL_TRUE; -} - -static GLboolean -intel_generate_mipmap_2d(GLcontext *ctx, - GLenum target, - struct gl_texture_object *texObj) -{ - struct intel_context *intel = intel_context(ctx); - GLint old_active_texture; - int level, max_levels, start_level, end_level; - GLuint fb_name; - GLboolean success = GL_FALSE; - struct gl_framebuffer *saved_fbo = NULL; - struct gl_buffer_object *saved_array_buffer = NULL; - struct gl_buffer_object *saved_element_buffer = NULL; - - _mesa_PushAttrib(GL_ENABLE_BIT | GL_TEXTURE_BIT | - GL_CURRENT_BIT | GL_COLOR_BUFFER_BIT | - GL_DEPTH_BUFFER_BIT); - _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT); - old_active_texture = ctx->Texture.CurrentUnit; - _mesa_reference_framebuffer(&saved_fbo, ctx->DrawBuffer); - - /* use default array/index buffers */ - _mesa_reference_buffer_object(ctx, &saved_array_buffer, - ctx->Array.ArrayBufferObj); - _mesa_reference_buffer_object(ctx, &ctx->Array.ArrayBufferObj, - ctx->Shared->NullBufferObj); - _mesa_reference_buffer_object(ctx, &saved_element_buffer, - ctx->Array.ElementArrayBufferObj); - _mesa_reference_buffer_object(ctx, &ctx->Array.ElementArrayBufferObj, - ctx->Shared->NullBufferObj); - - _mesa_Disable(GL_POLYGON_STIPPLE); - _mesa_Disable(GL_DEPTH_TEST); - _mesa_Disable(GL_STENCIL_TEST); - _mesa_ColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE); - _mesa_DepthMask(GL_FALSE); - - /* Bind the given texture to GL_TEXTURE_2D with linear filtering for our - * minification. - */ - _mesa_ActiveTextureARB(GL_TEXTURE0_ARB); - _mesa_Enable(GL_TEXTURE_2D); - _mesa_BindTexture(GL_TEXTURE_2D, texObj->Name); - _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, - GL_LINEAR_MIPMAP_NEAREST); - _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR); - - /* Bind the new renderbuffer to the color attachment point. */ - _mesa_GenFramebuffersEXT(1, &fb_name); - _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, fb_name); - - meta_set_fragment_program(&intel->meta, &intel->meta.tex2d_fp, - intel_fp_tex2d); - meta_set_passthrough_vertex_program(&intel->meta); - - max_levels = _mesa_max_texture_levels(ctx, texObj->Target); - start_level = texObj->BaseLevel; - end_level = texObj->MaxLevel; - - /* Loop generating level+1 from level. */ - for (level = start_level; level < end_level && level < max_levels - 1; level++) { - const struct gl_texture_image *srcImage; - int width, height; - - srcImage = _mesa_select_tex_image(ctx, texObj, target, level); - if (srcImage->Border != 0) - goto fail; - - width = srcImage->Width / 2; - if (width < 1) - width = 1; - height = srcImage->Height / 2; - if (height < 1) - height = 1; - - if (width == srcImage->Width && - height == srcImage->Height) { - /* Neither _mesa_max_texture_levels nor texObj->MaxLevel are the - * maximum texture level for the object, so break out when we've gone - * over the edge. - */ - break; - } - - /* Make sure that there's space allocated for the target level. - * We could skip this if there's already space allocated and save some - * time. - */ - _mesa_TexImage2D(GL_TEXTURE_2D, level + 1, srcImage->InternalFormat, - width, height, 0, - GL_RGBA, GL_UNSIGNED_INT, NULL); - - if (!intel_generate_mipmap_level(ctx, texObj->Name, level + 1, - width, height)) - goto fail; - } - - success = GL_TRUE; - -fail: - meta_restore_fragment_program(&intel->meta); - meta_restore_vertex_program(&intel->meta); - - /* restore array/index buffers */ - _mesa_reference_buffer_object(ctx, &ctx->Array.ArrayBufferObj, - saved_array_buffer); - _mesa_reference_buffer_object(ctx, &saved_array_buffer, NULL); - _mesa_reference_buffer_object(ctx, &ctx->Array.ElementArrayBufferObj, - saved_element_buffer); - _mesa_reference_buffer_object(ctx, &saved_element_buffer, NULL); - - - _mesa_DeleteFramebuffersEXT(1, &fb_name); - _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture); - if (saved_fbo) - _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, saved_fbo->Name); - _mesa_reference_framebuffer(&saved_fbo, NULL); - _mesa_PopClientAttrib(); - _mesa_PopAttrib(); - - return success; -} - - -/** - * Generate new mipmap data from BASE+1 to BASE+p (the minimally-sized mipmap - * level). - * - * The texture object's miptree must be mapped. - * - * It would be really nice if this was just called by Mesa whenever mipmaps - * needed to be regenerated, rather than us having to remember to do so in - * each texture image modification path. - * - * This function should also include an accelerated path. - */ -void -intel_generate_mipmap(GLcontext *ctx, GLenum target, - struct gl_texture_object *texObj) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_texture_object *intelObj = intel_texture_object(texObj); - GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; - int face, i; - - /* HW path */ - if (target == GL_TEXTURE_2D && - ctx->Extensions.EXT_framebuffer_object && - ctx->Extensions.ARB_fragment_program && - ctx->Extensions.ARB_vertex_program) { - GLboolean success; - - /* We'll be accessing this texture using GL entrypoints, which should - * be resilient against other access to this texture. - */ - _mesa_unlock_texture(ctx, texObj); - success = intel_generate_mipmap_2d(ctx, target, texObj); - _mesa_lock_texture(ctx, texObj); - - if (success) - return; - } - - /* SW path */ - intel_tex_map_level_images(intel, intelObj, texObj->BaseLevel); - _mesa_generate_mipmap(ctx, target, texObj); - intel_tex_unmap_level_images(intel, intelObj, texObj->BaseLevel); - - /* Update the level information in our private data in the new images, since - * it didn't get set as part of a normal TexImage path. - */ - for (face = 0; face < nr_faces; face++) { - for (i = texObj->BaseLevel + 1; i < texObj->MaxLevel; i++) { - struct intel_texture_image *intelImage; - - intelImage = intel_texture_image(texObj->Image[face][i]); - if (intelImage == NULL) - break; - - intelImage->level = i; - intelImage->face = face; - /* Unreference the miptree to signal that the new Data is a bare - * pointer from mesa. - */ - intel_miptree_release(intel, &intelImage->mt); - } - } -} diff --git a/shared/intel_mipmap_tree.c b/shared/intel_mipmap_tree.c index c985da5..abb3024 100644 --- a/shared/intel_mipmap_tree.c +++ b/shared/intel_mipmap_tree.c @@ -28,11 +28,16 @@ #include "intel_context.h" #include "intel_mipmap_tree.h" #include "intel_regions.h" +#include "intel_tex_layout.h" #include "intel_chipset.h" +#ifndef I915 +#include "brw_state.h" +#endif #include "main/enums.h" #define FILE_DEBUG_FLAG DEBUG_MIPTREE + static GLenum target_to_target(GLenum target) { @@ -49,6 +54,7 @@ target_to_target(GLenum target) } } + static struct intel_mipmap_tree * intel_miptree_create_internal(struct intel_context *intel, GLenum target, @@ -98,6 +104,7 @@ intel_miptree_create_internal(struct intel_context *intel, return mt; } + struct intel_mipmap_tree * intel_miptree_create(struct intel_context *intel, GLenum target, @@ -115,7 +122,7 @@ intel_miptree_create(struct intel_context *intel, if (intel->use_texture_tiling && compress_byte == 0 && intel->intelScreen->kernel_exec_fencing) { - if (IS_965(intel->intelScreen->deviceID) && + if (intel->gen >= 4 && (base_format == GL_DEPTH_COMPONENT || base_format == GL_DEPTH_STENCIL_EXT)) tiling = I915_TILING_Y; @@ -131,8 +138,10 @@ intel_miptree_create(struct intel_context *intel, /* * pitch == 0 || height == 0 indicates the null texture */ - if (!mt || !mt->pitch || !mt->total_height) + if (!mt || !mt->pitch || !mt->total_height) { + free(mt); return NULL; + } mt->region = intel_region_alloc(intel, tiling, @@ -150,6 +159,7 @@ intel_miptree_create(struct intel_context *intel, return mt; } + struct intel_mipmap_tree * intel_miptree_create_for_region(struct intel_context *intel, GLenum target, @@ -187,7 +197,8 @@ intel_miptree_create_for_region(struct intel_context *intel, intel_region_reference(&mt->region, region); return mt; - } +} + /** * intel_miptree_pitch_align: @@ -201,7 +212,6 @@ intel_miptree_create_for_region(struct intel_context *intel, * Given @pitch, compute a larger value which accounts for * any necessary alignment required by the device */ - int intel_miptree_pitch_align (struct intel_context *intel, struct intel_mipmap_tree *mt, uint32_t tiling, @@ -247,6 +257,7 @@ int intel_miptree_pitch_align (struct intel_context *intel, return pitch; } + void intel_miptree_reference(struct intel_mipmap_tree **dst, struct intel_mipmap_tree *src) @@ -256,6 +267,7 @@ intel_miptree_reference(struct intel_mipmap_tree **dst, DBG("%s %p refcount now %d\n", __FUNCTION__, src, src->refcount); } + void intel_miptree_release(struct intel_context *intel, struct intel_mipmap_tree **mt) @@ -269,11 +281,25 @@ intel_miptree_release(struct intel_context *intel, DBG("%s deleting %p\n", __FUNCTION__, *mt); +#ifndef I915 + /* Free up cached binding tables holding a reference on our buffer, to + * avoid excessive memory consumption. + * + * This isn't as aggressive as we could be, as we'd like to do + * it from any time we free the last ref on a region. But intel_region.c + * is context-agnostic. Perhaps our constant state cache should be, as + * well. + */ + brw_state_cache_bo_delete(&brw_context(&intel->ctx)->surface_cache, + (*mt)->region->buffer); +#endif + intel_region_release(&((*mt)->region)); - for (i = 0; i < MAX_TEXTURE_LEVELS; i++) - if ((*mt)->level[i].image_offset) - free((*mt)->level[i].image_offset); + for (i = 0; i < MAX_TEXTURE_LEVELS; i++) { + free((*mt)->level[i].x_offset); + free((*mt)->level[i].y_offset); + } free(*mt); } @@ -281,33 +307,31 @@ intel_miptree_release(struct intel_context *intel, } - - -/* Can the image be pulled into a unified mipmap tree. This mirrors +/** + * Can the image be pulled into a unified mipmap tree? This mirrors * the completeness test in a lot of ways. * * Not sure whether I want to pass gl_texture_image here. */ GLboolean intel_miptree_match_image(struct intel_mipmap_tree *mt, - struct gl_texture_image *image, - GLuint face, GLuint level) + struct gl_texture_image *image) { - /* Images with borders are never pulled into mipmap trees. - */ - if (image->Border || - ((image->_BaseFormat == GL_DEPTH_COMPONENT) && - ((image->TexObject->WrapS == GL_CLAMP_TO_BORDER) || - (image->TexObject->WrapT == GL_CLAMP_TO_BORDER)))) + GLboolean isCompressed = _mesa_is_format_compressed(image->TexFormat); + struct intel_texture_image *intelImage = intel_texture_image(image); + GLuint level = intelImage->level; + + /* Images with borders are never pulled into mipmap trees. */ + if (image->Border) return GL_FALSE; if (image->InternalFormat != mt->internal_format || - image->IsCompressed != mt->compressed) + isCompressed != mt->compressed) return GL_FALSE; - if (!image->IsCompressed && + if (!isCompressed && !mt->compressed && - image->TexFormat->TexelBytes != mt->cpp) + _mesa_get_format_bytes(image->TexFormat) != mt->cpp) return GL_FALSE; /* Test image dimensions against the base level image adjusted for @@ -334,82 +358,59 @@ intel_miptree_set_level_info(struct intel_mipmap_tree *mt, mt->level[level].height = h; mt->level[level].depth = d; mt->level[level].level_offset = (x + y * mt->pitch) * mt->cpp; + mt->level[level].level_x = x; + mt->level[level].level_y = y; mt->level[level].nr_images = nr_images; DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, level, w, h, d, x, y, mt->level[level].level_offset); - /* Not sure when this would happen, but anyway: - */ - if (mt->level[level].image_offset) { - free(mt->level[level].image_offset); - mt->level[level].image_offset = NULL; - } - assert(nr_images); + assert(!mt->level[level].x_offset); - mt->level[level].image_offset = malloc(nr_images * sizeof(GLuint)); - mt->level[level].image_offset[0] = 0; + mt->level[level].x_offset = malloc(nr_images * sizeof(GLuint)); + mt->level[level].x_offset[0] = mt->level[level].level_x; + mt->level[level].y_offset = malloc(nr_images * sizeof(GLuint)); + mt->level[level].y_offset[0] = mt->level[level].level_y; } void -intel_miptree_set_image_offset_ex(struct intel_mipmap_tree *mt, - GLuint level, GLuint img, - GLuint x, GLuint y, - GLuint offset) -{ - if (img == 0 && level == 0) - assert(x == 0 && y == 0); - - assert(img < mt->level[level].nr_images); - - mt->level[level].image_offset[img] = (x + y * mt->pitch) * mt->cpp + offset; - - DBG("%s level %d img %d pos %d,%d image_offset %x\n", - __FUNCTION__, level, img, x, y, mt->level[level].image_offset[img]); -} - -void intel_miptree_set_image_offset(struct intel_mipmap_tree *mt, GLuint level, GLuint img, GLuint x, GLuint y) { - intel_miptree_set_image_offset_ex(mt, level, img, x, y, 0); -} + if (img == 0 && level == 0) + assert(x == 0 && y == 0); + assert(img < mt->level[level].nr_images); -/* Although we use the image_offset[] array to store relative offsets - * to cube faces, Mesa doesn't know anything about this and expects - * each cube face to be treated as a separate image. - * - * These functions present that view to mesa: - */ -const GLuint * -intel_miptree_depth_offsets(struct intel_mipmap_tree *mt, GLuint level) -{ - static const GLuint zero = 0; + mt->level[level].x_offset[img] = mt->level[level].level_x + x; + mt->level[level].y_offset[img] = mt->level[level].level_y + y; - if (mt->target != GL_TEXTURE_3D || mt->level[level].nr_images == 1) - return &zero; - else - return mt->level[level].image_offset; + DBG("%s level %d img %d pos %d,%d\n", + __FUNCTION__, level, img, + mt->level[level].x_offset[img], mt->level[level].y_offset[img]); } -GLuint -intel_miptree_image_offset(struct intel_mipmap_tree *mt, - GLuint face, GLuint level) +void +intel_miptree_get_image_offset(struct intel_mipmap_tree *mt, + GLuint level, GLuint face, GLuint depth, + GLuint *x, GLuint *y) { - if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) - return (mt->level[level].level_offset + - mt->level[level].image_offset[face]); - else - return mt->level[level].level_offset; + if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) { + *x = mt->level[level].x_offset[face]; + *y = mt->level[level].y_offset[face]; + } else if (mt->target == GL_TEXTURE_3D) { + *x = mt->level[level].x_offset[depth]; + *y = mt->level[level].y_offset[depth]; + } else { + *x = mt->level[level].x_offset[0]; + *y = mt->level[level].y_offset[0]; + } } - - /** * Map a teximage in a mipmap tree. * \param row_stride returns row stride in bytes @@ -425,6 +426,7 @@ intel_miptree_image_map(struct intel_context * intel, GLuint level, GLuint * row_stride, GLuint * image_offsets) { + GLuint x, y; DBG("%s \n", __FUNCTION__); if (row_stride) @@ -433,19 +435,26 @@ intel_miptree_image_map(struct intel_context * intel, if (mt->target == GL_TEXTURE_3D) { int i; - for (i = 0; i < mt->level[level].depth; i++) - image_offsets[i] = mt->level[level].image_offset[i] / mt->cpp; + for (i = 0; i < mt->level[level].depth; i++) { + + intel_miptree_get_image_offset(mt, level, face, i, + &x, &y); + image_offsets[i] = x + y * mt->pitch; + } + + return intel_region_map(intel, mt->region); } else { assert(mt->level[level].depth == 1); - assert(mt->target == GL_TEXTURE_CUBE_MAP || - mt->level[level].image_offset[0] == 0); + intel_miptree_get_image_offset(mt, level, face, 0, + &x, &y); image_offsets[0] = 0; - } - return (intel_region_map(intel, mt->region) + - intel_miptree_image_offset(mt, face, level)); + return intel_region_map(intel, mt->region) + + (x + y * mt->pitch) * mt->cpp; + } } + void intel_miptree_image_unmap(struct intel_context *intel, struct intel_mipmap_tree *mt) @@ -455,8 +464,8 @@ intel_miptree_image_unmap(struct intel_context *intel, } - -/* Upload data for a particular image. +/** + * Upload data for a particular image. */ void intel_miptree_image_data(struct intel_context *intel, @@ -467,21 +476,21 @@ intel_miptree_image_data(struct intel_context *intel, GLuint src_row_pitch, GLuint src_image_pitch) { - GLuint depth = dst->level[level].depth; - GLuint dst_offset = intel_miptree_image_offset(dst, face, level); - const GLuint *dst_depth_offset = intel_miptree_depth_offsets(dst, level); + const GLuint depth = dst->level[level].depth; GLuint i; - GLuint height = 0; DBG("%s: %d/%d\n", __FUNCTION__, face, level); for (i = 0; i < depth; i++) { + GLuint dst_x, dst_y, height; + + intel_miptree_get_image_offset(dst, level, face, i, &dst_x, &dst_y); + height = dst->level[level].height; if(dst->compressed) height = (height + 3) / 4; + intel_region_data(intel, - dst->region, - dst_offset + dst_depth_offset[i], /* dst_offset */ - 0, 0, /* dstx, dsty */ + dst->region, 0, dst_x, dst_y, src, src_row_pitch, 0, 0, /* source x, y */ @@ -491,8 +500,9 @@ intel_miptree_image_data(struct intel_context *intel, } } -extern void intel_get_texture_alignment_unit(GLenum, GLuint *, GLuint *); -/* Copy mipmap image between trees + +/** + * Copy mipmap image between trees */ void intel_miptree_image_copy(struct intel_context *intel, @@ -503,38 +513,37 @@ intel_miptree_image_copy(struct intel_context *intel, GLuint width = src->level[level].width; GLuint height = src->level[level].height; GLuint depth = src->level[level].depth; - GLuint dst_offset = intel_miptree_image_offset(dst, face, level); - GLuint src_offset = intel_miptree_image_offset(src, face, level); - const GLuint *dst_depth_offset = intel_miptree_depth_offsets(dst, level); - const GLuint *src_depth_offset = intel_miptree_depth_offsets(src, level); + GLuint src_x, src_y, dst_x, dst_y; GLuint i; GLboolean success; if (dst->compressed) { GLuint align_w, align_h; - intel_get_texture_alignment_unit(dst->internal_format, &align_w, &align_h); + intel_get_texture_alignment_unit(dst->internal_format, + &align_w, &align_h); height = (height + 3) / 4; width = ALIGN(width, align_w); } for (i = 0; i < depth; i++) { + intel_miptree_get_image_offset(src, level, face, i, &src_x, &src_y); + intel_miptree_get_image_offset(dst, level, face, i, &dst_x, &dst_y); success = intel_region_copy(intel, - dst->region, dst_offset + dst_depth_offset[i], - 0, 0, - src->region, src_offset + src_depth_offset[i], - 0, 0, width, height, GL_COPY); + dst->region, 0, dst_x, dst_y, + src->region, 0, src_x, src_y, width, height, + GL_COPY); if (!success) { GLubyte *src_ptr, *dst_ptr; src_ptr = intel_region_map(intel, src->region); dst_ptr = intel_region_map(intel, dst->region); - _mesa_copy_rect(dst_ptr + dst_offset + dst_depth_offset[i], + _mesa_copy_rect(dst_ptr + dst->cpp * (dst_x + dst_y * dst->pitch), dst->cpp, dst->pitch, 0, 0, width, height, - src_ptr + src_offset + src_depth_offset[i], + src_ptr + src->cpp * (src_x + src_y * src->pitch), src->pitch, 0, 0); intel_region_unmap(intel, src->region); diff --git a/shared/intel_mipmap_tree.h b/shared/intel_mipmap_tree.h index c890b2a..b19c548 100644 --- a/shared/intel_mipmap_tree.h +++ b/shared/intel_mipmap_tree.h @@ -70,6 +70,10 @@ struct intel_mipmap_level * always zero in that case. */ GLuint level_offset; + /** Offset to this miptree level, used in computing x_offset. */ + GLuint level_x; + /** Offset to this miptree level, used in computing y_offset. */ + GLuint level_y; GLuint width; GLuint height; /** Depth of the mipmap at this level: 1 for 1D/2D/CUBE, n for 3D. */ @@ -86,7 +90,7 @@ struct intel_mipmap_level * compute the offsets of depth/cube images within a mipmap level, * so have to store them as a lookup table. */ - GLuint *image_offset; + GLuint *x_offset, *y_offset; }; struct intel_mipmap_tree @@ -161,8 +165,7 @@ void intel_miptree_release(struct intel_context *intel, /* Check if an image fits an existing mipmap tree layout */ GLboolean intel_miptree_match_image(struct intel_mipmap_tree *mt, - struct gl_texture_image *image, - GLuint face, GLuint level); + struct gl_texture_image *image); /* Return a pointer to an image within a tree. Return image stride as * well. @@ -176,19 +179,10 @@ GLubyte *intel_miptree_image_map(struct intel_context *intel, void intel_miptree_image_unmap(struct intel_context *intel, struct intel_mipmap_tree *mt); - -/* Return the linear offset of an image relative to the start of the - * tree: - */ -GLuint intel_miptree_image_offset(struct intel_mipmap_tree *mt, - GLuint face, GLuint level); - -/* Return pointers to each 2d slice within an image. Indexed by depth - * value. - */ -const GLuint *intel_miptree_depth_offsets(struct intel_mipmap_tree *mt, - GLuint level); - +void +intel_miptree_get_image_offset(struct intel_mipmap_tree *mt, + GLuint level, GLuint face, GLuint depth, + GLuint *x, GLuint *y); void intel_miptree_set_level_info(struct intel_mipmap_tree *mt, GLuint level, @@ -196,16 +190,10 @@ void intel_miptree_set_level_info(struct intel_mipmap_tree *mt, GLuint x, GLuint y, GLuint w, GLuint h, GLuint d); -void intel_miptree_set_image_offset_ex(struct intel_mipmap_tree *mt, - GLuint level, - GLuint img, GLuint x, GLuint y, - GLuint offset); - void intel_miptree_set_image_offset(struct intel_mipmap_tree *mt, GLuint level, GLuint img, GLuint x, GLuint y); - /* Upload an image into a tree */ void intel_miptree_image_data(struct intel_context *intel, diff --git a/shared/intel_pixel.c b/shared/intel_pixel.c index a300141..993e427 100644 --- a/shared/intel_pixel.c +++ b/shared/intel_pixel.c @@ -129,20 +129,6 @@ intel_check_blit_fragment_ops(GLcontext * ctx, GLboolean src_alpha_is_one) return GL_TRUE; } - -GLboolean -intel_check_meta_tex_fragment_ops(GLcontext * ctx) -{ - if (ctx->NewState) - _mesa_update_state(ctx); - - /* Some of _ImageTransferState (scale, bias) could be done with - * fragment programs on i915. - */ - return !