diff options
Diffstat (limited to 'src/gallium/auxiliary')
23 files changed, 1822 insertions, 1194 deletions
diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp index 5cafe8c3f0..8f7d3b7100 100644 --- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp +++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp @@ -552,7 +552,7 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_SHL: break; - case TGSI_OPCODE_SHR: + case TGSI_OPCODE_ISHR: break; case TGSI_OPCODE_AND: break; @@ -919,7 +919,7 @@ translate_instructionir(llvm::Module *module, break; case TGSI_OPCODE_SHL: break; - case TGSI_OPCODE_SHR: + case TGSI_OPCODE_ISHR: break; case TGSI_OPCODE_AND: break; diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index a9375abd21..ba6f7b15f9 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -80,7 +80,7 @@ struct fenced_buffer_list */ struct fenced_buffer { - /* + /* * Immutable members. */ @@ -126,8 +126,8 @@ fenced_buffer(struct pb_buffer *buf) /** * Add the buffer to the fenced list. * - * fenced_buffer_list::mutex and fenced_buffer::mutex must be held, in this - * order before calling this function. + * fenced_buffer_list::mutex and fenced_buffer::mutex must be held, in this + * order, before calling this function. * * Reference count should be incremented before calling this function. */ @@ -191,7 +191,7 @@ fenced_buffer_remove_locked(struct fenced_buffer_list *fenced_list, * Wait for the fence to expire, and remove it from the fenced list. * * fenced_buffer::mutex must be held. fenced_buffer_list::mutex must not be - * held -- it will + * held -- it will be acquired internally. */ static INLINE enum pipe_error fenced_buffer_finish_locked(struct fenced_buffer_list *fenced_list, @@ -207,7 +207,10 @@ fenced_buffer_finish_locked(struct fenced_buffer_list *fenced_list, assert(pipe_is_referenced(&fenced_buf->base.base.reference)); assert(fenced_buf->fence); - /* Acquire the global lock */ + /* + * Acquire the global lock. Must release buffer mutex first to preserve + * lock order. + */ pipe_mutex_unlock(fenced_buf->mutex); pipe_mutex_lock(fenced_list->mutex); pipe_mutex_lock(fenced_buf->mutex); @@ -217,7 +220,7 @@ fenced_buffer_finish_locked(struct fenced_buffer_list *fenced_list, /* Remove from the fenced list */ /* TODO: remove consequents */ fenced_buffer_remove_locked(fenced_list, fenced_buf); - + p_atomic_dec(&fenced_buf->base.base.reference.count); assert(pipe_is_referenced(&fenced_buf->base.base.reference)); @@ -238,7 +241,7 @@ fenced_buffer_finish_locked(struct fenced_buffer_list *fenced_list, */ static void fenced_buffer_list_check_free_locked(struct fenced_buffer_list *fenced_list, - int wait) + int wait) { struct pb_fence_ops *ops = fenced_list->ops; struct list_head *curr, *next; @@ -274,7 +277,6 @@ fenced_buffer_list_check_free_locked(struct fenced_buffer_list *fenced_list, pb_buf = &fenced_buf->base; pb_reference(&pb_buf, NULL); - curr = next; next = curr->next; @@ -329,7 +331,7 @@ fenced_buffer_map(struct pb_buffer *buf, if((flags & PIPE_BUFFER_USAGE_DONTBLOCK) && ops->fence_signalled(ops, fenced_buf->fence, 0) == 0) { /* Don't wait for the GPU to finish writing */ - goto finish; + goto done; } /* Wait for the GPU to finish writing */ @@ -350,7 +352,7 @@ fenced_buffer_map(struct pb_buffer *buf, fenced_buf->flags |= flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE; } -finish: +done: pipe_mutex_unlock(fenced_buf->mutex); return map; @@ -391,7 +393,7 @@ fenced_buffer_validate(struct pb_buffer *buf, fenced_buf->vl = NULL; fenced_buf->validation_flags = 0; ret = PIPE_OK; - goto finish; + goto done; } assert(flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE); @@ -401,7 +403,7 @@ fenced_buffer_validate(struct pb_buffer *buf, /* Buffer cannot be validated in two different lists */ if(fenced_buf->vl && fenced_buf->vl != vl) { ret = PIPE_ERROR_RETRY; - goto finish; + goto done; } #if 0 @@ -409,7 +411,7 @@ fenced_buffer_validate(struct pb_buffer *buf, if(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE) { /* TODO: wait for the thread that mapped the buffer to unmap it */ ret = PIPE_ERROR_RETRY; - goto finish; + goto done; } /* Final sanity checking */ assert(!(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE)); @@ -420,17 +422,17 @@ fenced_buffer_validate(struct pb_buffer *buf, (fenced_buf->validation_flags & flags) == flags) { /* Nothing to do -- buffer already validated */ ret = PIPE_OK; - goto finish; + goto done; } ret = pb_validate(fenced_buf->buffer, vl, flags); if (ret != PIPE_OK) - goto finish; + goto done; fenced_buf->vl = vl; fenced_buf->validation_flags |= flags; -finish: +done: pipe_mutex_unlock(fenced_buf->mutex); return ret; diff --git a/src/gallium/auxiliary/rtasm/rtasm_execmem.c b/src/gallium/auxiliary/rtasm/rtasm_execmem.c index 01811d5011..ffed768f97 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_execmem.c +++ b/src/gallium/auxiliary/rtasm/rtasm_execmem.c @@ -41,6 +41,12 @@ #define MAP_ANONYMOUS MAP_ANON #endif +#if defined(PIPE_OS_WINDOWS) +#ifndef WIN32_LEAN_AND_MEAN +#define WIN32_LEAN_AND_MEAN 1 +#endif +#include <windows.h> +#endif #if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) @@ -118,7 +124,29 @@ rtasm_exec_free(void *addr) } -#else /* PIPE_OS_LINUX || PIPE_OS_BSD || PIPE_OS_SOLARIS */ +#elif defined(PIPE_OS_WINDOWS) + + +/* + * Avoid Data Execution Prevention. + */ + +void * +rtasm_exec_malloc(size_t size) +{ + return VirtualAlloc(NULL, size, MEM_COMMIT, PAGE_EXECUTE_READWRITE); +} + + +void +rtasm_exec_free(void *addr) +{ + VirtualFree(addr, 0, MEM_RELEASE); +} + + +#else + /* * Just use regular memory. @@ -138,4 +166,4 @@ rtasm_exec_free(void *addr) } -#endif /* PIPE_OS_LINUX || PIPE_OS_BSD || PIPE_OS_SOLARIS */ +#endif diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c index 2c65ff16d8..e2e5394f86 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c @@ -128,7 +128,9 @@ static const char *semantic_names[] = static const char *immediate_type_names[] = { - "FLT32" + "FLT32", + "UINT32", + "INT32" }; static const char *swizzle_names[] = @@ -412,6 +414,12 @@ iter_immediate( case TGSI_IMM_FLOAT32: FLT( imm->u[i].Float ); break; + case TGSI_IMM_UINT32: + UID(imm->u[i].Uint); + break; + case TGSI_IMM_INT32: + SID(imm->u[i].Int); + break; default: assert( 0 ); } diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index ba89f2fbc3..2bcb33392a 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -2,6 +2,7 @@ * * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. + * Copyright 2009-2010 VMware, Inc. All rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the @@ -60,6 +61,7 @@ #include "util/u_memory.h" #include "util/u_math.h" + #define FAST_MATH 1 #define TILE_TOP_LEFT 0 @@ -67,11 +69,329 @@ #define TILE_BOTTOM_LEFT 2 #define TILE_BOTTOM_RIGHT 3 +static void +micro_abs(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = fabsf(src->f[0]); + dst->f[1] = fabsf(src->f[1]); + dst->f[2] = fabsf(src->f[2]); + dst->f[3] = fabsf(src->f[3]); +} + +static void +micro_arl(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->i[0] = (int)floorf(src->f[0]); + dst->i[1] = (int)floorf(src->f[1]); + dst->i[2] = (int)floorf(src->f[2]); + dst->i[3] = (int)floorf(src->f[3]); +} + +static void +micro_arr(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->i[0] = (int)floorf(src->f[0] + 0.5f); + dst->i[1] = (int)floorf(src->f[1] + 0.5f); + dst->i[2] = (int)floorf(src->f[2] + 0.5f); + dst->i[3] = (int)floorf(src->f[3] + 0.5f); +} + +static void +micro_ceil(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = ceilf(src->f[0]); + dst->f[1] = ceilf(src->f[1]); + dst->f[2] = ceilf(src->f[2]); + dst->f[3] = ceilf(src->f[3]); +} + +static void +micro_cos(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = cosf(src->f[0]); + dst->f[1] = cosf(src->f[1]); + dst->f[2] = cosf(src->f[2]); + dst->f[3] = cosf(src->f[3]); +} + +static void +micro_ddx(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = + dst->f[1] = + dst->f[2] = + dst->f[3] = src->f[TILE_BOTTOM_RIGHT] - src->f[TILE_BOTTOM_LEFT]; +} + +static void +micro_ddy(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = + dst->f[1] = + dst->f[2] = + dst->f[3] = src->f[TILE_BOTTOM_LEFT] - src->f[TILE_TOP_LEFT]; +} + +static void +micro_exp2(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ +#if FAST_MATH + dst->f[0] = util_fast_exp2(src->f[0]); + dst->f[1] = util_fast_exp2(src->f[1]); + dst->f[2] = util_fast_exp2(src->f[2]); + dst->f[3] = util_fast_exp2(src->f[3]); +#else +#if DEBUG + /* Inf is okay for this instruction, so clamp it to silence assertions. */ + uint i; + union tgsi_exec_channel clamped; + + for (i = 0; i < 4; i++) { + if (src->f[i] > 127.99999f) { + clamped.f[i] = 127.99999f; + } else if (src->f[i] < -126.99999f) { + clamped.f[i] = -126.99999f; + } else { + clamped.f[i] = src->f[i]; + } + } + src = &clamped; +#endif /* DEBUG */ + + dst->f[0] = powf(2.0f, src->f[0]); + dst->f[1] = powf(2.0f, src->f[1]); + dst->f[2] = powf(2.0f, src->f[2]); + dst->f[3] = powf(2.0f, src->f[3]); +#endif /* FAST_MATH */ +} + +static void +micro_flr(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = floorf(src->f[0]); + dst->f[1] = floorf(src->f[1]); + dst->f[2] = floorf(src->f[2]); + dst->f[3] = floorf(src->f[3]); +} + +static void +micro_frc(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = src->f[0] - floorf(src->f[0]); + dst->f[1] = src->f[1] - floorf(src->f[1]); + dst->f[2] = src->f[2] - floorf(src->f[2]); + dst->f[3] = src->f[3] - floorf(src->f[3]); +} + +static void +micro_iabs(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->i[0] = src->i[0] >= 0 ? src->i[0] : -src->i[0]; + dst->i[1] = src->i[1] >= 0 ? src->i[1] : -src->i[1]; + dst->i[2] = src->i[2] >= 0 ? src->i[2] : -src->i[2]; + dst->i[3] = src->i[3] >= 0 ? src->i[3] : -src->i[3]; +} + +static void +micro_ineg(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->i[0] = -src->i[0]; + dst->i[1] = -src->i[1]; + dst->i[2] = -src->i[2]; + dst->i[3] = -src->i[3]; +} + +static void +micro_lg2(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ +#if FAST_MATH + dst->f[0] = util_fast_log2(src->f[0]); + dst->f[1] = util_fast_log2(src->f[1]); + dst->f[2] = util_fast_log2(src->f[2]); + dst->f[3] = util_fast_log2(src->f[3]); +#else + dst->f[0] = logf(src->f[0]) * 1.442695f; + dst->f[1] = logf(src->f[1]) * 1.442695f; + dst->f[2] = logf(src->f[2]) * 1.442695f; + dst->f[3] = logf(src->f[3]) * 1.442695f; +#endif +} + +static void +micro_lrp(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = src[0].f[0] * (src[1].f[0] - src[2].f[0]) + src[2].f[0]; + dst->f[1] = src[0].f[1] * (src[1].f[1] - src[2].f[1]) + src[2].f[1]; + dst->f[2] = src[0].f[2] * (src[1].f[2] - src[2].f[2]) + src[2].f[2]; + dst->f[3] = src[0].f[3] * (src[1].f[3] - src[2].f[3]) + src[2].f[3]; +} + +static void +micro_mad(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = src[0].f[0] * src[1].f[0] + src[2].f[0]; + dst->f[1] = src[0].f[1] * src[1].f[1] + src[2].f[1]; + dst->f[2] = src[0].f[2] * src[1].f[2] + src[2].f[2]; + dst->f[3] = src[0].f[3] * src[1].f[3] + src[2].f[3]; +} + +static void +micro_mov(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = src->u[0]; + dst->u[1] = src->u[1]; + dst->u[2] = src->u[2]; + dst->u[3] = src->u[3]; +} + +static void +micro_rcp(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = 1.0f / src->f[0]; + dst->f[1] = 1.0f / src->f[1]; + dst->f[2] = 1.0f / src->f[2]; + dst->f[3] = 1.0f / src->f[3]; +} + +static void +micro_rnd(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = floorf(src->f[0] + 0.5f); + dst->f[1] = floorf(src->f[1] + 0.5f); + dst->f[2] = floorf(src->f[2] + 0.5f); + dst->f[3] = floorf(src->f[3] + 0.5f); +} + +static void +micro_rsq(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = 1.0f / sqrtf(fabsf(src->f[0])); + dst->f[1] = 1.0f / sqrtf(fabsf(src->f[1])); + dst->f[2] = 1.0f / sqrtf(fabsf(src->f[2])); + dst->f[3] = 1.0f / sqrtf(fabsf(src->f[3])); +} + +static void +micro_seq(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = src[0].f[0] == src[1].f[0] ? 1.0f : 0.0f; + dst->f[1] = src[0].f[1] == src[1].f[1] ? 1.0f : 0.0f; + dst->f[2] = src[0].f[2] == src[1].f[2] ? 1.0f : 0.0f; + dst->f[3] = src[0].f[3] == src[1].f[3] ? 1.0f : 0.0f; +} + +static void +micro_sge(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = src[0].f[0] >= src[1].f[0] ? 1.0f : 0.0f; + dst->f[1] = src[0].f[1] >= src[1].f[1] ? 1.0f : 0.0f; + dst->f[2] = src[0].f[2] >= src[1].f[2] ? 1.0f : 0.0f; + dst->f[3] = src[0].f[3] >= src[1].f[3] ? 1.0f : 0.0f; +} + +static void +micro_sgn(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = src->f[0] < 0.0f ? -1.0f : src->f[0] > 0.0f ? 1.0f : 0.0f; + dst->f[1] = src->f[1] < 0.0f ? -1.0f : src->f[1] > 0.0f ? 1.0f : 0.0f; + dst->f[2] = src->f[2] < 0.0f ? -1.0f : src->f[2] > 0.0f ? 1.0f : 0.0f; + dst->f[3] = src->f[3] < 0.0f ? -1.0f : src->f[3] > 0.0f ? 1.0f : 0.0f; +} + +static void +micro_sgt(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = src[0].f[0] > src[1].f[0] ? 1.0f : 0.0f; + dst->f[1] = src[0].f[1] > src[1].f[1] ? 1.0f : 0.0f; + dst->f[2] = src[0].f[2] > src[1].f[2] ? 1.0f : 0.0f; + dst->f[3] = src[0].f[3] > src[1].f[3] ? 1.0f : 0.0f; +} + +static void +micro_sin(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = sinf(src->f[0]); + dst->f[1] = sinf(src->f[1]); + dst->f[2] = sinf(src->f[2]); + dst->f[3] = sinf(src->f[3]); +} + +static void +micro_sle(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = src[0].f[0] <= src[1].f[0] ? 1.0f : 0.0f; + dst->f[1] = src[0].f[1] <= src[1].f[1] ? 1.0f : 0.0f; + dst->f[2] = src[0].f[2] <= src[1].f[2] ? 1.0f : 0.0f; + dst->f[3] = src[0].f[3] <= src[1].f[3] ? 1.0f : 0.0f; +} + +static void +micro_slt(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = src[0].f[0] < src[1].f[0] ? 1.0f : 0.0f; + dst->f[1] = src[0].f[1] < src[1].f[1] ? 1.0f : 0.0f; + dst->f[2] = src[0].f[2] < src[1].f[2] ? 1.0f : 0.0f; + dst->f[3] = src[0].f[3] < src[1].f[3] ? 1.0f : 0.0f; +} + +static void +micro_sne(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = src[0].f[0] != src[1].f[0] ? 1.0f : 0.0f; + dst->f[1] = src[0].f[1] != src[1].f[1] ? 1.0f : 0.0f; + dst->f[2] = src[0].f[2] != src[1].f[2] ? 1.0f : 0.0f; + dst->f[3] = src[0].f[3] != src[1].f[3] ? 