From 1bc84102ad4df377df6c8bf5734b886b7683b939 Mon Sep 17 00:00:00 2001 From: Michel Dänzer Date: Wed, 3 Oct 2007 10:48:56 +0200 Subject: i915: Work around texture pitch related performance drops on i915 at least. --- src/mesa/drivers/dri/i915/intel_mipmap_tree.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) (limited to 'src/mesa/drivers/dri/i915') diff --git a/src/mesa/drivers/dri/i915/intel_mipmap_tree.c b/src/mesa/drivers/dri/i915/intel_mipmap_tree.c index 74f6b2d851..aefb89ac83 100644 --- a/src/mesa/drivers/dri/i915/intel_mipmap_tree.c +++ b/src/mesa/drivers/dri/i915/intel_mipmap_tree.c @@ -104,7 +104,16 @@ intel_miptree_create(struct intel_context *intel, * replaced at some point by some scheme to only do this when really * necessary. */ - mt->pitch = ((mt->pitch * cpp + 63) & ~63) / cpp; + mt->pitch = (mt->pitch * cpp + 63) & ~63; + + /* XXX: At least the i915 seems very upset when the pitch is a multiple + * of 1024 and sometimes 512 bytes - performance can drop by several + * times. Go to the next multiple of 64 for now. + */ + if (!(mt->pitch & 511)) + mt->pitch += 64; + + mt->pitch /= cpp; } mt->region = intel_region_alloc(intel->intelScreen, -- cgit v1.2.3 From 58cdd1dc520d211b65f05fd06b5ba472f552853c Mon Sep 17 00:00:00 2001 From: Michel Dänzer Date: Wed, 3 Oct 2007 11:25:59 +0200 Subject: i915: Only align texture pitch to 64 bytes when textures can be render targets. --- src/mesa/drivers/dri/i915/intel_mipmap_tree.c | 24 ++++++++++++++++-------- 1 file changed, 16 insertions(+), 8 deletions(-) (limited to 'src/mesa/drivers/dri/i915') diff --git a/src/mesa/drivers/dri/i915/intel_mipmap_tree.c b/src/mesa/drivers/dri/i915/intel_mipmap_tree.c index aefb89ac83..2c167a9ab7 100644 --- a/src/mesa/drivers/dri/i915/intel_mipmap_tree.c +++ b/src/mesa/drivers/dri/i915/intel_mipmap_tree.c @@ -99,19 +99,27 @@ intel_miptree_create(struct intel_context *intel, if (ok) { if (!mt->compressed) { - /* XXX: Align pitch to multiple of 64 bytes for now to allow - * render-to-texture to work in all cases. This should probably be - * replaced at some point by some scheme to only do this when really - * necessary. - */ - mt->pitch = (mt->pitch * cpp + 63) & ~63; + int align; + + if (intel->intelScreen->ttm) { + /* XXX: Align pitch to multiple of 64 bytes for now to allow + * render-to-texture to work in all cases. This should probably be + * replaced at some point by some scheme to only do this when really + * necessary. + */ + align = 63; + } else { + align = 3; + } + + mt->pitch = (mt->pitch * cpp + align) & ~align; /* XXX: At least the i915 seems very upset when the pitch is a multiple * of 1024 and sometimes 512 bytes - performance can drop by several - * times. Go to the next multiple of 64 for now. + * times. Go to the next multiple of the required alignment for now. */ if (!(mt->pitch & 511)) - mt->pitch += 64; + mt->pitch += align + 1; mt->pitch /= cpp; } -- cgit v1.2.3 From 6bac9478c39fbe7955d10a21a2d7743697427a56 Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Wed, 3 Oct 2007 16:59:01 -0700 Subject: Replace duplicated intel_reg.h with a shared header. --- src/mesa/drivers/dri/i915/i915_reg.h | 2 - src/mesa/drivers/dri/i915/intel_batchbuffer.c | 2 +- src/mesa/drivers/dri/i915/intel_blit.c | 18 ++--- src/mesa/drivers/dri/i915/intel_context.h | 3 - src/mesa/drivers/dri/i915/intel_reg.h | 88 ------------------------- src/mesa/drivers/dri/i965/intel_blit.c | 29 +++----- src/mesa/drivers/dri/i965/intel_reg.h | 95 --------------------------- src/mesa/drivers/dri/intel/intel_reg.h | 76 +++++++++++++++++++++ 8 files changed, 94 insertions(+), 219 deletions(-) delete mode 100644 src/mesa/drivers/dri/i915/intel_reg.h delete mode 100644 src/mesa/drivers/dri/i965/intel_reg.h create mode 100644 src/mesa/drivers/dri/intel/intel_reg.h (limited to 'src/mesa/drivers/dri/i915') diff --git a/src/mesa/drivers/dri/i915/i915_reg.h b/src/mesa/drivers/dri/i915/i915_reg.h