From 5eb6b1a981f87a39323604f5496aafdb38b126a9 Mon Sep 17 00:00:00 2001 From: Luca Barbieri Date: Tue, 23 Feb 2010 15:52:24 +0100 Subject: [PATCH] nvfx: so->RING_3D: vertprog We could use an sb here instead, but there is no obvious performance advantage, and perhaps there could be a disadvantage. --- src/gallium/drivers/nvfx/nvfx_state.h | 1 - src/gallium/drivers/nvfx/nvfx_vertprog.c | 45 ++++++++++-------------- 2 files changed, 19 insertions(+), 27 deletions(-) diff --git a/src/gallium/drivers/nvfx/nvfx_state.h b/src/gallium/drivers/nvfx/nvfx_state.h index f1f9fb775b8..555513a6428 100644 --- a/src/gallium/drivers/nvfx/nvfx_state.h +++ b/src/gallium/drivers/nvfx/nvfx_state.h @@ -39,7 +39,6 @@ struct nvfx_vertex_program { uint32_t ir; uint32_t or; uint32_t clip_ctrl; - struct nouveau_stateobj *so; }; struct nvfx_fragment_program_data { diff --git a/src/gallium/drivers/nvfx/nvfx_vertprog.c b/src/gallium/drivers/nvfx/nvfx_vertprog.c index 78ba10c31c5..3d25ab761a3 100644 --- a/src/gallium/drivers/nvfx/nvfx_vertprog.c +++ b/src/gallium/drivers/nvfx/nvfx_vertprog.c @@ -847,6 +847,7 @@ nvfx_vertprog_validate(struct nvfx_context *nvfx) vp = nvfx->vertprog; constbuf = nvfx->constbuf[PIPE_SHADER_VERTEX]; + // TODO: ouch! can't we just use constant slots for these?! if ((nvfx->dirty & NVFX_NEW_UCP) || memcmp(&nvfx->clip, &vp->ucp, sizeof(vp->ucp))) { nvfx_vertprog_destroy(nvfx, vp); @@ -858,21 +859,19 @@ nvfx_vertprog_validate(struct nvfx_context *nvfx) } /* Translate TGSI shader into hw bytecode */ - if (vp->translated) - goto check_gpu_resources; - - nvfx->fallback_swtnl &= ~NVFX_NEW_VERTPROG; + if (!vp->translated) + { + nvfx->fallback_swtnl &= ~NVFX_NEW_VERTPROG; nvfx_vertprog_translate(nvfx, vp); - if (!vp->translated) { - nvfx->fallback_swtnl |= NVFX_NEW_VERTPROG; + if (!vp->translated) { + nvfx->fallback_swtnl |= NVFX_NEW_VERTPROG; return FALSE; + } } -check_gpu_resources: /* Allocate hw vtxprog exec slots */ if (!vp->exec) { struct nouveau_resource *heap = nvfx->screen->vp_exec_heap; - struct nouveau_stateobj *so; uint vplen = vp->nr_insns; if (nouveau_resource_alloc(heap, vplen, vp, &vp->exec)) { @@ -887,19 +886,6 @@ check_gpu_resources: assert(0); } - so = so_new(3, 4, 0); - so_method(so, eng3d, NV34TCL_VP_START_FROM_ID, 1); - so_data (so, vp->exec->start); - if(nvfx->is_nv4x) { - so_method(so, eng3d, NV40TCL_VP_ATTRIB_EN, 2); - so_data (so, vp->ir); - so_data (so, vp->or); - } - so_method(so, eng3d, NV34TCL_VP_CLIP_PLANES_ENABLE, 1); - so_data (so, vp->clip_ctrl); - so_ref(so, &vp->so); - so_ref(NULL, &so); - upload_code = TRUE; } @@ -1007,9 +993,18 @@ check_gpu_resources: } } - if (vp->so != nvfx->state.hw[NVFX_STATE_VERTPROG]) { - so_ref(vp->so, &nvfx->state.hw[NVFX_STATE_VERTPROG]); - return TRUE; + if(nvfx->dirty & (NVFX_NEW_VERTPROG | NVFX_NEW_UCP)) + { + WAIT_RING(chan, 7); + OUT_RING(chan, RING_3D(NV34TCL_VP_START_FROM_ID, 1)); + OUT_RING(chan, vp->exec->start); + if(nvfx->is_nv4x) { + OUT_RING(chan, RING_3D(NV40TCL_VP_ATTRIB_EN, 2)); + OUT_RING(chan, vp->ir); + OUT_RING(chan, vp->or); + } + OUT_RING(chan, RING_3D(NV34TCL_VP_CLIP_PLANES_ENABLE, 1)); + OUT_RING(chan, vp->clip_ctrl); } return FALSE; @@ -1039,13 +1034,11 @@ nvfx_vertprog_destroy(struct nvfx_context *nvfx, struct nvfx_vertex_program *vp) vp->data_start_min = 0; vp->ir = vp->or = vp->clip_ctrl = 0; - so_ref(NULL, &vp->so); } struct nvfx_state_entry nvfx_state_vertprog = { .validate = nvfx_vertprog_validate, .dirty = { .pipe = NVFX_NEW_VERTPROG | NVFX_NEW_UCP, - .hw = NVFX_STATE_VERTPROG, } }; -- 2.30.2