aboutsummaryrefslogtreecommitdiff
path: root/emu8k.c
diff options
context:
space:
mode:
Diffstat (limited to 'emu8k.c')
-rw-r--r--emu8k.c2427
1 files changed, 2427 insertions, 0 deletions
diff --git a/emu8k.c b/emu8k.c
new file mode 100644
index 0000000..fdcf773
--- /dev/null
+++ b/emu8k.c
@@ -0,0 +1,2427 @@
+/*
+ * PCem - IBM PC emulator
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
+ */
+
+/*
+ * Portions:
+ * VMusic - a VirtualBox extension pack with various music devices
+ * Copyright (C) 2022 Javier S. Pedro
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
+ */
+
+#include <math.h>
+
+#define LOG_ENABLED 1
+#define LOG_ENABLE_FLOW 1
+#define LOG_GROUP LOG_GROUP_DEV_SB16
+#include <VBox/log.h>
+#include <iprt/assert.h>
+#include <iprt/string.h>
+#include <iprt/mem.h>
+
+#include "emu8k_internal.h"
+
+#define pclog(...) LogFlow((__VA_ARGS__))
+
+#if !defined FILTER_INITIAL && !defined FILTER_MOOG && !defined FILTER_CONSTANT
+//#define FILTER_INITIAL
+#define FILTER_MOOG
+//#define FILTER_CONSTANT
+#endif
+
+#if !defined RESAMPLER_LINEAR && !defined RESAMPLER_CUBIC
+//#define RESAMPLER_LINEAR
+#define RESAMPLER_CUBIC
+#endif
+
+//#define EMU8K_DEBUG_REGISTERS
+
+char* PORT_NAMES[][8] =
+ {
+ /* Data 0 ( 0x620/0x622) */
+ { "AWE_CPF",
+ "AWE_PTRX",
+ "AWE_CVCF",
+ "AWE_VTFT",
+ "Unk-620-4",
+ "Unk-620-5",
+ "AWE_PSST",
+ "AWE_CSL",
+ },
+ /* Data 1 0xA20 */
+ { "AWE_CCCA",
+ 0,
+ /*
+ "AWE_HWCF4"
+ "AWE_HWCF5"
+ "AWE_HWCF6"
+ "AWE_HWCF7"
+ "AWE_SMALR"
+ "AWE_SMARR"
+ "AWE_SMALW"
+ "AWE_SMARW"
+ "AWE_SMLD"
+ "AWE_SMRD"
+ "AWE_WC"
+ "AWE_HWCF1"
+ "AWE_HWCF2"
+ "AWE_HWCF3"
+ */
+ 0,//"AWE_INIT1",
+ 0,//"AWE_INIT3",
+ "AWE_ENVVOL",
+ "AWE_DCYSUSV",
+ "AWE_ENVVAL",
+ "AWE_DCYSUS",
+ },
+ /* Data 2 0xA22 */
+ { "AWE_CCCA",
+ 0,
+ 0,//"AWE_INIT2",
+ 0,//"AWE_INIT4",
+ "AWE_ATKHLDV",
+ "AWE_LFO1VAL",
+ "AWE_ATKHLD",
+ "AWE_LFO2VAL",
+ },
+ /* Data 3 0xE20 */
+ { "AWE_IP",
+ "AWE_IFATN",
+ "AWE_PEFE",
+ "AWE_FMMOD",
+ "AWE_TREMFRQ",
+ "AWE_FM2FRQ2",
+ 0,
+ 0,
+ },
+ };
+
+enum
+{
+ ENV_STOPPED = 0,
+ ENV_DELAY = 1,
+ ENV_ATTACK = 2,
+ ENV_HOLD = 3,
+ //ENV_DECAY = 4,
+ ENV_SUSTAIN = 5,
+ //ENV_RELEASE = 6,
+ ENV_RAMP_DOWN = 7,
+ ENV_RAMP_UP = 8
+};
+
+static int random_helper = 0;
+static int dmareadbit = 0;
+static int dmawritebit = 0;
+
+
+/* cubic and linear tables resolution. Note: higher than 10 does not improve the result. */
+#define CUBIC_RESOLUTION_LOG 10
+#define CUBIC_RESOLUTION (1<<CUBIC_RESOLUTION_LOG)
+/* cubic_table coefficients. */
+static float cubic_table[CUBIC_RESOLUTION * 4];
+
+/* conversion from current pitch to linear frequency change (in 32.32 fixed point). */
+static int64_t freqtable[65536];
+/* Conversion from initial attenuation to 16 bit unsigned lineal amplitude (currently only a way to update volume target register) */
+static int32_t attentable[256];
+/* Conversion from envelope dbs (once rigth shifted) (0 = 0dBFS, 65535 = -96dbFS and silence ) to 16 bit unsigned lineal amplitude,
+ * to convert to current volume. (0 to 65536) */
+static int32_t env_vol_db_to_vol_target[65537];
+/* Same as above, but to convert amplitude (once rigth shifted) (0 to 65536) to db (0 = 0dBFS, 65535 = -96dbFS and silence ).
+ * it is needed so that the delay, attack and hold phase can be added to initial attenuation and tremolo */
+static int32_t env_vol_amplitude_to_db[65537];
+/* Conversion from envelope herts (once right shifted) to octave . it is needed so that the delay, attack and hold phase can be
+ * added to initial pitch ,lfos pitch , initial filter and lfo filter */
+static int32_t env_mod_hertz_to_octave[65537];
+/* Conversion from envelope amount to time in samples. */
+static int32_t env_attack_to_samples[128];
+/* This table has been generated using the following formula:
+ * Get the amount of dBs that have to be added each sample to reach 96dBs in the amount
+ * of time determined by the encoded value "i".
+ * float d = 1.0/((env_decay_to_millis[i]/96.0)*44.1);
+ * int result = round(d*21845);
+ * The multiplication by 21845 gives a minimum value of 1, and a maximum accumulated value of 1<<21
+ * The accumulated value has to be converted to amplitude, and that can be done with the
+ * env_vol_db_to_vol_target and shifting by 8
+ * In other words, the unit of the table is the 1/21845th of a dB per sample frame, to be added or
+ * substracted to the accumulating value_db of the envelope. */
+static int32_t env_decay_to_dbs_or_oct[128] =
+ {
+ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
+ 16, 17, 18, 19, 20, 20, 21, 22, 23, 24, 25, 27, 28, 29, 30, 32,
+ 33, 34, 36, 38, 39, 41, 43, 45, 49, 51, 53, 55, 58, 60, 63, 66,
+ 69, 72, 75, 78, 82, 85, 89, 93, 97, 102, 106, 111, 116, 121, 126, 132,
+ 138, 144, 150, 157, 164, 171, 179, 186, 195, 203, 212, 222, 232, 243, 253, 264,
+ 276, 288, 301, 315, 328, 342, 358, 374, 390, 406, 425, 444, 466, 485, 506, 528,
+ 553, 580, 602, 634, 660, 689, 721, 755, 780, 820, 849, 897, 932, 970, 1012, 1057,
+ 1106, 1160, 1219, 1285, 1321, 1399, 1441, 1534, 1585, 1640, 1698, 1829, 1902, 1981, 2068, 2162
+ };
+/* The table "env_decay_to_millis" is based on the table "decay_time_tbl" found in the freebsd/linux
+ * AWE32 driver.
+ * I tried calculating it using the instructions in awe32p10 from Judge Dredd, but the formula there
+ * is wrong.
+ *
+static int32_t env_decay_to_millis[128] = {
+0, 45120, 22614, 15990, 11307, 9508, 7995, 6723, 5653, 5184, 4754, 4359, 3997, 3665, 3361, 3082,
+2828, 2765, 2648, 2535, 2428, 2325, 2226, 2132, 2042, 1955, 1872, 1793, 1717, 1644, 1574, 1507,
+1443, 1382, 1324, 1267, 1214, 1162, 1113, 1066, 978, 936, 897, 859, 822, 787, 754, 722,
+691, 662, 634, 607, 581, 557, 533, 510, 489, 468, 448, 429, 411, 393, 377, 361,
+345, 331, 317, 303, 290, 278, 266, 255, 244, 234, 224, 214, 205, 196, 188, 180,
+172, 165, 158, 151, 145, 139, 133, 127, 122, 117, 112, 107, 102, 98, 94, 90,
+86, 82, 79, 75, 72, 69, 66, 63, 61, 58, 56, 53, 51, 49, 47, 45,
+43, 41, 39, 37, 36, 34, 33, 31, 30, 29, 28, 26, 25, 24, 23, 22,
+};
+*/
+
+/* Table represeting the LFO waveform (signed 16bits with 32768 max int. >> 15 to move back to +/-1 range). */
+static int32_t lfotable[65536];
+/* Table to transform the speed parameter to emu8k_mem_internal_t range. */
+static int64_t lfofreqtospeed[256];
+
+/* LFO used for the chorus. a sine wave.(signed 16bits with 32768 max int. >> 15 to move back to +/-1 range). */
+static double chortable[65536];
+
+static const int REV_BUFSIZE_STEP = 242;
+
+/* These lines come from the awe32faq, describing the NRPN control for the initial filter
+ * where it describes a linear increment filter instead of an octave-incremented one.
+ * NRPN LSB 21 (Initial Filter Cutoff)
+ * Range : [0, 127]
+ * Unit : 62Hz
+ * Filter cutoff from 100Hz to 8000Hz
+
+ * This table comes from the awe32faq, describing the NRPN control for the filter Q.
+ * I don't know if is meant to be interpreted as the actual measured output of the
+ * filter or what. Especially, I don't understand the "low" and "high" ranges.
+ * What is otherwise documented is that the Q ranges from 0dB to 24dB and the attenuation
+ * is half of the Q ( i.e. for 12dB Q, attenuate the input signal with -6dB)
+Coeff Low Fc(Hz)Low Q(dB)High Fc(kHz)High Q(dB)DC Attenuation(dB)
+* 0 92 5 Flat Flat -0.0
+* 1 93 6 8.5 0.5 -0.5
+* 2 94 8 8.3 1 -1.2
+* 3 95 10 8.2 2 -1.8
+* 4 96 11 8.1 3 -2.5
+* 5 97 13 8.0 4 -3.3
+* 6 98 14 7.9 5 -4.1
+* 7 99 16 7.8 6 -5.5
+* 8 100 17 7.7 7 -6.0
+* 9 100 19 7.5 9 -6.6
+* 10 100 20 7.4 10 -7.2
+* 11 100 22 7.3 11 -7.9
+* 12 100 23 7.2 13 -8.5
+* 13 100 25 7.1 15 -9.3
+* 14 100 26 7.1 16 -10.1
+* 15 100 28 7.0 18 -11.0
+*
+* Attenuation as above, codified in amplitude.*/
+static int32_t filter_atten[16] =
+ {
+ 65536, 61869, 57079, 53269, 49145, 44820, 40877, 34792, 32845, 30653, 28607,
+ 26392, 24630, 22463, 20487, 18470
+ };
+
+/*Coefficients for the filters for a defined Q and cutoff.*/
+static int32_t filt_coeffs[16][256][3];
+
+#define READ16_SWITCH(addr, var) switch ((addr) & 2) \
+ { \
+ case 0: ret = (var) & 0xffff; break; \
+ case 2: ret = ((var) >> 16) & 0xffff; break; \
+ }
+
+#define WRITE16_SWITCH(addr, var, val) switch ((addr) & 2) \
+ { \
+ case 0: var = (var & 0xffff0000) | (val); break; \
+ case 2: var = (var & 0x0000ffff) | ((val) << 16); break; \
+ }
+
+#ifdef EMU8K_DEBUG_REGISTERS
+static uint32_t last_read = 0;
+static uint32_t last_write = 0;
+static uint32_t rep_count_r = 0;
+static uint32_t rep_count_w = 0;
+
+# define READ16(addr, var) READ16_SWITCH(addr, var) \
+ { \
+ const char *name=0; \
+ switch(addr&0xF02) \
+ { \
+ case 0x600: case 0x602: \
+ name = PORT_NAMES[0][emu8k->cur_reg]; \
+ break; \
+ case 0xA00: \
+ name = PORT_NAMES[1][emu8k->cur_reg]; \
+ break; \
+ case 0xA02: \
+ name = PORT_NAMES[2][emu8k->cur_reg]; \
+ break; \
+ } \
+ if (name == 0) \
+ { \
+ /*pclog("EMU8K READ %04X-%02X(%d): %04X\n",addr,(emu8k->cur_reg)<<5|emu8k->cur_voice, emu8k->cur_voice,ret);*/ \
+ } \
+ else \
+ { \
+ pclog("EMU8K READ %s(%d) (%d): %04X\n",name, (addr&0x2), emu8k->cur_voice, ret); \
+ }\
+ }
+# define WRITE16(addr, var, val) WRITE16_SWITCH(addr, var, val) \
+ { \
+ const char *name=0; \
+ switch(addr&0xF02) \
+ { \
+ case 0x600: case 0x602: \
+ name = PORT_NAMES[0][emu8k->cur_reg]; \
+ break; \
+ case 0xA00: \
+ name = PORT_NAMES[1][emu8k->cur_reg]; \
+ break; \
+ case 0xA02: \
+ name = PORT_NAMES[2][emu8k->cur_reg]; \
+ break; \
+ } \
+ if (name == 0) \
+ { \
+ /*pclog("EMU8K WRITE %04X-%02X(%d): %04X\n",addr,(emu8k->cur_reg)<<5|emu8k->cur_voice,emu8k->cur_voice, val);*/ \
+ } \
+ else \
+ { \
+ pclog("EMU8K WRITE %s(%d) (%d): %04X\n",name, (addr&0x2), emu8k->cur_voice,val); \
+ }\
+ }
+
+#else
+# define READ16(addr, var) READ16_SWITCH(addr, var)
+# define WRITE16(addr, var, val) WRITE16_SWITCH(addr, var, val)
+#endif //EMU8K_DEBUG_REGISTERS
+
+static inline int16_t EMU8K_READ(emu8k_t* emu8k, uint32_t addr)
+{
+ const emu8k_mem_pointers_t addrmem = {{ addr }};
+ return emu8k->ram_pointers[addrmem.hb_address][addrmem.lw_address];
+}
+
+static inline int16_t EMU8K_READ_INTERP_LINEAR(emu8k_t* emu8k, uint32_t int_addr, uint16_t fract)
+{
+ /* The interpolation in AWE32 used a so-called patented 3-point interpolation
+ * ( I guess some sort of spline having one point before and one point after).