index a9fa56e8a6..b5585e70e7 100644 --- a/src/mesa/drivers/dri/i915/i915_reg.h +++ b/src/mesa/drivers/dri/i915/i915_reg.h @@ -34,8 +34,6 @@ #define I915_SET_FIELD( var, mask, value ) (var &= ~(mask), var |= value) -#define CMD_3D (0x3<<29) - #define PRIM3D_INLINE (CMD_3D | (0x1f<<24)) #define PRIM3D_TRILIST (0x0<<18) #define PRIM3D_TRISTRIP (0x1<<18) diff --git a/src/mesa/drivers/dri/i915/intel_batchbuffer.c b/src/mesa/drivers/dri/i915/intel_batchbuffer.c index 045ff0a5b0..639457d44a 100644 --- a/src/mesa/drivers/dri/i915/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/i915/intel_batchbuffer.c @@ -28,7 +28,7 @@ #include "intel_batchbuffer.h" #include "intel_ioctl.h" #include "intel_decode.h" -#include "i915_debug.h" +#include "intel_reg.h" /* Relocations in kernel space: * - pass dma buffer seperately diff --git a/src/mesa/drivers/dri/i915/intel_blit.c b/src/mesa/drivers/dri/i915/intel_blit.c index 5d97f08434..b00b0d9e1b 100644 --- a/src/mesa/drivers/dri/i915/intel_blit.c +++ b/src/mesa/drivers/dri/i915/intel_blit.c @@ -105,8 +105,7 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv, } else { BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); - CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); + CMD = (XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB); } for (i = 0; i < nbox; i++, pbox++) { @@ -184,8 +183,7 @@ intelEmitFillBlit(struct intel_context *intel, break; case 4: BR13 = dst_pitch | (0xF0 << 16) | (1 << 24) | (1 << 25); - CMD = (XY_COLOR_BLT_CMD | XY_COLOR_BLT_WRITE_ALPHA | - XY_COLOR_BLT_WRITE_RGB); + CMD = (XY_COLOR_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB); break; default: return; @@ -273,8 +271,7 @@ intelEmitCopyBlit(struct intel_context *intel, (((GLint) dst_pitch) & 0xffff) | (translate_raster_op(logic_op) << 16) | (1 << 24) | (1 << 25); CMD = - (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); + (XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB); break; default: return; @@ -443,15 +440,14 @@ intelClearWithBlit(GLcontext * ctx, GLbitfield mask) if (buf == BUFFER_DEPTH || buf == BUFFER_STENCIL) { CMD = XY_COLOR_BLT_CMD; if (clearMask & BUFFER_BIT_DEPTH) - CMD |= XY_COLOR_BLT_WRITE_RGB; + CMD |= XY_BLT_WRITE_RGB; if (clearMask & BUFFER_BIT_STENCIL) - CMD |= XY_COLOR_BLT_WRITE_ALPHA; + CMD |= XY_BLT_WRITE_ALPHA; } else { /* clearing RGBA */ - CMD = (XY_COLOR_BLT_CMD | - XY_COLOR_BLT_WRITE_ALPHA | - XY_COLOR_BLT_WRITE_RGB); + CMD = XY_COLOR_BLT_CMD | + XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB; } } else { diff --git a/src/mesa/drivers/dri/i915/intel_context.h b/src/mesa/drivers/dri/i915/intel_context.h index 1e9ccd5cdc..c8298dd9c4 100644 --- a/src/mesa/drivers/dri/i915/intel_context.h +++ b/src/mesa/drivers/dri/i915/intel_context.h @@ -460,9 +460,6 @@ extern void intelInitStateFuncs(struct dd_function_table *functions); #define BLENDFACT_INV_CONST_ALPHA 0x0f #define BLENDFACT_MASK 0x0f -#define MI_BATCH_BUFFER_END (0xA<<23) - - extern int intel_translate_compare_func(GLenum func); extern int intel_translate_stencil_op(GLenum op); extern int intel_translate_blend_factor(GLenum factor); diff --git a/src/mesa/drivers/dri/i915/intel_reg.h b/src/mesa/drivers/dri/i915/intel_reg.h deleted file mode 100644 index 7828ba6ad3..0000000000 --- a/src/mesa/drivers/dri/i915/intel_reg.h +++ /dev/null @@ -1,88 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * 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, sub license, 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 NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS 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. - * - **************************************************************************/ - - -#ifndef _INTEL_REG_H_ -#define _INTEL_REG_H_ - - - -#define CMD_3D (0x3<<29) - - -#define _3DPRIMITIVE ((0x3<<29)|(0x1f<<24)) -#define PRIM_INDIRECT (1<<23) -#define PRIM_INLINE (0<<23) -#define PRIM_INDIRECT_SEQUENTIAL (0<<17) -#define PRIM_INDIRECT_ELTS (1<<17) - -#define PRIM3D_TRILIST (0x0<<18) -#define PRIM3D_TRISTRIP (0x1<<18) -#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) -#define PRIM3D_TRIFAN (0x3<<18) -#define PRIM3D_POLY (0x4<<18) -#define PRIM3D_LINELIST (0x5<<18) -#define PRIM3D_LINESTRIP (0x6<<18) -#define PRIM3D_RECTLIST (0x7<<18) -#define PRIM3D_POINTLIST (0x8<<18) -#define PRIM3D_DIB (0x9<<18) -#define PRIM3D_MASK (0x1f<<18) - -#define I915PACKCOLOR4444(r,g,b,a) \ - ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) - -#define I915PACKCOLOR1555(r,g,b,a) \ - ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ - ((a) ? 0x8000 : 0)) - -#define I915PACKCOLOR565(r,g,b) \ - ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) - -#define I915PACKCOLOR8888(r,g,b,a) \ - ((a<<24) | (r<<16) | (g<<8) | b) - - - - -#define BR00_BITBLT_CLIENT 0x40000000 -#define BR00_OP_COLOR_BLT 0x10000000 -#define BR00_OP_SRC_COPY_BLT 0x10C00000 -#define BR13_SOLID_PATTERN 0x80000000 - -#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|0x4) -#define XY_COLOR_BLT_WRITE_ALPHA (1<<21) -#define XY_COLOR_BLT_WRITE_RGB (1<<20) - -#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) -#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) -#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) - -#define MI_WAIT_FOR_EVENT ((0x3<<23)) -#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) -#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) - -#endif diff --git a/src/mesa/drivers/dri/i965/intel_blit.c b/src/mesa/drivers/dri/i965/intel_blit.c index da6b68c22b..962bb24f67 100644 --- a/src/mesa/drivers/dri/i965/intel_blit.c +++ b/src/mesa/drivers/dri/i965/intel_blit.c @@ -110,8 +110,7 @@ void intelCopyBuffer( const __DRIdrawablePrivate *dPriv, } else { BR13 = (0xCC << 16) | (1<<24) | (1<<25); - CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); + CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB; } if (src->tiled) { @@ -199,8 +198,7 @@ void intelEmitFillBlit( struct intel_context *intel, break; case 4: BR13 = (0xF0 << 16) | (1<<24) | (1<<25); - CMD = (XY_COLOR_BLT_CMD | XY_COLOR_BLT_WRITE_ALPHA | - XY_COLOR_BLT_WRITE_RGB); + CMD = XY_COLOR_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB; break; default: return; @@ -290,8 +288,7 @@ void intelEmitCopyBlit( struct intel_context *intel, case 4: BR13 = (translate_raster_op(logic_op) << 16) | (1<<24) | (1<<25); - CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); + CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB; break; default: return; @@ -388,12 +385,11 @@ void intelClearWithBlit(GLcontext *ctx, GLbitfield flags) break; case 4: BR13 = (0xF0 << 16) | (1<<24) | (1<<25); - BACK_CMD = FRONT_CMD = (XY_COLOR_BLT_CMD | - XY_COLOR_BLT_WRITE_ALPHA | - XY_COLOR_BLT_WRITE_RGB); + BACK_CMD = FRONT_CMD = XY_COLOR_BLT_CMD | + XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB; DEPTH_CMD = XY_COLOR_BLT_CMD; - if (flags & BUFFER_BIT_DEPTH) DEPTH_CMD |= XY_COLOR_BLT_WRITE_RGB; - if (flags & BUFFER_BIT_STENCIL) DEPTH_CMD |= XY_COLOR_BLT_WRITE_ALPHA; + if (flags & BUFFER_BIT_DEPTH) DEPTH_CMD |= XY_BLT_WRITE_RGB; + if (flags & BUFFER_BIT_STENCIL) DEPTH_CMD |= XY_BLT_WRITE_ALPHA; break; default: return; @@ -517,11 +513,6 @@ void intelClearWithBlit(GLcontext *ctx, GLbitfield flags) } - -#define BR13_565 0x1 -#define BR13_8888 0x3 - - void intelEmitImmediateColorExpandBlit(struct intel_context *intel, GLuint cpp, @@ -576,15 +567,15 @@ intelEmitImmediateColorExpandBlit(struct intel_context *intel, opcode = XY_SETUP_BLT_CMD; if (cpp == 4) - opcode |= XY_COLOR_BLT_WRITE_ALPHA | XY_COLOR_BLT_WRITE_RGB; + opcode |= XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB; if (dst_tiled) opcode |= XY_DST_TILED; br13 = dst_pitch | (translate_raster_op(logic_op) << 16) | (1 << 29); if (cpp == 2) - br13 |= BR13_565 << 24; + br13 |= BR13_565; else - br13 |= BR13_8888 << 24; + br13 |= BR13_8888; BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); OUT_BATCH(opcode); diff --git a/src/mesa/drivers/dri/i965/intel_reg.h b/src/mesa/drivers/dri/i965/intel_reg.h deleted file mode 100644 index 618afb6cbb..0000000000 --- a/src/mesa/drivers/dri/i965/intel_reg.h +++ /dev/null @@ -1,95 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * 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, sub license, 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 NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS 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. - * - **************************************************************************/ - - -#ifndef _INTEL_REG_H_ -#define _INTEL_REG_H_ - - - -#define CMD_3D (0x3<<29) -#define CMD_2D (0x2<<29) - -#define _3DPRIMITIVE ((0x3<<29)|(0x1f<<24)) -#define PRIM_INDIRECT (1<<23) -#define PRIM_INLINE (0<<23) -#define PRIM_INDIRECT_SEQUENTIAL (0<<17) -#define PRIM_INDIRECT_ELTS (1<<17) - -#define PRIM3D_TRILIST (0x0<<18) -#define PRIM3D_TRISTRIP (0x1<<18) -#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) -#define PRIM3D_TRIFAN (0x3<<18) -#define PRIM3D_POLY (0x4<<18) -#define PRIM3D_LINELIST (0x5<<18) -#define PRIM3D_LINESTRIP (0x6<<18) -#define PRIM3D_RECTLIST (0x7<<18) -#define PRIM3D_POINTLIST (0x8<<18) -#define PRIM3D_DIB (0x9<<18) -#define PRIM3D_MASK (0x1f<<18) - -#define I915PACKCOLOR4444(r,g,b,a) \ - ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) - -#define I915PACKCOLOR1555(r,g,b,a) \ - ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ - ((a) ? 0x8000 : 0)) - -#define I915PACKCOLOR565(r,g,b) \ - ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) - -#define I915PACKCOLOR8888(r,g,b,a) \ - ((a<<24) | (r<<16) | (g<<8) | b) - - - - -#define BR00_BITBLT_CLIENT 0x40000000 -#define BR00_OP_COLOR_BLT 0x10000000 -#define BR00_OP_SRC_COPY_BLT 0x10C00000 -#define BR13_SOLID_PATTERN 0x80000000 - -#define XY_SETUP_BLT_CMD (CMD_2D | (1 << 22) | 6) - -#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|0x4) -#define XY_COLOR_BLT_WRITE_ALPHA (1<<21) -#define XY_COLOR_BLT_WRITE_RGB (1<<20) - -#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) -#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) -#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) - -#define XY_BLT_WRITE_ALPHA (1<<21) -#define XY_BLT_WRITE_RGB (1<<20) -#define XY_SRC_TILED (1<<15) -#define XY_DST_TILED (1<<11) - -#define FENCE_LINEAR 0 -#define FENCE_XMAJOR 1 -#define FENCE_YMAJOR 2 - -#endif diff --git a/src/mesa/drivers/dri/intel/intel_reg.h b/src/mesa/drivers/dri/intel/intel_reg.h new file mode 100644 index 0000000000..b9bbcdc704 --- /dev/null +++ b/src/mesa/drivers/dri/intel/intel_reg.h @@ -0,0 +1,76 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * 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, sub license, 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 NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS 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. + * + **************************************************************************/ + +#define CMD_MI (0x0 << 29) +#define CMD_2D (0x2 << 29) +#define CMD_3D (0x3 << 29) + +#define MI_BATCH_BUFFER_END (CMD_MI | 0xA << 23) + +/* Stalls command execution waiting for the given events to have occurred. */ +#define MI_WAIT_FOR_EVENT (CMD_MI | (0x3 << 23)) +#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) +#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) + +/* Primitive dispatch on 830-945 */ +#define _3DPRIMITIVE (CMD_3D | (0x1f << 24)) +#define PRIM_INDIRECT (1<<23) +#define PRIM_INLINE (0<<23) +#define PRIM_INDIRECT_SEQUENTIAL (0<<17) +#define PRIM_INDIRECT_ELTS (1<<17) + +#define PRIM3D_TRILIST (0x0<<18) +#define PRIM3D_TRISTRIP (0x1<<18) +#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) +#define PRIM3D_TRIFAN (0x3<<18) +#define PRIM3D_POLY (0x4<<18) +#define PRIM3D_LINELIST (0x5<<18) +#define PRIM3D_LINESTRIP (0x6<<18) +#define PRIM3D_RECTLIST (0x7<<18) +#define PRIM3D_POINTLIST (0x8<<18) +#define PRIM3D_DIB (0x9<<18) +#define PRIM3D_MASK (0x1f<<18) + +#define XY_SETUP_BLT_CMD (CMD_2D | (0x011 << 22) | 6) + +#define XY_COLOR_BLT_CMD (CMD_2D | (0x50 << 22) | 4) + +#define XY_SRC_COPY_BLT_CMD (CMD_2D | (0x53 << 22) | 6) + +/* BR00 */ +#define XY_BLT_WRITE_ALPHA (1 << 21) +#define XY_BLT_WRITE_RGB (1 << 20) +#define XY_SRC_TILED (1 << 15) +#define XY_DST_TILED (1 << 11) + +/* BR13 */ +#define BR13_565 (0x1 << 24) +#define BR13_8888 (0x3 << 24) + +#define FENCE_LINEAR 0 +#define FENCE_XMAJOR 1 +#define FENCE_YMAJOR 2 -- cgit v1.2.3 From 77e0523fb7769df4bf43747e136b1653b2421b97 Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Thu, 4 Oct 2007 12:07:25 -0700 Subject: [965] Replace various alignment code with a shared ALIGN() macro. In the process, fix some alignment issues: - Scratch space allocation was aligned into units of 1KB, while the allocation wanted units of bytes, so we never allocated enough space for scratch. - GRF register count was programmed as ALIGN(val - 1, 16) / 16 instead of ALIGN(val, 16) / 16 - 1, which overcounted for val != 16n+1. --- src/mesa/drivers/dri/i915/intel_context.h | 2 ++ src/mesa/drivers/dri/i965/brw_clip_state.c | 3 ++- src/mesa/drivers/dri/i965/brw_curbe.c | 2 +- src/mesa/drivers/dri/i965/brw_draw_upload.c | 2 +- src/mesa/drivers/dri/i965/brw_gs_state.c | 3 ++- src/mesa/drivers/dri/i965/brw_sf_state.c | 2 +- src/mesa/drivers/dri/i965/brw_state_cache.c | 2 +- src/mesa/drivers/dri/i965/brw_state_pool.c | 5 ++--- src/mesa/drivers/dri/i965/brw_tex_layout.c | 4 +--- src/mesa/drivers/dri/i965/brw_vs_state.c | 2 +- src/mesa/drivers/dri/i965/brw_wm_state.c | 4 ++-- src/mesa/drivers/dri/i965/bufmgr_fake.c | 2 +- src/mesa/drivers/dri/i965/intel_batchbuffer.c | 4 ++-- src/mesa/drivers/dri/i965/intel_blit.c | 2 +- src/mesa/drivers/dri/i965/intel_context.h | 2 ++ src/mesa/drivers/dri/i965/intel_mipmap_tree.c | 4 ++-- src/mesa/drivers/dri/i965/intel_pixel_bitmap.c | 9 ++------- src/mesa/drivers/dri/intel/intel_tex_layout.c | 20 +++++++------------- 18 files changed, 33 insertions(+), 41 deletions(-) (limited to 'src/mesa/drivers/dri/i915') diff --git a/src/mesa/drivers/dri/i915/intel_context.h b/src/mesa/drivers/dri/i915/intel_context.h index c8298dd9c4..ce9a362944 100644 --- a/src/mesa/drivers/dri/i915/intel_context.h +++ b/src/mesa/drivers/dri/i915/intel_context.h @@ -292,6 +292,8 @@ extern char *__progname; #define SUBPIXEL_X 0.125 #define SUBPIXEL_Y 0.125 +#define ALIGN(value, alignment) ((value + alignment - 1) & ~(alignment - 1)) + #define INTEL_FIREVERTICES(intel) \ do { \ if ((intel)->prim.flush) \ diff --git a/src/mesa/drivers/dri/i965/brw_clip_state.c b/src/mesa/drivers/dri/i965/brw_clip_state.c index ae46d7a86e..ba2f0edf51 100644 --- a/src/mesa/drivers/dri/i965/brw_clip_state.c +++ b/src/mesa/drivers/dri/i965/brw_clip_state.c @@ -43,7 +43,8 @@ static void upload_clip_unit( struct brw_context *brw ) memset(&clip, 0, sizeof(clip)); /* CACHE_NEW_CLIP_PROG */ - clip.thread0.grf_reg_count = ((brw->clip.prog_data->total_grf-1) & ~15) / 16; + clip.thread0.grf_reg_count = + ALIGN(brw->clip.prog_data->total_grf, 16) / 16 - 1; clip.thread0.kernel_start_pointer = brw->clip.prog_gs_offset >> 6; clip.thread3.urb_entry_read_length = brw->clip.prog_data->urb_read_length; clip.thread3.const_urb_entry_read_length = brw->clip.prog_data->curb_read_length; diff --git a/src/mesa/drivers/dri/i965/brw_curbe.c b/src/mesa/drivers/dri/i965/brw_curbe.c index d3c88c1dca..fa4ea42aa6 100644 --- a/src/mesa/drivers/dri/i965/brw_curbe.c +++ b/src/mesa/drivers/dri/i965/brw_curbe.c @@ -304,7 +304,7 @@ static void upload_constant_buffer(struct brw_context *brw) if (!brw_pool_alloc(pool, bufsz, - 6, + 1 << 6, &brw->curbe.gs_offset)) { _mesa_printf("out of GS memory for curbe\n"); assert(0); diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c index 