xemu-project/nv2a_vsh_cpu
1
0
Fork 0
mirror of https://github.com/xemu-project/nv2a_vsh_cpu.git synced 2024-11-23 09:49:39 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Matches HW behavior for slt/sge.

Browse Source
This commit is contained in:
Erik Abair 2022-06-23 12:50:26 -07:00
parent 4d63c0a1a6
commit 155105ce3a
2 changed files with 54 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
.gitignore vendored
View File

@ -1,3 +1,4 @@
*.c.d
*.cpp.d *.cpp.d
*.obj *.obj
*.iso *.iso

65
src/nv2a_vsh_cpu.c
View File

@ -1,6 +1,8 @@
#include "nv2a_vsh_cpu.h" #include "nv2a_vsh_cpu.h"
#include <math.h> #include <math.h>
#include <stdbool.h>
#include <stdint.h>
#include <string.h> #include <string.h>
#define _X 0 #define _X 0
@ -101,25 +103,50 @@ void nv2a_vsh_cpu_max(float *out, const float *inputs) {
COMP(inputs, 0, _W) > COMP(inputs, 1, _W) ? COMP(inputs, 0, _W) : COMP(inputs, 1, _W); COMP(inputs, 0, _W) > COMP(inputs, 1, _W) ? COMP(inputs, 0, _W) : COMP(inputs, 1, _W);
} }
static inline float nv2a_less_than(float a, float b) {
if (a < b) {
return 1.0f;
}
// nv2a hardware treats -0 as < 0.
uint32_t a_int = *(uint32_t*)&a;
uint32_t b_int = *(uint32_t*)&b;
if (a_int == 0x80000000 && !b_int) {
return 1.0f;
}
return 0.0f;
}
void nv2a_vsh_cpu_slt(float *out, const float *inputs) { void nv2a_vsh_cpu_slt(float *out, const float *inputs) {
out[0] = COMP(inputs, 0, _X) < COMP(inputs, 1, _X) ? 1.0f : 0.0f; out[0] = nv2a_less_than(COMP(inputs, 0, _X), COMP(inputs, 1, _X));
out[1] = COMP(inputs, 0, _Y) < COMP(inputs, 1, _Y) ? 1.0f : 0.0f; out[1] = nv2a_less_than(COMP(inputs, 0, _Y), COMP(inputs, 1, _Y));
out[2] = COMP(inputs, 0, _Z) < COMP(inputs, 1, _Z) ? 1.0f : 0.0f; out[2] = nv2a_less_than(COMP(inputs, 0, _Z), COMP(inputs, 1, _Z));
out[3] = COMP(inputs, 0, _W) < COMP(inputs, 1, _W) ? 1.0f : 0.0f; out[3] = nv2a_less_than(COMP(inputs, 0, _W), COMP(inputs, 1, _W));
} }
void nv2a_vsh_cpu_sge(float *out, const float *inputs) { void nv2a_vsh_cpu_sge(float *out, const float *inputs) {
out[0] = COMP(inputs, 0, _X) >= COMP(inputs, 1, _X) ? 1.0f : 0.0f; out[0] = -1.0f * nv2a_less_than(COMP(inputs, 0, _X), COMP(inputs, 1, _X)) + 1.0f;
out[1] = COMP(inputs, 0, _Y) >= COMP(inputs, 1, _Y) ? 1.0f : 0.0f; out[1] = -1.0f * nv2a_less_than(COMP(inputs, 0, _Y), COMP(inputs, 1, _Y)) + 1.0f;
out[2] = COMP(inputs, 0, _Z) >= COMP(inputs, 1, _Z) ? 1.0f : 0.0f; out[2] = -1.0f * nv2a_less_than(COMP(inputs, 0, _Z), COMP(inputs, 1, _Z)) + 1.0f;
out[3] = COMP(inputs, 0, _W) >= COMP(inputs, 1, _W) ? 1.0f : 0.0f; out[3] = -1.0f * nv2a_less_than(COMP(inputs, 0, _W), COMP(inputs, 1, _W)) + 1.0f;
} }
void nv2a_vsh_cpu_rcp(float *out, const float *inputs) { void nv2a_vsh_cpu_rcp(float *out, const float *inputs) {
float result = float in = COMP(inputs, 0, _X);
(COMP(inputs, 0, _X) == 1.0f float result;
? 1.0f if (in == 1.0f) {
: (COMP(inputs, 0, _X) == 0.0f ? INFINITY : 1.0f / COMP(inputs, 0, _X))); result = 1.0f;
} else if (in == 0.0f) {
// nv2a preserves the sign.
if (*(uint32_t*)&in & 0x80000000) {
result = -INFINITY;
} else {
result = INFINITY;
}
} else {
result = 1.0f / in;
}
out[0] = result; out[0] = result;
out[1] = result; out[1] = result;
out[2] = result; out[2] = result;
@ -151,6 +178,20 @@ void nv2a_vsh_cpu_rcc(float *out, const float *inputs) {
} }
void nv2a_vsh_cpu_rsq(float *out, const float *inputs) { void nv2a_vsh_cpu_rsq(float *out, const float *inputs) {
// float in = COMP(inputs, 0, _X);
// float result;
// if (in == 1.0f) {
// result = 1.0f;
// } else if (in == 0.0f) {
// // nv2a preserves the sign.
// if (*(uint32_t*)&in & 0x80000000) {
// result = -INFINITY;
// } else {
// result = INFINITY;
// }
// } else {
// result = 1.0f / in;
// }
float result = float result =
(inputs[0] == 1.0f (inputs[0] == 1.0f
? 1.0f ? 1.0f