+ * Also, it has the consequence that the playback is delayed by one sample.
+ * I simulate the "one sample later" than the address with addr+1 and addr+2
+ * instead of +0 and +1 */
+ int16_t dat1 = EMU8K_READ(emu8k, int_addr + 1);
+ int32_t dat2 = EMU8K_READ(emu8k, int_addr + 2);
+ dat1 += ((dat2 - (int32_t)dat1) * fract) >> 16;
+ return dat1;
+}
+
+static inline int32_t EMU8K_READ_INTERP_CUBIC(emu8k_t* emu8k, uint32_t int_addr, uint16_t fract)
+{
+ /*Since there are four floats in the table for each fraction, the position is 16byte aligned. */
+ fract >>= 16 - CUBIC_RESOLUTION_LOG;
+ fract <<= 2;
+
+ /* TODO: I still have to verify how this works, but I think that
+ * the card could use two oscillators (usually 31 and 32) where it would
+ * be writing the OPL3 output, and to which, chorus and reverb could be applied to get
+ * those effects for OPL3 sounds.*/
+// if ((addr & EMU8K_FM_MEM_ADDRESS) == EMU8K_FM_MEM_ADDRESS) {}
+
+ /* This is cubic interpolation.
+ * Not the same than 3-point interpolation, but a better approximation than linear
+ * interpolation.
+ * Also, it takes into account the "Note that the actual audio location is the point
+ * 1 word higher than this value due to interpolation offset".
+ * That's why the pointers are 0, 1, 2, 3 and not -1, 0, 1, 2 */
+ int32_t dat2 = EMU8K_READ(emu8k, int_addr + 1);
+ const float* table = &cubic_table[fract];
+ const int32_t dat1 = EMU8K_READ(emu8k, int_addr);
+ const int32_t dat3 = EMU8K_READ(emu8k, int_addr + 2);
+ const int32_t dat4 = EMU8K_READ(emu8k, int_addr + 3);
+ /* Note: I've ended using float for the table values to avoid some cases of integer overflow. */
+ dat2 = dat1 * table[0] + dat2 * table[1] + dat3 * table[2] + dat4 * table[3];
+ return dat2;
+}
+
+static inline void EMU8K_WRITE(emu8k_t* emu8k, uint32_t addr, uint16_t val)
+{
+ addr &= EMU8K_MEM_ADDRESS_MASK;
+ if (!emu8k->ram || addr < EMU8K_RAM_MEM_START || addr >= EMU8K_FM_MEM_ADDRESS)
+ return;
+
+ /* It looks like if an application writes to a memory part outside of the available
+ * amount on the card, it wraps, and opencubicplayer uses that to detect the amount
+ * of memory, as opposed to simply check at the address that it has just tried to write. */
+ while (addr >= emu8k->ram_end_addr)
+ addr -= emu8k->ram_end_addr - EMU8K_RAM_MEM_START;
+
+ emu8k->ram[addr - EMU8K_RAM_MEM_START] = val;
+}
+
+uint16_t emu8k_inw(emu8k_t *emu8k, uint16_t addr)
+{
+ uint16_t ret = 0xffff;
+
+#ifdef EMU8K_DEBUG_REGISTERS
+ if (addr == 0xE22)
+ {
+ pclog("EMU8K READ POINTER: %d\n",
+ ((0x80 | ((random_helper + 1) & 0x1F)) << 8) | (emu8k->cur_reg << 5) | emu8k->cur_voice);
+ }
+ else if ((addr&0xF00) == 0x600)
+ {
+ /* These are automatically reported by READ16 */
+ if (rep_count_r>1)
+ {
+ pclog("EMU8K ...... for %d times\n", rep_count_r);
+ rep_count_r=0;
+ }
+ last_read=0;
+ }
+ else if ((addr&0xF00) == 0xA00 && emu8k->cur_reg == 0)
+ {
+ /* These are automatically reported by READ16 */
+ if (rep_count_r>1)
+ {
+ pclog("EMU8K ...... for %d times\n", rep_count_r);
+ rep_count_r=0;
+ }
+ last_read=0;
+ }
+ else if ((addr&0xF00) == 0xA00 && emu8k->cur_reg == 1)
+ {
+ uint32_t tmpz = ((addr&0xF00) << 16)|(emu8k->cur_reg<<5);
+ if (tmpz != last_read)
+ {
+ if (rep_count_r>1)
+ {
+ pclog("EMU8K ...... for %d times\n", rep_count_r);
+ rep_count_r=0;
+ }
+ last_read=tmpz;
+ pclog("EMU8K READ RAM I/O or configuration or clock \n");
+ }
+ //pclog("EMU8K READ %04X-%02X(%d/%d)\n",addr,(emu8k->cur_reg)<<5|emu8k->cur_voice, emu8k->cur_reg, emu8k->cur_voice);
+ }
+ else if ((addr&0xF00) == 0xA00 && (emu8k->cur_reg == 2 || emu8k->cur_reg == 3))
+ {
+ uint32_t tmpz = ((addr&0xF00) << 16);
+ if (tmpz != last_read)
+ {
+ if (rep_count_r>1)
+ {
+ pclog("EMU8K ...... for %d times\n", rep_count_r);
+ rep_count_r=0;
+ }
+ last_read=tmpz;
+ pclog("EMU8K READ INIT \n");
+ }
+ //pclog("EMU8K READ %04X-%02X(%d/%d)\n",addr,(emu8k->cur_reg)<<5|emu8k->cur_voice, emu8k->cur_reg, emu8k->cur_voice);
+ }
+ else
+ {
+ uint32_t tmpz = (addr << 16)|(emu8k->cur_reg<<5)| emu8k->cur_voice;
+ if (tmpz != last_read)
+ {
+ char* name = 0;
+ uint16_t val = 0xBAAD;
+ if (addr == 0xA20)
+ {
+ name = PORT_NAMES[1][emu8k->cur_reg];
+ switch (emu8k->cur_reg)
+ {
+ case 2: val = emu8k->init1[emu8k->cur_voice]; break;
+ case 3: val = emu8k->init3[emu8k->cur_voice]; break;
+ case 4: val = emu8k->voice[emu8k->cur_voice].envvol; break;
+ case 5: val = emu8k->voice[emu8k->cur_voice].dcysusv; break;
+ case 6: val = emu8k->voice[emu8k->cur_voice].envval; break;
+ case 7: val = emu8k->voice[emu8k->cur_voice].dcysus; break;
+ }
+ }
+ else if (addr == 0xA22)
+ {
+ name = PORT_NAMES[2][emu8k->cur_reg];
+ switch (emu8k->cur_reg)
+ {
+ case 2: val = emu8k->init2[emu8k->cur_voice]; break;
+ case 3: val = emu8k->init4[emu8k->cur_voice]; break;
+ case 4: val = emu8k->voice[emu8k->cur_voice].atkhldv; break;
+ case 5: val = emu8k->voice[emu8k->cur_voice].lfo1val; break;
+ case 6: val = emu8k->voice[emu8k->cur_voice].atkhld; break;
+ case 7: val = emu8k->voice[emu8k->cur_voice].lfo2val; break;
+ }
+ }
+ else if (addr == 0xE20)
+ {
+ name = PORT_NAMES[3][emu8k->cur_reg];
+ switch (emu8k->cur_reg)
+ {
+ case 0: val = emu8k->voice[emu8k->cur_voice].ip; break;
+ case 1: val = emu8k->voice[emu8k->cur_voice].ifatn; break;
+ case 2: val = emu8k->voice[emu8k->cur_voice].pefe; break;
+ case 3: val = emu8k->voice[emu8k->cur_voice].fmmod; break;
+ case 4: val = emu8k->voice[emu8k->cur_voice].tremfrq; break;
+ case 5: val = emu8k->voice[emu8k->cur_voice].fm2frq2;break;
+ case 6: val = 0xffff; break;
+ case 7: val = 0x1c | ((emu8k->id & 0x0002) ? 0xff02 : 0); break;
+ }
+ }
+ if (rep_count_r>1)
+ {
+ pclog("EMU8K ...... for %d times\n", rep_count_r);
+ }
+ if (name == 0)
+ {
+ pclog("EMU8K READ %04X-%02X(%d/%d): %04X\n",addr,(emu8k->cur_reg)<<5|emu8k->cur_voice, emu8k->cur_reg, emu8k->cur_voice,val);
+ }
+ else
+ {
+ pclog("EMU8K READ %s (%d): %04X\n",name,emu8k->cur_voice, val);
+ }
+
+ rep_count_r=0;
+ last_read=tmpz;
+ }
+ rep_count_r++;
+ }
+#endif // EMU8K_DEBUG_REGISTERS
+
+ switch (addr & 0xF02)
+ {
+ case 0x600:
+ case 0x602: /*Data0. also known as BLASTER+0x400 and EMU+0x000 */
+ switch (emu8k->cur_reg)
+ {
+ case 0:
+ READ16(addr, emu8k->voice[emu8k->cur_voice].cpf);
+ return ret;
+
+ case 1:
+ READ16(addr, emu8k->voice[emu8k->cur_voice].ptrx);
+ return ret;
+
+ case 2:
+ READ16(addr, emu8k->voice[emu8k->cur_voice].cvcf);
+ return ret;
+
+ case 3:
+ READ16(addr, emu8k->voice[emu8k->cur_voice].vtft);
+ return ret;
+
+ case 4:
+ READ16(addr, emu8k->voice[emu8k->cur_voice].unknown_data0_4);
+ return ret;
+
+ case 5:
+ READ16(addr, emu8k->voice[emu8k->cur_voice].unknown_data0_5);
+ return ret;
+
+ case 6:
+ READ16(addr, emu8k->voice[emu8k->cur_voice].psst);
+ return ret;
+
+ case 7:
+ READ16(addr, emu8k->voice[emu8k->cur_voice].csl);
+ return ret;
+ }
+ break;
+
+ case 0xA00: /*Data1. also known as BLASTER+0x800 and EMU+0x400 */
+ switch (emu8k->cur_reg)
+ {
+ case 0:
+ READ16(addr, emu8k->voice[emu8k->cur_voice].ccca);
+ return ret;
+
+ case 1:
+ switch (emu8k->cur_voice)
+ {
+ case 9:
+ READ16(addr, emu8k->hwcf4);
+ return ret;
+ case 10:
+ READ16(addr, emu8k->hwcf5);
+ return ret;
+ /* Actually, these two might be command words rather than registers, or some LFO position/buffer reset.*/
+ case 13:
+ READ16(addr, emu8k->hwcf6);
+ return ret;
+ case 14:
+ READ16(addr, emu8k->hwcf7);
+ return ret;
+
+ case 20:
+ READ16(addr, emu8k->smalr);
+ return ret;
+ case 21:
+ READ16(addr, emu8k->smarr);
+ return ret;
+ case 22:
+ READ16(addr, emu8k->smalw);
+ return ret;
+ case 23:
+ READ16(addr, emu8k->smarw);
+ return ret;
+
+ case 26:
+ {
+ uint16_t val = emu8k->smld_buffer;
+ emu8k->smld_buffer = EMU8K_READ(emu8k, emu8k->smalr);
+ emu8k->smalr = (emu8k->smalr + 1) & EMU8K_MEM_ADDRESS_MASK;
+ return val;
+ }
+
+ /*The EMU8000 PGM describes the return values of these registers as 'a VLSI error'*/
+ case 29: /*Configuration Word 1*/
+ return (emu8k->hwcf1 & 0xfe) | (emu8k->hwcf3 & 0x01);
+ case 30: /*Configuration Word 2*/
+ return ((emu8k->hwcf2 >> 4) & 0x0e) | (emu8k->hwcf1 & 0x01) | ((emu8k->hwcf3 & 0x02) ? 0x10 : 0) | ((emu8k->hwcf3 & 0x04) ? 0x40 : 0)
