ShuriZma
/
xenia
mirror of https://github.com/xenia-project/xenia.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

[GPU] ucode doc: A2xx ALU names, DX9 rules, cube

This commit is contained in:
Triang3l 2020-04-26 16:53:13 +03:00
parent 96a61bc623
commit 07e4f84c2b
1 changed files with 104 additions and 85 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

189
src/xenia/gpu/ucode.h
View File

@ -653,8 +653,8 @@ static_assert_size(TextureFetchInstruction, 12);
// What follows is largely a mash up of the microcode assembly naming and the // What follows is largely a mash up of the microcode assembly naming and the
// R600 docs that have a near 1:1 with the instructions available in the xenos // R600 docs that have a near 1:1 with the instructions available in the xenos
// GPU. Some of the behavior has been experimentally verified. Some has been // GPU, and Adreno 2xx instruction names found in Freedreno. Some of the
// guessed. // behavior has been experimentally verified. Some has been guessed.
// Docs: https://www.x.org/docs/AMD/old/r600isa.pdf // Docs: https://www.x.org/docs/AMD/old/r600isa.pdf
// //
// Conventions: // Conventions:
@ -665,100 +665,101 @@ static_assert_size(TextureFetchInstruction, 12);
// - Scalar ops write the result to the entire destination register. // - Scalar ops write the result to the entire destination register.
// - pv and ps are the previous results of a vector or scalar ALU operation. // - pv and ps are the previous results of a vector or scalar ALU operation.
// Both are valid only within the current ALU clause. They are not modified // Both are valid only within the current ALU clause. They are not modified
// when write masks are disabled or the instruction that would write them // when the instruction that would write them fails its predication check.
// fails its predication check. // - Direct3D 9 rules (like in GCN v_*_legacy_f32 instructions) for
// multiplication (0 * anything = 0) and for NaN in min/max.
enum class AluScalarOpcode : uint32_t { enum class AluScalarOpcode : uint32_t {
// Floating-Point Add // Floating-Point Add
// adds dest, src0.ab // adds/ADDs dest, src0.ab
// dest.xyzw = src0.a + src0.b; // dest.xyzw = src0.a + src0.b;
kAdds = 0, kAdds = 0,
// Floating-Point Add (with Previous) // Floating-Point Add (with Previous)
// adds_prev dest, src0.a // adds_prev/ADD_PREVs dest, src0.a
// dest.xyzw = src0.a + ps; // dest.xyzw = src0.a + ps;
kAddsPrev = 1, kAddsPrev = 1,
// Floating-Point Multiply // Floating-Point Multiply
// muls dest, src0.ab // muls/MULs dest, src0.ab
// dest.xyzw = src0.a * src0.b; // dest.xyzw = src0.a * src0.b;
kMuls = 2, kMuls = 2,
// Floating-Point Multiply (with Previous) // Floating-Point Multiply (with Previous)
// muls_prev dest, src0.a // muls_prev/MUL_PREVs dest, src0.a
// dest.xyzw = src0.a * ps; // dest.xyzw = src0.a * ps;
kMulsPrev = 3, kMulsPrev = 3,
// Scalar Multiply Emulating LIT Operation // Scalar Multiply Emulating LIT Operation
// muls_prev2 dest, src0.ab // muls_prev2/MUL_PREV2s dest, src0.ab
// dest.xyzw = // dest.xyzw =
// ps == -FLT_MAX || !isfinite(ps) || !isfinite(src0.b) || src0.b <= 0 // ps == -FLT_MAX || !isfinite(ps) || !isfinite(src0.b) || src0.b <= 0
// ? -FLT_MAX : src0.a * ps; // ? -FLT_MAX : src0.a * ps;
kMulsPrev2 = 4, kMulsPrev2 = 4,
// Floating-Point Maximum // Floating-Point Maximum