1.0f : 0.0f; +} + +static void +micro_trunc(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = (float)(int)src->f[0]; + dst->f[1] = (float)(int)src->f[1]; + dst->f[2] = (float)(int)src->f[2]; + dst->f[3] = (float)(int)src->f[3]; +} + + #define CHAN_X 0 #define CHAN_Y 1 #define CHAN_Z 2 #define CHAN_W 3 +enum tgsi_exec_datatype { + TGSI_EXEC_DATA_FLOAT, + TGSI_EXEC_DATA_INT, + TGSI_EXEC_DATA_UINT +}; + /* * Shorthand locations of various utility registers (_I = Index, _C = Channel) */ @@ -123,23 +443,19 @@ /** The execution mask depends on the conditional mask and the loop mask */ #define UPDATE_EXEC_MASK(MACH) \ - MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask & MACH->FuncMask + MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask & MACH->Switch.mask & MACH->FuncMask static const union tgsi_exec_channel ZeroVec = { { 0.0, 0.0, 0.0, 0.0 } }; -#ifdef DEBUG -static void -check_inf_or_nan(const union tgsi_exec_channel *chan) -{ - assert(!util_is_inf_or_nan(chan->f[0])); - assert(!util_is_inf_or_nan(chan->f[1])); - assert(!util_is_inf_or_nan(chan->f[2])); - assert(!util_is_inf_or_nan(chan->f[3])); -} -#endif +#define CHECK_INF_OR_NAN(chan) do {\ + assert(!util_is_inf_or_nan((chan)->f[0]));\ + assert(!util_is_inf_or_nan((chan)->f[1]));\ + assert(!util_is_inf_or_nan((chan)->f[2]));\ + assert(!util_is_inf_or_nan((chan)->f[3]));\ + } while (0) #ifdef DEBUG @@ -422,18 +738,6 @@ tgsi_exec_machine_destroy(struct tgsi_exec_machine *mach) align_free(mach); } - -static void -micro_abs( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = fabsf( src->f[0] ); - dst->f[1] = fabsf( src->f[1] ); - dst->f[2] = fabsf( src->f[2] ); - dst->f[3] = fabsf( src->f[3] ); -} - static void micro_add( union tgsi_exec_channel *dst, @@ -446,76 +750,6 @@ micro_add( dst->f[3] = src0->f[3] + src1->f[3]; } -#if 0 -static void -micro_iadd( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] + src1->i[0]; - dst->i[1] = src0->i[1] + src1->i[1]; - dst->i[2] = src0->i[2] + src1->i[2]; - dst->i[3] = src0->i[3] + src1->i[3]; -} -#endif - -static void -micro_and( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] & src1->u[0]; - dst->u[1] = src0->u[1] & src1->u[1]; - dst->u[2] = src0->u[2] & src1->u[2]; - dst->u[3] = src0->u[3] & src1->u[3]; -} - -static void -micro_ceil( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = ceilf( src->f[0] ); - dst->f[1] = ceilf( src->f[1] ); - dst->f[2] = ceilf( src->f[2] ); - dst->f[3] = ceilf( src->f[3] ); -} - -static void -micro_cos( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = cosf( src->f[0] ); - dst->f[1] = cosf( src->f[1] ); - dst->f[2] = cosf( src->f[2] ); - dst->f[3] = cosf( src->f[3] ); -} - -static void -micro_ddx( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = - dst->f[1] = - dst->f[2] = - dst->f[3] = src->f[TILE_BOTTOM_RIGHT] - src->f[TILE_BOTTOM_LEFT]; -} - -static void -micro_ddy( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = - dst->f[1] = - dst->f[2] = - dst->f[3] = src->f[TILE_BOTTOM_LEFT] - src->f[TILE_TOP_LEFT]; -} - static void micro_div( union tgsi_exec_channel *dst, @@ -536,99 +770,6 @@ micro_div( } } -#if 0 -static void -micro_udiv( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] / src1->u[0]; - dst->u[1] = src0->u[1] / src1->u[1]; - dst->u[2] = src0->u[2] / src1->u[2]; - dst->u[3] = src0->u[3] / src1->u[3]; -} -#endif - -static void -micro_eq( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] == src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] == src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] == src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] == src1->f[3] ? src2->f[3] : src3->f[3]; -} - -#if 0 -static void -micro_ieq( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->i[0] = src0->i[0] == src1->i[0] ? src2->i[0] : src3->i[0]; - dst->i[1] = src0->i[1] == src1->i[1] ? src2->i[1] : src3->i[1]; - dst->i[2] = src0->i[2] == src1->i[2] ? src2->i[2] : src3->i[2]; - dst->i[3] = src0->i[3] == src1->i[3] ? src2->i[3] : src3->i[3]; -} -#endif - -static void -micro_exp2( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src) -{ -#if FAST_MATH - dst->f[0] = util_fast_exp2( src->f[0] ); - dst->f[1] = util_fast_exp2( src->f[1] ); - dst->f[2] = util_fast_exp2( src->f[2] ); - dst->f[3] = util_fast_exp2( src->f[3] ); -#else - -#if DEBUG - /* Inf is okay for this instruction, so clamp it to silence assertions. */ - uint i; - union tgsi_exec_channel clamped; - - for (i = 0; i < 4; i++) { - if (src->f[i] > 127.99999f) { - clamped.f[i] = 127.99999f; - } else if (src->f[i] < -126.99999f) { - clamped.f[i] = -126.99999f; - } else { - clamped.f[i] = src->f[i]; - } - } - src = &clamped; -#endif - - dst->f[0] = powf( 2.0f, src->f[0] ); - dst->f[1] = powf( 2.0f, src->f[1] ); - dst->f[2] = powf( 2.0f, src->f[2] ); - dst->f[3] = powf( 2.0f, src->f[3] ); -#endif -} - -#if 0 -static void -micro_f2ut( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->u[0] = (uint) src->f[0]; - dst->u[1] = (uint) src->f[1]; - dst->u[2] = (uint) src->f[2]; - dst->u[3] = (uint) src->f[3]; -} -#endif - static void micro_float_clamp(union tgsi_exec_channel *dst, const union tgsi_exec_channel *src) @@ -656,71 +797,6 @@ micro_float_clamp(union tgsi_exec_channel *dst, } static void -micro_flr( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = floorf( src->f[0] ); - dst->f[1] = floorf( src->f[1] ); - dst->f[2] = floorf( src->f[2] ); - dst->f[3] = floorf( src->f[3] ); -} - -static void -micro_frc( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = src->f[0] - floorf( src->f[0] ); - dst->f[1] = src->f[1] - floorf( src->f[1] ); - dst->f[2] = src->f[2] - floorf( src->f[2] ); - dst->f[3] = src->f[3] - floorf( src->f[3] ); -} - -static void -micro_i2f( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) src->i[0]; - dst->f[1] = (float) src->i[1]; - dst->f[2] = (float) src->i[2]; - dst->f[3] = (float) src->i[3]; -} - -static void -micro_lg2( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ -#if FAST_MATH - dst->f[0] = util_fast_log2( src->f[0] ); - dst->f[1] = util_fast_log2( src->f[1] ); - dst->f[2] = util_fast_log2( src->f[2] ); - dst->f[3] = util_fast_log2( src->f[3] ); -#else - dst->f[0] = logf( src->f[0] ) * 1.442695f; - dst->f[1] = logf( src->f[1] ) * 1.442695f; - dst->f[2] = logf( src->f[2] ) * 1.442695f; - dst->f[3] = logf( src->f[3] ) * 1.442695f; -#endif -} - -static void -micro_le( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] <= src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] <= src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] <= src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] <= src1->f[3] ? src2->f[3] : src3->f[3]; -} - -static void micro_lt( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src0, @@ -734,38 +810,6 @@ micro_lt( dst->f[3] = src0->f[3] < src1->f[3] ? src2->f[3] : src3->f[3]; } -#if 0 -static void -micro_ilt( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->i[0] = src0->i[0] < src1->i[0] ? src2->i[0] : src3->i[0]; - dst->i[1] = src0->i[1] < src1->i[1] ? src2->i[1] : src3->i[1]; - dst->i[2] = src0->i[2] < src1->i[2] ? src2->i[2] : src3->i[2]; - dst->i[3] = src0->i[3] < src1->i[3] ? src2->i[3] : src3->i[3]; -} -#endif - -#if 0 -static void -micro_ult( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->u[0] = src0->u[0] < src1->u[0] ? src2->u[0] : src3->u[0]; - dst->u[1] = src0->u[1] < src1->u[1] ? src2->u[1] : src3->u[1]; - dst->u[2] = src0->u[2] < src1->u[2] ? src2->u[2] : src3->u[2]; - dst->u[3] = src0->u[3] < src1->u[3] ? src2->u[3] : src3->u[3]; -} -#endif - static void micro_max( union tgsi_exec_channel *dst, @@ -778,34 +822,6 @@ micro_max( dst->f[3] = src0->f[3] > src1->f[3] ? src0->f[3] : src1->f[3]; } -#if 0 -static void -micro_imax( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] > src1->i[0] ? src0->i[0] : src1->i[0]; - dst->i[1] = src0->i[1] > src1->i[1] ? src0->i[1] : src1->i[1]; - dst->i[2] = src0->i[2] > src1->i[2] ? src0->i[2] : src1->i[2]; - dst->i[3] = src0->i[3] > src1->i[3] ? src0->i[3] : src1->i[3]; -} -#endif - -#if 0 -static void -micro_umax( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] > src1->u[0] ? src0->u[0] : src1->u[0]; - dst->u[1] = src0->u[1] > src1->u[1] ? src0->u[1] : src1->u[1]; - dst->u[2] = src0->u[2] > src1->u[2] ? src0->u[2] : src1->u[2]; - dst->u[3] = src0->u[3] > src1->u[3] ? src0->u[3] : src1->u[3]; -} -#endif - static void micro_min( union tgsi_exec_channel *dst, @@ -818,48 +834,6 @@ micro_min( dst->f[3] = src0->f[3] < src1->f[3] ? src0->f[3] : src1->f[3]; } -#if 0 -static void -micro_imin( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] < src1->i[0] ? src0->i[0] : src1->i[0]; - dst->i[1] = src0->i[1] < src1->i[1] ? src0->i[1] : src1->i[1]; - dst->i[2] = src0->i[2] < src1->i[2] ? src0->i[2] : src1->i[2]; - dst->i[3] = src0->i[3] < src1->i[3] ? src0->i[3] : src1->i[3]; -} -#endif - -#if 0 -static void -micro_umin( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] < src1->u[0] ? src0->u[0] : src1->u[0]; - dst->u[1] = src0->u[1] < src1->u[1] ? src0->u[1] : src1->u[1]; - dst->u[2] = src0->u[2] < src1->u[2] ? src0->u[2] : src1->u[2]; - dst->u[3] = src0->u[3] < src1->u[3] ? src0->u[3] : src1->u[3]; -} -#endif - -#if 0 -static void -micro_umod( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] % src1->u[0]; - dst->u[1] = src0->u[1] % src1->u[1]; - dst->u[2] = src0->u[2] % src1->u[2]; - dst->u[3] = src0->u[3] % src1->u[3]; -} -#endif - static void micro_mul( union tgsi_exec_channel *dst, @@ -874,20 +848,6 @@ micro_mul( #if 0 static void -micro_imul( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] * src1->i[0]; - dst->i[1] = src0->i[1] * src1->i[1]; - dst->i[2] = src0->i[2] * src1->i[2]; - dst->i[3] = src0->i[3] * src1->i[3]; -} -#endif - -#if 0 -static void micro_imul64( union tgsi_exec_channel *dst0, union tgsi_exec_channel *dst1, @@ -951,42 +911,6 @@ micro_neg( dst->f[3] = -src->f[3]; } -#if 0 -static void -micro_ineg( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->i[0] = -src->i[0]; - dst->i[1] = -src->i[1]; - dst->i[2] = -src->i[2]; - dst->i[3] = -src->i[3]; -} -#endif - -static void -micro_not( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->u[0] = ~src->u[0]; - dst->u[1] = ~src->u[1]; - dst->u[2] = ~src->u[2]; - dst->u[3] = ~src->u[3]; -} - -static void -micro_or( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] | src1->u[0]; - dst->u[1] = src0->u[1] | src1->u[1]; - dst->u[2] = src0->u[2] | src1->u[2]; - dst->u[3] = src0->u[3] | src1->u[3]; -} - static void micro_pow( union tgsi_exec_channel *dst, @@ -1007,88 +931,6 @@ micro_pow( } static void -micro_rnd( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = floorf( src->f[0] + 0.5f ); - dst->f[1] = floorf( src->f[1] + 0.5f ); - dst->f[2] = floorf( src->f[2] + 0.5f ); - dst->f[3] = floorf( src->f[3] + 0.5f ); -} - -static void -micro_sgn( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = src->f[0] < 0.0f ? -1.0f : src->f[0] > 0.0f ? 1.0f : 0.0f; - dst->f[1] = src->f[1] < 0.0f ? -1.0f : src->f[1] > 0.0f ? 1.0f : 0.0f; - dst->f[2] = src->f[2] < 0.0f ? -1.0f : src->f[2] > 0.0f ? 1.0f : 0.0f; - dst->f[3] = src->f[3] < 0.0f ? -1.0f : src->f[3] > 0.0f ? 1.0f : 0.0f; -} - -static void -micro_shl( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] << src1->i[0]; - dst->i[1] = src0->i[1] << src1->i[1]; - dst->i[2] = src0->i[2] << src1->i[2]; - dst->i[3] = src0->i[3] << src1->i[3]; -} - -static void -micro_ishr( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] >> src1->i[0]; - dst->i[1] = src0->i[1] >> src1->i[1]; - dst->i[2] = src0->i[2] >> src1->i[2]; - dst->i[3] = src0->i[3] >> src1->i[3]; -} - -static void -micro_trunc( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0 ) -{ - dst->f[0] = (float) (int) src0->f[0]; - dst->f[1] = (float) (int) src0->f[1]; - dst->f[2] = (float) (int) src0->f[2]; - dst->f[3] = (float) (int) src0->f[3]; -} - -#if 0 -static void -micro_ushr( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] >> src1->u[0]; - dst->u[1] = src0->u[1] >> src1->u[1]; - dst->u[2] = src0->u[2] >> src1->u[2]; - dst->u[3] = src0->u[3] >> src1->u[3]; -} -#endif - -static void -micro_sin( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = sinf( src->f[0] ); - dst->f[1] = sinf( src->f[1] ); - dst->f[2] = sinf( src->f[2] ); - dst->f[3] = sinf( src->f[3] ); -} - -static void micro_sqrt( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { @@ -1110,31 +952,6 @@ micro_sub( dst->f[3] = src0->f[3] - src1->f[3]; } -#if 0 -static void -micro_u2f( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) src->u[0]; - dst->f[1] = (float) src->u[1]; - dst->f[2] = (float) src->u[2]; - dst->f[3] = (float) src->u[3]; -} -#endif - -static void -micro_xor( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] ^ src1->u[0]; - dst->u[1] = src0->u[1] ^ src1->u[1]; - dst->u[2] = src0->u[2] ^ src1->u[2]; - dst->u[3] = src0->u[3] ^ src1->u[3]; -} - static void fetch_src_file_channel( const struct tgsi_exec_machine *mach, @@ -1233,11 +1050,11 @@ fetch_src_file_channel( } static void -fetch_source( - const struct tgsi_exec_machine *mach, - union tgsi_exec_channel *chan, - const struct tgsi_full_src_register *reg, - const uint chan_index ) +fetch_source(const struct tgsi_exec_machine *mach, + union tgsi_exec_channel *chan, + const struct tgsi_full_src_register *reg, + const uint chan_index, + enum tgsi_exec_datatype src_datatype) { union tgsi_exec_channel index; uint swizzle; @@ -1286,10 +1103,10 @@ fetch_source( &indir_index ); /* add value of address register to the offset */ - index.i[0] += (int) indir_index.f[0]; - index.i[1] += (int) indir_index.f[1]; - index.i[2] += (int) indir_index.f[2]; - index.i[3] += (int) indir_index.f[3]; + index.i[0] += indir_index.i[0]; + index.i[1] += indir_index.i[1]; + index.i[2] += indir_index.i[2]; + index.i[3] += indir_index.i[3]; /* for disabled execution channels, zero-out the index to * avoid using a potential garbage value. @@ -1366,10 +1183,10 @@ fetch_source( &index2, &indir_index ); - index.i[0] += (int) indir_index.f[0]; - index.i[1] += (int) indir_index.f[1]; - index.i[2] += (int) indir_index.f[2]; - index.i[3] += (int) indir_index.f[3]; + index.i[0] += indir_index.i[0]; + index.i[1] += indir_index.i[1]; + index.i[2] += indir_index.i[2]; + index.i[3] += indir_index.i[3]; /* for