+ | ((emu8k->hwcf3 & 0x08) ? 0x20 : 0) | ((emu8k->hwcf3 & 0x10) ? 0x80 : 0);
+ case 31: /*Configuration Word 3*/
+ return emu8k->hwcf2 & 0x1f;
+ }
+ break;
+
+ case 2:
+ return emu8k->init1[emu8k->cur_voice];
+
+ case 3:
+ return emu8k->init3[emu8k->cur_voice];
+
+ case 4:
+ return emu8k->voice[emu8k->cur_voice].envvol;
+
+ case 5:
+ return emu8k->voice[emu8k->cur_voice].dcysusv;
+
+ case 6:
+ return emu8k->voice[emu8k->cur_voice].envval;
+
+ case 7:
+ return emu8k->voice[emu8k->cur_voice].dcysus;
+ }
+ break;
+
+ case 0xA02: /*Data2. also known as BLASTER+0x802 and EMU+0x402 */
+ switch (emu8k->cur_reg)
+ {
+ case 0:
+ READ16(addr, emu8k->voice[emu8k->cur_voice].ccca);
+ return ret;
+
+ case 1:
+ switch (emu8k->cur_voice)
+ {
+ case 9:
+ READ16(addr, emu8k->hwcf4);
+ return ret;
+ case 10:
+ READ16(addr, emu8k->hwcf5);
+ return ret;
+ /* Actually, these two might be command words rather than registers, or some LFO position/buffer reset. */
+ case 13:
+ READ16(addr, emu8k->hwcf6);
+ return ret;
+ case 14:
+ READ16(addr, emu8k->hwcf7);
+ return ret;
+
+ /* Simulating empty/full bits by unsetting it once read. */
+ case 20:
+ READ16(addr, emu8k->smalr | dmareadbit);
+ /* xor with itself to set to zero faster. */
+ dmareadbit ^= dmareadbit;
+ return ret;
+ case 21:
+ READ16(addr, emu8k->smarr | dmareadbit);
+ /* xor with itself to set to zero faster.*/
+ dmareadbit ^= dmareadbit;
+ return ret;
+ case 22:
+ READ16(addr, emu8k->smalw | dmawritebit);
+ /*xor with itself to set to zero faster.*/
+ dmawritebit ^= dmawritebit;
+ return ret;
+ case 23:
+ READ16(addr, emu8k->smarw | dmawritebit);
+ /*xor with itself to set to zero faster.*/
+ dmawritebit ^= dmawritebit;
+ return ret;
+
+ case 26:
+ {
+ uint16_t val = emu8k->smrd_buffer;
+ emu8k->smrd_buffer = EMU8K_READ(emu8k, emu8k->smarr);
+ emu8k->smarr = (emu8k->smarr + 1) & EMU8K_MEM_ADDRESS_MASK;
+ return val;
+ }
+ /*TODO: We need to improve the precision of this clock, since
+ it is used by programs to wait. Not critical, but should help reduce
+ the amount of calls and wait time */
+ case 27: /*Sample Counter ( 44Khz clock) */
+ return emu8k->wc;
+ }
+ break;
+
+ case 2:
+ return emu8k->init2[emu8k->cur_voice];
+
+ case 3:
+ return emu8k->init4[emu8k->cur_voice];
+
+ case 4:
+ return emu8k->voice[emu8k->cur_voice].atkhldv;
+
+ case 5:
+ return emu8k->voice[emu8k->cur_voice].lfo1val;
+
+ case 6:
+ return emu8k->voice[emu8k->cur_voice].atkhld;
+
+ case 7:
+ return emu8k->voice[emu8k->cur_voice].lfo2val;
+ }
+ break;
+
+ case 0xE00: /*Data3. also known as BLASTER+0xC00 and EMU+0x800 */
+ switch (emu8k->cur_reg)
+ {
+ case 0:
+ return emu8k->voice[emu8k->cur_voice].ip;
+
+ case 1:
+ return emu8k->voice[emu8k->cur_voice].ifatn;
+
+ case 2:
+ return emu8k->voice[emu8k->cur_voice].pefe;
+
+ case 3:
+ return emu8k->voice[emu8k->cur_voice].fmmod;
+
+ case 4:
+ return emu8k->voice[emu8k->cur_voice].tremfrq;
+
+ case 5:
+ return emu8k->voice[emu8k->cur_voice].fm2frq2;
+
+ case 6:
+ return 0xffff;
+
+ case 7: /*ID?*/
+ return 0x1c | ((emu8k->id & 0x0002) ? 0xff02 : 0);
+ }
+ break;
+
+ case 0xE02: /* Pointer. also known as BLASTER+0xC02 and EMU+0x802 */
+ /* LS five bits = channel number, next 3 bits = register number
+ * and MS 8 bits = VLSI test register.
+ * Impulse tracker tests the non variability of the LS byte that it has set, and the variability
+ * of the MS byte to determine that it really is an AWE32.
+ * cubic player has a similar code, where it waits until value & 0x1000 is nonzero, and then waits again until it changes to zero.*/
+ random_helper = (random_helper + 1) & 0x1F;
+ return ((0x80 | random_helper) << 8) | (emu8k->cur_reg << 5) | emu8k->cur_voice;
+ }
+ pclog("EMU8K READ : Unknown register read: %04X-%02X(%d/%d) \n", addr, (emu8k->cur_reg << 5) | emu8k->cur_voice, emu8k->cur_reg, emu8k->cur_voice);
+ return 0xffff;
+}
+
+void emu8k_outw(emu8k_t *emu8k, uint16_t addr, uint16_t val)
+{
+ /*TODO: I would like to not call this here, but i found it was needed or else cubic player would not finish opening (take a looot more of time than usual).
+ * Basically, being here means that the audio is generated in the emulation thread, instead of the audio thread.*/
+ // TODO emu8k_update(emu8k);
+
+#ifdef EMU8K_DEBUG_REGISTERS
+ if (addr == 0xE22)
+ {
+ //pclog("EMU8K WRITE POINTER: %d\n", val);
+ }
+ else if ((addr&0xF00) == 0x600)
+ {
+ /* These are automatically reported by WRITE16 */
+ if (rep_count_w>1)
+ {
+ pclog("EMU8K ...... for %d times\n", rep_count_w);
+ rep_count_w=0;
+ }
+ last_write=0;
+ }
+ else if ((addr&0xF00) == 0xA00 && emu8k->cur_reg == 0)
+ {
+ /* These are automatically reported by WRITE16 */
+ if (rep_count_w>1)
+ {
+ pclog("EMU8K ...... for %d times\n", rep_count_w);
+ rep_count_w=0;
+ }
+ last_write=0;
+ }
+ else if ((addr&0xF00) == 0xA00 && emu8k->cur_reg == 1)
+ {
+ uint32_t tmpz = ((addr&0xF00) << 16)|(emu8k->cur_reg<<5);
+ if (tmpz != last_write)
+ {
+ if (rep_count_w>1)
+ {
+ pclog("EMU8K ...... for %d times\n", rep_count_w);
+ rep_count_w=0;
+ }
+ last_write=tmpz;
+ pclog("EMU8K WRITE RAM I/O or configuration \n");
+ }
+ //pclog("EMU8K WRITE %04X-%02X(%d/%d): %04X\n",addr,(emu8k->cur_reg)<<5|emu8k->cur_voice,emu8k->cur_reg,emu8k->cur_voice, val);
+ }
+ else if ((addr&0xF00) == 0xA00 && (emu8k->cur_reg == 2 || emu8k->cur_reg == 3))
+ {
+ uint32_t tmpz = ((addr&0xF00) << 16);
+ if (tmpz != last_write)
+ {
+ if (rep_count_w>1)
+ {
+ pclog("EMU8K ...... for %d times\n", rep_count_w);
+ rep_count_w=0;
+ }
+ last_write=tmpz;
+ pclog("EMU8K WRITE INIT \n");
+ }
+ //pclog("EMU8K WRITE %04X-%02X(%d/%d): %04X\n",addr,(emu8k->cur_reg)<<5|emu8k->cur_voice,emu8k->cur_reg,emu8k->cur_voice, val);
+ }
+ else if (addr != 0xE22)
+ {
+ uint32_t tmpz = (addr << 16)|(emu8k->cur_reg<<5)| emu8k->cur_voice;
+ //if (tmpz != last_write)
+ if(1)
+ {
+ char* name = 0;
+ if (addr == 0xA20)
+ {
+ name = PORT_NAMES[1][emu8k->cur_reg];
+ }
+ else if (addr == 0xA22)
+ {
+ name = PORT_NAMES[2][emu8k->cur_reg];
+ }
+ else if (addr == 0xE20)
+ {
+ name = PORT_NAMES[3][emu8k->cur_reg];
+ }
+
+ if (rep_count_w>1)
+ {
+ pclog("EMU8K ...... for %d times\n", rep_count_w);
+ }
+ if (name == 0)
+ {
+ pclog("EMU8K WRITE %04X-%02X(%d/%d): %04X\n",addr,(emu8k->cur_reg)<<5|emu8k->cur_voice,emu8k->cur_reg,emu8k->cur_voice, val);
+ }
+ else
+ {
+ pclog("EMU8K WRITE %s (%d): %04X\n",name,emu8k->cur_voice, val);
+ }
+
+ rep_count_w=0;
+ last_write=tmpz;
+ }
+ rep_count_w++;
+ }
+#endif //EMU8K_DEBUG_REGISTERS
+
+ switch (addr & 0xF02)
+ {
+ case 0x600:
+ case 0x602: /*Data0. also known as BLASTER+0x400 and EMU+0x000 */
+ switch (emu8k->cur_reg)
+ {
+ case 0:
+ /* The docs says that this value is constantly updating, and it should have no actual effect. Actions should be done over ptrx */
+ WRITE16(addr, emu8k->voice[emu8k->cur_voice].cpf, val);
+ return;
+
+ case 1:
+ WRITE16(addr, emu8k->voice[emu8k->cur_voice].ptrx, val);
+ return;
+
+ case 2:
+ /* The docs says that this value is constantly updating, and it should have no actual effect. Actions should be done over vtft */
+ WRITE16(addr, emu8k->voice[emu8k->cur_voice].cvcf, val);
+ return;
+
+ case 3:
+ WRITE16(addr, emu8k->voice[emu8k->cur_voice].vtft, val);
+ return;
+
+ case 4:
+ WRITE16(addr, emu8k->voice[emu8k->cur_voice].unknown_data0_4, val);
+ return;
+
+ case 5:
+ WRITE16(addr, emu8k->voice[emu8k->cur_voice].unknown_data0_5, val);
+ return;
+
+ case 6:
+ {
+ emu8k_voice_t* emu_voice = &emu8k->voice[emu8k->cur_voice];
+ WRITE16(addr, emu_voice->psst, val);
+ /* TODO: Should we update only on MSB update, or this could be used as some sort of hack by applications? */
+ emu_voice->loop_start.int_address = emu_voice->psst & EMU8K_MEM_ADDRESS_MASK;
+ if (addr & 2)
+ {
+ emu_voice->vol_l = emu_voice->psst_pan;
+ emu_voice->vol_r = 255 - (emu_voice->psst_pan);
+ }
+ }
+ return;
+
+ case 7:
+ WRITE16(addr, emu8k->voice[emu8k->cur_voice].csl, val);
+ /* TODO: Should we update only on MSB update, or this could be used as some sort of hack by applications? */
+ emu8k->voice[emu8k->cur_voice].loop_end.int_address = emu8k->voice[emu8k->cur_voice].csl & EMU8K_MEM_ADDRESS_MASK;
