perflvl->vdec = read_clk(dev, 0x0e);
return 0;
}
+
+struct nvc0_pm_clock {
+ u32 freq;
+ u32 ssel;
+ u32 mdiv;
+ u32 dsrc;
+ u32 ddiv;
+ u32 coef;
+};
+
+struct nvc0_pm_state {
+ struct nvc0_pm_clock eng[16];
+};
+
+static u32
+calc_div(struct drm_device *dev, int clk, u32 ref, u32 freq, u32 *ddiv)
+{
+ u32 div = min((ref * 2) / freq, (u32)65);
+ if (div < 2)
+ div = 2;
+
+ *ddiv = div - 2;
+ return (ref * 2) / div;
+}
+
+static u32
+calc_src(struct drm_device *dev, int clk, u32 freq, u32 *dsrc, u32 *ddiv)
+{
+ u32 sclk;
+
+ /* use one of the fixed frequencies if possible */
+ *ddiv = 0x00000000;
+ switch (freq) {
+ case 27000:
+ case 108000:
+ *dsrc = 0x00000000;
+ if (freq == 108000)
+ *dsrc |= 0x00030000;
+ return freq;
+ case 100000:
+ *dsrc = 0x00000002;
+ return freq;
+ default:
+ *dsrc = 0x00000003;
+ break;
+ }
+
+ /* otherwise, calculate the closest divider */
+ sclk = read_vco(dev, clk);
+ if (clk < 7)
+ sclk = calc_div(dev, clk, sclk, freq, ddiv);
+ return sclk;
+}
+
+static u32
+calc_pll(struct drm_device *dev, int clk, u32 freq, u32 *coef)
+{
+ struct pll_lims limits;
+ int N, M, P, ret;
+
+ ret = get_pll_limits(dev, 0x137000 + (clk * 0x20), &limits);
+ if (ret)
+ return 0;
+
+ limits.refclk = read_div(dev, clk, 0x137120, 0x137140);
+ if (!limits.refclk)
+ return 0;
+
+ ret = nva3_calc_pll(dev, &limits, freq, &N, NULL, &M, &P);
+ if (ret <= 0)
+ return 0;
+
+ *coef = (P << 16) | (N << 8) | M;
+ return ret;
+}
+
+/* A (likely rather simplified and incomplete) view of the clock tree
+ *
+ * Key:
+ *
+ * S: source select
+ * D: divider
+ * P: pll
+ * F: switch
+ *
+ * Engine clocks:
+ *
+ * 137250(D) ---- 137100(F0) ---- 137160(S)/1371d0(D) ------------------- ref
+ * (F1) ---- 1370X0(P) ---- 137120(S)/137140(D) ---- ref
+ *
+ * Not all registers exist for all clocks. For example: clocks >= 8 don't
+ * have their own PLL (all tied to clock 7's PLL when in PLL mode), nor do
+ * they have the divider at 1371d0, though the source selection at 137160
+ * still exists. You must use the divider at 137250 for these instead.
+ *
+ * Memory clock:
+ *
+ * TBD, read_mem() above is likely very wrong...
+ *
+ */
+
+static int
+calc_clk(struct drm_device *dev, int clk, struct nvc0_pm_clock *info, u32 freq)
+{
+ u32 src0, div0, div1D, div1P = 0;
+ u32 clk0, clk1 = 0;
+
+ /* invalid clock domain */
+ if (!freq)
+ return 0;
+
+ /* first possible path, using only dividers */
+ clk0 = calc_src(dev, clk, freq, &src0, &div0);
+ clk0 = calc_div(dev, clk, clk0, freq, &div1D);
+
+ /* see if we can get any closer using PLLs */
+ if (clk0 != freq) {
+ if (clk < 7)
+ clk1 = calc_pll(dev, clk, freq, &info->coef);
+ else
+ clk1 = read_pll(dev, 0x1370e0);
+ clk1 = calc_div(dev, clk, clk1, freq, &div1P);
+ }
+
+ /* select the method which gets closest to target freq */
+ if (abs((int)freq - clk0) <= abs((int)freq - clk1)) {
+ info->dsrc = src0;
+ if (div0) {
+ info->ddiv |= 0x80000000;
+ info->ddiv |= div0 << 8;
+ info->ddiv |= div0;
+ }
+ if (div1D) {
+ info->mdiv |= 0x80000000;
+ info->mdiv |= div1D;
+ }
+ info->ssel = 0;
+ info->freq = clk0;
+ } else {
+ if (div1P) {
+ info->mdiv |= 0x80000000;
+ info->mdiv |= div1P << 8;
+ }
+ info->ssel = (1 << clk);
+ info->freq = clk1;
+ }
+
+ return 0;
+}
+
+void *
+nvc0_pm_clocks_pre(struct drm_device *dev, struct nouveau_pm_level *perflvl)
+{
+ struct drm_nouveau_private *dev_priv = dev->dev_private;
+ struct nvc0_pm_state *info;
+ int ret;
+
+ info = kzalloc(sizeof(*info), GFP_KERNEL);
+ if (!info)
+ return ERR_PTR(-ENOMEM);
+
+ /* NFI why this is still in the performance table, the ROPCs appear
+ * to get their clock from clock 2 ("hub07", actually hub05 on this
+ * chip, but, anyway...) as well. nvatiming confirms hub05 and ROP
+ * are always the same freq with the binary driver even when the
+ * performance table says they should differ.