disabled execution channels, zero-out the index to * avoid using a potential garbage value. @@ -1394,32 +1211,30 @@ fetch_source( &index, chan ); - switch (tgsi_util_get_full_src_register_sign_mode( reg, chan_index )) { - case TGSI_UTIL_SIGN_CLEAR: - micro_abs( chan, chan ); - break; - - case TGSI_UTIL_SIGN_SET: - micro_abs( chan, chan ); - micro_neg( chan, chan ); - break; - - case TGSI_UTIL_SIGN_TOGGLE: - micro_neg( chan, chan ); - break; + if (reg->Register.Absolute) { + if (src_datatype == TGSI_EXEC_DATA_FLOAT) { + micro_abs(chan, chan); + } else { + micro_iabs(chan, chan); + } + } - case TGSI_UTIL_SIGN_KEEP: - break; + if (reg->Register.Negate) { + if (src_datatype == TGSI_EXEC_DATA_FLOAT) { + micro_neg(chan, chan); + } else { + micro_ineg(chan, chan); + } } } static void -store_dest( - struct tgsi_exec_machine *mach, - const union tgsi_exec_channel *chan, - const struct tgsi_full_dst_register *reg, - const struct tgsi_full_instruction *inst, - uint chan_index ) +store_dest(struct tgsi_exec_machine *mach, + const union tgsi_exec_channel *chan, + const struct tgsi_full_dst_register *reg, + const struct tgsi_full_instruction *inst, + uint chan_index, + enum tgsi_exec_datatype dst_datatype) { uint i; union tgsi_exec_channel null; @@ -1428,9 +1243,9 @@ store_dest( int offset = 0; /* indirection offset */ int index; -#ifdef DEBUG - check_inf_or_nan(chan); -#endif + if (dst_datatype == TGSI_EXEC_DATA_FLOAT) { + CHECK_INF_OR_NAN(chan); + } /* There is an extra source register that indirectly subscripts * a register file. The direct index now becomes an offset @@ -1465,7 +1280,7 @@ store_dest( &indir_index ); /* save indirection offset */ - offset = (int) indir_index.f[0]; + offset = indir_index.i[0]; } switch (reg->Register.File) { @@ -1595,10 +1410,10 @@ store_dest( } #define FETCH(VAL,INDEX,CHAN)\ - fetch_source (mach, VAL, &inst->Src[INDEX], CHAN) + fetch_source(mach, VAL, &inst->Src[INDEX], CHAN, TGSI_EXEC_DATA_FLOAT) #define STORE(VAL,INDEX,CHAN)\ - store_dest (mach, VAL, &inst->Dst[INDEX], inst, CHAN ) + store_dest(mach, VAL, &inst->Dst[INDEX], inst, CHAN, TGSI_EXEC_DATA_FLOAT) /** @@ -1694,7 +1509,8 @@ fetch_texel( struct tgsi_sampler *sampler, const union tgsi_exec_channel *s, const union tgsi_exec_channel *t, const union tgsi_exec_channel *p, - float lodbias, /* XXX should be float[4] */ + const union tgsi_exec_channel *c0, + enum tgsi_sampler_control control, union tgsi_exec_channel *r, union tgsi_exec_channel *g, union tgsi_exec_channel *b, @@ -1703,7 +1519,7 @@ fetch_texel( struct tgsi_sampler *sampler, uint j; float rgba[NUM_CHANNELS][QUAD_SIZE]; - sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, rgba); + sampler->get_samples(sampler, s->f, t->f, p->f, c0->f, control, rgba); for (j = 0; j < 4; j++) { r->f[j] = rgba[0][j]; @@ -1714,102 +1530,95 @@ fetch_texel( struct tgsi_sampler *sampler, } +#define TEX_MODIFIER_NONE 0 +#define TEX_MODIFIER_PROJECTED 1 +#define TEX_MODIFIER_LOD_BIAS 2 +#define TEX_MODIFIER_EXPLICIT_LOD 3 + + static void exec_tex(struct tgsi_exec_machine *mach, const struct tgsi_full_instruction *inst, - boolean biasLod, - boolean projected) + uint modifier) { const uint unit = inst->Src[1].Register.Index; union tgsi_exec_channel r[4]; + const union tgsi_exec_channel *lod = &ZeroVec; + enum tgsi_sampler_control control; uint chan_index; - float lodBias; - /* debug_printf("Sampler %u unit %u\n", sampler, unit); */ + if (modifier != TEX_MODIFIER_NONE) { + FETCH(&r[3], 0, CHAN_W); + if (modifier != TEX_MODIFIER_PROJECTED) { + lod = &r[3]; + } + } + + if (modifier == TEX_MODIFIER_EXPLICIT_LOD) { + control = tgsi_sampler_lod_explicit; + } else { + control = tgsi_sampler_lod_bias; + } switch (inst->Texture.Texture) { case TGSI_TEXTURE_1D: case TGSI_TEXTURE_SHADOW1D: - FETCH(&r[0], 0, CHAN_X); - if (projected) { - FETCH(&r[1], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[1] ); - } - - if (biasLod) { - FETCH(&r[1], 0, CHAN_W); - lodBias = r[2].f[0]; + if (modifier == TEX_MODIFIER_PROJECTED) { + micro_div(&r[0], &r[0], &r[3]); } - else - lodBias = 0.0; fetch_texel(mach->Samplers[unit], - &r[0], &ZeroVec, &ZeroVec, lodBias, /* S, T, P, BIAS */ - &r[0], &r[1], &r[2], &r[3]); /* R, G, B, A */ + &r[0], &ZeroVec, &ZeroVec, lod, /* S, T, P, LOD */ + control, + &r[0], &r[1], &r[2], &r[3]); /* R, G, B, A */ break; case TGSI_TEXTURE_2D: case TGSI_TEXTURE_RECT: case TGSI_TEXTURE_SHADOW2D: case TGSI_TEXTURE_SHADOWRECT: - FETCH(&r[0], 0, CHAN_X); FETCH(&r[1], 0, CHAN_Y); FETCH(&r[2], 0, CHAN_Z); - if (projected) { - FETCH(&r[3], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[3] ); - micro_div( &r[1], &r[1], &r[3] ); - micro_div( &r[2], &r[2], &r[3] ); - } - - if (biasLod) { - FETCH(&r[3], 0, CHAN_W); - lodBias = r[3].f[0]; + if (modifier == TEX_MODIFIER_PROJECTED) { + micro_div(&r[0], &r[0], &r[3]); + micro_div(&r[1], &r[1], &r[3]); + micro_div(&r[2], &r[2], &r[3]); } - else - lodBias = 0.0; fetch_texel(mach->Samplers[unit], - &r[0], &r[1], &r[2], lodBias, /* inputs */ + &r[0], &r[1], &r[2], lod, /* S, T, P, LOD */ + control, &r[0], &r[1], &r[2], &r[3]); /* outputs */ break; case TGSI_TEXTURE_3D: case TGSI_TEXTURE_CUBE: - FETCH(&r[0], 0, CHAN_X); FETCH(&r[1], 0, CHAN_Y); FETCH(&r[2], 0, CHAN_Z); - if (projected) { - FETCH(&r[3], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[3] ); - micro_div( &r[1], &r[1], &r[3] ); - micro_div( &r[2], &r[2], &r[3] ); - } - - if (biasLod) { - FETCH(&r[3], 0, CHAN_W); - lodBias = r[3].f[0]; + if (modifier == TEX_MODIFIER_PROJECTED) { + micro_div(&r[0], &r[0], &r[3]); + micro_div(&r[1], &r[1], &r[3]); + micro_div(&r[2], &r[2], &r[3]); } - else - lodBias = 0.0; fetch_texel(mach->Samplers[unit], - &r[0], &r[1], &r[2], lodBias, + &r[0], &r[1], &r[2], lod, + control, &r[0], &r[1], &r[2], &r[3]); break; default: - assert (0); + assert(0); } - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[chan_index], 0, chan_index ); + FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { + STORE(&r[chan_index], 0, chan_index); } } @@ -1832,8 +1641,9 @@ exec_txd(struct tgsi_exec_machine *mach, FETCH(&r[0], 0, CHAN_X); fetch_texel(mach->Samplers[unit], - &r[0], &ZeroVec, &ZeroVec, 0.0f, /* S, T, P, BIAS */ - &r[0], &r[1], &r[2], &r[3]); /* R, G, B, A */ + &r[0], &ZeroVec, &ZeroVec, &ZeroVec, /* S, T, P, BIAS */ + tgsi_sampler_lod_bias, + &r[0], &r[1], &r[2], &r[3]); /* R, G, B, A */ break; case TGSI_TEXTURE_2D: @@ -1846,8 +1656,9 @@ exec_txd(struct tgsi_exec_machine *mach, FETCH(&r[2], 0, CHAN_Z); fetch_texel(mach->Samplers[unit], - &r[0], &r[1], &r[2], 0.0f, /* inputs */ - &r[0], &r[1], &r[2], &r[3]); /* outputs */ + &r[0], &r[1], &r[2], &ZeroVec, /* inputs */ + tgsi_sampler_lod_bias, + &r[0], &r[1], &r[2], &r[3]); /* outputs */ break; case TGSI_TEXTURE_3D: @@ -1858,7 +1669,8 @@ exec_txd(struct tgsi_exec_machine *mach, FETCH(&r[2], 0, CHAN_Z); fetch_texel(mach->Samplers[unit], - &r[0], &r[1], &r[2], 0.0f, + &r[0], &r[1], &r[2], &ZeroVec, + tgsi_sampler_lod_bias, &r[0], &r[1], &r[2], &r[3]); break; @@ -1955,7 +1767,7 @@ exec_declaration(struct tgsi_exec_machine *mach, if (decl->Semantic.Name == TGSI_SEMANTIC_POSITION) { assert(decl->Semantic.Index == 0); assert(first == last); - assert(mask = TGSI_WRITEMASK_XYZW); + assert(mask == TGSI_WRITEMASK_XYZW); mach->Inputs[first] = mach->QuadPos; } else if (decl->Semantic.Name == TGSI_SEMANTIC_FACE) { @@ -2001,6 +1813,585 @@ exec_declaration(struct tgsi_exec_machine *mach, } } +typedef void (* micro_op)(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src); + +static void +exec_scalar_unary(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst, + micro_op op, + enum tgsi_exec_datatype dst_datatype, + enum tgsi_exec_datatype src_datatype) +{ + unsigned int chan; + union tgsi_exec_channel src; + union tgsi_exec_channel dst; + + fetch_source(mach, &src, &inst->Src[0], CHAN_X, src_datatype); + op(&dst, &src); + for (chan = 0; chan < NUM_CHANNELS; chan++) { + if (inst->Dst[0].Register.WriteMask & (1 << chan)) { + store_dest(mach, &dst, &inst->Dst[0], inst, chan, dst_datatype); + } + } +} + +static void +exec_vector_unary(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst, + micro_op op, + enum tgsi_exec_datatype dst_datatype, + enum tgsi_exec_datatype src_datatype) +{ + unsigned int chan; + struct tgsi_exec_vector dst; + + for (chan = 0; chan < NUM_CHANNELS; chan++) { + if (inst->Dst[0].Register.WriteMask & (1 << chan)) { + union tgsi_exec_channel src; + + fetch_source(mach, &src, &inst->Src[0], chan, src_datatype); + op(&dst.xyzw[chan], &src); + } + } + for (chan = 0; chan < NUM_CHANNELS; chan++) { + if (inst->Dst[0].Register.WriteMask & (1 << chan)) { + store_dest(mach, &dst.xyzw[chan], &inst->Dst[0], inst, chan, dst_datatype); + } + } +} + +static void +exec_vector_binary(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst, + micro_op op, + enum tgsi_exec_datatype dst_datatype, + enum tgsi_exec_datatype src_datatype) +{ + unsigned int chan; + struct tgsi_exec_vector dst; + + for (chan = 0; chan < NUM_CHANNELS; chan++) { + if (inst->Dst[0].Register.WriteMask & (1 << chan)) { + union tgsi_exec_channel src[2]; + + fetch_source(mach, &src[0], &inst->Src[0], chan, src_datatype); + fetch_source(mach, &src[1], &inst->Src[1], chan, src_datatype); + op(&dst.xyzw[chan], src); + } + } + for (chan = 0; chan < NUM_CHANNELS; chan++) { + if (inst->Dst[0].Register.WriteMask & (1 << chan)) { + store_dest(mach, &dst.xyzw[chan], &inst->Dst[0], inst, chan, dst_datatype); + } + } +} + +static void +exec_vector_trinary(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst, + micro_op op, + enum tgsi_exec_datatype dst_datatype, + enum tgsi_exec_datatype src_datatype) +{ + unsigned int chan; + struct tgsi_exec_vector dst; + + for (chan = 0; chan < NUM_CHANNELS; chan++) { + if (inst->Dst[0].Register.WriteMask & (1 << chan)) { + union tgsi_exec_channel src[3]; + + fetch_source(mach, &src[0], &inst->Src[0], chan, src_datatype); + fetch_source(mach, &src[1], &inst->Src[1], chan, src_datatype); + fetch_source(mach, &src[2], &inst->Src[2], chan, src_datatype); + op(&dst.xyzw[chan], src); + } + } + for (chan = 0; chan < NUM_CHANNELS; chan++) { + if (inst->Dst[0].Register.WriteMask & (1 << chan)) { + store_dest(mach, &dst.xyzw[chan], &inst->Dst[0], inst, chan, dst_datatype); + } + } +} + +static void +exec_dp3(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst) +{ + unsigned int chan; + union tgsi_exec_channel arg[3]; + + fetch_source(mach, &arg[0], &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_FLOAT); + fetch_source(mach, &arg[1], &inst->Src[1], CHAN_X, TGSI_EXEC_DATA_FLOAT); + micro_mul(&arg[2], &arg[0], &arg[1]); + + for (chan = CHAN_Y; chan <= CHAN_Z; chan++) { + fetch_source(mach, &arg[0], &inst->Src[0], chan, TGSI_EXEC_DATA_FLOAT); + fetch_source(mach, &arg[1], &inst->Src[1], chan, TGSI_EXEC_DATA_FLOAT); + micro_mad(&arg[2], arg); + } + + for (chan = 0; chan < NUM_CHANNELS; chan++) { + if (inst->Dst[0].Register.WriteMask & (1 << chan)) { + store_dest(mach, &arg[2], &inst->Dst[0], inst, chan, TGSI_EXEC_DATA_FLOAT); + } + } +} + +static void +exec_dp4(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst) +{ + unsigned int chan; + union tgsi_exec_channel arg[3]; + + fetch_source(mach, &arg[0], &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_FLOAT); + fetch_source(mach, &arg[1], &inst->Src[1], CHAN_X, TGSI_EXEC_DATA_FLOAT); + micro_mul(&arg[2], &arg[0], &arg[1]); + + for (chan = CHAN_Y; chan <= CHAN_W; chan++) { + fetch_source(mach, &arg[0], &inst->Src[0], chan, TGSI_EXEC_DATA_FLOAT); + fetch_source(mach, &arg[1], &inst->Src[1], chan, TGSI_EXEC_DATA_FLOAT); + micro_mad(&arg[2], arg); + } + + for (chan = 0; chan < NUM_CHANNELS; chan++) { + if (inst->Dst[0].Register.WriteMask & (1 << chan)) { + store_dest(mach, &arg[2], &inst->Dst[0], inst, chan, TGSI_EXEC_DATA_FLOAT); + } + } +} + +static void +exec_dp2a(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst) +{ + unsigned int chan; + union tgsi_exec_channel arg[3]; + + fetch_source(mach, &arg[0], &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_FLOAT); + fetch_source(mach, &arg[1], &inst->Src[1], CHAN_X, TGSI_EXEC_DATA_FLOAT); + micro_mul(&arg[2], &arg[0], &arg[1]); + + fetch_source(mach, &arg[0], &inst->Src[0], CHAN_Y, TGSI_EXEC_DATA_FLOAT); + fetch_source(mach, &arg[1], &inst->Src[1], CHAN_Y, TGSI_EXEC_DATA_FLOAT); + micro_mad(&arg[0], arg); + + fetch_source(mach, &arg[1], &inst->Src[2], CHAN_X, TGSI_EXEC_DATA_FLOAT); + micro_add(&arg[0], &arg[0], &arg[1]); + + for (chan = 0; chan < NUM_CHANNELS; chan++) { + if (inst->Dst[0].Register.WriteMask & (1 << chan)) { + store_dest(mach, &arg[0], &inst->Dst[0], inst, chan, TGSI_EXEC_DATA_FLOAT); + } + } +} + +static void +exec_dph(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst) +{ + unsigned int chan; + union tgsi_exec_channel arg[3]; + + fetch_source(mach, &arg[0], &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_FLOAT); + fetch_source(mach, &arg[1], &inst->Src[1], CHAN_X, TGSI_EXEC_DATA_FLOAT); + micro_mul(&arg[2], &arg[0], &arg[1]); + + fetch_source(mach, &arg[0], &inst->Src[0], CHAN_Y, TGSI_EXEC_DATA_FLOAT); + fetch_source(mach, &arg[1], &inst->Src[1], CHAN_Y, TGSI_EXEC_DATA_FLOAT); + micro_mad(&arg[2], arg); + + fetch_source(mach, &arg[0], &inst->Src[0], CHAN_Z, TGSI_EXEC_DATA_FLOAT); + fetch_source(mach, &arg[1], &inst->Src[1], CHAN_Z, TGSI_EXEC_DATA_FLOAT); + micro_mad(&arg[0], arg); + + fetch_source(mach, &arg[1], &inst->Src[1], CHAN_W, TGSI_EXEC_DATA_FLOAT); + micro_add(&arg[0], &arg[0], &arg[1]); + + for (chan = 0; chan < NUM_CHANNELS; chan++) { + if (inst->Dst[0].Register.WriteMask & (1 << chan)) { + store_dest(mach, &arg[0], &inst->Dst[0], inst, chan, TGSI_EXEC_DATA_FLOAT); + } + } +} + +static void +exec_dp2(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst) +{ + unsigned int chan; + union tgsi_exec_channel arg[3]; + + fetch_source(mach, &arg[0], &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_FLOAT); + fetch_source(mach, &arg[1], &inst->Src[1], CHAN_X, TGSI_EXEC_DATA_FLOAT); + micro_mul(&arg[2], &arg[0], &arg[1]); + + fetch_source(mach, &arg[0], &inst->Src[0], CHAN_Y, TGSI_EXEC_DATA_FLOAT); + fetch_source(mach, &arg[1], &inst->Src[1], CHAN_Y, TGSI_EXEC_DATA_FLOAT); + micro_mad(&arg[2], arg); + + for (chan = 0; chan < NUM_CHANNELS; chan++) { + if (inst->Dst[0].Register.WriteMask & (1 << chan)) { + store_dest(mach, &arg[2], &inst->Dst[0], inst, chan, TGSI_EXEC_DATA_FLOAT); + } + } +} + +static void +exec_break(struct tgsi_exec_machine *mach) +{ + if (mach->BreakType == TGSI_EXEC_BREAK_INSIDE_LOOP) { + /* turn off loop channels for each enabled exec channel */ + mach->LoopMask &= ~mach->ExecMask; + /* Todo: if mach->LoopMask == 0, jump to end of loop */ + UPDATE_EXEC_MASK(mach); + } else { + assert(mach->BreakType == TGSI_EXEC_BREAK_INSIDE_SWITCH); + + mach->Switch.mask = 0x0; + + UPDATE_EXEC_MASK(mach); + } +} + +static void +exec_switch(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst) +{ + assert(mach->SwitchStackTop < TGSI_EXEC_MAX_SWITCH_NESTING); + assert(mach->BreakStackTop < TGSI_EXEC_MAX_BREAK_STACK); + + mach->SwitchStack[mach->SwitchStackTop++] = mach->Switch; + fetch_source(mach, &mach->Switch.selector, &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_UINT); + mach->Switch.mask = 0x0; + mach->Switch.defaultMask = 0x0; + + mach->BreakStack[mach->BreakStackTop++] = mach->BreakType; + mach->BreakType = TGSI_EXEC_BREAK_INSIDE_SWITCH; + + UPDATE_EXEC_MASK(mach); +} + +static void +exec_case(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst) +{ + uint prevMask = mach->SwitchStack[mach->SwitchStackTop - 1].mask; + union tgsi_exec_channel src; + uint mask = 0; + + fetch_source(mach, &src, &inst->Src[0], CHAN_X, TGSI_EXEC_DATA_UINT); + + if (mach->Switch.selector.u[0] == src.u[0]) { + mask |= 0x1; + } + if (mach->Switch.selector.u[1] == src.u[1]) { + mask |= 0x2; + } + if (mach->Switch.selector.u[2] == src.u[2]) { + mask |= 0x4; + } + if (mach->Switch.selector.u[3] == src.u[3]) { + mask |= 0x8; + } + + mach->Switch.defaultMask |= mask; + + mach->Switch.mask |= mask & prevMask; + + UPDATE_EXEC_MASK(mach); +} + +static void +exec_default(struct tgsi_exec_machine *mach) +{ + uint prevMask = mach->SwitchStack[mach->SwitchStackTop - 1].mask; + + mach->Switch.mask |= ~mach->Switch.defaultMask & prevMask; + + UPDATE_EXEC_MASK(mach); +} + +static void +exec_endswitch(struct tgsi_exec_machine *mach) +{ + mach->Switch = mach->SwitchStack[--mach->SwitchStackTop]; + mach->BreakType = mach->BreakStack[--mach->BreakStackTop]; + + UPDATE_EXEC_MASK(mach); +} + +static void +micro_i2f(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = (float)src->i[0]; + dst->f[1] = (float)src->i[1]; + dst->f[2] = (float)src->i[2]; + dst->f[3] = (float)src->i[3]; +} + +static void +micro_not(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = ~src->u[0]; + dst->u[1] = ~src->u[1]; + dst->u[2] = ~src->u[2]; + dst->u[3] = ~src->u[3]; +} + +static void +micro_shl(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = src[0].u[0] << src[1].u[0]; + dst->u[1] = src[0].u[1] << src[1].u[1]; + dst->u[2] = src[0].u[2] << src[1].u[2]; + dst->u[3] = src[0].u[3] << src[1].u[3]; +} + +static void +micro_and(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = src[0].u[0] & src[1].u[0]; + dst->u[1] = src[0].u[1] & src[1].u[1]; + dst->u[2] = src[0].u[2] & src[1].u[2]; + dst->u[3] = src[0].u[3] & src[1].u[3]; +} + +static void +micro_or(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = src[0].u[0] | src[1].u[0]; + dst->u[1] = src[0].u[1] | src[1].u[1]; + dst->u[2] = src[0].u[2] | src[1].u[2]; + dst->u[3] = src[0].u[3] | src[1].u[3]; +} + +static void +micro_xor(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = src[0].u[0] ^ src[1].u[0]; + dst->u[1] = src[0].u[1] ^ src[1].u[1]; + dst->u[2] = src[0].u[2] ^ src[1].u[2]; + dst->u[3] = src[0].u[3] ^ src[1].u[3]; +} + +static void +micro_f2i(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->i[0] = (int)src->f[0]; + dst->i[1] = (int)src->f[1]; + dst->i[2] = (int)src->f[2]; + dst->i[3] = (int)src->f[3]; +} + +static void +micro_idiv(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->i[0] = src[0].i[0] / src[1].i[0]; + dst->i[1] = src[0].i[1] / src[1].i[1]; + dst->i[2] = src[0].i[2] / src[1].i[2]; + dst->i[3] = src[0].i[3] / src[1].i[3]; +} + +static void +micro_imax(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->i[0] = src[0].i[0] > src[1].i[0] ? src[0].i[0] : src[1].i[0]; + dst->i[1] = src[0].i[1] > src[1].i[1] ? src[0].i[1] : src[1].i[1]; + dst->i[2] = src[0].i[2] > src[1].i[2] ? src[0].i[2] : src[1].i[2]; + dst->i[3] = src[0].i[3] > src[1].i[3] ? src[0].i[3] : src[1].i[3]; +} + +static void +micro_imin(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->i[0] = src[0].i[0] < src[1].i[0] ? src[0].i[0] : src[1].i[0]; + dst->i[1] = src[0].i[1] < src[1].i[1] ? src[0].i[1] : src[1].i[1]; + dst->i[2] = src[0].i[2] < src[1].i[2] ? src[0].i[2] : src[1].i[2]; + dst->i[3] = src[0].i[3] < src[1].i[3] ? src[0].i[3] : src[1].i[3]; +} + +static void +micro_isge(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->i[0] = src[0].i[0] >= src[1].i[0] ? -1 : 0; + dst->i[1] = src[0].i[1] >= src[1].i[1] ? -1 : 0; + dst->i[2] = src[0].i[2] >= src[1].i[2] ? -1 : 0; + dst->i[3] = src[0].i[3] >= src[1].i[3] ? -1 : 0; +} + +static void +micro_ishr(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->i[0] = src[0].i[0] >> src[1].i[0]; + dst->i[1] = src[0].i[1] >> src[1].i[1]; + dst->i[2] = src[0].i[2] >> src[1].i[2]; + dst->i[3] = src[0].i[3] >> src[1].i[3]; +} + +static void +micro_islt(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->i[0] = src[0].i[0] < src[1].i[0] ? -1 : 0; + dst->i[1] = src[0].i[1] < src[1].i[1] ? -1 : 0; + dst->i[2] = src[0].i[2] < src[1].i[2] ? -1 : 0; + dst->i[3] = src[0].i[3] < src[1].i[3] ? -1 : 0; +} + +static void +micro_f2u(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = (uint)src->f[0]; + dst->u[1] = (uint)src->f[1]; + dst->u[2] = (uint)src->f[2]; + dst->u[3] = (uint)src->f[3]; +} + +static void +micro_u2f(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = (float)src->u[0]; + dst->f[1] = (float)src->u[1]; + dst->f[2] = (float)src->u[2]; + dst->f[3] = (float)src->u[3]; +} + +static void +micro_uadd(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = src[0].u[0] + src[1].u[0]; + dst->u[1] = src[0].u[1] + src[1].u[1]; + dst->u[2] = src[0].u[2] + src[1].u[2]; + dst->u[3] = src[0].u[3] + src[1].u[3]; +} + +static void +micro_udiv(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = src[0].u[0] / src[1].u[0]; + dst->u[1] = src[0].u[1] / src[1].u[1]; + dst->u[2] = src[0].u[2] / src[1].u[2]; + dst->u[3] = src[0].u[3] / src[1].u[3]; +} + +static void +micro_umad(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = src[0].u[0] * src[1].u[0] + src[2].u[0]; + dst->u[1] = src[0].u[1] * src[1].u[1] + src[2].u[1]; + dst->u[2] = src[0].u[2] * src[1].u[2] + src[2].u[2]; + dst->u[3] = src[0].u[3] * src[1].u[3] + src[2].u[3]; +} + +static void +micro_umax(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = src[0].u[0] > src[1].u[0] ? src[0].u[0] : src[1].u[0]; + dst->u[1] = src[0].u[1] > src[1].u[1] ? src[0].u[1] : src[1].u[1]; + dst->u[2] = src[0].u[2] > src[1].u[2] ? src[0].u[2] : src[1].u[2]; + dst->u[3] = src[0].u[3] > src[1].u[3] ? src[0].u[3] : src[1].u[3]; +} + +static void +micro_umin(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = src[0].u[0] < src[1].u[0] ? src[0].u[0] : src[1].u[0]; + dst->u[1] = src[0].u[1] < src[1].u[1] ? src[0].u[1] : src[1].u[1]; + dst->u[2] = src[0].u[2] < src[1].u[2] ? src[0].u[2] : src[1].u[2]; + dst->u[3] = src[0].u[3] < src[1].u[3] ? src[0].u[3] : src[1].u[3]; +} + +static void +micro_umod(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = src[0].u[0] % src[1].u[0]; + dst->u[1] = src[0].u[1] % src[1].u[1]; + dst->u[2] = src[0].u[2] % src[1].u[2]; + dst->u[3] = src[0].u[3] % src[1].u[3]; +} + +static void +micro_umul(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = src[0].u[0] * src[1].u[0]; + dst->u[1] = src[0].u[1] * src[1].u[1]; + dst->u[2] = src[0].u[2] * src[1].u[2]; + dst->u[3] = src[0].u[3] * src[1].u[3]; +} + +static void +micro_useq(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = src[0].u[0] == src[1].u[0] ? ~0 : 0; + dst->u[1] = src[0].u[1] == src[1].u[1] ? ~0 : 0; + dst->u[2] = src[0].u[2] == src[1].u[2] ? ~0 : 0; + dst->u[3] = src[0].u[3] == src[1].u[3] ? ~0 : 0; +} + +static void +micro_usge(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = src[0].u[0] >= src[1].u[0] ? ~0 : 0; + dst->u[1] = src[0].u[1] >= src[1].u[1] ? ~0 : 0; + dst->u[2] = src[0].u[2] >= src[1].u[2] ? ~0 : 0; + dst->u[3] = src[0].u[3] >= src[1].u[3] ? ~0 : 0; +} + +static void +micro_ushr(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = src[0].u[0] >> src[1].u[0]; + dst->u[1] = src[0].u[1] >> src[1].u[1]; + dst->u[2] = src[0].u[2] >> src[1].u[2]; + dst->u[3] = src[0].u[3] >> src[1].u[3]; +} + +static void +micro_uslt(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = src[0].u[0] < src[1].u[0] ? ~0 : 0; + dst->u[1] = src[0].u[1] < src[1].u[1] ? ~0 : 0; + dst->u[2] = src[0].u[2] < src[1].u[2] ? ~0 : 0; + dst->u[3] = src[0].u[3] < src[1].u[3] ? ~0 : 0; +} + +static void +micro_usne(union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->u[0] = src[0].u[0] != src[1].u[0] ? ~0 : 0; + dst->u[1] = src[0].u[1] != src[1].u[1] ? ~0 : 0; + dst->u[2] = src[0].u[2] != src[1].u[2] ? ~0 : 0; + dst->u[3] = src[0].u[3] != src[1].u[3] ? ~0 : 0; +} + static void exec_instruction( struct tgsi_exec_machine *mach, @@ -2015,23 +2406,11 @@ exec_instruction( switch (inst->Instruction.Opcode) { case TGSI_OPCODE_ARL: - case TGSI_OPCODE_FLR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_flr(&d[chan_index], &r[0]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_unary(mach, inst, micro_arl, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_MOV: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&d[chan_index], 0, chan_index); - } - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_unary(mach, inst, micro_mov, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_LIT: @@ -2068,23 +2447,11 @@ exec_instruction( break; case TGSI_OPCODE_RCP: - /* TGSI_OPCODE_RECIP */ - FETCH( &r[0], 0, CHAN_X ); - micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } + exec_scalar_unary(mach, inst, micro_rcp, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_RSQ: - /* TGSI_OPCODE_RECIPSQRT */ - FETCH( &r[0], 0, CHAN_X ); - micro_abs( &r[0], &r[0] ); - micro_sqrt( &r[0], &r[0] ); - micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } + exec_scalar_unary(mach, inst, micro_rsq, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_EXP: @@ -2151,54 +2518,11 @@ exec_instruction( break; case TGSI_OPCODE_DP3: - /* TGSI_OPCODE_DOT3 */ - FETCH( &r[0], 0, CHAN_X ); - FETCH( &r[1], 1, CHAN_X ); - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH( &r[1], 0, CHAN_Y ); - FETCH( &r[2], 1, CHAN_Y ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH( &r[1], 0, CHAN_Z ); - FETCH( &r[2], 1, CHAN_Z ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } + exec_dp3(mach, inst); break; - case TGSI_OPCODE_DP4: - /* TGSI_OPCODE_DOT4 */ - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); - - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 1, CHAN_Y); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Z); - FETCH(&r[2], 1, CHAN_Z); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_W); - FETCH(&r[2], 1, CHAN_W); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } + case TGSI_OPCODE_DP4: + exec_dp4(mach, inst); break; case TGSI_OPCODE_DST: @@ -2255,41 +2579,15 @@ exec_instruction( break; case TGSI_OPCODE_SLT: - /* TGSI_OPCODE_SETLT */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_lt(&d[chan_index], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_binary(mach, inst, micro_slt, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_SGE: - /* TGSI_OPCODE_SETGE */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_le(&d[chan_index], &r[1], &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_binary(mach, inst, micro_sge, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_MAD: - /* TGSI_OPCODE_MADD */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_mul( &r[0], &r[0], &r[1] ); - FETCH( &r[1], 2, chan_index ); - micro_add(&d[chan_index], &r[0], &r[1]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_trinary(mach, inst, micro_mad, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_SUB: @@ -2304,17 +2602,7 @@ exec_instruction( break; case TGSI_OPCODE_LRP: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - FETCH(&r[2], 2, chan_index); - micro_sub( &r[1], &r[1], &r[2] ); - micro_mul( &r[0], &r[0], &r[1] ); - micro_add(&d[chan_index], &r[0], &r[2]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_trinary(mach, inst, micro_lrp, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_CND: @@ -2330,31 +2618,11 @@ exec_instruction( break; case TGSI_OPCODE_DP2A: - FETCH( &r[0], 0, CHAN_X ); - FETCH( &r[1], 1, CHAN_X ); - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH( &r[1], 0, CHAN_Y ); - FETCH( &r[2], 1, CHAN_Y ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH( &r[2], 2, CHAN_X ); - micro_add( &r[0], &r[0], &r[2] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } + exec_dp2a(mach, inst); break; case TGSI_OPCODE_FRC: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_frc(&d[chan_index], &r[0]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_unary(mach, inst, micro_frc, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_CLAMP: @@ -2370,33 +2638,20 @@ exec_instruction( } break; + case TGSI_OPCODE_FLR: + exec_vector_unary(mach, inst, micro_flr, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); + break; + case TGSI_OPCODE_ROUND: - case TGSI_OPCODE_ARR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_rnd(&d[chan_index], &r[0]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_unary(mach, inst, micro_rnd, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_EX2: - FETCH(&r[0], 0, CHAN_X); - - micro_exp2( &r[0], &r[0] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } + exec_scalar_unary(mach, inst, micro_exp2, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_LG2: - FETCH( &r[0], 0, CHAN_X ); - micro_lg2( &r[0], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } + exec_scalar_unary(mach, inst, micro_lg2, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_POW: @@ -2449,15 +2704,9 @@ exec_instruction( } break; - case