+ return;
+ }
+ break;
+
+ case 0xA00: /*Data1. also known as BLASTER+0x800 and EMU+0x400 */
+ switch (emu8k->cur_reg)
+ {
+ case 0:
+ WRITE16(addr, emu8k->voice[emu8k->cur_voice].ccca, val);
+ /* TODO: Should we update only on MSB update, or this could be used as some sort of hack by applications? */
+ emu8k->voice[emu8k->cur_voice].addr.int_address = emu8k->voice[emu8k->cur_voice].ccca & EMU8K_MEM_ADDRESS_MASK;
+ return;
+
+ case 1:
+ switch (emu8k->cur_voice)
+ {
+ case 9:
+ WRITE16(addr, emu8k->hwcf4, val);
+ return;
+ case 10:
+ WRITE16(addr, emu8k->hwcf5, val);
+ return;
+ /* Actually, these two might be command words rather than registers, or some LFO position/buffer reset. */
+ case 13:
+ WRITE16(addr, emu8k->hwcf6, val);
+ return;
+ case 14:
+ WRITE16(addr, emu8k->hwcf7, val);
+ return;
+
+ case 20:
+ WRITE16(addr, emu8k->smalr, val);
+ return;
+ case 21:
+ WRITE16(addr, emu8k->smarr, val);
+ return;
+ case 22:
+ WRITE16(addr, emu8k->smalw, val);
+ return;
+ case 23:
+ WRITE16(addr, emu8k->smarw, val);
+ return;
+
+ case 26:
+ EMU8K_WRITE(emu8k, emu8k->smalw, val);
+ emu8k->smalw = (emu8k->smalw + 1) & EMU8K_MEM_ADDRESS_MASK;
+ return;
+
+ case 29:
+ emu8k->hwcf1 = val;
+ return;
+ case 30:
+ emu8k->hwcf2 = val;
+ return;
+ case 31:
+ emu8k->hwcf3 = val;
+ return;
+ }
+ break;
+
+ case 2:
+ emu8k->init1[emu8k->cur_voice] = val;
+ /* Skip if in first/second initialization step */
+ if (emu8k->init1[0] != 0x03FF)
+ {
+ switch (emu8k->cur_voice)
+ {
+ case 0x3:
+ emu8k->reverb_engine.out_mix = val & 0xFF;
+ break;
+ case 0x5:
+ {
+ int c;
+ for (c = 0; c < 8; c++)
+ {
+ emu8k->reverb_engine.allpass[c].feedback = (val & 0xFF) / ((float)0xFF);
+ }
+ }
+ break;
+ case 0x7:
+ emu8k->reverb_engine.link_return_type = (val == 0x8474) ? 1 : 0;
+ break;
+ case 0xF:
+ emu8k->reverb_engine.reflections[0].output_gain = ((val & 0xF0) >> 4) / 15.0;
+ break;
+ case 0x17:
+ emu8k->reverb_engine.reflections[1].output_gain = ((val & 0xF0) >> 4) / 15.0;
+ break;
+ case 0x1F:
+ emu8k->reverb_engine.reflections[2].output_gain = ((val & 0xF0) >> 4) / 15.0;
+ break;
+ case 0x9:
+ emu8k->reverb_engine.reflections[0].feedback = (val & 0xF) / 15.0;
+ break;
+ case 0xB: //emu8k->reverb_engine.reflections[0].feedback_r = (val&0xF)/15.0;
+ break;
+ case 0x11:
+ emu8k->reverb_engine.reflections[1].feedback = (val & 0xF) / 15.0;
+ break;
+ case 0x13: //emu8k->reverb_engine.reflections[1].feedback_r = (val&0xF)/15.0;
+ break;
+ case 0x19:
+ emu8k->reverb_engine.reflections[2].feedback = (val & 0xF) / 15.0;
+ break;
+ case 0x1B: //emu8k->reverb_engine.reflections[2].feedback_r = (val&0xF)/15.0;
+ break;
+ }
+ }
+ return;
+
+ case 3:
+ emu8k->init3[emu8k->cur_voice] = val;
+ /* Skip if in first/second initialization step */
+ if (emu8k->init1[0] != 0x03FF)
+ {
+ switch (emu8k->cur_voice)
+ {
+ case 9:
+ emu8k->chorus_engine.feedback = (val & 0xFF);
+ break;
+ case 12:
+ /* Limiting this to a sane value given our buffer. */
+ emu8k->chorus_engine.delay_samples_central = (val & 0x1FFF);
+ break;
+
+ case 1:
+ emu8k->reverb_engine.refl_in_amp = val & 0xFF;
+ break;
+ case 3: //emu8k->reverb_engine.refl_in_amp_r = val&0xFF;
+ break;
+ }
+ }
+ return;
+
+ case 4:
+ emu8k->voice[emu8k->cur_voice].envvol = val;
+ emu8k->voice[emu8k->cur_voice].vol_envelope.delay_samples = ENVVOL_TO_EMU_SAMPLES(val);
+ return;
+
+ case 5:
+ {
+ emu8k->voice[emu8k->cur_voice].dcysusv = val;
+ emu8k_envelope_t* const vol_env = &emu8k->voice[emu8k->cur_voice].vol_envelope;
+ int old_on = emu8k->voice[emu8k->cur_voice].env_engine_on;
+ emu8k->voice[emu8k->cur_voice].env_engine_on = DCYSUSV_GENERATOR_ENGINE_ON(val);
+
+ if (emu8k->voice[emu8k->cur_voice].env_engine_on &&
+ old_on != emu8k->voice[emu8k->cur_voice].env_engine_on)
+ {
+ if (emu8k->hwcf3 != 0x04)
+ {
+ /* This is a hack for some programs like Doom or cubic player 1.7 that don't initialize
+ the hwcfg and init registers (doom does not init the card at all. only tests the cfg registers) */
+ emu8k->hwcf3 = 0x04;
+ }
+
+ //reset lfos.
+ emu8k->voice[emu8k->cur_voice].lfo1_count.addr = 0;
+ emu8k->voice[emu8k->cur_voice].lfo2_count.addr = 0;
+ // Trigger envelopes
+ if (ATKHLDV_TRIGGER(emu8k->voice[emu8k->cur_voice].atkhldv))
+ {
+ vol_env->value_amp_hz = 0;
+ if (vol_env->delay_samples)
+ {
+ vol_env->state = ENV_DELAY;
+ }
+ else if (vol_env->attack_amount_amp_hz == 0)
+ {
+ vol_env->state = ENV_STOPPED;
+ }
+ else
+ {
+ vol_env->state = ENV_ATTACK;
+ /* TODO: Verify if "never attack" means eternal mute,
+ * or it means skip attack, go to hold".
+ if (vol_env->attack_amount == 0)
+ {
+ vol_env->value = (1 << 21);
+ vol_env->state = ENV_HOLD;
+ }*/
+ }
+ }
+
+ if (ATKHLD_TRIGGER(emu8k->voice[emu8k->cur_voice].atkhld))
+ {
+ emu8k_envelope_t* const mod_env = &emu8k->voice[emu8k->cur_voice].mod_envelope;
+ mod_env->value_amp_hz = 0;
+ mod_env->value_db_oct = 0;
+ if (mod_env->delay_samples)
+ {
+ mod_env->state = ENV_DELAY;
+ }
+ else if (mod_env->attack_amount_amp_hz == 0)
+ {
+ mod_env->state = ENV_STOPPED;
+ }
+ else
+ {
+ mod_env->state = ENV_ATTACK;
+ /* TODO: Verify if "never attack" means eternal start,
+ * or it means skip attack, go to hold".
+ if (mod_env->attack_amount == 0)
+ {
+ mod_env->value = (1 << 21);
+ mod_env->state = ENV_HOLD;
+ }*/
+ }
+ }
+ }
+
+
+ /* Converting the input in dBs to envelope value range. */
+ vol_env->sustain_value_db_oct = DCYSUSV_SUS_TO_ENV_RANGE(DCYSUSV_SUSVALUE_GET(val));
+ vol_env->ramp_amount_db_oct = env_decay_to_dbs_or_oct[DCYSUSV_DECAYRELEASE_GET(val)];
+ if (DCYSUSV_IS_RELEASE(val))
+ {
+ if (vol_env->state == ENV_DELAY || vol_env->state == ENV_ATTACK || vol_env->state == ENV_HOLD)
+ {
+ vol_env->value_db_oct = env_vol_amplitude_to_db[vol_env->value_amp_hz >> 5] << 5;
+ if (vol_env->value_db_oct > (1 << 21))
+ vol_env->value_db_oct = 1 << 21;
+ }
+
+ vol_env->state = (vol_env->value_db_oct >= vol_env->sustain_value_db_oct) ? ENV_RAMP_DOWN : ENV_RAMP_UP;
+ }
+ }
+ return;
+
+ case 6:
+ emu8k->voice[emu8k->cur_voice].envval = val;
+ emu8k->voice[emu8k->cur_voice].mod_envelope.delay_samples = ENVVAL_TO_EMU_SAMPLES(val);
+ return;
+
+ case 7:
+ {
+ //TODO: Look for a bug on delay (first trigger it works, next trigger it doesn't)
+ emu8k->voice[emu8k->cur_voice].dcysus = val;
+ emu8k_envelope_t* const mod_env = &emu8k->voice[emu8k->cur_voice].mod_envelope;
+ /* Converting the input in octaves to envelope value range. */
+ mod_env->sustain_value_db_oct = DCYSUS_SUS_TO_ENV_RANGE(DCYSUS_SUSVALUE_GET(val));
+ mod_env->ramp_amount_db_oct = env_decay_to_dbs_or_oct[DCYSUS_DECAYRELEASE_GET(val)];
+ if (DCYSUS_IS_RELEASE(val))
+ {
+ if (mod_env->state == ENV_DELAY || mod_env->state == ENV_ATTACK || mod_env->state == ENV_HOLD)
+ {
+ mod_env->value_db_oct = env_mod_hertz_to_octave[mod_env->value_amp_hz >> 9] << 9;
+ if (mod_env->value_db_oct >= (1 << 21))
+ mod_env->value_db_oct = (1 << 21) - 1;
+ }
+
+ mod_env->state = (mod_env->value_db_oct >= mod_env->sustain_value_db_oct) ? ENV_RAMP_DOWN : ENV_RAMP_UP;
+ }
+ }
+ return;
+ }
+ break;
+
+ case 0xA02: /*Data2. also known as BLASTER+0x802 and EMU+0x402 */
+ switch (emu8k->cur_reg)
+ {
+ case 0:
+ {
+ emu8k_voice_t* emu_voice = &emu8k->voice[emu8k->cur_voice];
+ WRITE16(addr, emu_voice->ccca, val);
+ emu_voice->addr.int_address = emu_voice->ccca & EMU8K_MEM_ADDRESS_MASK;
+ uint32_t paramq = CCCA_FILTQ_GET(emu_voice->ccca);
+ emu_voice->filt_att = filter_atten[paramq];
+ emu_voice->filterq_idx = paramq;
+ }
+ return;
+
+ case 1:
+ switch (emu8k->cur_voice)
+ {
+ case 9:
+ WRITE16(addr, emu8k->hwcf4, val);
+ /* Skip if in first/second initialization step */
+ if (emu8k->init1[0] != 0x03FF)
+ {
+ /*(1/256th of a 44Khz sample) */
+ /* clip the value to a reasonable value given our buffer */
+ int32_t tmp = emu8k->hwcf4 & 0x1FFFFF;
+ emu8k->chorus_engine.delay_offset_samples_right = ((double)tmp) / 256.0;
+ }
+ return;
+ case 10:
+ WRITE16(addr, emu8k->hwcf5, val);
+ /* Skip if in first/second initialization step */
+ if (emu8k->init1[0] != 0x03FF)
+ {
+ /* The scale of this value is unknown. I've taken it as milliHz.