// maxs dest, src0.ab // maxs/MAXs dest, src0.ab
// dest.xyzw = src0.a >= src0.b ? src0.a : src0.b; // dest.xyzw = src0.a >= src0.b ? src0.a : src0.b;
kMaxs = 5, kMaxs = 5,
// Floating-Point Minimum // Floating-Point Minimum
// mins dest, src0.ab // mins/MINs dest, src0.ab
// dest.xyzw = src0.a < src0.b ? src0.a : src0.b; // dest.xyzw = src0.a < src0.b ? src0.a : src0.b;
kMins = 6, kMins = 6,
// Floating-Point Set If Equal // Floating-Point Set If Equal
// seqs dest, src0.a // seqs/SETEs dest, src0.a
// dest.xyzw = src0.a == 0.0 ? 1.0 : 0.0; // dest.xyzw = src0.a == 0.0 ? 1.0 : 0.0;
kSeqs = 7, kSeqs = 7,
// Floating-Point Set If Greater Than // Floating-Point Set If Greater Than
// sgts dest, src0.a // sgts/SETGTs dest, src0.a
// dest.xyzw = src0.a > 0.0 ? 1.0 : 0.0; // dest.xyzw = src0.a > 0.0 ? 1.0 : 0.0;
kSgts = 8, kSgts = 8,
// Floating-Point Set If Greater Than Or Equal // Floating-Point Set If Greater Than Or Equal
// sges dest, src0.a // sges/SETGTEs dest, src0.a
// dest.xyzw = src0.a >= 0.0 ? 1.0 : 0.0; // dest.xyzw = src0.a >= 0.0 ? 1.0 : 0.0;
kSges = 9, kSges = 9,
// Floating-Point Set If Not Equal // Floating-Point Set If Not Equal
// snes dest, src0.a // snes/SETNEs dest, src0.a
// dest.xyzw = src0.a != 0.0 ? 1.0 : 0.0; // dest.xyzw = src0.a != 0.0 ? 1.0 : 0.0;
kSnes = 10, kSnes = 10,
// Floating-Point Fractional // Floating-Point Fractional
// frcs dest, src0.a // frcs/FRACs dest, src0.a
// dest.xyzw = src0.a - floor(src0.a); // dest.xyzw = src0.a - floor(src0.a);
kFrcs = 11, kFrcs = 11,
// Floating-Point Truncate // Floating-Point Truncate
// truncs dest, src0.a // truncs/TRUNCs dest, src0.a
// dest.xyzw = src0.a >= 0 ? floor(src0.a) : -floor(-src0.a); // dest.xyzw = src0.a >= 0 ? floor(src0.a) : -floor(-src0.a);
kTruncs = 12, kTruncs = 12,
// Floating-Point Floor // Floating-Point Floor
// floors dest, src0.a // floors/FLOORs dest, src0.a
// dest.xyzw = floor(src0.a); // dest.xyzw = floor(src0.a);
kFloors = 13, kFloors = 13,
// Scalar Base-2 Exponent, IEEE // Scalar Base-2 Exponent, IEEE
// exp dest, src0.a // exp/EXP_IEEE dest, src0.a
// dest.xyzw = src0.a == 0.0 ? 1.0 : pow(2, src0.a); // dest.xyzw = src0.a == 0.0 ? 1.0 : pow(2, src0.a);
kExp = 14, kExp = 14,
// Scalar Base-2 Log // Scalar Base-2 Log
// logc dest, src0.a // logc/LOG_CLAMP dest, src0.a
// float t = src0.a == 1.0 ? 0.0 : log(src0.a) / log(2.0); // float t = src0.a == 1.0 ? 0.0 : log(src0.a) / log(2.0);
// dest.xyzw = t == -INF ? -FLT_MAX : t; // dest.xyzw = t == -INF ? -FLT_MAX : t;
kLogc = 15, kLogc = 15,
// Scalar Base-2 IEEE Log // Scalar Base-2 IEEE Log
// log dest, src0.a // log/LOG_IEEE dest, src0.a
// dest.xyzw = src0.a == 1.0 ? 0.0 : log(src0.a) / log(2.0); // dest.xyzw = src0.a == 1.0 ? 0.0 : log(src0.a) / log(2.0);
kLog = 16, kLog = 16,
// Scalar Reciprocal, Clamp to Maximum // Scalar Reciprocal, Clamp to Maximum
// rcpc dest, src0.a // rcpc/RECIP_CLAMP dest, src0.a
// float t = src0.a == 1.0 ? 1.0 : 1.0 / src0.a; // float t = src0.a == 1.0 ? 1.0 : 1.0 / src0.a;
// if (t == -INF) t = -FLT_MAX; // if (t == -INF) t = -FLT_MAX;
// else if (t == INF) t = FLT_MAX; // else if (t == INF) t = FLT_MAX;
@ -766,7 +767,8 @@ enum class AluScalarOpcode : uint32_t {
kRcpc = 17, kRcpc = 17,
// Scalar Reciprocal, Clamp to Zero // Scalar Reciprocal, Clamp to Zero
// rcpf dest, src0.a // Mimicking the behavior of the fixed-function pipeline.