TGSI_OPCODE_ABS: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - micro_abs(&d[chan_index], &r[0]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } - break; + case TGSI_OPCODE_ABS: + exec_vector_unary(mach, inst, micro_abs, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); + break; case TGSI_OPCODE_RCC: FETCH(&r[0], 0, CHAN_X); @@ -2469,60 +2718,19 @@ exec_instruction( break; case TGSI_OPCODE_DPH: - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); - - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 1, CHAN_Y); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Z); - FETCH(&r[2], 1, CHAN_Z); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 1, CHAN_W); - - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } + exec_dph(mach, inst); break; case TGSI_OPCODE_COS: - FETCH(&r[0], 0, CHAN_X); - - micro_cos( &r[0], &r[0] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } + exec_scalar_unary(mach, inst, micro_cos, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_DDX: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_ddx(&d[chan_index], &r[0]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_unary(mach, inst, micro_ddx, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_DDY: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_ddy(&d[chan_index], &r[0]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_unary(mach, inst, micro_ddy, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_KILP: @@ -2599,14 +2807,7 @@ exec_instruction( break; case TGSI_OPCODE_SEQ: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_eq(&d[chan_index], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_binary(mach, inst, micro_seq, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_SFL: @@ -2616,44 +2817,19 @@ exec_instruction( break; case TGSI_OPCODE_SGT: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_le(&d[chan_index], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_binary(mach, inst, micro_sgt, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_SIN: - FETCH( &r[0], 0, CHAN_X ); - micro_sin( &r[0], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } + exec_scalar_unary(mach, inst, micro_sin, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_SLE: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_le(&d[chan_index], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_binary(mach, inst, micro_sle, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_SNE: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_eq(&d[chan_index], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_binary(mach, inst, micro_sne, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_STR: @@ -2666,14 +2842,14 @@ exec_instruction( /* simple texture lookup */ /* src[0] = texcoord */ /* src[1] = sampler unit */ - exec_tex(mach, inst, FALSE, FALSE); + exec_tex(mach, inst, TEX_MODIFIER_NONE); break; case TGSI_OPCODE_TXB: /* Texture lookup with lod bias */ /* src[0] = texcoord (src[0].w = LOD bias) */ /* src[1] = sampler unit */ - exec_tex(mach, inst, TRUE, FALSE); + exec_tex(mach, inst, TEX_MODIFIER_LOD_BIAS); break; case TGSI_OPCODE_TXD: @@ -2689,14 +2865,14 @@ exec_instruction( /* Texture lookup with explit LOD */ /* src[0] = texcoord (src[0].w = LOD) */ /* src[1] = sampler unit */ - exec_tex(mach, inst, TRUE, FALSE); + exec_tex(mach, inst, TEX_MODIFIER_EXPLICIT_LOD); break; case TGSI_OPCODE_TXP: /* Texture lookup with projection */ /* src[0] = texcoord (src[0].w = projection) */ /* src[1] = sampler unit */ - exec_tex(mach, inst, FALSE, TRUE); + exec_tex(mach, inst, TEX_MODIFIER_PROJECTED); break; case TGSI_OPCODE_UP2H: @@ -2758,6 +2934,10 @@ exec_instruction( assert (0); break; + case TGSI_OPCODE_ARR: + exec_vector_unary(mach, inst, micro_arr, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_FLOAT); + break; + case TGSI_OPCODE_BRA: assert (0); break; @@ -2777,6 +2957,8 @@ exec_instruction( mach->CallStack[mach->CallStackTop].CondStackTop = mach->CondStackTop; mach->CallStack[mach->CallStackTop].LoopStackTop = mach->LoopStackTop; mach->CallStack[mach->CallStackTop].ContStackTop = mach->ContStackTop; + mach->CallStack[mach->CallStackTop].SwitchStackTop = mach->SwitchStackTop; + mach->CallStack[mach->CallStackTop].BreakStackTop = mach->BreakStackTop; /* note that PC was already incremented above */ mach->CallStack[mach->CallStackTop].ReturnAddr = *pc; @@ -2784,12 +2966,17 @@ exec_instruction( /* Second, push the Cond, Loop, Cont, Func stacks */ assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); - mach->CondStack[mach->CondStackTop++] = mach->CondMask; assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->ContStack[mach->ContStackTop++] = mach->ContMask; + assert(mach->SwitchStackTop < TGSI_EXEC_MAX_SWITCH_NESTING); + assert(mach->BreakStackTop < TGSI_EXEC_MAX_BREAK_STACK); assert(mach->FuncStackTop < TGSI_EXEC_MAX_CALL_NESTING); + + mach->CondStack[mach->CondStackTop++] = mach->CondMask; + mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; + mach->ContStack[mach->ContStackTop++] = mach->ContMask; + mach->SwitchStack[mach->SwitchStackTop++] = mach->Switch; + mach->BreakStack[mach->BreakStackTop++] = mach->BreakType; mach->FuncStack[mach->FuncStackTop++] = mach->FuncMask; /* Finally, jump to the subroutine */ @@ -2822,6 +3009,12 @@ exec_instruction( mach->ContStackTop = mach->CallStack[mach->CallStackTop].ContStackTop; mach->ContMask = mach->ContStack[mach->ContStackTop]; + mach->SwitchStackTop = mach->CallStack[mach->CallStackTop].SwitchStackTop; + mach->Switch = mach->SwitchStack[mach->SwitchStackTop]; + + mach->BreakStackTop = mach->CallStack[mach->CallStackTop].BreakStackTop; + mach->BreakType = mach->BreakStack[mach->BreakStackTop]; + assert(mach->FuncStackTop > 0); mach->FuncMask = mach->FuncStack[--mach->FuncStackTop]; @@ -2832,14 +3025,7 @@ exec_instruction( break; case TGSI_OPCODE_SSG: - /* TGSI_OPCODE_SGN */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_sgn(&d[chan_index], &r[0]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_unary(mach, inst, micro_sgn, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_CMP: @@ -2946,18 +3132,7 @@ exec_instruction( break; case TGSI_OPCODE_DP2: - FETCH( &r[0], 0, CHAN_X ); - FETCH( &r[1], 1, CHAN_X ); - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH( &r[1], 0, CHAN_Y ); - FETCH( &r[2], 1, CHAN_Y ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } + exec_dp2(mach, inst); break; case TGSI_OPCODE_IF: @@ -3023,87 +3198,31 @@ exec_instruction( break; case TGSI_OPCODE_CEIL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_ceil(&d[chan_index], &r[0]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_unary(mach, inst, micro_ceil, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_I2F: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_i2f(&d[chan_index], &r[0]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_unary(mach, inst, micro_i2f, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_INT); break; case TGSI_OPCODE_NOT: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_not(&d[chan_index], &r[0]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_unary(mach, inst, micro_not, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT); break; case TGSI_OPCODE_TRUNC: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_trunc(&d[chan_index], &r[0]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_unary(mach, inst, micro_trunc, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_FLOAT); break; case TGSI_OPCODE_SHL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_shl(&d[chan_index], &r[0], &r[1]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } - break; - - case TGSI_OPCODE_SHR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_ishr(&d[chan_index], &r[0], &r[1]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_binary(mach, inst, micro_shl, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT); break; case TGSI_OPCODE_AND: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_and(&d[chan_index], &r[0], &r[1]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_binary(mach, inst, micro_and, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT); break; case TGSI_OPCODE_OR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_or(&d[chan_index], &r[0], &r[1]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_binary(mach, inst, micro_or, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT); break; case TGSI_OPCODE_MOD: @@ -3111,14 +3230,7 @@ exec_instruction( break; case TGSI_OPCODE_XOR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_xor(&d[chan_index], &r[0], &r[1]); - } - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - STORE(&d[chan_index], 0, chan_index); - } + exec_vector_binary(mach, inst, micro_xor, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT); break; case TGSI_OPCODE_SAD: @@ -3167,11 +3279,15 @@ exec_instruction( case TGSI_OPCODE_BGNLOOP: /* push LoopMask and ContMasks */ assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->ContStack[mach->ContStackTop++] = mach->ContMask; assert(mach->LoopLabelStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + assert(mach->BreakStackTop < TGSI_EXEC_MAX_BREAK_STACK); + + mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; + mach->ContStack[mach->ContStackTop++] = mach->ContMask; mach->LoopLabelStack[mach->LoopLabelStackTop++] = *pc - 1; + mach->BreakStack[mach->BreakStackTop++] = mach->BreakType; + mach->BreakType = TGSI_EXEC_BREAK_INSIDE_LOOP; break; case TGSI_OPCODE_ENDFOR: @@ -3218,6 +3334,8 @@ exec_instruction( --mach->LoopLabelStackTop; assert(mach->LoopCounterStackTop > 0); --mach->LoopCounterStackTop; + + mach->BreakType = mach->BreakStack[--mach->BreakStackTop]; } UPDATE_EXEC_MASK(mach); break; @@ -3241,15 +3359,14 @@ exec_instruction( mach->ContMask = mach->ContStack[--mach->ContStackTop]; assert(mach->LoopLabelStackTop > 0); --mach->LoopLabelStackTop; + + mach->BreakType = mach->BreakStack[--mach->BreakStackTop]; } UPDATE_EXEC_MASK(mach); break; case TGSI_OPCODE_BRK: - /* turn off loop channels for each enabled exec channel */ - mach->LoopMask &= ~mach->ExecMask; - /* Todo: if mach->LoopMask == 0, jump to end of loop */ - UPDATE_EXEC_MASK(mach); + exec_break(mach); break; case TGSI_OPCODE_CONT: @@ -3280,6 +3397,12 @@ exec_instruction( mach->ContStackTop = mach->CallStack[mach->CallStackTop].ContStackTop; mach->ContMask = mach->ContStack[mach->ContStackTop]; + mach->SwitchStackTop = mach->CallStack[mach->CallStackTop].SwitchStackTop; + mach->Switch = mach->SwitchStack[mach->SwitchStackTop]; + + mach->BreakStackTop = mach->CallStack[mach->CallStackTop].BreakStackTop; + mach->BreakType = mach->BreakStack[mach->BreakStackTop]; + assert(mach->FuncStackTop > 0); mach->FuncMask = mach->FuncStack[--mach->FuncStackTop]; @@ -3310,11 +3433,116 @@ exec_instruction( UPDATE_EXEC_MASK(mach); break; + case TGSI_OPCODE_F2I: + exec_vector_unary(mach, inst, micro_f2i, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_FLOAT); + break; + + case TGSI_OPCODE_IDIV: + exec_vector_binary(mach, inst, micro_idiv, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_INT); + break; + + case TGSI_OPCODE_IMAX: + exec_vector_binary(mach, inst, micro_imax, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_INT); + break; + + case TGSI_OPCODE_IMIN: + exec_vector_binary(mach, inst, micro_imin, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_INT); + break; + + case TGSI_OPCODE_INEG: + exec_vector_unary(mach, inst, micro_ineg, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_INT); + break; + + case TGSI_OPCODE_ISGE: + exec_vector_binary(mach, inst, micro_isge, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_INT); + break; + + case TGSI_OPCODE_ISHR: + exec_vector_binary(mach, inst, micro_ishr, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_INT); + break; + + case TGSI_OPCODE_ISLT: + exec_vector_binary(mach, inst, micro_islt, TGSI_EXEC_DATA_INT, TGSI_EXEC_DATA_INT); + break; + + case TGSI_OPCODE_F2U: + exec_vector_unary(mach, inst, micro_f2u, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_FLOAT); + break; + + case TGSI_OPCODE_U2F: + exec_vector_unary(mach, inst, micro_u2f, TGSI_EXEC_DATA_FLOAT, TGSI_EXEC_DATA_UINT); + break; + + case TGSI_OPCODE_UADD: + exec_vector_binary(mach, inst, micro_uadd, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT); + break; + + case TGSI_OPCODE_UDIV: + exec_vector_binary(mach, inst, micro_udiv, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT); + break; + + case TGSI_OPCODE_UMAD: + exec_vector_trinary(mach, inst, micro_umad, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT); + break; + + case TGSI_OPCODE_UMAX: + exec_vector_binary(mach, inst, micro_umax, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT); + break; + + case TGSI_OPCODE_UMIN: + exec_vector_binary(mach, inst, micro_umin, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT); + break; + + case TGSI_OPCODE_UMOD: + exec_vector_binary(mach, inst, micro_umod, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT); + break; + + case TGSI_OPCODE_UMUL: + exec_vector_binary(mach, inst, micro_umul, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT); + break; + + case TGSI_OPCODE_USEQ: + exec_vector_binary(mach, inst, micro_useq, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT); + break; + + case TGSI_OPCODE_USGE: + exec_vector_binary(mach, inst, micro_usge, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT); + break; + + case TGSI_OPCODE_USHR: + exec_vector_binary(mach, inst, micro_ushr, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT); + break; + + case TGSI_OPCODE_USLT: + exec_vector_binary(mach, inst, micro_uslt, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT); + break; + + case TGSI_OPCODE_USNE: + exec_vector_binary(mach, inst, micro_usne, TGSI_EXEC_DATA_UINT, TGSI_EXEC_DATA_UINT); + break; + + case TGSI_OPCODE_SWITCH: + exec_switch(mach, inst); + break; + + case TGSI_OPCODE_CASE: + exec_case(mach, inst); + break; + + case TGSI_OPCODE_DEFAULT: + exec_default(mach); + break; + + case TGSI_OPCODE_ENDSWITCH: + exec_endswitch(mach); + break; + default: assert( 0 ); } } + #define DEBUG_EXECUTION 0 @@ -3334,9 +3562,13 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) mach->FuncMask = 0xf; mach->ExecMask = 