+ * Another interpretation could be periods. (and so, Hz = 1/period)*/
+ double osc_speed = emu8k->hwcf5;//*1.316;
+#if 1 // milliHz
+ /*milliHz to lfotable samples.*/
+ osc_speed *= 65.536 / 44100.0;
+#elif 0 //periods
+ /* 44.1Khz ticks to lfotable samples.*/
+ osc_speed = 65.536/osc_speed;
+#endif
+ /*left shift 32bits for 32.32 fixed.point*/
+ osc_speed *= 65536.0 * 65536.0;
+ emu8k->chorus_engine.lfo_inc.addr = (uint64_t)osc_speed;
+ }
+ return;
+ /* Actually, these two might be command words rather than registers, or some LFO position/buffer reset.*/
+ case 13:
+ WRITE16(addr, emu8k->hwcf6, val);
+ return;
+ case 14:
+ WRITE16(addr, emu8k->hwcf7, val);
+ return;
+
+ case 20: /*Top 8 bits are for Empty (MT) bit or non-addressable.*/
+ WRITE16(addr, emu8k->smalr, val & 0xFF);
+ dmareadbit = 0x8000;
+ return;
+ case 21: /*Top 8 bits are for Empty (MT) bit or non-addressable.*/
+ WRITE16(addr, emu8k->smarr, val & 0xFF);
+ dmareadbit = 0x8000;
+ return;
+ case 22: /*Top 8 bits are for full bit or non-addressable.*/
+ WRITE16(addr, emu8k->smalw, val & 0xFF);
+ return;
+ case 23: /*Top 8 bits are for full bit or non-addressable.*/
+ WRITE16(addr, emu8k->smarw, val & 0xFF);
+ return;
+
+ case 26:
+ dmawritebit = 0x8000;
+ EMU8K_WRITE(emu8k, emu8k->smarw, val);
+ emu8k->smarw++;
+ return;
+ }
+ break;
+
+ case 2:
+ emu8k->init2[emu8k->cur_voice] = val;
+ /* Skip if in first/second initialization step */
+ if (emu8k->init1[0] != 0x03FF)
+ {
+ switch (emu8k->cur_voice)
+ {
+ case 0x14:
+ {
+ int multip = ((val & 0xF00) >> 8) + 18;
+ emu8k->reverb_engine.reflections[5].bufsize = multip * REV_BUFSIZE_STEP;
+ emu8k->reverb_engine.tailL.bufsize = (multip + 1) * REV_BUFSIZE_STEP;
+ if (emu8k->reverb_engine.link_return_type == 0)
+ {
+ emu8k->reverb_engine.tailR.bufsize = (multip + 1) * REV_BUFSIZE_STEP;
+ }
+ }
+ break;
+ case 0x16:
+ if (emu8k->reverb_engine.link_return_type == 1)
+ {
+ int multip = ((val & 0xF00) >> 8) + 18;
+ emu8k->reverb_engine.tailR.bufsize = (multip + 1) * REV_BUFSIZE_STEP;
+ }
+ break;
+ case 0x7:
+ emu8k->reverb_engine.reflections[3].output_gain = ((val & 0xF0) >> 4) / 15.0;
+ break;
+ case 0xf:
+ emu8k->reverb_engine.reflections[4].output_gain = ((val & 0xF0) >> 4) / 15.0;
+ break;
+ case 0x17:
+ emu8k->reverb_engine.reflections[5].output_gain = ((val & 0xF0) >> 4) / 15.0;
+ break;
+ case 0x1d:
+ {
+ int c;
+ for (c = 0; c < 6; c++)
+ {
+ emu8k->reverb_engine.reflections[c].damp1 = (val & 0xFF) / 255.0;
+ emu8k->reverb_engine.reflections[c].damp2 = (0xFF - (val & 0xFF)) / 255.0;
+ emu8k->reverb_engine.reflections[c].filterstore = 0;
+ }
+ emu8k->reverb_engine.damper.damp1 = (val & 0xFF) / 255.0;
+ emu8k->reverb_engine.damper.damp2 = (0xFF - (val & 0xFF)) / 255.0;
+ emu8k->reverb_engine.damper.filterstore = 0;
+ }
+ break;
+ case 0x1f: /* filter r */
+ break;
+ case 0x1:
+ emu8k->reverb_engine.reflections[3].feedback = (val & 0xF) / 15.0;
+ break;
+ case 0x3: //emu8k->reverb_engine.reflections[3].feedback_r = (val&0xF)/15.0;
+ break;
+ case 0x9:
+ emu8k->reverb_engine.reflections[4].feedback = (val & 0xF) / 15.0;
+ break;
+ case 0xb: //emu8k->reverb_engine.reflections[4].feedback_r = (val&0xF)/15.0;
+ break;
+ case 0x11:
+ emu8k->reverb_engine.reflections[5].feedback = (val & 0xF) / 15.0;
+ break;
+ case 0x13: //emu8k->reverb_engine.reflections[5].feedback_r = (val&0xF)/15.0;
+ break;
+ }
+ }
+ return;
+
+ case 3:
+ emu8k->init4[emu8k->cur_voice] = val;
+ /* Skip if in first/second initialization step */
+ if (emu8k->init1[0] != 0x03FF)
+ {
+ switch (emu8k->cur_voice)
+ {
+ case 0x3:
+ {
+ int32_t samples = ((val & 0xFF) * emu8k->chorus_engine.delay_samples_central) >> 8;
+ emu8k->chorus_engine.lfodepth_multip = samples;
+
+ }
+ break;
+
+ case 0x1F:
+ emu8k->reverb_engine.link_return_amp = val & 0xFF;
+ break;
+ }
+ }
+ return;
+
+ case 4:
+ {
+ emu8k->voice[emu8k->cur_voice].atkhldv = val;
+ emu8k_envelope_t* const vol_env = &emu8k->voice[emu8k->cur_voice].vol_envelope;
+ vol_env->attack_samples = env_attack_to_samples[ATKHLDV_ATTACK(val)];
+ if (vol_env->attack_samples == 0)
+ {
+ vol_env->attack_amount_amp_hz = 0;
+ }
+ else
+ {
+ /* Linear amplitude increase each sample. */
+ vol_env->attack_amount_amp_hz = (1 << 21) / vol_env->attack_samples;
+ }
+ vol_env->hold_samples = ATKHLDV_HOLD_TO_EMU_SAMPLES(val);
+ if (ATKHLDV_TRIGGER(val) && emu8k->voice[emu8k->cur_voice].env_engine_on)
+ {
+ /*TODO: I assume that "envelope trigger" is the same as new note
+ * (since changing the IP can be done when modulating pitch too) */
+ emu8k->voice[emu8k->cur_voice].lfo1_count.addr = 0;
+ emu8k->voice[emu8k->cur_voice].lfo2_count.addr = 0;
+
+ vol_env->value_amp_hz = 0;
+ if (vol_env->delay_samples)
+ {
+ vol_env->state = ENV_DELAY;
+ }
+ else if (vol_env->attack_amount_amp_hz == 0)
+ {
+ vol_env->state = ENV_STOPPED;
+ }
+ else
+ {
+ vol_env->state = ENV_ATTACK;
+ /* TODO: Verify if "never attack" means eternal mute,
+ * or it means skip attack, go to hold".
+ if (vol_env->attack_amount == 0)
+ {
+ vol_env->value = (1 << 21);
+ vol_env->state = ENV_HOLD;
+ }*/
+ }
+ }
+ }
+ return;
+
+ case 5:
+ emu8k->voice[emu8k->cur_voice].lfo1val = val;
+ /* TODO: verify if this is set once, or set every time. */
+ emu8k->voice[emu8k->cur_voice].lfo1_delay_samples = LFOxVAL_TO_EMU_SAMPLES(val);
+ return;
+
+ case 6:
+ {
+ emu8k->voice[emu8k->cur_voice].atkhld = val;
+ emu8k_envelope_t* const mod_env = &emu8k->voice[emu8k->cur_voice].mod_envelope;
+ mod_env->attack_samples = env_attack_to_samples[ATKHLD_ATTACK(val)];
+ if (mod_env->attack_samples == 0)
+ {
+ mod_env->attack_amount_amp_hz = 0;
+ }
+ else
+ {
+ /* Linear amplitude increase each sample. */
+ mod_env->attack_amount_amp_hz = (1 << 21) / mod_env->attack_samples;
+ }
+ mod_env->hold_samples = ATKHLD_HOLD_TO_EMU_SAMPLES(val);
+ if (ATKHLD_TRIGGER(val) && emu8k->voice[emu8k->cur_voice].env_engine_on)
+ {
+ mod_env->value_amp_hz = 0;
+ mod_env->value_db_oct = 0;
+ if (mod_env->delay_samples)
+ {
+ mod_env->state = ENV_DELAY;
+ }
+ else if (mod_env->attack_amount_amp_hz == 0)
+ {
+ mod_env->state = ENV_STOPPED;
+ }
+ else
+ {
+ mod_env->state = ENV_ATTACK;
+ /* TODO: Verify if "never attack" means eternal start,
+ * or it means skip attack, go to hold".
+ if (mod_env->attack_amount == 0)
+ {
+ mod_env->value = (1 << 21);
+ mod_env->state = ENV_HOLD;
+ }*/
+ }
+ }
+ }
+ return;
+
+ case 7:
+ emu8k->voice[emu8k->cur_voice].lfo2val = val;
+ emu8k->voice[emu8k->cur_voice].lfo2_delay_samples = LFOxVAL_TO_EMU_SAMPLES(val);
+
+ return;
+ }
+ break;
+
+ case 0xE00: /*Data3. also known as BLASTER+0xC00 and EMU+0x800 */
+ switch (emu8k->cur_reg)
+ {
+ case 0:
+ emu8k->voice[emu8k->cur_voice].ip = val;
+ emu8k->voice[emu8k->cur_voice].ptrx_pit_target = freqtable[val] >> 18;
+ return;
+
+ case 1:
+ {
+ emu8k_voice_t* const the_voice = &emu8k->voice[emu8k->cur_voice];
+ if ((val & 0xFF) == 0 && the_voice->cvcf_curr_volume == 0 && the_voice->vtft_vol_target == 0
+ && the_voice->dcysusv == 0x80 && the_voice->ip == 0)
+ {
+ // Patch to avoid some clicking noises with Impulse tracker or other software that sets
+ // different values to 0 to set noteoff, but here, 0 means no attenuation = full volume.
+ return;
+ }
+ the_voice->ifatn = val;
+ the_voice->initial_att = (((int32_t)the_voice->ifatn_attenuation << 21) / 0xFF);
+ the_voice->vtft_vol_target = attentable[the_voice->ifatn_attenuation];
+
+ the_voice->initial_filter = (((int32_t)the_voice->ifatn_init_filter << 21) / 0xFF);
+ if (the_voice->ifatn_init_filter == 0xFF)
+ {
+ the_voice->vtft_filter_target = 0xFFFF;
+ }
+ else
+ {
+ the_voice->vtft_filter_target = the_voice->initial_filter >> 5;
+ }
+ }
+ return;
+
+ case 2:
+ {
+ emu8k_voice_t* const the_voice = &emu8k->voice[emu8k->cur_voice];
+ the_voice->pefe = val;
+
+ int divider = (the_voice->pefe_modenv_filter_height < 0) ? 0x80 : 0x7F;
+ the_voice->fixed_modenv_filter_height = ((int32_t)the_voice->pefe_modenv_filter_height) * 0x4000 / divider;
+
+ divider = (the_voice->pefe_modenv_pitch_height < 0) ? 0x80 : 0x7F;
+ the_voice->fixed_modenv_pitch_height = ((int32_t)the_voice->pefe_modenv_pitch_height) * 0x4000 / divider;
+ }
+ return;
+
+ case 3:
+ {
+ emu8k_voice_t* const the_voice = &emu8k->voice[emu8k->cur_voice];
+ the_voice->fmmod = val;
+
+ int divider = (the_voice->fmmod_lfo1_filt_mod < 0) ? 0x80 : 0x7F;
+ the_voice->fixed_lfo1_filt_mod = ((int32_t)the_voice->fmmod_lfo1_filt_mod) * 0x4000 / divider;
+
+ divider = (the_voice->fmmod_lfo1_vibrato < 0) ? 0x80 : 0x7F;
+ the_voice->fixed_lfo1_vibrato = ((int32_t)the_voice->fmmod_lfo1_vibrato) * 0x4000 / divider;
+ }
+ return;
+
+ case 4:
+ {
+ emu8k_voice_t* const the_voice = &emu8k->voice[emu8k->cur_voice];
+ the_voice->tremfrq = val;
+ the_voice->lfo1_speed = lfofreqtospeed[the_voice->tremfrq_lfo1_freq];
+
+ int divider = (the_voice->tremfrq_lfo1_tremolo < 0) ? 0x80 : 0x7F;
+ the_voice->fixed_lfo1_tremolo = ((int32_t)the_voice->tremfrq_lfo1_tremolo) * 0x4000 / divider;
+ }
+ return;
+
+ case 5:
+ {
+ emu8k_voice_t* const the_voice = &emu8k->voice[emu8k->cur_voice];
+ the_voice->fm2frq2 = val;
+ the_voice->lfo2_speed = lfofreqtospeed[the_voice->fm2frq2_lfo2_freq];
+
+ int divider = (the_voice->fm2frq2_lfo2_vibrato < 0) ? 0x80 : 0x7F;
+ the_voice->fixed_lfo2_vibrato = ((int32_t)the_voice->fm2frq2_lfo2_vibrato) * 0x4000 / divider;
+ }
+ return;
+
+ case 7: /*ID? I believe that this allows applications to know if the emu is in use by another application */
+ emu8k->id = val;
+ return;
+ }
+ break;
+
+ case 0xE02: /* Pointer. also known as BLASTER+0xC02 and EMU+0x802 */
+ emu8k->cur_voice = (val & 31);
+ emu8k->cur_reg = ((val >> 5) & 7);
+ return;
+ }
+ pclog("EMU8K WRITE: Unknown register write: %04X-%02X(%d/%d): %04X \n", addr, (emu8k->cur_reg) << 5 | emu8k->cur_voice,
+ emu8k->cur_reg, emu8k->cur_voice, val);
+
+}
+
+uint8_t emu8k_inb(emu8k_t *emu8k, uint16_t addr)
+{
+ /* Reading a single byte is a feature that at least Impulse tracker uses,
+ * but only on detection code and not for odd addresses.*/
+ if (addr & 1)
+ return emu8k_inw(emu8k, addr & ~1) >> 1;
+ return emu8k_inw(emu8k, addr) & 0xff;
+}
+
+void emu8k_outb(emu8k_t *emu8k, uint16_t addr, uint8_t val)
+{
+ /* TODO: AWE32 docs says that you cannot write in bytes, but if
+ * an app were to use this implementation, the content of the LS Byte would be lost.*/