// rcpf/RECIP_FF dest, src0.a
// float t = src0.a == 1.0 ? 1.0 : 1.0 / src0.a; // float t = src0.a == 1.0 ? 1.0 : 1.0 / src0.a;
// if (t == -INF) t = -0.0; // if (t == -INF) t = -0.0;
// else if (t == INF) t = 0.0; // else if (t == INF) t = 0.0;
@ -774,12 +776,12 @@ enum class AluScalarOpcode : uint32_t {
kRcpf = 18, kRcpf = 18,
// Scalar Reciprocal, IEEE Approximation // Scalar Reciprocal, IEEE Approximation
// rcp dest, src0.a // rcp/RECIP_IEEE dest, src0.a
// dest.xyzw = src0.a == 1.0 ? 1.0 : 1.0 / src0.a; // dest.xyzw = src0.a == 1.0 ? 1.0 : 1.0 / src0.a;
kRcp = 19, kRcp = 19,
// Scalar Reciprocal Square Root, Clamp to Maximum // Scalar Reciprocal Square Root, Clamp to Maximum
// rsqc dest, src0.a // rsqc/RECIPSQ_CLAMP dest, src0.a
// float t = src0.a == 1.0 ? 1.0 : 1.0 / sqrt(src0.a); // float t = src0.a == 1.0 ? 1.0 : 1.0 / sqrt(src0.a);
// if (t == -INF) t = -FLT_MAX; // if (t == -INF) t = -FLT_MAX;
// else if (t == INF) t = FLT_MAX; // else if (t == INF) t = FLT_MAX;
@ -787,7 +789,7 @@ enum class AluScalarOpcode : uint32_t {
kRsqc = 20, kRsqc = 20,
// Scalar Reciprocal Square Root, Clamp to Zero // Scalar Reciprocal Square Root, Clamp to Zero
// rsqc dest, src0.a // rsqf/RECIPSQ_FF dest, src0.a
// float t = src0.a == 1.0 ? 1.0 : 1.0 / sqrt(src0.a); // float t = src0.a == 1.0 ? 1.0 : 1.0 / sqrt(src0.a);
// if (t == -INF) t = -0.0; // if (t == -INF) t = -0.0;
// else if (t == INF) t = 0.0; // else if (t == INF) t = 0.0;
@ -795,13 +797,13 @@ enum class AluScalarOpcode : uint32_t {
kRsqf = 21, kRsqf = 21,
// Scalar Reciprocal Square Root, IEEE Approximation // Scalar Reciprocal Square Root, IEEE Approximation
// rsq dest, src0.a // rsq/RECIPSQ_IEEE dest, src0.a
// dest.xyzw = src0.a == 1.0 ? 1.0 : 1.0 / sqrt(src0.a); // dest.xyzw = src0.a == 1.0 ? 1.0 : 1.0 / sqrt(src0.a);
kRsq = 22, kRsq = 22,
// Floating-Point Maximum with Copy To Integer in AR // Floating-Point Maximum with Copy To Integer in AR
// maxas dest, src0.ab // maxas dest, src0.ab
// movas dest, src0.aa // movas/MOVAs dest, src0.aa
// int result = (int)floor(src0.a + 0.5); // int result = (int)floor(src0.a + 0.5);
// a0 = clamp(result, -256, 255); // a0 = clamp(result, -256, 255);
// dest.xyzw = src0.a >= src0.b ? src0.a : src0.b; // dest.xyzw = src0.a >= src0.b ? src0.a : src0.b;
@ -809,24 +811,24 @@ enum class AluScalarOpcode : uint32_t {
// Floating-Point Maximum with Copy Truncated To Integer in AR // Floating-Point Maximum with Copy Truncated To Integer in AR
// maxasf dest, src0.ab // maxasf dest, src0.ab
// movasf dest, src0.aa // movasf/MOVA_FLOORs dest, src0.aa
// int result = (int)floor(src0.a); // int result = (int)floor(src0.a);
// a0 = clamp(result, -256, 255); // a0 = clamp(result, -256, 255);
// dest.xyzw = src0.a >= src0.b ? src0.a : src0.b; // dest.xyzw = src0.a >= src0.b ? src0.a : src0.b;
kMaxAsf = 24, kMaxAsf = 24,
// Floating-Point Subtract // Floating-Point Subtract
// subs dest, src0.ab // subs/SUBs dest, src0.ab
// dest.xyzw = src0.a - src0.b; // dest.xyzw = src0.a - src0.b;
kSubs = 25, kSubs = 25,
// Floating-Point Subtract (with Previous) // Floating-Point Subtract (with Previous)
// subs_prev dest, src0.a // subs_prev/SUB_PREVs dest, src0.a
// dest.xyzw = src0.a - ps; // dest.xyzw = src0.a - ps;
kSubsPrev = 26, kSubsPrev = 26,
// Floating-Point Predicate Set If Equal // Floating-Point Predicate Set If Equal
// setp_eq dest, src0.a // setp_eq/PRED_SETEs dest, src0.a
// if (src0.a == 0.0) { // if (src0.a == 0.0) {
// dest.xyzw = 0.0; // dest.xyzw = 0.0;
// p0 = 1; // p0 = 1;
@ -837,7 +839,7 @@ enum class AluScalarOpcode : uint32_t {
kSetpEq = 27, kSetpEq = 27,
// Floating-Point Predicate Set If Not Equal // Floating-Point Predicate Set If Not Equal