(ctx->_ImageTransferState || ctx->Fog.Enabled || /* not done yet */ - ctx->Texture._EnabledUnits || ctx->FragmentProgram._Enabled); -} - /* The intel_region struct doesn't really do enough to capture the * format of the pixels in the region. For now this code assumes that * the region is a display surface and hence is either ARGB8888 or diff --git a/shared/intel_pixel.h b/shared/intel_pixel.h index 96a6dd1..743b649 100644 --- a/shared/intel_pixel.h +++ b/shared/intel_pixel.h @@ -34,8 +34,6 @@ void intelInitPixelFuncs(struct dd_function_table *functions); GLboolean intel_check_blit_fragment_ops(GLcontext * ctx, GLboolean src_alpha_is_one); -GLboolean intel_check_meta_tex_fragment_ops(GLcontext * ctx); - GLboolean intel_check_blit_format(struct intel_region *region, GLenum format, GLenum type); diff --git a/shared/intel_pixel_bitmap.c b/shared/intel_pixel_bitmap.c index b543a0b..204a233 100644 --- a/shared/intel_pixel_bitmap.c +++ b/shared/intel_pixel_bitmap.c @@ -33,6 +33,7 @@ #include "main/macros.h" #include "main/bufferobj.h" #include "main/pixelstore.h" +#include "main/polygon.h" #include "main/state.h" #include "main/teximage.h" #include "main/texenv.h" @@ -209,7 +210,7 @@ do_blit_bitmap( GLcontext *ctx, if (!dst) return GL_FALSE; - if (unpack->BufferObj->Name) { + if (_mesa_is_bufferobj(unpack->BufferObj)) { bitmap = map_pbo(ctx, width, height, unpack, bitmap); if (bitmap == NULL) return GL_TRUE; /* even though this is an error, we're done */ @@ -329,12 +330,14 @@ out: if (INTEL_DEBUG & DEBUG_SYNC) intel_batchbuffer_flush(intel->batch); - if (unpack->BufferObj->Name) { + if (_mesa_is_bufferobj(unpack->BufferObj)) { /* done with PBO so unmap it now */ ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, unpack->BufferObj); } + intel_check_front_buffer_rendering(intel); + return GL_TRUE; } @@ -418,7 +421,7 @@ intel_texture_bitmap(GLcontext * ctx, return GL_FALSE; } - if (unpack->BufferObj->Name) { + if (_mesa_is_bufferobj(unpack->BufferObj)) { bitmap = map_pbo(ctx, width, height, unpack, bitmap); if (bitmap == NULL) return GL_TRUE; /* even though this is an error, we're done */ @@ -428,20 +431,21 @@ intel_texture_bitmap(GLcontext * ctx, a8_bitmap = _mesa_calloc(width * height); _mesa_expand_bitmap(width, height, unpack, bitmap, a8_bitmap, width, 0xff); - if (unpack->BufferObj->Name) { + if (_mesa_is_bufferobj(unpack->BufferObj)) { /* done with PBO so unmap it now */ ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, unpack->BufferObj); } /* Save GL state before we start setting up our drawing */ - _mesa_PushAttrib(GL_ENABLE_BIT | GL_CURRENT_BIT | - GL_VIEWPORT_BIT); + _mesa_PushAttrib(GL_ENABLE_BIT | GL_CURRENT_BIT | GL_POLYGON_BIT | + GL_TEXTURE_BIT | GL_VIEWPORT_BIT); _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT | GL_CLIENT_PIXEL_STORE_BIT); old_active_texture = ctx->Texture.CurrentUnit; _mesa_Disable(GL_POLYGON_STIPPLE); + _mesa_PolygonMode(GL_FRONT_AND_BACK, GL_FILL); /* Upload our bitmap data to an alpha texture */ _mesa_ActiveTextureARB(GL_TEXTURE0_ARB); @@ -501,8 +505,6 @@ intel_texture_bitmap(GLcontext * ctx, meta_restore_vertex_program(&intel->meta); _mesa_PopClientAttrib(); - _mesa_Disable(GL_TEXTURE_2D); /* asserted that it was disabled at entry */ - _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture); _mesa_PopAttrib(); _mesa_DeleteTextures(1, &texname); diff --git a/shared/intel_pixel_copy.c b/shared/intel_pixel_copy.c index 07ca8f7..622aaa2 100644 --- a/shared/intel_pixel_copy.c +++ b/shared/intel_pixel_copy.c @@ -222,6 +222,8 @@ do_blit_copypixels(GLcontext * ctx, out: UNLOCK_HARDWARE(intel); + intel_check_front_buffer_rendering(intel); + DBG("%s: success\n", __FUNCTION__); return GL_TRUE; } @@ -240,5 +242,5 @@ intelCopyPixels(GLcontext * ctx, return; /* this will use swrast if needed */ - _mesa_meta_copy_pixels(ctx, srcx, srcy, width, height, destx, desty, type); + _mesa_meta_CopyPixels(ctx, srcx, srcy, width, height, destx, desty, type); } diff --git a/shared/intel_pixel_draw.c b/shared/intel_pixel_draw.c index 7fbb89f..9b382e3 100644 --- a/shared/intel_pixel_draw.c +++ b/shared/intel_pixel_draw.c @@ -54,7 +54,7 @@ #include "intel_fbo.h" -/** XXX compare perf of this vs. _mesa_meta_draw_pixels(STENCIL) */ +/** XXX compare perf of this vs. _mesa_meta_DrawPixels(STENCIL) */ static GLboolean intel_stencil_drawpixels(GLcontext * ctx, GLint x, GLint y, @@ -169,7 +169,7 @@ intel_stencil_drawpixels(GLcontext * ctx, * buffer. */ depth_irb = intel_get_renderbuffer(ctx->DrawBuffer, BUFFER_DEPTH); - irb = intel_create_renderbuffer(GL_RGBA8); + irb = intel_create_renderbuffer(MESA_FORMAT_ARGB8888); rb = &irb->Base; irb->Base.Width = depth_irb->Base.Width; irb->Base.Height = depth_irb->Base.Height; @@ -265,7 +265,7 @@ intelDrawPixels(GLcontext * ctx, /* XXX this function doesn't seem to work reliably even when all * the pre-requisite conditions are met. * Note that this function is never hit with conform. - * Fall back to swrast because even the _mesa_meta_draw_pixels() approach + * Fall back to swrast because even the _mesa_meta_DrawPixels() approach * isn't working because of an apparent stencil bug. */ if (intel_stencil_drawpixels(ctx, x, y, width, height, format, type, @@ -280,6 +280,6 @@ intelDrawPixels(GLcontext * ctx, } #endif - _mesa_meta_draw_pixels(ctx, x, y, width, height, format, type, - unpack, pixels); + _mesa_meta_DrawPixels(ctx, x, y, width, height, format, type, + unpack, pixels); } diff --git a/shared/intel_pixel_read.c b/shared/intel_pixel_read.c index 8713463..20424e2 100644 --- a/shared/intel_pixel_read.c +++ b/shared/intel_pixel_read.c @@ -180,16 +180,7 @@ do_blit_readpixels(GLcontext * ctx, if (!src) return GL_FALSE; - if (pack->BufferObj->Name) { - /* XXX This validation should be done by core mesa: - */ - if (!_mesa_validate_pbo_access(2, pack, width, height, 1, - format, type, pixels)) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawPixels"); - return GL_TRUE; - } - } - else { + if (!_mesa_is_bufferobj(pack->BufferObj)) { /* PBO only for now: */ if (INTEL_DEBUG & DEBUG_PIXEL) @@ -225,9 +216,8 @@ do_blit_readpixels(GLcontext * ctx, rowLength = -rowLength; } - /* XXX 64-bit cast? */ - dst_offset = (GLuint) _mesa_image_address(2, pack, pixels, width, height, - format, type, 0, 0, 0); + dst_offset = (GLintptr) _mesa_image_address(2, pack, pixels, width, height, + format, type, 0, 0, 0); /* Although the blits go on the command buffer, need to do this and @@ -236,14 +226,14 @@ do_blit_readpixels(GLcontext * ctx, intelFlush(&intel->ctx); LOCK_HARDWARE(intel); - if (intel->driDrawable->numClipRects) { + if (intel->driReadDrawable->numClipRects) { GLboolean all = (width * height * src->cpp == dst->Base.Size && x == 0 && dst_offset == 0); dri_bo *dst_buffer = intel_bufferobj_buffer(intel, dst, all ? INTEL_WRITE_FULL : INTEL_WRITE_PART); - __DRIdrawablePrivate *dPriv = intel->driDrawable; + __DRIdrawablePrivate *dPriv = intel->driReadDrawable; int nbox = dPriv->numClipRects; drm_clip_rect_t *box = dPriv->pClipRects; drm_clip_rect_t rect; @@ -295,11 +285,11 @@ intelReadPixels(GLcontext * ctx, intelFlush(ctx); -#ifdef I915 if (do_blit_readpixels (ctx, x, y, width, height, format, type, pack, pixels)) return; +#ifdef I915 if (do_texture_readpixels (ctx, x, y, width, height, format, type, pack, pixels)) return; diff --git a/shared/intel_regions.c b/shared/intel_regions.c index a86c66a..8097516 100644 --- a/shared/intel_regions.c +++ b/shared/intel_regions.c @@ -582,8 +582,7 @@ intel_recreate_static(struct intel_context *intel, * instead of which tiling mode it is. Guess. */ if (region_desc->tiled) { - if (IS_965(intel->intelScreen->deviceID) && - region_desc == &intelScreen->depth) + if (intel->gen >= 4 && region_desc == &intelScreen->depth) region->tiling = I915_TILING_Y; else region->tiling = I915_TILING_X; diff --git a/shared/intel_regions.h b/shared/intel_regions.h index 0d379bd..535fcd7 100644 --- a/shared/intel_regions.h +++ b/shared/intel_regions.h @@ -62,6 +62,8 @@ struct intel_region GLuint map_refcount; /**< Reference count for mapping */ GLuint draw_offset; /**< Offset of drawing address within the region */ + GLuint draw_x, draw_y; /**< Offset of drawing within the region */ + uint32_t tiling; /**< Which tiling mode the region is in */ uint32_t bit_6_swizzle; /**< GEM flag for address swizzling requirement */ drmAddress classic_map; /**< drmMap of the region when not in GEM mode */ diff --git a/shared/intel_screen.c b/shared/intel_screen.c index 1b8c56e..789135b 100644 --- a/shared/intel_screen.c +++ b/shared/intel_screen.c @@ -79,6 +79,10 @@ PUBLIC const char __driConfigOptions[] = DRI_CONF_DESC(en, "Enable early Z in classic mode (unstable, 945-only).") DRI_CONF_OPT_END + DRI_CONF_OPT_BEGIN(fragment_shader, bool, false) + DRI_CONF_DESC(en, "Enable limited ARB_fragment_shader support on 915/945.") + DRI_CONF_OPT_END + DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY DRI_CONF_FORCE_S3TC_ENABLE(false) @@ -88,10 +92,14 @@ PUBLIC const char __driConfigOptions[] = DRI_CONF_NO_RAST(false) DRI_CONF_ALWAYS_FLUSH_BATCH(false) DRI_CONF_ALWAYS_FLUSH_CACHE(false) + + DRI_CONF_OPT_BEGIN(stub_occlusion_query, bool, false) + DRI_CONF_DESC(en, "Enable stub ARB_occlusion_query support on 915/945.") + DRI_CONF_OPT_END DRI_CONF_SECTION_END DRI_CONF_END; -const GLuint __driNConfigOptions = 10; +const GLuint __driNConfigOptions = 12; #ifdef USE_NEW_INTERFACE static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; @@ -341,7 +349,7 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, else { GLboolean swStencil = (mesaVis->stencilBits > 0 && mesaVis->depthBits != 24); - GLenum rgbFormat; + gl_format rgbFormat; struct intel_framebuffer *intel_fb = CALLOC_STRUCT(intel_framebuffer); @@ -351,11 +359,11 