+ if (addr & 1)
+ emu8k_outw(emu8k, addr & ~1, val << 8);
+ else
+ emu8k_outw(emu8k, addr, val);
+}
+
+/* TODO: This is not a correct emulation, just a workalike implementation. */
+void emu8k_work_chorus(int32_t* inbuf, int32_t* outbuf, emu8k_chorus_eng_t* engine, int count)
+{
+ int pos;
+ for (pos = 0; pos < count; pos++)
+ {
+ double lfo_inter1 = chortable[engine->lfo_pos.int_address];
+ // double lfo_inter2 = chortable[(engine->lfo_pos.int_address+1)&0xFFFF];
+
+ double offset_lfo = lfo_inter1; //= lfo_inter1 + ((lfo_inter2-lfo_inter1)*engine->lfo_pos.fract_address/65536.0);
+ offset_lfo *= engine->lfodepth_multip;
+
+ /* Work left */
+ double readdouble = (double)engine->write - (double)engine->delay_samples_central - offset_lfo;
+ int read = (int32_t)floor(readdouble);
+ int fraction_part = (readdouble - (double)read) * 65536.0;
+ int next_value = read + 1;
+ if (read < 0)
+ {
+ read += EMU8K_LFOCHORUS_SIZE;
+ if (next_value < 0) next_value += EMU8K_LFOCHORUS_SIZE;
+ }
+ else if (next_value >= EMU8K_LFOCHORUS_SIZE)
+ {
+ next_value -= EMU8K_LFOCHORUS_SIZE;
+ if (read >= EMU8K_LFOCHORUS_SIZE) read -= EMU8K_LFOCHORUS_SIZE;
+ }
+ int32_t dat1 = engine->chorus_left_buffer[read];
+ int32_t dat2 = engine->chorus_left_buffer[next_value];
+ dat1 += ((dat2 - dat1) * fraction_part) >> 16;
+
+ engine->chorus_left_buffer[engine->write] = *inbuf + ((dat1 * engine->feedback) >> 8);
+
+
+ /* Work right */
+ readdouble = (double)engine->write - (double)engine->delay_samples_central - engine->delay_offset_samples_right - offset_lfo;
+ read = (int32_t)floor(readdouble);
+ next_value = read + 1;
+ if (read < 0)
+ {
+ read += EMU8K_LFOCHORUS_SIZE;
+ if (next_value < 0) next_value += EMU8K_LFOCHORUS_SIZE;
+ }
+ else if (next_value >= EMU8K_LFOCHORUS_SIZE)
+ {
+ next_value -= EMU8K_LFOCHORUS_SIZE;
+ if (read >= EMU8K_LFOCHORUS_SIZE) read -= EMU8K_LFOCHORUS_SIZE;
+ }
+ int32_t dat3 = engine->chorus_right_buffer[read];
+ int32_t dat4 = engine->chorus_right_buffer[next_value];
+ dat3 += ((dat4 - dat3) * fraction_part) >> 16;
+
+ engine->chorus_right_buffer[engine->write] = *inbuf + ((dat3 * engine->feedback) >> 8);
+
+ ++engine->write;
+ engine->write %= EMU8K_LFOCHORUS_SIZE;
+ engine->lfo_pos.addr += engine->lfo_inc.addr;
+ engine->lfo_pos.int_address &= 0xFFFF;
+
+ (*outbuf++) += dat1;
+ (*outbuf++) += dat3;
+ inbuf++;
+ }
+
+}
+
+int32_t emu8k_reverb_comb_work(emu8k_reverb_combfilter_t* comb, int32_t in)
+{
+
+ int32_t bufin;
+ /* get echo */
+ int32_t output = comb->reflection[comb->read_pos];
+ /* apply lowpass */
+ comb->filterstore = (output * comb->damp2) + (comb->filterstore * comb->damp1);
+ /* appply feedback */
+ bufin = in - (comb->filterstore * comb->feedback);
+ /* store new value in delayed buffer */
+ comb->reflection[comb->read_pos] = bufin;
+
+ if (++comb->read_pos >= comb->bufsize) comb->read_pos = 0;
+
+ return output * comb->output_gain;
+}
+
+int32_t emu8k_reverb_diffuser_work(emu8k_reverb_combfilter_t* comb, int32_t in)
+{
+
+ int32_t bufout = comb->reflection[comb->read_pos];
+ /*diffuse*/
+ int32_t bufin = -in + (bufout * comb->feedback);
+ int32_t output = bufout - (bufin * comb->feedback);
+ /* store new value in delayed buffer */
+ comb->reflection[comb->read_pos] = bufin;
+
+ if (++comb->read_pos >= comb->bufsize) comb->read_pos = 0;
+
+ return output;
+}
+
+int32_t emu8k_reverb_tail_work(emu8k_reverb_combfilter_t* comb, emu8k_reverb_combfilter_t* allpasses, int32_t in)
+{
+ int32_t output = comb->reflection[comb->read_pos];
+ /* store new value in delayed buffer */
+ comb->reflection[comb->read_pos] = in;
+
+ //output = emu8k_reverb_allpass_work(&allpasses[0],output);
+ output = emu8k_reverb_diffuser_work(&allpasses[1], output);
+ output = emu8k_reverb_diffuser_work(&allpasses[2], output);
+ //output = emu8k_reverb_allpass_work(&allpasses[3],output);
+
+ if (++comb->read_pos >= comb->bufsize) comb->read_pos = 0;
+
+ return output;
+}
+int32_t emu8k_reverb_damper_work(emu8k_reverb_combfilter_t* comb, int32_t in)
+{
+ /* apply lowpass */
+ comb->filterstore = (in * comb->damp2) + (comb->filterstore * comb->damp1);
+ return comb->filterstore;
+}
+
+/* TODO: This is not a correct emulation, just a workalike implementation. */
+void emu8k_work_reverb(int32_t* inbuf, int32_t* outbuf, emu8k_reverb_eng_t* engine, int count)
+{
+ int pos;
+ if (engine->link_return_type)
+ {
+ for (pos = 0; pos < count; pos++)
+ {
+ int32_t dat1, dat2, in, in2;
+ in = emu8k_reverb_damper_work(&engine->damper, inbuf[pos]);
+ in2 = (in * engine->refl_in_amp) >> 8;
+ dat2 = emu8k_reverb_comb_work(&engine->reflections[0], in2);
+ dat2 += emu8k_reverb_comb_work(&engine->reflections[1], in2);
+ dat1 = emu8k_reverb_comb_work(&engine->reflections[2], in2);
+ dat2 += emu8k_reverb_comb_work(&engine->reflections[3], in2);
+ dat1 += emu8k_reverb_comb_work(&engine->reflections[4], in2);
+ dat2 += emu8k_reverb_comb_work(&engine->reflections[5], in2);
+
+ dat1 += (emu8k_reverb_tail_work(&engine->tailL, &engine->allpass[0], in + dat1) * engine->link_return_amp) >> 8;
+ dat2 += (emu8k_reverb_tail_work(&engine->tailR, &engine->allpass[4], in + dat2) * engine->link_return_amp) >> 8;
+
+ (*outbuf++) += (dat1 * engine->out_mix) >> 8;
+ (*outbuf++) += (dat2 * engine->out_mix) >> 8;
+ }
+ }
+ else
+ {
+ for (pos = 0; pos < count; pos++)
+ {
+ int32_t dat1, dat2, in, in2;
+ in = emu8k_reverb_damper_work(&engine->damper, inbuf[pos]);
+ in2 = (in * engine->refl_in_amp) >> 8;
+ dat1 = emu8k_reverb_comb_work(&engine->reflections[0], in2);
+ dat1 += emu8k_reverb_comb_work(&engine->reflections[1], in2);
+ dat1 += emu8k_reverb_comb_work(&engine->reflections[2], in2);
+ dat1 += emu8k_reverb_comb_work(&engine->reflections[3], in2);
+ dat1 += emu8k_reverb_comb_work(&engine->reflections[4], in2);
+ dat1 += emu8k_reverb_comb_work(&engine->reflections[5], in2);
+ dat2 = dat1;
+
+ dat1 += (emu8k_reverb_tail_work(&engine->tailL, &engine->allpass[0], in + dat1) * engine->link_return_amp) >> 8;
+ dat2 += (emu8k_reverb_tail_work(&engine->tailR, &engine->allpass[4], in + dat2) * engine->link_return_amp) >> 8;
+
+ (*outbuf++) += (dat1 * engine->out_mix) >> 8;
+ (*outbuf++) += (dat2 * engine->out_mix) >> 8;
+ }
+ }
+}
+void emu8k_work_eq(int32_t* inoutbuf, int count)
+{
+ // TODO: Work EQ over buf
+ NOREF(inoutbuf);
+ NOREF(count);
+}
+
+int32_t emu8k_vol_slide(emu8k_slide_t* slide, int32_t target)
+{
+ if (slide->last < target)
+ {
+ slide->last += 0x400;
+ if (slide->last > target) slide->last = target;
+ }
+ else if (slide->last > target)
+ {
+ slide->last -= 0x400;
+ if (slide->last < target) slide->last = target;
+ }
+ return slide->last;
+}
+
+//int32_t old_pitch[32]={0};
+//int32_t old_cut[32]={0};
+//int32_t old_vol[32]={0};
+void emu8k_update(emu8k_t* emu8k, int new_pos)
+{
+ if (emu8k->pos >= new_pos)
+ return;
+
+ AssertLogRelReturnVoid(new_pos <= MAXSOUNDBUFLEN);
+
+ int32_t* buf;
+ emu8k_voice_t* emu_voice;
+ int pos;
+ int c;
+
+ /* Clean the buffers since we will accumulate into them. */
+ buf = &emu8k->buffer[emu8k->pos * 2];
+ memset(buf, 0, 2 * (new_pos - emu8k->pos) * sizeof(emu8k->buffer[0]));
+ memset(&emu8k->chorus_in_buffer[emu8k->pos], 0, (new_pos - emu8k->pos) * sizeof(emu8k->chorus_in_buffer[0]));
+ memset(&emu8k->reverb_in_buffer[emu8k->pos], 0, (new_pos - emu8k->pos) * sizeof(emu8k->reverb_in_buffer[0]));
+
+ /* Voices section */
+ for (c = 0; c < 32; c++)
+ {
+ emu_voice = &emu8k->voice[c];
+ buf = &emu8k->buffer[emu8k->pos * 2];
+
+ for (pos = emu8k->pos; pos < new_pos; pos++)
+ {
+ int32_t dat;
+
+ if (emu_voice->cvcf_curr_volume)
+ {
+ /* Waveform oscillator */
+#ifdef RESAMPLER_LINEAR
+ dat = EMU8K_READ_INTERP_LINEAR(emu8k, emu_voice->addr.int_address,
+ emu_voice->addr.fract_address);
+
+#elif defined RESAMPLER_CUBIC
+ dat = EMU8K_READ_INTERP_CUBIC(emu8k, emu_voice->addr.int_address,
+ emu_voice->addr.fract_address);
+#endif
+
+ /* Filter section */
+ if (emu_voice->filterq_idx || emu_voice->cvcf_curr_filt_ctoff != 0xFFFF)
+ {
+ int cutoff = emu_voice->cvcf_curr_filt_ctoff >> 8;
+ const int64_t coef0 = filt_coeffs[emu_voice->filterq_idx][cutoff][0];
+ const int64_t coef1 = filt_coeffs[emu_voice->filterq_idx][cutoff][1];
+ const int64_t coef2 = filt_coeffs[emu_voice->filterq_idx][cutoff][2];
+ /* clip at twice the range */
+#define ClipBuffer(buf) (buf < -16777216) ? -16777216 : (buf > 16777216) ? 16777216 : buf
+
+#ifdef FILTER_INITIAL
+#define NOOP(x) (void)x;
+ NOOP(coef1)
+ /* Apply expected attenuation. (FILTER_MOOG does it implicitly, but this one doesn't).
+ * Work in 24bits. */
+ dat = (dat * emu_voice->filt_att) >> 8;
+
+ int64_t vhp = ((-emu_voice->filt_buffer[0] * coef2) >> 24) - emu_voice->filt_buffer[1] - dat;
+ emu_voice->filt_buffer[1] += (emu_voice->filt_buffer[0] * coef0) >> 24;
+ emu_voice->filt_buffer[0] += (vhp * coef0) >> 24;
+ dat = (int32_t)(emu_voice->filt_buffer[1] >> 8);
+ if (dat > 32767) { dat = 32767; }
+ else if (dat < -32768) { dat = -32768; }
+
+#elif defined FILTER_MOOG
+
+ /*move to 24bits*/
+ dat <<= 8;
+
+ dat -= (coef2 * emu_voice->filt_buffer[4]) >> 24; /*feedback*/
+ int64_t t1 = emu_voice->filt_buffer[1];
+ emu_voice->filt_buffer[1] = ((dat + emu_voice->filt_buffer[0]) * coef0 - emu_voice->filt_buffer[1] * coef1) >> 24;
+ emu_voice->filt_buffer[1] = ClipBuffer(emu_voice->filt_buffer[1]);
+
+ int64_t t2 = emu_voice->filt_buffer[2];
+ emu_voice->filt_buffer[2] = ((emu_voice->filt_buffer[1] + t1) * coef0 - emu_voice->filt_buffer[2] * coef1) >> 24;
+ emu_voice->filt_buffer[2] = ClipBuffer(emu_voice->filt_buffer[2]);
+
+ int64_t t3 = emu_voice->filt_buffer[3];
+ emu_voice->filt_buffer[3] = ((emu_voice->filt_buffer[2] + t2) * coef0 - emu_voice->filt_buffer[3] * coef1) >> 24;
+ emu_voice->filt_buffer[3] = ClipBuffer(emu_voice->filt_buffer[3]);
+
+ emu_voice->filt_buffer[4] = ((emu_voice->filt_buffer[3] + t3) * coef0 - emu_voice->filt_buffer[4] * coef1) >> 24;
+ emu_voice->filt_buffer[4] = ClipBuffer(emu_voice->filt_buffer[4]);
+
+ emu_voice->filt_buffer[0] = ClipBuffer(dat);
+
+ dat = (int32_t)(emu_voice->filt_buffer[4] >> 8);
+ if (dat > 32767)
+ { dat = 32767; }
+ else if (dat < -32768)
+ { dat = -32768; }
+
+#elif defined FILTER_CONSTANT
+
+ /* Apply expected attenuation. (FILTER_MOOG does it implicitly, but this one is constant gain).