// setp_ne dest, src0.a // setp_ne/PRED_SETNEs dest, src0.a
// if (src0.a != 0.0) { // if (src0.a != 0.0) {
// dest.xyzw = 0.0; // dest.xyzw = 0.0;
// p0 = 1; // p0 = 1;
@ -848,7 +850,7 @@ enum class AluScalarOpcode : uint32_t {
kSetpNe = 28, kSetpNe = 28,
// Floating-Point Predicate Set If Greater Than // Floating-Point Predicate Set If Greater Than
// setp_gt dest, src0.a // setp_gt/PRED_SETGTs dest, src0.a
// if (src0.a > 0.0) { // if (src0.a > 0.0) {
// dest.xyzw = 0.0; // dest.xyzw = 0.0;
// p0 = 1; // p0 = 1;
@ -859,7 +861,7 @@ enum class AluScalarOpcode : uint32_t {
kSetpGt = 29, kSetpGt = 29,
// Floating-Point Predicate Set If Greater Than Or Equal // Floating-Point Predicate Set If Greater Than Or Equal
// setp_ge dest, src0.a // setp_ge/PRED_SETGTEs dest, src0.a
// if (src0.a >= 0.0) { // if (src0.a >= 0.0) {
// dest.xyzw = 0.0; // dest.xyzw = 0.0;
// p0 = 1; // p0 = 1;
@ -870,7 +872,7 @@ enum class AluScalarOpcode : uint32_t {
kSetpGe = 30, kSetpGe = 30,
// Predicate Counter Invert // Predicate Counter Invert
// setp_inv dest, src0.a // setp_inv/PRED_SET_INVs dest, src0.a
// if (src0.a == 1.0) { // if (src0.a == 1.0) {
// dest.xyzw = 0.0; // dest.xyzw = 0.0;
// p0 = 1; // p0 = 1;
@ -885,7 +887,7 @@ enum class AluScalarOpcode : uint32_t {
kSetpInv = 31, kSetpInv = 31,
// Predicate Counter Pop // Predicate Counter Pop
// setp_pop dest, src0.a // setp_pop/PRED_SET_POPs dest, src0.a
// if (src0.a - 1.0 <= 0.0) { // if (src0.a - 1.0 <= 0.0) {
// dest.xyzw = 0.0; // dest.xyzw = 0.0;
// p0 = 1; // p0 = 1;
@ -896,13 +898,13 @@ enum class AluScalarOpcode : uint32_t {
kSetpPop = 32, kSetpPop = 32,
// Predicate Counter Clear // Predicate Counter Clear
// setp_clr dest // setp_clr/PRED_SET_CLRs dest
// dest.xyzw = FLT_MAX; // dest.xyzw = FLT_MAX;
// p0 = 0; // p0 = 0;
kSetpClr = 33, kSetpClr = 33,
// Predicate Counter Restore // Predicate Counter Restore
// setp_rstr dest, src0.a // setp_rstr/PRED_SET_RESTOREs dest, src0.a
// if (src0.a == 0.0) { // if (src0.a == 0.0) {
// dest.xyzw = 0.0; // dest.xyzw = 0.0;
// p0 = 1; // p0 = 1;
@ -913,7 +915,7 @@ enum class AluScalarOpcode : uint32_t {
kSetpRstr = 34, kSetpRstr = 34,
// Floating-Point Pixel Kill If Equal // Floating-Point Pixel Kill If Equal
// kills_eq dest, src0.a // kills_eq/KILLEs dest, src0.a
// if (src0.a == 0.0) { // if (src0.a == 0.0) {
// dest.xyzw = 1.0; // dest.xyzw = 1.0;
// discard; // discard;
@ -923,7 +925,7 @@ enum class AluScalarOpcode : uint32_t {
kKillsEq = 35, kKillsEq = 35,
// Floating-Point Pixel Kill If Greater Than // Floating-Point Pixel Kill If Greater Than
// kills_gt dest, src0.a // kills_gt/KILLGTs dest, src0.a
// if (src0.a > 0.0) { // if (src0.a > 0.0) {
// dest.xyzw = 1.0; // dest.xyzw = 1.0;
// discard; // discard;
@ -933,7 +935,7 @@ enum class AluScalarOpcode : uint32_t {
kKillsGt = 36, kKillsGt = 36,
// Floating-Point Pixel Kill If Greater Than Or Equal // Floating-Point Pixel Kill If Greater Than Or Equal
// kills_ge dest, src0.a // kills_ge/KILLGTEs dest, src0.a
// if (src0.a >= 0.0) { // if (src0.a >= 0.0) {
// dest.xyzw = 1.0; // dest.xyzw = 1.0;
// discard; // discard;
@ -943,7 +945,7 @@ enum class AluScalarOpcode : uint32_t {
kKillsGe = 37, kKillsGe = 37,
// Floating-Point Pixel Kill If Not Equal // Floating-Point Pixel Kill If Not Equal
// kills_ne dest, src0.a // kills_ne/KILLNEs dest, src0.a
// if (src0.a != 0.0) { // if (src0.a != 0.0) {
// dest.xyzw = 1.0; // dest.xyzw = 1.0;
// discard; // discard;
@ -953,7 +955,7 @@ enum class AluScalarOpcode : uint32_t {
kKillsNe = 38, kKillsNe = 38,
// Floating-Point Pixel Kill If One // Floating-Point Pixel Kill If One
// kills_one dest, src0.a // kills_one/KILLONEs dest, src0.a
// if (src0.a == 1.0) { // if (src0.a == 1.0) {