0xf; + mach->Switch.mask = 0xf; + assert(mach->CondStackTop == 0); assert(mach->LoopStackTop == 0); assert(mach->ContStackTop == 0); + assert(mach->SwitchStackTop == 0); + assert(mach->BreakStackTop == 0); assert(mach->CallStackTop == 0); mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; @@ -3393,11 +3625,11 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) if (j > 0) { debug_printf(" "); } - debug_printf("(%6f, %6f, %6f, %6f)\n", - temps[i].xyzw[0].f[j], - temps[i].xyzw[1].f[j], - temps[i].xyzw[2].f[j], - temps[i].xyzw[3].f[j]); + debug_printf("(%6f %u, %6f %u, %6f %u, %6f %u)\n", + temps[i].xyzw[0].f[j], temps[i].xyzw[0].u[j], + temps[i].xyzw[1].f[j], temps[i].xyzw[1].u[j], + temps[i].xyzw[2].f[j], temps[i].xyzw[2].u[j], + temps[i].xyzw[3].f[j], temps[i].xyzw[3].u[j]); } } } @@ -3411,11 +3643,11 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) if (j > 0) { debug_printf(" "); } - debug_printf("{%6f, %6f, %6f, %6f}\n", - outputs[i].xyzw[0].f[j], - outputs[i].xyzw[1].f[j], - outputs[i].xyzw[2].f[j], - outputs[i].xyzw[3].f[j]); + debug_printf("(%6f %u, %6f %u, %6f %u, %6f %u)\n", + outputs[i].xyzw[0].f[j], outputs[i].xyzw[0].u[j], + outputs[i].xyzw[1].f[j], outputs[i].xyzw[1].u[j], + outputs[i].xyzw[2].f[j], outputs[i].xyzw[2].u[j], + outputs[i].xyzw[3].f[j], outputs[i].xyzw[3].u[j]); } } } @@ -3437,6 +3669,8 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) assert(mach->CondStackTop == 0); assert(mach->LoopStackTop == 0); assert(mach->ContStackTop == 0); + assert(mach->SwitchStackTop == 0); + assert(mach->BreakStackTop == 0); assert(mach->CallStackTop == 0); return ~mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0]; diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.h b/src/gallium/auxiliary/tgsi/tgsi_exec.h index afaf5c39c4..59e3b445cc 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.h +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.h @@ -2,6 +2,7 @@ * * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. + * Copyright 2009-2010 VMware, Inc. All rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the @@ -35,11 +36,13 @@ extern "C" { #endif + #define MAX_LABELS (4 * 1024) /**< basically, max instructions */ #define NUM_CHANNELS 4 /* R,G,B,A */ #define QUAD_SIZE 4 /* 4 pixel/quad */ + /** * Registers may be treated as float, signed int or unsigned int. */ @@ -69,6 +72,11 @@ struct tgsi_interp_coef float dady[NUM_CHANNELS]; }; +enum tgsi_sampler_control { + tgsi_sampler_lod_bias, + tgsi_sampler_lod_explicit +}; + /** * Information for sampling textures, which must be implemented * by code outside the TGSI executor. @@ -80,7 +88,8 @@ struct tgsi_sampler const float s[QUAD_SIZE], const float t[QUAD_SIZE], const float p[QUAD_SIZE], - float lodbias, + const float c0[QUAD_SIZE], + enum tgsi_sampler_control control, float rgba[NUM_CHANNELS][QUAD_SIZE]); }; @@ -179,6 +188,7 @@ struct tgsi_exec_labels #define TGSI_EXEC_MAX_COND_NESTING 32 #define TGSI_EXEC_MAX_LOOP_NESTING 32 +#define TGSI_EXEC_MAX_SWITCH_NESTING 32 #define TGSI_EXEC_MAX_CALL_NESTING 32 /* The maximum number of input attributes per vertex. For 2D @@ -206,9 +216,29 @@ struct tgsi_call_record uint CondStackTop; uint LoopStackTop; uint ContStackTop; + int SwitchStackTop; + int BreakStackTop; uint ReturnAddr; }; + +/* Switch-case block state. */ +struct tgsi_switch_record { + uint mask; /**< execution mask */ + union tgsi_exec_channel selector; /**< a value case statements are compared to */ + uint defaultMask; /**< non-execute mask for default case */ +}; + + +enum tgsi_break_type { + TGSI_EXEC_BREAK_INSIDE_LOOP, + TGSI_EXEC_BREAK_INSIDE_SWITCH +}; + + +#define TGSI_EXEC_MAX_BREAK_STACK (TGSI_EXEC_MAX_LOOP_NESTING + TGSI_EXEC_MAX_SWITCH_NESTING) + + /** * Run-time virtual machine state for executing TGSI shader. */ @@ -251,6 +281,12 @@ struct tgsi_exec_machine uint FuncMask; /**< For function calls */ uint ExecMask; /**< = CondMask & LoopMask */ + /* Current switch-case state. */ + struct tgsi_switch_record Switch; + + /* Current break type. */ + enum tgsi_break_type BreakType; + /** Condition mask stack (for nested conditionals) */ uint CondStack[TGSI_EXEC_MAX_COND_NESTING]; int CondStackTop; @@ -271,6 +307,13 @@ struct tgsi_exec_machine uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING]; int ContStackTop; + /** Switch case stack */ + struct tgsi_switch_record SwitchStack[TGSI_EXEC_MAX_SWITCH_NESTING]; + int SwitchStackTop; + + enum tgsi_break_type BreakStack[TGSI_EXEC_MAX_BREAK_STACK]; + int BreakStackTop; + /** Function execution mask stack (for executing subroutine code) */ uint FuncStack[TGSI_EXEC_MAX_CALL_NESTING]; int FuncStackTop; diff --git a/src/gallium/auxiliary/tgsi/tgsi_info.c b/src/gallium/auxiliary/tgsi/tgsi_info.c index be375cabb8..de0e09cdba 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_info.c +++ b/src/gallium/auxiliary/tgsi/tgsi_info.c @@ -119,7 +119,7 @@ static const struct tgsi_opcode_info opcode_info[TGSI_OPCODE_LAST] = { 1, 1, 0, 0, 0, 0, "NOT", TGSI_OPCODE_NOT }, { 1, 1, 0, 0, 0, 0, "TRUNC", TGSI_OPCODE_TRUNC }, { 1, 2, 0, 0, 0, 0, "SHL", TGSI_OPCODE_SHL }, - { 1, 2, 0, 0, 0, 0, "SHR", TGSI_OPCODE_SHR }, + { 0, 0, 0, 0, 0, 0, "", 88 }, /* removed */ { 1, 2, 0, 0, 0, 0, "AND", TGSI_OPCODE_AND }, { 1, 2, 0, 0, 0, 0, "OR", TGSI_OPCODE_OR }, { 1, 2, 0, 0, 0, 0, "MOD", TGSI_OPCODE_MOD }, @@ -149,7 +149,33 @@ static const struct tgsi_opcode_info opcode_info[TGSI_OPCODE_LAST] = { 0, 1, 0, 0, 0, 0, "BREAKC", TGSI_OPCODE_BREAKC }, { 0, 1, 0, 0, 0, 0, "KIL", TGSI_OPCODE_KIL }, { 0, 0, 0, 0, 0, 0, "END", TGSI_OPCODE_END }, - { 0, 0, 0, 0, 0, 0, "", 118 } /* removed */ + { 0, 0, 0, 0, 0, 0, "", 118 }, /* removed */ + { 1, 1, 0, 0, 0, 0, "F2I", TGSI_OPCODE_F2I }, + { 1, 2, 0, 0, 0, 0, "IDIV", TGSI_OPCODE_IDIV }, + { 1, 2, 0, 0, 0, 0, "IMAX", TGSI_OPCODE_IMAX }, + { 1, 2, 0, 0, 0, 0, "IMIN", TGSI_OPCODE_IMIN }, + { 1, 1, 0, 0, 0, 0, "INEG", TGSI_OPCODE_INEG }, + { 1, 2, 0, 0, 0, 0, "ISGE", TGSI_OPCODE_ISGE }, + { 1, 2, 0, 0, 0, 0, "ISHR", TGSI_OPCODE_ISHR }, + { 1, 2, 0, 0, 0, 0, "ISLT", TGSI_OPCODE_ISLT }, + { 1, 1, 0, 0, 0, 0, "F2U", TGSI_OPCODE_F2U }, + { 1, 1, 0, 0, 0, 0, "U2F", TGSI_OPCODE_U2F }, + { 1, 2, 0, 0, 0, 0, "UADD", TGSI_OPCODE_UADD }, + { 1, 2, 0, 0, 0, 0, "UDIV", TGSI_OPCODE_UDIV }, + { 1, 3, 0, 0, 0, 0, "UMAD", TGSI_OPCODE_UMAD }, + { 1, 2, 0, 0, 0, 0, "UMAX", TGSI_OPCODE_UMAX }, + { 1, 2, 0, 0, 0, 0, "UMIN", TGSI_OPCODE_UMIN }, + { 1, 2, 0, 0, 0, 0, "UMOD", TGSI_OPCODE_UMOD }, + { 1, 2, 0, 0, 0, 0, "UMUL", TGSI_OPCODE_UMUL }, + { 1, 2, 0, 0, 0, 0, "USEQ", TGSI_OPCODE_USEQ }, + { 1, 2, 0, 0, 0, 0, "USGE", TGSI_OPCODE_USGE }, + { 1, 2, 0, 0, 0, 0, "USHR", TGSI_OPCODE_USHR }, + { 1, 2, 0, 0, 0, 0, "USLT", TGSI_OPCODE_USLT }, + { 1, 2, 0, 0, 0, 0, "USNE", TGSI_OPCODE_USNE }, + { 0, 1, 0, 0, 0, 0, "SWITCH", TGSI_OPCODE_SWITCH }, + { 0, 1, 0, 0, 0, 0, "CASE", TGSI_OPCODE_CASE }, + { 0, 0, 0, 0, 0, 0, "DEFAULT", TGSI_OPCODE_DEFAULT }, + { 0, 0, 0, 0, 0, 0, "ENDSWITCH", TGSI_OPCODE_ENDSWITCH } }; const struct tgsi_opcode_info * diff --git a/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h b/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h index b34263da48..e4af15c156 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h +++ b/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h @@ -124,7 +124,6 @@ OP11(I2F) OP11(NOT) OP11(TRUNC) OP12(SHL) -OP12(SHR) OP12(AND) OP12(OR) OP12(MOD) @@ -146,6 +145,28 @@ OP01(IFC) OP01(BREAKC) OP01(KIL) OP00(END) +OP11(F2I) +OP12(IDIV) +OP12(IMAX) +OP12(IMIN) +OP11(INEG) +OP12(ISGE) +OP12(ISHR) +OP12(ISLT) +OP11(F2U) +OP11(U2F) +OP12(UADD) +OP12(UDIV) +OP13(UMAD) +OP12(UMAX) +OP12(UMIN) +OP12(UMOD) +OP12(UMUL) +OP12(USEQ) +OP12(USGE) +OP12(USHR) +OP12(USLT) +OP12(USNE) #undef OP00 diff --git a/src/gallium/auxiliary/tgsi/tgsi_parse.c b/src/gallium/auxiliary/tgsi/tgsi_parse.c index fa65ecb997..8c7062d850 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_parse.c +++ b/src/gallium/auxiliary/tgsi/tgsi_parse.c @@ -119,17 +119,29 @@ tgsi_parse_token( case TGSI_TOKEN_TYPE_IMMEDIATE: { struct tgsi_full_immediate *imm = &ctx->FullToken.FullImmediate; + uint imm_count; memset(imm, 0, sizeof *imm); copy_token(&imm->Immediate, &token); + imm_count = imm->Immediate.NrTokens - 1; + switch (imm->Immediate.DataType) { case TGSI_IMM_FLOAT32: - { - uint imm_count = imm->Immediate.NrTokens - 1; - for (i = 0; i < imm_count; i++) { - next_token(ctx, &imm->u[i]); - } + for (i = 0; i < imm_count; i++) { + next_token(ctx, &imm->u[i].Float); + } + break; + + case TGSI_IMM_UINT32: + for (i = 0; i < imm_count; i++) { + next_token(ctx, &imm->u[i].Uint); + } + break; + + case TGSI_IMM_INT32: + for (i = 0; i < imm_count; i++) { + next_token(ctx, &imm->u[i].Int); } break; diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/tgsi_sanity.c index 16b8ec6051..7f1c8e5dd6 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sanity.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.c @@ -195,7 +195,7 @@ is_any_register_declared( struct cso_hash_iter iter = cso_hash_first_node(ctx->regs_decl); - while (cso_hash_iter_is_null(iter)) { + while (!cso_hash_iter_is_null(iter)) { scan_register *reg = (scan_register *)cso_hash_iter_data(iter); if (reg->file == file) return TRUE; @@ -247,7 +247,7 @@ check_register_usage( boolean indirect_access ) { if (!check_file_name( ctx, reg->file )) { - free(reg); + FREE(reg); return FALSE; } @@ -261,21 +261,23 @@ check_register_usage( if (!is_ind_register_used(ctx, reg)) cso_hash_insert(ctx->regs_ind_used, reg->file, reg); else - free(reg); + FREE(reg); } else { if (!is_register_declared( ctx, reg )) { - if (reg->dimensions == 2) + if (reg->dimensions == 2) { report_error( ctx, "%s[%d][%d]: Undeclared %s register", file_names[reg->file], reg->indices[0], reg->indices[1], name ); - else + } + else { report_error( ctx, "%s[%d]: Undeclared %s register", file_names[reg->file], reg->indices[0], name ); } + } if (!is_register_used( ctx, reg )) cso_hash_insert(ctx->regs_used, scan_register_key(reg), reg); else - free(reg); + FREE(reg); } return TRUE; } @@ -333,15 +335,15 @@ iter_instruction( fill_scan_register1d(ind_reg, inst->Src[i].Indirect.File, inst->Src[i].Indirect.Index); + if (!(reg->file == TGSI_FILE_ADDRESS || reg->file == TGSI_FILE_LOOP) || + reg->indices[0] != 0) { + report_warning(ctx, "Indirect register neither ADDR[0] nor LOOP[0]"); + } check_register_usage( ctx, reg, "indirect", FALSE ); - if (!(reg->file == TGSI_FILE_ADDRESS || reg->file == TGSI_FILE_LOOP) || - reg->indices[0] != 0) { - report_warning(ctx, "Indirect register neither ADDR[0] nor LOOP[0]"); - } } } @@ -445,7 +447,9 @@ iter_immediate( /* Check data type validity. */ - if (imm->Immediate.DataType != TGSI_IMM_FLOAT32) { + if (imm->Immediate.DataType != TGSI_IMM_FLOAT32 && + imm->Immediate.DataType != TGSI_IMM_UINT32 && + imm->Immediate.DataType != TGSI_IMM_INT32) { report_error( ctx, "(%u): Invalid immediate data type", imm->Immediate.DataType ); return TRUE; } @@ -486,7 +490,7 @@ epilog( struct cso_hash_iter iter = cso_hash_first_node(ctx->regs_decl); - while (cso_hash_iter_is_null(iter)) { + while (!cso_hash_iter_is_null(iter)) { scan_register *reg = (scan_register *)cso_hash_iter_data(iter); if (!is_register_used(ctx, reg) && !is_ind_register_used(ctx, reg)) { report_warning( ctx, "%s[%u]: Register never used", @@ -511,7 +515,8 @@ regs_hash_destroy(struct cso_hash *hash) while (!cso_hash_iter_is_null(iter)) { scan_register *reg = (scan_register *)cso_hash_iter_data(iter); iter = cso_hash_erase(hash, iter); - free(reg); + assert(reg->file < TGSI_FILE_COUNT); + FREE(reg); } cso_hash_delete(hash); } diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c index d63c75dafb..a85cc4659e 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c @@ -2,6 +2,7 @@ * * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. + * Copyright 2009-2010 VMware, Inc. All rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the @@ -1418,13 +1419,13 @@ fetch_texel( struct tgsi_sampler **sampler, sampler, *sampler, store ); - debug_printf("lodbias %f\n", store[12]); - for (j = 0; j < 4; j++) - debug_printf("sample %d texcoord %f %f\n", + debug_printf("sample %d texcoord %f %f %f lodbias %f\n", j, store[0+j], - store[4+j]); + store[4+j], + store[8 + j], + store[12 + j]); #endif { @@ -1433,7 +1434,8 @@ fetch_texel( struct tgsi_sampler **sampler, &store[0], /* s */ &store[4], /* t */ &store[8], /* r */ - store[12], /* lodbias */ + &store[12], /* lodbias */ + tgsi_sampler_lod_bias, rgba); /* results */ memcpy( store, rgba, 16 * sizeof(float)); @@ -2144,40 +2146,50 @@ emit_instruction( break; case TGSI_OPCODE_XPD: + /* Note: we do all stores after all operands have been fetched + * to avoid src/dst register aliasing issues for an instruction + * such as: XPD TEMP[2].xyz, TEMP[0], TEMP[2]; + */ if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - FETCH( func, *inst, 1, 1, CHAN_Z ); - FETCH( func, *inst, 3, 0, CHAN_Z ); + FETCH( func, *inst, 1, 1, CHAN_Z ); /* xmm[1] = src[1].z */ + FETCH( func, *inst, 3, 0, CHAN_Z ); /* xmm[3] = src[0].z */ } if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - FETCH( func, *inst, 0, 0, CHAN_Y ); - FETCH( func, *inst, 4, 1, CHAN_Y ); + FETCH( func, *inst, 0, 0, CHAN_Y ); /* xmm[0] = src[0].y */ + FETCH( func, *inst, 4, 1, CHAN_Y ); /* xmm[4] = src[1].y */ } IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { - emit_MOV( func, 2, 0 ); - emit_mul( func, 2, 1 ); - emit_MOV( func, 5, 3 ); - emit_mul( func, 5, 4 ); - emit_sub( func, 2, 5 ); - STORE( func, *inst, 2, 0, CHAN_X ); + emit_MOV( func, 7, 0 ); /* xmm[7] = xmm[0] */ + emit_mul( func, 7, 1 ); /* xmm[7] = xmm[2] * xmm[1] */ + emit_MOV( func, 5, 3 ); /* xmm[5] = xmm[3] */ + emit_mul( func, 5, 4 ); /* xmm[5] = xmm[5] * xmm[4] */ + emit_sub( func, 7, 5 ); /* xmm[7] = xmm[2] - xmm[5] */ + /* store xmm[7] in dst.x below */ } if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - FETCH( func, *inst, 2, 1, CHAN_X ); - FETCH( func, *inst, 5, 0, CHAN_X ); + FETCH( func, *inst, 2, 1, CHAN_X ); /* xmm[2] = src[1].x */ + FETCH( func, *inst, 5, 0, CHAN_X ); /* xmm[5] = src[0].x */ } IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { - emit_mul( func, 3, 2 ); - emit_mul( func, 1, 5 ); - emit_sub( func, 3, 1 ); - STORE( func, *inst, 3, 0, CHAN_Y ); + emit_mul( func, 3, 2 ); /* xmm[3] = xmm[3] * xmm[2] */ + emit_mul( func, 1, 5 ); /* xmm[1] = xmm[1] * xmm[5] */ + emit_sub( func, 3, 1 ); /* xmm[3] = xmm[3] - xmm[1] */ + /* store xmm[3] in dst.y below */ } IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { - emit_mul( func, 5, 4 ); - emit_mul( func, 0, 2 ); - emit_sub( func, 5, 0 ); - STORE( func, *inst, 5, 0, CHAN_Z ); + emit_mul( func, 5, 4 ); /* xmm[5] = xmm[5] * xmm[4] */ + emit_mul( func, 0, 2 ); /* xmm[0] = xmm[0] * xmm[2] */ + emit_sub( func, 5, 0 ); /* xmm[5] = xmm[5] - xmm[0] */ + STORE( func, *inst, 5, 0, CHAN_Z ); /* dst.z = xmm[5] */ + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { + STORE( func, *inst, 7, 0, CHAN_X ); /* dst.x = xmm[7] */ + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { + STORE( func, *inst, 3, 0, CHAN_Y ); /* dst.y = xmm[3] */ } IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { emit_tempf( @@ -2506,7 +2518,7 @@ emit_instruction( break; case TGSI_OPCODE_TXL: - emit_tex( func, inst, TRUE, FALSE ); + return 0; break; case TGSI_OPCODE_TXP: @@ -2578,7 +2590,7 @@ emit_instruction( return 0; break; - case TGSI_OPCODE_SHR: + case TGSI_OPCODE_ISHR: return 0; break; diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.c b/src/gallium/auxiliary/tgsi/tgsi_ureg.c index 6a0af664dd..e64e2b731d 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_ureg.c +++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.c @@ -101,8 +101,13 @@ struct ureg_program unsigned nr_outputs; struct { - float v[4]; + union { + float f[4]; + unsigned u[4]; + int i[4]; + } value; unsigned nr; + unsigned type; } immediate[UREG_MAX_IMMEDIATE]; unsigned nr_immediates; @@ -486,22 +491,22 @@ struct ureg_src ureg_DECL_sampler( struct ureg_program *ureg, } - - -static int match_or_expand_immediate( const float *v, - unsigned nr, - float *v2, - unsigned *nr2, - unsigned *swizzle ) +static int +match_or_expand_immediate( const unsigned *v, + unsigned nr, + unsigned *v2, + unsigned *pnr2, + unsigned *swizzle ) { + unsigned nr2 = *pnr2; unsigned i, j; - + *swizzle = 0; for (i = 0; i < nr; i++) { boolean found = FALSE; - for (j = 0; j < *nr2 && !found; j++) { + for (j = 0; j < nr2 && !found; j++) { if (v[i] == v2[j]) { *swizzle |= j << (i * 2); found = TRUE; @@ -509,24 +514,28 @@ static int match_or_expand_immediate( const float *v, } if (!found) { - if (*nr2 >= 4) + if (nr2 >= 4) { return FALSE; + } - v2[*nr2] = v[i]; - *swizzle |= *nr2 << (i * 2); - (*nr2)++; + v2[nr2] = v[i]; + *swizzle |= nr2 << (i * 2); + nr2++; } } + /* Actually expand immediate only when fully succeeded. + */ + *pnr2 = nr2; return TRUE; } - - -struct ureg_src ureg_DECL_immediate( struct ureg_program *ureg, - const float *v, - unsigned nr ) +static struct ureg_src +decl_immediate( struct ureg_program *ureg, + const unsigned *v, + unsigned nr, + unsigned type ) { unsigned i, j; unsigned swizzle; @@ -536,38 +545,82 @@ struct ureg_src ureg_DECL_immediate( struct ureg_program *ureg, */ for (i = 0; i < ureg->nr_immediates; i++) { - if (match_or_expand_immediate( v, - nr, - ureg->immediate[i].v, - &ureg->immediate[i].nr, - &swizzle )) + if (ureg->immediate[i].type != type) { + continue; + } + if (match_or_expand_immediate(v, + nr, + ureg->immediate[i].value.u, + &ureg->immediate[i].nr, + &swizzle)) { goto out; + } } if (ureg->nr_immediates < UREG_MAX_IMMEDIATE) { i = ureg->nr_immediates++; - if (match_or_expand_immediate( v, - nr, - ureg->immediate[i].v, - &ureg->immediate[i].nr, - &swizzle )) + ureg->immediate[i].type = type; + if (match_or_expand_immediate(v, + nr, + ureg->immediate[i].value.u, + &ureg->immediate[i].nr, + &swizzle)) { goto out; + } } - set_bad( ureg ); + set_bad(ureg); out: /* Make sure that all referenced elements are from this immediate. * Has the effect of making size-one immediates into scalars. */ - for (j = nr; j < 4; j++) + for (j = nr; j < 4; j++) { swizzle |= (swizzle & 0x3) << (j * 2); + } + + return ureg_swizzle(ureg_src_register(TGSI_FILE_IMMEDIATE, i), + (swizzle >> 0) & 0x3, + (swizzle >> 2) & 0x3, + (swizzle >> 4) & 0x3, + (swizzle >> 6) & 0x3); +} + + +struct ureg_src +ureg_DECL_immediate( struct ureg_program *ureg, + const float *v, + unsigned nr ) +{ + union { + float f[4]; + unsigned u[4]; + } fu; + unsigned int i; + + for (i = 0; i < nr; i++) { + fu.f[i] = v[i]; + } + + return decl_immediate(ureg, fu.u, nr, TGSI_IMM_FLOAT32); +} + - return ureg_swizzle( ureg_src_register( TGSI_FILE_IMMEDIATE, i ), - (swizzle >> 0) & 0x3, - (swizzle >> 2) & 0x3, - (swizzle >> 4) & 0x3, - (swizzle >> 6) & 0x3); +struct ureg_src +ureg_DECL_immediate_uint( struct ureg_program *ureg, + const unsigned *v, + unsigned nr ) +{ + return decl_immediate(ureg, v, nr, TGSI_IMM_UINT32); +} + + +struct ureg_src +ureg_DECL_immediate_int( struct ureg_program *ureg, + const int *v, + unsigned nr ) +{ + return decl_immediate(ureg, (const unsigned *)v, nr, TGSI_IMM_INT32); } @@ -955,21 +1008,23 @@ static void emit_decl_range( struct ureg_program *ureg, out[1].decl_range.Last = first + count - 1; } -static void emit_immediate( struct ureg_program *ureg, - const float *v ) +static void +emit_immediate( struct ureg_program *ureg, + const unsigned *v, + unsigned type ) { union tgsi_any_token *out = get_tokens( ureg, DOMAIN_DECL, 5 ); out[0].value = 0; out[0].imm.Type = TGSI_TOKEN_TYPE_IMMEDIATE; out[0].imm.NrTokens = 5; - out[0].imm.DataType = TGSI_IMM_FLOAT32; + out[0].imm.DataType = type; out[0].imm.Padding = 0; - out[1].imm_data.Float = v[0]; - out[2].imm_data.Float = v[1]; - out[3].imm_data.Float = v[2]; - out[4].imm_data.Float = v[3]; + out[1].imm_data.Uint = v[0]; + out[2].imm_data.Uint = v[1]; + out[3].imm_data.Uint = v[2]; + out[4].imm_data.Uint = v[3]; } @@ -1055,7 +1110,8 @@ static void emit_decls( struct ureg_program *ureg ) for (i = 0; i < ureg->nr_immediates; i++) { emit_immediate( ureg, - ureg->immediate[i].v ); + ureg->immediate[i].value.u, + ureg->immediate[i].type ); } } diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.h b/src/gallium/auxiliary/tgsi/tgsi_ureg.h index 7e3e7bcf1d..6f11273320 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_ureg.h +++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.h @@ -148,6 +148,16 @@ ureg_DECL_immediate( struct ureg_program *, unsigned nr ); struct ureg_src +ureg_DECL_immediate_uint( struct ureg_program *, + const unsigned *v, + unsigned nr ); + +struct ureg_src +ureg_DECL_immediate_int( struct ureg_program *, + const int *v, + unsigned nr ); + +struct ureg_src ureg_DECL_constant( struct ureg_program *, unsigned index ); @@ -221,6 +231,90 @@ ureg_imm1f( struct ureg_program *ureg, return ureg_DECL_immediate( ureg, v, 1 ); } +static INLINE struct ureg_src +ureg_imm4u( struct ureg_program *ureg, + unsigned a, unsigned b, + unsigned c, unsigned d) +{ + unsigned v[4]; + v[0] = a; + v[1] = b; + v[2] = c; + v[3] = d; + return ureg_DECL_immediate_uint( ureg, v, 4 ); +} + +static INLINE struct ureg_src +ureg_imm3u( struct ureg_program *ureg, + unsigned a, unsigned b, + unsigned c) +{ + unsigned v[3]; + v[0] = a; + v[1] = b; + v[2] = c; + return ureg_DECL_immediate_uint( ureg, v, 3 ); +} + +static INLINE struct ureg_src +ureg_imm2u( struct ureg_program *ureg, + unsigned a, unsigned b) +{ + unsigned v[2]; + v[0] = a; + v[1] = b; + return ureg_DECL_immediate_uint( ureg, v, 2 ); +} + +static INLINE struct ureg_src +ureg_imm1u( struct ureg_program *ureg, + unsigned a) +{ + return ureg_DECL_immediate_uint( ureg, &a, 1 ); +} + +static INLINE struct ureg_src +ureg_imm4i( struct ureg_program *ureg, + int a, int b, + int c, int d) +{ + int v[4]; + v[0] = a; + v[1] = b; + v[2] = c; + v[3] = d; + return ureg_DECL_immediate_int( ureg, v, 4 ); +} + +static INLINE struct ureg_src +ureg_imm3i( struct ureg_program *ureg, + int a, int b, + int c) +{ + int v[3]; + v[0] = a; + v[1] = b; + v[2] = c; + return ureg_DECL_immediate_int( ureg, v, 3 ); +} + +static INLINE struct ureg_src +ureg_imm2i( struct ureg_program *ureg, + int a, int b) +{ + int v[2]; + v[0] = a; + v[1] = b; + return ureg_DECL_immediate_int( ureg, v, 2 ); +} + +static INLINE struct ureg_src +ureg_imm1i( struct ureg_program *ureg, + int a) +{ + return ureg_DECL_immediate_int( ureg, &a, 1 ); +} + /*********************************************************************** * Functions for patching up labels */ diff --git a/src/gallium/auxiliary/util/u_bitmask.c b/src/gallium/auxiliary/util/u_bitmask.c index 77587c07ec..23c93a3ebc 100644 --- a/src/gallium/auxiliary/util/u_bitmask.c +++ b/src/gallium/auxiliary/util/u_bitmask.c @@ -97,12 +97,12 @@ util_bitmask_resize(struct util_bitmask *bm, if(!minimum_size) return FALSE; - if(bm->size > minimum_size) + if(bm->size >= minimum_size) return TRUE; assert(bm->size % UTIL_BITMASK_BITS_PER_WORD == 0); new_size = bm->size; - while(!(new_size > minimum_size)) { + while(new_size < minimum_size) { new_size *= 2; /* Check integer overflow */ if(new_size < bm->size) @@ -136,7 +136,7 @@ util_bitmask_filled_set(struct util_bitmask *bm, unsigned index) { assert(bm->filled <= bm->size); - assert(index <= bm->size); + assert(index < bm->size); if(index == bm->filled) { ++bm->filled; @@ -149,7 +149,7 @@ util_bitmask_filled_unset(struct util_bitmask *bm, unsigned index) { assert(bm->filled <= bm->size); - assert(index <= bm->size); + assert(index < bm->size); if(index < bm->filled) bm->filled = index; @@ -182,7 +182,7 @@ util_bitmask_add(struct util_bitmask *bm) mask = 1; } found: - + /* grow the bitmask if necessary */ if(!util_bitmask_resize(bm, bm->filled)) return UTIL_BITMASK_INVALID_INDEX; @@ -198,9 +198,9 @@ unsigned util_bitmask_set(struct util_bitmask *bm, unsigned index) { - unsigned word = index / UTIL_BITMASK_BITS_PER_WORD; - unsigned bit = index % UTIL_BITMASK_BITS_PER_WORD; - util_bitmask_word mask = 1 << bit; + unsigned word; + unsigned bit; + util_bitmask_word mask; assert(bm); @@ -208,6 +208,10 @@ util_bitmask_set(struct util_bitmask *bm, if(!util_bitmask_resize(bm, index)) return UTIL_BITMASK_INVALID_INDEX; + word = index / UTIL_BITMASK_BITS_PER_WORD; + bit = index % UTIL_BITMASK_BITS_PER_WORD; + mask = 1 << bit; + bm->words[word] |= mask; util_bitmask_filled_set(bm, index); @@ -220,15 +224,19 @@ void util_bitmask_clear(struct util_bitmask *bm, unsigned index) { - unsigned word = index / UTIL_BITMASK_BITS_PER_WORD; - unsigned bit = index % UTIL_BITMASK_BITS_PER_WORD; - util_bitmask_word mask = 1 << bit; + unsigned word; + unsigned bit; + util_bitmask_word mask; assert(bm); if(index >= bm->size) return; + word = index / UTIL_BITMASK_BITS_PER_WORD; + bit = index % UTIL_BITMASK_BITS_PER_WORD; + mask = 1 << bit; + bm->words[word] &= ~mask; util_bitmask_filled_unset(bm, index); @@ -250,7 +258,7 @@ util_bitmask_get(struct util_bitmask *bm, return TRUE; } - if(index > bm->size) + if(index >= bm->size) return FALSE; if(bm->words[word] & mask) { diff --git a/src/gallium/auxiliary/util/u_blitter.c b/src/gallium/auxiliary/util/u_blitter.c index 1f794d39a1..46b4706b76 100644 --- a/src/gallium/auxiliary/util/u_blitter.c +++ b/src/gallium/auxiliary/util/u_blitter.c @@ -48,6 +48,8 @@ #include "util/u_simple_shaders.h" #include "util/u_texture.h" +#define INVALID_PTR ((void*)~0) + struct blitter_context_priv { struct blitter_context blitter; @@ -110,6 +112,11 @@ struct blitter_context *util_blitter_create(struct pipe_context *pipe) ctx->pipe = pipe; /* init state objects for them to be considered invalid */ + ctx->blitter.saved_blend_state = INVALID_PTR; + ctx->blitter.saved_dsa_state = INVALID_PTR; + ctx->blitter.saved_rs_state = INVALID_PTR; + ctx->blitter.saved_fs = INVALID_PTR; + ctx->blitter.saved_vs = INVALID_PTR; ctx->blitter.saved_fb_state.nr_cbufs = ~0; ctx->blitter.saved_num_textures = ~0; ctx->blitter.saved_num_sampler_states = ~0; @@ -156,6 +163,7 @@ struct blitter_context *util_blitter_create(struct pipe_context *pipe) rs_state.cull_mode = PIPE_WINDING_NONE; rs_state.bypass_vs_clip_and_viewport = 1; rs_state.gl_rasterization_rules = 1; + rs_state.flatshade = 1; ctx->rs_state = pipe->create_rasterizer_state(pipe, &rs_state); /* fragment shaders are created on-demand */ @@ -234,11 +242,11 @@ void util_blitter_destroy(struct blitter_context *blitter) static void blitter_check_saved_CSOs(struct blitter_context_priv *ctx) { /* make sure these CSOs have been saved */ - assert(ctx->blitter.saved_blend_state && - ctx->blitter.saved_dsa_state && - ctx->blitter.saved_rs_state && - ctx->blitter.saved_fs && - ctx->blitter.saved_vs); + assert(ctx->blitter.saved_blend_state != INVALID_PTR && + ctx->blitter.saved_dsa_state != INVALID_PTR && + ctx->blitter.saved_rs_state != INVALID_PTR && + ctx->blitter.saved_fs != INVALID_PTR && + ctx->blitter.saved_vs != INVALID_PTR); } static void blitter_restore_CSOs(struct blitter_context_priv *ctx) @@ -252,11 +260,11 @@ static void blitter_restore_CSOs(struct blitter_context_priv *ctx) pipe->bind_fs_state(pipe, ctx->blitter.saved_fs); pipe->bind_vs_state(pipe, ctx->blitter.saved_vs); - ctx->blitter.saved_blend_state = 0; - ctx->blitter.saved_dsa_state = 0; - ctx->blitter.saved_rs_state = 0; - ctx->blitter.saved_fs = 0; - ctx->blitter.saved_vs = 0; + ctx->blitter.saved_blend_state = INVALID_PTR; + ctx->blitter.saved_dsa_state = INVALID_PTR; + ctx->blitter.saved_rs_state = INVALID_PTR; + ctx->blitter.saved_fs = INVALID_PTR; + ctx->blitter.saved_vs = INVALID_PTR; /* restore the state objects which are required to be saved before copy/fill */ @@ -560,45 +568,29 @@ void util_blitter_clear(struct blitter_context *blitter, blitter_restore_CSOs(ctx); } -void util_blitter_copy(struct blitter_context *blitter, - struct pipe_surface *dst, - unsigned dstx, unsigned dsty, - struct pipe_surface *src, - unsigned srcx, unsigned srcy, - unsigned width, unsigned height, - boolean ignore_stencil) +static boolean +is_overlap(unsigned sx1, unsigned sx2, unsigned sy1, unsigned sy2, + unsigned dx1, unsigned dx2, unsigned dy1, unsigned dy2) +{ + if (sx1 >= dx2 || sx2 <= dx1 || sy1 >= dy2 || sy2 <= dy1) { + return FALSE; + } else { + return TRUE; + } +} + +static void util_blitter_do_copy(struct blitter_context *blitter, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height, + boolean is_depth) { struct blitter_context_priv *ctx = (struct blitter_context_priv*)blitter; struct