89cd063d46..b7795703fd 100644 --- a/src/mesa/drivers/dri/i965/brw_draw_upload.c +++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c @@ -290,7 +290,7 @@ static void get_space( struct brw_context *brw, struct gl_buffer_object **vbo_return, GLuint *offset_return ) { - size = (size + 63) & ~63; + size = ALIGN(size, 64); if (brw->vb.upload.offset + size > BRW_UPLOAD_INIT_SIZE) wrap_buffers(brw, size); diff --git a/src/mesa/drivers/dri/i965/brw_gs_state.c b/src/mesa/drivers/dri/i965/brw_gs_state.c index 5826c01d4f..5db4dd4603 100644 --- a/src/mesa/drivers/dri/i965/brw_gs_state.c +++ b/src/mesa/drivers/dri/i965/brw_gs_state.c @@ -46,7 +46,8 @@ static void upload_gs_unit( struct brw_context *brw ) /* CACHE_NEW_GS_PROG */ if (brw->gs.prog_active) { - gs.thread0.grf_reg_count = ((brw->gs.prog_data->total_grf-1) & ~15) / 16; + gs.thread0.grf_reg_count = + ALIGN(brw->gs.prog_data->total_grf, 16) / 16 - 1; gs.thread0.kernel_start_pointer = brw->gs.prog_gs_offset >> 6; gs.thread3.urb_entry_read_length = brw->gs.prog_data->urb_read_length; } diff --git a/src/mesa/drivers/dri/i965/brw_sf_state.c b/src/mesa/drivers/dri/i965/brw_sf_state.c index 236c6fd42a..2257916aae 100644 --- a/src/mesa/drivers/dri/i965/brw_sf_state.c +++ b/src/mesa/drivers/dri/i965/brw_sf_state.c @@ -118,7 +118,7 @@ static void upload_sf_unit( struct brw_context *brw ) memset(&sf, 0, sizeof(sf)); /* CACHE_NEW_SF_PROG */ - sf.thread0.grf_reg_count = ((brw->sf.prog_data->total_grf-1) & ~15) / 16; + sf.thread0.grf_reg_count = ALIGN(brw->sf.prog_data->total_grf, 16) / 16 - 1; sf.thread0.kernel_start_pointer = brw->sf.prog_gs_offset >> 6; sf.thread3.urb_entry_read_length = brw->sf.prog_data->urb_read_length; diff --git a/src/mesa/drivers/dri/i965/brw_state_cache.c b/src/mesa/drivers/dri/i965/brw_state_cache.c index 98d765ac0e..0e73ff8390 100644 --- a/src/mesa/drivers/dri/i965/brw_state_cache.c +++ b/src/mesa/drivers/dri/i965/brw_state_cache.c @@ -148,7 +148,7 @@ GLuint brw_upload_cache( struct brw_cache *cache, GLuint hash = hash_key(key, key_size); void *tmp = _mesa_malloc(key_size + cache->aux_size); - if (!brw_pool_alloc(cache->pool, data_size, 6, &offset)) { + if (!brw_pool_alloc(cache->pool, data_size, 1 << 6, &offset)) { /* Should not be possible: */ _mesa_printf("brw_pool_alloc failed\n"); diff --git a/src/mesa/drivers/dri/i965/brw_state_pool.c b/src/mesa/drivers/dri/i965/brw_state_pool.c index 708ae857ab..eda92a2fa8 100644 --- a/src/mesa/drivers/dri/i965/brw_state_pool.c +++ b/src/mesa/drivers/dri/i965/brw_state_pool.c @@ -41,10 +41,9 @@ GLboolean brw_pool_alloc( struct brw_mem_pool *pool, GLuint align, GLuint *offset_return) { - GLuint align_mask = (1<offset + align_mask) & ~align_mask) - pool->offset; + GLuint fixup = ALIGN(pool->offset, align) - pool->offset; - size = (size + 3) & ~3; + size = ALIGN(size, 4); if (pool->offset + fixup + size >= pool->size) { _mesa_printf("%s failed\n", __FUNCTION__); diff --git a/src/mesa/drivers/dri/i965/brw_tex_layout.c b/src/mesa/drivers/dri/i965/brw_tex_layout.c index 2094a1c8ad..e306c9cf10 100644 --- a/src/mesa/drivers/dri/i965/brw_tex_layout.c +++ b/src/mesa/drivers/dri/i965/brw_tex_layout.c @@ -37,8 +37,6 @@ #include "intel_tex_layout.h" #include "macros.h" -#define ALIGN(value, alignment) ((value + alignment - 1) & ~(alignment - 1)) - GLboolean brw_miptree_layout( struct intel_mipmap_tree *mt ) { /* XXX: these vary depending on image format: @@ -64,7 +62,7 @@ GLboolean brw_miptree_layout( struct intel_mipmap_tree *mt ) mt->pitch = ALIGN(width, align_w); pack_y_pitch = (height + 3) / 4; } else { - mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; + mt->pitch = ALIGN(mt->width0 * mt->cpp, 4) / mt->cpp; pack_y_pitch = ALIGN(mt->height0, align_h); } diff --git a/src/mesa/drivers/dri/i965/brw_vs_state.c b/src/mesa/drivers/dri/i965/brw_vs_state.c index c225bf8f5c..f561979138 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_state.c +++ b/src/mesa/drivers/dri/i965/brw_vs_state.c @@ -44,7 +44,7 @@ static void upload_vs_unit( struct brw_context *brw ) /* CACHE_NEW_VS_PROG */ vs.thread0.kernel_start_pointer = brw->vs.prog_gs_offset >> 6; - vs.thread0.grf_reg_count = ((brw->vs.prog_data->total_grf-1) & ~15) / 16; + vs.thread0.grf_reg_count = ALIGN(brw->vs.prog_data->total_grf, 16) / 16 - 1; vs.thread3.urb_entry_read_length = brw->vs.prog_data->urb_read_length; vs.thread3.const_urb_entry_read_length = brw->vs.prog_data->curb_read_length; vs.thread3.dispatch_grf_start_reg = 1; diff --git a/src/mesa/drivers/dri/i965/brw_wm_state.c b/src/mesa/drivers/dri/i965/brw_wm_state.c index 5b4f2abd0e..351de6d90e 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_state.c @@ -62,7 +62,7 @@ static void upload_wm_unit(struct brw_context *brw ) memset(&wm, 0, sizeof(wm)); /* CACHE_NEW_WM_PROG */ - wm.thread0.grf_reg_count = ((brw->wm.prog_data->total_grf-1) & ~15) / 16; + wm.thread0.grf_reg_count = ALIGN(brw->wm.prog_data->total_grf, 16) / 16 - 1; wm.thread0.kernel_start_pointer = brw->wm.prog_gs_offset >> 6; wm.thread3.dispatch_grf_start_reg = brw->wm.prog_data->first_curbe_grf; wm.thread3.urb_entry_read_length = brw->wm.prog_data->urb_read_length; @@ -71,7 +71,7 @@ static void upload_wm_unit(struct brw_context *brw ) wm.wm5.max_threads = max_threads; if (brw->wm.prog_data->total_scratch) { - GLuint per_thread = (brw->wm.prog_data->total_scratch + 1023) / 1024; + GLuint per_thread = ALIGN(brw->wm.prog_data->total_scratch, 1024); GLuint total = per_thread * (max_threads + 1); /* Scratch space -- just have to make sure there is sufficient diff --git a/src/mesa/drivers/dri/i965/bufmgr_fake.c b/src/mesa/drivers/dri/i965/bufmgr_fake.c index a85121122f..65760c40d4 100644 --- a/src/mesa/drivers/dri/i965/bufmgr_fake.c +++ b/src/mesa/drivers/dri/i965/bufmgr_fake.c @@ -168,7 +168,7 @@ static GLboolean alloc_from_pool( struct intel_context *intel, if (!block) return GL_FALSE; - sz = (buf->size + align-1) & ~(align-1); + sz = ALIGN(buf->size, align); block->mem = mmAllocMem(pool->heap, sz, diff --git a/src/mesa/drivers/dri/i965/intel_batchbuffer.c b/src/mesa/drivers/dri/i965/intel_batchbuffer.c index fb58c0e708..7a6293b557 100644 --- a/src/mesa/drivers/dri/i965/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/i965/intel_batchbuffer.c @@ -37,7 +37,7 @@ static void intel_batchbuffer_reset( struct intel_batchbuffer *batch ) assert(batch->map == NULL); batch->offset = (unsigned long)batch->ptr; - batch->offset = (batch->offset + 63) & ~63; + batch->offset = ALIGN(batch->offset, 64); batch->ptr = (unsigned char *) batch->offset; if (BATCH_SZ - batch->offset < BATCH_REFILL) { @@ -208,7 +208,7 @@ void intel_batchbuffer_align( struct intel_batchbuffer *batch, GLuint sz ) { unsigned long ptr = (unsigned long) batch->ptr; - unsigned long aptr = (ptr + align) & ~((unsigned long)align-1); + unsigned long aptr = ALIGN(ptr, align); GLuint fixup = aptr - ptr; if (intel_batchbuffer_space(batch) < fixup + sz) diff --git a/src/mesa/drivers/dri/i965/intel_blit.c b/src/mesa/drivers/dri/i965/intel_blit.c index 7a9e1a2a3f..d1c1c8afb6 100644 --- a/src/mesa/drivers/dri/i965/intel_blit.c +++ b/src/mesa/drivers/dri/i965/intel_blit.c @@ -533,7 +533,7 @@ intelEmitImmediateColorExpandBlit(struct intel_context *intel, GLenum logic_op) { struct xy_text_immediate_blit text; - int dwords = ((src_size + 7) & ~7) / 4; + int dwords = ALIGN(src_size, 8) / 4; uint32_t opcode, br13; assert( logic_op - GL_CLEAR >= 0 ); diff --git a/src/mesa/drivers/dri/i965/intel_context.h b/src/mesa/drivers/dri/i965/intel_context.h index f63c2f613d..65898caaa7 100644 --- a/src/mesa/drivers/dri/i965/intel_context.h +++ b/src/mesa/drivers/dri/i965/intel_context.h @@ -252,6 +252,8 @@ void UNLOCK_HARDWARE( struct intel_context *intel ); #define SUBPIXEL_X 0.125 #define SUBPIXEL_Y 0.125 +#define ALIGN(value, alignment) ((value + alignment - 1) & ~(alignment - 