// dest.xyzw = 1.0; // dest.xyzw = 1.0;
// discard; // discard;
@ -963,41 +965,41 @@ enum class AluScalarOpcode : uint32_t {
kKillsOne = 39, kKillsOne = 39,
// Scalar Square Root, IEEE Aproximation // Scalar Square Root, IEEE Aproximation
// sqrt dest, src0.a // sqrt/SQRT_IEEE dest, src0.a
// dest.xyzw = sqrt(src0.a); // dest.xyzw = sqrt(src0.a);
kSqrt = 40, kSqrt = 40,
// mulsc dest, src0.a, src1.a // mulsc/MUL_CONST_0 dest, src0.a, src1.a
kMulsc0 = 42, kMulsc0 = 42,
// mulsc dest, src0.a, src1.a // mulsc/MUL_CONST_1 dest, src0.a, src1.a
kMulsc1 = 43, kMulsc1 = 43,
// addsc dest, src0.a, src1.a // addsc/ADD_CONST_0 dest, src0.a, src1.a
kAddsc0 = 44, kAddsc0 = 44,
// addsc dest, src0.a, src1.a // addsc/ADD_CONST_1 dest, src0.a, src1.a
kAddsc1 = 45, kAddsc1 = 45,
// subsc dest, src0.a, src1.a // subsc/SUB_CONST_0 dest, src0.a, src1.a
kSubsc0 = 46, kSubsc0 = 46,
// subsc dest, src0.a, src1.a // subsc/SUB_CONST_1 dest, src0.a, src1.a
kSubsc1 = 47, kSubsc1 = 47,
// Scalar Sin // Scalar Sin
// sin dest, src0.a // sin/SIN dest, src0.a
// dest.xyzw = sin(src0.a); // dest.xyzw = sin(src0.a);
kSin = 48, kSin = 48,
// Scalar Cos // Scalar Cos
// cos dest, src0.a // cos/COS dest, src0.a
// dest.xyzw = cos(src0.a); // dest.xyzw = cos(src0.a);
kCos = 49, kCos = 49,
// retain_prev dest // retain_prev/RETAIN_PREV dest
// dest.xyzw = ps; // dest.xyzw = ps;
kRetainPrev = 50, kRetainPrev = 50,
}; };
enum class AluVectorOpcode : uint32_t { enum class AluVectorOpcode : uint32_t {
// Per-Component Floating-Point Add // Per-Component Floating-Point Add
// add dest, src0, src1 // add/ADDv dest, src0, src1
// dest.x = src0.x + src1.x; // dest.x = src0.x + src1.x;
// dest.y = src0.y + src1.y; // dest.y = src0.y + src1.y;
// dest.z = src0.z + src1.z; // dest.z = src0.z + src1.z;
@ -1005,7 +1007,7 @@ enum class AluVectorOpcode : uint32_t {
kAdd = 0, kAdd = 0,
// Per-Component Floating-Point Multiply // Per-Component Floating-Point Multiply
// mul dest, src0, src1 // mul/MULv dest, src0, src1
// dest.x = src0.x * src1.x; // dest.x = src0.x * src1.x;
// dest.y = src0.y * src1.y; // dest.y = src0.y * src1.y;
// dest.z = src0.z * src1.z; // dest.z = src0.z * src1.z;
@ -1013,7 +1015,7 @@ enum class AluVectorOpcode : uint32_t {
kMul = 1, kMul = 1,
// Per-Component Floating-Point Maximum // Per-Component Floating-Point Maximum
// max dest, src0, src1 // max/MAXv dest, src0, src1
// dest.x = src0.x >= src1.x ? src0.x : src1.x; // dest.x = src0.x >= src1.x ? src0.x : src1.x;
// dest.y = src0.x >= src1.y ? src0.y : src1.y; // dest.y = src0.x >= src1.y ? src0.y : src1.y;
// dest.z = src0.x >= src1.z ? src0.z : src1.z; // dest.z = src0.x >= src1.z ? src0.z : src1.z;
@ -1021,7 +1023,7 @@ enum class AluVectorOpcode : uint32_t {
kMax = 2, kMax = 2,
// Per-Component Floating-Point Minimum // Per-Component Floating-Point Minimum
// min dest, src0, src1 // min/MINv dest, src0, src1
// dest.x = src0.x < src1.x ? src0.x : src1.x; // dest.x = src0.x < src1.x ? src0.x : src1.x;
// dest.y = src0.x < src1.y ? src0.y : src1.y; // dest.y = src0.x < src1.y ? src0.y : src1.y;
// dest.z = src0.x < src1.z ? src0.z : src1.z; // dest.z = src0.x < src1.z ? src0.z : src1.z;
@ -1029,7 +1031,7 @@ enum class AluVectorOpcode : uint32_t {
kMin = 3, kMin = 3,
// Per-Component Floating-Point Set If Equal // Per-Component Floating-Point Set If Equal
// seq dest, src0, src1 // seq/SETEv dest, src0, src1
// dest.x = src0.x == src1.x ? 1.0 : 0.0; // dest.x = src0.x == src1.x ? 1.0 : 0.0;
// dest.y = src0.y == src1.y ? 1.0 : 0.0; // dest.y = src0.y == src1.y ? 1.0 : 0.0;
// dest.z = src0.z == src1.z ? 1.0 : 0.0; // dest.z = src0.z == src1.z ? 1.0 : 0.0;
@ -1037,7 +1039,7 @@ enum class AluVectorOpcode : uint32_t {
kSeq = 4, kSeq = 4,