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, _mesa_initialize_framebuffer(&intel_fb->Base, mesaVis); if (mesaVis->redBits == 5) - rgbFormat = GL_RGB5; + rgbFormat = MESA_FORMAT_RGB565; else if (mesaVis->alphaBits == 0) - rgbFormat = GL_RGB8; + rgbFormat = MESA_FORMAT_XRGB8888; else - rgbFormat = GL_RGBA8; + rgbFormat = MESA_FORMAT_ARGB8888; /* setup the hardware-based renderbuffers */ intel_fb->color_rb[0] = intel_create_renderbuffer(rgbFormat); @@ -374,7 +382,7 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, if (mesaVis->stencilBits == 8) { /* combined depth/stencil buffer */ struct intel_renderbuffer *depthStencilRb - = intel_create_renderbuffer(GL_DEPTH24_STENCIL8_EXT); + = intel_create_renderbuffer(MESA_FORMAT_S8_Z24); /* note: bind RB to two attachment points */ _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthStencilRb->Base); @@ -382,7 +390,7 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, &depthStencilRb->Base); } else { struct intel_renderbuffer *depthRb - = intel_create_renderbuffer(GL_DEPTH_COMPONENT24); + = intel_create_renderbuffer(MESA_FORMAT_X8_Z24); _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthRb->Base); } @@ -390,7 +398,7 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, else if (mesaVis->depthBits == 16) { /* just 16-bit depth buffer, no hw stencil */ struct intel_renderbuffer *depthRb - = intel_create_renderbuffer(GL_DEPTH_COMPONENT16); + = intel_create_renderbuffer(MESA_FORMAT_Z16); _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthRb->Base); } @@ -688,18 +696,6 @@ static const __DRIconfig **intelInitScreen(__DRIscreenPrivate *psp) return NULL; } - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is - * called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - intelInitExtensions(NULL, GL_TRUE); - if (!intelInitDriver(psp)) return NULL; @@ -752,18 +748,6 @@ __DRIconfig **intelInitScreen2(__DRIscreenPrivate *psp) int color; __DRIconfig **configs = NULL; - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is - * called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - intelInitExtensions(NULL, GL_TRUE); - /* Allocate the private area */ intelScreen = (intelScreenPrivate *) CALLOC(sizeof(intelScreenPrivate)); if (!intelScreen) { diff --git a/shared/intel_span.c b/shared/intel_span.c index 8df4990..34c3d9d 100644 --- a/shared/intel_span.c +++ b/shared/intel_span.c @@ -29,7 +29,6 @@ #include "main/macros.h" #include "main/mtypes.h" #include "main/colormac.h" -#include "main/texformat.h" #include "intel_buffers.h" #include "intel_fbo.h" @@ -132,18 +131,6 @@ pwrite_8(struct intel_renderbuffer *irb, uint32_t offset, uint8_t val) dri_bo_subdata(irb->region->buffer, offset, 1, &val); } -static uint32_t -z24s8_to_s8z24(uint32_t val) -{ - return (val << 24) | (val >> 8); -} - -static uint32_t -s8z24_to_z24s8(uint32_t val) -{ - return (val >> 24) | (val << 8); -} - static uint32_t no_tile_swizzle(struct intel_renderbuffer *irb, int x, int y) { @@ -163,6 +150,9 @@ static uint32_t x_tile_swizzle(struct intel_renderbuffer *irb, int x_tile_number, y_tile_number; int tile_off, tile_base; + x += irb->region->draw_x; + y += irb->region->draw_y; + tile_stride = (irb->region->pitch * irb->region->cpp) << 3; xbyte = x * irb->region->cpp; @@ -218,6 +208,9 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb, int x_tile_number, y_tile_number; int tile_off, tile_base; + x += irb->region->draw_x; + y += irb->region->draw_y; + tile_stride = (irb->region->pitch * irb->region->cpp) << 5; xbyte = x * irb->region->cpp; @@ -273,8 +266,11 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb, unsigned int num_cliprects; \ struct drm_clip_rect *cliprects; \ int x_off, y_off; \ + int pitch = irb->region->pitch * irb->region->cpp; \ + void *buf = irb->region->buffer->virtual; \ GLuint p; \ (void) p; \ + (void)buf; (void)pitch; /* unused for non-gttmap. */ \ intel_get_cliprects(intel, &cliprects, &num_cliprects, &x_off, &y_off); /* XXX FBO: this is identical to the macro in spantmp2.h except we get @@ -296,7 +292,6 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb, #define Y_FLIP(_y) ((_y) * yScale + yBias) -/* XXX with GEM, these need to tell the kernel */ #define HW_LOCK() #define HW_UNLOCK() @@ -339,7 +334,7 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb, #include "intel_spantmp.h" /* x8r8g8b8 color span and pixel functions */ -#define INTEL_PIXEL_FMT GL_BGRA +#define INTEL_PIXEL_FMT GL_BGR #define INTEL_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV #define INTEL_READ_VALUE(offset) pread_xrgb8888(irb, offset) #define INTEL_WRITE_VALUE(offset, v) pwrite_xrgb8888(irb, offset, v) @@ -354,6 +349,9 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb, unsigned int num_cliprects; \ struct drm_clip_rect *cliprects; \ int x_off, y_off; \ + int pitch = irb->region->pitch * irb->region->cpp; \ + void *buf = irb->region->buffer->virtual; \ + (void)buf; (void)pitch; /* unused for non-gttmap. */ \ intel_get_cliprects(intel, &cliprects, &num_cliprects, &x_off, &y_off); @@ -366,20 +364,22 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb, #define INTEL_TAG(name) name##_z16 #include "intel_depthtmp.h" -/* z24 depthbuffer functions. */ +/* z24x8 depthbuffer functions. */ #define INTEL_VALUE_TYPE GLuint #define INTEL_WRITE_DEPTH(offset, d) pwrite_32(irb, offset, d) #define INTEL_READ_DEPTH(offset) pread_32(irb, offset) -#define INTEL_TAG(name) name##_z24 +#define INTEL_TAG(name) name##_z24_x8 #include "intel_depthtmp.h" -/* z24s8 depthbuffer functions. */ -#define INTEL_VALUE_TYPE GLuint -#define INTEL_WRITE_DEPTH(offset, d) pwrite_32(irb, offset, z24s8_to_s8z24(d)) -#define INTEL_READ_DEPTH(offset) s8z24_to_z24s8(pread_32(irb, offset)) -#define INTEL_TAG(name) name##_z24_s8 -#include "intel_depthtmp.h" +/** + ** 8-bit stencil function (XXX FBO: This is obsolete) + **/ +/* XXX */ +#define WRITE_STENCIL(_x, _y, d) pwrite_8(irb, NO_TILE(_x, _y) + 3, d) +#define READ_STENCIL(d, _x, _y) d = pread_8(irb, NO_TILE(_x, _y) + 3); +#define TAG(x) intel_gttmap_##x##_z24_s8 +#include "stenciltmp.h" /** ** 8-bit stencil function (XXX FBO: This is obsolete) @@ -413,6 +413,9 @@ intel_renderbuffer_map(struct intel_context *intel, struct gl_renderbuffer *rb) if (irb == NULL || irb->region == NULL) return; + if (intel->intelScreen->kernel_exec_fencing) + drm_intel_gem_bo_map_gtt(irb->region->buffer); + intel_set_span_functions(intel, rb); } @@ -425,7 +428,10 @@ intel_renderbuffer_unmap(struct intel_context *intel, if (irb == NULL || irb->region == NULL) return; - clear_span_cache(irb); + if (intel->intelScreen->kernel_exec_fencing) + drm_intel_gem_bo_unmap_gtt(irb->region->buffer); + else + clear_span_cache(irb); rb->GetRow = NULL; rb->PutRow = NULL; @@ -444,23 +450,30 @@ intel_renderbuffer_unmap(struct intel_context *intel, * _ColorReadBuffer, _DepthBuffer or _StencilBuffer fields. */ static void -intel_map_unmap_buffers(struct intel_context *intel, GLboolean map) +intel_map_unmap_framebuffer(struct intel_context *intel, + struct gl_framebuffer *fb, + GLboolean map) { - GLcontext *ctx = &intel->ctx; - GLuint i, j; + GLuint i; /* color draw buffers */ - for (j = 0; j < ctx->DrawBuffer->_NumColorDrawBuffers; j++) { + for (i = 0; i < fb->_NumColorDrawBuffers; i++) { if (map) - intel_renderbuffer_map(intel, ctx->DrawBuffer->_ColorDrawBuffers[j]); + intel_renderbuffer_map(intel, fb->_ColorDrawBuffers[i]); else - intel_renderbuffer_unmap(intel, ctx->DrawBuffer->_ColorDrawBuffers[j]); + intel_renderbuffer_unmap(intel, fb->_ColorDrawBuffers[i]); } + /* color read buffer */ + if (map) + intel_renderbuffer_map(intel, fb->_ColorReadBuffer); + else + intel_renderbuffer_unmap(intel, fb->_ColorReadBuffer); + /* check for render to textures */ for (i = 0; i < BUFFER_COUNT; i++) { struct gl_renderbuffer_attachment *att = - ctx->DrawBuffer->Attachment + i; + fb->Attachment + i; struct gl_texture_object *tex = att->Texture; if (tex) { /* render to texture */ @@ -472,33 +485,24 @@ intel_map_unmap_buffers(struct intel_context *intel, GLboolean map) } } - /* color read buffers */ - if (map) - intel_renderbuffer_map(intel, ctx->ReadBuffer->_ColorReadBuffer); - else - intel_renderbuffer_unmap(intel, ctx->ReadBuffer->_ColorReadBuffer); - /* depth buffer (Note wrapper!) */ - if (ctx->DrawBuffer->_DepthBuffer) { + if (fb->_DepthBuffer) { if (map) - intel_renderbuffer_map(intel, ctx->DrawBuffer->_DepthBuffer->Wrapped); + intel_renderbuffer_map(intel, fb->_DepthBuffer->Wrapped); else - intel_renderbuffer_unmap(intel, - ctx->DrawBuffer->_DepthBuffer->Wrapped); + intel_renderbuffer_unmap(intel, fb->_DepthBuffer->Wrapped); } /* stencil buffer (Note wrapper!) */ - if (ctx->DrawBuffer->_StencilBuffer) { + if (fb->_StencilBuffer) { if (map) - intel_renderbuffer_map(intel, - ctx->DrawBuffer->_StencilBuffer->Wrapped); + intel_renderbuffer_map(intel, fb->_StencilBuffer->Wrapped); else - intel_renderbuffer_unmap(intel, - ctx->DrawBuffer->_StencilBuffer->Wrapped); + intel_renderbuffer_unmap(intel, fb->_StencilBuffer->Wrapped); } -} - + intel_check_front_buffer_rendering(intel); +} /** * Prepare for software rendering. Map current read/draw framebuffers' @@ -522,7 +526,9 @@ intelSpanRenderStart(GLcontext * ctx) } } - intel_map_unmap_buffers(intel, GL_TRUE); + intel_map_unmap_framebuffer(intel, ctx->DrawBuffer, GL_TRUE); + if (ctx->ReadBuffer != ctx->DrawBuffer) + intel_map_unmap_framebuffer(intel, ctx->ReadBuffer, GL_TRUE); } /** @@ -544,7 +550,9 @@ intelSpanRenderFinish(GLcontext * ctx) } } - intel_map_unmap_buffers(intel, GL_FALSE); + intel_map_unmap_framebuffer(intel, ctx->DrawBuffer, GL_FALSE); + if (ctx->ReadBuffer != ctx->DrawBuffer) + intel_map_unmap_framebuffer(intel, ctx->ReadBuffer, GL_FALSE); UNLOCK_HARDWARE(intel); } @@ -558,6 +566,43 @@ intelInitSpanFuncs(GLcontext * ctx) swdd->SpanRenderFinish = intelSpanRenderFinish; } +void +intel_map_vertex_shader_textures(GLcontext *ctx) +{ + struct intel_context *intel = intel_context(ctx); + int i; + + if (ctx->VertexProgram._Current == NULL) + return; + + for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) { + if (ctx->Texture.Unit[i]._ReallyEnabled && + ctx->VertexProgram._Current->Base.TexturesUsed[i] != 0) { + struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current; + + intel_tex_map_images(intel, intel_texture_object(texObj)); + } + } +} + +void +intel_unmap_vertex_shader_textures(GLcontext *ctx) +{ + struct intel_context *intel = intel_context(ctx); + int i; + + if (ctx->VertexProgram._Current == NULL) + return; + + for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) { + if (ctx->Texture.Unit[i]._ReallyEnabled && + ctx->VertexProgram._Current->Base.TexturesUsed[i] != 0) { + struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current; + + intel_tex_unmap_images(intel, intel_texture_object(texObj)); + } + } +} /** * Plug in appropriate span read/write functions for the given renderbuffer. @@ -578,7 +623,51 @@ intel_set_span_functions(struct intel_context *intel, else tiling = I915_TILING_NONE; - switch (irb->texformat->MesaFormat) { + if (intel->intelScreen->kernel_exec_fencing) { + switch (irb->Base.Format) { + case MESA_FORMAT_RGB565: + intel_gttmap_InitPointers_RGB565(rb); + break; + case MESA_FORMAT_ARGB4444: + intel_gttmap_InitPointers_ARGB4444(rb); + break; + case MESA_FORMAT_ARGB1555: + intel_gttmap_InitPointers_ARGB1555(rb); + break; + case MESA_FORMAT_XRGB8888: + intel_gttmap_InitPointers_xRGB8888(rb); + break; + case MESA_FORMAT_ARGB8888: + intel_gttmap_InitPointers_ARGB8888(rb); + break; + case MESA_FORMAT_Z16: + intel_gttmap_InitDepthPointers_z16(rb); + break; + case MESA_FORMAT_X8_Z24: + intel_gttmap_InitDepthPointers_z24_x8(rb); + break; + case MESA_FORMAT_S8_Z24: + /* There are a few different ways SW asks us to access the S8Z24 data: + * Z24 depth-only depth reads + * S8Z24 depth reads + * S8Z24 stencil reads. + */ + if (rb->Format == MESA_FORMAT_S8_Z24) { + intel_gttmap_InitDepthPointers_z24_x8(rb); + } else if (rb->Format == MESA_FORMAT_S8) { + intel_gttmap_InitStencilPointers_z24_s8(rb); + } + break; + default: + _mesa_problem(NULL, + "Unexpected MesaFormat %d in intelSetSpanFunctions", + irb->Base.Format); + break; + } + return; + } + + switch (irb->Base.Format) { case MESA_FORMAT_RGB565: switch (tiling) { case I915_TILING_NONE: @@ -621,35 +710,33 @@ intel_set_span_functions(struct intel_context *intel, break; } break; + case MESA_FORMAT_XRGB8888: + switch (tiling) { + case I915_TILING_NONE: + default: + intelInitPointers_xRGB8888(rb); + break; + case I915_TILING_X: + intel_XTile_InitPointers_xRGB8888(rb); + break; + case I915_TILING_Y: + intel_YTile_InitPointers_xRGB8888(rb); + break; + } + break; case MESA_FORMAT_ARGB8888: - if (rb->AlphaBits == 0) { /* XXX: Need xRGB8888 Mesa format */ - /* 8888 RGBx */ - switch (tiling) { - case I915_TILING_NONE: - default: - intelInitPointers_xRGB8888(rb); - break; - case I915_TILING_X: - intel_XTile_InitPointers_xRGB8888(rb); - break; - case I915_TILING_Y: - intel_YTile_InitPointers_xRGB8888(rb); - break; - } - } else { - /* 8888 RGBA */ - switch (tiling) { - case I915_TILING_NONE: - default: - intelInitPointers_ARGB8888(rb); - break; - case I915_TILING_X: - intel_XTile_InitPointers_ARGB8888(rb); - break; - case I915_TILING_Y: - intel_YTile_InitPointers_ARGB8888(rb); - break; - } + /* 8888 RGBA */ + switch (tiling) { + case I915_TILING_NONE: + default: + intelInitPointers_ARGB8888(rb); + break; + case I915_TILING_X: + intel_XTile_InitPointers_ARGB8888(rb); + break; + case I915_TILING_Y: + intel_YTile_InitPointers_ARGB8888(rb); + break; } break; case MESA_FORMAT_Z16: @@ -666,39 +753,27 @@ intel_set_span_functions(struct intel_context *intel, break; } break; + case MESA_FORMAT_X8_Z24: case MESA_FORMAT_S8_Z24: /* There are a few different ways SW asks us to access the S8Z24 data: * Z24 depth-only depth reads * S8Z24 depth reads * S8Z24 stencil reads. */ - if (rb->_ActualFormat == GL_DEPTH_COMPONENT24) { - switch (tiling) { - case I915_TILING_NONE: - default: - intelInitDepthPointers_z24(rb); - break; - case I915_TILING_X: - intel_XTile_InitDepthPointers_z24(rb); - break; - case I915_TILING_Y: - intel_YTile_InitDepthPointers_z24(rb); - break; - } - } else if (rb->_ActualFormat == GL_DEPTH24_STENCIL8_EXT) { + if (rb->Format == MESA_FORMAT_S8_Z24) { switch (tiling) { case I915_TILING_NONE: default: - intelInitDepthPointers_z24_s8(rb); + intelInitDepthPointers_z24_x8(rb); break; case I915_TILING_X: - intel_XTile_InitDepthPointers_z24_s8(rb); + intel_XTile_InitDepthPointers_z24_x8(rb); break; case I915_TILING_Y: - intel_YTile_InitDepthPointers_z24_s8(rb); + intel_YTile_InitDepthPointers_z24_x8(rb); break; } - } else if (rb->_ActualFormat == GL_STENCIL_INDEX8_EXT) { + } else if (rb->Format == MESA_FORMAT_S8) { switch (tiling) { case I915_TILING_NONE: default: @@ -711,6 +786,9 @@ intel_set_span_functions(struct intel_context *intel, intel_YTile_InitStencilPointers_z24_s8(rb); break; } + } else { + _mesa_problem(NULL, + "Unexpected ActualFormat in intelSetSpanFunctions"); } break; default: diff --git a/shared/intel_span.h b/shared/intel_span.h index acbeb4a..bffe109 100644 --- a/shared/intel_span.h +++ b/shared/intel_span.h @@ -36,5 +36,7 @@ void intel_renderbuffer_map(struct intel_context *intel, struct gl_renderbuffer *rb); void intel_renderbuffer_unmap(struct intel_context *intel, struct gl_renderbuffer *rb); +void intel_map_vertex_shader_textures(GLcontext *ctx); +void intel_unmap_vertex_shader_textures(GLcontext *ctx); #endif diff --git a/shared/intel_spantmp.h b/shared/intel_spantmp.h index ead0b1c..bad0339 100644 --- a/shared/intel_spantmp.h +++ b/shared/intel_spantmp.h @@ -32,6 +32,12 @@ #define SPANTMP_PIXEL_FMT INTEL_PIXEL_FMT #define SPANTMP_PIXEL_TYPE INTEL_PIXEL_TYPE +#define TAG(x) INTEL_TAG(intel_gttmap_##x) +#define TAG2(x, y) INTEL_TAG(intel_gttmap_##x##y) +#include "spantmp2.h" + +#define SPANTMP_PIXEL_FMT INTEL_PIXEL_FMT +#define SPANTMP_PIXEL_TYPE INTEL_PIXEL_TYPE #define PUT_VALUE(_x, _y, v) INTEL_WRITE_VALUE(NO_TILE(_x, _y), v) #define GET_VALUE(_x, _y) INTEL_READ_VALUE(NO_TILE(_x, _y)) #define TAG(x) INTEL_TAG(intel##x) @@ -48,8 +54,8 @@ #define SPANTMP_PIXEL_FMT INTEL_PIXEL_FMT #define SPANTMP_PIXEL_TYPE INTEL_PIXEL_TYPE -#define PUT_VALUE(_x, _y, v) INTEL_WRITE_VALUE(X_TILE(_x, _y), v) -#define GET_VALUE(_x, _y) INTEL_READ_VALUE(X_TILE(_x, _y)) +#define PUT_VALUE(_x, _y, v) INTEL_WRITE_VALUE(Y_TILE(_x, _y), v) +#define GET_VALUE(_x, _y) INTEL_READ_VALUE(Y_TILE(_x, _y)) #define TAG(x) INTEL_TAG(intel_YTile_##x) #define TAG2(x, y) INTEL_TAG(intel_YTile_##x)##y #include "spantmp2.h" diff --git a/shared/intel_syncobj.c b/shared/intel_syncobj.c index 1286fe9..0d7889d 100644 --- a/shared/intel_syncobj.c +++ b/shared/intel_syncobj.c @@ -114,7 +114,7 @@ static void intel_check_sync(GLcontext *ctx, struct gl_sync_object *s) { struct intel_sync_object *sync = (struct intel_sync_object *)s; - if (sync->bo && drm_intel_bo_busy(sync->bo)) { + if (sync->bo && !drm_intel_bo_busy(sync->bo)) { drm_intel_bo_unreference(sync->bo); sync->bo = NULL; s->StatusFlag = 1; diff --git a/shared/intel_tex.c b/shared/intel_tex.c index df63f29..215a534 100644 --- a/shared/intel_tex.c +++ b/shared/intel_tex.c @@ -2,6 +2,7 @@ #include "main/texobj.h" #include "main/teximage.h" #include "main/mipmap.h" +#include "drivers/common/meta.h" #include "intel_context.h" #include "intel_mipmap_tree.h" #include "intel_tex.h" @@ -158,11 +159,58 @@ timed_memcpy(void *dest, const void *src, size_t n) } #endif /* DO_DEBUG */ + +/** + * Called via ctx->Driver.GenerateMipmap() + * This is basically a wrapper for _mesa_meta_GenerateMipmap() which checks + * if we'll be using software mipmap generation. In that case, we need to + * map/unmap the base level texture image. + */ +static void +intelGenerateMipmap(GLcontext *ctx, GLenum target, + struct gl_texture_object *texObj) +{ + if (_mesa_meta_check_generate_mipmap_fallback(ctx, target, texObj)) { + /* sw path: need to map texture images */ + struct intel_context *intel = intel_context(ctx); + struct intel_texture_object *intelObj = intel_texture_object(texObj); + intel_tex_map_level_images(intel, intelObj, texObj->BaseLevel); + _mesa_generate_mipmap(ctx, target, texObj); + intel_tex_unmap_level_images(intel, intelObj, texObj->BaseLevel); + + { + GLuint nr_faces = (texObj->Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + GLuint face, i; + /* Update the level information in our private data in the new images, + * since it didn't get set as part of a normal TexImage path. + */ + for (face = 0; face < nr_faces; face++) { + for (i = texObj->BaseLevel + 1; i < texObj->MaxLevel; i++) { + struct intel_texture_image *intelImage = + intel_texture_image(texObj->Image[face][i]); + if (!intelImage) + break; + intelImage->level = i; + intelImage->face = face; + /* Unreference the miptree to signal that the new Data is a + * bare pointer from mesa. + */ + intel_miptree_release(intel, &intelImage->mt); + } + } + } + } + else { + _mesa_meta_GenerateMipmap(ctx, target, texObj); + } +} + + void intelInitTextureFuncs(struct dd_function_table *functions) { functions->ChooseTextureFormat = intelChooseTextureFormat; - functions->GenerateMipmap = intel_generate_mipmap; + functions->GenerateMipmap = intelGenerateMipmap; functions->NewTextureObject = intelNewTextureObject; functions->NewTextureImage = intelNewTextureImage; diff --git a/shared/intel_tex.h b/shared/intel_tex.h index 471aa2a..f3cc0ff 100644 --- a/shared/intel_tex.h +++ b/shared/intel_tex.h @@ -29,6 +29,7 @@ #define INTELTEX_INC #include "main/mtypes.h" +#include "main/formats.h" #include "intel_context.h" #include "texmem.h" @@ -41,10 +42,8 @@ void intelInitTextureSubImageFuncs(struct dd_function_table *functions); void intelInitTextureCopyImageFuncs(struct dd_function_table *functions); -const struct gl_texture_format *intelChooseTextureFormat(GLcontext * ctx, - GLint internalFormat, - GLenum format, - GLenum type); +gl_format intelChooseTextureFormat(GLcontext *ctx, GLint internalFormat, + GLenum format, GLenum type); void intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, unsigned long long offset, GLint depth, GLuint pitch); @@ -71,7 +70,4 @@ void intel_tex_unmap_images(struct intel_context *intel, int intel_compressed_num_bytes(GLuint mesaFormat); -void intel_generate_mipmap(GLcontext *ctx, GLenum target, - struct gl_texture_object *texObj); - #endif diff --git a/shared/intel_tex_copy.c b/shared/intel_tex_copy.c index 74f7f58..767d04d 100644 --- a/shared/intel_tex_copy.c +++ b/shared/intel_tex_copy.c @@ -29,8 +29,10 @@ #include "main/enums.h" #include "main/image.h" #include "main/teximage.h" +#include "main/texstate.h" #include "main/mipmap.h" -#include "swrast/swrast.h" + +#include "drivers/common/meta.h" #include "intel_screen.h" #include "intel_context.h" @@ -91,9 +93,7 @@ do_copy_texsubimage(struct intel_context *intel, GLint x, GLint y, GLsizei width, GLsizei height) { GLcontext *ctx = &intel->ctx; - struct gl_texture_object *texObj = intelImage->base.TexObject; - const struct intel_region *src = - get_teximage_source(intel, internalFormat); + const struct intel_region *src = get_teximage_source(intel, internalFormat); if (!intelImage->mt || !src) { if (INTEL_DEBUG & DEBUG_FALLBACKS) @@ -109,33 +109,36 @@ do_copy_texsubimage(struct intel_context *intel, return GL_FALSE; } - intelFlush(ctx); + // intelFlush(ctx); LOCK_HARDWARE(intel); { drm_intel_bo *dst_bo = intel_region_buffer(intel, intelImage->mt->region, INTEL_WRITE_PART); - GLuint image_offset = intel_miptree_image_offset(intelImage->mt, - intelImage->face, - intelImage->level); const GLint orig_x = x; const GLint orig_y = y; + GLuint image_x, image_y; GLshort src_pitch; + /* get dest x/y in destination texture */ + intel_miptree_get_image_offset(intelImage->mt, + intelImage->level, + intelImage->face, + 0, + &image_x, &image_y); /* Update dst for clipped src. Need to also clip the source rect. */ dstx += x - orig_x; dsty += y - orig_y; /* Can't blit to tiled buffers with non-tile-aligned offset. */ - if (intelImage->mt->region->tiling != I915_TILING_NONE && - (image_offset & 4095) != 0) { + if (intelImage->mt->region->tiling == I915_TILING_Y) { UNLOCK_HARDWARE(intel); return GL_FALSE; } if (ctx->ReadBuffer->Name == 0) { /* reading from a window, adjust x, y */ - __DRIdrawablePrivate *dPriv = intel->driDrawable; + const __DRIdrawablePrivate *dPriv = intel->driReadDrawable; y = dPriv->y + (dPriv->h - (y + height)); x += dPriv->x; @@ -152,17 +155,19 @@ do_copy_texsubimage(struct intel_context *intel, src_pitch = src->pitch; } + /* blit from src buffer to texture */ if (!intelEmitCopyBlit(intel, intelImage->mt->cpp, src_pitch, src->buffer, - 0, + src->draw_offset, src->tiling, intelImage->mt->pitch, dst_bo, - image_offset, + 0, intelImage->mt->region->tiling, - x, y, dstx, dsty, width, height, + x, y, image_x + dstx, image_y + dsty, + width, height, GL_COPY)) { UNLOCK_HARDWARE(intel); return GL_FALSE; @@ -171,11 +176,6 @@ do_copy_texsubimage(struct intel_context *intel, UNLOCK_HARDWARE(intel); - /* GL_SGIS_generate_mipmap */ - if (intelImage->level == texObj->BaseLevel && texObj->GenerateMipmap) { - intel_generate_mipmap(ctx, target, texObj); - } - return GL_TRUE; } @@ -185,8 +185,7 @@ intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, GLenum internalFormat, GLint x, GLint y, GLsizei width, GLint border) { - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_unit *texUnit = _mesa_get_current_tex_unit(ctx); struct gl_texture_object *texObj = _mesa_select_tex_object(ctx, texUnit, target); struct gl_texture_image *texImage = @@ -222,8 +221,10 @@ intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, return; fail: - _swrast_copy_teximage1d(ctx, target, level, internalFormat, x, y, - width, border); + if (INTEL_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__); + _mesa_meta_CopyTexImage1D(ctx, target, level, internalFormat, x, y, + width, border); } @@ -233,8 +234,7 @@ intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, GLint x, GLint y, GLsizei width, GLsizei height, GLint border) { - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_unit *texUnit = _mesa_get_current_tex_unit(ctx); struct gl_texture_object *texObj = _mesa_select_tex_object(ctx, texUnit, target); struct gl_texture_image *texImage = @@ -249,7 +249,7 @@ intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, */ ctx->Driver.TexImage2D(ctx, target, level, internalFormat, width, height, border, - GL_RGBA, CHAN_TYPE, NULL, + GL_RGBA, GL_UNSIGNED_BYTE, NULL, &ctx->DefaultPacking, texObj, texImage); srcx = x; @@ -270,8 +270,10 @@ intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, return; fail: - _swrast_copy_teximage2d(ctx, target, level, internalFormat, x, y, - width, height, border); + if (INTEL_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__); + _mesa_meta_CopyTexImage2D(ctx, target, level, internalFormat, x, y, + width, height, border); } @@ -279,8 +281,7 @@ static void intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, GLint xoffset, GLint x, GLint y, GLsizei width) { - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_unit *texUnit = _mesa_get_current_tex_unit(ctx); struct gl_texture_object *texObj = _mesa_select_tex_object(ctx, texUnit, target); struct gl_texture_image *texImage = @@ -295,7 +296,9 @@ intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, if (!do_copy_texsubimage(intel_context(ctx), target, intel_texture_image(texImage), internalFormat, xoffset, 0, x, y, width, 1)) { - _swrast_copy_texsubimage1d(ctx, target, level, xoffset, x, y, width); + if (INTEL_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__); + _mesa_meta_CopyTexSubImage1D(ctx, target, level, xoffset, x, y, width); } } @@ -305,8 +308,7 @@ intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint x, GLint y, GLsizei width, GLsizei height) { - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_unit *texUnit = _mesa_get_current_tex_unit(ctx); struct gl_texture_object *texObj = _mesa_select_tex_object(ctx, texUnit, target); struct gl_texture_image *texImage = @@ -321,10 +323,10 @@ intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, internalFormat, xoffset, yoffset, x, y, width, height)) { - DBG("%s - fallback to swrast\n", __FUNCTION__); - - _swrast_copy_texsubimage2d(ctx, target, level, - xoffset, yoffset, x, y, width, height); + if (INTEL_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__); + _mesa_meta_CopyTexSubImage2D(ctx, target, level, + xoffset, yoffset, x, y, width, height); } } diff --git a/shared/intel_tex_format.c b/shared/intel_tex_format.c index 3322a71..87efb72 100644 --- a/shared/intel_tex_format.c +++ b/shared/intel_tex_format.c @@ -1,7 +1,6 @@ #include "intel_context.h" #include "intel_tex.h" #include "intel_chipset.h" -#include "main/texformat.h" #include "main/enums.h" @@ -16,7 +15,7 @@ * these if we take the step of simply swizzling the colors * immediately after sampling... */ -const struct gl_texture_format * +gl_format intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, GLenum format, GLenum type) { @@ -34,48 +33,48 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, case GL_COMPRESSED_RGBA: if (format == GL_BGRA) { if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { - return &_mesa_texformat_argb8888; + return MESA_FORMAT_ARGB8888; } else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; } else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; } } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444; case 3: case GL_RGB: case GL_COMPRESSED_RGB: if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { - return &_mesa_texformat_rgb565; + return MESA_FORMAT_RGB565; } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; + return do32bpt ? MESA_FORMAT_XRGB8888 : MESA_FORMAT_RGB565; case GL_RGBA8: case GL_RGB10_A2: case GL_RGBA12: case GL_RGBA16: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444; case GL_RGBA4: case GL_RGBA2: - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case GL_RGB5_A1: - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; case GL_RGB8: case GL_RGB10: case GL_RGB12: case GL_RGB16: - return &_mesa_texformat_argb8888; + return MESA_FORMAT_XRGB8888; case GL_RGB5: case GL_RGB4: case GL_R3_G3_B2: - return &_mesa_texformat_rgb565; + return MESA_FORMAT_RGB565; case GL_ALPHA: case GL_ALPHA4: @@ -83,7 +82,7 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, case GL_ALPHA12: case GL_ALPHA16: case GL_COMPRESSED_ALPHA: - return &_mesa_texformat_a8; + return MESA_FORMAT_A8; case 1: case GL_LUMINANCE: @@ -92,18 +91,24 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, case GL_LUMINANCE12: case GL_LUMINANCE16: case GL_COMPRESSED_LUMINANCE: - return &_mesa_texformat_l8; + return MESA_FORMAT_L8; + + case GL_LUMINANCE12_ALPHA4: + case GL_LUMINANCE12_ALPHA12: + case GL_LUMINANCE16_ALPHA16: +#ifndef I915 + return MESA_FORMAT_AL1616; +#else + /* FALLTHROUGH */ +#endif case 2: case GL_LUMINANCE_ALPHA: case GL_LUMINANCE4_ALPHA4: case GL_LUMINANCE6_ALPHA2: case GL_LUMINANCE8_ALPHA8: - case GL_LUMINANCE12_ALPHA4: - case GL_LUMINANCE12_ALPHA12: - case GL_LUMINANCE16_ALPHA16: case GL_COMPRESSED_LUMINANCE_ALPHA: - return &_mesa_texformat_al88; + return MESA_FORMAT_AL88; case GL_INTENSITY: case GL_INTENSITY4: @@ -111,41 +116,41 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, case GL_INTENSITY12: case GL_INTENSITY16: case GL_COMPRESSED_INTENSITY: - return &_mesa_texformat_i8; + return MESA_FORMAT_I8; case GL_YCBCR_MESA: if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) - return &_mesa_texformat_ycbcr; + return MESA_FORMAT_YCBCR; else - return &_mesa_texformat_ycbcr_rev; + return MESA_FORMAT_YCBCR_REV; case GL_COMPRESSED_RGB_FXT1_3DFX: - return &_mesa_texformat_rgb_fxt1; + return MESA_FORMAT_RGB_FXT1; case GL_COMPRESSED_RGBA_FXT1_3DFX: - return &_mesa_texformat_rgba_fxt1; + return MESA_FORMAT_RGBA_FXT1; case GL_RGB_S3TC: case GL_RGB4_S3TC: case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: - return &_mesa_texformat_rgb_dxt1; + return MESA_FORMAT_RGB_DXT1; case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: - return &_mesa_texformat_rgba_dxt1; + return MESA_FORMAT_RGBA_DXT1; case GL_RGBA_S3TC: case GL_RGBA4_S3TC: case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: - return &_mesa_texformat_rgba_dxt3; + return MESA_FORMAT_RGBA_DXT3; case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: - return &_mesa_texformat_rgba_dxt5; + return MESA_FORMAT_RGBA_DXT5; case GL_DEPTH_COMPONENT: case GL_DEPTH_COMPONENT16: case GL_DEPTH_COMPONENT24: case GL_DEPTH_COMPONENT32: #if 0 - return &_mesa_texformat_z16; + return MESA_FORMAT_Z16; #else /* fall-through. * 16bpp depth texture can't be paired with a stencil buffer so @@ -154,7 +159,7 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, #endif case GL_DEPTH_STENCIL_EXT: case GL_DEPTH24_STENCIL8_EXT: - return &_mesa_texformat_s8_z24; + return MESA_FORMAT_S8_Z24; #ifndef I915 case GL_SRGB_EXT: @@ -165,41 +170,41 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, case GL_COMPRESSED_SRGB_ALPHA_EXT: case GL_COMPRESSED_SLUMINANCE_EXT: case GL_COMPRESSED_SLUMINANCE_ALPHA_EXT: - return &_mesa_texformat_sargb8; + return