pipe_context *pipe = ctx->pipe; - struct pipe_screen *screen = pipe->screen; struct pipe_framebuffer_state fb_state; - boolean is_stencil, is_depth; - unsigned dst_tex_usage; - - /* give up if textures are not set */ - assert(dst->texture && src->texture); - if (!dst->texture || !src->texture) - return; - is_depth = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 0) != 0; - is_stencil = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 1) != 0; - dst_tex_usage = is_depth || is_stencil ? PIPE_TEXTURE_USAGE_DEPTH_STENCIL : - PIPE_TEXTURE_USAGE_RENDER_TARGET; - - /* check if we can sample from and render to the surfaces */ - /* (assuming copying a stencil buffer is not possible) */ - if ((!ignore_stencil && is_stencil) || - !screen->is_format_supported(screen, dst->format, dst->texture->target, - dst_tex_usage, 0) || - !screen->is_format_supported(screen, src->format, src->texture->target, - PIPE_TEXTURE_USAGE_SAMPLER, 0)) { - util_surface_copy(pipe, FALSE, dst, dstx, dsty, src, srcx, srcy, - width, height); - return; - } - - /* check whether the states are properly saved */ - blitter_check_saved_CSOs(ctx); assert(blitter->saved_fb_state.nr_cbufs != ~0); assert(blitter->saved_num_textures != ~0); assert(blitter->saved_num_sampler_states != ~0); @@ -656,6 +648,108 @@ void util_blitter_copy(struct blitter_context *blitter, blitter_set_rectangle(ctx, dstx, dsty, dstx+width, dsty+height, 0); blitter_draw_quad(ctx); + +} + +static void util_blitter_overlap_copy(struct blitter_context *blitter, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height) +{ + struct blitter_context_priv *ctx = (struct blitter_context_priv*)blitter; + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + + struct pipe_texture texTemp; + struct pipe_texture *texture; + struct pipe_surface *tex_surf; + + /* check whether the states are properly saved */ + blitter_check_saved_CSOs(ctx); + + memset(&texTemp, 0, sizeof(texTemp)); + texTemp.target = PIPE_TEXTURE_2D; + texTemp.format = dst->texture->format; /* XXX verify supported by driver! */ + texTemp.last_level = 0; + texTemp.width0 = width; + texTemp.height0 = height; + texTemp.depth0 = 1; + + texture = screen->texture_create(screen, &texTemp); + if (!texture) + return; + + tex_surf = screen->get_tex_surface(screen, texture, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE); + + /* blit from the src to the temp */ + util_blitter_do_copy(blitter, tex_surf, 0, 0, + src, srcx, srcy, + width, height, + FALSE); + util_blitter_do_copy(blitter, dst, dstx, dsty, + tex_surf, 0, 0, + width, height, + FALSE); + pipe_surface_reference(&tex_surf, NULL); + pipe_texture_reference(&texture, NULL); + blitter_restore_CSOs(ctx); +} + +void util_blitter_copy(struct blitter_context *blitter, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height, + boolean ignore_stencil) +{ + struct blitter_context_priv *ctx = (struct blitter_context_priv*)blitter; + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + boolean is_stencil, is_depth; + unsigned dst_tex_usage; + + /* give up if textures are not set */ + assert(dst->texture && src->texture); + if (!dst->texture || !src->texture) + return; + + if (dst->texture == src->texture) { + if (is_overlap(srcx, srcx + width, srcy, srcy + height, + dstx, dstx + width, dsty, dsty + height)) { + util_blitter_overlap_copy(blitter, dst, dstx, dsty, src, srcx, srcy, + width, height); + return; + } + } + + is_depth = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 0) != 0; + is_stencil = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 1) != 0; + dst_tex_usage = is_depth || is_stencil ? PIPE_TEXTURE_USAGE_DEPTH_STENCIL : + PIPE_TEXTURE_USAGE_RENDER_TARGET; + + /* check if we can sample from and render to the surfaces */ + /* (assuming copying a stencil buffer is not possible) */ + if ((!ignore_stencil && is_stencil) || + !screen->is_format_supported(screen, dst->format, dst->texture->target, + dst_tex_usage, 0) || + !screen->is_format_supported(screen, src->format, src->texture->target, + PIPE_TEXTURE_USAGE_SAMPLER, 0)) { + util_surface_copy(pipe, FALSE, dst, dstx, dsty, src, srcx, srcy, + width, height); + return; + } + + /* check whether the states are properly saved */ + blitter_check_saved_CSOs(ctx); + util_blitter_do_copy(blitter, + dst, dstx, dsty, + src, srcx, srcy, + width, height, is_depth); blitter_restore_CSOs(ctx); } diff --git a/src/gallium/auxiliary/util/u_debug_dump.c b/src/gallium/auxiliary/util/u_debug_dump.c index 09866880ae..61624d05c0 100644 --- a/src/gallium/auxiliary/util/u_debug_dump.c +++ b/src/gallium/auxiliary/util/u_debug_dump.c @@ -255,15 +255,13 @@ DEFINE_DEBUG_DUMP_CONTINUOUS(tex_mipfilter) static const char * debug_dump_tex_filter_names[] = { "PIPE_TEX_FILTER_NEAREST", - "PIPE_TEX_FILTER_LINEAR", - "PIPE_TEX_FILTER_ANISO" + "PIPE_TEX_FILTER_LINEAR" }; static const char * debug_dump_tex_filter_short_names[] = { "nearest", - "linear", - "aniso" + "linear" }; DEFINE_DEBUG_DUMP_CONTINUOUS(tex_filter) diff --git a/src/gallium/auxiliary/util/u_debug_memory.c b/src/gallium/auxiliary/util/u_debug_memory.c index 7623cb9398..d6484f4ad5 100644 --- a/src/gallium/auxiliary/util/u_debug_memory.c +++ b/src/gallium/auxiliary/util/u_debug_memory.c @@ -297,9 +297,9 @@ debug_memory_end(unsigned long start_no) if((start_no <= hdr->no && hdr->no < last_no) || (last_no < start_no && (hdr->no < last_no || start_no <= hdr->no))) { - debug_printf("%s:%u:%s: %u bytes at %p not freed\n", + debug_printf("%s:%u:%s: %lu bytes at %p not freed\n", hdr->file, hdr->line, hdr->function, - hdr->size, ptr); + (unsigned long) hdr->size, ptr); #if DEBUG_MEMORY_STACK debug_backtrace_dump(hdr->backtrace, DEBUG_MEMORY_STACK); #endif @@ -315,8 +315,8 @@ debug_memory_end(unsigned long start_no) } if(total_size) { - debug_printf("Total of %u KB of system memory apparently leaked\n", - (total_size + 1023)/1024); + debug_printf("Total of %lu KB of system memory apparently leaked\n", + (unsigned long) (total_size + 1023)/1024); } else { debug_printf("No memory leaks detected.\n"); diff --git a/src/gallium/auxiliary/util/u_format.csv b/src/gallium/auxiliary/util/u_format.csv index 866b18ff16..9f16b42944 100644 --- a/src/gallium/auxiliary/util/u_format.csv +++ b/src/gallium/auxiliary/util/u_format.csv @@ -76,9 +76,9 @@ PIPE_FORMAT_R8G8_SNORM , array , 1, 1, sn8 , sn8 , , , xy01, PIPE_FORMAT_R8G8B8_SNORM , array , 1, 1, sn8 , sn8 , sn8 , , xyz1, rgb PIPE_FORMAT_R8G8B8A8_SNORM , array , 1, 1, sn8 , sn8 , sn8 , sn8 , xyzw, rgb PIPE_FORMAT_R8G8B8X8_SNORM , array , 1, 1, sn8 , sn8 , sn8 , sn8 , xyz1, rgb -PIPE_FORMAT_B6G5R5_SNORM , arith , 1, 1, sn5 , sn5 , sn6 , , zyx1, rgb -PIPE_FORMAT_A8B8G8R8_SNORM , arith , 1, 1, sn8 , sn8 , sn8 , sn8 , zyxw, rgb -PIPE_FORMAT_X8B8G8R8_SNORM , arith , 1, 1, sn8 , sn8 , sn8 , sn8 , zyx1, rgb +PIPE_FORMAT_B6G5R5_SNORM , arith , 1, 1, sn5 , sn5 , sn6 , , xyz1, rgb +PIPE_FORMAT_A8B8G8R8_SNORM , array , 1, 1, sn8 , sn8 , sn8 , sn8 , wzyx, rgb +PIPE_FORMAT_X8B8G8R8_SNORM , array , 1, 1, sn8 , sn8 , sn8 , sn8 , wzy1, rgb PIPE_FORMAT_R8_SSCALED , array , 1, 1, s8 , , , , x001, rgb PIPE_FORMAT_R8G8_SSCALED , array , 1, 1, s8 , s8 , , , xy01, rgb PIPE_FORMAT_R8G8B8_SSCALED , array , 1, 1, s8 , s8 , s8 , , xyz1, rgb @@ -90,14 +90,14 @@ PIPE_FORMAT_R32G32B32_FIXED , array , 1, 1, h32 , h32 , h32 , , xyz1, PIPE_FORMAT_R32G32B32A32_FIXED , array , 1, 1, h32 , h32 , h32 , h32 , xyzw, rgb PIPE_FORMAT_L8_SRGB , arith , 1, 1, u8 , , , , xxx1, srgb PIPE_FORMAT_A8L8_SRGB , arith , 1, 1, u8 , u8 , , , xxxy, srgb -PIPE_FORMAT_R8G8B8_SRGB , arith , 1, 1, u8 , u8 , u8 , , xyz1, srgb -PIPE_FORMAT_R8G8B8A8_SRGB , arith , 1, 1, u8 , u8 , u8 , u8 , xyzw, srgb -PIPE_FORMAT_R8G8B8X8_SRGB , arith , 1, 1, u8 , u8 , u8 , u8 , xyz1, srgb -PIPE_FORMAT_A8R8G8B8_SRGB , arith , 1, 1, u8 , u8 , u8 , u8 , wxyz, srgb -PIPE_FORMAT_X8R8G8B8_SRGB , arith , 1, 1, u8 , u8 , u8 , u8 , 1xyz, srgb -PIPE_FORMAT_B8G8R8A8_SRGB , arith , 1, 1, u8 , u8 , u8 , u8 , zyxw, srgb -PIPE_FORMAT_B8G8R8X8_SRGB , arith , 1, 1, u8 , u8 , u8 , u8 , zyx1, srgb -PIPE_FORMAT_X8UB8UG8SR8S_NORM , arith , 1, 1, sn8 , sn8 , un8 , x8 , 1zyx, rgb +PIPE_FORMAT_R8G8B8_SRGB , array , 1, 1, u8 , u8 , u8 , , xyz1, srgb +PIPE_FORMAT_R8G8B8A8_SRGB , array , 1, 1, u8 , u8 , u8 , u8 , xyzw, srgb +PIPE_FORMAT_R8G8B8X8_SRGB , array , 1, 1, u8 , u8 , u8 , u8 , xyz1, srgb +PIPE_FORMAT_A8R8G8B8_SRGB , array , 1, 1, u8 , u8 , u8 , u8 , yzwx, srgb +PIPE_FORMAT_X8R8G8B8_SRGB , array , 1, 1, u8 , u8 , u8 , u8 , yzw1, srgb +PIPE_FORMAT_B8G8R8A8_SRGB , array , 1, 1, u8 , u8 , u8 , u8 , zyxw, srgb +PIPE_FORMAT_B8G8R8X8_SRGB , array , 1, 1, u8 , u8 , u8 , u8 , zyx1, srgb +PIPE_FORMAT_X8UB8UG8SR8S_NORM , array , 1, 1, sn8 , sn8 , un8 , x8 , wzy1, rgb PIPE_FORMAT_B6UG5SR5S_NORM , arith , 1, 1, sn5 , sn5 , un6 , , xyz1, rgb PIPE_FORMAT_DXT1_RGB , dxt , 4, 4, x64 , , , , xyz1, rgb PIPE_FORMAT_DXT1_RGBA , dxt , 4, 4, x64 , , , , xyzw, rgb diff --git a/src/gallium/auxiliary/util/u_network.c b/src/gallium/auxiliary/util/u_network.c index 9eb8f309cd..87ee0e4768 100644 --- a/src/gallium/auxiliary/util/u_network.c +++ b/src/gallium/auxiliary/util/u_network.c @@ -117,7 +117,7 @@ u_socket_connect(const char *hostname, uint16_t port) if (!host) return -1; - memcpy((char *)&sa.sin_addr,host->h_addr,host->h_length); + memcpy((char *)&sa.sin_addr,host->h_addr_list[0],host->h_length); sa.sin_family= host->h_addrtype; sa.sin_port = htons(port); diff --git a/src/gallium/auxiliary/util/u_rect.c b/src/gallium/auxiliary/util/u_rect.c index 298fbacecb..8479161c74 100644 --- a/src/gallium/auxiliary/util/u_rect.c +++ b/src/gallium/auxiliary/util/u_rect.c @@ -41,7 +41,7 @@ /** * Copy 2D rect from one place to another. * Position and sizes are in pixels. - * src_pitch may be negative to do vertical flip of pixels from source. + * src_stride may be negative to do vertical flip of pixels from source. */ void util_copy_rect(ubyte * dst, @@ -54,7 +54,7 @@ util_copy_rect(ubyte * dst, const ubyte * src, int src_stride, unsigned src_x, - int src_y) + unsigned src_y) { unsigned i; int src_stride_pos = src_stride < 0 ? -src_stride : src_stride; @@ -65,10 +65,6 @@ util_copy_rect(ubyte * dst, assert(blocksize > 0); assert(blockwidth > 0); assert(blockheight > 0); - assert(src_x >= 0); - assert(src_y >= 0); - assert(dst_x >= 0); - assert(dst_y >= 0); dst_x /= blockwidth; dst_y /= blockheight; @@ -113,8 +109,6 @@ util_fill_rect(ubyte * dst, assert(blocksize > 0); assert(blockwidth > 0); assert(blockheight > 0); - assert(dst_x >= 0); - assert(dst_y >= 0); dst_x /= blockwidth; dst_y /= blockheight; diff --git a/src/gallium/auxiliary/util/u_rect.h b/src/gallium/auxiliary/util/u_rect.h index 5e444ffae2..b44d821904 100644 --- a/src/gallium/auxiliary/util/u_rect.h +++ b/src/gallium/auxiliary/util/u_rect.h @@ -45,7 +45,7 @@ extern void util_copy_rect(ubyte * dst, enum pipe_format format, unsigned dst_stride, unsigned dst_x, unsigned dst_y, unsigned width, unsigned height, const ubyte * src, - int src_stride, unsigned src_x, int src_y); + int src_stride, unsigned src_x, unsigned src_y); extern void util_fill_rect(ubyte * dst, enum pipe_format format, diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c index 8172ead020..b751e29ab6 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.c +++ b/src/gallium/auxiliary/util/u_simple_shaders.c @@ -44,13 +44,15 @@ /** * Make simple vertex pass-through shader. + * \param num_attribs number of attributes to pass through + * \param semantic_names array of semantic names for each attribute + * \param semantic_indexes array of semantic indexes for each attribute */ void * util_make_vertex_passthrough_shader(struct pipe_context *pipe, uint num_attribs, const uint *semantic_names, const uint *semantic_indexes) - { struct ureg_program *ureg; uint i; @@ -78,8 +80,6 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, } - - /** * Make simple fragment texture shader: * IMM {0,0,0,1} // (if writemask != 0xf) @@ -125,6 +125,12 @@ util_make_fragment_tex_shader_writemask(struct pipe_context *pipe, return ureg_create_shader_and_destroy( ureg, pipe ); } + +/** + * Make a simple fragment shader that sets the output color to a color + * taken from a texture. + * \param tex_target one of PIPE_TEXTURE_x + */ void * util_make_fragment_tex_shader(struct pipe_context *pipe, unsigned tex_target ) { @@ -133,6 +139,7 @@ util_make_fragment_tex_shader(struct pipe_context *pipe, unsigned tex_target ) TGSI_WRITEMASK_XYZW ); } + /** * Make a simple fragment texture shader which reads an X component from * a texture and writes it as depth. @@ -177,6 +184,7 @@ util_make_fragment_tex_shader_writedepth(struct pipe_context *pipe, return ureg_create_shader_and_destroy( ureg, pipe ); } + /** * Make simple fragment color pass-through shader. */ @@ -186,15 +194,19 @@ util_make_fragment_passthrough_shader(struct pipe_context *pipe) return util_make_fragment_clonecolor_shader(pipe, 1); } + +/** + * Make a fragment shader that copies the input color to N output colors. + */ void * util_make_fragment_clonecolor_shader(struct pipe_context *pipe, int num_cbufs) { struct ureg_program *ureg; struct ureg_src src; - struct ureg_dst dst[8]; + struct ureg_dst dst[PIPE_MAX_COLOR_BUFS]; int i; - assert(num_cbufs <= 8); + assert(num_cbufs <= PIPE_MAX_COLOR_BUFS); ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT ); if (ureg == NULL) diff --git a/src/gallium/auxiliary/util/u_tile.c b/src/gallium/auxiliary/util/u_tile.c index 5b8dd1abb9..1ba82bb21f 100644 --- a/src/gallium/auxiliary/util/u_tile.c +++ b/src/gallium/auxiliary/util/u_tile.c @@ -1155,27 +1155,6 @@ ycbcr_get_tile_rgba(const ushort *src, } -static void -fake_get_tile_rgba(const ushort *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = (i ^ j) & 1 ? 1.0f : 0.0f; - } - p += dst_stride; - } -} - - void pipe_tile_raw_to_rgba(enum pipe_format format, void *src, @@ -1258,8 +1237,10 @@ pipe_tile_raw_to_rgba(enum pipe_format format, ycbcr_get_tile_rgba((ushort *) src, w, h, dst, dst_stride, TRUE); break; default: - debug_printf("%s: unsupported format %s\n", __FUNCTION__, pf_name(format)); - fake_get_tile_rgba(src, w, h, dst, dst_stride); + util_format_read_4f(format, + dst, dst_stride * sizeof(float), + src, util_format_get_stride(format, w), + 0, 0, w, h); } } |