// Per-Component Floating-Point Set If Greater Than // Per-Component Floating-Point Set If Greater Than
// sgt dest, src0, src1 // sgt/SETGTv dest, src0, src1
// dest.x = src0.x > src1.x ? 1.0 : 0.0; // dest.x = src0.x > src1.x ? 1.0 : 0.0;
// dest.y = src0.y > src1.y ? 1.0 : 0.0; // dest.y = src0.y > src1.y ? 1.0 : 0.0;
// dest.z = src0.z > src1.z ? 1.0 : 0.0; // dest.z = src0.z > src1.z ? 1.0 : 0.0;
@ -1045,7 +1047,7 @@ enum class AluVectorOpcode : uint32_t {
kSgt = 5, kSgt = 5,
// Per-Component Floating-Point Set If Greater Than Or Equal // Per-Component Floating-Point Set If Greater Than Or Equal
// sge dest, src0, src1 // sge/SETGTEv dest, src0, src1
// dest.x = src0.x >= src1.x ? 1.0 : 0.0; // dest.x = src0.x >= src1.x ? 1.0 : 0.0;
// dest.y = src0.y >= src1.y ? 1.0 : 0.0; // dest.y = src0.y >= src1.y ? 1.0 : 0.0;
// dest.z = src0.z >= src1.z ? 1.0 : 0.0; // dest.z = src0.z >= src1.z ? 1.0 : 0.0;
@ -1053,7 +1055,7 @@ enum class AluVectorOpcode : uint32_t {
kSge = 6, kSge = 6,
// Per-Component Floating-Point Set If Not Equal // Per-Component Floating-Point Set If Not Equal
// sne dest, src0, src1 // sne/SETNEv dest, src0, src1
// dest.x = src0.x != src1.x ? 1.0 : 0.0; // dest.x = src0.x != src1.x ? 1.0 : 0.0;
// dest.y = src0.y != src1.y ? 1.0 : 0.0; // dest.y = src0.y != src1.y ? 1.0 : 0.0;
// dest.z = src0.z != src1.z ? 1.0 : 0.0; // dest.z = src0.z != src1.z ? 1.0 : 0.0;
@ -1061,7 +1063,7 @@ enum class AluVectorOpcode : uint32_t {
kSne = 7, kSne = 7,
// Per-Component Floating-Point Fractional // Per-Component Floating-Point Fractional
// frc dest, src0 // frc/FRACv dest, src0
// dest.x = src0.x - floor(src0.x); // dest.x = src0.x - floor(src0.x);
// dest.y = src0.y - floor(src0.y); // dest.y = src0.y - floor(src0.y);
// dest.z = src0.z - floor(src0.z); // dest.z = src0.z - floor(src0.z);
@ -1069,7 +1071,7 @@ enum class AluVectorOpcode : uint32_t {
kFrc = 8, kFrc = 8,
// Per-Component Floating-Point Truncate // Per-Component Floating-Point Truncate
// trunc dest, src0 // trunc/TRUNCv dest, src0
// dest.x = src0.x >= 0 ? floor(src0.x) : -floor(-src0.x); // dest.x = src0.x >= 0 ? floor(src0.x) : -floor(-src0.x);
// dest.y = src0.y >= 0 ? floor(src0.y) : -floor(-src0.y); // dest.y = src0.y >= 0 ? floor(src0.y) : -floor(-src0.y);
// dest.z = src0.z >= 0 ? floor(src0.z) : -floor(-src0.z); // dest.z = src0.z >= 0 ? floor(src0.z) : -floor(-src0.z);
@ -1077,7 +1079,7 @@ enum class AluVectorOpcode : uint32_t {
kTrunc = 9, kTrunc = 9,
// Per-Component Floating-Point Floor // Per-Component Floating-Point Floor
// floor dest, src0 // floor/FLOORv dest, src0
// dest.x = floor(src0.x); // dest.x = floor(src0.x);
// dest.y = floor(src0.y); // dest.y = floor(src0.y);
// dest.z = floor(src0.z); // dest.z = floor(src0.z);
@ -1085,7 +1087,7 @@ enum class AluVectorOpcode : uint32_t {
kFloor = 10, kFloor = 10,
// Per-Component Floating-Point Multiply-Add // Per-Component Floating-Point Multiply-Add
// mad dest, src0, src1, src2 // mad/MULADDv dest, src0, src1, src2
// dest.x = src0.x * src1.x + src2.x; // dest.x = src0.x * src1.x + src2.x;
// dest.y = src0.y * src1.y + src2.y; // dest.y = src0.y * src1.y + src2.y;
// dest.z = src0.z * src1.z + src2.z; // dest.z = src0.z * src1.z + src2.z;
@ -1093,7 +1095,7 @@ enum class AluVectorOpcode : uint32_t {
kMad = 11, kMad = 11,
// Per-Component Floating-Point Conditional Move If Equal // Per-Component Floating-Point Conditional Move If Equal
// cndeq dest, src0, src1, src2 // cndeq/CNDEv dest, src0, src1, src2
// dest.x = src0.x == 0.0 ? src1.x : src2.x; // dest.x = src0.x == 0.0 ? src1.x : src2.x;
// dest.y = src0.y == 0.0 ? src1.y : src2.y; // dest.y = src0.y == 0.0 ? src1.y : src2.y;
// dest.z = src0.z == 0.0 ? src1.z : src2.z; // dest.z = src0.z == 0.0 ? src1.z : src2.z;