MESA_FORMAT_SARGB8; case GL_SLUMINANCE_EXT: case GL_SLUMINANCE8_EXT: if (IS_G4X(intel->intelScreen->deviceID)) - return &_mesa_texformat_sl8; + return MESA_FORMAT_SL8; else - return &_mesa_texformat_sargb8; + return MESA_FORMAT_SARGB8; case GL_SLUMINANCE_ALPHA_EXT: case GL_SLUMINANCE8_ALPHA8_EXT: if (IS_G4X(intel->intelScreen->deviceID)) - return &_mesa_texformat_sla8; + return MESA_FORMAT_SLA8; else - return &_mesa_texformat_sargb8; + return MESA_FORMAT_SARGB8; case GL_COMPRESSED_SRGB_S3TC_DXT1_EXT: case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT: case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT: case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT: - return &_mesa_texformat_srgb_dxt1; + return MESA_FORMAT_SRGB_DXT1; /* i915 could also do this */ case GL_DUDV_ATI: case GL_DU8DV8_ATI: - return &_mesa_texformat_dudv8; + return MESA_FORMAT_DUDV8; case GL_RGBA_SNORM: case GL_RGBA8_SNORM: - return &_mesa_texformat_signed_rgba8888_rev; + return MESA_FORMAT_SIGNED_RGBA8888_REV; #endif default: fprintf(stderr, "unexpected texture format %s in %s\n", _mesa_lookup_enum_by_nr(internalFormat), __FUNCTION__); - return NULL; + return MESA_FORMAT_NONE; } - return NULL; /* never get here */ + return MESA_FORMAT_NONE; /* never get here */ } int intel_compressed_num_bytes(GLuint mesaFormat) diff --git a/shared/intel_tex_image.c b/shared/intel_tex_image.c index c5f5220..66d61f9 100644 --- a/shared/intel_tex_image.c +++ b/shared/intel_tex_image.c @@ -1,17 +1,15 @@ -#include <stdlib.h> -#include <stdio.h> - #include "main/glheader.h" #include "main/macros.h" #include "main/mtypes.h" #include "main/enums.h" -#include "main/colortab.h" +#include "main/bufferobj.h" #include "main/convolve.h" #include "main/context.h" -#include "main/simple_list.h" +#include "main/formats.h" +#include "main/image.h" #include "main/texcompress.h" -#include "main/texformat.h" +#include "main/texstore.h" #include "main/texgetimage.h" #include "main/texobj.h" #include "main/texstore.h" @@ -73,6 +71,7 @@ guess_and_alloc_mipmap_tree(struct intel_context *intel, GLuint depth = intelImage->base.Depth; GLuint l2width, l2height, l2depth; GLuint i, comp_byte = 0; + GLuint texelBytes; DBG("%s\n", __FUNCTION__); @@ -116,7 +115,8 @@ guess_and_alloc_mipmap_tree(struct intel_context *intel, */ if ((intelObj->base.MinFilter == GL_NEAREST || intelObj->base.MinFilter == GL_LINEAR) && - intelImage->level == firstLevel) { + intelImage->level == firstLevel && + (intel->gen < 4 || firstLevel == 0)) { lastLevel = firstLevel; } else { @@ -127,8 +127,11 @@ guess_and_alloc_mipmap_tree(struct intel_context *intel, } assert(!intelObj->mt); - if (intelImage->base.IsCompressed) - comp_byte = intel_compressed_num_bytes(intelImage->base.TexFormat->MesaFormat); + if (_mesa_is_format_compressed(intelImage->base.TexFormat)) + comp_byte = intel_compressed_num_bytes(intelImage->base.TexFormat); + + texelBytes = _mesa_get_format_bytes(intelImage->base.TexFormat); + intelObj->mt = intel_miptree_create(intel, intelObj->base.Target, intelImage->base._BaseFormat, @@ -138,7 +141,7 @@ guess_and_alloc_mipmap_tree(struct intel_context *intel, width, height, depth, - intelImage->base.TexFormat->TexelBytes, + texelBytes, comp_byte, expect_accelerated_upload); @@ -170,7 +173,7 @@ target_to_face(GLenum target) static GLboolean check_pbo_format(GLint internalFormat, GLenum format, GLenum type, - const struct gl_texture_format *mesa_format) + gl_format mesa_format) { switch (internalFormat) { case 4: @@ -178,12 +181,12 @@ check_pbo_format(GLint internalFormat, return (format == GL_BGRA && (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) && - mesa_format == &_mesa_texformat_argb8888); + mesa_format == MESA_FORMAT_ARGB8888); case 3: case GL_RGB: return (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5 && - mesa_format == &_mesa_texformat_rgb565); + mesa_format == MESA_FORMAT_RGB565); case GL_YCBCR_MESA: return (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE); default: @@ -204,9 +207,12 @@ try_pbo_upload(struct intel_context *intel, { struct intel_buffer_object *pbo = intel_buffer_object(unpack->BufferObj); GLuint src_offset, src_stride; - GLuint dst_offset, dst_stride; + GLuint dst_x, dst_y, dst_stride; + dri_bo *dst_buffer = intel_region_buffer(intel, + intelImage->mt->region, + INTEL_WRITE_FULL); - if (unpack->BufferObj->Name == 0 || + if (!_mesa_is_bufferobj(unpack->BufferObj) || intel->ctx._ImageTransferState || unpack->SkipPixels || unpack->SkipRows) { DBG("%s: failure 1\n", __FUNCTION__); @@ -221,26 +227,23 @@ try_pbo_upload(struct intel_context *intel, else src_stride = width; - dst_offset = intel_miptree_image_offset(intelImage->mt, - intelImage->face, - intelImage->level); + intel_miptree_get_image_offset(intelImage->mt, intelImage->level, + intelImage->face, 0, + &dst_x, &dst_y); dst_stride = intelImage->mt->pitch; - intelFlush(&intel->ctx); + if (drm_intel_bo_references(intel->batch->buf, dst_buffer)) + intelFlush(&intel->ctx); LOCK_HARDWARE(intel); { dri_bo *src_buffer = intel_bufferobj_buffer(intel, pbo, INTEL_READ); - dri_bo *dst_buffer = intel_region_buffer(intel, - intelImage->mt->region, - INTEL_WRITE_FULL); - if (!intelEmitCopyBlit(intel, intelImage->mt->cpp, src_stride, src_buffer, src_offset, GL_FALSE, - dst_stride, dst_buffer, dst_offset, GL_FALSE, - 0, 0, 0, 0, width, height, + dst_stride, dst_buffer, 0, GL_FALSE, + 0, 0, dst_x, dst_y, width, height, GL_COPY)) { UNLOCK_HARDWARE(intel); return GL_FALSE; @@ -262,9 +265,9 @@ try_pbo_zcopy(struct intel_context *intel, { struct intel_buffer_object *pbo = intel_buffer_object(unpack->BufferObj); GLuint src_offset, src_stride; - GLuint dst_offset, dst_stride; + GLuint dst_x, dst_y, dst_stride; - if (unpack->BufferObj->Name == 0 || + if (!_mesa_is_bufferobj(unpack->BufferObj) || intel->ctx._ImageTransferState || unpack->SkipPixels || unpack->SkipRows) { DBG("%s: failure 1\n", __FUNCTION__); @@ -279,13 +282,14 @@ try_pbo_zcopy(struct intel_context *intel, else src_stride = width; - dst_offset = intel_miptree_image_offset(intelImage->mt, - intelImage->face, - intelImage->level); + intel_miptree_get_image_offset(intelImage->mt, intelImage->level, + intelImage->face, 0, + &dst_x, &dst_y); dst_stride = intelImage->mt->pitch; - if (src_stride != dst_stride || dst_offset != 0 || src_offset != 0) { + if (src_stride != dst_stride || dst_x != 0 || dst_y != 0 || + src_offset != 0) { DBG("%s: failure 2\n", __FUNCTION__); return GL_FALSE; } @@ -320,8 +324,6 @@ intelTexImage(GLcontext * ctx, DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__, _mesa_lookup_enum_by_nr(target), level, width, height, depth, border); - intelFlush(ctx); - intelImage->face = target_to_face(target); intelImage->level = level; @@ -330,22 +332,11 @@ intelTexImage(GLcontext * ctx, &postConvHeight); } - /* choose the texture format */ - texImage->TexFormat = intelChooseTextureFormat(ctx, internalFormat, - format, type); - - _mesa_set_fetch_functions(texImage, dims); - - if (texImage->TexFormat->TexelBytes == 0) { - /* must be a compressed format */ + if (_mesa_is_format_compressed(texImage->TexFormat)) { texelBytes = 0; - texImage->IsCompressed = GL_TRUE; - texImage->CompressedSize = - ctx->Driver.CompressedTextureSize(ctx, texImage->Width, - texImage->Height, texImage->Depth, - texImage->TexFormat->MesaFormat); - } else { - texelBytes = texImage->TexFormat->TexelBytes; + } + else { + texelBytes = _mesa_get_format_bytes(texImage->TexFormat); /* Minimum pitch of 32 bytes */ if (postConvWidth * texelBytes < 32) { @@ -378,8 +369,7 @@ intelTexImage(GLcontext * ctx, intelObj->mt->first_level == level && intelObj->mt->last_level == level && intelObj->mt->target != GL_TEXTURE_CUBE_MAP_ARB && - !intel_miptree_match_image(intelObj->mt, &intelImage->base, - intelImage->face, intelImage->level)) { + !intel_miptree_match_image(intelObj->mt, &intelImage->base)) { DBG("release it\n"); intel_miptree_release(intel, &intelObj->mt); @@ -396,17 +386,17 @@ intelTexImage(GLcontext * ctx, assert(!intelImage->mt); if (intelObj->mt && - intel_miptree_match_image(intelObj->mt, &intelImage->base, - intelImage->face, intelImage->level)) { + intel_miptree_match_image(intelObj->mt, &intelImage->base)) { intel_miptree_reference(&intelImage->mt, intelObj->mt); assert(intelImage->mt); } else if (intelImage->base.Border == 0) { int comp_byte = 0; - - if (intelImage->base.IsCompressed) { + GLuint texelBytes = _mesa_get_format_bytes(intelImage->base.TexFormat); + GLenum baseFormat = _mesa_get_format_base_format(intelImage->base.TexFormat); + if (_mesa_is_format_compressed(intelImage->base.TexFormat)) { comp_byte = - intel_compressed_num_bytes(intelImage->base.TexFormat->MesaFormat); + intel_compressed_num_bytes(intelImage->base.TexFormat); } /* Didn't fit in the object miptree, but it's suitable for inclusion in @@ -414,11 +404,11 @@ intelTexImage(GLcontext * ctx, * It'll get moved into the object miptree at validate time. */ intelImage->mt = intel_miptree_create(intel, target, - intelImage->base.TexFormat->BaseFormat, + baseFormat, internalFormat, level, level, width, height, depth, - intelImage->base.TexFormat->TexelBytes, + texelBytes, comp_byte, pixels == NULL); } @@ -427,7 +417,7 @@ intelTexImage(GLcontext * ctx, */ if (dims <= 2 && intelImage->mt && - unpack->BufferObj->Name != 0 && + _mesa_is_bufferobj(unpack->BufferObj) && check_pbo_format(internalFormat, format, type, intelImage->base.TexFormat)) { @@ -482,21 +472,31 @@ intelTexImage(GLcontext * ctx, LOCK_HARDWARE(intel); if (intelImage->mt) { - if (pixels != NULL) + if (pixels != NULL) { + /* Flush any queued rendering with the texture before mapping. */ + if (drm_intel_bo_references(intel->batch->buf, + intelImage->mt->region->buffer)) { + intelFlush(ctx); + } texImage->Data = intel_miptree_image_map(intel, intelImage->mt, intelImage->face, intelImage->level, &dstRowStride, intelImage->base.ImageOffsets); + } + texImage->RowStride = dstRowStride / intelImage->mt->cpp; } else { /* Allocate regular memory and store the image there temporarily. */ - if (texImage->IsCompressed) { - sizeInBytes = texImage->CompressedSize; + if (_mesa_is_format_compressed(texImage->TexFormat)) { + sizeInBytes = _mesa_format_image_size(texImage->TexFormat, + texImage->Width, + texImage->Height, + texImage->Depth); dstRowStride = - _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width); + _mesa_format_row_stride(texImage->TexFormat, width); assert(dims != 3); } else { @@ -527,17 +527,20 @@ intelTexImage(GLcontext * ctx, pixels, srcRowStride, 0, 0); - } else + } + else { memcpy(texImage->Data, pixels, imageSize); - } else if (!texImage->TexFormat->StoreImage(ctx, dims, - texImage->_BaseFormat, - texImage->TexFormat, - texImage->Data, 0, 0, 0, /* dstX/Y/Zoffset */ - dstRowStride, - texImage->ImageOffsets, - width, height, depth, - format, type, pixels, unpack)) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); + } + } + else if (!