diff --git a/src/add-ons/accelerants/neomagic/engine/nm_general.c b/src/add-ons/accelerants/neomagic/engine/nm_general.c index 8a80adbe24..03f891d37a 100644 --- a/src/add-ons/accelerants/neomagic/engine/nm_general.c +++ b/src/add-ons/accelerants/neomagic/engine/nm_general.c @@ -47,7 +47,7 @@ status_t nm_general_powerup() { status_t status; - LOG(1,("POWERUP: Neomagic (open)BeOS Accelerant 0.06-10 running.\n")); + LOG(1,("POWERUP: Neomagic (open)BeOS Accelerant 0.06-11 running.\n")); /* detect card type and power it up */ switch(CFGR(DEVID)) @@ -409,20 +409,41 @@ status_t nm_general_validate_pic_size (display_mode *target, uint32 *bytes_per_r * (Note: Don't mix this up with CRTC timing contraints! Those are * multiples of 8 for horizontal, 1 for vertical timing.) * - * CRTC pitch constraints are the same for all Neomagic cards */ - switch (target->space) + * CRTC pitch constraints are 'officially' the same for all Neomagic cards */ + switch (si->ps.card_type) { - case B_CMAP8: crtc_mask = 0x07; depth = 8; break; - case B_RGB15: crtc_mask = 0x03; depth = 16; break; - case B_RGB16: crtc_mask = 0x03; depth = 16; break; - case B_RGB24: crtc_mask = 0x07; depth = 24; break; - default: - LOG(8,("INIT: unsupported colorspace: 0x%08x\n", target->space)); - return B_ERROR; + case NM2070: + switch (target->space) + { + case B_CMAP8: crtc_mask = 0x07; depth = 8; break; + /* Note for NM2070 only: + * The following depths have bandwidth trouble (pixel noise) with the + * 'official' crtc_masks (used as defaults below). Masks of 0x1f are + * needed (confirmed 15 and 16 bit spaces) to prevent this from occuring. */ + //fixme: doublecheck for NM2090 and NM2093: correct code if needed! + case B_RGB15: crtc_mask = 0x1f; depth = 16; break; + case B_RGB16: crtc_mask = 0x1f; depth = 16; break; + case B_RGB24: crtc_mask = 0x1f; depth = 24; break; + default: + LOG(8,("INIT: unsupported colorspace: 0x%08x\n", target->space)); + return B_ERROR; + } + break; + default: + switch (target->space) + { + case B_CMAP8: crtc_mask = 0x07; depth = 8; break; + case B_RGB15: crtc_mask = 0x03; depth = 16; break; + case B_RGB16: crtc_mask = 0x03; depth = 16; break; + case B_RGB24: crtc_mask = 0x07; depth = 24; break; + default: + LOG(8,("INIT: unsupported colorspace: 0x%08x\n", target->space)); + return B_ERROR; + } + break; } - /* check if we can setup this mode with acceleration: - * constraints are same for all cards */ + /* check if we can setup this mode with acceleration: */ *acc_mode = true; switch (si->ps.card_type) @@ -439,12 +460,21 @@ status_t nm_general_validate_pic_size (display_mode *target, uint32 *bytes_per_r } /* virtual_width */ - if (target->virtual_width > 1600) *acc_mode = false; + if (si->ps.card_type == NM2070) + { + /* confirmed NM2070 */ + if (target->virtual_width > 1024) *acc_mode = false; + } + else + { + /* confirmed for all cards */ + if (target->virtual_width > 1600) *acc_mode = false; + } /* virtual_height */ if (si->ps.card_type < NM2200) { - /* confirmed NM2097 and NM2160 */ + /* confirmed NM2070, NM2097 and NM2160 */ if (target->virtual_height > 1024) *acc_mode = false; } else @@ -483,66 +513,89 @@ status_t nm_general_validate_pic_size (display_mode *target, uint32 *bytes_per_r /* OK, now we know that virtual_width is valid, and it's needing no slopspace if * it was confined above, so we can finally calculate safely if we need slopspace * for this mode... */ - /* note: - * we prefer unaccelerated modes above accelerated ones if not enough RAM exists - * and the mode can be closer matched to the requested one if unaccelerated. */ if (*acc_mode) { - uint32 mem_avail, bytes_X_height; + if (si->ps.card_type == NM2070) + { + uint32 acc_mask = 0; - /* calculate amount of available memory */ - mem_avail = (si->ps.memory_size * 1024); - if (si->settings.hardcursor) mem_avail -= si->ps.curmem_size; - /* helper */ - bytes_X_height = (depth >> 3) * target->virtual_height; - - /* Accelerated modes work with a table, there are very few fixed settings.. */ - if (target->virtual_width == 640) video_pitch = 640; - else - if (target->virtual_width <= 800) + /* determine pixel multiple based on 2D engine constraints */ + switch (target->space) { - if (((800 * bytes_X_height) > mem_avail) && - (target->virtual_width < (800 - crtc_mask))) - *acc_mode = false; - else - video_pitch = 800; + case B_CMAP8: acc_mask = 0x07; break; + /* Note: + * The following depths have actual acc_masks of 0x03. 0x1f is used + * because on lower acc_masks bandwidth trouble arises (pixel noise). */ + //fixme: doublecheck for NM2090 and NM2093: correct code if needed! + case B_RGB15: acc_mask = 0x1f; break; + case B_RGB16: acc_mask = 0x1f; break; } + video_pitch = ((target->virtual_width + acc_mask) & ~acc_mask); + } + else + { + /* Note: + * We prefer unaccelerated modes above accelerated ones here if not enough + * RAM exists and the mode can be closer matched to the requested one if + * unaccelerated. We do this because of the large amount of slopspace + * sometimes needed here to keep a mode accelerated. */ + + uint32 mem_avail, bytes_X_height; + + /* calculate amount of available memory */ + mem_avail = (si->ps.memory_size * 1024); + if (si->settings.hardcursor) mem_avail -= si->ps.curmem_size; + /* helper */ + bytes_X_height = (depth >> 3) * target->virtual_height; + + /* Accelerated modes work with a table, there are very few fixed settings.. */ + if (target->virtual_width == 640) video_pitch = 640; else - if (target->virtual_width <= 1024) + if (target->virtual_width <= 800) { - if (((1024 * bytes_X_height) > mem_avail) && - (target->virtual_width < (1024 - crtc_mask))) + if (((800 * bytes_X_height) > mem_avail) && + (target->virtual_width < (800 - crtc_mask))) *acc_mode = false; else - video_pitch = 1024; + video_pitch = 800; } else - if (target->virtual_width <= 1152) + if (target->virtual_width <= 1024) { - if (((1152 * bytes_X_height) > mem_avail) && - (target->virtual_width < (1152 - crtc_mask))) + if (((1024 * bytes_X_height) > mem_avail) && + (target->virtual_width < (1024 - crtc_mask))) *acc_mode = false; else - video_pitch = 1152; + video_pitch = 1024; } else - if (target->virtual_width <= 1280) + if (target->virtual_width <= 1152) { - if (((1280 * bytes_X_height) > mem_avail) && - (target->virtual_width < (1280 - crtc_mask))) + if (((1152 * bytes_X_height) > mem_avail) && + (target->virtual_width < (1152 - crtc_mask))) *acc_mode = false; else - video_pitch = 1280; + video_pitch = 1152; } else - if (target->virtual_width <= 1600) + if (target->virtual_width <= 1280) { - if (((1600 * bytes_X_height) > mem_avail) && - (target->virtual_width < (1600 - crtc_mask))) + if (((1280 * bytes_X_height) > mem_avail) && + (target->virtual_width < (1280 - crtc_mask))) *acc_mode = false; else - video_pitch = 1600; + video_pitch = 1280; } + else + if (target->virtual_width <= 1600) + { + if (((1600 * bytes_X_height) > mem_avail) && + (target->virtual_width < (1600 - crtc_mask))) + *acc_mode = false; + else + video_pitch = 1600; + } + } } if (!*acc_mode) video_pitch = ((target->virtual_width + crtc_mask) & ~crtc_mask); diff --git a/src/add-ons/kernel/drivers/graphics/neomagic/UPDATE.html b/src/add-ons/kernel/drivers/graphics/neomagic/UPDATE.html index c9c23a3ac8..dbbef22326 100644 --- a/src/add-ons/kernel/drivers/graphics/neomagic/UPDATE.html +++ b/src/add-ons/kernel/drivers/graphics/neomagic/UPDATE.html @@ -4,7 +4,7 @@

Changes done for each driverversion:

-

head (0.06-10, Rudolf)

+

head (0.06-11, Rudolf)