1)) + /* ================================================================ * Color packing: */ diff --git a/src/mesa/drivers/dri/i965/intel_mipmap_tree.c b/src/mesa/drivers/dri/i965/intel_mipmap_tree.c index 0fb33e27f4..268a982a97 100644 --- a/src/mesa/drivers/dri/i965/intel_mipmap_tree.c +++ b/src/mesa/drivers/dri/i965/intel_mipmap_tree.c @@ -233,8 +233,8 @@ GLboolean intel_miptree_image_data(struct intel_context *intel, if (dst->compressed) { alignment = intel_compressed_alignment(dst->internal_format); - src_row_pitch = ((src_row_pitch + alignment - 1) & ~(alignment - 1)); - width = ((width + alignment - 1) & ~(alignment - 1)); + src_row_pitch = ALIGN(src_row_pitch, alignment); + width = ALIGN(width, alignment); height = (height + 3) / 4; } diff --git a/src/mesa/drivers/dri/i965/intel_pixel_bitmap.c b/src/mesa/drivers/dri/i965/intel_pixel_bitmap.c index 79c1fee9c0..3777422619 100644 --- a/src/mesa/drivers/dri/i965/intel_pixel_bitmap.c +++ b/src/mesa/drivers/dri/i965/intel_pixel_bitmap.c @@ -91,11 +91,6 @@ static void set_bit( GLubyte *dest, dest[bit/8] |= 1 << (bit % 8); } -static int align(int x, int align) -{ - return (x + align - 1) & ~(align - 1); -} - /* Extract a rectangle's worth of data from the bitmap. Called * per-cliprect. */ @@ -147,7 +142,7 @@ static GLuint get_bitmap_rect(GLsizei width, GLsizei height, } if (row_align) - bit = (bit + row_align - 1) & ~(row_align - 1); + bit = ALIGN(bit, row_align); } return count; @@ -268,7 +263,7 @@ do_blit_bitmap( GLcontext *ctx, for (px = 0; px < box_w; px += DX) { int h = MIN2(DY, box_h - py); int w = MIN2(DX, box_w - px); - GLuint sz = align(align(w,8) * h, 64)/8; + GLuint sz = ALIGN(ALIGN(w,8) * h, 64)/8; GLenum logic_op = ctx->Color.ColorLogicOpEnabled ? ctx->Color.LogicOp : GL_COPY; diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.c b/src/mesa/drivers/dri/intel/intel_tex_layout.c index fdecd3e186..e3c6e1c17c 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_layout.c +++ b/src/mesa/drivers/dri/intel/intel_tex_layout.c @@ -34,12 +34,6 @@ #include "intel_tex_layout.h" #include "macros.h" - -static int align(int value, int alignment) -{ - return (value + alignment - 1) & ~(alignment - 1); -} - GLuint intel_compressed_alignment(GLenum internalFormat) { GLuint alignment = 4; @@ -70,7 +64,7 @@ void i945_miptree_layout_2d( struct intel_mipmap_tree *mt ) if (mt->compressed) { align_w = intel_compressed_alignment(mt->internal_format); - mt->pitch = align(mt->width0, align_w); + mt->pitch = ALIGN(mt->width0, align_w); } /* May need to adjust pitch to accomodate the placement of @@ -82,10 +76,10 @@ void i945_miptree_layout_2d( struct intel_mipmap_tree *mt ) GLuint mip1_width; if (mt->compressed) { - mip1_width = align(minify(mt->width0), align_w) - + align(minify(minify(mt->width0)), align_w); + mip1_width = ALIGN(minify(mt->width0), align_w) + + ALIGN(minify(minify(mt->width0)), align_w); } else { - mip1_width = align(minify(mt->width0), align_w) + mip1_width = ALIGN(minify(mt->width0), align_w) + minify(minify(mt->width0)); } @@ -97,7 +91,7 @@ void i945_miptree_layout_2d( struct intel_mipmap_tree *mt ) /* Pitch must be a whole number of dwords, even though we * express it in texels. */ - mt->pitch = align(mt->pitch * mt->cpp, 4) / mt->cpp; + mt->pitch = ALIGN(mt->pitch * mt->cpp, 4) / mt->cpp; mt->total_height = 0; for ( level = mt->first_level ; level <= mt->last_level ; level++ ) { @@ -109,7 +103,7 @@ void i945_miptree_layout_2d( struct intel_mipmap_tree *mt ) if (mt->compressed) img_height = MAX2(1, height/4); else - img_height = align(height, align_h); + img_height = ALIGN(height, align_h); /* Because the images are packed better, the final offset @@ -120,7 +114,7 @@ void i945_miptree_layout_2d( struct intel_mipmap_tree *mt ) /* Layout_below: step right after second mipmap. */ if (level == mt->first_level + 1) { - x += align(width, align_w); + x += ALIGN(width, align_w); } else { y += img_height; -- cgit v1.2.3