+ */
+ if (dev_priv->chipset == 0xd9)
+ perflvl->rop = 0;
+
+ if ((ret = calc_clk(dev, 0x00, &info->eng[0x00], perflvl->shader)) ||
+ (ret = calc_clk(dev, 0x01, &info->eng[0x01], perflvl->rop)) ||
+ (ret = calc_clk(dev, 0x02, &info->eng[0x02], perflvl->hub07)) ||
+ (ret = calc_clk(dev, 0x07, &info->eng[0x07], perflvl->hub06)) ||
+ (ret = calc_clk(dev, 0x08, &info->eng[0x08], perflvl->hub01)) ||
+ (ret = calc_clk(dev, 0x09, &info->eng[0x09], perflvl->copy)) ||
+ (ret = calc_clk(dev, 0x0c, &info->eng[0x0c], perflvl->daemon)) ||
+ (ret = calc_clk(dev, 0x0e, &info->eng[0x0e], perflvl->vdec))) {
+ kfree(info);
+ return ERR_PTR(ret);
+ }
+
+ return info;
+}
+
+static void
+prog_clk(struct drm_device *dev, int clk, struct nvc0_pm_clock *info)
+{
+ /* program dividers at 137160/1371d0 first */
+ if (clk < 7 && !info->ssel) {
+ nv_mask(dev, 0x1371d0 + (clk * 0x04), 0x80003f3f, info->ddiv);
+ nv_wr32(dev, 0x137160 + (clk * 0x04), info->dsrc);
+ }
+
+ /* switch clock to non-pll mode */
+ nv_mask(dev, 0x137100, (1 << clk), 0x00000000);
+ nv_wait(dev, 0x137100, (1 << clk), 0x00000000);
+
+ /* reprogram pll */
+ if (clk < 7) {
+ /* make sure it's disabled first... */
+ u32 base = 0x137000 + (clk * 0x20);
+ u32 ctrl = nv_rd32(dev, base + 0x00);
+ if (ctrl & 0x00000001) {
+ nv_mask(dev, base + 0x00, 0x00000004, 0x00000000);
+ nv_mask(dev, base + 0x00, 0x00000001, 0x00000000);
+ }
+ /* program it to new values, if necessary */
+ if (info->ssel) {
+ nv_wr32(dev, base + 0x04, info->coef);
+ nv_mask(dev, base + 0x00, 0x00000001, 0x00000001);
+ nv_wait(dev, base + 0x00, 0x00020000, 0x00020000);
+ nv_mask(dev, base + 0x00, 0x00020004, 0x00000004);
+ }
+ }
+
+ /* select pll/non-pll mode, and program final clock divider */
+ nv_mask(dev, 0x137100, (1 << clk), info->ssel);
+ nv_wait(dev, 0x137100, (1 << clk), info->ssel);
+ nv_mask(dev, 0x137250 + (clk * 0x04), 0x00003f3f, info->mdiv);
+}
+
+int
+nvc0_pm_clocks_set(struct drm_device *dev, void *data)
+{
+ struct nvc0_pm_state *info = data;
+ int i;
+
+ for (i = 0; i < 16; i++) {
+ if (!info->eng[i].freq)
+ continue;
+ prog_clk(dev, i, &info->eng[i]);
+ }
+
+ kfree(info);
+ return 0;
+}
projects / karo-tx-linux.git / commitdiff