+ * Also stay at 24bits.*/
+ dat = (dat * emu_voice->filt_att) >> 8;
+
+ emu_voice->filt_buffer[0] = (coef1 * emu_voice->filt_buffer[0]
+ + coef0 * (dat +
+ ((coef2 * (emu_voice->filt_buffer[0] - emu_voice->filt_buffer[1]))>>24))
+ ) >> 24;
+ emu_voice->filt_buffer[1] = (coef1 * emu_voice->filt_buffer[1]
+ + coef0 * emu_voice->filt_buffer[0]) >> 24;
+
+ emu_voice->filt_buffer[0] = ClipBuffer(emu_voice->filt_buffer[0]);
+ emu_voice->filt_buffer[1] = ClipBuffer(emu_voice->filt_buffer[1]);
+
+ dat = (int32_t)(emu_voice->filt_buffer[1] >> 8);
+ if (dat > 32767) { dat = 32767; }
+ else if (dat < -32768) { dat = -32768; }
+
+#endif
+
+ }
+ if ((emu8k->hwcf3 & 0x04) && !CCCA_DMA_ACTIVE(emu_voice->ccca))
+ {
+ /*volume and pan*/
+ dat = (dat * emu_voice->cvcf_curr_volume) >> 16;
+
+ (*buf++) += (dat * emu_voice->vol_l) >> 8;
+ (*buf++) += (dat * emu_voice->vol_r) >> 8;
+
+ /* Effects section */
+ if (emu_voice->ptrx_revb_send > 0)
+ {
+ emu8k->reverb_in_buffer[pos] += (dat * emu_voice->ptrx_revb_send) >> 8;
+ }
+ if (emu_voice->csl_chor_send > 0)
+ {
+ emu8k->chorus_in_buffer[pos] += (dat * emu_voice->csl_chor_send) >> 8;
+ }
+ }
+ }
+
+ if (emu_voice->env_engine_on)
+ {
+ int32_t attenuation = emu_voice->initial_att;
+ int32_t filtercut = emu_voice->initial_filter;
+ int32_t currentpitch = emu_voice->ip;
+ /* run envelopes */
+ emu8k_envelope_t* volenv = &emu_voice->vol_envelope;
+ switch (volenv->state)
+ {
+ case ENV_DELAY:
+ volenv->delay_samples--;
+ if (volenv->delay_samples <= 0)
+ {
+ volenv->state = ENV_ATTACK;
+ volenv->delay_samples = 0;
+ }
+ attenuation = 0x1FFFFF;
+ break;
+
+ case ENV_ATTACK:
+ /* Attack amount is in linear amplitude */
+ volenv->value_amp_hz += volenv->attack_amount_amp_hz;
+ if (volenv->value_amp_hz >= (1 << 21))
+ {
+ volenv->value_amp_hz = 1 << 21;
+ volenv->value_db_oct = 0;
+ if (volenv->hold_samples)
+ {
+ volenv->state = ENV_HOLD;
+ }
+ else
+ {
+ /* RAMP_UP since db value is inverted and it is 0 at this point. */
+ volenv->state = ENV_RAMP_UP;
+ }
+ }
+ attenuation += env_vol_amplitude_to_db[volenv->value_amp_hz >> 5] << 5;
+ break;
+
+ case ENV_HOLD:
+ volenv->hold_samples--;
+ if (volenv->hold_samples <= 0)
+ {
+ volenv->state = ENV_RAMP_UP;
+ }
+ attenuation += volenv->value_db_oct;
+ break;
+
+ case ENV_RAMP_DOWN:
+ /* Decay/release amount is in fraction of dBs and is always positive */
+ volenv->value_db_oct -= volenv->ramp_amount_db_oct;
+ if (volenv->value_db_oct <= volenv->sustain_value_db_oct)
+ {
+ volenv->value_db_oct = volenv->sustain_value_db_oct;
+ volenv->state = ENV_SUSTAIN;
+ }
+ attenuation += volenv->value_db_oct;
+ break;
+
+ case ENV_RAMP_UP:
+ /* Decay/release amount is in fraction of dBs and is always positive */
+ volenv->value_db_oct += volenv->ramp_amount_db_oct;
+ if (volenv->value_db_oct >= volenv->sustain_value_db_oct)
+ {
+ volenv->value_db_oct = volenv->sustain_value_db_oct;
+ volenv->state = ENV_SUSTAIN;
+ }
+ attenuation += volenv->value_db_oct;
+ break;
+
+ case ENV_SUSTAIN:
+ attenuation += volenv->value_db_oct;
+ break;
+
+ case ENV_STOPPED:
+ attenuation = 0x1FFFFF;
+ break;
+ }
+
+ emu8k_envelope_t* modenv = &emu_voice->mod_envelope;
+ switch (modenv->state)
+ {
+ case ENV_DELAY:
+ modenv->delay_samples--;
+ if (modenv->delay_samples <= 0)
+ {
+ modenv->state = ENV_ATTACK;
+ modenv->delay_samples = 0;
+ }
+ break;
+
+ case ENV_ATTACK:
+ /* Attack amount is in linear amplitude */
+ modenv->value_amp_hz += modenv->attack_amount_amp_hz;
+ modenv->value_db_oct = env_mod_hertz_to_octave[modenv->value_amp_hz >> 5] << 5;
+ if (modenv->value_amp_hz >= (1 << 21))
+ {
+ modenv->value_amp_hz = 1 << 21;
+ modenv->value_db_oct = 1 << 21;
+ if (modenv->hold_samples)
+ {
+ modenv->state = ENV_HOLD;
+ }
+ else
+ {
+ modenv->state = ENV_RAMP_DOWN;
+ }
+ }
+ break;
+
+ case ENV_HOLD:
+ modenv->hold_samples--;
+ if (modenv->hold_samples <= 0)
+ {
+ modenv->state = ENV_RAMP_UP;
+ }
+ break;
+
+ case ENV_RAMP_DOWN:
+ /* Decay/release amount is in fraction of octave and is always positive */
+ modenv->value_db_oct -= modenv->ramp_amount_db_oct;
+ if (modenv->value_db_oct <= modenv->sustain_value_db_oct)
+ {
+ modenv->value_db_oct = modenv->sustain_value_db_oct;
+ modenv->state = ENV_SUSTAIN;
+ }
+ break;
+
+ case ENV_RAMP_UP:
+ /* Decay/release amount is in fraction of octave and is always positive */
+ modenv->value_db_oct += modenv->ramp_amount_db_oct;
+ if (modenv->value_db_oct >= modenv->sustain_value_db_oct)
+ {
+ modenv->value_db_oct = modenv->sustain_value_db_oct;
+ modenv->state = ENV_SUSTAIN;
+ }
+ break;
+ }
+
+ /* run lfos */
+ if (emu_voice->lfo1_delay_samples)
+ {
+ emu_voice->lfo1_delay_samples--;
+ }
+ else
+ {
+ emu_voice->lfo1_count.addr += emu_voice->lfo1_speed;
+ emu_voice->lfo1_count.int_address &= 0xFFFF;
+ }
+ if (emu_voice->lfo2_delay_samples)
+ {
+ emu_voice->lfo2_delay_samples--;
+ }
+ else
+ {
+ emu_voice->lfo2_count.addr += emu_voice->lfo2_speed;
+ emu_voice->lfo2_count.int_address &= 0xFFFF;
+ }
+
+ if (emu_voice->fixed_modenv_pitch_height)
+ {
+ /* modenv range 1<<21, pitch height range 1<<14 desired range 0x1000 (+/-one octave) */
+ currentpitch += ((modenv->value_db_oct >> 9) * emu_voice->fixed_modenv_pitch_height) >> 14;
+ }
+
+ if (emu_voice->fixed_lfo1_vibrato)
+ {
+ /* table range 1<<15, pitch mod range 1<<14 desired range 0x1000 (+/-one octave) */
+ int32_t lfo1_vibrato = (lfotable[emu_voice->lfo1_count.int_address] * emu_voice->fixed_lfo1_vibrato) >> 17;
+ currentpitch += lfo1_vibrato;
+ }
+ if (emu_voice->fixed_lfo2_vibrato)
+ {
+ /* table range 1<<15, pitch mod range 1<<14 desired range 0x1000 (+/-one octave) */
+ int32_t lfo2_vibrato = (lfotable[emu_voice->lfo2_count.int_address] * emu_voice->fixed_lfo2_vibrato) >> 17;
+ currentpitch += lfo2_vibrato;
+ }
+
+ if (emu_voice->fixed_modenv_filter_height)
+ {
+ /* modenv range 1<<21, pitch height range 1<<14 desired range 0x200000 (+/-full filter range) */
+ filtercut += ((modenv->value_db_oct >> 9) * emu_voice->fixed_modenv_filter_height) >> 5;
+ }
+
+ if (emu_voice->fixed_lfo1_filt_mod)
+ {
+ /* table range 1<<15, pitch mod range 1<<14 desired range 0x100000 (+/-three octaves) */
+ int32_t lfo1_filtmod = (lfotable[emu_voice->lfo1_count.int_address] * emu_voice->fixed_lfo1_filt_mod) >> 9;
+ filtercut += lfo1_filtmod;
+ }
+
+ if (emu_voice->fixed_lfo1_tremolo)
+ {
+ /* table range 1<<15, pitch mod range 1<<14 desired range 0x40000 (+/-12dBs). */
+ int32_t lfo1_tremolo = (lfotable[emu_voice->lfo1_count.int_address] * emu_voice->fixed_lfo1_tremolo) >> 11;
+ attenuation += lfo1_tremolo;
+ }
+
+ if (currentpitch > 0xFFFF) currentpitch = 0xFFFF;
+ if (currentpitch < 0) currentpitch = 0;
+ if (attenuation > 0x1FFFFF) attenuation = 0x1FFFFF;
+ if (attenuation < 0) attenuation = 0;
+ if (filtercut > 0x1FFFFF) filtercut = 0x1FFFFF;
+ if (filtercut < 0) filtercut = 0;
+
+ emu_voice->vtft_vol_target = env_vol_db_to_vol_target[attenuation >> 5];
+ emu_voice->vtft_filter_target = filtercut >> 5;
+ emu_voice->ptrx_pit_target = freqtable[currentpitch] >> 18;
+
+ }
+/*
+I've recopilated these sentences to get an idea of how to loop
+
+- Set its PSST register and its CLS register to zero to cause no loops to occur.
+-Setting the Loop Start Offset and the Loop End Offset to the same value, will cause the oscillator to loop the entire memory.
+
+-Setting the PlayPosition greater than the Loop End Offset, will cause the oscillator to play in reverse, back to the Loop End Offset.
+ It's pretty neat, but appears to be uncontrollable (the rate at which the samples are played in reverse).
+
+-Note that due to interpolator offset, the actual loop point is one greater than the start address
+-Note that due to interpolator offset, the actual loop point will end at an address one greater than the loop address
+-Note that the actual audio location is the point 1 word higher than this value due to interpolation offset
+-In programs that use the awe, they generally set the loop address as "loopaddress -1" to compensate for the above.
+(Note: I am already using address+1 in the interpolators so these things are already as they should.)
+*/
+ emu_voice->addr.addr += ((uint64_t)emu_voice->cpf_curr_pitch) << 18;
+ if (emu_voice->addr.addr >= emu_voice->loop_end.addr)
+ {
+ emu_voice->addr.int_address -= (emu_voice->loop_end.int_address - emu_voice->loop_start.int_address);
+ emu_voice->addr.int_address &= EMU8K_MEM_ADDRESS_MASK;
+ }
+
+ /* TODO: How and when are the target and current values updated */
+ emu_voice->cpf_curr_pitch = emu_voice->ptrx_pit_target;
+ emu_voice->cvcf_curr_volume = emu8k_vol_slide(&emu_voice->volumeslide, emu_voice->vtft_vol_target);
+ emu_voice->cvcf_curr_filt_ctoff = emu_voice->vtft_filter_target;
+ }
+
+ /* Update EMU voice registers. */
+ emu_voice->ccca = (((uint32_t)emu_voice->ccca_qcontrol) << 24) | emu_voice->addr.int_address;
+ emu_voice->cpf_curr_frac_addr = emu_voice->addr.fract_address;
+
+ //if ( emu_voice->cvcf_curr_volume != old_vol[c]) {
+ // pclog("EMUVOL (%d):%d\n", c, emu_voice->cvcf_curr_volume);
+ // old_vol[c]=emu_voice->cvcf_curr_volume;
+ //}
+ //pclog("EMUFILT :%d\n", emu_voice->cvcf_curr_filt_ctoff);
+ }
+
+ buf = &emu8k->buffer[emu8k->pos * 2];
+ emu8k_work_reverb(&emu8k->reverb_in_buffer[emu8k->pos], buf, &emu8k->reverb_engine, new_pos - emu8k->pos);
+ emu8k_work_chorus(&emu8k->chorus_in_buffer[emu8k->pos], buf, &emu8k->chorus_engine, new_pos - emu8k->pos);
+ emu8k_work_eq(buf, new_pos - emu8k->pos);
+
+ // Clip signal
+ for (pos = emu8k->pos; pos < new_pos; pos++)
+ {
+ if (buf[0] < -32768)
+ buf[0] = -32768;
+ else if (buf[0] > 32767)
+ buf[0] = 32767;
+
+ if (buf[1] < -32768)
+ buf[1] = -32768;
+ else if (buf[1] > 32767)
+ buf[1] = 32767;
+
+ buf += 2;
+ }
+
+ /* Update EMU clock. */
+ emu8k->wc += (new_pos - emu8k->pos);
+
+ emu8k->pos = new_pos;
+}
+
+static void emu8k_init_globals()
+{
+ int c;
+ double out;
+
+ /*Create frequency table. (Convert initial pitch register value to a linear speed change)
+ * The input is encoded such as 0xe000 is center note (no pitch shift)
+ * and from then on , changing up or down 0x1000 (4096) increments/decrements an octave.