@ -1101,7 +1103,7 @@ enum class AluVectorOpcode : uint32_t {
kCndEq = 12, kCndEq = 12,
// Per-Component Floating-Point Conditional Move If Greater Than Or Equal // Per-Component Floating-Point Conditional Move If Greater Than Or Equal
// cndge dest, src0, src1, src2 // cndge/CNDGTEv dest, src0, src1, src2
// dest.x = src0.x >= 0.0 ? src1.x : src2.x; // dest.x = src0.x >= 0.0 ? src1.x : src2.x;
// dest.y = src0.y >= 0.0 ? src1.y : src2.y; // dest.y = src0.y >= 0.0 ? src1.y : src2.y;
// dest.z = src0.z >= 0.0 ? src1.z : src2.z; // dest.z = src0.z >= 0.0 ? src1.z : src2.z;
@ -1109,7 +1111,7 @@ enum class AluVectorOpcode : uint32_t {
kCndGe = 13, kCndGe = 13,
// Per-Component Floating-Point Conditional Move If Greater Than // Per-Component Floating-Point Conditional Move If Greater Than
// cndgt dest, src0, src1, src2 // cndgt/CNDGTv dest, src0, src1, src2
// dest.x = src0.x > 0.0 ? src1.x : src2.x; // dest.x = src0.x > 0.0 ? src1.x : src2.x;
// dest.y = src0.y > 0.0 ? src1.y : src2.y; // dest.y = src0.y > 0.0 ? src1.y : src2.y;
// dest.z = src0.z > 0.0 ? src1.z : src2.z; // dest.z = src0.z > 0.0 ? src1.z : src2.z;
@ -1117,43 +1119,60 @@ enum class AluVectorOpcode : uint32_t {
kCndGt = 14, kCndGt = 14,
// Four-Element Dot Product // Four-Element Dot Product
// dp4 dest, src0, src1 // dp4/DOT4v dest, src0, src1
// dest.xyzw = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + // dest.xyzw = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z +
// src0.w * src1.w; // src0.w * src1.w;
// Note: only pv.x contains the value. // Note: only pv.x contains the value.
kDp4 = 15, kDp4 = 15,
// Three-Element Dot Product // Three-Element Dot Product
// dp3 dest, src0, src1 // dp3/DOT3v dest, src0, src1
// dest.xyzw = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z; // dest.xyzw = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z;
// Note: only pv.x contains the value. // Note: only pv.x contains the value.
kDp3 = 16, kDp3 = 16,
// Two-Element Dot Product and Add // Two-Element Dot Product and Add
// dp2add dest, src0, src1, src2 // dp2add/DOT2ADDv dest, src0, src1, src2
// dest.xyzw = src0.x * src1.x + src0.y * src1.y + src2.x; // dest.xyzw = src0.x * src1.x + src0.y * src1.y + src2.x;
// Note: only pv.x contains the value. // Note: only pv.x contains the value.
kDp2Add = 17, kDp2Add = 17,
// Cube Map // Cube Map
// cube dest, src0, src1 // cube/CUBEv dest, src0, src1
// dest.x = T cube coordinate; // dest.x = T cube coordinate:
// dest.y = S cube coordinate; // dest.y = S cube coordinate;
// dest.z = 2.0 * MajorAxis; // dest.z = 2.0 * MajorAxis;
// dest.w = FaceID; // dest.w = FaceID;
// https://developer.amd.com/wordpress/media/2012/12/AMD_Southern_Islands_Instruction_Set_Architecture.pdf
// if (abs(z) >= abs(x) && abs(z) >= abs(y)) {
// tc = -y;
// sc = z < 0.0 ? -x : x;
// ma = 2.0 * z;
// id = z < 0.0 ? 5.0 : 4.0;
// } else if (abs(y) >= abs(x)) {
// tc = y < 0.0 ? -z : z;
// sc = x;
// ma = 2.0 * y;
// id = y < 0.0 ? 3.0 : 2.0;
// } else {
// tc = -y;
// sc = x < 0.0 ? z : -z;
// ma = 2.0 * x;
// id = x < 0.0 ? 1.0 : 0.0;
// }
// Expects src0.zzxy and src1.yxzz swizzles. // Expects src0.zzxy and src1.yxzz swizzles.
// FaceID is D3DCUBEMAP_FACES: // FaceID is D3DCUBEMAP_FACES:
// https://msdn.microsoft.com/en-us/library/windows/desktop/bb172528(v=vs.85).aspx // https://msdn.microsoft.com/en-us/library/windows/desktop/bb172528(v=vs.85).aspx
kCube = 18, kCube = 18,
// Four-Element Maximum // Four-Element Maximum
// max4 dest, src0 // max4/MAX4v dest, src0
// dest.xyzw = max(src0.x, src0.y, src0.z, src0.w); // dest.xyzw = max(src0.x, src0.y, src0.z, src0.w);
// Note: only pv.x contains the value. // Note: only pv.x contains the value.