_mesa_texstore(ctx, dims, + texImage->_BaseFormat, + texImage->TexFormat, + texImage->Data, 0, 0, 0, /* dstX/Y/Zoffset */ + dstRowStride, + texImage->ImageOffsets, + width, height, depth, + format, type, pixels, unpack)) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); } } @@ -550,11 +553,6 @@ intelTexImage(GLcontext * ctx, } UNLOCK_HARDWARE(intel); - - /* GL_SGIS_generate_mipmap */ - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - intel_generate_mipmap(ctx, target, texObj); - } } @@ -670,9 +668,10 @@ intel_get_tex_image(GLcontext * ctx, GLenum target, GLint level, if (compressed) { _mesa_get_compressed_teximage(ctx, target, level, pixels, texObj, texImage); - } else { + } + else { _mesa_get_teximage(ctx, target, level, format, type, pixels, - texObj, texImage); + texObj, texImage); } @@ -735,17 +734,16 @@ intelSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, { struct intel_framebuffer *intel_fb = dPriv->driverPrivate; struct intel_context *intel = pDRICtx->driverPrivate; + GLcontext *ctx = &intel->ctx; struct intel_texture_object *intelObj; struct intel_texture_image *intelImage; struct intel_mipmap_tree *mt; struct intel_renderbuffer *rb; - struct gl_texture_unit *texUnit; struct gl_texture_object *texObj; struct gl_texture_image *texImage; - int level = 0, type, format, internalFormat; + int level = 0, internalFormat; - texUnit = &intel->ctx.Texture.Unit[intel->ctx.Texture.CurrentUnit]; - texObj = _mesa_select_tex_object(&intel->ctx, texUnit, target); + texObj = _mesa_get_current_tex_object(ctx, target); intelObj = intel_texture_object(texObj); if (!intelObj) @@ -760,8 +758,6 @@ intelSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, if (rb->region == NULL) return; - type = GL_BGRA; - format = GL_UNSIGNED_BYTE; if (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT) internalFormat = GL_RGB; else @@ -792,14 +788,14 @@ intelSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, intelImage->face = target_to_face(target); intelImage->level = level; - texImage->TexFormat = intelChooseTextureFormat(&intel->ctx, internalFormat, - type, format); - _mesa_set_fetch_functions(texImage, 2); + if (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT) + texImage->TexFormat = MESA_FORMAT_XRGB8888; + else + texImage->TexFormat = MESA_FORMAT_ARGB8888; texImage->RowStride = rb->region->pitch; intel_miptree_reference(&intelImage->mt, intelObj->mt); - if (!intel_miptree_match_image(intelObj->mt, &intelImage->base, - intelImage->face, intelImage->level)) { + if (!intel_miptree_match_image(intelObj->mt, &intelImage->base)) { fprintf(stderr, "miptree doesn't match image\n"); } diff --git a/shared/intel_tex_layout.h b/shared/intel_tex_layout.h index c9de9b5..a9ac9e7 100644 --- a/shared/intel_tex_layout.h +++ b/shared/intel_tex_layout.h @@ -33,7 +33,7 @@ #include "main/macros.h" -static GLuint minify( GLuint d ) +static INLINE GLuint minify( GLuint d ) { return MAX2(1, d>>1); } diff --git a/shared/intel_tex_obj.h b/shared/intel_tex_obj.h index 5a93461..3ad10d3 100644 --- a/shared/intel_tex_obj.h +++ b/shared/intel_tex_obj.h @@ -66,6 +66,7 @@ struct intel_texture_image * Else there is no image data. */ struct intel_mipmap_tree *mt; + GLboolean used_as_render_target; }; static INLINE struct intel_texture_object * diff --git a/shared/intel_tex_subimage.c b/shared/intel_tex_subimage.c index 8903707..1f68208 100644 --- a/shared/intel_tex_subimage.c +++ b/shared/intel_tex_subimage.c @@ -85,13 +85,13 @@ intelTexSubimage(GLcontext * ctx, &dstRowStride, texImage->ImageOffsets); else { - if (texImage->IsCompressed) { + if (_mesa_is_format_compressed(texImage->TexFormat)) { dstRowStride = - _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width); + _mesa_format_row_stride(texImage->TexFormat, width); assert(dims != 3); } else { - dstRowStride = texImage->RowStride * texImage->TexFormat->TexelBytes; + dstRowStride = texImage->RowStride * _mesa_get_format_bytes(texImage->TexFormat); } } @@ -105,18 +105,20 @@ intelTexSubimage(GLcontext * ctx, xoffset, yoffset / 4, (width + 3) & ~3, (height + 3) / 4, pixels, (width + 3) & ~3, 0, 0); - } else + } + else { memcpy(texImage->Data, pixels, imageSize); + } } else { - if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, - texImage->TexFormat, - texImage->Data, - xoffset, yoffset, zoffset, - dstRowStride, - texImage->ImageOffsets, - width, height, depth, - format, type, pixels, packing)) { + if (!_mesa_texstore(ctx, dims, texImage->_BaseFormat, + texImage->TexFormat, + texImage->Data, + xoffset, yoffset, zoffset, + dstRowStride, + texImage->ImageOffsets, + width, height, depth, + format, type, pixels, packing)) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage"); } } @@ -129,11 +131,6 @@ intelTexSubimage(GLcontext * ctx, } UNLOCK_HARDWARE(intel); - - /* GL_SGIS_generate_mipmap */ - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - intel_generate_mipmap(ctx, target, texObj); - } } diff --git a/shared/intel_tex_validate.c b/shared/intel_tex_validate.c index a284d54..c9a24ac 100644 --- a/shared/intel_tex_validate.c +++ b/shared/intel_tex_validate.c @@ -5,6 +5,7 @@ #include "intel_batchbuffer.h" #include "intel_mipmap_tree.h" #include "intel_tex.h" +#include "intel_chipset.h" #define FILE_DEBUG_FLAG DEBUG_TEXTURE @@ -14,7 +15,8 @@ * GL_TEXTURE_MAX_LOD, GL_TEXTURE_BASE_LEVEL, and GL_TEXTURE_MAX_LEVEL. */ static void -intel_calculate_first_last_level(struct intel_texture_object *intelObj) +intel_calculate_first_last_level(struct intel_context *intel, + struct intel_texture_object *intelObj) { struct gl_texture_object *tObj = &intelObj->base; const struct gl_texture_image *const baseImage = @@ -40,27 +42,27 @@ intel_calculate_first_last_level(struct intel_texture_object *intelObj) firstLevel = lastLevel = tObj->BaseLevel; } else { -#ifdef I915 - firstLevel = tObj->BaseLevel + (GLint) (tObj->MinLod + 0.5); - firstLevel = MAX2(firstLevel, tObj->BaseLevel); - firstLevel = MIN2(firstLevel, tObj->BaseLevel + baseImage->MaxLog2); - lastLevel = tObj->BaseLevel + (GLint) (tObj->MaxLod + 0.5); - lastLevel = MAX2(lastLevel, tObj->BaseLevel); - lastLevel = MIN2(lastLevel, tObj->BaseLevel + baseImage->MaxLog2); - lastLevel = MIN2(lastLevel, tObj->MaxLevel); - lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */ -#else - /* Currently not taking min/max lod into account here, those - * values are programmed as sampler state elsewhere and we - * upload the same mipmap levels regardless. Not sure if - * this makes sense as it means it isn't possible for the app - * to use min/max lod to reduce texture memory pressure: - */ - firstLevel = tObj->BaseLevel; - lastLevel = MIN2(tObj->BaseLevel + baseImage->MaxLog2, - tObj->MaxLevel); - lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */ -#endif + if (intel->gen == 2) { + firstLevel = tObj->BaseLevel + (GLint) (tObj->MinLod + 0.5); + firstLevel = MAX2(firstLevel, tObj->BaseLevel); + firstLevel = MIN2(firstLevel, tObj->BaseLevel + baseImage->MaxLog2); + lastLevel = tObj->BaseLevel + (GLint) (tObj->MaxLod + 0.5); + lastLevel = MAX2(lastLevel, tObj->BaseLevel); + lastLevel = MIN2(lastLevel, tObj->BaseLevel + baseImage->MaxLog2); + lastLevel = MIN2(lastLevel, tObj->MaxLevel); + lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */ + } else { + /* Min/max LOD are taken into account in sampler state. We don't + * want to re-layout textures just because clamping has been applied + * since it means a bunch of blitting around and probably no memory + * savings (since we have to keep the other levels around anyway). + */ + firstLevel = tObj->BaseLevel; + lastLevel = MIN2(tObj->BaseLevel + baseImage->MaxLog2, + tObj->MaxLevel); + /* need at least one level */ + lastLevel = MAX2(firstLevel, lastLevel); + } } break; case GL_TEXTURE_RECTANGLE_NV: @@ -135,9 +137,8 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) /* What levels must the tree include at a minimum? */ - intel_calculate_first_last_level(intelObj); - firstImage = - intel_texture_image(intelObj->base.Image[0][intelObj->firstLevel]); + intel_calculate_first_last_level(intel, intelObj); + firstImage = intel_texture_image(tObj->Image[0][intelObj->firstLevel]); /* Fallback case: */ @@ -165,11 +166,12 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) intel_miptree_reference(&intelObj->mt, firstImage->mt); } - if (firstImage->base.IsCompressed) { - comp_byte = intel_compressed_num_bytes(firstImage->base.TexFormat->MesaFormat); + if (_mesa_is_format_compressed(firstImage->base.TexFormat)) { + comp_byte = intel_compressed_num_bytes(firstImage->base.TexFormat); cpp = comp_byte; } - else cpp = firstImage->base.TexFormat->TexelBytes; + else + cpp = _mesa_get_format_bytes(firstImage->base.TexFormat); /* Check tree can hold all active levels. Check tree matches * target, imageFormat, etc. @@ -189,7 +191,7 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) intelObj->mt->height0 != firstImage->base.Height || intelObj->mt->depth0 != firstImage->base.Depth || intelObj->mt->cpp != cpp || - intelObj->mt->compressed != firstImage->base.IsCompressed)) { + intelObj->mt->compressed != _mesa_is_format_compressed(firstImage->base.TexFormat))) { intel_miptree_release(intel, &intelObj->mt); } @@ -220,8 +222,13 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) intel_texture_image(intelObj->base.Image[face][i]); /* Need to import images in main memory or held in other trees. + * If it's a render target, then its data isn't needed to be in + * the object tree (otherwise we'd be FBO incomplete), and we need + * to keep track of the image's MT as needing to be pulled in still, + * or we'll lose the rendering that's done to it. */ - if (intelObj->mt != intelImage->mt) { + if (intelObj->mt != intelImage->mt && + !intelImage->used_as_render_target) { copy_image_data_to_tree(intel, intelObj, intelImage); } } |