Skip to content

@DPZZxlz: Validate submission bd18622f-bf25-443a-93b8-2cf6b89214cb #740

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
yukon-autoresearch wants to merge 1 commit into from
+35 −6
Closed

@DPZZxlz: Validate submission bd18622f-bf25-443a-93b8-2cf6b89214cb #740

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions candidates/subset/subset.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1 +1,4 @@
#ifndef QSB_REMEASURE_TAG_09201153
#define QSB_REMEASURE_TAG_09201153 1 /* no-op: frontier re-measurement, see submission note */
#endif
#include "tests/gpu_epochs/tree.cu"
38 changes: 32 additions & 6 deletions candidates/subset/tests/gpu_epochs/pair_shared.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,24 +58,50 @@ __device__ __forceinline__ uint32_t qsb_k2s_post(
#define ZLAB_K2S3M 1
#endif
#if ZLAB_K2S3M
/* QSB_SPEC_PREPARE (kill switch; 0 restores the promoted code exactly).
* Carried from owizdom's public submission 4f367236-4a3c-49de-a81b-1730e6f0d889
* (commit 878eb25e4470da77446b398889aab2f65ba0eadd), which moved the pre-inverse
* prepare and pre3 off the exact field ops and onto the same unguarded speculative
* ops that the filter chain (qsb_filter_point_add, qsb_filter_last_add) and the
* post-inverse qsb_k2s_post3 on this frontier already use. Same formulas, same
* operand order, same operation count; only the carry/borrow guard differs.
* Correctness argument is the one already stated at the top of filter_tail_sc.cuh:
* a dropped carry here can only corrupt this candidate's (or, through the block
* inverse product, this block's) speculative x-coordinates, which loses tentative
* hits; it can never publish one, because every tentative hit is recomputed by
* kernel_verify_pair_hits on the unchanged exact chain.
* Not touched: qsb_k2s_front, qsb_k2s_front_exact, qsb_k2s_pre,
* qsb_xyzz_finish_prepare, qsb_k2s_post, qsb_pair_verify_candidate. */
#ifndef QSB_SPEC_PREPARE
#define QSB_SPEC_PREPARE 1
#endif
#if QSB_SPEC_PREPARE
#define QSB_PRE_FMUL(r,a,b) QSB_FMUL(r,a,b)
#define QSB_PRE_FSUB(r,a,b) QSB_FSUB(r,a,b)
#define QSB_PRE_FADD(r,a,b) QSB_FADD(r,a,b)
#else
#define QSB_PRE_FMUL(r,a,b) X_FMUL(r,a,b)
#define QSB_PRE_FSUB(r,a,b) X_FSUB(r,a,b)
#define QSB_PRE_FADD(r,a,b) X_FADD(r,a,b)
#endif
__device__ __forceinline__ void qsb_k2s_pre3(
uint64_t *Y, uint64_t *ZZ, uint64_t *ZZZ, uint64_t *yR, uint64_t *n
) {
uint64_t yb[4];
X_FMUL(yb, yR, ZZZ);
X_FSUB(n, yb, Y);
X_FADD(n + 4, yb, Y);
QSB_PRE_FMUL(yb, yR, ZZZ);
QSB_PRE_FSUB(n, yb, Y);
QSB_PRE_FADD(n + 4, yb, Y);
Load256(n + 8, ZZ);
}
/* Filter-only copy of qsb_xyzz_finish_prepare (the exact front keeps the original). */
__device__ __forceinline__ void qsb_xyzz_finish_prepare_f(
uint64_t *X_D, uint64_t *ZZ, uint64_t *ZZZ, uint64_t *xR, uint64_t *W
) {
uint64_t t[4];
X_FMUL(t, xR, ZZ);
X_FSUB(t, t, X_D);
QSB_PRE_FMUL(t, xR, ZZ);
QSB_PRE_FSUB(t, t, X_D);
Load256(X_D, t); /* X_D becomes d */
X_FMUL(W, ZZZ, X_D); /* W = ZZZ*d */
QSB_PRE_FMUL(W, ZZZ, X_D); /* W = ZZZ*d */
W[4] = 0;
}
/* h = ZZ*inv is the common slope scale: m1 = n[0..3]*h, m2 = n[4..7]*h. The
Expand Down
Loading