kMax4 = 19, kMax4 = 19,
// Floating-Point Predicate Counter Increment If Equal // Floating-Point Predicate Counter Increment If Equal
// setp_eq_push dest, src0, src1 // setp_eq_push/PRED_SETE_PUSHv dest, src0, src1
// if (src0.w == 0.0 && src1.w == 0.0) { // if (src0.w == 0.0 && src1.w == 0.0) {
// p0 = 1; // p0 = 1;
// } else { // } else {
@ -1167,7 +1186,7 @@ enum class AluVectorOpcode : uint32_t {
kSetpEqPush = 20, kSetpEqPush = 20,
// Floating-Point Predicate Counter Increment If Not Equal // Floating-Point Predicate Counter Increment If Not Equal
// setp_ne_push dest, src0, src1 // setp_ne_push/PRED_SETNE_PUSHv dest, src0, src1
// if (src0.w == 0.0 && src1.w != 0.0) { // if (src0.w == 0.0 && src1.w != 0.0) {
// p0 = 1; // p0 = 1;
// } else { // } else {
@ -1181,7 +1200,7 @@ enum class AluVectorOpcode : uint32_t {
kSetpNePush = 21, kSetpNePush = 21,
// Floating-Point Predicate Counter Increment If Greater Than // Floating-Point Predicate Counter Increment If Greater Than
// setp_gt_push dest, src0, src1 // setp_gt_push/PRED_SETGT_PUSHv dest, src0, src1
// if (src0.w == 0.0 && src1.w > 0.0) { // if (src0.w == 0.0 && src1.w > 0.0) {
// p0 = 1; // p0 = 1;
// } else { // } else {
@ -1195,7 +1214,7 @@ enum class AluVectorOpcode : uint32_t {
kSetpGtPush = 22, kSetpGtPush = 22,
// Floating-Point Predicate Counter Increment If Greater Than Or Equal // Floating-Point Predicate Counter Increment If Greater Than Or Equal
// setp_ge_push dest, src0, src1 // setp_ge_push/PRED_SETGTE_PUSHv dest, src0, src1
// if (src0.w == 0.0 && src1.w >= 0.0) { // if (src0.w == 0.0 && src1.w >= 0.0) {
// p0 = 1; // p0 = 1;
// } else { // } else {
@ -1209,7 +1228,7 @@ enum class AluVectorOpcode : uint32_t {
kSetpGePush = 23, kSetpGePush = 23,
// Floating-Point Pixel Kill If Equal // Floating-Point Pixel Kill If Equal
// kill_eq dest, src0, src1 // kill_eq/KILLEv dest, src0, src1
// if (src0.x == src1.x || // if (src0.x == src1.x ||
// src0.y == src1.y || // src0.y == src1.y ||
// src0.z == src1.z || // src0.z == src1.z ||
@ -1222,7 +1241,7 @@ enum class AluVectorOpcode : uint32_t {
kKillEq = 24, kKillEq = 24,
// Floating-Point Pixel Kill If Greater Than // Floating-Point Pixel Kill If Greater Than
// kill_gt dest, src0, src1 // kill_gt/KILLGTv dest, src0, src1
// if (src0.x > src1.x || // if (src0.x > src1.x ||
// src0.y > src1.y || // src0.y > src1.y ||
// src0.z > src1.z || // src0.z > src1.z ||
@ -1235,7 +1254,7 @@ enum class AluVectorOpcode : uint32_t {
kKillGt = 25, kKillGt = 25,
// Floating-Point Pixel Kill If Equal // Floating-Point Pixel Kill If Equal
// kill_ge dest, src0, src1 // kill_ge/KILLGTEv dest, src0, src1
// if (src0.x >= src1.x || // if (src0.x >= src1.x ||
// src0.y >= src1.y || // src0.y >= src1.y ||
// src0.z >= src1.z || // src0.z >= src1.z ||
@ -1248,7 +1267,7 @@ enum class AluVectorOpcode : uint32_t {
kKillGe = 26, kKillGe = 26,
// Floating-Point Pixel Kill If Equal // Floating-Point Pixel Kill If Equal
// kill_ne dest, src0, src1 // kill_ne/KILLNEv dest, src0, src1
// if (src0.x != src1.x || // if (src0.x != src1.x ||
// src0.y != src1.y || // src0.y != src1.y ||
// src0.z != src1.z || // src0.z != src1.z ||
@ -1260,7 +1279,7 @@ enum class AluVectorOpcode : uint32_t {
// } // }
kKillNe = 27, kKillNe = 27,
// dst dest, src0, src1 // dst/DSTv dest, src0, src1
// dest.x = 1.0; // dest.x = 1.0;
// dest.y = src0.y * src1.y; // dest.y = src0.y * src1.y;
// dest.z = src0.z; // dest.z = src0.z;
@ -1269,7 +1288,7 @@ enum class AluVectorOpcode : uint32_t {
// Per-Component Floating-Point Maximum with Copy To Integer in AR // Per-Component Floating-Point Maximum with Copy To Integer in AR
// maxa dest, src0, src1 // maxa dest, src0, src1
// This is a combined max + mova. // This is a combined max + mova/MOVAv.
// int result = (int)floor(src0.w + 0.5); // int result = (int)floor(src0.w + 0.5);
// a0 = clamp(result, -256, 255); // a0 = clamp(result, -256, 255);
// dest.x = src0.x >= src1.x ? src0.x : src1.x; // dest.x = src0.x >= src1.x ? src0.x : src1.x;