+ * Note that this is in reference to the 44.1Khz clock that the channels play at.
+ * The 65536 * 65536 is in order to left-shift the 32bit value to a 64bit value as a 32.32 fixed point.
+ */
+ for (c = 0; c < 0x10000; c++)
+ {
+ freqtable[c] = (uint64_t)(exp2((double)(c - 0xe000) / 4096.0) * 65536.0 * 65536.0);
+ }
+ /* Shortcut: minimum pitch equals stopped. I don't really know if this is true, but it's better
+ * since some programs set the pitch to 0 for unused channels. */
+ freqtable[0] = 0;
+
+ /* starting at 65535 because it is used for "volume target" register conversion. */
+ out = 65535.0;
+ for (c = 0; c < 256; c++)
+ {
+ attentable[c] = (int32_t)out;
+ out /= sqrt(1.09018); /*0.375 dB steps*/
+ }
+ /* Shortcut: max attenuation is silent, not -96dB. */
+ attentable[255] = 0;
+
+ /* Note: these two tables have "db" inverted: 0 dB is max volume, 65535 "db" (-96.32dBFS) is silence.
+ * Important: Using 65535 as max output value because this is intended to be used with the volume target register! */
+ out = 65535.0;
+ for (c = 0; c < 0x10000; c++)
+ {
+ //double db = -(c*6.0205999/65535.0)*16.0;
+ //out = powf(10.f,db/20.f) * 65536.0;
+ env_vol_db_to_vol_target[c] = (int32_t)out;
+ /* calculated from the 65536th root of 65536 */
+ out /= 1.00016923970;
+ }
+ /* Shortcut: max attenuation is silent, not -96dB. */
+ env_vol_db_to_vol_target[0x10000 - 1] = 0;
+ /* One more position to accept max value being 65536. */
+ env_vol_db_to_vol_target[0x10000] = 0;
+
+ for (c = 1; c < 0x10000; c++)
+ {
+ out = -680.32142884264 * 20.0 * log10(((double)c) / 65535.0);
+ env_vol_amplitude_to_db[c] = (int32_t)out;
+ }
+ /*Shortcut: max attenuation is silent, not -96dB.*/
+ env_vol_amplitude_to_db[0] = 65535;
+ /* One more position to acceMpt max value being 65536. */
+ env_vol_amplitude_to_db[0x10000] = 0;
+
+ for (c = 1; c < 0x10000; c++)
+ {
+ out = log2((((double)c) / 0x10000) + 1.0) * 65536.0;
+ env_mod_hertz_to_octave[c] = (int32_t)out;
+ }
+ /*No hertz change, no octave change. */
+ env_mod_hertz_to_octave[0] = 0;
+ /* One more position to accept max value being 65536. */
+ env_mod_hertz_to_octave[0x10000] = 65536;
+
+
+ /* This formula comes from vince vu/judge dredd's awe32p10 and corresponds to what the freebsd/linux AWE32 driver has. */
+ float millis;
+ for (c = 0; c < 128; c++)
+ {
+ if (c == 0)
+ millis = 0; /* This means never attack. */
+ else if (c < 32)
+ millis = 11878.0 / c;
+ else
+ millis = 360 * exp((c - 32) / (16.0 / log(1.0 / 2.0)));
+
+ env_attack_to_samples[c] = 44.1 * millis;
+ /* This is an alternate formula with linear increments, but probably incorrect:
+ * millis = (256+4096*(0x7F-c)) */
+ }
+
+ /* The LFOs use a triangular waveform starting at zero and going 1/-1/1/-1.
+ * This table is stored in signed 16bits precision, with a period of 65536 samples */
+ for (c = 0; c < 65536; c++)
+ {
+ int d = (c + 16384) & 65535;
+ if (d >= 32768)
+ lfotable[c] = 32768 + ((32768 - d) * 2);
+ else
+ lfotable[c] = (d * 2) - 32768;
+ }
+ /* The 65536 * 65536 is in order to left-shift the 32bit value to a 64bit value as a 32.32 fixed point. */
+ out = 0.01;
+ for (c = 0; c < 256; c++)
+ {
+ lfofreqtospeed[c] = (uint64_t)(out * 65536.0 / 44100.0 * 65536.0 * 65536.0);
+ out += 0.042;
+ }
+
+ for (c = 0; c < 65536; c++)
+ {
+ chortable[c] = sin(c * M_PI / 32768.0);
+ }
+
+
+ /* Filter coefficients tables. Note: Values are multiplied by *16777216 to left shift 24 bits. (i.e. 8.24 fixed point) */
+ int qidx;
+ for (qidx = 0; qidx < 16; qidx++)
+ {
+ out = 125.0; /* Start at 125Hz */
+ for (c = 0; c < 256; c++)
+ {
+#ifdef FILTER_INITIAL
+ float w0 = sin(2.0*M_PI*out / 44100.0);
+ /* The value 102.5f has been selected a bit randomly. Pretends to reach 0.2929 at w0 = 1.0 */
+ float q = (qidx / 102.5f) * (1.0 + 1.0 / w0);
+ /* Limit max value. Else it would be 470. */
+ if (q > 200) q=200;
+ filt_coeffs[qidx][c][0] = (int32_t)(w0 * 16777216.0);
+ filt_coeffs[qidx][c][1] = 16777216.0;
+ filt_coeffs[qidx][c][2] = (int32_t)((1.0f / (0.7071f + q)) * 16777216.0);
+#elif defined FILTER_MOOG
+ float w0 = sin(2.0 * M_PI * out / 44100.0);
+ float q_factor = 1.0f - w0;
+ float p = w0 + 0.8f * w0 * q_factor;
+ float f = p + p - 1.0f;
+ float resonance = (1.0 - pow(2.0, -qidx * 24.0 / 90.0)) * 0.8;
+ float q = resonance * (1.0f + 0.5f * q_factor * (w0 + 5.6f * q_factor * q_factor));
+ filt_coeffs[qidx][c][0] = (int32_t)(p * 16777216.0);
+ filt_coeffs[qidx][c][1] = (int32_t)(f * 16777216.0);
+ filt_coeffs[qidx][c][2] = (int32_t)(q * 16777216.0);
+#elif defined FILTER_CONSTANT
+ float q = (1.0-pow(2.0,-qidx*24.0/90.0))*0.8;
+ float coef0 = sin(2.0*M_PI*out / 44100.0);
+ float coef1 = 1.0 - coef0;
+ float coef2 = q * (1.0 + 1.0 / coef1);
+ filt_coeffs[qidx][c][0] = (int32_t)(coef0 * 16777216.0);
+ filt_coeffs[qidx][c][1] = (int32_t)(coef1 * 16777216.0);
+ filt_coeffs[qidx][c][2] = (int32_t)(coef2 * 16777216.0);
+#endif //FILTER_TYPE
+ /* 42.66 divisions per octave (the doc says quarter seminotes which is 48, but then it would be almost an octave less) */
+ out *= 1.016378315;
+ /* 42 divisions. This moves the max frequency to 8.5Khz.*/
+ //out *= 1.0166404394;
+ /* This is a linear increment method, that corresponds to the NRPN table, but contradicts the EMU8KPRM doc: */
+ //out = 100.0 + (c+1.0)*31.25; //31.25Hz steps */
+ }
+ }
+
+ /* Cubic Resampling ( 4point cubic spline) */
+ double const resdouble = 1.0 / (double)CUBIC_RESOLUTION;
+ for (c = 0; c < CUBIC_RESOLUTION; c++)
+ {
+ double x = (double)c * resdouble;
+ /* Cubic resolution is made of four table, but I've put them all in one table to optimize memory access. */
+ cubic_table[c * 4] = (-0.5 * x * x * x + x * x - 0.5 * x);
+ cubic_table[c * 4 + 1] = (1.5 * x * x * x - 2.5 * x * x + 1.0);
+ cubic_table[c * 4 + 2] = (-1.5 * x * x * x + 2.0 * x * x + 0.5 * x);
+ cubic_table[c * 4 + 3] = (0.5 * x * x * x - 0.5 * x * x);
+ }
+}
+
+emu8k_t* emu8k_alloc(void *rom, size_t onboard_ram)
+{
+ emu8k_t *emu8k = RTMemAlloc(sizeof(emu8k_t));
+ AssertPtrReturn(emu8k, NULL);
+
+ emu8k_init_globals();
+
+ emu8k->rom = rom;
+
+ /*AWE-DUMP creates ROM images offset by 2 bytes, so if we detect this
+ then correct it*/
+ if (emu8k->rom[3] == 0x314d && emu8k->rom[4] == 0x474d)
+ {
+ memmove(&emu8k->rom[0], &emu8k->rom[1], (1024 * 1024) - 2);
+ emu8k->rom[0x7ffff] = 0;
+ }
+
+ emu8k->empty = RTMemAllocZ(2*BLOCK_SIZE_WORDS);
+ AssertPtr(emu8k->empty);
+
+ // Initialize ram_pointers
+ int j = 0;
+ for (; j < 0x8; j++)
+ {
+ emu8k->ram_pointers[j] = emu8k->rom + (j * BLOCK_SIZE_WORDS);
+ }
+ for (; j < 0x20; j++)
+ {
+ emu8k->ram_pointers[j] = emu8k->empty;
+ }
+
+ if (onboard_ram > 0)
+ {
+ /*Clip to 28MB, since that's the max that we can address. */
+ Assert(onboard_ram <= 0x7000);
+ emu8k->ram = RTMemAllocZ(onboard_ram * _1K);
+ AssertPtr(emu8k->ram);
+
+ const int i_end = onboard_ram >> 7;
+ int i = 0;
+ for (; i < i_end; i++, j++)
+ {
+ emu8k->ram_pointers[j] = emu8k->ram + (i * BLOCK_SIZE_WORDS);
+ }
+ emu8k->ram_end_addr = EMU8K_RAM_MEM_START + (onboard_ram << 9);
+ }
+ else
+ {
+ emu8k->ram = NULL;
+ emu8k->ram_end_addr = EMU8K_RAM_MEM_START;
+ }
+ for (; j < 0x100; j++)
+ {
+ emu8k->ram_pointers[j] = emu8k->empty;
+
+ }
+
+ return emu8k;
+}
+
+void emu8k_free(emu8k_t* emu8k)
+{
+ RTMemFree(emu8k->empty);
+ RTMemFree(emu8k->ram);
+
+ RTMemFree(emu8k);
+}
+
+void emu8k_reset(emu8k_t* emu8k)
+{
+ /* NOTE! read_pos and buffer content is implicitly initialized to zero by the sb_t structure memset on sb_awe32_init() */
+ emu8k->reverb_engine.reflections[0].bufsize = 2 * REV_BUFSIZE_STEP;
+ emu8k->reverb_engine.reflections[1].bufsize = 4 * REV_BUFSIZE_STEP;
+ emu8k->reverb_engine.reflections[2].bufsize = 8 * REV_BUFSIZE_STEP;
+ emu8k->reverb_engine.reflections[3].bufsize = 13 * REV_BUFSIZE_STEP;
+ emu8k->reverb_engine.reflections[4].bufsize = 19 * REV_BUFSIZE_STEP;
+ emu8k->reverb_engine.reflections[5].bufsize = 26 * REV_BUFSIZE_STEP;
+
+ /*This is a bit random.*/
+ for (int c = 0; c < 4; c++)
+ {
+ emu8k->reverb_engine.allpass[3 - c].feedback = 0.5;
+ emu8k->reverb_engine.allpass[3 - c].bufsize = (4 * c) * REV_BUFSIZE_STEP + 55;
+ emu8k->reverb_engine.allpass[7 - c].feedback = 0.5;
+ emu8k->reverb_engine.allpass[7 - c].bufsize = (4 * c) * REV_BUFSIZE_STEP + 55;
+ }
+
+ /* Even when the documentation says that this has to be written by applications to initialize the card,
+ * several applications and drivers ( aweman on windows, linux oss driver..) read it to detect an AWE card. */
+ emu8k->hwcf1 = 0x59;
+ emu8k->hwcf2 = 0x20;
+ /* Initial state is muted. 0x04 is unmuted. */
+ emu8k->hwcf3 = 0x00;
+}
+
+void emu8k_render(emu8k_t *emu8k, int16_t *buf, size_t frames)
+{
+ emu8k_update(emu8k, frames);
+
+ // Convert from int32_t samples to int16_t
+ for (unsigned int i = 0; i < frames * 2; i++)
+ {
+ buf[i] = RT_CLAMP(emu8k->buffer[i], INT16_MIN, INT16_MAX);
+ }
+
+ emu8k->pos = 0;
+}